diff options
Diffstat (limited to 'src/grp-system/libcore')
121 files changed, 57085 insertions, 0 deletions
diff --git a/src/grp-system/libcore/.gitignore b/src/grp-system/libcore/.gitignore new file mode 100644 index 0000000000..465b4fcc20 --- /dev/null +++ b/src/grp-system/libcore/.gitignore @@ -0,0 +1,3 @@ +/macros.systemd +/triggers.systemd +/systemd.pc diff --git a/src/grp-system/libcore/GNUmakefile b/src/grp-system/libcore/GNUmakefile new file mode 120000 index 0000000000..95e5924740 --- /dev/null +++ b/src/grp-system/libcore/GNUmakefile @@ -0,0 +1 @@ +../../../GNUmakefile
\ No newline at end of file diff --git a/src/grp-system/libcore/Makefile b/src/grp-system/libcore/Makefile new file mode 100644 index 0000000000..76e6e9ddee --- /dev/null +++ b/src/grp-system/libcore/Makefile @@ -0,0 +1,28 @@ +# -*- Mode: makefile; indent-tabs-mode: t -*- +# +# This file is part of systemd. +# +# Copyright 2010-2012 Lennart Poettering +# Copyright 2010-2012 Kay Sievers +# Copyright 2013 Zbigniew Jędrzejewski-Szmek +# Copyright 2013 David Strauss +# Copyright 2016 Luke Shumaker +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# systemd is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with systemd; If not, see <http://www.gnu.org/licenses/>. +include $(dir $(lastword $(MAKEFILE_LIST)))/../../../config.mk +include $(topsrcdir)/build-aux/Makefile.head.mk + +nested.subdirs += src + +include $(topsrcdir)/build-aux/Makefile.tail.mk diff --git a/src/grp-system/libcore/include/core/automount.h b/src/grp-system/libcore/include/core/automount.h new file mode 100644 index 0000000000..76a201178e --- /dev/null +++ b/src/grp-system/libcore/include/core/automount.h @@ -0,0 +1,59 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Automount Automount; + +#include "unit.h" + +typedef enum AutomountResult { + AUTOMOUNT_SUCCESS, + AUTOMOUNT_FAILURE_RESOURCES, + AUTOMOUNT_FAILURE_START_LIMIT_HIT, + AUTOMOUNT_FAILURE_MOUNT_START_LIMIT_HIT, + _AUTOMOUNT_RESULT_MAX, + _AUTOMOUNT_RESULT_INVALID = -1 +} AutomountResult; + +struct Automount { + Unit meta; + + AutomountState state, deserialized_state; + + char *where; + usec_t timeout_idle_usec; + + int pipe_fd; + sd_event_source *pipe_event_source; + mode_t directory_mode; + dev_t dev_id; + + Set *tokens; + Set *expire_tokens; + + sd_event_source *expire_event_source; + + AutomountResult result; +}; + +extern const UnitVTable automount_vtable; + +const char* automount_result_to_string(AutomountResult i) _const_; +AutomountResult automount_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/bus-policy.h b/src/grp-system/libcore/include/core/bus-policy.h new file mode 100644 index 0000000000..a338f29af6 --- /dev/null +++ b/src/grp-system/libcore/include/core/bus-policy.h @@ -0,0 +1,64 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2014 Daniel Mack + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "sd-bus/kdbus.h" +#include "systemd-basic/list.h" +#include "systemd-basic/macro.h" + +typedef struct BusNamePolicy BusNamePolicy; + +typedef enum BusPolicyAccess { + BUS_POLICY_ACCESS_SEE, + BUS_POLICY_ACCESS_TALK, + BUS_POLICY_ACCESS_OWN, + _BUS_POLICY_ACCESS_MAX, + _BUS_POLICY_ACCESS_INVALID = -1 +} BusPolicyAccess; + +typedef enum BusNamePolicyType { + BUSNAME_POLICY_TYPE_USER, + BUSNAME_POLICY_TYPE_GROUP, + _BUSNAME_POLICY_TYPE_MAX, + _BUSNAME_POLICY_TYPE_INVALID = -1 +} BusNamePolicyType; + +struct BusNamePolicy { + BusNamePolicyType type; + BusPolicyAccess access; + + char *name; + + LIST_FIELDS(BusNamePolicy, policy); +}; + +int bus_kernel_translate_access(BusPolicyAccess access); +int bus_kernel_translate_policy(const BusNamePolicy *policy, struct kdbus_item *item); + +const char* bus_policy_access_to_string(BusPolicyAccess i) _const_; +BusPolicyAccess bus_policy_access_from_string(const char *s) _pure_; + +int bus_kernel_make_starter( + int fd, + const char *name, + bool activating, + bool accept_fd, + BusNamePolicy *policy, + BusPolicyAccess world_policy); diff --git a/src/grp-system/libcore/include/core/busname.h b/src/grp-system/libcore/include/core/busname.h new file mode 100644 index 0000000000..aa7f0ecb1b --- /dev/null +++ b/src/grp-system/libcore/include/core/busname.h @@ -0,0 +1,69 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct BusName BusName; +typedef struct BusNamePolicy BusNamePolicy; + +#include "bus-policy.h" +#include "unit.h" + +typedef enum BusNameResult { + BUSNAME_SUCCESS, + BUSNAME_FAILURE_RESOURCES, + BUSNAME_FAILURE_TIMEOUT, + BUSNAME_FAILURE_EXIT_CODE, + BUSNAME_FAILURE_SIGNAL, + BUSNAME_FAILURE_CORE_DUMP, + BUSNAME_FAILURE_START_LIMIT_HIT, + BUSNAME_FAILURE_SERVICE_START_LIMIT_HIT, + _BUSNAME_RESULT_MAX, + _BUSNAME_RESULT_INVALID = -1 +} BusNameResult; + +struct BusName { + Unit meta; + + char *name; + int starter_fd; + + bool activating; + bool accept_fd; + + UnitRef service; + + BusNameState state, deserialized_state; + BusNameResult result; + + usec_t timeout_usec; + + sd_event_source *starter_event_source; + sd_event_source *timer_event_source; + + pid_t control_pid; + + LIST_HEAD(BusNamePolicy, policy); + BusPolicyAccess policy_world; +}; + +extern const UnitVTable busname_vtable; + +const char* busname_result_to_string(BusNameResult i) _const_; +BusNameResult busname_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/cgroup.h b/src/grp-system/libcore/include/core/cgroup.h new file mode 100644 index 0000000000..6293b84cd7 --- /dev/null +++ b/src/grp-system/libcore/include/core/cgroup.h @@ -0,0 +1,187 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/list.h" +#include "systemd-basic/time-util.h" + +typedef struct CGroupContext CGroupContext; +typedef struct CGroupDeviceAllow CGroupDeviceAllow; +typedef struct CGroupIODeviceWeight CGroupIODeviceWeight; +typedef struct CGroupIODeviceLimit CGroupIODeviceLimit; +typedef struct CGroupBlockIODeviceWeight CGroupBlockIODeviceWeight; +typedef struct CGroupBlockIODeviceBandwidth CGroupBlockIODeviceBandwidth; + +typedef enum CGroupDevicePolicy { + + /* When devices listed, will allow those, plus built-in ones, + if none are listed will allow everything. */ + CGROUP_AUTO, + + /* Everything forbidden, except built-in ones and listed ones. */ + CGROUP_CLOSED, + + /* Everythings forbidden, except for the listed devices */ + CGROUP_STRICT, + + _CGROUP_DEVICE_POLICY_MAX, + _CGROUP_DEVICE_POLICY_INVALID = -1 +} CGroupDevicePolicy; + +struct CGroupDeviceAllow { + LIST_FIELDS(CGroupDeviceAllow, device_allow); + char *path; + bool r:1; + bool w:1; + bool m:1; +}; + +struct CGroupIODeviceWeight { + LIST_FIELDS(CGroupIODeviceWeight, device_weights); + char *path; + uint64_t weight; +}; + +struct CGroupIODeviceLimit { + LIST_FIELDS(CGroupIODeviceLimit, device_limits); + char *path; + uint64_t limits[_CGROUP_IO_LIMIT_TYPE_MAX]; +}; + +struct CGroupBlockIODeviceWeight { + LIST_FIELDS(CGroupBlockIODeviceWeight, device_weights); + char *path; + uint64_t weight; +}; + +struct CGroupBlockIODeviceBandwidth { + LIST_FIELDS(CGroupBlockIODeviceBandwidth, device_bandwidths); + char *path; + uint64_t rbps; + uint64_t wbps; +}; + +struct CGroupContext { + bool cpu_accounting; + bool io_accounting; + bool blockio_accounting; + bool memory_accounting; + bool tasks_accounting; + + /* For unified hierarchy */ + uint64_t cpu_weight; + uint64_t startup_cpu_weight; + usec_t cpu_quota_per_sec_usec; + + uint64_t io_weight; + uint64_t startup_io_weight; + LIST_HEAD(CGroupIODeviceWeight, io_device_weights); + LIST_HEAD(CGroupIODeviceLimit, io_device_limits); + + uint64_t memory_low; + uint64_t memory_high; + uint64_t memory_max; + uint64_t memory_swap_max; + + /* For legacy hierarchies */ + uint64_t cpu_shares; + uint64_t startup_cpu_shares; + + uint64_t blockio_weight; + uint64_t startup_blockio_weight; + LIST_HEAD(CGroupBlockIODeviceWeight, blockio_device_weights); + LIST_HEAD(CGroupBlockIODeviceBandwidth, blockio_device_bandwidths); + + uint64_t memory_limit; + + CGroupDevicePolicy device_policy; + LIST_HEAD(CGroupDeviceAllow, device_allow); + + /* Common */ + uint64_t tasks_max; + + bool delegate; +}; + +#include "unit.h" + +void cgroup_context_init(CGroupContext *c); +void cgroup_context_done(CGroupContext *c); +void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix); + +CGroupMask cgroup_context_get_mask(CGroupContext *c); + +void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a); +void cgroup_context_free_io_device_weight(CGroupContext *c, CGroupIODeviceWeight *w); +void cgroup_context_free_io_device_limit(CGroupContext *c, CGroupIODeviceLimit *l); +void cgroup_context_free_blockio_device_weight(CGroupContext *c, CGroupBlockIODeviceWeight *w); +void cgroup_context_free_blockio_device_bandwidth(CGroupContext *c, CGroupBlockIODeviceBandwidth *b); + +CGroupMask unit_get_own_mask(Unit *u); +CGroupMask unit_get_siblings_mask(Unit *u); +CGroupMask unit_get_members_mask(Unit *u); +CGroupMask unit_get_subtree_mask(Unit *u); + +CGroupMask unit_get_target_mask(Unit *u); +CGroupMask unit_get_enable_mask(Unit *u); + +void unit_update_cgroup_members_masks(Unit *u); + +char *unit_default_cgroup_path(Unit *u); +int unit_set_cgroup_path(Unit *u, const char *path); + +int unit_realize_cgroup(Unit *u); +void unit_release_cgroup(Unit *u); +void unit_prune_cgroup(Unit *u); +int unit_watch_cgroup(Unit *u); + +int unit_attach_pids_to_cgroup(Unit *u); + +int manager_setup_cgroup(Manager *m); +void manager_shutdown_cgroup(Manager *m, bool delete); + +unsigned manager_dispatch_cgroup_queue(Manager *m); + +Unit *manager_get_unit_by_cgroup(Manager *m, const char *cgroup); +Unit *manager_get_unit_by_pid_cgroup(Manager *m, pid_t pid); +Unit* manager_get_unit_by_pid(Manager *m, pid_t pid); + +int unit_search_main_pid(Unit *u, pid_t *ret); +int unit_watch_all_pids(Unit *u); + +int unit_get_memory_current(Unit *u, uint64_t *ret); +int unit_get_tasks_current(Unit *u, uint64_t *ret); +int unit_get_cpu_usage(Unit *u, nsec_t *ret); +int unit_reset_cpu_usage(Unit *u); + +bool unit_cgroup_delegate(Unit *u); + +int unit_notify_cgroup_empty(Unit *u); +int manager_notify_cgroup_empty(Manager *m, const char *group); + +void unit_invalidate_cgroup(Unit *u, CGroupMask m); + +void manager_invalidate_startup_units(Manager *m); + +const char* cgroup_device_policy_to_string(CGroupDevicePolicy i) _const_; +CGroupDevicePolicy cgroup_device_policy_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/dbus-manager.h b/src/grp-system/libcore/include/core/dbus-manager.h new file mode 100644 index 0000000000..36a2e9481b --- /dev/null +++ b/src/grp-system/libcore/include/core/dbus-manager.h @@ -0,0 +1,28 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "manager.h" + +extern const sd_bus_vtable bus_manager_vtable[]; + +void bus_manager_send_finished(Manager *m, usec_t firmware_usec, usec_t loader_usec, usec_t kernel_usec, usec_t initrd_usec, usec_t userspace_usec, usec_t total_usec); +void bus_manager_send_reloading(Manager *m, bool active); +void bus_manager_send_change_signal(Manager *m); diff --git a/src/grp-system/libcore/include/core/device.h b/src/grp-system/libcore/include/core/device.h new file mode 100644 index 0000000000..184a1a349b --- /dev/null +++ b/src/grp-system/libcore/include/core/device.h @@ -0,0 +1,47 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Device Device; + +typedef enum DeviceFound { + DEVICE_NOT_FOUND = 0, + DEVICE_FOUND_UDEV = 1, + DEVICE_FOUND_MOUNT = 2, + DEVICE_FOUND_SWAP = 4, +} DeviceFound; + +struct Device { + Unit meta; + + char *sysfs; + DeviceFound found; + + /* In order to be able to distinguish dependencies on + different device nodes we might end up creating multiple + devices for the same sysfs path. We chain them up here. */ + LIST_FIELDS(struct Device, same_sysfs); + + DeviceState state, deserialized_state; +}; + +extern const UnitVTable device_vtable; + +int device_found_node(Manager *m, const char *node, bool add, DeviceFound found, bool now); diff --git a/src/grp-system/libcore/include/core/dynamic-user.h b/src/grp-system/libcore/include/core/dynamic-user.h new file mode 100644 index 0000000000..0b8bce1a72 --- /dev/null +++ b/src/grp-system/libcore/include/core/dynamic-user.h @@ -0,0 +1,66 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct DynamicUser DynamicUser; + +typedef struct DynamicCreds { + /* A combination of a dynamic user and group */ + DynamicUser *user; + DynamicUser *group; +} DynamicCreds; + +#include "manager.h" + +/* Note that this object always allocates a pair of user and group under the same name, even if one of them isn't + * used. This means, if you want to allocate a group and user pair, and they might have two different names, then you + * need to allocated two of these objects. DynamicCreds below makes that easy. */ +struct DynamicUser { + int n_ref; + Manager *manager; + + /* An AF_UNIX socket pair that contains a datagram containing both the numeric ID assigned, as well as a lock + * file fd locking the user ID we picked. */ + int storage_socket[2]; + + char name[]; +}; + +int dynamic_user_acquire(Manager *m, const char *name, DynamicUser **ret); + +int dynamic_user_realize(DynamicUser *d, uid_t *ret); +int dynamic_user_current(DynamicUser *d, uid_t *ret); + +DynamicUser* dynamic_user_ref(DynamicUser *d); +DynamicUser* dynamic_user_unref(DynamicUser *d); +DynamicUser* dynamic_user_destroy(DynamicUser *d); + +int dynamic_user_serialize(Manager *m, FILE *f, FDSet *fds); +void dynamic_user_deserialize_one(Manager *m, const char *value, FDSet *fds); +void dynamic_user_vacuum(Manager *m, bool close_user); + +int dynamic_user_lookup_uid(Manager *m, uid_t uid, char **ret); +int dynamic_user_lookup_name(Manager *m, const char *name, uid_t *ret); + +int dynamic_creds_acquire(DynamicCreds *creds, Manager *m, const char *user, const char *group); +int dynamic_creds_realize(DynamicCreds *creds, uid_t *uid, gid_t *gid); + +void dynamic_creds_unref(DynamicCreds *creds); +void dynamic_creds_destroy(DynamicCreds *creds); diff --git a/src/grp-system/libcore/include/core/emergency-action.h b/src/grp-system/libcore/include/core/emergency-action.h new file mode 100644 index 0000000000..c463b892bc --- /dev/null +++ b/src/grp-system/libcore/include/core/emergency-action.h @@ -0,0 +1,42 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2014 Lennart Poettering + Copyright 2012 Michael Olbrich + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef enum EmergencyAction { + EMERGENCY_ACTION_NONE, + EMERGENCY_ACTION_REBOOT, + EMERGENCY_ACTION_REBOOT_FORCE, + EMERGENCY_ACTION_REBOOT_IMMEDIATE, + EMERGENCY_ACTION_POWEROFF, + EMERGENCY_ACTION_POWEROFF_FORCE, + EMERGENCY_ACTION_POWEROFF_IMMEDIATE, + _EMERGENCY_ACTION_MAX, + _EMERGENCY_ACTION_INVALID = -1 +} EmergencyAction; + +#include "systemd-basic/macro.h" + +#include "manager.h" + +int emergency_action(Manager *m, EmergencyAction action, const char *reboot_arg, const char *reason); + +const char* emergency_action_to_string(EmergencyAction i) _const_; +EmergencyAction emergency_action_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/execute.h b/src/grp-system/libcore/include/core/execute.h new file mode 100644 index 0000000000..4dad8713fc --- /dev/null +++ b/src/grp-system/libcore/include/core/execute.h @@ -0,0 +1,317 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sched.h> +#include <stdbool.h> +#include <stdio.h> +#include <sys/capability.h> + +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/list.h" +#include "systemd-basic/missing.h" +#include "systemd-shared/fdset.h" + +typedef struct ExecCommand ExecCommand; +typedef struct ExecContext ExecContext; +typedef struct ExecParameters ExecParameters; +typedef struct ExecRuntime ExecRuntime; +typedef struct ExecStatus ExecStatus; + +#include "namespace.h" + +typedef enum ExecUtmpMode { + EXEC_UTMP_INIT, + EXEC_UTMP_LOGIN, + EXEC_UTMP_USER, + _EXEC_UTMP_MODE_MAX, + _EXEC_UTMP_MODE_INVALID = -1 +} ExecUtmpMode; + +typedef enum ExecInput { + EXEC_INPUT_NULL, + EXEC_INPUT_TTY, + EXEC_INPUT_TTY_FORCE, + EXEC_INPUT_TTY_FAIL, + EXEC_INPUT_SOCKET, + EXEC_INPUT_NAMED_FD, + _EXEC_INPUT_MAX, + _EXEC_INPUT_INVALID = -1 +} ExecInput; + +typedef enum ExecOutput { + EXEC_OUTPUT_INHERIT, + EXEC_OUTPUT_NULL, + EXEC_OUTPUT_TTY, + EXEC_OUTPUT_SYSLOG, + EXEC_OUTPUT_SYSLOG_AND_CONSOLE, + EXEC_OUTPUT_KMSG, + EXEC_OUTPUT_KMSG_AND_CONSOLE, + EXEC_OUTPUT_JOURNAL, + EXEC_OUTPUT_JOURNAL_AND_CONSOLE, + EXEC_OUTPUT_SOCKET, + EXEC_OUTPUT_NAMED_FD, + _EXEC_OUTPUT_MAX, + _EXEC_OUTPUT_INVALID = -1 +} ExecOutput; + +struct ExecStatus { + dual_timestamp start_timestamp; + dual_timestamp exit_timestamp; + pid_t pid; + int code; /* as in siginfo_t::si_code */ + int status; /* as in sigingo_t::si_status */ +}; + +struct ExecCommand { + char *path; + char **argv; + ExecStatus exec_status; + LIST_FIELDS(ExecCommand, command); /* useful for chaining commands */ + bool ignore:1; + bool privileged:1; +}; + +struct ExecRuntime { + int n_ref; + + char *tmp_dir; + char *var_tmp_dir; + + /* An AF_UNIX socket pair, that contains a datagram containing a file descriptor referring to the network + * namespace. */ + int netns_storage_socket[2]; +}; + +struct ExecContext { + char **environment; + char **environment_files; + char **pass_environment; + + struct rlimit *rlimit[_RLIMIT_MAX]; + char *working_directory, *root_directory; + bool working_directory_missing_ok; + bool working_directory_home; + + mode_t umask; + int oom_score_adjust; + int nice; + int ioprio; + int cpu_sched_policy; + int cpu_sched_priority; + + cpu_set_t *cpuset; + unsigned cpuset_ncpus; + + ExecInput std_input; + ExecOutput std_output; + ExecOutput std_error; + char *stdio_fdname[3]; + + nsec_t timer_slack_nsec; + + bool stdio_as_fds; + + char *tty_path; + + bool tty_reset; + bool tty_vhangup; + bool tty_vt_disallocate; + + bool ignore_sigpipe; + + /* Since resolving these names might involve socket + * connections and we don't want to deadlock ourselves these + * names are resolved on execution only and in the child + * process. */ + char *user; + char *group; + char **supplementary_groups; + + char *pam_name; + + char *utmp_id; + ExecUtmpMode utmp_mode; + + bool selinux_context_ignore; + char *selinux_context; + + bool apparmor_profile_ignore; + char *apparmor_profile; + + bool smack_process_label_ignore; + char *smack_process_label; + + char **read_write_paths, **read_only_paths, **inaccessible_paths; + unsigned long mount_flags; + + uint64_t capability_bounding_set; + uint64_t capability_ambient_set; + int secure_bits; + + int syslog_priority; + char *syslog_identifier; + bool syslog_level_prefix; + + bool cpu_sched_reset_on_fork; + bool non_blocking; + bool private_tmp; + bool private_network; + bool private_devices; + bool private_users; + ProtectSystem protect_system; + ProtectHome protect_home; + bool protect_kernel_tunables; + bool protect_kernel_modules; + bool protect_control_groups; + + bool no_new_privileges; + + bool dynamic_user; + bool remove_ipc; + + /* This is not exposed to the user but available + * internally. We need it to make sure that whenever we spawn + * /usr/bin/mount it is run in the same process group as us so + * that the autofs logic detects that it belongs to us and we + * don't enter a trigger loop. */ + bool same_pgrp; + + unsigned long personality; + + Set *syscall_filter; + Set *syscall_archs; + int syscall_errno; + bool syscall_whitelist:1; + + Set *address_families; + bool address_families_whitelist:1; + + char **runtime_directory; + mode_t runtime_directory_mode; + + bool memory_deny_write_execute; + bool restrict_realtime; + + bool oom_score_adjust_set:1; + bool nice_set:1; + bool ioprio_set:1; + bool cpu_sched_set:1; + bool no_new_privileges_set:1; +}; + +typedef enum ExecFlags { + EXEC_CONFIRM_SPAWN = 1U << 0, + EXEC_APPLY_PERMISSIONS = 1U << 1, + EXEC_APPLY_CHROOT = 1U << 2, + EXEC_APPLY_TTY_STDIN = 1U << 3, + + /* The following are not used by execute.c, but by consumers internally */ + EXEC_PASS_FDS = 1U << 4, + EXEC_IS_CONTROL = 1U << 5, + EXEC_SETENV_RESULT = 1U << 6, + EXEC_SET_WATCHDOG = 1U << 7, +} ExecFlags; + +struct ExecParameters { + char **argv; + char **environment; + + int *fds; + char **fd_names; + unsigned n_fds; + + ExecFlags flags; + bool selinux_context_net:1; + + bool cgroup_delegate:1; + CGroupMask cgroup_supported; + const char *cgroup_path; + + const char *runtime_prefix; + + usec_t watchdog_usec; + + int *idle_pipe; + + int stdin_fd; + int stdout_fd; + int stderr_fd; +}; + +#include "dynamic-user.h" +#include "unit.h" + +int exec_spawn(Unit *unit, + ExecCommand *command, + const ExecContext *context, + const ExecParameters *exec_params, + ExecRuntime *runtime, + DynamicCreds *dynamic_creds, + pid_t *ret); + +void exec_command_done(ExecCommand *c); +void exec_command_done_array(ExecCommand *c, unsigned n); + +ExecCommand* exec_command_free_list(ExecCommand *c); +void exec_command_free_array(ExecCommand **c, unsigned n); + +char *exec_command_line(char **argv); + +void exec_command_dump(ExecCommand *c, FILE *f, const char *prefix); +void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix); +void exec_command_append_list(ExecCommand **l, ExecCommand *e); +int exec_command_set(ExecCommand *c, const char *path, ...); +int exec_command_append(ExecCommand *c, const char *path, ...); + +void exec_context_init(ExecContext *c); +void exec_context_done(ExecContext *c); +void exec_context_dump(ExecContext *c, FILE* f, const char *prefix); + +int exec_context_destroy_runtime_directory(ExecContext *c, const char *runtime_root); + +int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l); +int exec_context_named_iofds(Unit *unit, const ExecContext *c, const ExecParameters *p, int named_iofds[3]); +const char* exec_context_fdname(const ExecContext *c, int fd_index); + +bool exec_context_may_touch_console(ExecContext *c); +bool exec_context_maintains_privileges(ExecContext *c); + +void exec_status_start(ExecStatus *s, pid_t pid); +void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, int status); +void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix); + +int exec_runtime_make(ExecRuntime **rt, ExecContext *c, const char *id); +ExecRuntime *exec_runtime_ref(ExecRuntime *r); +ExecRuntime *exec_runtime_unref(ExecRuntime *r); + +int exec_runtime_serialize(Unit *unit, ExecRuntime *rt, FILE *f, FDSet *fds); +int exec_runtime_deserialize_item(Unit *unit, ExecRuntime **rt, const char *key, const char *value, FDSet *fds); + +void exec_runtime_destroy(ExecRuntime *rt); + +const char* exec_output_to_string(ExecOutput i) _const_; +ExecOutput exec_output_from_string(const char *s) _pure_; + +const char* exec_input_to_string(ExecInput i) _const_; +ExecInput exec_input_from_string(const char *s) _pure_; + +const char* exec_utmp_mode_to_string(ExecUtmpMode i) _const_; +ExecUtmpMode exec_utmp_mode_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/hostname-setup.h b/src/grp-system/libcore/include/core/hostname-setup.h new file mode 100644 index 0000000000..73e8c75c71 --- /dev/null +++ b/src/grp-system/libcore/include/core/hostname-setup.h @@ -0,0 +1,22 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int hostname_setup(void); diff --git a/src/grp-system/libcore/include/core/ima-setup.h b/src/grp-system/libcore/include/core/ima-setup.h new file mode 100644 index 0000000000..472b58cb00 --- /dev/null +++ b/src/grp-system/libcore/include/core/ima-setup.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright (C) 2012 Roberto Sassu - Politecnico di Torino, Italy + TORSEC group — http://security.polito.it + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int ima_setup(void); diff --git a/src/grp-system/libcore/include/core/job.h b/src/grp-system/libcore/include/core/job.h new file mode 100644 index 0000000000..3e62465695 --- /dev/null +++ b/src/grp-system/libcore/include/core/job.h @@ -0,0 +1,242 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include <systemd/sd-event.h> + +#include "systemd-basic/list.h" +#include "systemd-basic/unit-name.h" + +typedef struct Job Job; +typedef struct JobDependency JobDependency; +typedef enum JobType JobType; +typedef enum JobState JobState; +typedef enum JobMode JobMode; +typedef enum JobResult JobResult; + +/* Be careful when changing the job types! Adjust job_merging_table[] accordingly! */ +enum JobType { + JOB_START, /* if a unit does not support being started, we'll just wait until it becomes active */ + JOB_VERIFY_ACTIVE, + + JOB_STOP, + + JOB_RELOAD, /* if running, reload */ + + /* Note that restarts are first treated like JOB_STOP, but + * then instead of finishing are patched to become + * JOB_START. */ + JOB_RESTART, /* If running, stop. Then start unconditionally. */ + + _JOB_TYPE_MAX_MERGING, + + /* JOB_NOP can enter into a transaction, but as it won't pull in + * any dependencies and it uses the special 'nop_job' slot in Unit, + * it won't have to merge with anything (except possibly into another + * JOB_NOP, previously installed). JOB_NOP is special-cased in + * job_type_is_*() functions so that the transaction can be + * activated. */ + JOB_NOP = _JOB_TYPE_MAX_MERGING, /* do nothing */ + + _JOB_TYPE_MAX_IN_TRANSACTION, + + /* JOB_TRY_RESTART can never appear in a transaction, because + * it always collapses into JOB_RESTART or JOB_NOP before entering. + * Thus we never need to merge it with anything. */ + JOB_TRY_RESTART = _JOB_TYPE_MAX_IN_TRANSACTION, /* if running, stop and then start */ + + /* Similar to JOB_TRY_RESTART but collapses to JOB_RELOAD or JOB_NOP */ + JOB_TRY_RELOAD, + + /* JOB_RELOAD_OR_START won't enter into a transaction and cannot result + * from transaction merging (there's no way for JOB_RELOAD and + * JOB_START to meet in one transaction). It can result from a merge + * during job installation, but then it will immediately collapse into + * one of the two simpler types. */ + JOB_RELOAD_OR_START, /* if running, reload, otherwise start */ + + _JOB_TYPE_MAX, + _JOB_TYPE_INVALID = -1 +}; + +enum JobState { + JOB_WAITING, + JOB_RUNNING, + _JOB_STATE_MAX, + _JOB_STATE_INVALID = -1 +}; + +enum JobMode { + JOB_FAIL, /* Fail if a conflicting job is already queued */ + JOB_REPLACE, /* Replace an existing conflicting job */ + JOB_REPLACE_IRREVERSIBLY,/* Like JOB_REPLACE + produce irreversible jobs */ + JOB_ISOLATE, /* Start a unit, and stop all others */ + JOB_FLUSH, /* Flush out all other queued jobs when queing this one */ + JOB_IGNORE_DEPENDENCIES, /* Ignore both requirement and ordering dependencies */ + JOB_IGNORE_REQUIREMENTS, /* Ignore requirement dependencies */ + _JOB_MODE_MAX, + _JOB_MODE_INVALID = -1 +}; + +enum JobResult { + JOB_DONE, /* Job completed successfully */ + JOB_CANCELED, /* Job canceled by a conflicting job installation or by explicit cancel request */ + JOB_TIMEOUT, /* Job timeout elapsed */ + JOB_FAILED, /* Job failed */ + JOB_DEPENDENCY, /* A required dependency job did not result in JOB_DONE */ + JOB_SKIPPED, /* Negative result of JOB_VERIFY_ACTIVE */ + JOB_INVALID, /* JOB_RELOAD of inactive unit */ + JOB_ASSERT, /* Couldn't start a unit, because an assert didn't hold */ + JOB_UNSUPPORTED, /* Couldn't start a unit, because the unit type is not supported on the system */ + _JOB_RESULT_MAX, + _JOB_RESULT_INVALID = -1 +}; + +#include "unit.h" + +struct JobDependency { + /* Encodes that the 'subject' job needs the 'object' job in + * some way. This structure is used only while building a transaction. */ + Job *subject; + Job *object; + + LIST_FIELDS(JobDependency, subject); + LIST_FIELDS(JobDependency, object); + + bool matters; + bool conflicts; +}; + +struct Job { + Manager *manager; + Unit *unit; + + LIST_FIELDS(Job, transaction); + LIST_FIELDS(Job, run_queue); + LIST_FIELDS(Job, dbus_queue); + + LIST_HEAD(JobDependency, subject_list); + LIST_HEAD(JobDependency, object_list); + + /* Used for graph algs as a "I have been here" marker */ + Job* marker; + unsigned generation; + + uint32_t id; + + JobType type; + JobState state; + + sd_event_source *timer_event_source; + usec_t begin_usec; + + /* + * This tracks where to send signals, and also which clients + * are allowed to call DBus methods on the job (other than + * root). + * + * There can be more than one client, because of job merging. + */ + sd_bus_track *clients; + char **deserialized_clients; + + JobResult result; + + bool installed:1; + bool in_run_queue:1; + bool matters_to_anchor:1; + bool in_dbus_queue:1; + bool sent_dbus_new_signal:1; + bool ignore_order:1; + bool irreversible:1; +}; + +Job* job_new(Unit *unit, JobType type); +Job* job_new_raw(Unit *unit); +void job_free(Job *job); +Job* job_install(Job *j); +int job_install_deserialized(Job *j); +void job_uninstall(Job *j); +void job_dump(Job *j, FILE*f, const char *prefix); +int job_serialize(Job *j, FILE *f); +int job_deserialize(Job *j, FILE *f); +int job_coldplug(Job *j); + +JobDependency* job_dependency_new(Job *subject, Job *object, bool matters, bool conflicts); +void job_dependency_free(JobDependency *l); + +int job_merge(Job *j, Job *other); + +JobType job_type_lookup_merge(JobType a, JobType b) _pure_; + +_pure_ static inline bool job_type_is_mergeable(JobType a, JobType b) { + return job_type_lookup_merge(a, b) >= 0; +} + +_pure_ static inline bool job_type_is_conflicting(JobType a, JobType b) { + return a != JOB_NOP && b != JOB_NOP && !job_type_is_mergeable(a, b); +} + +_pure_ static inline bool job_type_is_superset(JobType a, JobType b) { + /* Checks whether operation a is a "superset" of b in its actions */ + if (b == JOB_NOP) + return true; + if (a == JOB_NOP) + return false; + return a == job_type_lookup_merge(a, b); +} + +bool job_type_is_redundant(JobType a, UnitActiveState b) _pure_; + +/* Collapses a state-dependent job type into a simpler type by observing + * the state of the unit which it is going to be applied to. */ +JobType job_type_collapse(JobType t, Unit *u); + +int job_type_merge_and_collapse(JobType *a, JobType b, Unit *u); + +void job_add_to_run_queue(Job *j); +void job_add_to_dbus_queue(Job *j); + +int job_start_timer(Job *j); + +int job_run_and_invalidate(Job *j); +int job_finish_and_invalidate(Job *j, JobResult result, bool recursive, bool already); + +char *job_dbus_path(Job *j); + +void job_shutdown_magic(Job *j); + +int job_get_timeout(Job *j, usec_t *timeout) _pure_; + +const char* job_type_to_string(JobType t) _const_; +JobType job_type_from_string(const char *s) _pure_; + +const char* job_state_to_string(JobState t) _const_; +JobState job_state_from_string(const char *s) _pure_; + +const char* job_mode_to_string(JobMode t) _const_; +JobMode job_mode_from_string(const char *s) _pure_; + +const char* job_result_to_string(JobResult t) _const_; +JobResult job_result_from_string(const char *s) _pure_; + +const char* job_type_to_access_method(JobType t); diff --git a/src/grp-system/libcore/include/core/kill.h b/src/grp-system/libcore/include/core/kill.h new file mode 100644 index 0000000000..ad8583b9b0 --- /dev/null +++ b/src/grp-system/libcore/include/core/kill.h @@ -0,0 +1,65 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <stdio.h> + +#include "systemd-basic/macro.h" + +typedef struct KillContext KillContext; + +typedef enum KillMode { + /* The kill mode is a property of a unit. */ + KILL_CONTROL_GROUP = 0, + KILL_PROCESS, + KILL_MIXED, + KILL_NONE, + _KILL_MODE_MAX, + _KILL_MODE_INVALID = -1 +} KillMode; + +struct KillContext { + KillMode kill_mode; + int kill_signal; + bool send_sigkill; + bool send_sighup; +}; + +typedef enum KillWho { + /* Kill who is a property of an operation */ + KILL_MAIN, + KILL_CONTROL, + KILL_ALL, + KILL_MAIN_FAIL, + KILL_CONTROL_FAIL, + KILL_ALL_FAIL, + _KILL_WHO_MAX, + _KILL_WHO_INVALID = -1 +} KillWho; + +void kill_context_init(KillContext *c); +void kill_context_dump(KillContext *c, FILE *f, const char *prefix); + +const char *kill_mode_to_string(KillMode k) _const_; +KillMode kill_mode_from_string(const char *s) _pure_; + +const char *kill_who_to_string(KillWho k) _const_; +KillWho kill_who_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/killall.h b/src/grp-system/libcore/include/core/killall.h new file mode 100644 index 0000000000..acc2439f00 --- /dev/null +++ b/src/grp-system/libcore/include/core/killall.h @@ -0,0 +1,22 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +void broadcast_signal(int sig, bool wait_for_exit, bool send_sighup); diff --git a/src/grp-system/libcore/include/core/kmod-setup.h b/src/grp-system/libcore/include/core/kmod-setup.h new file mode 100644 index 0000000000..685f4df301 --- /dev/null +++ b/src/grp-system/libcore/include/core/kmod-setup.h @@ -0,0 +1,22 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int kmod_setup(void); diff --git a/src/grp-system/libcore/include/core/load-fragment.h b/src/grp-system/libcore/include/core/load-fragment.h new file mode 100644 index 0000000000..ede6b1f735 --- /dev/null +++ b/src/grp-system/libcore/include/core/load-fragment.h @@ -0,0 +1,128 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "unit.h" + +/* Read service data from .desktop file style configuration fragments */ + +int unit_load_fragment(Unit *u); + +void unit_dump_config_items(FILE *f); + +int config_parse_warn_compat(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_deps(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_obsolete_unit_deps(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_string_printf(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_strv_printf(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_path_printf(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_path_strv_printf(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_documentation(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_listen(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_protocol(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_bind(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_nice(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_oom_score_adjust(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_timeout(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_type(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_restart(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_bindtodevice(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_output(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_output(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_input(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_input(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_io_class(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_io_priority(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_sched_policy(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_sched_prio(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_cpu_affinity(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_secure_bits(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_capability_set(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_limit(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_sysv_priority(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_kill_signal(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_mount_flags(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_timer(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_trigger_unit(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_path_spec(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_socket_service(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_service_sockets(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_busname_service(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_bus_policy(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_bus_policy_world(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_env_file(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_ip_tos(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_path(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_string(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_condition_null(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_kill_mode(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_notify_access(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_emergency_action(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_requires_mounts_for(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_syscall_filter(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_syscall_archs(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_syscall_errno(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_environ(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_pass_environ(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_unit_slice(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_cpu_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_cpu_shares(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_memory_limit(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_tasks_max(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_device_policy(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_device_allow(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_io_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_io_device_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_io_limit(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_blockio_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_blockio_device_weight(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_blockio_bandwidth(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_netclass(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_job_mode(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_job_mode_isolate(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_selinux_context(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_apparmor_profile(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_smack_process_label(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_address_families(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_runtime_directory(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_set_status(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_namespace_path_strv(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_no_new_privileges(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_cpu_quota(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_protect_home(const char* unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_protect_system(const char* unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_bus_name(const char* unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_exec_utmp_mode(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_working_directory(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_fdname(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_sec_fix_0(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_user_group(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); +int config_parse_user_group_strv(const char *unit, const char *filename, unsigned line, const char *section, unsigned section_line, const char *lvalue, int ltype, const char *rvalue, void *data, void *userdata); + +/* gperf prototypes */ +const struct ConfigPerfItem* load_fragment_gperf_lookup(const char *key, GPERF_LEN_TYPE length); +extern const char load_fragment_gperf_nulstr[]; + +typedef enum Disabled { + DISABLED_CONFIGURATION, + DISABLED_LEGACY, + DISABLED_EXPERIMENTAL, +} Disabled; diff --git a/src/grp-system/libcore/include/core/loopback-setup.h b/src/grp-system/libcore/include/core/loopback-setup.h new file mode 100644 index 0000000000..e7547b8a26 --- /dev/null +++ b/src/grp-system/libcore/include/core/loopback-setup.h @@ -0,0 +1,22 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int loopback_setup(void); diff --git a/src/grp-system/libcore/include/core/manager.h b/src/grp-system/libcore/include/core/manager.h new file mode 100644 index 0000000000..429b86a94c --- /dev/null +++ b/src/grp-system/libcore/include/core/manager.h @@ -0,0 +1,406 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <libmount.h> +#include <stdbool.h> +#include <stdio.h> + +#include <systemd/sd-bus.h> +#include <systemd/sd-event.h> + +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/hashmap.h" +#include "systemd-basic/list.h" +#include "systemd-basic/ratelimit.h" +#include "systemd-shared/fdset.h" + +/* Enforce upper limit how many names we allow */ +#define MANAGER_MAX_NAMES 131072 /* 128K */ + +typedef struct Manager Manager; + +typedef enum ManagerState { + MANAGER_INITIALIZING, + MANAGER_STARTING, + MANAGER_RUNNING, + MANAGER_DEGRADED, + MANAGER_MAINTENANCE, + MANAGER_STOPPING, + _MANAGER_STATE_MAX, + _MANAGER_STATE_INVALID = -1 +} ManagerState; + +typedef enum ManagerExitCode { + MANAGER_OK, + MANAGER_EXIT, + MANAGER_RELOAD, + MANAGER_REEXECUTE, + MANAGER_REBOOT, + MANAGER_POWEROFF, + MANAGER_HALT, + MANAGER_KEXEC, + MANAGER_SWITCH_ROOT, + _MANAGER_EXIT_CODE_MAX, + _MANAGER_EXIT_CODE_INVALID = -1 +} ManagerExitCode; + +typedef enum StatusType { + STATUS_TYPE_EPHEMERAL, + STATUS_TYPE_NORMAL, + STATUS_TYPE_EMERGENCY, +} StatusType; + +#include "systemd-basic/unit-name.h" +#include "systemd-shared/path-lookup.h" + +#include "execute.h" +#include "job.h" +#include "show-status.h" + +struct Manager { + /* Note that the set of units we know of is allowed to be + * inconsistent. However the subset of it that is loaded may + * not, and the list of jobs may neither. */ + + /* Active jobs and units */ + Hashmap *units; /* name string => Unit object n:1 */ + Hashmap *units_by_invocation_id; + Hashmap *jobs; /* job id => Job object 1:1 */ + + /* To make it easy to iterate through the units of a specific + * type we maintain a per type linked list */ + LIST_HEAD(Unit, units_by_type[_UNIT_TYPE_MAX]); + + /* Units that need to be loaded */ + LIST_HEAD(Unit, load_queue); /* this is actually more a stack than a queue, but uh. */ + + /* Jobs that need to be run */ + LIST_HEAD(Job, run_queue); /* more a stack than a queue, too */ + + /* Units and jobs that have not yet been announced via + * D-Bus. When something about a job changes it is added here + * if it is not in there yet. This allows easy coalescing of + * D-Bus change signals. */ + LIST_HEAD(Unit, dbus_unit_queue); + LIST_HEAD(Job, dbus_job_queue); + + /* Units to remove */ + LIST_HEAD(Unit, cleanup_queue); + + /* Units to check when doing GC */ + LIST_HEAD(Unit, gc_queue); + + /* Units that should be realized */ + LIST_HEAD(Unit, cgroup_queue); + + sd_event *event; + + /* We use two hash tables here, since the same PID might be + * watched by two different units: once the unit that forked + * it off, and possibly a different unit to which it was + * joined as cgroup member. Since we know that it is either + * one or two units for each PID we just use to hashmaps + * here. */ + Hashmap *watch_pids1; /* pid => Unit object n:1 */ + Hashmap *watch_pids2; /* pid => Unit object n:1 */ + + /* A set contains all units which cgroup should be refreshed after startup */ + Set *startup_units; + + /* A set which contains all currently failed units */ + Set *failed_units; + + sd_event_source *run_queue_event_source; + + char *notify_socket; + int notify_fd; + sd_event_source *notify_event_source; + + int cgroups_agent_fd; + sd_event_source *cgroups_agent_event_source; + + int signal_fd; + sd_event_source *signal_event_source; + + int time_change_fd; + sd_event_source *time_change_event_source; + + sd_event_source *jobs_in_progress_event_source; + + int user_lookup_fds[2]; + sd_event_source *user_lookup_event_source; + + UnitFileScope unit_file_scope; + LookupPaths lookup_paths; + Set *unit_path_cache; + + char **environment; + + usec_t runtime_watchdog; + usec_t shutdown_watchdog; + + dual_timestamp firmware_timestamp; + dual_timestamp loader_timestamp; + dual_timestamp kernel_timestamp; + dual_timestamp initrd_timestamp; + dual_timestamp userspace_timestamp; + dual_timestamp finish_timestamp; + + dual_timestamp security_start_timestamp; + dual_timestamp security_finish_timestamp; + dual_timestamp generators_start_timestamp; + dual_timestamp generators_finish_timestamp; + dual_timestamp units_load_start_timestamp; + dual_timestamp units_load_finish_timestamp; + + struct udev* udev; + + /* Data specific to the device subsystem */ + struct udev_monitor* udev_monitor; + sd_event_source *udev_event_source; + Hashmap *devices_by_sysfs; + + /* Data specific to the mount subsystem */ + struct libmnt_monitor *mount_monitor; + sd_event_source *mount_event_source; + + /* Data specific to the swap filesystem */ + FILE *proc_swaps; + sd_event_source *swap_event_source; + Hashmap *swaps_by_devnode; + + /* Data specific to the D-Bus subsystem */ + sd_bus *api_bus, *system_bus; + Set *private_buses; + int private_listen_fd; + sd_event_source *private_listen_event_source; + + /* Contains all the clients that are subscribed to signals via + the API bus. Note that private bus connections are always + considered subscribes, since they last for very short only, + and it is much simpler that way. */ + sd_bus_track *subscribed; + char **deserialized_subscribed; + + /* This is used during reloading: before the reload we queue + * the reply message here, and afterwards we send it */ + sd_bus_message *queued_message; + + Hashmap *watch_bus; /* D-Bus names => Unit object n:1 */ + + bool send_reloading_done; + + uint32_t current_job_id; + uint32_t default_unit_job_id; + + /* Data specific to the Automount subsystem */ + int dev_autofs_fd; + + /* Data specific to the cgroup subsystem */ + Hashmap *cgroup_unit; + CGroupMask cgroup_supported; + char *cgroup_root; + + /* Notifications from cgroups, when the unified hierarchy is + * used is done via inotify. */ + int cgroup_inotify_fd; + sd_event_source *cgroup_inotify_event_source; + Hashmap *cgroup_inotify_wd_unit; + + /* Make sure the user cannot accidentally unmount our cgroup + * file system */ + int pin_cgroupfs_fd; + + int gc_marker; + unsigned n_in_gc_queue; + + /* Flags */ + ManagerExitCode exit_code:5; + + bool dispatching_load_queue:1; + bool dispatching_dbus_queue:1; + + bool taint_usr:1; + bool test_run:1; + + /* If non-zero, exit with the following value when the systemd + * process terminate. Useful for containers: systemd-nspawn could get + * the return value. */ + uint8_t return_value; + + ShowStatus show_status; + bool confirm_spawn; + bool no_console_output; + + ExecOutput default_std_output, default_std_error; + + usec_t default_restart_usec, default_timeout_start_usec, default_timeout_stop_usec; + + usec_t default_start_limit_interval; + unsigned default_start_limit_burst; + + bool default_cpu_accounting; + bool default_memory_accounting; + bool default_io_accounting; + bool default_blockio_accounting; + bool default_tasks_accounting; + + uint64_t default_tasks_max; + usec_t default_timer_accuracy_usec; + + struct rlimit *rlimit[_RLIMIT_MAX]; + + /* non-zero if we are reloading or reexecuting, */ + int n_reloading; + + unsigned n_installed_jobs; + unsigned n_failed_jobs; + + /* Jobs in progress watching */ + unsigned n_running_jobs; + unsigned n_on_console; + unsigned jobs_in_progress_iteration; + + /* Do we have any outstanding password prompts? */ + int have_ask_password; + int ask_password_inotify_fd; + sd_event_source *ask_password_event_source; + + /* Type=idle pipes */ + int idle_pipe[4]; + sd_event_source *idle_pipe_event_source; + + char *switch_root; + char *switch_root_init; + + /* This maps all possible path prefixes to the units needing + * them. It's a hashmap with a path string as key and a Set as + * value where Unit objects are contained. */ + Hashmap *units_requiring_mounts_for; + + /* Used for processing polkit authorization responses */ + Hashmap *polkit_registry; + + /* Dynamic users/groups, indexed by their name */ + Hashmap *dynamic_users; + + /* Keep track of all UIDs and GIDs any of our services currently use. This is useful for the RemoveIPC= logic. */ + Hashmap *uid_refs; + Hashmap *gid_refs; + + /* When the user hits C-A-D more than 7 times per 2s, do something immediately... */ + RateLimit ctrl_alt_del_ratelimit; + EmergencyAction cad_burst_action; + + const char *unit_log_field; + const char *unit_log_format_string; + + const char *invocation_log_field; + const char *invocation_log_format_string; + + int first_boot; /* tri-state */ +}; + +#define MANAGER_IS_SYSTEM(m) ((m)->unit_file_scope == UNIT_FILE_SYSTEM) +#define MANAGER_IS_USER(m) ((m)->unit_file_scope != UNIT_FILE_SYSTEM) + +#define MANAGER_IS_RELOADING(m) ((m)->n_reloading > 0) + +int manager_new(UnitFileScope scope, bool test_run, Manager **m); +Manager* manager_free(Manager *m); + +void manager_enumerate(Manager *m); +int manager_startup(Manager *m, FILE *serialization, FDSet *fds); + +Job *manager_get_job(Manager *m, uint32_t id); +Unit *manager_get_unit(Manager *m, const char *name); + +int manager_get_job_from_dbus_path(Manager *m, const char *s, Job **_j); + +int manager_load_unit_prepare(Manager *m, const char *name, const char *path, sd_bus_error *e, Unit **_ret); +int manager_load_unit(Manager *m, const char *name, const char *path, sd_bus_error *e, Unit **_ret); +int manager_load_unit_from_dbus_path(Manager *m, const char *s, sd_bus_error *e, Unit **_u); + +int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, sd_bus_error *e, Job **_ret); +int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, sd_bus_error *e, Job **_ret); +int manager_add_job_by_name_and_warn(Manager *m, JobType type, const char *name, JobMode mode, Job **ret); + +void manager_dump_units(Manager *s, FILE *f, const char *prefix); +void manager_dump_jobs(Manager *s, FILE *f, const char *prefix); + +void manager_clear_jobs(Manager *m); + +unsigned manager_dispatch_load_queue(Manager *m); + +int manager_environment_add(Manager *m, char **minus, char **plus); +int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit); + +int manager_loop(Manager *m); + +int manager_open_serialization(Manager *m, FILE **_f); + +int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root); +int manager_deserialize(Manager *m, FILE *f, FDSet *fds); + +int manager_reload(Manager *m); + +void manager_reset_failed(Manager *m); + +void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success); +void manager_send_unit_plymouth(Manager *m, Unit *u); + +bool manager_unit_inactive_or_pending(Manager *m, const char *name); + +void manager_check_finished(Manager *m); + +void manager_recheck_journal(Manager *m); + +void manager_set_show_status(Manager *m, ShowStatus mode); +void manager_set_first_boot(Manager *m, bool b); + +void manager_status_printf(Manager *m, StatusType type, const char *status, const char *format, ...) _printf_(4,5); +void manager_flip_auto_status(Manager *m, bool enable); + +Set *manager_get_units_requiring_mounts_for(Manager *m, const char *path); + +const char *manager_get_runtime_prefix(Manager *m); + +ManagerState manager_state(Manager *m); + +int manager_update_failed_units(Manager *m, Unit *u, bool failed); + +void manager_unref_uid(Manager *m, uid_t uid, bool destroy_now); +int manager_ref_uid(Manager *m, uid_t uid, bool clean_ipc); + +void manager_unref_gid(Manager *m, gid_t gid, bool destroy_now); +int manager_ref_gid(Manager *m, gid_t gid, bool destroy_now); + +void manager_vacuum_uid_refs(Manager *m); +void manager_vacuum_gid_refs(Manager *m); + +void manager_serialize_uid_refs(Manager *m, FILE *f); +void manager_deserialize_uid_refs_one(Manager *m, const char *value); + +void manager_serialize_gid_refs(Manager *m, FILE *f); +void manager_deserialize_gid_refs_one(Manager *m, const char *value); + +const char *manager_state_to_string(ManagerState m) _const_; +ManagerState manager_state_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/mount-setup.h b/src/grp-system/libcore/include/core/mount-setup.h new file mode 100644 index 0000000000..647bd770ae --- /dev/null +++ b/src/grp-system/libcore/include/core/mount-setup.h @@ -0,0 +1,30 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +int mount_setup_early(void); +int mount_setup(bool loaded_policy); + +int mount_cgroup_controllers(char ***join_controllers); + +bool mount_point_is_api(const char *path); +bool mount_point_ignore(const char *path); diff --git a/src/grp-system/libcore/include/core/mount.h b/src/grp-system/libcore/include/core/mount.h new file mode 100644 index 0000000000..148fedf354 --- /dev/null +++ b/src/grp-system/libcore/include/core/mount.h @@ -0,0 +1,112 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Mount Mount; + +#include "dynamic-user.h" +#include "kill.h" + +typedef enum MountExecCommand { + MOUNT_EXEC_MOUNT, + MOUNT_EXEC_UNMOUNT, + MOUNT_EXEC_REMOUNT, + _MOUNT_EXEC_COMMAND_MAX, + _MOUNT_EXEC_COMMAND_INVALID = -1 +} MountExecCommand; + +typedef enum MountResult { + MOUNT_SUCCESS, + MOUNT_FAILURE_RESOURCES, + MOUNT_FAILURE_TIMEOUT, + MOUNT_FAILURE_EXIT_CODE, + MOUNT_FAILURE_SIGNAL, + MOUNT_FAILURE_CORE_DUMP, + MOUNT_FAILURE_START_LIMIT_HIT, + _MOUNT_RESULT_MAX, + _MOUNT_RESULT_INVALID = -1 +} MountResult; + +typedef struct MountParameters { + char *what; + char *options; + char *fstype; +} MountParameters; + +struct Mount { + Unit meta; + + char *where; + + MountParameters parameters_proc_self_mountinfo; + MountParameters parameters_fragment; + + bool from_proc_self_mountinfo:1; + bool from_fragment:1; + + /* Used while looking for mount points that vanished or got + * added from/to /proc/self/mountinfo */ + bool is_mounted:1; + bool just_mounted:1; + bool just_changed:1; + + bool reset_cpu_usage:1; + + bool sloppy_options; + + bool lazy_unmount; + bool force_unmount; + + MountResult result; + MountResult reload_result; + + mode_t directory_mode; + + usec_t timeout_usec; + + ExecCommand exec_command[_MOUNT_EXEC_COMMAND_MAX]; + + ExecContext exec_context; + KillContext kill_context; + CGroupContext cgroup_context; + + ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; + + MountState state, deserialized_state; + + ExecCommand* control_command; + MountExecCommand control_command_id; + pid_t control_pid; + + sd_event_source *timer_event_source; + + unsigned n_retry_umount; +}; + +extern const UnitVTable mount_vtable; + +void mount_fd_event(Manager *m, int events); + +const char* mount_exec_command_to_string(MountExecCommand i) _const_; +MountExecCommand mount_exec_command_from_string(const char *s) _pure_; + +const char* mount_result_to_string(MountResult i) _const_; +MountResult mount_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/namespace.h b/src/grp-system/libcore/include/core/namespace.h new file mode 100644 index 0000000000..8e80e2f38e --- /dev/null +++ b/src/grp-system/libcore/include/core/namespace.h @@ -0,0 +1,74 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2016 Djalal Harouni + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "systemd-basic/macro.h" + +typedef struct NameSpaceInfo NameSpaceInfo; + +typedef enum ProtectHome { + PROTECT_HOME_NO, + PROTECT_HOME_YES, + PROTECT_HOME_READ_ONLY, + _PROTECT_HOME_MAX, + _PROTECT_HOME_INVALID = -1 +} ProtectHome; + +typedef enum ProtectSystem { + PROTECT_SYSTEM_NO, + PROTECT_SYSTEM_YES, + PROTECT_SYSTEM_FULL, + PROTECT_SYSTEM_STRICT, + _PROTECT_SYSTEM_MAX, + _PROTECT_SYSTEM_INVALID = -1 +} ProtectSystem; + +struct NameSpaceInfo { + bool private_dev:1; + bool protect_control_groups:1; + bool protect_kernel_tunables:1; + bool protect_kernel_modules:1; +}; + +int setup_namespace(const char *chroot, + const NameSpaceInfo *ns_info, + char **read_write_paths, + char **read_only_paths, + char **inaccessible_paths, + const char *tmp_dir, + const char *var_tmp_dir, + ProtectHome protect_home, + ProtectSystem protect_system, + unsigned long mount_flags); + +int setup_tmp_dirs(const char *id, + char **tmp_dir, + char **var_tmp_dir); + +int setup_netns(int netns_storage_socket[2]); + +const char* protect_home_to_string(ProtectHome p) _const_; +ProtectHome protect_home_from_string(const char *s) _pure_; + +const char* protect_system_to_string(ProtectSystem p) _const_; +ProtectSystem protect_system_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/path.h b/src/grp-system/libcore/include/core/path.h new file mode 100644 index 0000000000..4230c8fb99 --- /dev/null +++ b/src/grp-system/libcore/include/core/path.h @@ -0,0 +1,93 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Path Path; +typedef struct PathSpec PathSpec; + +#include "unit.h" + +typedef enum PathType { + PATH_EXISTS, + PATH_EXISTS_GLOB, + PATH_DIRECTORY_NOT_EMPTY, + PATH_CHANGED, + PATH_MODIFIED, + _PATH_TYPE_MAX, + _PATH_TYPE_INVALID = -1 +} PathType; + +typedef struct PathSpec { + Unit *unit; + + char *path; + + sd_event_source *event_source; + + LIST_FIELDS(struct PathSpec, spec); + + PathType type; + int inotify_fd; + int primary_wd; + + bool previous_exists; +} PathSpec; + +int path_spec_watch(PathSpec *s, sd_event_io_handler_t handler); +void path_spec_unwatch(PathSpec *s); +int path_spec_fd_event(PathSpec *s, uint32_t events); +void path_spec_done(PathSpec *s); + +static inline bool path_spec_owns_inotify_fd(PathSpec *s, int fd) { + return s->inotify_fd == fd; +} + +typedef enum PathResult { + PATH_SUCCESS, + PATH_FAILURE_RESOURCES, + PATH_FAILURE_START_LIMIT_HIT, + _PATH_RESULT_MAX, + _PATH_RESULT_INVALID = -1 +} PathResult; + +struct Path { + Unit meta; + + LIST_HEAD(PathSpec, specs); + + PathState state, deserialized_state; + + bool inotify_triggered; + + bool make_directory; + mode_t directory_mode; + + PathResult result; +}; + +void path_free_specs(Path *p); + +extern const UnitVTable path_vtable; + +const char* path_type_to_string(PathType i) _const_; +PathType path_type_from_string(const char *s) _pure_; + +const char* path_result_to_string(PathResult i) _const_; +PathResult path_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/scope.h b/src/grp-system/libcore/include/core/scope.h new file mode 100644 index 0000000000..eaf8e8b447 --- /dev/null +++ b/src/grp-system/libcore/include/core/scope.h @@ -0,0 +1,58 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Scope Scope; + +#include "cgroup.h" +#include "kill.h" +#include "unit.h" + +typedef enum ScopeResult { + SCOPE_SUCCESS, + SCOPE_FAILURE_RESOURCES, + SCOPE_FAILURE_TIMEOUT, + _SCOPE_RESULT_MAX, + _SCOPE_RESULT_INVALID = -1 +} ScopeResult; + +struct Scope { + Unit meta; + + CGroupContext cgroup_context; + KillContext kill_context; + + ScopeState state, deserialized_state; + ScopeResult result; + + usec_t timeout_stop_usec; + + char *controller; + bool was_abandoned; + + sd_event_source *timer_event_source; +}; + +extern const UnitVTable scope_vtable; + +int scope_abandon(Scope *s); + +const char* scope_result_to_string(ScopeResult i) _const_; +ScopeResult scope_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/selinux-setup.h b/src/grp-system/libcore/include/core/selinux-setup.h new file mode 100644 index 0000000000..7b613249b0 --- /dev/null +++ b/src/grp-system/libcore/include/core/selinux-setup.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +int mac_selinux_setup(bool *loaded_policy); diff --git a/src/grp-system/libcore/include/core/service.h b/src/grp-system/libcore/include/core/service.h new file mode 100644 index 0000000000..1ebf62eed6 --- /dev/null +++ b/src/grp-system/libcore/include/core/service.h @@ -0,0 +1,225 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "systemd-basic/exit-status.h" +#include "systemd-basic/ratelimit.h" + +typedef struct Service Service; +typedef struct ServiceFDStore ServiceFDStore; + +#include "kill.h" +#include "path.h" + +typedef enum ServiceRestart { + SERVICE_RESTART_NO, + SERVICE_RESTART_ON_SUCCESS, + SERVICE_RESTART_ON_FAILURE, + SERVICE_RESTART_ON_ABNORMAL, + SERVICE_RESTART_ON_WATCHDOG, + SERVICE_RESTART_ON_ABORT, + SERVICE_RESTART_ALWAYS, + _SERVICE_RESTART_MAX, + _SERVICE_RESTART_INVALID = -1 +} ServiceRestart; + +typedef enum ServiceType { + SERVICE_SIMPLE, /* we fork and go on right-away (i.e. modern socket activated daemons) */ + SERVICE_FORKING, /* forks by itself (i.e. traditional daemons) */ + SERVICE_ONESHOT, /* we fork and wait until the program finishes (i.e. programs like fsck which run and need to finish before we continue) */ + SERVICE_DBUS, /* we fork and wait until a specific D-Bus name appears on the bus */ + SERVICE_NOTIFY, /* we fork and wait until a daemon sends us a ready message with sd_notify() */ + SERVICE_IDLE, /* much like simple, but delay exec() until all jobs are dispatched. */ + _SERVICE_TYPE_MAX, + _SERVICE_TYPE_INVALID = -1 +} ServiceType; + +typedef enum ServiceExecCommand { + SERVICE_EXEC_START_PRE, + SERVICE_EXEC_START, + SERVICE_EXEC_START_POST, + SERVICE_EXEC_RELOAD, + SERVICE_EXEC_STOP, + SERVICE_EXEC_STOP_POST, + _SERVICE_EXEC_COMMAND_MAX, + _SERVICE_EXEC_COMMAND_INVALID = -1 +} ServiceExecCommand; + +typedef enum NotifyAccess { + NOTIFY_NONE, + NOTIFY_ALL, + NOTIFY_MAIN, + _NOTIFY_ACCESS_MAX, + _NOTIFY_ACCESS_INVALID = -1 +} NotifyAccess; + +typedef enum NotifyState { + NOTIFY_UNKNOWN, + NOTIFY_READY, + NOTIFY_RELOADING, + NOTIFY_STOPPING, + _NOTIFY_STATE_MAX, + _NOTIFY_STATE_INVALID = -1 +} NotifyState; + +typedef enum ServiceResult { + SERVICE_SUCCESS, + SERVICE_FAILURE_RESOURCES, /* a bit of a misnomer, just our catch-all error for errnos we didn't expect */ + SERVICE_FAILURE_TIMEOUT, + SERVICE_FAILURE_EXIT_CODE, + SERVICE_FAILURE_SIGNAL, + SERVICE_FAILURE_CORE_DUMP, + SERVICE_FAILURE_WATCHDOG, + SERVICE_FAILURE_START_LIMIT_HIT, + _SERVICE_RESULT_MAX, + _SERVICE_RESULT_INVALID = -1 +} ServiceResult; + +struct ServiceFDStore { + Service *service; + + int fd; + char *fdname; + sd_event_source *event_source; + + LIST_FIELDS(ServiceFDStore, fd_store); +}; + +struct Service { + Unit meta; + + ServiceType type; + ServiceRestart restart; + ExitStatusSet restart_prevent_status; + ExitStatusSet restart_force_status; + ExitStatusSet success_status; + + /* If set we'll read the main daemon PID from this file */ + char *pid_file; + + usec_t restart_usec; + usec_t timeout_start_usec; + usec_t timeout_stop_usec; + usec_t runtime_max_usec; + + dual_timestamp watchdog_timestamp; + usec_t watchdog_usec; + usec_t watchdog_override_usec; + bool watchdog_override_enable; + sd_event_source *watchdog_event_source; + + ExecCommand* exec_command[_SERVICE_EXEC_COMMAND_MAX]; + + ExecContext exec_context; + KillContext kill_context; + CGroupContext cgroup_context; + + ServiceState state, deserialized_state; + + /* The exit status of the real main process */ + ExecStatus main_exec_status; + + /* The currently executed control process */ + ExecCommand *control_command; + + /* The currently executed main process, which may be NULL if + * the main process got started via forking mode and not by + * us */ + ExecCommand *main_command; + + /* The ID of the control command currently being executed */ + ServiceExecCommand control_command_id; + + /* Runtime data of the execution context */ + ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; + + pid_t main_pid, control_pid; + int socket_fd; + SocketPeer *peer; + bool socket_fd_selinux_context_net; + + bool permissions_start_only; + bool root_directory_start_only; + bool remain_after_exit; + bool guess_main_pid; + + /* If we shut down, remember why */ + ServiceResult result; + ServiceResult reload_result; + + bool main_pid_known:1; + bool main_pid_alien:1; + bool bus_name_good:1; + bool forbid_restart:1; + bool start_timeout_defined:1; + + bool reset_cpu_usage:1; + + char *bus_name; + char *bus_name_owner; /* unique name of the current owner */ + + char *status_text; + int status_errno; + + EmergencyAction emergency_action; + + UnitRef accept_socket; + + sd_event_source *timer_event_source; + PathSpec *pid_file_pathspec; + + NotifyAccess notify_access; + NotifyState notify_state; + + ServiceFDStore *fd_store; + unsigned n_fd_store; + unsigned n_fd_store_max; + + char *usb_function_descriptors; + char *usb_function_strings; + + int stdin_fd; + int stdout_fd; + int stderr_fd; +}; + +extern const UnitVTable service_vtable; + +int service_set_socket_fd(Service *s, int fd, struct Socket *socket, bool selinux_context_net); +void service_close_socket_fd(Service *s); + +const char* service_restart_to_string(ServiceRestart i) _const_; +ServiceRestart service_restart_from_string(const char *s) _pure_; + +const char* service_type_to_string(ServiceType i) _const_; +ServiceType service_type_from_string(const char *s) _pure_; + +const char* service_exec_command_to_string(ServiceExecCommand i) _const_; +ServiceExecCommand service_exec_command_from_string(const char *s) _pure_; + +const char* notify_access_to_string(NotifyAccess i) _const_; +NotifyAccess notify_access_from_string(const char *s) _pure_; + +const char* notify_state_to_string(NotifyState i) _const_; +NotifyState notify_state_from_string(const char *s) _pure_; + +const char* service_result_to_string(ServiceResult i) _const_; +ServiceResult service_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/show-status.h b/src/grp-system/libcore/include/core/show-status.h new file mode 100644 index 0000000000..08d6b7f6e1 --- /dev/null +++ b/src/grp-system/libcore/include/core/show-status.h @@ -0,0 +1,39 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2014 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> + +#include "systemd-basic/macro.h" + +/* Manager status */ + +typedef enum ShowStatus { + _SHOW_STATUS_UNSET = -2, + SHOW_STATUS_AUTO = -1, + SHOW_STATUS_NO = 0, + SHOW_STATUS_YES = 1, + SHOW_STATUS_TEMPORARY = 2, +} ShowStatus; + +int parse_show_status(const char *v, ShowStatus *ret); + +int status_vprintf(const char *status, bool ellipse, bool ephemeral, const char *format, va_list ap) _printf_(4,0); +int status_printf(const char *status, bool ellipse, bool ephemeral, const char *format, ...) _printf_(4,5); diff --git a/src/grp-system/libcore/include/core/slice.h b/src/grp-system/libcore/include/core/slice.h new file mode 100644 index 0000000000..c9f3f61067 --- /dev/null +++ b/src/grp-system/libcore/include/core/slice.h @@ -0,0 +1,32 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Slice Slice; + +struct Slice { + Unit meta; + + SliceState state, deserialized_state; + + CGroupContext cgroup_context; +}; + +extern const UnitVTable slice_vtable; diff --git a/src/grp-system/libcore/include/core/smack-setup.h b/src/grp-system/libcore/include/core/smack-setup.h new file mode 100644 index 0000000000..78164c85e6 --- /dev/null +++ b/src/grp-system/libcore/include/core/smack-setup.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright (C) 2013 Intel Corporation + Authors: + Nathaniel Chen <nathaniel.chen@intel.com> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published + by the Free Software Foundation; either version 2.1 of the License, + or (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int mac_smack_setup(bool *loaded_policy); diff --git a/src/grp-system/libcore/include/core/socket.h b/src/grp-system/libcore/include/core/socket.h new file mode 100644 index 0000000000..feba21dab3 --- /dev/null +++ b/src/grp-system/libcore/include/core/socket.h @@ -0,0 +1,198 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "systemd-basic/socket-util.h" + +typedef struct Socket Socket; +typedef struct SocketPeer SocketPeer; + +#include "mount.h" +#include "service.h" + +typedef enum SocketExecCommand { + SOCKET_EXEC_START_PRE, + SOCKET_EXEC_START_CHOWN, + SOCKET_EXEC_START_POST, + SOCKET_EXEC_STOP_PRE, + SOCKET_EXEC_STOP_POST, + _SOCKET_EXEC_COMMAND_MAX, + _SOCKET_EXEC_COMMAND_INVALID = -1 +} SocketExecCommand; + +typedef enum SocketType { + SOCKET_SOCKET, + SOCKET_FIFO, + SOCKET_SPECIAL, + SOCKET_MQUEUE, + SOCKET_USB_FUNCTION, + _SOCKET_FIFO_MAX, + _SOCKET_FIFO_INVALID = -1 +} SocketType; + +typedef enum SocketResult { + SOCKET_SUCCESS, + SOCKET_FAILURE_RESOURCES, + SOCKET_FAILURE_TIMEOUT, + SOCKET_FAILURE_EXIT_CODE, + SOCKET_FAILURE_SIGNAL, + SOCKET_FAILURE_CORE_DUMP, + SOCKET_FAILURE_START_LIMIT_HIT, + SOCKET_FAILURE_TRIGGER_LIMIT_HIT, + SOCKET_FAILURE_SERVICE_START_LIMIT_HIT, + _SOCKET_RESULT_MAX, + _SOCKET_RESULT_INVALID = -1 +} SocketResult; + +typedef struct SocketPort { + Socket *socket; + + SocketType type; + int fd; + int *auxiliary_fds; + int n_auxiliary_fds; + + SocketAddress address; + char *path; + sd_event_source *event_source; + + LIST_FIELDS(struct SocketPort, port); +} SocketPort; + +struct Socket { + Unit meta; + + LIST_HEAD(SocketPort, ports); + + Set *peers_by_address; + + unsigned n_accepted; + unsigned n_connections; + unsigned max_connections; + unsigned max_connections_per_source; + + unsigned backlog; + unsigned keep_alive_cnt; + usec_t timeout_usec; + usec_t keep_alive_time; + usec_t keep_alive_interval; + usec_t defer_accept; + + ExecCommand* exec_command[_SOCKET_EXEC_COMMAND_MAX]; + ExecContext exec_context; + KillContext kill_context; + CGroupContext cgroup_context; + + ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; + + /* For Accept=no sockets refers to the one service we'll + activate. For Accept=yes sockets is either NULL, or filled + when the next service we spawn. */ + UnitRef service; + + SocketState state, deserialized_state; + + sd_event_source *timer_event_source; + + ExecCommand* control_command; + SocketExecCommand control_command_id; + pid_t control_pid; + + mode_t directory_mode; + mode_t socket_mode; + + SocketResult result; + + char **symlinks; + + bool accept; + bool remove_on_stop; + bool writable; + + int socket_protocol; + + /* Socket options */ + bool keep_alive; + bool no_delay; + bool free_bind; + bool transparent; + bool broadcast; + bool pass_cred; + bool pass_sec; + + /* Only for INET6 sockets: issue IPV6_V6ONLY sockopt */ + SocketAddressBindIPv6Only bind_ipv6_only; + + int priority; + int mark; + size_t receive_buffer; + size_t send_buffer; + int ip_tos; + int ip_ttl; + size_t pipe_size; + char *bind_to_device; + char *tcp_congestion; + bool reuse_port; + long mq_maxmsg; + long mq_msgsize; + + char *smack; + char *smack_ip_in; + char *smack_ip_out; + + bool selinux_context_from_net; + + char *user, *group; + + bool reset_cpu_usage:1; + + char *fdname; + + RateLimit trigger_limit; +}; + +SocketPeer *socket_peer_ref(SocketPeer *p); +SocketPeer *socket_peer_unref(SocketPeer *p); +int socket_acquire_peer(Socket *s, int fd, SocketPeer **p); + +DEFINE_TRIVIAL_CLEANUP_FUNC(SocketPeer*, socket_peer_unref); + +/* Called from the service code when collecting fds */ +int socket_collect_fds(Socket *s, int **fds); + +/* Called from the service code when a per-connection service ended */ +void socket_connection_unref(Socket *s); + +void socket_free_ports(Socket *s); + +int socket_instantiate_service(Socket *s); + +char *socket_fdname(Socket *s); + +extern const UnitVTable socket_vtable; + +const char* socket_exec_command_to_string(SocketExecCommand i) _const_; +SocketExecCommand socket_exec_command_from_string(const char *s) _pure_; + +const char* socket_result_to_string(SocketResult i) _const_; +SocketResult socket_result_from_string(const char *s) _pure_; + +const char* socket_port_type_to_string(SocketPort *p) _pure_; diff --git a/src/grp-system/libcore/include/core/swap.h b/src/grp-system/libcore/include/core/swap.h new file mode 100644 index 0000000000..64db3267b2 --- /dev/null +++ b/src/grp-system/libcore/include/core/swap.h @@ -0,0 +1,111 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <libudev.h> + +typedef struct Swap Swap; + +typedef enum SwapExecCommand { + SWAP_EXEC_ACTIVATE, + SWAP_EXEC_DEACTIVATE, + _SWAP_EXEC_COMMAND_MAX, + _SWAP_EXEC_COMMAND_INVALID = -1 +} SwapExecCommand; + +typedef enum SwapResult { + SWAP_SUCCESS, + SWAP_FAILURE_RESOURCES, + SWAP_FAILURE_TIMEOUT, + SWAP_FAILURE_EXIT_CODE, + SWAP_FAILURE_SIGNAL, + SWAP_FAILURE_CORE_DUMP, + SWAP_FAILURE_START_LIMIT_HIT, + _SWAP_RESULT_MAX, + _SWAP_RESULT_INVALID = -1 +} SwapResult; + +typedef struct SwapParameters { + char *what; + char *options; + int priority; +} SwapParameters; + +struct Swap { + Unit meta; + + char *what; + + /* If the device has already shown up, this is the device + * node, which might be different from what, due to + * symlinks */ + char *devnode; + + SwapParameters parameters_proc_swaps; + SwapParameters parameters_fragment; + + bool from_proc_swaps:1; + bool from_fragment:1; + + /* Used while looking for swaps that vanished or got added + * from/to /proc/swaps */ + bool is_active:1; + bool just_activated:1; + + bool reset_cpu_usage:1; + + SwapResult result; + + usec_t timeout_usec; + + ExecCommand exec_command[_SWAP_EXEC_COMMAND_MAX]; + ExecContext exec_context; + KillContext kill_context; + CGroupContext cgroup_context; + + ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; + + SwapState state, deserialized_state; + + ExecCommand* control_command; + SwapExecCommand control_command_id; + pid_t control_pid; + + sd_event_source *timer_event_source; + + /* In order to be able to distinguish dependencies on + different device nodes we might end up creating multiple + devices for the same swap. We chain them up here. */ + + LIST_FIELDS(struct Swap, same_devnode); +}; + +extern const UnitVTable swap_vtable; + +int swap_process_device_new(Manager *m, struct udev_device *dev); +int swap_process_device_remove(Manager *m, struct udev_device *dev); + +const char* swap_exec_command_to_string(SwapExecCommand i) _const_; +SwapExecCommand swap_exec_command_from_string(const char *s) _pure_; + +const char* swap_result_to_string(SwapResult i) _const_; +SwapResult swap_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/target.h b/src/grp-system/libcore/include/core/target.h new file mode 100644 index 0000000000..339aea154e --- /dev/null +++ b/src/grp-system/libcore/include/core/target.h @@ -0,0 +1,30 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +typedef struct Target Target; + +struct Target { + Unit meta; + + TargetState state, deserialized_state; +}; + +extern const UnitVTable target_vtable; diff --git a/src/grp-system/libcore/include/core/timer.h b/src/grp-system/libcore/include/core/timer.h new file mode 100644 index 0000000000..9cb30249e3 --- /dev/null +++ b/src/grp-system/libcore/include/core/timer.h @@ -0,0 +1,89 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "systemd-basic/calendarspec.h" + +typedef struct Timer Timer; + +typedef enum TimerBase { + TIMER_ACTIVE, + TIMER_BOOT, + TIMER_STARTUP, + TIMER_UNIT_ACTIVE, + TIMER_UNIT_INACTIVE, + TIMER_CALENDAR, + _TIMER_BASE_MAX, + _TIMER_BASE_INVALID = -1 +} TimerBase; + +typedef struct TimerValue { + TimerBase base; + bool disabled; + + usec_t value; /* only for monotonic events */ + CalendarSpec *calendar_spec; /* only for calendar events */ + usec_t next_elapse; + + LIST_FIELDS(struct TimerValue, value); +} TimerValue; + +typedef enum TimerResult { + TIMER_SUCCESS, + TIMER_FAILURE_RESOURCES, + TIMER_FAILURE_START_LIMIT_HIT, + _TIMER_RESULT_MAX, + _TIMER_RESULT_INVALID = -1 +} TimerResult; + +struct Timer { + Unit meta; + + usec_t accuracy_usec; + usec_t random_usec; + + LIST_HEAD(TimerValue, values); + usec_t next_elapse_realtime; + usec_t next_elapse_monotonic_or_boottime; + dual_timestamp last_trigger; + + TimerState state, deserialized_state; + + sd_event_source *monotonic_event_source; + sd_event_source *realtime_event_source; + + TimerResult result; + + bool persistent; + bool wake_system; + bool remain_after_elapse; + + char *stamp_path; +}; + +void timer_free_values(Timer *t); + +extern const UnitVTable timer_vtable; + +const char *timer_base_to_string(TimerBase i) _const_; +TimerBase timer_base_from_string(const char *s) _pure_; + +const char* timer_result_to_string(TimerResult i) _const_; +TimerResult timer_result_from_string(const char *s) _pure_; diff --git a/src/grp-system/libcore/include/core/unit.h b/src/grp-system/libcore/include/core/unit.h new file mode 100644 index 0000000000..6864b71183 --- /dev/null +++ b/src/grp-system/libcore/include/core/unit.h @@ -0,0 +1,680 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdbool.h> +#include <stdlib.h> +#include <unistd.h> + +#include "systemd-basic/list.h" +#include "systemd-basic/unit-name.h" +#include "systemd-shared/condition.h" +#include "systemd-shared/install.h" + +typedef struct Unit Unit; +typedef struct UnitRef UnitRef; +typedef struct UnitStatusMessageFormats UnitStatusMessageFormats; +typedef struct UnitVTable UnitVTable; + +#include "emergency-action.h" + +typedef enum KillOperation { + KILL_TERMINATE, + KILL_TERMINATE_AND_LOG, + KILL_KILL, + KILL_ABORT, + _KILL_OPERATION_MAX, + _KILL_OPERATION_INVALID = -1 +} KillOperation; + +static inline bool UNIT_IS_ACTIVE_OR_RELOADING(UnitActiveState t) { + return t == UNIT_ACTIVE || t == UNIT_RELOADING; +} + +static inline bool UNIT_IS_ACTIVE_OR_ACTIVATING(UnitActiveState t) { + return t == UNIT_ACTIVE || t == UNIT_ACTIVATING || t == UNIT_RELOADING; +} + +static inline bool UNIT_IS_INACTIVE_OR_DEACTIVATING(UnitActiveState t) { + return t == UNIT_INACTIVE || t == UNIT_FAILED || t == UNIT_DEACTIVATING; +} + +static inline bool UNIT_IS_INACTIVE_OR_FAILED(UnitActiveState t) { + return t == UNIT_INACTIVE || t == UNIT_FAILED; +} + +#include "job.h" + +struct UnitRef { + /* Keeps tracks of references to a unit. This is useful so + * that we can merge two units if necessary and correct all + * references to them */ + + Unit* unit; + LIST_FIELDS(UnitRef, refs); +}; + +struct Unit { + Manager *manager; + + UnitType type; + UnitLoadState load_state; + Unit *merged_into; + + char *id; /* One name is special because we use it for identification. Points to an entry in the names set */ + char *instance; + + Set *names; + Set *dependencies[_UNIT_DEPENDENCY_MAX]; + + char **requires_mounts_for; + + char *description; + char **documentation; + + char *fragment_path; /* if loaded from a config file this is the primary path to it */ + char *source_path; /* if converted, the source file */ + char **dropin_paths; + + usec_t fragment_mtime; + usec_t source_mtime; + usec_t dropin_mtime; + + /* If this is a transient unit we are currently writing, this is where we are writing it to */ + FILE *transient_file; + + /* If there is something to do with this unit, then this is the installed job for it */ + Job *job; + + /* JOB_NOP jobs are special and can be installed without disturbing the real job. */ + Job *nop_job; + + /* The slot used for watching NameOwnerChanged signals */ + sd_bus_slot *match_bus_slot; + + /* References to this unit from clients */ + sd_bus_track *bus_track; + char **deserialized_refs; + + /* Job timeout and action to take */ + usec_t job_timeout; + EmergencyAction job_timeout_action; + char *job_timeout_reboot_arg; + + /* References to this */ + LIST_HEAD(UnitRef, refs); + + /* Conditions to check */ + LIST_HEAD(Condition, conditions); + LIST_HEAD(Condition, asserts); + + dual_timestamp condition_timestamp; + dual_timestamp assert_timestamp; + + /* Updated whenever the low-level state changes */ + dual_timestamp state_change_timestamp; + + /* Updated whenever the (high-level) active state enters or leaves the active or inactive states */ + dual_timestamp inactive_exit_timestamp; + dual_timestamp active_enter_timestamp; + dual_timestamp active_exit_timestamp; + dual_timestamp inactive_enter_timestamp; + + UnitRef slice; + + /* Per type list */ + LIST_FIELDS(Unit, units_by_type); + + /* All units which have requires_mounts_for set */ + LIST_FIELDS(Unit, has_requires_mounts_for); + + /* Load queue */ + LIST_FIELDS(Unit, load_queue); + + /* D-Bus queue */ + LIST_FIELDS(Unit, dbus_queue); + + /* Cleanup queue */ + LIST_FIELDS(Unit, cleanup_queue); + + /* GC queue */ + LIST_FIELDS(Unit, gc_queue); + + /* CGroup realize members queue */ + LIST_FIELDS(Unit, cgroup_queue); + + /* Units with the same CGroup netclass */ + LIST_FIELDS(Unit, cgroup_netclass); + + /* PIDs we keep an eye on. Note that a unit might have many + * more, but these are the ones we care enough about to + * process SIGCHLD for */ + Set *pids; + + /* Used in sigchld event invocation to avoid repeat events being invoked */ + uint64_t sigchldgen; + + /* Used during GC sweeps */ + unsigned gc_marker; + + /* Error code when we didn't manage to load the unit (negative) */ + int load_error; + + /* Put a ratelimit on unit starting */ + RateLimit start_limit; + EmergencyAction start_limit_action; + char *reboot_arg; + + /* Make sure we never enter endless loops with the check unneeded logic, or the BindsTo= logic */ + RateLimit auto_stop_ratelimit; + + /* Reference to a specific UID/GID */ + uid_t ref_uid; + gid_t ref_gid; + + /* Cached unit file state and preset */ + UnitFileState unit_file_state; + int unit_file_preset; + + /* Where the cpu.stat or cpuacct.usage was at the time the unit was started */ + nsec_t cpu_usage_base; + nsec_t cpu_usage_last; /* the most recently read value */ + + /* Counterparts in the cgroup filesystem */ + char *cgroup_path; + CGroupMask cgroup_realized_mask; + CGroupMask cgroup_enabled_mask; + CGroupMask cgroup_subtree_mask; + CGroupMask cgroup_members_mask; + int cgroup_inotify_wd; + + /* How to start OnFailure units */ + JobMode on_failure_job_mode; + + /* The current invocation ID */ + sd_id128_t invocation_id; + char invocation_id_string[SD_ID128_STRING_MAX]; /* useful when logging */ + + /* Garbage collect us we nobody wants or requires us anymore */ + bool stop_when_unneeded; + + /* Create default dependencies */ + bool default_dependencies; + + /* Refuse manual starting, allow starting only indirectly via dependency. */ + bool refuse_manual_start; + + /* Don't allow the user to stop this unit manually, allow stopping only indirectly via dependency. */ + bool refuse_manual_stop; + + /* Allow isolation requests */ + bool allow_isolate; + + /* Ignore this unit when isolating */ + bool ignore_on_isolate; + + /* Did the last condition check succeed? */ + bool condition_result; + bool assert_result; + + /* Is this a transient unit? */ + bool transient; + + /* Is this a unit that is always running and cannot be stopped? */ + bool perpetual; + + bool in_load_queue:1; + bool in_dbus_queue:1; + bool in_cleanup_queue:1; + bool in_gc_queue:1; + bool in_cgroup_queue:1; + + bool sent_dbus_new_signal:1; + + bool in_audit:1; + + bool cgroup_realized:1; + bool cgroup_members_mask_valid:1; + bool cgroup_subtree_mask_valid:1; + + bool start_limit_hit:1; + + /* Did we already invoke unit_coldplug() for this unit? */ + bool coldplugged:1; + + /* For transient units: whether to add a bus track reference after creating the unit */ + bool bus_track_add:1; +}; + +struct UnitStatusMessageFormats { + const char *starting_stopping[2]; + const char *finished_start_job[_JOB_RESULT_MAX]; + const char *finished_stop_job[_JOB_RESULT_MAX]; +}; + +typedef enum UnitSetPropertiesMode { + UNIT_CHECK = 0, + UNIT_RUNTIME = 1, + UNIT_PERSISTENT = 2, +} UnitSetPropertiesMode; + +#include "automount.h" +#include "busname.h" +#include "device.h" +#include "path.h" +#include "scope.h" +#include "slice.h" +#include "socket.h" +#include "swap.h" +#include "target.h" +#include "timer.h" + +struct UnitVTable { + /* How much memory does an object of this unit type need */ + size_t object_size; + + /* If greater than 0, the offset into the object where + * ExecContext is found, if the unit type has that */ + size_t exec_context_offset; + + /* If greater than 0, the offset into the object where + * CGroupContext is found, if the unit type has that */ + size_t cgroup_context_offset; + + /* If greater than 0, the offset into the object where + * KillContext is found, if the unit type has that */ + size_t kill_context_offset; + + /* If greater than 0, the offset into the object where the + * pointer to ExecRuntime is found, if the unit type has + * that */ + size_t exec_runtime_offset; + + /* If greater than 0, the offset into the object where the pointer to DynamicCreds is found, if the unit type + * has that. */ + size_t dynamic_creds_offset; + + /* The name of the configuration file section with the private settings of this unit */ + const char *private_section; + + /* Config file sections this unit type understands, separated + * by NUL chars */ + const char *sections; + + /* This should reset all type-specific variables. This should + * not allocate memory, and is called with zero-initialized + * data. It should hence only initialize variables that need + * to be set != 0. */ + void (*init)(Unit *u); + + /* This should free all type-specific variables. It should be + * idempotent. */ + void (*done)(Unit *u); + + /* Actually load data from disk. This may fail, and should set + * load_state to UNIT_LOADED, UNIT_MERGED or leave it at + * UNIT_STUB if no configuration could be found. */ + int (*load)(Unit *u); + + /* If a lot of units got created via enumerate(), this is + * where to actually set the state and call unit_notify(). */ + int (*coldplug)(Unit *u); + + void (*dump)(Unit *u, FILE *f, const char *prefix); + + int (*start)(Unit *u); + int (*stop)(Unit *u); + int (*reload)(Unit *u); + + int (*kill)(Unit *u, KillWho w, int signo, sd_bus_error *error); + + bool (*can_reload)(Unit *u); + + /* Write all data that cannot be restored from other sources + * away using unit_serialize_item() */ + int (*serialize)(Unit *u, FILE *f, FDSet *fds); + + /* Restore one item from the serialization */ + int (*deserialize_item)(Unit *u, const char *key, const char *data, FDSet *fds); + + /* Try to match up fds with what we need for this unit */ + void (*distribute_fds)(Unit *u, FDSet *fds); + + /* Boils down the more complex internal state of this unit to + * a simpler one that the engine can understand */ + UnitActiveState (*active_state)(Unit *u); + + /* Returns the substate specific to this unit type as + * string. This is purely information so that we can give the + * user a more fine grained explanation in which actual state a + * unit is in. */ + const char* (*sub_state_to_string)(Unit *u); + + /* Return true when there is reason to keep this entry around + * even nothing references it and it isn't active in any + * way */ + bool (*check_gc)(Unit *u); + + /* When the unit is not running and no job for it queued we + * shall release its runtime resources */ + void (*release_resources)(Unit *u, bool inactive); + + /* Invoked on every child that died */ + void (*sigchld_event)(Unit *u, pid_t pid, int code, int status); + + /* Reset failed state if we are in failed state */ + void (*reset_failed)(Unit *u); + + /* Called whenever any of the cgroups this unit watches for + * ran empty */ + void (*notify_cgroup_empty)(Unit *u); + + /* Called whenever a process of this unit sends us a message */ + void (*notify_message)(Unit *u, pid_t pid, char **tags, FDSet *fds); + + /* Called whenever a name this Unit registered for comes or goes away. */ + void (*bus_name_owner_change)(Unit *u, const char *name, const char *old_owner, const char *new_owner); + + /* Called for each property that is being set */ + int (*bus_set_property)(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); + + /* Called after at least one property got changed to apply the necessary change */ + int (*bus_commit_properties)(Unit *u); + + /* Return the unit this unit is following */ + Unit *(*following)(Unit *u); + + /* Return the set of units that are following each other */ + int (*following_set)(Unit *u, Set **s); + + /* Invoked each time a unit this unit is triggering changes + * state or gains/loses a job */ + void (*trigger_notify)(Unit *u, Unit *trigger); + + /* Called whenever CLOCK_REALTIME made a jump */ + void (*time_change)(Unit *u); + + /* Returns the next timeout of a unit */ + int (*get_timeout)(Unit *u, usec_t *timeout); + + /* Returns the main PID if there is any defined, or 0. */ + pid_t (*main_pid)(Unit *u); + + /* Returns the main PID if there is any defined, or 0. */ + pid_t (*control_pid)(Unit *u); + + /* This is called for each unit type and should be used to + * enumerate existing devices and load them. However, + * everything that is loaded here should still stay in + * inactive state. It is the job of the coldplug() call above + * to put the units into the initial state. */ + void (*enumerate)(Manager *m); + + /* Type specific cleanups. */ + void (*shutdown)(Manager *m); + + /* If this function is set and return false all jobs for units + * of this type will immediately fail. */ + bool (*supported)(void); + + /* The bus vtable */ + const sd_bus_vtable *bus_vtable; + + /* The strings to print in status messages */ + UnitStatusMessageFormats status_message_formats; + + /* True if transient units of this type are OK */ + bool can_transient:1; +}; + +extern const UnitVTable * const unit_vtable[_UNIT_TYPE_MAX]; + +#define UNIT_VTABLE(u) unit_vtable[(u)->type] + +/* For casting a unit into the various unit types */ +#define DEFINE_CAST(UPPERCASE, MixedCase) \ + static inline MixedCase* UPPERCASE(Unit *u) { \ + if (_unlikely_(!u || u->type != UNIT_##UPPERCASE)) \ + return NULL; \ + \ + return (MixedCase*) u; \ + } + +/* For casting the various unit types into a unit */ +#define UNIT(u) (&(u)->meta) + +#define UNIT_HAS_EXEC_CONTEXT(u) (UNIT_VTABLE(u)->exec_context_offset > 0) +#define UNIT_HAS_CGROUP_CONTEXT(u) (UNIT_VTABLE(u)->cgroup_context_offset > 0) +#define UNIT_HAS_KILL_CONTEXT(u) (UNIT_VTABLE(u)->kill_context_offset > 0) + +#define UNIT_TRIGGER(u) ((Unit*) set_first((u)->dependencies[UNIT_TRIGGERS])) + +DEFINE_CAST(SERVICE, Service); +DEFINE_CAST(SOCKET, Socket); +DEFINE_CAST(BUSNAME, BusName); +DEFINE_CAST(TARGET, Target); +DEFINE_CAST(DEVICE, Device); +DEFINE_CAST(MOUNT, Mount); +DEFINE_CAST(AUTOMOUNT, Automount); +DEFINE_CAST(SWAP, Swap); +DEFINE_CAST(TIMER, Timer); +DEFINE_CAST(PATH, Path); +DEFINE_CAST(SLICE, Slice); +DEFINE_CAST(SCOPE, Scope); + +Unit *unit_new(Manager *m, size_t size); +void unit_free(Unit *u); + +int unit_new_for_name(Manager *m, size_t size, const char *name, Unit **ret); +int unit_add_name(Unit *u, const char *name); + +int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_reference); +int unit_add_two_dependencies(Unit *u, UnitDependency d, UnitDependency e, Unit *other, bool add_reference); + +int unit_add_dependency_by_name(Unit *u, UnitDependency d, const char *name, const char *filename, bool add_reference); +int unit_add_two_dependencies_by_name(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference); + +int unit_add_exec_dependencies(Unit *u, ExecContext *c); + +int unit_choose_id(Unit *u, const char *name); +int unit_set_description(Unit *u, const char *description); + +bool unit_check_gc(Unit *u); + +void unit_add_to_load_queue(Unit *u); +void unit_add_to_dbus_queue(Unit *u); +void unit_add_to_cleanup_queue(Unit *u); +void unit_add_to_gc_queue(Unit *u); + +int unit_merge(Unit *u, Unit *other); +int unit_merge_by_name(Unit *u, const char *other); + +Unit *unit_follow_merge(Unit *u) _pure_; + +int unit_load_fragment_and_dropin(Unit *u); +int unit_load_fragment_and_dropin_optional(Unit *u); +int unit_load(Unit *unit); + +int unit_set_slice(Unit *u, Unit *slice); +int unit_set_default_slice(Unit *u); + +const char *unit_description(Unit *u) _pure_; + +bool unit_has_name(Unit *u, const char *name); + +UnitActiveState unit_active_state(Unit *u); + +const char* unit_sub_state_to_string(Unit *u); + +void unit_dump(Unit *u, FILE *f, const char *prefix); + +bool unit_can_reload(Unit *u) _pure_; +bool unit_can_start(Unit *u) _pure_; +bool unit_can_stop(Unit *u) _pure_; +bool unit_can_isolate(Unit *u) _pure_; + +int unit_start(Unit *u); +int unit_stop(Unit *u); +int unit_reload(Unit *u); + +int unit_kill(Unit *u, KillWho w, int signo, sd_bus_error *error); +int unit_kill_common(Unit *u, KillWho who, int signo, pid_t main_pid, pid_t control_pid, sd_bus_error *error); + +void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_success); + +int unit_watch_pid(Unit *u, pid_t pid); +void unit_unwatch_pid(Unit *u, pid_t pid); +void unit_unwatch_all_pids(Unit *u); + +void unit_tidy_watch_pids(Unit *u, pid_t except1, pid_t except2); + +int unit_install_bus_match(Unit *u, sd_bus *bus, const char *name); +int unit_watch_bus_name(Unit *u, const char *name); +void unit_unwatch_bus_name(Unit *u, const char *name); + +bool unit_job_is_applicable(Unit *u, JobType j); + +int set_unit_path(const char *p); + +char *unit_dbus_path(Unit *u); +char *unit_dbus_path_invocation_id(Unit *u); + +int unit_load_related_unit(Unit *u, const char *type, Unit **_found); + +bool unit_can_serialize(Unit *u) _pure_; + +int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs); +int unit_deserialize(Unit *u, FILE *f, FDSet *fds); + +int unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value); +int unit_serialize_item_escaped(Unit *u, FILE *f, const char *key, const char *value); +int unit_serialize_item_fd(Unit *u, FILE *f, FDSet *fds, const char *key, int fd); +void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *value, ...) _printf_(4,5); + +int unit_add_node_link(Unit *u, const char *what, bool wants, UnitDependency d); + +int unit_coldplug(Unit *u); + +void unit_status_printf(Unit *u, const char *status, const char *unit_status_msg_format) _printf_(3, 0); +void unit_status_emit_starting_stopping_reloading(Unit *u, JobType t); + +bool unit_need_daemon_reload(Unit *u); + +void unit_reset_failed(Unit *u); + +Unit *unit_following(Unit *u); +int unit_following_set(Unit *u, Set **s); + +const char *unit_slice_name(Unit *u); + +bool unit_stop_pending(Unit *u) _pure_; +bool unit_inactive_or_pending(Unit *u) _pure_; +bool unit_active_or_pending(Unit *u); + +int unit_add_default_target_dependency(Unit *u, Unit *target); + +void unit_start_on_failure(Unit *u); +void unit_trigger_notify(Unit *u); + +UnitFileState unit_get_unit_file_state(Unit *u); +int unit_get_unit_file_preset(Unit *u); + +Unit* unit_ref_set(UnitRef *ref, Unit *u); +void unit_ref_unset(UnitRef *ref); + +#define UNIT_DEREF(ref) ((ref).unit) +#define UNIT_ISSET(ref) (!!(ref).unit) + +int unit_patch_contexts(Unit *u); + +ExecContext *unit_get_exec_context(Unit *u) _pure_; +KillContext *unit_get_kill_context(Unit *u) _pure_; +CGroupContext *unit_get_cgroup_context(Unit *u) _pure_; + +ExecRuntime *unit_get_exec_runtime(Unit *u) _pure_; + +int unit_setup_exec_runtime(Unit *u); +int unit_setup_dynamic_creds(Unit *u); + +int unit_write_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data); +int unit_write_drop_in_format(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *format, ...) _printf_(4,5); + +int unit_write_drop_in_private(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data); +int unit_write_drop_in_private_format(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *format, ...) _printf_(4,5); + +int unit_kill_context(Unit *u, KillContext *c, KillOperation k, pid_t main_pid, pid_t control_pid, bool main_pid_alien); + +int unit_make_transient(Unit *u); + +int unit_require_mounts_for(Unit *u, const char *path); + +bool unit_type_supported(UnitType t); + +bool unit_is_pristine(Unit *u); + +pid_t unit_control_pid(Unit *u); +pid_t unit_main_pid(Unit *u); + +static inline bool unit_supported(Unit *u) { + return unit_type_supported(u->type); +} + +void unit_warn_if_dir_nonempty(Unit *u, const char* where); +int unit_fail_if_symlink(Unit *u, const char* where); + +int unit_start_limit_test(Unit *u); + +void unit_unref_uid(Unit *u, bool destroy_now); +int unit_ref_uid(Unit *u, uid_t uid, bool clean_ipc); + +void unit_unref_gid(Unit *u, bool destroy_now); +int unit_ref_gid(Unit *u, gid_t gid, bool clean_ipc); + +int unit_ref_uid_gid(Unit *u, uid_t uid, gid_t gid); +void unit_unref_uid_gid(Unit *u, bool destroy_now); + +void unit_notify_user_lookup(Unit *u, uid_t uid, gid_t gid); + +int unit_set_invocation_id(Unit *u, sd_id128_t id); +int unit_acquire_invocation_id(Unit *u); + +/* Macros which append UNIT= or USER_UNIT= to the message */ + +#define log_unit_full(unit, level, error, ...) \ + ({ \ + const Unit *_u = (unit); \ + _u ? log_object_internal(level, error, __FILE__, __LINE__, __func__, _u->manager->unit_log_field, _u->id, _u->manager->invocation_log_field, _u->invocation_id_string, ##__VA_ARGS__) : \ + log_internal(level, error, __FILE__, __LINE__, __func__, ##__VA_ARGS__); \ + }) + +#define log_unit_debug(unit, ...) log_unit_full(unit, LOG_DEBUG, 0, ##__VA_ARGS__) +#define log_unit_info(unit, ...) log_unit_full(unit, LOG_INFO, 0, ##__VA_ARGS__) +#define log_unit_notice(unit, ...) log_unit_full(unit, LOG_NOTICE, 0, ##__VA_ARGS__) +#define log_unit_warning(unit, ...) log_unit_full(unit, LOG_WARNING, 0, ##__VA_ARGS__) +#define log_unit_error(unit, ...) log_unit_full(unit, LOG_ERR, 0, ##__VA_ARGS__) + +#define log_unit_debug_errno(unit, error, ...) log_unit_full(unit, LOG_DEBUG, error, ##__VA_ARGS__) +#define log_unit_info_errno(unit, error, ...) log_unit_full(unit, LOG_INFO, error, ##__VA_ARGS__) +#define log_unit_notice_errno(unit, error, ...) log_unit_full(unit, LOG_NOTICE, error, ##__VA_ARGS__) +#define log_unit_warning_errno(unit, error, ...) log_unit_full(unit, LOG_WARNING, error, ##__VA_ARGS__) +#define log_unit_error_errno(unit, error, ...) log_unit_full(unit, LOG_ERR, error, ##__VA_ARGS__) + +#define LOG_UNIT_MESSAGE(unit, fmt, ...) "MESSAGE=%s: " fmt, (unit)->id, ##__VA_ARGS__ +#define LOG_UNIT_ID(unit) (unit)->manager->unit_log_format_string, (unit)->id diff --git a/src/grp-system/libcore/src/GNUmakefile b/src/grp-system/libcore/src/GNUmakefile new file mode 120000 index 0000000000..13308a50cd --- /dev/null +++ b/src/grp-system/libcore/src/GNUmakefile @@ -0,0 +1 @@ +../../../../GNUmakefile
\ No newline at end of file diff --git a/src/grp-system/libcore/src/Makefile b/src/grp-system/libcore/src/Makefile new file mode 100644 index 0000000000..cd63651f2a --- /dev/null +++ b/src/grp-system/libcore/src/Makefile @@ -0,0 +1,170 @@ +# -*- Mode: makefile; indent-tabs-mode: t -*- +# +# This file is part of systemd. +# +# Copyright 2010-2012 Lennart Poettering +# Copyright 2010-2012 Kay Sievers +# Copyright 2013 Zbigniew Jędrzejewski-Szmek +# Copyright 2013 David Strauss +# Copyright 2016 Luke Shumaker +# +# systemd is free software; you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published by +# the Free Software Foundation; either version 2.1 of the License, or +# (at your option) any later version. +# +# systemd is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with systemd; If not, see <http://www.gnu.org/licenses/>. +include $(dir $(lastword $(MAKEFILE_LIST)))/../../../../config.mk +include $(topsrcdir)/build-aux/Makefile.head.mk + +noinst_LTLIBRARIES += \ + libcore.la + +libcore_la_SOURCES = \ + src/core/unit.c \ + src/core/unit.h \ + src/core/unit-printf.c \ + src/core/unit-printf.h \ + src/core/job.c \ + src/core/job.h \ + src/core/manager.c \ + src/core/manager.h \ + src/core/transaction.c \ + src/core/transaction.h \ + src/core/load-fragment.c \ + src/core/load-fragment.h \ + src/core/service.c \ + src/core/service.h \ + src/core/socket.c \ + src/core/socket.h \ + src/core/busname.c \ + src/core/busname.h \ + src/core/bus-policy.c \ + src/core/bus-policy.h \ + src/core/target.c \ + src/core/target.h \ + src/core/device.c \ + src/core/device.h \ + src/core/mount.c \ + src/core/mount.h \ + src/core/automount.c \ + src/core/automount.h \ + src/core/swap.c \ + src/core/swap.h \ + src/core/timer.c \ + src/core/timer.h \ + src/core/path.c \ + src/core/path.h \ + src/core/slice.c \ + src/core/slice.h \ + src/core/scope.c \ + src/core/scope.h \ + src/core/load-dropin.c \ + src/core/load-dropin.h \ + src/core/execute.c \ + src/core/execute.h \ + src/core/dynamic-user.c \ + src/core/dynamic-user.h \ + src/core/kill.c \ + src/core/kill.h \ + src/core/dbus.c \ + src/core/dbus.h \ + src/core/dbus-manager.c \ + src/core/dbus-manager.h \ + src/core/dbus-unit.c \ + src/core/dbus-unit.h \ + src/core/dbus-job.c \ + src/core/dbus-job.h \ + src/core/dbus-service.c \ + src/core/dbus-service.h \ + src/core/dbus-socket.c \ + src/core/dbus-socket.h \ + src/core/dbus-busname.c \ + src/core/dbus-busname.h \ + src/core/dbus-target.c \ + src/core/dbus-target.h \ + src/core/dbus-device.c \ + src/core/dbus-device.h \ + src/core/dbus-mount.c \ + src/core/dbus-mount.h \ + src/core/dbus-automount.c \ + src/core/dbus-automount.h \ + src/core/dbus-swap.c \ + src/core/dbus-swap.h \ + src/core/dbus-timer.c \ + src/core/dbus-timer.h \ + src/core/dbus-path.c \ + src/core/dbus-path.h \ + src/core/dbus-slice.c \ + src/core/dbus-slice.h \ + src/core/dbus-scope.c \ + src/core/dbus-scope.h \ + src/core/dbus-execute.c \ + src/core/dbus-execute.h \ + src/core/dbus-kill.c \ + src/core/dbus-kill.h \ + src/core/dbus-cgroup.c \ + src/core/dbus-cgroup.h \ + src/core/cgroup.c \ + src/core/cgroup.h \ + src/core/selinux-access.c \ + src/core/selinux-access.h \ + src/core/selinux-setup.c \ + src/core/selinux-setup.h \ + src/core/smack-setup.c \ + src/core/smack-setup.h \ + src/core/ima-setup.c \ + src/core/ima-setup.h \ + src/core/locale-setup.h \ + src/core/locale-setup.c \ + src/core/hostname-setup.c \ + src/core/hostname-setup.h \ + src/core/machine-id-setup.c \ + src/core/machine-id-setup.h \ + src/core/mount-setup.c \ + src/core/mount-setup.h \ + src/core/kmod-setup.c \ + src/core/kmod-setup.h \ + src/core/loopback-setup.h \ + src/core/loopback-setup.c \ + src/core/namespace.c \ + src/core/namespace.h \ + src/core/killall.h \ + src/core/killall.c \ + src/core/audit-fd.c \ + src/core/audit-fd.h \ + src/core/show-status.c \ + src/core/show-status.h \ + src/core/emergency-action.c \ + src/core/emergency-action.h + +nodist_libcore_la_SOURCES = \ + src/core/load-fragment-gperf.c \ + src/core/load-fragment-gperf-nulstr.c + +libcore_la_CFLAGS = \ + $(PAM_CFLAGS) \ + $(AUDIT_CFLAGS) \ + $(KMOD_CFLAGS) \ + $(APPARMOR_CFLAGS) \ + $(MOUNT_CFLAGS) \ + $(SECCOMP_CFLAGS) + +libcore_la_LIBADD = \ + libsystemd-shared.la \ + $(PAM_LIBS) \ + $(AUDIT_LIBS) \ + $(KMOD_LIBS) \ + $(APPARMOR_LIBS) \ + $(MOUNT_LIBS) + +$(outdir)/load-fragment-gperf-nulstr.c: src/core/load-fragment-gperf.gperf + $(AM_V_GEN)$(AWK) 'BEGIN{ keywords=0 ; FS="," ; print "extern const char load_fragment_gperf_nulstr[];" ; print "const char load_fragment_gperf_nulstr[] ="} ; keyword==1 { print "\"" $$1 "\\0\"" } ; /%%/ { keyword=1} ; END { print ";" }' < $< > $@ + +include $(topsrcdir)/build-aux/Makefile.tail.mk diff --git a/src/grp-system/libcore/src/audit-fd.c b/src/grp-system/libcore/src/audit-fd.c new file mode 100644 index 0000000000..12509951b7 --- /dev/null +++ b/src/grp-system/libcore/src/audit-fd.c @@ -0,0 +1,73 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + +#include <errno.h> + +#include "audit-fd.h" + +#ifdef HAVE_AUDIT + +#include <libaudit.h> +#include <stdbool.h> + +#include "systemd-basic/fd-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/util.h" + +static bool initialized = false; +static int audit_fd; + +int get_audit_fd(void) { + + if (!initialized) { + audit_fd = audit_open(); + + if (audit_fd < 0) { + if (errno != EAFNOSUPPORT && errno != EPROTONOSUPPORT) + log_error_errno(errno, "Failed to connect to audit log: %m"); + + audit_fd = errno ? -errno : -EINVAL; + } + + initialized = true; + } + + return audit_fd; +} + +void close_audit_fd(void) { + + if (initialized && audit_fd >= 0) + safe_close(audit_fd); + + initialized = true; + audit_fd = -ECONNRESET; +} + +#else + +int get_audit_fd(void) { + return -EAFNOSUPPORT; +} + +void close_audit_fd(void) { +} + +#endif diff --git a/src/grp-system/libcore/src/audit-fd.h b/src/grp-system/libcore/src/audit-fd.h new file mode 100644 index 0000000000..0eccb59210 --- /dev/null +++ b/src/grp-system/libcore/src/audit-fd.h @@ -0,0 +1,23 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int get_audit_fd(void); +void close_audit_fd(void); diff --git a/src/grp-system/libcore/src/automount.c b/src/grp-system/libcore/src/automount.c new file mode 100644 index 0000000000..e0ba6f68a9 --- /dev/null +++ b/src/grp-system/libcore/src/automount.c @@ -0,0 +1,1136 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <limits.h> +#include <sys/epoll.h> +#include <sys/mount.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <linux/auto_dev-ioctl.h> +#include <linux/auto_fs4.h> + +#include "core/automount.h" +#include "core/mount.h" +#include "core/unit.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/async.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/label.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/mount-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" + +#include "dbus-automount.h" + +static const UnitActiveState state_translation_table[_AUTOMOUNT_STATE_MAX] = { + [AUTOMOUNT_DEAD] = UNIT_INACTIVE, + [AUTOMOUNT_WAITING] = UNIT_ACTIVE, + [AUTOMOUNT_RUNNING] = UNIT_ACTIVE, + [AUTOMOUNT_FAILED] = UNIT_FAILED +}; + +struct expire_data { + int dev_autofs_fd; + int ioctl_fd; +}; + +static inline void expire_data_free(struct expire_data *data) { + if (!data) + return; + + safe_close(data->dev_autofs_fd); + safe_close(data->ioctl_fd); + free(data); +} + +DEFINE_TRIVIAL_CLEANUP_FUNC(struct expire_data*, expire_data_free); + +static int open_dev_autofs(Manager *m); +static int automount_dispatch_io(sd_event_source *s, int fd, uint32_t events, void *userdata); +static int automount_start_expire(Automount *a); +static void automount_stop_expire(Automount *a); +static int automount_send_ready(Automount *a, Set *tokens, int status); + +static void automount_init(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + a->pipe_fd = -1; + a->directory_mode = 0755; + UNIT(a)->ignore_on_isolate = true; +} + +static void unmount_autofs(Automount *a) { + int r; + + assert(a); + + if (a->pipe_fd < 0) + return; + + a->pipe_event_source = sd_event_source_unref(a->pipe_event_source); + a->pipe_fd = safe_close(a->pipe_fd); + + /* If we reload/reexecute things we keep the mount point + * around */ + if (a->where && + (UNIT(a)->manager->exit_code != MANAGER_RELOAD && + UNIT(a)->manager->exit_code != MANAGER_REEXECUTE)) { + automount_send_ready(a, a->tokens, -EHOSTDOWN); + automount_send_ready(a, a->expire_tokens, -EHOSTDOWN); + + r = repeat_unmount(a->where, MNT_DETACH); + if (r < 0) + log_error_errno(r, "Failed to unmount: %m"); + } +} + +static void automount_done(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + + unmount_autofs(a); + + a->where = mfree(a->where); + + a->tokens = set_free(a->tokens); + a->expire_tokens = set_free(a->expire_tokens); + + a->expire_event_source = sd_event_source_unref(a->expire_event_source); +} + +static int automount_add_mount_links(Automount *a) { + _cleanup_free_ char *parent = NULL; + + assert(a); + + parent = dirname_malloc(a->where); + if (!parent) + return -ENOMEM; + + return unit_require_mounts_for(UNIT(a), parent); +} + +static int automount_add_default_dependencies(Automount *a) { + int r; + + assert(a); + + if (!UNIT(a)->default_dependencies) + return 0; + + if (!MANAGER_IS_SYSTEM(UNIT(a)->manager)) + return 0; + + r = unit_add_two_dependencies_by_name(UNIT(a), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int automount_verify(Automount *a) { + _cleanup_free_ char *e = NULL; + int r; + + assert(a); + + if (UNIT(a)->load_state != UNIT_LOADED) + return 0; + + if (path_equal(a->where, "/")) { + log_unit_error(UNIT(a), "Cannot have an automount unit for the root directory. Refusing."); + return -EINVAL; + } + + r = unit_name_from_path(a->where, ".automount", &e); + if (r < 0) + return log_unit_error(UNIT(a), "Failed to generate unit name from path: %m"); + + if (!unit_has_name(UNIT(a), e)) { + log_unit_error(UNIT(a), "Where= setting doesn't match unit name. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int automount_load(Unit *u) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + /* Load a .automount file */ + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + Unit *x; + + if (!a->where) { + r = unit_name_to_path(u->id, &a->where); + if (r < 0) + return r; + } + + path_kill_slashes(a->where); + + r = unit_load_related_unit(u, ".mount", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + + r = automount_add_mount_links(a); + if (r < 0) + return r; + + r = automount_add_default_dependencies(a); + if (r < 0) + return r; + } + + return automount_verify(a); +} + +static void automount_set_state(Automount *a, AutomountState state) { + AutomountState old_state; + assert(a); + + old_state = a->state; + a->state = state; + + if (state != AUTOMOUNT_RUNNING) + automount_stop_expire(a); + + if (state != AUTOMOUNT_WAITING && + state != AUTOMOUNT_RUNNING) + unmount_autofs(a); + + if (state != old_state) + log_unit_debug(UNIT(a), "Changed %s -> %s", automount_state_to_string(old_state), automount_state_to_string(state)); + + unit_notify(UNIT(a), state_translation_table[old_state], state_translation_table[state], true); +} + +static int automount_coldplug(Unit *u) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(a); + assert(a->state == AUTOMOUNT_DEAD); + + if (a->deserialized_state != a->state) { + + r = open_dev_autofs(u->manager); + if (r < 0) + return r; + + if (a->deserialized_state == AUTOMOUNT_WAITING || + a->deserialized_state == AUTOMOUNT_RUNNING) { + assert(a->pipe_fd >= 0); + + r = sd_event_add_io(u->manager->event, &a->pipe_event_source, a->pipe_fd, EPOLLIN, automount_dispatch_io, u); + if (r < 0) + return r; + + (void) sd_event_source_set_description(a->pipe_event_source, "automount-io"); + if (a->deserialized_state == AUTOMOUNT_RUNNING) { + r = automount_start_expire(a); + if (r < 0) + log_unit_warning_errno(UNIT(a), r, "Failed to start expiration timer, ignoring: %m"); + } + } + + automount_set_state(a, a->deserialized_state); + } + + return 0; +} + +static void automount_dump(Unit *u, FILE *f, const char *prefix) { + char time_string[FORMAT_TIMESPAN_MAX]; + Automount *a = AUTOMOUNT(u); + + assert(a); + + fprintf(f, + "%sAutomount State: %s\n" + "%sResult: %s\n" + "%sWhere: %s\n" + "%sDirectoryMode: %04o\n" + "%sTimeoutIdleUSec: %s\n", + prefix, automount_state_to_string(a->state), + prefix, automount_result_to_string(a->result), + prefix, a->where, + prefix, a->directory_mode, + prefix, format_timespan(time_string, FORMAT_TIMESPAN_MAX, a->timeout_idle_usec, USEC_PER_SEC)); +} + +static void automount_enter_dead(Automount *a, AutomountResult f) { + assert(a); + + if (a->result == AUTOMOUNT_SUCCESS) + a->result = f; + + automount_set_state(a, a->result != AUTOMOUNT_SUCCESS ? AUTOMOUNT_FAILED : AUTOMOUNT_DEAD); +} + +static int open_dev_autofs(Manager *m) { + struct autofs_dev_ioctl param; + + assert(m); + + if (m->dev_autofs_fd >= 0) + return m->dev_autofs_fd; + + label_fix("/dev/autofs", false, false); + + m->dev_autofs_fd = open("/dev/autofs", O_CLOEXEC|O_RDONLY); + if (m->dev_autofs_fd < 0) + return log_error_errno(errno, "Failed to open /dev/autofs: %m"); + + init_autofs_dev_ioctl(¶m); + if (ioctl(m->dev_autofs_fd, AUTOFS_DEV_IOCTL_VERSION, ¶m) < 0) { + m->dev_autofs_fd = safe_close(m->dev_autofs_fd); + return -errno; + } + + log_debug("Autofs kernel version %i.%i", param.ver_major, param.ver_minor); + + return m->dev_autofs_fd; +} + +static int open_ioctl_fd(int dev_autofs_fd, const char *where, dev_t devid) { + struct autofs_dev_ioctl *param; + size_t l; + + assert(dev_autofs_fd >= 0); + assert(where); + + l = sizeof(struct autofs_dev_ioctl) + strlen(where) + 1; + param = alloca(l); + + init_autofs_dev_ioctl(param); + param->size = l; + param->ioctlfd = -1; + param->openmount.devid = devid; + strcpy(param->path, where); + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_OPENMOUNT, param) < 0) + return -errno; + + if (param->ioctlfd < 0) + return -EIO; + + (void) fd_cloexec(param->ioctlfd, true); + return param->ioctlfd; +} + +static int autofs_protocol(int dev_autofs_fd, int ioctl_fd) { + uint32_t major, minor; + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_PROTOVER, ¶m) < 0) + return -errno; + + major = param.protover.version; + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_PROTOSUBVER, ¶m) < 0) + return -errno; + + minor = param.protosubver.sub_version; + + log_debug("Autofs protocol version %i.%i", major, minor); + return 0; +} + +static int autofs_set_timeout(int dev_autofs_fd, int ioctl_fd, usec_t usec) { + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + /* Convert to seconds, rounding up. */ + param.timeout.timeout = (usec + USEC_PER_SEC - 1) / USEC_PER_SEC; + + if (ioctl(dev_autofs_fd, AUTOFS_DEV_IOCTL_TIMEOUT, ¶m) < 0) + return -errno; + + return 0; +} + +static int autofs_send_ready(int dev_autofs_fd, int ioctl_fd, uint32_t token, int status) { + struct autofs_dev_ioctl param; + + assert(dev_autofs_fd >= 0); + assert(ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = ioctl_fd; + + if (status != 0) { + param.fail.token = token; + param.fail.status = status; + } else + param.ready.token = token; + + if (ioctl(dev_autofs_fd, status ? AUTOFS_DEV_IOCTL_FAIL : AUTOFS_DEV_IOCTL_READY, ¶m) < 0) + return -errno; + + return 0; +} + +static int automount_send_ready(Automount *a, Set *tokens, int status) { + _cleanup_close_ int ioctl_fd = -1; + unsigned token; + int r; + + assert(a); + assert(status <= 0); + + if (set_isempty(tokens)) + return 0; + + ioctl_fd = open_ioctl_fd(UNIT(a)->manager->dev_autofs_fd, a->where, a->dev_id); + if (ioctl_fd < 0) + return ioctl_fd; + + if (status != 0) + log_unit_debug_errno(UNIT(a), status, "Sending failure: %m"); + else + log_unit_debug(UNIT(a), "Sending success."); + + r = 0; + + /* Autofs thankfully does not hand out 0 as a token */ + while ((token = PTR_TO_UINT(set_steal_first(tokens)))) { + int k; + + /* Autofs fun fact II: + * + * if you pass a positive status code here, the kernel will + * freeze! Yay! */ + + k = autofs_send_ready(UNIT(a)->manager->dev_autofs_fd, + ioctl_fd, + token, + status); + if (k < 0) + r = k; + } + + return r; +} + +static void automount_trigger_notify(Unit *u, Unit *other) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(a); + assert(other); + + /* Filter out invocations with bogus state */ + if (other->load_state != UNIT_LOADED || other->type != UNIT_MOUNT) + return; + + /* Don't propagate state changes from the mount if we are already down */ + if (!IN_SET(a->state, AUTOMOUNT_WAITING, AUTOMOUNT_RUNNING)) + return; + + /* Propagate start limit hit state */ + if (other->start_limit_hit) { + automount_enter_dead(a, AUTOMOUNT_FAILURE_MOUNT_START_LIMIT_HIT); + return; + } + + /* Don't propagate anything if there's still a job queued */ + if (other->job) + return; + + /* The mount is successfully established */ + if (IN_SET(MOUNT(other)->state, MOUNT_MOUNTED, MOUNT_REMOUNTING)) { + (void) automount_send_ready(a, a->tokens, 0); + + r = automount_start_expire(a); + if (r < 0) + log_unit_warning_errno(UNIT(a), r, "Failed to start expiration timer, ignoring: %m"); + + automount_set_state(a, AUTOMOUNT_RUNNING); + } + + if (IN_SET(MOUNT(other)->state, + MOUNT_MOUNTING, MOUNT_MOUNTING_DONE, + MOUNT_MOUNTED, MOUNT_REMOUNTING, + MOUNT_MOUNTING_SIGTERM, MOUNT_MOUNTING_SIGKILL, + MOUNT_REMOUNTING_SIGTERM, MOUNT_REMOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, MOUNT_UNMOUNTING_SIGKILL, + MOUNT_FAILED)) { + + (void) automount_send_ready(a, a->expire_tokens, -ENODEV); + } + + if (MOUNT(other)->state == MOUNT_DEAD) + (void) automount_send_ready(a, a->expire_tokens, 0); + + /* The mount is in some unhappy state now, let's unfreeze any waiting clients */ + if (IN_SET(MOUNT(other)->state, + MOUNT_DEAD, MOUNT_UNMOUNTING, + MOUNT_MOUNTING_SIGTERM, MOUNT_MOUNTING_SIGKILL, + MOUNT_REMOUNTING_SIGTERM, MOUNT_REMOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, MOUNT_UNMOUNTING_SIGKILL, + MOUNT_FAILED)) { + + (void) automount_send_ready(a, a->tokens, -ENODEV); + + automount_set_state(a, AUTOMOUNT_WAITING); + } +} + +static void automount_enter_waiting(Automount *a) { + _cleanup_close_ int ioctl_fd = -1; + int p[2] = { -1, -1 }; + char name[sizeof("systemd-")-1 + DECIMAL_STR_MAX(pid_t) + 1]; + char options[sizeof("fd=,pgrp=,minproto=5,maxproto=5,direct")-1 + + DECIMAL_STR_MAX(int) + DECIMAL_STR_MAX(gid_t) + 1]; + bool mounted = false; + int r, dev_autofs_fd; + struct stat st; + + assert(a); + assert(a->pipe_fd < 0); + assert(a->where); + + set_clear(a->tokens); + + r = unit_fail_if_symlink(UNIT(a), a->where); + if (r < 0) + goto fail; + + (void) mkdir_p_label(a->where, 0555); + + unit_warn_if_dir_nonempty(UNIT(a), a->where); + + dev_autofs_fd = open_dev_autofs(UNIT(a)->manager); + if (dev_autofs_fd < 0) { + r = dev_autofs_fd; + goto fail; + } + + if (pipe2(p, O_NONBLOCK|O_CLOEXEC) < 0) { + r = -errno; + goto fail; + } + + xsprintf(options, "fd=%i,pgrp="PID_FMT",minproto=5,maxproto=5,direct", p[1], getpgrp()); + xsprintf(name, "systemd-"PID_FMT, getpid()); + if (mount(name, a->where, "autofs", 0, options) < 0) { + r = -errno; + goto fail; + } + + mounted = true; + + p[1] = safe_close(p[1]); + + if (stat(a->where, &st) < 0) { + r = -errno; + goto fail; + } + + ioctl_fd = open_ioctl_fd(dev_autofs_fd, a->where, st.st_dev); + if (ioctl_fd < 0) { + r = ioctl_fd; + goto fail; + } + + r = autofs_protocol(dev_autofs_fd, ioctl_fd); + if (r < 0) + goto fail; + + r = autofs_set_timeout(dev_autofs_fd, ioctl_fd, a->timeout_idle_usec); + if (r < 0) + goto fail; + + /* Autofs fun fact: + * + * Unless we close the ioctl fd here, for some weird reason + * the direct mount will not receive events from the + * kernel. */ + + r = sd_event_add_io(UNIT(a)->manager->event, &a->pipe_event_source, p[0], EPOLLIN, automount_dispatch_io, a); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(a->pipe_event_source, "automount-io"); + + a->pipe_fd = p[0]; + a->dev_id = st.st_dev; + + automount_set_state(a, AUTOMOUNT_WAITING); + + return; + +fail: + log_unit_error_errno(UNIT(a), r, "Failed to initialize automounter: %m"); + + safe_close_pair(p); + + if (mounted) { + r = repeat_unmount(a->where, MNT_DETACH); + if (r < 0) + log_error_errno(r, "Failed to unmount, ignoring: %m"); + } + + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); +} + +static void *expire_thread(void *p) { + struct autofs_dev_ioctl param; + _cleanup_(expire_data_freep) struct expire_data *data = (struct expire_data*)p; + int r; + + assert(data->dev_autofs_fd >= 0); + assert(data->ioctl_fd >= 0); + + init_autofs_dev_ioctl(¶m); + param.ioctlfd = data->ioctl_fd; + + do { + r = ioctl(data->dev_autofs_fd, AUTOFS_DEV_IOCTL_EXPIRE, ¶m); + } while (r >= 0); + + if (errno != EAGAIN) + log_warning_errno(errno, "Failed to expire automount, ignoring: %m"); + + return NULL; +} + +static int automount_dispatch_expire(sd_event_source *source, usec_t usec, void *userdata) { + Automount *a = AUTOMOUNT(userdata); + _cleanup_(expire_data_freep) struct expire_data *data = NULL; + int r; + + assert(a); + assert(source == a->expire_event_source); + + data = new0(struct expire_data, 1); + if (!data) + return log_oom(); + + data->ioctl_fd = -1; + + data->dev_autofs_fd = fcntl(UNIT(a)->manager->dev_autofs_fd, F_DUPFD_CLOEXEC, 3); + if (data->dev_autofs_fd < 0) + return log_unit_error_errno(UNIT(a), errno, "Failed to duplicate autofs fd: %m"); + + data->ioctl_fd = open_ioctl_fd(UNIT(a)->manager->dev_autofs_fd, a->where, a->dev_id); + if (data->ioctl_fd < 0) + return log_unit_error_errno(UNIT(a), data->ioctl_fd, "Couldn't open autofs ioctl fd: %m"); + + r = asynchronous_job(expire_thread, data); + if (r < 0) + return log_unit_error_errno(UNIT(a), r, "Failed to start expire job: %m"); + + data = NULL; + + return automount_start_expire(a); +} + +static int automount_start_expire(Automount *a) { + int r; + usec_t timeout; + + assert(a); + + if (a->timeout_idle_usec == 0) + return 0; + + timeout = now(CLOCK_MONOTONIC) + MAX(a->timeout_idle_usec/3, USEC_PER_SEC); + + if (a->expire_event_source) { + r = sd_event_source_set_time(a->expire_event_source, timeout); + if (r < 0) + return r; + + return sd_event_source_set_enabled(a->expire_event_source, SD_EVENT_ONESHOT); + } + + r = sd_event_add_time( + UNIT(a)->manager->event, + &a->expire_event_source, + CLOCK_MONOTONIC, timeout, 0, + automount_dispatch_expire, a); + if (r < 0) + return r; + + (void) sd_event_source_set_description(a->expire_event_source, "automount-expire"); + + return 0; +} + +static void automount_stop_expire(Automount *a) { + assert(a); + + if (!a->expire_event_source) + return; + + (void) sd_event_source_set_enabled(a->expire_event_source, SD_EVENT_OFF); +} + +static void automount_enter_runnning(Automount *a) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + struct stat st; + int r; + + assert(a); + + /* We don't take mount requests anymore if we are supposed to + * shut down anyway */ + if (unit_stop_pending(UNIT(a))) { + log_unit_debug(UNIT(a), "Suppressing automount request since unit stop is scheduled."); + automount_send_ready(a, a->tokens, -EHOSTDOWN); + automount_send_ready(a, a->expire_tokens, -EHOSTDOWN); + return; + } + + mkdir_p_label(a->where, a->directory_mode); + + /* Before we do anything, let's see if somebody is playing games with us? */ + if (lstat(a->where, &st) < 0) { + log_unit_warning_errno(UNIT(a), errno, "Failed to stat automount point: %m"); + goto fail; + } + + if (!S_ISDIR(st.st_mode) || st.st_dev != a->dev_id) + log_unit_info(UNIT(a), "Automount point already active?"); + else { + Unit *trigger; + + trigger = UNIT_TRIGGER(UNIT(a)); + if (!trigger) { + log_unit_error(UNIT(a), "Unit to trigger vanished."); + goto fail; + } + + r = manager_add_job(UNIT(a)->manager, JOB_START, trigger, JOB_REPLACE, &error, NULL); + if (r < 0) { + log_unit_warning(UNIT(a), "Failed to queue mount startup job: %s", bus_error_message(&error, r)); + goto fail; + } + } + + automount_set_state(a, AUTOMOUNT_RUNNING); + return; + +fail: + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); +} + +static int automount_start(Unit *u) { + Automount *a = AUTOMOUNT(u); + Unit *trigger; + int r; + + assert(a); + assert(a->state == AUTOMOUNT_DEAD || a->state == AUTOMOUNT_FAILED); + + if (path_is_mount_point(a->where, 0) > 0) { + log_unit_error(u, "Path %s is already a mount point, refusing start.", a->where); + return -EEXIST; + } + + trigger = UNIT_TRIGGER(u); + if (!trigger || trigger->load_state != UNIT_LOADED) { + log_unit_error(u, "Refusing to start, unit to trigger not loaded."); + return -ENOENT; + } + + r = unit_start_limit_test(u); + if (r < 0) { + automount_enter_dead(a, AUTOMOUNT_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + a->result = AUTOMOUNT_SUCCESS; + automount_enter_waiting(a); + return 1; +} + +static int automount_stop(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + assert(a->state == AUTOMOUNT_WAITING || a->state == AUTOMOUNT_RUNNING); + + automount_enter_dead(a, AUTOMOUNT_SUCCESS); + return 1; +} + +static int automount_serialize(Unit *u, FILE *f, FDSet *fds) { + Automount *a = AUTOMOUNT(u); + Iterator i; + void *p; + int r; + + assert(a); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", automount_state_to_string(a->state)); + unit_serialize_item(u, f, "result", automount_result_to_string(a->result)); + unit_serialize_item_format(u, f, "dev-id", "%u", (unsigned) a->dev_id); + + SET_FOREACH(p, a->tokens, i) + unit_serialize_item_format(u, f, "token", "%u", PTR_TO_UINT(p)); + SET_FOREACH(p, a->expire_tokens, i) + unit_serialize_item_format(u, f, "expire-token", "%u", PTR_TO_UINT(p)); + + r = unit_serialize_item_fd(u, f, fds, "pipe-fd", a->pipe_fd); + if (r < 0) + return r; + + return 0; +} + +static int automount_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Automount *a = AUTOMOUNT(u); + int r; + + assert(a); + assert(fds); + + if (streq(key, "state")) { + AutomountState state; + + state = automount_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + a->deserialized_state = state; + } else if (streq(key, "result")) { + AutomountResult f; + + f = automount_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != AUTOMOUNT_SUCCESS) + a->result = f; + + } else if (streq(key, "dev-id")) { + unsigned d; + + if (safe_atou(value, &d) < 0) + log_unit_debug(u, "Failed to parse dev-id value: %s", value); + else + a->dev_id = (unsigned) d; + } else if (streq(key, "token")) { + unsigned token; + + if (safe_atou(value, &token) < 0) + log_unit_debug(u, "Failed to parse token value: %s", value); + else { + r = set_ensure_allocated(&a->tokens, NULL); + if (r < 0) { + log_oom(); + return 0; + } + + r = set_put(a->tokens, UINT_TO_PTR(token)); + if (r < 0) + log_unit_error_errno(u, r, "Failed to add token to set: %m"); + } + } else if (streq(key, "expire-token")) { + unsigned token; + + if (safe_atou(value, &token) < 0) + log_unit_debug(u, "Failed to parse token value: %s", value); + else { + r = set_ensure_allocated(&a->expire_tokens, NULL); + if (r < 0) { + log_oom(); + return 0; + } + + r = set_put(a->expire_tokens, UINT_TO_PTR(token)); + if (r < 0) + log_unit_error_errno(u, r, "Failed to add expire token to set: %m"); + } + } else if (streq(key, "pipe-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse pipe-fd value: %s", value); + else { + safe_close(a->pipe_fd); + a->pipe_fd = fdset_remove(fds, fd); + } + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +static UnitActiveState automount_active_state(Unit *u) { + assert(u); + + return state_translation_table[AUTOMOUNT(u)->state]; +} + +static const char *automount_sub_state_to_string(Unit *u) { + assert(u); + + return automount_state_to_string(AUTOMOUNT(u)->state); +} + +static bool automount_check_gc(Unit *u) { + assert(u); + + if (!UNIT_TRIGGER(u)) + return false; + + return UNIT_VTABLE(UNIT_TRIGGER(u))->check_gc(UNIT_TRIGGER(u)); +} + +static int automount_dispatch_io(sd_event_source *s, int fd, uint32_t events, void *userdata) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + union autofs_v5_packet_union packet; + Automount *a = AUTOMOUNT(userdata); + struct stat st; + Unit *trigger; + int r; + + assert(a); + assert(fd == a->pipe_fd); + + if (events != EPOLLIN) { + log_unit_error(UNIT(a), "Got invalid poll event %"PRIu32" on pipe (fd=%d)", events, fd); + goto fail; + } + + r = loop_read_exact(a->pipe_fd, &packet, sizeof(packet), true); + if (r < 0) { + log_unit_error_errno(UNIT(a), r, "Invalid read from pipe: %m"); + goto fail; + } + + switch (packet.hdr.type) { + + case autofs_ptype_missing_direct: + + if (packet.v5_packet.pid > 0) { + _cleanup_free_ char *p = NULL; + + get_process_comm(packet.v5_packet.pid, &p); + log_unit_info(UNIT(a), "Got automount request for %s, triggered by %"PRIu32" (%s)", a->where, packet.v5_packet.pid, strna(p)); + } else + log_unit_debug(UNIT(a), "Got direct mount request on %s", a->where); + + r = set_ensure_allocated(&a->tokens, NULL); + if (r < 0) { + log_unit_error(UNIT(a), "Failed to allocate token set."); + goto fail; + } + + r = set_put(a->tokens, UINT_TO_PTR(packet.v5_packet.wait_queue_token)); + if (r < 0) { + log_unit_error_errno(UNIT(a), r, "Failed to remember token: %m"); + goto fail; + } + + automount_enter_runnning(a); + break; + + case autofs_ptype_expire_direct: + log_unit_debug(UNIT(a), "Got direct umount request on %s", a->where); + + automount_stop_expire(a); + + r = set_ensure_allocated(&a->expire_tokens, NULL); + if (r < 0) { + log_unit_error(UNIT(a), "Failed to allocate token set."); + goto fail; + } + + r = set_put(a->expire_tokens, UINT_TO_PTR(packet.v5_packet.wait_queue_token)); + if (r < 0) { + log_unit_error_errno(UNIT(a), r, "Failed to remember token: %m"); + goto fail; + } + + /* Before we do anything, let's see if somebody is playing games with us? */ + if (lstat(a->where, &st) < 0) { + log_unit_warning_errno(UNIT(a), errno, "Failed to stat automount point: %m"); + goto fail; + } + + if (!S_ISDIR(st.st_mode) || st.st_dev == a->dev_id) { + log_unit_info(UNIT(a), "Automount point already unmounted?"); + automount_send_ready(a, a->expire_tokens, 0); + break; + } + + trigger = UNIT_TRIGGER(UNIT(a)); + if (!trigger) { + log_unit_error(UNIT(a), "Unit to trigger vanished."); + goto fail; + } + + r = manager_add_job(UNIT(a)->manager, JOB_STOP, trigger, JOB_REPLACE, &error, NULL); + if (r < 0) { + log_unit_warning(UNIT(a), "Failed to queue umount startup job: %s", bus_error_message(&error, r)); + goto fail; + } + break; + + default: + log_unit_error(UNIT(a), "Received unknown automount request %i", packet.hdr.type); + break; + } + + return 0; + +fail: + automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); + return 0; +} + +static void automount_shutdown(Manager *m) { + assert(m); + + m->dev_autofs_fd = safe_close(m->dev_autofs_fd); +} + +static void automount_reset_failed(Unit *u) { + Automount *a = AUTOMOUNT(u); + + assert(a); + + if (a->state == AUTOMOUNT_FAILED) + automount_set_state(a, AUTOMOUNT_DEAD); + + a->result = AUTOMOUNT_SUCCESS; +} + +static bool automount_supported(void) { + static int supported = -1; + + if (supported < 0) + supported = access("/dev/autofs", F_OK) >= 0; + + return supported; +} + +static const char* const automount_result_table[_AUTOMOUNT_RESULT_MAX] = { + [AUTOMOUNT_SUCCESS] = "success", + [AUTOMOUNT_FAILURE_RESOURCES] = "resources", + [AUTOMOUNT_FAILURE_START_LIMIT_HIT] = "start-limit-hit", + [AUTOMOUNT_FAILURE_MOUNT_START_LIMIT_HIT] = "mount-start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(automount_result, AutomountResult); + +const UnitVTable automount_vtable = { + .object_size = sizeof(Automount), + + .sections = + "Unit\0" + "Automount\0" + "Install\0", + + .init = automount_init, + .load = automount_load, + .done = automount_done, + + .coldplug = automount_coldplug, + + .dump = automount_dump, + + .start = automount_start, + .stop = automount_stop, + + .serialize = automount_serialize, + .deserialize_item = automount_deserialize_item, + + .active_state = automount_active_state, + .sub_state_to_string = automount_sub_state_to_string, + + .check_gc = automount_check_gc, + + .trigger_notify = automount_trigger_notify, + + .reset_failed = automount_reset_failed, + + .bus_vtable = bus_automount_vtable, + .bus_set_property = bus_automount_set_property, + + .can_transient = true, + + .shutdown = automount_shutdown, + .supported = automount_supported, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Set up automount %s.", + [JOB_FAILED] = "Failed to set up automount %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Unset automount %s.", + [JOB_FAILED] = "Failed to unset automount %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/bus-policy.c b/src/grp-system/libcore/src/bus-policy.c new file mode 100644 index 0000000000..d5bba61a93 --- /dev/null +++ b/src/grp-system/libcore/src/bus-policy.c @@ -0,0 +1,180 @@ +/*** + This file is part of systemd. + + Copyright 2014 Daniel Mack + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <stdlib.h> + +#include "core/bus-policy.h" +#include "sd-bus/bus-kernel.h" +#include "sd-bus/kdbus.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/util.h" + +int bus_kernel_translate_access(BusPolicyAccess access) { + assert(access >= 0); + assert(access < _BUS_POLICY_ACCESS_MAX); + + switch (access) { + + case BUS_POLICY_ACCESS_SEE: + return KDBUS_POLICY_SEE; + + case BUS_POLICY_ACCESS_TALK: + return KDBUS_POLICY_TALK; + + case BUS_POLICY_ACCESS_OWN: + return KDBUS_POLICY_OWN; + + default: + assert_not_reached("Unknown policy access"); + } +} + +int bus_kernel_translate_policy(const BusNamePolicy *policy, struct kdbus_item *item) { + int r; + + assert(policy); + assert(item); + + switch (policy->type) { + + case BUSNAME_POLICY_TYPE_USER: { + const char *user = policy->name; + uid_t uid; + + r = get_user_creds(&user, &uid, NULL, NULL, NULL); + if (r < 0) + return r; + + item->policy_access.type = KDBUS_POLICY_ACCESS_USER; + item->policy_access.id = uid; + break; + } + + case BUSNAME_POLICY_TYPE_GROUP: { + const char *group = policy->name; + gid_t gid; + + r = get_group_creds(&group, &gid); + if (r < 0) + return r; + + item->policy_access.type = KDBUS_POLICY_ACCESS_GROUP; + item->policy_access.id = gid; + break; + } + + default: + assert_not_reached("Unknown policy type"); + } + + item->policy_access.access = bus_kernel_translate_access(policy->access); + + return 0; +} + +int bus_kernel_make_starter( + int fd, + const char *name, + bool activating, + bool accept_fd, + BusNamePolicy *policy, + BusPolicyAccess world_policy) { + + struct kdbus_cmd_free cmd_free = { .size = sizeof(cmd_free) }; + struct kdbus_cmd_hello *hello; + struct kdbus_item *n; + size_t policy_cnt = 0; + BusNamePolicy *po; + size_t size; + int r; + + assert(fd >= 0); + assert(name); + + LIST_FOREACH(policy, po, policy) + policy_cnt++; + + if (world_policy >= 0) + policy_cnt++; + + size = offsetof(struct kdbus_cmd_hello, items) + + ALIGN8(offsetof(struct kdbus_item, str) + strlen(name) + 1) + + policy_cnt * ALIGN8(offsetof(struct kdbus_item, policy_access) + sizeof(struct kdbus_policy_access)); + + hello = alloca0_align(size, 8); + + n = hello->items; + strcpy(n->str, name); + n->size = offsetof(struct kdbus_item, str) + strlen(n->str) + 1; + n->type = KDBUS_ITEM_NAME; + n = KDBUS_ITEM_NEXT(n); + + LIST_FOREACH(policy, po, policy) { + n->type = KDBUS_ITEM_POLICY_ACCESS; + n->size = offsetof(struct kdbus_item, policy_access) + sizeof(struct kdbus_policy_access); + + r = bus_kernel_translate_policy(po, n); + if (r < 0) + return r; + + n = KDBUS_ITEM_NEXT(n); + } + + if (world_policy >= 0) { + n->type = KDBUS_ITEM_POLICY_ACCESS; + n->size = offsetof(struct kdbus_item, policy_access) + sizeof(struct kdbus_policy_access); + n->policy_access.type = KDBUS_POLICY_ACCESS_WORLD; + n->policy_access.access = bus_kernel_translate_access(world_policy); + } + + hello->size = size; + hello->flags = + (activating ? KDBUS_HELLO_ACTIVATOR : KDBUS_HELLO_POLICY_HOLDER) | + (accept_fd ? KDBUS_HELLO_ACCEPT_FD : 0); + hello->pool_size = KDBUS_POOL_SIZE; + hello->attach_flags_send = _KDBUS_ATTACH_ANY; + hello->attach_flags_recv = _KDBUS_ATTACH_ANY; + + if (ioctl(fd, KDBUS_CMD_HELLO, hello) < 0) { + if (errno == ENOTTY) /* Major API change */ + return -ESOCKTNOSUPPORT; + return -errno; + } + + /* not interested in any output values */ + cmd_free.offset = hello->offset; + (void) ioctl(fd, KDBUS_CMD_FREE, &cmd_free); + + /* The higher 32bit of the bus_flags fields are considered + * 'incompatible flags'. Refuse them all for now. */ + if (hello->bus_flags > 0xFFFFFFFFULL) + return -ESOCKTNOSUPPORT; + + return fd; +} + +static const char* const bus_policy_access_table[_BUS_POLICY_ACCESS_MAX] = { + [BUS_POLICY_ACCESS_SEE] = "see", + [BUS_POLICY_ACCESS_TALK] = "talk", + [BUS_POLICY_ACCESS_OWN] = "own", +}; + +DEFINE_STRING_TABLE_LOOKUP(bus_policy_access, BusPolicyAccess); diff --git a/src/grp-system/libcore/src/busname.c b/src/grp-system/libcore/src/busname.c new file mode 100644 index 0000000000..3d365530bf --- /dev/null +++ b/src/grp-system/libcore/src/busname.c @@ -0,0 +1,1082 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/mman.h> + +#include "core/bus-policy.h" +#include "core/busname.h" +#include "core/service.h" +#include "sd-bus/bus-internal.h" +#include "sd-bus/bus-kernel.h" +#include "sd-bus/bus-util.h" +#include "sd-bus/kdbus.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" + +#include "dbus-busname.h" + +static const UnitActiveState state_translation_table[_BUSNAME_STATE_MAX] = { + [BUSNAME_DEAD] = UNIT_INACTIVE, + [BUSNAME_MAKING] = UNIT_ACTIVATING, + [BUSNAME_REGISTERED] = UNIT_ACTIVE, + [BUSNAME_LISTENING] = UNIT_ACTIVE, + [BUSNAME_RUNNING] = UNIT_ACTIVE, + [BUSNAME_SIGTERM] = UNIT_DEACTIVATING, + [BUSNAME_SIGKILL] = UNIT_DEACTIVATING, + [BUSNAME_FAILED] = UNIT_FAILED +}; + +static int busname_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int busname_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); + +static void busname_init(Unit *u) { + BusName *n = BUSNAME(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + n->starter_fd = -1; + n->accept_fd = true; + n->activating = true; + + n->timeout_usec = u->manager->default_timeout_start_usec; +} + +static void busname_unwatch_control_pid(BusName *n) { + assert(n); + + if (n->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(n), n->control_pid); + n->control_pid = 0; +} + +static void busname_free_policy(BusName *n) { + BusNamePolicy *p; + + assert(n); + + while ((p = n->policy)) { + LIST_REMOVE(policy, n->policy, p); + + free(p->name); + free(p); + } +} + +static void busname_close_fd(BusName *n) { + assert(n); + + n->starter_event_source = sd_event_source_unref(n->starter_event_source); + n->starter_fd = safe_close(n->starter_fd); +} + +static void busname_done(Unit *u) { + BusName *n = BUSNAME(u); + + assert(n); + + n->name = mfree(n->name); + + busname_free_policy(n); + busname_unwatch_control_pid(n); + busname_close_fd(n); + + unit_ref_unset(&n->service); + + n->timer_event_source = sd_event_source_unref(n->timer_event_source); +} + +static int busname_arm_timer(BusName *n, usec_t usec) { + int r; + + assert(n); + + if (n->timer_event_source) { + r = sd_event_source_set_time(n->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(n->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(n)->manager->event, + &n->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + busname_dispatch_timer, n); + if (r < 0) + return r; + + (void) sd_event_source_set_description(n->timer_event_source, "busname-timer"); + + return 0; +} + +static int busname_add_default_default_dependencies(BusName *n) { + int r; + + assert(n); + + r = unit_add_dependency_by_name(UNIT(n), UNIT_BEFORE, SPECIAL_BUSNAMES_TARGET, NULL, true); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(UNIT(n)->manager)) { + r = unit_add_two_dependencies_by_name(UNIT(n), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(n), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int busname_add_extras(BusName *n) { + Unit *u = UNIT(n); + int r; + + assert(n); + + if (!n->name) { + r = unit_name_to_prefix(u->id, &n->name); + if (r < 0) + return r; + } + + if (!u->description) { + r = unit_set_description(u, n->name); + if (r < 0) + return r; + } + + if (n->activating) { + if (!UNIT_DEREF(n->service)) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + unit_ref_set(&n->service, x); + } + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, UNIT_DEREF(n->service), true); + if (r < 0) + return r; + } + + if (u->default_dependencies) { + r = busname_add_default_default_dependencies(n); + if (r < 0) + return r; + } + + return 0; +} + +static int busname_verify(BusName *n) { + char *e; + + assert(n); + + if (UNIT(n)->load_state != UNIT_LOADED) + return 0; + + if (!service_name_is_valid(n->name)) { + log_unit_error(UNIT(n), "Name= setting is not a valid service name Refusing."); + return -EINVAL; + } + + e = strjoina(n->name, ".busname"); + if (!unit_has_name(UNIT(n), e)) { + log_unit_error(UNIT(n), "Name= setting doesn't match unit name. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int busname_load(Unit *u) { + BusName *n = BUSNAME(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + /* This is a new unit? Then let's add in some extras */ + r = busname_add_extras(n); + if (r < 0) + return r; + } + + return busname_verify(n); +} + +static void busname_dump(Unit *u, FILE *f, const char *prefix) { + BusName *n = BUSNAME(u); + + assert(n); + assert(f); + + fprintf(f, + "%sBus Name State: %s\n" + "%sResult: %s\n" + "%sName: %s\n" + "%sActivating: %s\n" + "%sAccept FD: %s\n", + prefix, busname_state_to_string(n->state), + prefix, busname_result_to_string(n->result), + prefix, n->name, + prefix, yes_no(n->activating), + prefix, yes_no(n->accept_fd)); + + if (n->control_pid > 0) + fprintf(f, + "%sControl PID: "PID_FMT"\n", + prefix, n->control_pid); +} + +static void busname_unwatch_fd(BusName *n) { + int r; + + assert(n); + + if (!n->starter_event_source) + return; + + r = sd_event_source_set_enabled(n->starter_event_source, SD_EVENT_OFF); + if (r < 0) + log_unit_debug_errno(UNIT(n), r, "Failed to disable event source: %m"); +} + +static int busname_watch_fd(BusName *n) { + int r; + + assert(n); + + if (n->starter_fd < 0) + return 0; + + if (n->starter_event_source) { + r = sd_event_source_set_enabled(n->starter_event_source, SD_EVENT_ON); + if (r < 0) + goto fail; + } else { + r = sd_event_add_io(UNIT(n)->manager->event, &n->starter_event_source, n->starter_fd, EPOLLIN, busname_dispatch_io, n); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(n->starter_event_source, "busname-starter"); + } + + return 0; + +fail: + log_unit_warning_errno(UNIT(n), r, "Failed to watch starter fd: %m"); + busname_unwatch_fd(n); + return r; +} + +static int busname_open_fd(BusName *n) { + _cleanup_free_ char *path = NULL; + const char *mode; + + assert(n); + + if (n->starter_fd >= 0) + return 0; + + mode = MANAGER_IS_SYSTEM(UNIT(n)->manager) ? "system" : "user"; + n->starter_fd = bus_kernel_open_bus_fd(mode, &path); + if (n->starter_fd < 0) + return log_unit_warning_errno(UNIT(n), n->starter_fd, "Failed to open %s: %m", path ?: "kdbus"); + + return 0; +} + +static void busname_set_state(BusName *n, BusNameState state) { + BusNameState old_state; + assert(n); + + old_state = n->state; + n->state = state; + + if (!IN_SET(state, BUSNAME_MAKING, BUSNAME_SIGTERM, BUSNAME_SIGKILL)) { + n->timer_event_source = sd_event_source_unref(n->timer_event_source); + busname_unwatch_control_pid(n); + } + + if (state != BUSNAME_LISTENING) + busname_unwatch_fd(n); + + if (!IN_SET(state, BUSNAME_LISTENING, BUSNAME_MAKING, BUSNAME_REGISTERED, BUSNAME_RUNNING)) + busname_close_fd(n); + + if (state != old_state) + log_unit_debug(UNIT(n), "Changed %s -> %s", busname_state_to_string(old_state), busname_state_to_string(state)); + + unit_notify(UNIT(n), state_translation_table[old_state], state_translation_table[state], true); +} + +static int busname_coldplug(Unit *u) { + BusName *n = BUSNAME(u); + int r; + + assert(n); + assert(n->state == BUSNAME_DEAD); + + if (n->deserialized_state == n->state) + return 0; + + if (n->control_pid > 0 && + pid_is_unwaited(n->control_pid) && + IN_SET(n->deserialized_state, BUSNAME_MAKING, BUSNAME_SIGTERM, BUSNAME_SIGKILL)) { + + r = unit_watch_pid(UNIT(n), n->control_pid); + if (r < 0) + return r; + + r = busname_arm_timer(n, usec_add(u->state_change_timestamp.monotonic, n->timeout_usec)); + if (r < 0) + return r; + } + + if (IN_SET(n->deserialized_state, BUSNAME_MAKING, BUSNAME_LISTENING, BUSNAME_REGISTERED, BUSNAME_RUNNING)) { + r = busname_open_fd(n); + if (r < 0) + return r; + } + + if (n->deserialized_state == BUSNAME_LISTENING) { + r = busname_watch_fd(n); + if (r < 0) + return r; + } + + busname_set_state(n, n->deserialized_state); + return 0; +} + +static int busname_make_starter(BusName *n, pid_t *_pid) { + pid_t pid; + int r; + + r = busname_arm_timer(n, usec_add(now(CLOCK_MONOTONIC), n->timeout_usec)); + if (r < 0) + goto fail; + + /* We have to resolve the user/group names out-of-process, + * hence let's fork here. It's messy, but well, what can we + * do? */ + + pid = fork(); + if (pid < 0) + return -errno; + + if (pid == 0) { + int ret; + + (void) default_signals(SIGNALS_CRASH_HANDLER, SIGNALS_IGNORE, -1); + (void) ignore_signals(SIGPIPE, -1); + log_forget_fds(); + + r = bus_kernel_make_starter(n->starter_fd, n->name, n->activating, n->accept_fd, n->policy, n->policy_world); + if (r < 0) { + ret = EXIT_MAKE_STARTER; + goto fail_child; + } + + _exit(0); + + fail_child: + log_open(); + log_error_errno(r, "Failed to create starter connection at step %s: %m", exit_status_to_string(ret, EXIT_STATUS_SYSTEMD)); + + _exit(ret); + } + + r = unit_watch_pid(UNIT(n), pid); + if (r < 0) + goto fail; + + *_pid = pid; + return 0; + +fail: + n->timer_event_source = sd_event_source_unref(n->timer_event_source); + return r; +} + +static void busname_enter_dead(BusName *n, BusNameResult f) { + assert(n); + + if (n->result == BUSNAME_SUCCESS) + n->result = f; + + busname_set_state(n, n->result != BUSNAME_SUCCESS ? BUSNAME_FAILED : BUSNAME_DEAD); +} + +static void busname_enter_signal(BusName *n, BusNameState state, BusNameResult f) { + KillContext kill_context = {}; + int r; + + assert(n); + + if (n->result == BUSNAME_SUCCESS) + n->result = f; + + kill_context_init(&kill_context); + + r = unit_kill_context(UNIT(n), + &kill_context, + state != BUSNAME_SIGTERM ? KILL_KILL : KILL_TERMINATE, + -1, + n->control_pid, + false); + if (r < 0) { + log_unit_warning_errno(UNIT(n), r, "Failed to kill control process: %m"); + goto fail; + } + + if (r > 0) { + r = busname_arm_timer(n, usec_add(now(CLOCK_MONOTONIC), n->timeout_usec)); + if (r < 0) { + log_unit_warning_errno(UNIT(n), r, "Failed to arm timer: %m"); + goto fail; + } + + busname_set_state(n, state); + } else if (state == BUSNAME_SIGTERM) + busname_enter_signal(n, BUSNAME_SIGKILL, BUSNAME_SUCCESS); + else + busname_enter_dead(n, BUSNAME_SUCCESS); + + return; + +fail: + busname_enter_dead(n, BUSNAME_FAILURE_RESOURCES); +} + +static void busname_enter_listening(BusName *n) { + int r; + + assert(n); + + if (n->activating) { + r = busname_watch_fd(n); + if (r < 0) { + log_unit_warning_errno(UNIT(n), r, "Failed to watch names: %m"); + goto fail; + } + + busname_set_state(n, BUSNAME_LISTENING); + } else + busname_set_state(n, BUSNAME_REGISTERED); + + return; + +fail: + busname_enter_signal(n, BUSNAME_SIGTERM, BUSNAME_FAILURE_RESOURCES); +} + +static void busname_enter_making(BusName *n) { + int r; + + assert(n); + + r = busname_open_fd(n); + if (r < 0) + goto fail; + + if (n->policy) { + /* If there is a policy, we need to resolve user/group + * names, which we can't do from PID1, hence let's + * fork. */ + busname_unwatch_control_pid(n); + + r = busname_make_starter(n, &n->control_pid); + if (r < 0) { + log_unit_warning_errno(UNIT(n), r, "Failed to fork 'making' task: %m"); + goto fail; + } + + busname_set_state(n, BUSNAME_MAKING); + } else { + /* If there is no policy, we can do everything + * directly from PID 1, hence do so. */ + + r = bus_kernel_make_starter(n->starter_fd, n->name, n->activating, n->accept_fd, NULL, n->policy_world); + if (r < 0) { + log_unit_warning_errno(UNIT(n), r, "Failed to make starter: %m"); + goto fail; + } + + busname_enter_listening(n); + } + + return; + +fail: + busname_enter_dead(n, BUSNAME_FAILURE_RESOURCES); +} + +static void busname_enter_running(BusName *n) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + bool pending = false; + Unit *other; + Iterator i; + int r; + + assert(n); + + if (!n->activating) + return; + + /* We don't take connections anymore if we are supposed to + * shut down anyway */ + + if (unit_stop_pending(UNIT(n))) { + log_unit_debug(UNIT(n), "Suppressing activation request since unit stop is scheduled."); + + /* Flush all queued activation reqeuest by closing and reopening the connection */ + bus_kernel_drop_one(n->starter_fd); + + busname_enter_listening(n); + return; + } + + /* If there's already a start pending don't bother to do + * anything */ + SET_FOREACH(other, UNIT(n)->dependencies[UNIT_TRIGGERS], i) + if (unit_active_or_pending(other)) { + pending = true; + break; + } + + if (!pending) { + if (!UNIT_ISSET(n->service)) { + log_unit_error(UNIT(n), "Service to activate vanished, refusing activation."); + r = -ENOENT; + goto fail; + } + + r = manager_add_job(UNIT(n)->manager, JOB_START, UNIT_DEREF(n->service), JOB_REPLACE, &error, NULL); + if (r < 0) + goto fail; + } + + busname_set_state(n, BUSNAME_RUNNING); + return; + +fail: + log_unit_warning(UNIT(n), "Failed to queue service startup job: %s", bus_error_message(&error, r)); + busname_enter_dead(n, BUSNAME_FAILURE_RESOURCES); +} + +static int busname_start(Unit *u) { + BusName *n = BUSNAME(u); + int r; + + assert(n); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (IN_SET(n->state, BUSNAME_SIGTERM, BUSNAME_SIGKILL)) + return -EAGAIN; + + /* Already on it! */ + if (n->state == BUSNAME_MAKING) + return 0; + + if (n->activating && UNIT_ISSET(n->service)) { + Service *service; + + service = SERVICE(UNIT_DEREF(n->service)); + + if (UNIT(service)->load_state != UNIT_LOADED) { + log_unit_error(u, "Bus service %s not loaded, refusing.", UNIT(service)->id); + return -ENOENT; + } + } + + assert(IN_SET(n->state, BUSNAME_DEAD, BUSNAME_FAILED)); + + r = unit_start_limit_test(u); + if (r < 0) { + busname_enter_dead(n, BUSNAME_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + n->result = BUSNAME_SUCCESS; + busname_enter_making(n); + + return 1; +} + +static int busname_stop(Unit *u) { + BusName *n = BUSNAME(u); + + assert(n); + + /* Already on it */ + if (IN_SET(n->state, BUSNAME_SIGTERM, BUSNAME_SIGKILL)) + return 0; + + /* If there's already something running, we go directly into + * kill mode. */ + + if (n->state == BUSNAME_MAKING) { + busname_enter_signal(n, BUSNAME_SIGTERM, BUSNAME_SUCCESS); + return -EAGAIN; + } + + assert(IN_SET(n->state, BUSNAME_REGISTERED, BUSNAME_LISTENING, BUSNAME_RUNNING)); + + busname_enter_dead(n, BUSNAME_SUCCESS); + return 1; +} + +static int busname_serialize(Unit *u, FILE *f, FDSet *fds) { + BusName *n = BUSNAME(u); + int r; + + assert(n); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", busname_state_to_string(n->state)); + unit_serialize_item(u, f, "result", busname_result_to_string(n->result)); + + if (n->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", PID_FMT, n->control_pid); + + r = unit_serialize_item_fd(u, f, fds, "starter-fd", n->starter_fd); + if (r < 0) + return r; + + return 0; +} + +static int busname_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + BusName *n = BUSNAME(u); + + assert(n); + assert(key); + assert(value); + + if (streq(key, "state")) { + BusNameState state; + + state = busname_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + n->deserialized_state = state; + + } else if (streq(key, "result")) { + BusNameResult f; + + f = busname_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != BUSNAME_SUCCESS) + n->result = f; + + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse control-pid value: %s", value); + else + n->control_pid = pid; + } else if (streq(key, "starter-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse starter fd value: %s", value); + else { + safe_close(n->starter_fd); + n->starter_fd = fdset_remove(fds, fd); + } + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState busname_active_state(Unit *u) { + assert(u); + + return state_translation_table[BUSNAME(u)->state]; +} + +_pure_ static const char *busname_sub_state_to_string(Unit *u) { + assert(u); + + return busname_state_to_string(BUSNAME(u)->state); +} + +static int busname_peek_message(BusName *n) { + struct kdbus_cmd_recv cmd_recv = { + .size = sizeof(cmd_recv), + .flags = KDBUS_RECV_PEEK, + }; + struct kdbus_cmd_free cmd_free = { + .size = sizeof(cmd_free), + }; + const char *comm = NULL; + struct kdbus_item *d; + struct kdbus_msg *k; + size_t start, ps, sz, delta; + void *p = NULL; + pid_t pid = 0; + int r; + + /* Generate a friendly debug log message about which process + * caused triggering of this bus name. This simply peeks the + * metadata of the first queued message and logs it. */ + + assert(n); + + /* Let's shortcut things a bit, if debug logging is turned off + * anyway. */ + + if (log_get_max_level() < LOG_DEBUG) + return 0; + + r = ioctl(n->starter_fd, KDBUS_CMD_RECV, &cmd_recv); + if (r < 0) { + if (errno == EINTR || errno == EAGAIN) + return 0; + + return log_unit_error_errno(UNIT(n), errno, "Failed to query activation message: %m"); + } + + /* We map as late as possible, and unmap imemdiately after + * use. On 32bit address space is scarce and we want to be + * able to handle a lot of activator connections at the same + * time, and hence shouldn't keep the mmap()s around for + * longer than necessary. */ + + ps = page_size(); + start = (cmd_recv.msg.offset / ps) * ps; + delta = cmd_recv.msg.offset - start; + sz = PAGE_ALIGN(delta + cmd_recv.msg.msg_size); + + p = mmap(NULL, sz, PROT_READ, MAP_SHARED, n->starter_fd, start); + if (p == MAP_FAILED) { + r = log_unit_error_errno(UNIT(n), errno, "Failed to map activation message: %m"); + goto finish; + } + + k = (struct kdbus_msg *) ((uint8_t *) p + delta); + KDBUS_ITEM_FOREACH(d, k, items) { + switch (d->type) { + + case KDBUS_ITEM_PIDS: + pid = d->pids.pid; + break; + + case KDBUS_ITEM_PID_COMM: + comm = d->str; + break; + } + } + + if (pid > 0) + log_unit_debug(UNIT(n), "Activation triggered by process " PID_FMT " (%s)", pid, strna(comm)); + + r = 0; + +finish: + if (p) + (void) munmap(p, sz); + + cmd_free.offset = cmd_recv.msg.offset; + if (ioctl(n->starter_fd, KDBUS_CMD_FREE, &cmd_free) < 0) + log_unit_warning(UNIT(n), "Failed to free peeked message, ignoring: %m"); + + return r; +} + +static int busname_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + BusName *n = userdata; + + assert(n); + assert(fd >= 0); + + if (n->state != BUSNAME_LISTENING) + return 0; + + log_unit_debug(UNIT(n), "Activation request"); + + if (revents != EPOLLIN) { + log_unit_error(UNIT(n), "Got unexpected poll event (0x%x) on starter fd.", revents); + goto fail; + } + + busname_peek_message(n); + busname_enter_running(n); + return 0; +fail: + + busname_enter_dead(n, BUSNAME_FAILURE_RESOURCES); + return 0; +} + +static void busname_sigchld_event(Unit *u, pid_t pid, int code, int status) { + BusName *n = BUSNAME(u); + BusNameResult f; + + assert(n); + assert(pid >= 0); + + if (pid != n->control_pid) + return; + + n->control_pid = 0; + + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) + f = BUSNAME_SUCCESS; + else if (code == CLD_EXITED) + f = BUSNAME_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = BUSNAME_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = BUSNAME_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown sigchld code"); + + log_unit_full(u, f == BUSNAME_SUCCESS ? LOG_DEBUG : LOG_NOTICE, 0, + "Control process exited, code=%s status=%i", sigchld_code_to_string(code), status); + + if (n->result == BUSNAME_SUCCESS) + n->result = f; + + switch (n->state) { + + case BUSNAME_MAKING: + if (f == BUSNAME_SUCCESS) + busname_enter_listening(n); + else + busname_enter_signal(n, BUSNAME_SIGTERM, f); + break; + + case BUSNAME_SIGTERM: + case BUSNAME_SIGKILL: + busname_enter_dead(n, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static int busname_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + BusName *n = BUSNAME(userdata); + + assert(n); + assert(n->timer_event_source == source); + + switch (n->state) { + + case BUSNAME_MAKING: + log_unit_warning(UNIT(n), "Making timed out. Terminating."); + busname_enter_signal(n, BUSNAME_SIGTERM, BUSNAME_FAILURE_TIMEOUT); + break; + + case BUSNAME_SIGTERM: + log_unit_warning(UNIT(n), "Stopping timed out. Killing."); + busname_enter_signal(n, BUSNAME_SIGKILL, BUSNAME_FAILURE_TIMEOUT); + break; + + case BUSNAME_SIGKILL: + log_unit_warning(UNIT(n), "Processes still around after SIGKILL. Ignoring."); + busname_enter_dead(n, BUSNAME_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +static void busname_reset_failed(Unit *u) { + BusName *n = BUSNAME(u); + + assert(n); + + if (n->state == BUSNAME_FAILED) + busname_set_state(n, BUSNAME_DEAD); + + n->result = BUSNAME_SUCCESS; +} + +static void busname_trigger_notify(Unit *u, Unit *other) { + BusName *n = BUSNAME(u); + + assert(n); + assert(other); + + if (!IN_SET(n->state, BUSNAME_RUNNING, BUSNAME_LISTENING)) + return; + + if (other->start_limit_hit) { + busname_enter_dead(n, BUSNAME_FAILURE_SERVICE_START_LIMIT_HIT); + return; + } + + if (other->load_state != UNIT_LOADED || other->type != UNIT_SERVICE) + return; + + if (IN_SET(SERVICE(other)->state, + SERVICE_DEAD, SERVICE_FAILED, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL, + SERVICE_AUTO_RESTART)) + busname_enter_listening(n); + + if (SERVICE(other)->state == SERVICE_RUNNING) + busname_set_state(n, BUSNAME_RUNNING); +} + +static int busname_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, BUSNAME(u)->control_pid, error); +} + +static int busname_get_timeout(Unit *u, usec_t *timeout) { + BusName *n = BUSNAME(u); + usec_t t; + int r; + + if (!n->timer_event_source) + return 0; + + r = sd_event_source_get_time(n->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +static bool busname_supported(void) { + return false; +} + +static int busname_control_pid(Unit *u) { + BusName *n = BUSNAME(u); + + assert(n); + + return n->control_pid; +} + +static const char* const busname_result_table[_BUSNAME_RESULT_MAX] = { + [BUSNAME_SUCCESS] = "success", + [BUSNAME_FAILURE_RESOURCES] = "resources", + [BUSNAME_FAILURE_TIMEOUT] = "timeout", + [BUSNAME_FAILURE_EXIT_CODE] = "exit-code", + [BUSNAME_FAILURE_SIGNAL] = "signal", + [BUSNAME_FAILURE_CORE_DUMP] = "core-dump", + [BUSNAME_FAILURE_START_LIMIT_HIT] = "start-limit-hit", + [BUSNAME_FAILURE_SERVICE_START_LIMIT_HIT] = "service-start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(busname_result, BusNameResult); + +const UnitVTable busname_vtable = { + .object_size = sizeof(BusName), + + .sections = + "Unit\0" + "BusName\0" + "Install\0", + .private_section = "BusName", + + .init = busname_init, + .done = busname_done, + .load = busname_load, + + .coldplug = busname_coldplug, + + .dump = busname_dump, + + .start = busname_start, + .stop = busname_stop, + + .kill = busname_kill, + + .get_timeout = busname_get_timeout, + + .serialize = busname_serialize, + .deserialize_item = busname_deserialize_item, + + .active_state = busname_active_state, + .sub_state_to_string = busname_sub_state_to_string, + + .sigchld_event = busname_sigchld_event, + + .trigger_notify = busname_trigger_notify, + + .reset_failed = busname_reset_failed, + + .supported = busname_supported, + + .control_pid = busname_control_pid, + + .bus_vtable = bus_busname_vtable, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Listening on %s.", + [JOB_FAILED] = "Failed to listen on %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Closed %s.", + [JOB_FAILED] = "Failed stopping %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/cgroup.c b/src/grp-system/libcore/src/cgroup.c new file mode 100644 index 0000000000..62222903fe --- /dev/null +++ b/src/grp-system/libcore/src/cgroup.c @@ -0,0 +1,2170 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <fnmatch.h> + +#include "core/cgroup.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" + +#define CGROUP_CPU_QUOTA_PERIOD_USEC ((usec_t) 100 * USEC_PER_MSEC) + +static void cgroup_compat_warn(void) { + static bool cgroup_compat_warned = false; + + if (cgroup_compat_warned) + return; + + log_warning("cgroup compatibility translation between legacy and unified hierarchy settings activated. See cgroup-compat debug messages for details."); + cgroup_compat_warned = true; +} + +#define log_cgroup_compat(unit, fmt, ...) do { \ + cgroup_compat_warn(); \ + log_unit_debug(unit, "cgroup-compat: " fmt, ##__VA_ARGS__); \ + } while (false) + +void cgroup_context_init(CGroupContext *c) { + assert(c); + + /* Initialize everything to the kernel defaults, assuming the + * structure is preinitialized to 0 */ + + c->cpu_weight = CGROUP_WEIGHT_INVALID; + c->startup_cpu_weight = CGROUP_WEIGHT_INVALID; + c->cpu_quota_per_sec_usec = USEC_INFINITY; + + c->cpu_shares = CGROUP_CPU_SHARES_INVALID; + c->startup_cpu_shares = CGROUP_CPU_SHARES_INVALID; + + c->memory_high = CGROUP_LIMIT_MAX; + c->memory_max = CGROUP_LIMIT_MAX; + c->memory_swap_max = CGROUP_LIMIT_MAX; + + c->memory_limit = CGROUP_LIMIT_MAX; + + c->io_weight = CGROUP_WEIGHT_INVALID; + c->startup_io_weight = CGROUP_WEIGHT_INVALID; + + c->blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + c->startup_blockio_weight = CGROUP_BLKIO_WEIGHT_INVALID; + + c->tasks_max = (uint64_t) -1; +} + +void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) { + assert(c); + assert(a); + + LIST_REMOVE(device_allow, c->device_allow, a); + free(a->path); + free(a); +} + +void cgroup_context_free_io_device_weight(CGroupContext *c, CGroupIODeviceWeight *w) { + assert(c); + assert(w); + + LIST_REMOVE(device_weights, c->io_device_weights, w); + free(w->path); + free(w); +} + +void cgroup_context_free_io_device_limit(CGroupContext *c, CGroupIODeviceLimit *l) { + assert(c); + assert(l); + + LIST_REMOVE(device_limits, c->io_device_limits, l); + free(l->path); + free(l); +} + +void cgroup_context_free_blockio_device_weight(CGroupContext *c, CGroupBlockIODeviceWeight *w) { + assert(c); + assert(w); + + LIST_REMOVE(device_weights, c->blockio_device_weights, w); + free(w->path); + free(w); +} + +void cgroup_context_free_blockio_device_bandwidth(CGroupContext *c, CGroupBlockIODeviceBandwidth *b) { + assert(c); + assert(b); + + LIST_REMOVE(device_bandwidths, c->blockio_device_bandwidths, b); + free(b->path); + free(b); +} + +void cgroup_context_done(CGroupContext *c) { + assert(c); + + while (c->io_device_weights) + cgroup_context_free_io_device_weight(c, c->io_device_weights); + + while (c->io_device_limits) + cgroup_context_free_io_device_limit(c, c->io_device_limits); + + while (c->blockio_device_weights) + cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); + + while (c->blockio_device_bandwidths) + cgroup_context_free_blockio_device_bandwidth(c, c->blockio_device_bandwidths); + + while (c->device_allow) + cgroup_context_free_device_allow(c, c->device_allow); +} + +void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { + CGroupIODeviceLimit *il; + CGroupIODeviceWeight *iw; + CGroupBlockIODeviceBandwidth *b; + CGroupBlockIODeviceWeight *w; + CGroupDeviceAllow *a; + char u[FORMAT_TIMESPAN_MAX]; + + assert(c); + assert(f); + + prefix = strempty(prefix); + + fprintf(f, + "%sCPUAccounting=%s\n" + "%sIOAccounting=%s\n" + "%sBlockIOAccounting=%s\n" + "%sMemoryAccounting=%s\n" + "%sTasksAccounting=%s\n" + "%sCPUWeight=%" PRIu64 "\n" + "%sStartupCPUWeight=%" PRIu64 "\n" + "%sCPUShares=%" PRIu64 "\n" + "%sStartupCPUShares=%" PRIu64 "\n" + "%sCPUQuotaPerSecSec=%s\n" + "%sIOWeight=%" PRIu64 "\n" + "%sStartupIOWeight=%" PRIu64 "\n" + "%sBlockIOWeight=%" PRIu64 "\n" + "%sStartupBlockIOWeight=%" PRIu64 "\n" + "%sMemoryLow=%" PRIu64 "\n" + "%sMemoryHigh=%" PRIu64 "\n" + "%sMemoryMax=%" PRIu64 "\n" + "%sMemorySwapMax=%" PRIu64 "\n" + "%sMemoryLimit=%" PRIu64 "\n" + "%sTasksMax=%" PRIu64 "\n" + "%sDevicePolicy=%s\n" + "%sDelegate=%s\n", + prefix, yes_no(c->cpu_accounting), + prefix, yes_no(c->io_accounting), + prefix, yes_no(c->blockio_accounting), + prefix, yes_no(c->memory_accounting), + prefix, yes_no(c->tasks_accounting), + prefix, c->cpu_weight, + prefix, c->startup_cpu_weight, + prefix, c->cpu_shares, + prefix, c->startup_cpu_shares, + prefix, format_timespan(u, sizeof(u), c->cpu_quota_per_sec_usec, 1), + prefix, c->io_weight, + prefix, c->startup_io_weight, + prefix, c->blockio_weight, + prefix, c->startup_blockio_weight, + prefix, c->memory_low, + prefix, c->memory_high, + prefix, c->memory_max, + prefix, c->memory_swap_max, + prefix, c->memory_limit, + prefix, c->tasks_max, + prefix, cgroup_device_policy_to_string(c->device_policy), + prefix, yes_no(c->delegate)); + + LIST_FOREACH(device_allow, a, c->device_allow) + fprintf(f, + "%sDeviceAllow=%s %s%s%s\n", + prefix, + a->path, + a->r ? "r" : "", a->w ? "w" : "", a->m ? "m" : ""); + + LIST_FOREACH(device_weights, iw, c->io_device_weights) + fprintf(f, + "%sIODeviceWeight=%s %" PRIu64, + prefix, + iw->path, + iw->weight); + + LIST_FOREACH(device_limits, il, c->io_device_limits) { + char buf[FORMAT_BYTES_MAX]; + CGroupIOLimitType type; + + for (type = 0; type < _CGROUP_IO_LIMIT_TYPE_MAX; type++) + if (il->limits[type] != cgroup_io_limit_defaults[type]) + fprintf(f, + "%s%s=%s %s\n", + prefix, + cgroup_io_limit_type_to_string(type), + il->path, + format_bytes(buf, sizeof(buf), il->limits[type])); + } + + LIST_FOREACH(device_weights, w, c->blockio_device_weights) + fprintf(f, + "%sBlockIODeviceWeight=%s %" PRIu64, + prefix, + w->path, + w->weight); + + LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { + char buf[FORMAT_BYTES_MAX]; + + if (b->rbps != CGROUP_LIMIT_MAX) + fprintf(f, + "%sBlockIOReadBandwidth=%s %s\n", + prefix, + b->path, + format_bytes(buf, sizeof(buf), b->rbps)); + if (b->wbps != CGROUP_LIMIT_MAX) + fprintf(f, + "%sBlockIOWriteBandwidth=%s %s\n", + prefix, + b->path, + format_bytes(buf, sizeof(buf), b->wbps)); + } +} + +static int lookup_block_device(const char *p, dev_t *dev) { + struct stat st; + int r; + + assert(p); + assert(dev); + + r = stat(p, &st); + if (r < 0) + return log_warning_errno(errno, "Couldn't stat device %s: %m", p); + + if (S_ISBLK(st.st_mode)) + *dev = st.st_rdev; + else if (major(st.st_dev) != 0) { + /* If this is not a device node then find the block + * device this file is stored on */ + *dev = st.st_dev; + + /* If this is a partition, try to get the originating + * block device */ + block_get_whole_disk(*dev, dev); + } else { + log_warning("%s is not a block device and file system block device cannot be determined or is not local.", p); + return -ENODEV; + } + + return 0; +} + +static int whitelist_device(const char *path, const char *node, const char *acc) { + char buf[2+DECIMAL_STR_MAX(dev_t)*2+2+4]; + struct stat st; + int r; + + assert(path); + assert(acc); + + if (stat(node, &st) < 0) { + log_warning("Couldn't stat device %s", node); + return -errno; + } + + if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) { + log_warning("%s is not a device.", node); + return -ENODEV; + } + + sprintf(buf, + "%c %u:%u %s", + S_ISCHR(st.st_mode) ? 'c' : 'b', + major(st.st_rdev), minor(st.st_rdev), + acc); + + r = cg_set_attribute("devices", path, "devices.allow", buf); + if (r < 0) + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set devices.allow on %s: %m", path); + + return r; +} + +static int whitelist_major(const char *path, const char *name, char type, const char *acc) { + _cleanup_fclose_ FILE *f = NULL; + char line[LINE_MAX]; + bool good = false; + int r; + + assert(path); + assert(acc); + assert(type == 'b' || type == 'c'); + + f = fopen("/proc/devices", "re"); + if (!f) + return log_warning_errno(errno, "Cannot open /proc/devices to resolve %s (%c): %m", name, type); + + FOREACH_LINE(line, f, goto fail) { + char buf[2+DECIMAL_STR_MAX(unsigned)+3+4], *p, *w; + unsigned maj; + + truncate_nl(line); + + if (type == 'c' && streq(line, "Character devices:")) { + good = true; + continue; + } + + if (type == 'b' && streq(line, "Block devices:")) { + good = true; + continue; + } + + if (isempty(line)) { + good = false; + continue; + } + + if (!good) + continue; + + p = strstrip(line); + + w = strpbrk(p, WHITESPACE); + if (!w) + continue; + *w = 0; + + r = safe_atou(p, &maj); + if (r < 0) + continue; + if (maj <= 0) + continue; + + w++; + w += strspn(w, WHITESPACE); + + if (fnmatch(name, w, 0) != 0) + continue; + + sprintf(buf, + "%c %u:* %s", + type, + maj, + acc); + + r = cg_set_attribute("devices", path, "devices.allow", buf); + if (r < 0) + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set devices.allow on %s: %m", path); + } + + return 0; + +fail: + log_warning_errno(errno, "Failed to read /proc/devices: %m"); + return -errno; +} + +static bool cgroup_context_has_cpu_weight(CGroupContext *c) { + return c->cpu_weight != CGROUP_WEIGHT_INVALID || + c->startup_cpu_weight != CGROUP_WEIGHT_INVALID; +} + +static bool cgroup_context_has_cpu_shares(CGroupContext *c) { + return c->cpu_shares != CGROUP_CPU_SHARES_INVALID || + c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID; +} + +static uint64_t cgroup_context_cpu_weight(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_cpu_weight != CGROUP_WEIGHT_INVALID) + return c->startup_cpu_weight; + else if (c->cpu_weight != CGROUP_WEIGHT_INVALID) + return c->cpu_weight; + else + return CGROUP_WEIGHT_DEFAULT; +} + +static uint64_t cgroup_context_cpu_shares(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_cpu_shares != CGROUP_CPU_SHARES_INVALID) + return c->startup_cpu_shares; + else if (c->cpu_shares != CGROUP_CPU_SHARES_INVALID) + return c->cpu_shares; + else + return CGROUP_CPU_SHARES_DEFAULT; +} + +static void cgroup_apply_unified_cpu_config(Unit *u, uint64_t weight, uint64_t quota) { + char buf[MAX(DECIMAL_STR_MAX(uint64_t) + 1, (DECIMAL_STR_MAX(usec_t) + 1) * 2)]; + int r; + + xsprintf(buf, "%" PRIu64 "\n", weight); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.weight", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.weight: %m"); + + if (quota != USEC_INFINITY) + xsprintf(buf, USEC_FMT " " USEC_FMT "\n", + quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC, CGROUP_CPU_QUOTA_PERIOD_USEC); + else + xsprintf(buf, "max " USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); + + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.max", buf); + + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.max: %m"); +} + +static void cgroup_apply_legacy_cpu_config(Unit *u, uint64_t shares, uint64_t quota) { + char buf[MAX(DECIMAL_STR_MAX(uint64_t), DECIMAL_STR_MAX(usec_t)) + 1]; + int r; + + xsprintf(buf, "%" PRIu64 "\n", shares); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.shares", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.shares: %m"); + + xsprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_period_us", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.cfs_period_us: %m"); + + if (quota != USEC_INFINITY) { + xsprintf(buf, USEC_FMT "\n", quota * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC); + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_quota_us", buf); + } else + r = cg_set_attribute("cpu", u->cgroup_path, "cpu.cfs_quota_us", "-1"); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set cpu.cfs_quota_us: %m"); +} + +static uint64_t cgroup_cpu_shares_to_weight(uint64_t shares) { + return CLAMP(shares * CGROUP_WEIGHT_DEFAULT / CGROUP_CPU_SHARES_DEFAULT, + CGROUP_WEIGHT_MIN, CGROUP_WEIGHT_MAX); +} + +static uint64_t cgroup_cpu_weight_to_shares(uint64_t weight) { + return CLAMP(weight * CGROUP_CPU_SHARES_DEFAULT / CGROUP_WEIGHT_DEFAULT, + CGROUP_CPU_SHARES_MIN, CGROUP_CPU_SHARES_MAX); +} + +static bool cgroup_context_has_io_config(CGroupContext *c) { + return c->io_accounting || + c->io_weight != CGROUP_WEIGHT_INVALID || + c->startup_io_weight != CGROUP_WEIGHT_INVALID || + c->io_device_weights || + c->io_device_limits; +} + +static bool cgroup_context_has_blockio_config(CGroupContext *c) { + return c->blockio_accounting || + c->blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID || + c->startup_blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID || + c->blockio_device_weights || + c->blockio_device_bandwidths; +} + +static uint64_t cgroup_context_io_weight(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_io_weight != CGROUP_WEIGHT_INVALID) + return c->startup_io_weight; + else if (c->io_weight != CGROUP_WEIGHT_INVALID) + return c->io_weight; + else + return CGROUP_WEIGHT_DEFAULT; +} + +static uint64_t cgroup_context_blkio_weight(CGroupContext *c, ManagerState state) { + if (IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && + c->startup_blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID) + return c->startup_blockio_weight; + else if (c->blockio_weight != CGROUP_BLKIO_WEIGHT_INVALID) + return c->blockio_weight; + else + return CGROUP_BLKIO_WEIGHT_DEFAULT; +} + +static uint64_t cgroup_weight_blkio_to_io(uint64_t blkio_weight) { + return CLAMP(blkio_weight * CGROUP_WEIGHT_DEFAULT / CGROUP_BLKIO_WEIGHT_DEFAULT, + CGROUP_WEIGHT_MIN, CGROUP_WEIGHT_MAX); +} + +static uint64_t cgroup_weight_io_to_blkio(uint64_t io_weight) { + return CLAMP(io_weight * CGROUP_BLKIO_WEIGHT_DEFAULT / CGROUP_WEIGHT_DEFAULT, + CGROUP_BLKIO_WEIGHT_MIN, CGROUP_BLKIO_WEIGHT_MAX); +} + +static void cgroup_apply_io_device_weight(Unit *u, const char *dev_path, uint64_t io_weight) { + char buf[DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+1]; + dev_t dev; + int r; + + r = lookup_block_device(dev_path, &dev); + if (r < 0) + return; + + xsprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), io_weight); + r = cg_set_attribute("io", u->cgroup_path, "io.weight", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set io.weight: %m"); +} + +static void cgroup_apply_blkio_device_weight(Unit *u, const char *dev_path, uint64_t blkio_weight) { + char buf[DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+1]; + dev_t dev; + int r; + + r = lookup_block_device(dev_path, &dev); + if (r < 0) + return; + + xsprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), blkio_weight); + r = cg_set_attribute("blkio", u->cgroup_path, "blkio.weight_device", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set blkio.weight_device: %m"); +} + +static unsigned cgroup_apply_io_device_limit(Unit *u, const char *dev_path, uint64_t *limits) { + char limit_bufs[_CGROUP_IO_LIMIT_TYPE_MAX][DECIMAL_STR_MAX(uint64_t)]; + char buf[DECIMAL_STR_MAX(dev_t)*2+2+(6+DECIMAL_STR_MAX(uint64_t)+1)*4]; + CGroupIOLimitType type; + dev_t dev; + unsigned n = 0; + int r; + + r = lookup_block_device(dev_path, &dev); + if (r < 0) + return 0; + + for (type = 0; type < _CGROUP_IO_LIMIT_TYPE_MAX; type++) { + if (limits[type] != cgroup_io_limit_defaults[type]) { + xsprintf(limit_bufs[type], "%" PRIu64, limits[type]); + n++; + } else { + xsprintf(limit_bufs[type], "%s", limits[type] == CGROUP_LIMIT_MAX ? "max" : "0"); + } + } + + xsprintf(buf, "%u:%u rbps=%s wbps=%s riops=%s wiops=%s\n", major(dev), minor(dev), + limit_bufs[CGROUP_IO_RBPS_MAX], limit_bufs[CGROUP_IO_WBPS_MAX], + limit_bufs[CGROUP_IO_RIOPS_MAX], limit_bufs[CGROUP_IO_WIOPS_MAX]); + r = cg_set_attribute("io", u->cgroup_path, "io.max", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set io.max: %m"); + return n; +} + +static unsigned cgroup_apply_blkio_device_limit(Unit *u, const char *dev_path, uint64_t rbps, uint64_t wbps) { + char buf[DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+1]; + dev_t dev; + unsigned n = 0; + int r; + + r = lookup_block_device(dev_path, &dev); + if (r < 0) + return 0; + + if (rbps != CGROUP_LIMIT_MAX) + n++; + sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), rbps); + r = cg_set_attribute("blkio", u->cgroup_path, "blkio.throttle.read_bps_device", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set blkio.throttle.read_bps_device: %m"); + + if (wbps != CGROUP_LIMIT_MAX) + n++; + sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), wbps); + r = cg_set_attribute("blkio", u->cgroup_path, "blkio.throttle.write_bps_device", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set blkio.throttle.write_bps_device: %m"); + + return n; +} + +static bool cgroup_context_has_unified_memory_config(CGroupContext *c) { + return c->memory_low > 0 || c->memory_high != CGROUP_LIMIT_MAX || c->memory_max != CGROUP_LIMIT_MAX || c->memory_swap_max != CGROUP_LIMIT_MAX; +} + +static void cgroup_apply_unified_memory_limit(Unit *u, const char *file, uint64_t v) { + char buf[DECIMAL_STR_MAX(uint64_t) + 1] = "max"; + int r; + + if (v != CGROUP_LIMIT_MAX) + xsprintf(buf, "%" PRIu64 "\n", v); + + r = cg_set_attribute("memory", u->cgroup_path, file, buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set %s: %m", file); +} + +static void cgroup_context_apply(Unit *u, CGroupMask mask, ManagerState state) { + const char *path; + CGroupContext *c; + bool is_root; + int r; + + assert(u); + + c = unit_get_cgroup_context(u); + path = u->cgroup_path; + + assert(c); + assert(path); + + if (mask == 0) + return; + + /* Some cgroup attributes are not supported on the root cgroup, + * hence silently ignore */ + is_root = isempty(path) || path_equal(path, "/"); + if (is_root) + /* Make sure we don't try to display messages with an empty path. */ + path = "/"; + + /* We generally ignore errors caused by read-only mounted + * cgroup trees (assuming we are running in a container then), + * and missing cgroups, i.e. EROFS and ENOENT. */ + + if ((mask & CGROUP_MASK_CPU) && !is_root) { + bool has_weight = cgroup_context_has_cpu_weight(c); + bool has_shares = cgroup_context_has_cpu_shares(c); + + if (cg_all_unified() > 0) { + uint64_t weight; + + if (has_weight) + weight = cgroup_context_cpu_weight(c, state); + else if (has_shares) { + uint64_t shares = cgroup_context_cpu_shares(c, state); + + weight = cgroup_cpu_shares_to_weight(shares); + + log_cgroup_compat(u, "Applying [Startup]CpuShares %" PRIu64 " as [Startup]CpuWeight %" PRIu64 " on %s", + shares, weight, path); + } else + weight = CGROUP_WEIGHT_DEFAULT; + + cgroup_apply_unified_cpu_config(u, weight, c->cpu_quota_per_sec_usec); + } else { + uint64_t shares; + + if (has_weight) { + uint64_t weight = cgroup_context_cpu_weight(c, state); + + shares = cgroup_cpu_weight_to_shares(weight); + + log_cgroup_compat(u, "Applying [Startup]CpuWeight %" PRIu64 " as [Startup]CpuShares %" PRIu64 " on %s", + weight, shares, path); + } else if (has_shares) + shares = cgroup_context_cpu_shares(c, state); + else + shares = CGROUP_CPU_SHARES_DEFAULT; + + cgroup_apply_legacy_cpu_config(u, shares, c->cpu_quota_per_sec_usec); + } + } + + if (mask & CGROUP_MASK_IO) { + bool has_io = cgroup_context_has_io_config(c); + bool has_blockio = cgroup_context_has_blockio_config(c); + + if (!is_root) { + char buf[8+DECIMAL_STR_MAX(uint64_t)+1]; + uint64_t weight; + + if (has_io) + weight = cgroup_context_io_weight(c, state); + else if (has_blockio) { + uint64_t blkio_weight = cgroup_context_blkio_weight(c, state); + + weight = cgroup_weight_blkio_to_io(blkio_weight); + + log_cgroup_compat(u, "Applying [Startup]BlockIOWeight %" PRIu64 " as [Startup]IOWeight %" PRIu64, + blkio_weight, weight); + } else + weight = CGROUP_WEIGHT_DEFAULT; + + xsprintf(buf, "default %" PRIu64 "\n", weight); + r = cg_set_attribute("io", path, "io.weight", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set io.weight: %m"); + + if (has_io) { + CGroupIODeviceWeight *w; + + /* FIXME: no way to reset this list */ + LIST_FOREACH(device_weights, w, c->io_device_weights) + cgroup_apply_io_device_weight(u, w->path, w->weight); + } else if (has_blockio) { + CGroupBlockIODeviceWeight *w; + + /* FIXME: no way to reset this list */ + LIST_FOREACH(device_weights, w, c->blockio_device_weights) { + weight = cgroup_weight_blkio_to_io(w->weight); + + log_cgroup_compat(u, "Applying BlockIODeviceWeight %" PRIu64 " as IODeviceWeight %" PRIu64 " for %s", + w->weight, weight, w->path); + + cgroup_apply_io_device_weight(u, w->path, weight); + } + } + } + + /* Apply limits and free ones without config. */ + if (has_io) { + CGroupIODeviceLimit *l, *next; + + LIST_FOREACH_SAFE(device_limits, l, next, c->io_device_limits) { + if (!cgroup_apply_io_device_limit(u, l->path, l->limits)) + cgroup_context_free_io_device_limit(c, l); + } + } else if (has_blockio) { + CGroupBlockIODeviceBandwidth *b, *next; + + LIST_FOREACH_SAFE(device_bandwidths, b, next, c->blockio_device_bandwidths) { + uint64_t limits[_CGROUP_IO_LIMIT_TYPE_MAX]; + CGroupIOLimitType type; + + for (type = 0; type < _CGROUP_IO_LIMIT_TYPE_MAX; type++) + limits[type] = cgroup_io_limit_defaults[type]; + + limits[CGROUP_IO_RBPS_MAX] = b->rbps; + limits[CGROUP_IO_WBPS_MAX] = b->wbps; + + log_cgroup_compat(u, "Applying BlockIO{Read|Write}Bandwidth %" PRIu64 " %" PRIu64 " as IO{Read|Write}BandwidthMax for %s", + b->rbps, b->wbps, b->path); + + if (!cgroup_apply_io_device_limit(u, b->path, limits)) + cgroup_context_free_blockio_device_bandwidth(c, b); + } + } + } + + if (mask & CGROUP_MASK_BLKIO) { + bool has_io = cgroup_context_has_io_config(c); + bool has_blockio = cgroup_context_has_blockio_config(c); + + if (!is_root) { + char buf[DECIMAL_STR_MAX(uint64_t)+1]; + uint64_t weight; + + if (has_io) { + uint64_t io_weight = cgroup_context_io_weight(c, state); + + weight = cgroup_weight_io_to_blkio(cgroup_context_io_weight(c, state)); + + log_cgroup_compat(u, "Applying [Startup]IOWeight %" PRIu64 " as [Startup]BlockIOWeight %" PRIu64, + io_weight, weight); + } else if (has_blockio) + weight = cgroup_context_blkio_weight(c, state); + else + weight = CGROUP_BLKIO_WEIGHT_DEFAULT; + + xsprintf(buf, "%" PRIu64 "\n", weight); + r = cg_set_attribute("blkio", path, "blkio.weight", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set blkio.weight: %m"); + + if (has_io) { + CGroupIODeviceWeight *w; + + /* FIXME: no way to reset this list */ + LIST_FOREACH(device_weights, w, c->io_device_weights) { + weight = cgroup_weight_io_to_blkio(w->weight); + + log_cgroup_compat(u, "Applying IODeviceWeight %" PRIu64 " as BlockIODeviceWeight %" PRIu64 " for %s", + w->weight, weight, w->path); + + cgroup_apply_blkio_device_weight(u, w->path, weight); + } + } else if (has_blockio) { + CGroupBlockIODeviceWeight *w; + + /* FIXME: no way to reset this list */ + LIST_FOREACH(device_weights, w, c->blockio_device_weights) + cgroup_apply_blkio_device_weight(u, w->path, w->weight); + } + } + + /* Apply limits and free ones without config. */ + if (has_io) { + CGroupIODeviceLimit *l, *next; + + LIST_FOREACH_SAFE(device_limits, l, next, c->io_device_limits) { + log_cgroup_compat(u, "Applying IO{Read|Write}Bandwidth %" PRIu64 " %" PRIu64 " as BlockIO{Read|Write}BandwidthMax for %s", + l->limits[CGROUP_IO_RBPS_MAX], l->limits[CGROUP_IO_WBPS_MAX], l->path); + + if (!cgroup_apply_blkio_device_limit(u, l->path, l->limits[CGROUP_IO_RBPS_MAX], l->limits[CGROUP_IO_WBPS_MAX])) + cgroup_context_free_io_device_limit(c, l); + } + } else if (has_blockio) { + CGroupBlockIODeviceBandwidth *b, *next; + + LIST_FOREACH_SAFE(device_bandwidths, b, next, c->blockio_device_bandwidths) + if (!cgroup_apply_blkio_device_limit(u, b->path, b->rbps, b->wbps)) + cgroup_context_free_blockio_device_bandwidth(c, b); + } + } + + if ((mask & CGROUP_MASK_MEMORY) && !is_root) { + if (cg_all_unified() > 0) { + uint64_t max; + uint64_t swap_max = CGROUP_LIMIT_MAX; + + if (cgroup_context_has_unified_memory_config(c)) { + max = c->memory_max; + swap_max = c->memory_swap_max; + } else { + max = c->memory_limit; + + if (max != CGROUP_LIMIT_MAX) + log_cgroup_compat(u, "Applying MemoryLimit %" PRIu64 " as MemoryMax", max); + } + + cgroup_apply_unified_memory_limit(u, "memory.low", c->memory_low); + cgroup_apply_unified_memory_limit(u, "memory.high", c->memory_high); + cgroup_apply_unified_memory_limit(u, "memory.max", max); + cgroup_apply_unified_memory_limit(u, "memory.swap.max", swap_max); + } else { + char buf[DECIMAL_STR_MAX(uint64_t) + 1]; + uint64_t val; + + if (cgroup_context_has_unified_memory_config(c)) { + val = c->memory_max; + log_cgroup_compat(u, "Applying MemoryMax %" PRIi64 " as MemoryLimit", val); + } else + val = c->memory_limit; + + if (val == CGROUP_LIMIT_MAX) + strncpy(buf, "-1\n", sizeof(buf)); + else + xsprintf(buf, "%" PRIu64 "\n", val); + + r = cg_set_attribute("memory", path, "memory.limit_in_bytes", buf); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set memory.limit_in_bytes: %m"); + } + } + + if ((mask & CGROUP_MASK_DEVICES) && !is_root) { + CGroupDeviceAllow *a; + + /* Changing the devices list of a populated cgroup + * might result in EINVAL, hence ignore EINVAL + * here. */ + + if (c->device_allow || c->device_policy != CGROUP_AUTO) + r = cg_set_attribute("devices", path, "devices.deny", "a"); + else + r = cg_set_attribute("devices", path, "devices.allow", "a"); + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to reset devices.list: %m"); + + if (c->device_policy == CGROUP_CLOSED || + (c->device_policy == CGROUP_AUTO && c->device_allow)) { + static const char auto_devices[] = + "/dev/null\0" "rwm\0" + "/dev/zero\0" "rwm\0" + "/dev/full\0" "rwm\0" + "/dev/random\0" "rwm\0" + "/dev/urandom\0" "rwm\0" + "/dev/tty\0" "rwm\0" + "/dev/pts/ptmx\0" "rw\0" /* /dev/pts/ptmx may not be duplicated, but accessed */ + /* Allow /run/systemd/inaccessible/{chr,blk} devices for mapping InaccessiblePaths */ + "/run/systemd/inaccessible/chr\0" "rwm\0" + "/run/systemd/inaccessible/blk\0" "rwm\0"; + + const char *x, *y; + + NULSTR_FOREACH_PAIR(x, y, auto_devices) + whitelist_device(path, x, y); + + whitelist_major(path, "pts", 'c', "rw"); + whitelist_major(path, "kdbus", 'c', "rw"); + whitelist_major(path, "kdbus/*", 'c', "rw"); + } + + LIST_FOREACH(device_allow, a, c->device_allow) { + char acc[4]; + unsigned k = 0; + + if (a->r) + acc[k++] = 'r'; + if (a->w) + acc[k++] = 'w'; + if (a->m) + acc[k++] = 'm'; + + if (k == 0) + continue; + + acc[k++] = 0; + + if (startswith(a->path, "/dev/")) + whitelist_device(path, a->path, acc); + else if (startswith(a->path, "block-")) + whitelist_major(path, a->path + 6, 'b', acc); + else if (startswith(a->path, "char-")) + whitelist_major(path, a->path + 5, 'c', acc); + else + log_unit_debug(u, "Ignoring device %s while writing cgroup attribute.", a->path); + } + } + + if ((mask & CGROUP_MASK_PIDS) && !is_root) { + + if (c->tasks_max != CGROUP_LIMIT_MAX) { + char buf[DECIMAL_STR_MAX(uint64_t) + 2]; + + sprintf(buf, "%" PRIu64 "\n", c->tasks_max); + r = cg_set_attribute("pids", path, "pids.max", buf); + } else + r = cg_set_attribute("pids", path, "pids.max", "max"); + + if (r < 0) + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set pids.max: %m"); + } +} + +CGroupMask cgroup_context_get_mask(CGroupContext *c) { + CGroupMask mask = 0; + + /* Figure out which controllers we need */ + + if (c->cpu_accounting || + cgroup_context_has_cpu_weight(c) || + cgroup_context_has_cpu_shares(c) || + c->cpu_quota_per_sec_usec != USEC_INFINITY) + mask |= CGROUP_MASK_CPUACCT | CGROUP_MASK_CPU; + + if (cgroup_context_has_io_config(c) || cgroup_context_has_blockio_config(c)) + mask |= CGROUP_MASK_IO | CGROUP_MASK_BLKIO; + + if (c->memory_accounting || + c->memory_limit != CGROUP_LIMIT_MAX || + cgroup_context_has_unified_memory_config(c)) + mask |= CGROUP_MASK_MEMORY; + + if (c->device_allow || + c->device_policy != CGROUP_AUTO) + mask |= CGROUP_MASK_DEVICES; + + if (c->tasks_accounting || + c->tasks_max != (uint64_t) -1) + mask |= CGROUP_MASK_PIDS; + + return mask; +} + +CGroupMask unit_get_own_mask(Unit *u) { + CGroupContext *c; + + /* Returns the mask of controllers the unit needs for itself */ + + c = unit_get_cgroup_context(u); + if (!c) + return 0; + + /* If delegation is turned on, then turn on all cgroups, + * unless we are on the legacy hierarchy and the process we + * fork into it is known to drop privileges, and hence + * shouldn't get access to the controllers. + * + * Note that on the unified hierarchy it is safe to delegate + * controllers to unprivileged services. */ + + if (c->delegate) { + ExecContext *e; + + e = unit_get_exec_context(u); + if (!e || + exec_context_maintains_privileges(e) || + cg_all_unified() > 0) + return _CGROUP_MASK_ALL; + } + + return cgroup_context_get_mask(c); +} + +CGroupMask unit_get_members_mask(Unit *u) { + assert(u); + + /* Returns the mask of controllers all of the unit's children + * require, merged */ + + if (u->cgroup_members_mask_valid) + return u->cgroup_members_mask; + + u->cgroup_members_mask = 0; + + if (u->type == UNIT_SLICE) { + Unit *member; + Iterator i; + + SET_FOREACH(member, u->dependencies[UNIT_BEFORE], i) { + + if (member == u) + continue; + + if (UNIT_DEREF(member->slice) != u) + continue; + + u->cgroup_members_mask |= + unit_get_own_mask(member) | + unit_get_members_mask(member); + } + } + + u->cgroup_members_mask_valid = true; + return u->cgroup_members_mask; +} + +CGroupMask unit_get_siblings_mask(Unit *u) { + assert(u); + + /* Returns the mask of controllers all of the unit's siblings + * require, i.e. the members mask of the unit's parent slice + * if there is one. */ + + if (UNIT_ISSET(u->slice)) + return unit_get_members_mask(UNIT_DEREF(u->slice)); + + return unit_get_own_mask(u) | unit_get_members_mask(u); +} + +CGroupMask unit_get_subtree_mask(Unit *u) { + + /* Returns the mask of this subtree, meaning of the group + * itself and its children. */ + + return unit_get_own_mask(u) | unit_get_members_mask(u); +} + +CGroupMask unit_get_target_mask(Unit *u) { + CGroupMask mask; + + /* This returns the cgroup mask of all controllers to enable + * for a specific cgroup, i.e. everything it needs itself, + * plus all that its children need, plus all that its siblings + * need. This is primarily useful on the legacy cgroup + * hierarchy, where we need to duplicate each cgroup in each + * hierarchy that shall be enabled for it. */ + + mask = unit_get_own_mask(u) | unit_get_members_mask(u) | unit_get_siblings_mask(u); + mask &= u->manager->cgroup_supported; + + return mask; +} + +CGroupMask unit_get_enable_mask(Unit *u) { + CGroupMask mask; + + /* This returns the cgroup mask of all controllers to enable + * for the children of a specific cgroup. This is primarily + * useful for the unified cgroup hierarchy, where each cgroup + * controls which controllers are enabled for its children. */ + + mask = unit_get_members_mask(u); + mask &= u->manager->cgroup_supported; + + return mask; +} + +/* Recurse from a unit up through its containing slices, propagating + * mask bits upward. A unit is also member of itself. */ +void unit_update_cgroup_members_masks(Unit *u) { + CGroupMask m; + bool more; + + assert(u); + + /* Calculate subtree mask */ + m = unit_get_subtree_mask(u); + + /* See if anything changed from the previous invocation. If + * not, we're done. */ + if (u->cgroup_subtree_mask_valid && m == u->cgroup_subtree_mask) + return; + + more = + u->cgroup_subtree_mask_valid && + ((m & ~u->cgroup_subtree_mask) != 0) && + ((~m & u->cgroup_subtree_mask) == 0); + + u->cgroup_subtree_mask = m; + u->cgroup_subtree_mask_valid = true; + + if (UNIT_ISSET(u->slice)) { + Unit *s = UNIT_DEREF(u->slice); + + if (more) + /* There's more set now than before. We + * propagate the new mask to the parent's mask + * (not caring if it actually was valid or + * not). */ + + s->cgroup_members_mask |= m; + + else + /* There's less set now than before (or we + * don't know), we need to recalculate + * everything, so let's invalidate the + * parent's members mask */ + + s->cgroup_members_mask_valid = false; + + /* And now make sure that this change also hits our + * grandparents */ + unit_update_cgroup_members_masks(s); + } +} + +static const char *migrate_callback(CGroupMask mask, void *userdata) { + Unit *u = userdata; + + assert(mask != 0); + assert(u); + + while (u) { + if (u->cgroup_path && + u->cgroup_realized && + (u->cgroup_realized_mask & mask) == mask) + return u->cgroup_path; + + u = UNIT_DEREF(u->slice); + } + + return NULL; +} + +char *unit_default_cgroup_path(Unit *u) { + _cleanup_free_ char *escaped = NULL, *slice = NULL; + int r; + + assert(u); + + if (unit_has_name(u, SPECIAL_ROOT_SLICE)) + return strdup(u->manager->cgroup_root); + + if (UNIT_ISSET(u->slice) && !unit_has_name(UNIT_DEREF(u->slice), SPECIAL_ROOT_SLICE)) { + r = cg_slice_to_path(UNIT_DEREF(u->slice)->id, &slice); + if (r < 0) + return NULL; + } + + escaped = cg_escape(u->id); + if (!escaped) + return NULL; + + if (slice) + return strjoin(u->manager->cgroup_root, "/", slice, "/", escaped, NULL); + else + return strjoin(u->manager->cgroup_root, "/", escaped, NULL); +} + +int unit_set_cgroup_path(Unit *u, const char *path) { + _cleanup_free_ char *p = NULL; + int r; + + assert(u); + + if (path) { + p = strdup(path); + if (!p) + return -ENOMEM; + } else + p = NULL; + + if (streq_ptr(u->cgroup_path, p)) + return 0; + + if (p) { + r = hashmap_put(u->manager->cgroup_unit, p, u); + if (r < 0) + return r; + } + + unit_release_cgroup(u); + + u->cgroup_path = p; + p = NULL; + + return 1; +} + +int unit_watch_cgroup(Unit *u) { + _cleanup_free_ char *events = NULL; + int r; + + assert(u); + + if (!u->cgroup_path) + return 0; + + if (u->cgroup_inotify_wd >= 0) + return 0; + + /* Only applies to the unified hierarchy */ + r = cg_unified(SYSTEMD_CGROUP_CONTROLLER); + if (r < 0) + return log_unit_error_errno(u, r, "Failed detect whether the unified hierarchy is used: %m"); + if (r == 0) + return 0; + + /* Don't watch the root slice, it's pointless. */ + if (unit_has_name(u, SPECIAL_ROOT_SLICE)) + return 0; + + r = hashmap_ensure_allocated(&u->manager->cgroup_inotify_wd_unit, &trivial_hash_ops); + if (r < 0) + return log_oom(); + + r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, "cgroup.events", &events); + if (r < 0) + return log_oom(); + + u->cgroup_inotify_wd = inotify_add_watch(u->manager->cgroup_inotify_fd, events, IN_MODIFY); + if (u->cgroup_inotify_wd < 0) { + + if (errno == ENOENT) /* If the directory is already + * gone we don't need to track + * it, so this is not an error */ + return 0; + + return log_unit_error_errno(u, errno, "Failed to add inotify watch descriptor for control group %s: %m", u->cgroup_path); + } + + r = hashmap_put(u->manager->cgroup_inotify_wd_unit, INT_TO_PTR(u->cgroup_inotify_wd), u); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to add inotify watch descriptor to hash map: %m"); + + return 0; +} + +static int unit_create_cgroup( + Unit *u, + CGroupMask target_mask, + CGroupMask enable_mask) { + + CGroupContext *c; + int r; + + assert(u); + + c = unit_get_cgroup_context(u); + if (!c) + return 0; + + if (!u->cgroup_path) { + _cleanup_free_ char *path = NULL; + + path = unit_default_cgroup_path(u); + if (!path) + return log_oom(); + + r = unit_set_cgroup_path(u, path); + if (r == -EEXIST) + return log_unit_error_errno(u, r, "Control group %s exists already.", path); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to set unit's control group path to %s: %m", path); + } + + /* First, create our own group */ + r = cg_create_everywhere(u->manager->cgroup_supported, target_mask, u->cgroup_path); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to create cgroup %s: %m", u->cgroup_path); + + /* Start watching it */ + (void) unit_watch_cgroup(u); + + /* Enable all controllers we need */ + r = cg_enable_everywhere(u->manager->cgroup_supported, enable_mask, u->cgroup_path); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to enable controllers on cgroup %s, ignoring: %m", u->cgroup_path); + + /* Keep track that this is now realized */ + u->cgroup_realized = true; + u->cgroup_realized_mask = target_mask; + u->cgroup_enabled_mask = enable_mask; + + if (u->type != UNIT_SLICE && !c->delegate) { + + /* Then, possibly move things over, but not if + * subgroups may contain processes, which is the case + * for slice and delegation units. */ + r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to migrate cgroup from to %s, ignoring: %m", u->cgroup_path); + } + + return 0; +} + +int unit_attach_pids_to_cgroup(Unit *u) { + int r; + assert(u); + + r = unit_realize_cgroup(u); + if (r < 0) + return r; + + r = cg_attach_many_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->pids, migrate_callback, u); + if (r < 0) + return r; + + return 0; +} + +static void cgroup_xattr_apply(Unit *u) { + char ids[SD_ID128_STRING_MAX]; + int r; + + assert(u); + + if (!MANAGER_IS_SYSTEM(u->manager)) + return; + + if (sd_id128_is_null(u->invocation_id)) + return; + + r = cg_set_xattr(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, + "trusted.invocation_id", + sd_id128_to_string(u->invocation_id, ids), 32, + 0); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to set invocation ID on control group %s, ignoring: %m", u->cgroup_path); +} + +static bool unit_has_mask_realized(Unit *u, CGroupMask target_mask, CGroupMask enable_mask) { + assert(u); + + return u->cgroup_realized && u->cgroup_realized_mask == target_mask && u->cgroup_enabled_mask == enable_mask; +} + +/* Check if necessary controllers and attributes for a unit are in place. + * + * If so, do nothing. + * If not, create paths, move processes over, and set attributes. + * + * Returns 0 on success and < 0 on failure. */ +static int unit_realize_cgroup_now(Unit *u, ManagerState state) { + CGroupMask target_mask, enable_mask; + int r; + + assert(u); + + if (u->in_cgroup_queue) { + LIST_REMOVE(cgroup_queue, u->manager->cgroup_queue, u); + u->in_cgroup_queue = false; + } + + target_mask = unit_get_target_mask(u); + enable_mask = unit_get_enable_mask(u); + + if (unit_has_mask_realized(u, target_mask, enable_mask)) + return 0; + + /* First, realize parents */ + if (UNIT_ISSET(u->slice)) { + r = unit_realize_cgroup_now(UNIT_DEREF(u->slice), state); + if (r < 0) + return r; + } + + /* And then do the real work */ + r = unit_create_cgroup(u, target_mask, enable_mask); + if (r < 0) + return r; + + /* Finally, apply the necessary attributes. */ + cgroup_context_apply(u, target_mask, state); + cgroup_xattr_apply(u); + + return 0; +} + +static void unit_add_to_cgroup_queue(Unit *u) { + + if (u->in_cgroup_queue) + return; + + LIST_PREPEND(cgroup_queue, u->manager->cgroup_queue, u); + u->in_cgroup_queue = true; +} + +unsigned manager_dispatch_cgroup_queue(Manager *m) { + ManagerState state; + unsigned n = 0; + Unit *i; + int r; + + state = manager_state(m); + + while ((i = m->cgroup_queue)) { + assert(i->in_cgroup_queue); + + r = unit_realize_cgroup_now(i, state); + if (r < 0) + log_warning_errno(r, "Failed to realize cgroups for queued unit %s, ignoring: %m", i->id); + + n++; + } + + return n; +} + +static void unit_queue_siblings(Unit *u) { + Unit *slice; + + /* This adds the siblings of the specified unit and the + * siblings of all parent units to the cgroup queue. (But + * neither the specified unit itself nor the parents.) */ + + while ((slice = UNIT_DEREF(u->slice))) { + Iterator i; + Unit *m; + + SET_FOREACH(m, slice->dependencies[UNIT_BEFORE], i) { + if (m == u) + continue; + + /* Skip units that have a dependency on the slice + * but aren't actually in it. */ + if (UNIT_DEREF(m->slice) != slice) + continue; + + /* No point in doing cgroup application for units + * without active processes. */ + if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(m))) + continue; + + /* If the unit doesn't need any new controllers + * and has current ones realized, it doesn't need + * any changes. */ + if (unit_has_mask_realized(m, unit_get_target_mask(m), unit_get_enable_mask(m))) + continue; + + unit_add_to_cgroup_queue(m); + } + + u = slice; + } +} + +int unit_realize_cgroup(Unit *u) { + assert(u); + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return 0; + + /* So, here's the deal: when realizing the cgroups for this + * unit, we need to first create all parents, but there's more + * actually: for the weight-based controllers we also need to + * make sure that all our siblings (i.e. units that are in the + * same slice as we are) have cgroups, too. Otherwise, things + * would become very uneven as each of their processes would + * get as much resources as all our group together. This call + * will synchronously create the parent cgroups, but will + * defer work on the siblings to the next event loop + * iteration. */ + + /* Add all sibling slices to the cgroup queue. */ + unit_queue_siblings(u); + + /* And realize this one now (and apply the values) */ + return unit_realize_cgroup_now(u, manager_state(u->manager)); +} + +void unit_release_cgroup(Unit *u) { + assert(u); + + /* Forgets all cgroup details for this cgroup */ + + if (u->cgroup_path) { + (void) hashmap_remove(u->manager->cgroup_unit, u->cgroup_path); + u->cgroup_path = mfree(u->cgroup_path); + } + + if (u->cgroup_inotify_wd >= 0) { + if (inotify_rm_watch(u->manager->cgroup_inotify_fd, u->cgroup_inotify_wd) < 0) + log_unit_debug_errno(u, errno, "Failed to remove cgroup inotify watch %i for %s, ignoring", u->cgroup_inotify_wd, u->id); + + (void) hashmap_remove(u->manager->cgroup_inotify_wd_unit, INT_TO_PTR(u->cgroup_inotify_wd)); + u->cgroup_inotify_wd = -1; + } +} + +void unit_prune_cgroup(Unit *u) { + int r; + bool is_root_slice; + + assert(u); + + /* Removes the cgroup, if empty and possible, and stops watching it. */ + + if (!u->cgroup_path) + return; + + (void) unit_get_cpu_usage(u, NULL); /* Cache the last CPU usage value before we destroy the cgroup */ + + is_root_slice = unit_has_name(u, SPECIAL_ROOT_SLICE); + + r = cg_trim_everywhere(u->manager->cgroup_supported, u->cgroup_path, !is_root_slice); + if (r < 0) { + log_unit_debug_errno(u, r, "Failed to destroy cgroup %s, ignoring: %m", u->cgroup_path); + return; + } + + if (is_root_slice) + return; + + unit_release_cgroup(u); + + u->cgroup_realized = false; + u->cgroup_realized_mask = 0; + u->cgroup_enabled_mask = 0; +} + +int unit_search_main_pid(Unit *u, pid_t *ret) { + _cleanup_fclose_ FILE *f = NULL; + pid_t pid = 0, npid, mypid; + int r; + + assert(u); + assert(ret); + + if (!u->cgroup_path) + return -ENXIO; + + r = cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, &f); + if (r < 0) + return r; + + mypid = getpid(); + while (cg_read_pid(f, &npid) > 0) { + pid_t ppid; + + if (npid == pid) + continue; + + /* Ignore processes that aren't our kids */ + if (get_process_ppid(npid, &ppid) >= 0 && ppid != mypid) + continue; + + if (pid != 0) + /* Dang, there's more than one daemonized PID + in this group, so we don't know what process + is the main process. */ + + return -ENODATA; + + pid = npid; + } + + *ret = pid; + return 0; +} + +static int unit_watch_pids_in_path(Unit *u, const char *path) { + _cleanup_closedir_ DIR *d = NULL; + _cleanup_fclose_ FILE *f = NULL; + int ret = 0, r; + + assert(u); + assert(path); + + r = cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, path, &f); + if (r < 0) + ret = r; + else { + pid_t pid; + + while ((r = cg_read_pid(f, &pid)) > 0) { + r = unit_watch_pid(u, pid); + if (r < 0 && ret >= 0) + ret = r; + } + + if (r < 0 && ret >= 0) + ret = r; + } + + r = cg_enumerate_subgroups(SYSTEMD_CGROUP_CONTROLLER, path, &d); + if (r < 0) { + if (ret >= 0) + ret = r; + } else { + char *fn; + + while ((r = cg_read_subgroup(d, &fn)) > 0) { + _cleanup_free_ char *p = NULL; + + p = strjoin(path, "/", fn, NULL); + free(fn); + + if (!p) + return -ENOMEM; + + r = unit_watch_pids_in_path(u, p); + if (r < 0 && ret >= 0) + ret = r; + } + + if (r < 0 && ret >= 0) + ret = r; + } + + return ret; +} + +int unit_watch_all_pids(Unit *u) { + assert(u); + + /* Adds all PIDs from our cgroup to the set of PIDs we + * watch. This is a fallback logic for cases where we do not + * get reliable cgroup empty notifications: we try to use + * SIGCHLD as replacement. */ + + if (!u->cgroup_path) + return -ENOENT; + + if (cg_unified(SYSTEMD_CGROUP_CONTROLLER) > 0) /* On unified we can use proper notifications */ + return 0; + + return unit_watch_pids_in_path(u, u->cgroup_path); +} + +int unit_notify_cgroup_empty(Unit *u) { + int r; + + assert(u); + + if (!u->cgroup_path) + return 0; + + r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path); + if (r <= 0) + return r; + + unit_add_to_gc_queue(u); + + if (UNIT_VTABLE(u)->notify_cgroup_empty) + UNIT_VTABLE(u)->notify_cgroup_empty(u); + + return 0; +} + +static int on_cgroup_inotify_event(sd_event_source *s, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + + assert(s); + assert(fd >= 0); + assert(m); + + for (;;) { + union inotify_event_buffer buffer; + struct inotify_event *e; + ssize_t l; + + l = read(fd, &buffer, sizeof(buffer)); + if (l < 0) { + if (errno == EINTR || errno == EAGAIN) + return 0; + + return log_error_errno(errno, "Failed to read control group inotify events: %m"); + } + + FOREACH_INOTIFY_EVENT(e, buffer, l) { + Unit *u; + + if (e->wd < 0) + /* Queue overflow has no watch descriptor */ + continue; + + if (e->mask & IN_IGNORED) + /* The watch was just removed */ + continue; + + u = hashmap_get(m->cgroup_inotify_wd_unit, INT_TO_PTR(e->wd)); + if (!u) /* Not that inotify might deliver + * events for a watch even after it + * was removed, because it was queued + * before the removal. Let's ignore + * this here safely. */ + continue; + + (void) unit_notify_cgroup_empty(u); + } + } +} + +int manager_setup_cgroup(Manager *m) { + _cleanup_free_ char *path = NULL; + CGroupController c; + int r, all_unified, systemd_unified; + char *e; + + assert(m); + + /* 1. Determine hierarchy */ + m->cgroup_root = mfree(m->cgroup_root); + r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &m->cgroup_root); + if (r < 0) + return log_error_errno(r, "Cannot determine cgroup we are running in: %m"); + + /* Chop off the init scope, if we are already located in it */ + e = endswith(m->cgroup_root, "/" SPECIAL_INIT_SCOPE); + + /* LEGACY: Also chop off the system slice if we are in + * it. This is to support live upgrades from older systemd + * versions where PID 1 was moved there. Also see + * cg_get_root_path(). */ + if (!e && MANAGER_IS_SYSTEM(m)) { + e = endswith(m->cgroup_root, "/" SPECIAL_SYSTEM_SLICE); + if (!e) + e = endswith(m->cgroup_root, "/system"); /* even more legacy */ + } + if (e) + *e = 0; + + /* And make sure to store away the root value without trailing + * slash, even for the root dir, so that we can easily prepend + * it everywhere. */ + while ((e = endswith(m->cgroup_root, "/"))) + *e = 0; + + /* 2. Show data */ + r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, NULL, &path); + if (r < 0) + return log_error_errno(r, "Cannot find cgroup mount point: %m"); + + all_unified = cg_all_unified(); + systemd_unified = cg_unified(SYSTEMD_CGROUP_CONTROLLER); + + if (all_unified < 0 || systemd_unified < 0) + return log_error_errno(all_unified < 0 ? all_unified : systemd_unified, + "Couldn't determine if we are running in the unified hierarchy: %m"); + + if (all_unified > 0) + log_debug("Unified cgroup hierarchy is located at %s.", path); + else if (systemd_unified > 0) + log_debug("Unified cgroup hierarchy is located at %s. Controllers are on legacy hierarchies.", path); + else + log_debug("Using cgroup controller " SYSTEMD_CGROUP_CONTROLLER ". File system hierarchy is at %s.", path); + + if (!m->test_run) { + const char *scope_path; + + /* 3. Install agent */ + if (systemd_unified) { + + /* In the unified hierarchy we can get + * cgroup empty notifications via inotify. */ + + m->cgroup_inotify_event_source = sd_event_source_unref(m->cgroup_inotify_event_source); + safe_close(m->cgroup_inotify_fd); + + m->cgroup_inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (m->cgroup_inotify_fd < 0) + return log_error_errno(errno, "Failed to create control group inotify object: %m"); + + r = sd_event_add_io(m->event, &m->cgroup_inotify_event_source, m->cgroup_inotify_fd, EPOLLIN, on_cgroup_inotify_event, m); + if (r < 0) + return log_error_errno(r, "Failed to watch control group inotify object: %m"); + + /* Process cgroup empty notifications early, but after service notifications and SIGCHLD. Also + * see handling of cgroup agent notifications, for the classic cgroup hierarchy support. */ + r = sd_event_source_set_priority(m->cgroup_inotify_event_source, SD_EVENT_PRIORITY_NORMAL-5); + if (r < 0) + return log_error_errno(r, "Failed to set priority of inotify event source: %m"); + + (void) sd_event_source_set_description(m->cgroup_inotify_event_source, "cgroup-inotify"); + + } else if (MANAGER_IS_SYSTEM(m)) { + + /* On the legacy hierarchy we only get + * notifications via cgroup agents. (Which + * isn't really reliable, since it does not + * generate events when control groups with + * children run empty. */ + + r = cg_install_release_agent(SYSTEMD_CGROUP_CONTROLLER, SYSTEMD_CGROUP_AGENT_PATH); + if (r < 0) + log_warning_errno(r, "Failed to install release agent, ignoring: %m"); + else if (r > 0) + log_debug("Installed release agent."); + else if (r == 0) + log_debug("Release agent already installed."); + } + + /* 4. Make sure we are in the special "init.scope" unit in the root slice. */ + scope_path = strjoina(m->cgroup_root, "/" SPECIAL_INIT_SCOPE); + r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, scope_path, 0); + if (r < 0) + return log_error_errno(r, "Failed to create %s control group: %m", scope_path); + + /* also, move all other userspace processes remaining + * in the root cgroup into that scope. */ + r = cg_migrate(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, SYSTEMD_CGROUP_CONTROLLER, scope_path, 0); + if (r < 0) + log_warning_errno(r, "Couldn't move remaining userspace processes, ignoring: %m"); + + /* 5. And pin it, so that it cannot be unmounted */ + safe_close(m->pin_cgroupfs_fd); + m->pin_cgroupfs_fd = open(path, O_RDONLY|O_CLOEXEC|O_DIRECTORY|O_NOCTTY|O_NONBLOCK); + if (m->pin_cgroupfs_fd < 0) + return log_error_errno(errno, "Failed to open pin file: %m"); + + /* 6. Always enable hierarchical support if it exists... */ + if (!all_unified) + (void) cg_set_attribute("memory", "/", "memory.use_hierarchy", "1"); + } + + /* 7. Figure out which controllers are supported */ + r = cg_mask_supported(&m->cgroup_supported); + if (r < 0) + return log_error_errno(r, "Failed to determine supported controllers: %m"); + + for (c = 0; c < _CGROUP_CONTROLLER_MAX; c++) + log_debug("Controller '%s' supported: %s", cgroup_controller_to_string(c), yes_no(m->cgroup_supported & CGROUP_CONTROLLER_TO_MASK(c))); + + return 0; +} + +void manager_shutdown_cgroup(Manager *m, bool delete) { + assert(m); + + /* We can't really delete the group, since we are in it. But + * let's trim it. */ + if (delete && m->cgroup_root) + (void) cg_trim(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, false); + + m->cgroup_inotify_wd_unit = hashmap_free(m->cgroup_inotify_wd_unit); + + m->cgroup_inotify_event_source = sd_event_source_unref(m->cgroup_inotify_event_source); + m->cgroup_inotify_fd = safe_close(m->cgroup_inotify_fd); + + m->pin_cgroupfs_fd = safe_close(m->pin_cgroupfs_fd); + + m->cgroup_root = mfree(m->cgroup_root); +} + +Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) { + char *p; + Unit *u; + + assert(m); + assert(cgroup); + + u = hashmap_get(m->cgroup_unit, cgroup); + if (u) + return u; + + p = strdupa(cgroup); + for (;;) { + char *e; + + e = strrchr(p, '/'); + if (!e || e == p) + return hashmap_get(m->cgroup_unit, SPECIAL_ROOT_SLICE); + + *e = 0; + + u = hashmap_get(m->cgroup_unit, p); + if (u) + return u; + } +} + +Unit *manager_get_unit_by_pid_cgroup(Manager *m, pid_t pid) { + _cleanup_free_ char *cgroup = NULL; + int r; + + assert(m); + + if (pid <= 0) + return NULL; + + r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &cgroup); + if (r < 0) + return NULL; + + return manager_get_unit_by_cgroup(m, cgroup); +} + +Unit *manager_get_unit_by_pid(Manager *m, pid_t pid) { + Unit *u; + + assert(m); + + if (pid <= 0) + return NULL; + + if (pid == 1) + return hashmap_get(m->units, SPECIAL_INIT_SCOPE); + + u = hashmap_get(m->watch_pids1, PID_TO_PTR(pid)); + if (u) + return u; + + u = hashmap_get(m->watch_pids2, PID_TO_PTR(pid)); + if (u) + return u; + + return manager_get_unit_by_pid_cgroup(m, pid); +} + +int manager_notify_cgroup_empty(Manager *m, const char *cgroup) { + Unit *u; + + assert(m); + assert(cgroup); + + log_debug("Got cgroup empty notification for: %s", cgroup); + + u = manager_get_unit_by_cgroup(m, cgroup); + if (!u) + return 0; + + return unit_notify_cgroup_empty(u); +} + +int unit_get_memory_current(Unit *u, uint64_t *ret) { + _cleanup_free_ char *v = NULL; + int r; + + assert(u); + assert(ret); + + if (!u->cgroup_path) + return -ENODATA; + + if ((u->cgroup_realized_mask & CGROUP_MASK_MEMORY) == 0) + return -ENODATA; + + if (cg_all_unified() <= 0) + r = cg_get_attribute("memory", u->cgroup_path, "memory.usage_in_bytes", &v); + else + r = cg_get_attribute("memory", u->cgroup_path, "memory.current", &v); + if (r == -ENOENT) + return -ENODATA; + if (r < 0) + return r; + + return safe_atou64(v, ret); +} + +int unit_get_tasks_current(Unit *u, uint64_t *ret) { + _cleanup_free_ char *v = NULL; + int r; + + assert(u); + assert(ret); + + if (!u->cgroup_path) + return -ENODATA; + + if ((u->cgroup_realized_mask & CGROUP_MASK_PIDS) == 0) + return -ENODATA; + + r = cg_get_attribute("pids", u->cgroup_path, "pids.current", &v); + if (r == -ENOENT) + return -ENODATA; + if (r < 0) + return r; + + return safe_atou64(v, ret); +} + +static int unit_get_cpu_usage_raw(Unit *u, nsec_t *ret) { + _cleanup_free_ char *v = NULL; + uint64_t ns; + int r; + + assert(u); + assert(ret); + + if (!u->cgroup_path) + return -ENODATA; + + if (cg_all_unified() > 0) { + const char *keys[] = { "usage_usec", NULL }; + _cleanup_free_ char *val = NULL; + uint64_t us; + + if ((u->cgroup_realized_mask & CGROUP_MASK_CPU) == 0) + return -ENODATA; + + r = cg_get_keyed_attribute("cpu", u->cgroup_path, "cpu.stat", keys, &val); + if (r < 0) + return r; + + r = safe_atou64(val, &us); + if (r < 0) + return r; + + ns = us * NSEC_PER_USEC; + } else { + if ((u->cgroup_realized_mask & CGROUP_MASK_CPUACCT) == 0) + return -ENODATA; + + r = cg_get_attribute("cpuacct", u->cgroup_path, "cpuacct.usage", &v); + if (r == -ENOENT) + return -ENODATA; + if (r < 0) + return r; + + r = safe_atou64(v, &ns); + if (r < 0) + return r; + } + + *ret = ns; + return 0; +} + +int unit_get_cpu_usage(Unit *u, nsec_t *ret) { + nsec_t ns; + int r; + + assert(u); + + /* Retrieve the current CPU usage counter. This will subtract the CPU counter taken when the unit was + * started. If the cgroup has been removed already, returns the last cached value. To cache the value, simply + * call this function with a NULL return value. */ + + r = unit_get_cpu_usage_raw(u, &ns); + if (r == -ENODATA && u->cpu_usage_last != NSEC_INFINITY) { + /* If we can't get the CPU usage anymore (because the cgroup was already removed, for example), use our + * cached value. */ + + if (ret) + *ret = u->cpu_usage_last; + return 0; + } + if (r < 0) + return r; + + if (ns > u->cpu_usage_base) + ns -= u->cpu_usage_base; + else + ns = 0; + + u->cpu_usage_last = ns; + if (ret) + *ret = ns; + + return 0; +} + +int unit_reset_cpu_usage(Unit *u) { + nsec_t ns; + int r; + + assert(u); + + u->cpu_usage_last = NSEC_INFINITY; + + r = unit_get_cpu_usage_raw(u, &ns); + if (r < 0) { + u->cpu_usage_base = 0; + return r; + } + + u->cpu_usage_base = ns; + return 0; +} + +bool unit_cgroup_delegate(Unit *u) { + CGroupContext *c; + + assert(u); + + c = unit_get_cgroup_context(u); + if (!c) + return false; + + return c->delegate; +} + +void unit_invalidate_cgroup(Unit *u, CGroupMask m) { + assert(u); + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return; + + if (m == 0) + return; + + /* always invalidate compat pairs together */ + if (m & (CGROUP_MASK_IO | CGROUP_MASK_BLKIO)) + m |= CGROUP_MASK_IO | CGROUP_MASK_BLKIO; + + if ((u->cgroup_realized_mask & m) == 0) + return; + + u->cgroup_realized_mask &= ~m; + unit_add_to_cgroup_queue(u); +} + +void manager_invalidate_startup_units(Manager *m) { + Iterator i; + Unit *u; + + assert(m); + + SET_FOREACH(u, m->startup_units, i) + unit_invalidate_cgroup(u, CGROUP_MASK_CPU|CGROUP_MASK_IO|CGROUP_MASK_BLKIO); +} + +static const char* const cgroup_device_policy_table[_CGROUP_DEVICE_POLICY_MAX] = { + [CGROUP_AUTO] = "auto", + [CGROUP_CLOSED] = "closed", + [CGROUP_STRICT] = "strict", +}; + +DEFINE_STRING_TABLE_LOOKUP(cgroup_device_policy, CGroupDevicePolicy); diff --git a/src/grp-system/libcore/src/dbus-automount.c b/src/grp-system/libcore/src/dbus-automount.c new file mode 100644 index 0000000000..05e248758f --- /dev/null +++ b/src/grp-system/libcore/src/dbus-automount.c @@ -0,0 +1,89 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/automount.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-automount.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, automount_result, AutomountResult); + +const sd_bus_vtable bus_automount_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Where", "s", NULL, offsetof(Automount, where), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DirectoryMode", "u", bus_property_get_mode, offsetof(Automount, directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Automount, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("TimeoutIdleUSec", "t", bus_property_get_usec, offsetof(Automount, timeout_idle_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_VTABLE_END +}; + +static int bus_automount_set_transient_property( + Automount *a, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(a); + assert(name); + assert(message); + + if (streq(name, "TimeoutIdleUSec")) { + usec_t timeout_idle_usec; + r = sd_bus_message_read(message, "t", &timeout_idle_usec); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + char time[FORMAT_TIMESPAN_MAX]; + + a->timeout_idle_usec = timeout_idle_usec; + unit_write_drop_in_format(UNIT(a), mode, name, "[Automount]\nTimeoutIdleSec=%s\n", + format_timespan(time, sizeof(time), timeout_idle_usec, USEC_PER_MSEC)); + } + } else + return 0; + + return 1; +} + +int bus_automount_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Automount *a = AUTOMOUNT(u); + int r = 0; + + assert(a); + assert(name); + assert(message); + + if (u->transient && u->load_state == UNIT_STUB) + /* This is a transient unit, let's load a little more */ + + r = bus_automount_set_transient_property(a, name, message, mode, error); + + return r; +} diff --git a/src/grp-system/libcore/src/dbus-automount.h b/src/grp-system/libcore/src/dbus-automount.h new file mode 100644 index 0000000000..f41adda2a6 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-automount.h @@ -0,0 +1,25 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + +extern const sd_bus_vtable bus_automount_vtable[]; + +int bus_automount_set_property(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); diff --git a/src/grp-system/libcore/src/dbus-busname.c b/src/grp-system/libcore/src/dbus-busname.c new file mode 100644 index 0000000000..d1324e66bd --- /dev/null +++ b/src/grp-system/libcore/src/dbus-busname.c @@ -0,0 +1,38 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/busname.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-busname.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, busname_result, BusNameResult); + +const sd_bus_vtable bus_busname_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Name", "s", NULL, offsetof(BusName, name), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimeoutUSec", "t", bus_property_get_usec, offsetof(BusName, timeout_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(BusName, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(BusName, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Activating", "b", bus_property_get_bool, offsetof(BusName, activating), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("AcceptFileDescriptors", "b", bus_property_get_bool, offsetof(BusName, accept_fd), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_VTABLE_END +}; diff --git a/src/grp-system/libcore/src/dbus-busname.h b/src/grp-system/libcore/src/dbus-busname.h new file mode 100644 index 0000000000..8643d1a404 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-busname.h @@ -0,0 +1,23 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + +extern const sd_bus_vtable bus_busname_vtable[]; diff --git a/src/grp-system/libcore/src/dbus-cgroup.c b/src/grp-system/libcore/src/dbus-cgroup.c new file mode 100644 index 0000000000..875dc35e87 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-cgroup.c @@ -0,0 +1,1159 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/cgroup.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/path-util.h" + +#include "dbus-cgroup.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_cgroup_device_policy, cgroup_device_policy, CGroupDevicePolicy); + +static int property_get_io_device_weight( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + CGroupContext *c = userdata; + CGroupIODeviceWeight *w; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(st)"); + if (r < 0) + return r; + + LIST_FOREACH(device_weights, w, c->io_device_weights) { + r = sd_bus_message_append(reply, "(st)", w->path, w->weight); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_io_device_limits( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + CGroupContext *c = userdata; + CGroupIODeviceLimit *l; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(st)"); + if (r < 0) + return r; + + LIST_FOREACH(device_limits, l, c->io_device_limits) { + CGroupIOLimitType type; + + type = cgroup_io_limit_type_from_string(property); + if (type < 0 || l->limits[type] == cgroup_io_limit_defaults[type]) + continue; + + r = sd_bus_message_append(reply, "(st)", l->path, l->limits[type]); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_blockio_device_weight( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + CGroupContext *c = userdata; + CGroupBlockIODeviceWeight *w; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(st)"); + if (r < 0) + return r; + + LIST_FOREACH(device_weights, w, c->blockio_device_weights) { + r = sd_bus_message_append(reply, "(st)", w->path, w->weight); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_blockio_device_bandwidths( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + CGroupContext *c = userdata; + CGroupBlockIODeviceBandwidth *b; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(st)"); + if (r < 0) + return r; + + LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { + uint64_t v; + + if (streq(property, "BlockIOReadBandwidth")) + v = b->rbps; + else + v = b->wbps; + + if (v == CGROUP_LIMIT_MAX) + continue; + + r = sd_bus_message_append(reply, "(st)", b->path, v); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_device_allow( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + CGroupContext *c = userdata; + CGroupDeviceAllow *a; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(ss)"); + if (r < 0) + return r; + + LIST_FOREACH(device_allow, a, c->device_allow) { + unsigned k = 0; + char rwm[4]; + + if (a->r) + rwm[k++] = 'r'; + if (a->w) + rwm[k++] = 'w'; + if (a->m) + rwm[k++] = 'm'; + + rwm[k] = 0; + + r = sd_bus_message_append(reply, "(ss)", a->path, rwm); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +const sd_bus_vtable bus_cgroup_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Delegate", "b", bus_property_get_bool, offsetof(CGroupContext, delegate), 0), + SD_BUS_PROPERTY("CPUAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, cpu_accounting), 0), + SD_BUS_PROPERTY("CPUWeight", "t", NULL, offsetof(CGroupContext, cpu_weight), 0), + SD_BUS_PROPERTY("StartupCPUWeight", "t", NULL, offsetof(CGroupContext, startup_cpu_weight), 0), + SD_BUS_PROPERTY("CPUShares", "t", NULL, offsetof(CGroupContext, cpu_shares), 0), + SD_BUS_PROPERTY("StartupCPUShares", "t", NULL, offsetof(CGroupContext, startup_cpu_shares), 0), + SD_BUS_PROPERTY("CPUQuotaPerSecUSec", "t", bus_property_get_usec, offsetof(CGroupContext, cpu_quota_per_sec_usec), 0), + SD_BUS_PROPERTY("IOAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, io_accounting), 0), + SD_BUS_PROPERTY("IOWeight", "t", NULL, offsetof(CGroupContext, io_weight), 0), + SD_BUS_PROPERTY("StartupIOWeight", "t", NULL, offsetof(CGroupContext, startup_io_weight), 0), + SD_BUS_PROPERTY("IODeviceWeight", "a(st)", property_get_io_device_weight, 0, 0), + SD_BUS_PROPERTY("IOReadBandwidthMax", "a(st)", property_get_io_device_limits, 0, 0), + SD_BUS_PROPERTY("IOWriteBandwidthMax", "a(st)", property_get_io_device_limits, 0, 0), + SD_BUS_PROPERTY("IOReadIOPSMax", "a(st)", property_get_io_device_limits, 0, 0), + SD_BUS_PROPERTY("IOWriteIOPSMax", "a(st)", property_get_io_device_limits, 0, 0), + SD_BUS_PROPERTY("BlockIOAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, blockio_accounting), 0), + SD_BUS_PROPERTY("BlockIOWeight", "t", NULL, offsetof(CGroupContext, blockio_weight), 0), + SD_BUS_PROPERTY("StartupBlockIOWeight", "t", NULL, offsetof(CGroupContext, startup_blockio_weight), 0), + SD_BUS_PROPERTY("BlockIODeviceWeight", "a(st)", property_get_blockio_device_weight, 0, 0), + SD_BUS_PROPERTY("BlockIOReadBandwidth", "a(st)", property_get_blockio_device_bandwidths, 0, 0), + SD_BUS_PROPERTY("BlockIOWriteBandwidth", "a(st)", property_get_blockio_device_bandwidths, 0, 0), + SD_BUS_PROPERTY("MemoryAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, memory_accounting), 0), + SD_BUS_PROPERTY("MemoryLow", "t", NULL, offsetof(CGroupContext, memory_low), 0), + SD_BUS_PROPERTY("MemoryHigh", "t", NULL, offsetof(CGroupContext, memory_high), 0), + SD_BUS_PROPERTY("MemoryMax", "t", NULL, offsetof(CGroupContext, memory_max), 0), + SD_BUS_PROPERTY("MemorySwapMax", "t", NULL, offsetof(CGroupContext, memory_swap_max), 0), + SD_BUS_PROPERTY("MemoryLimit", "t", NULL, offsetof(CGroupContext, memory_limit), 0), + SD_BUS_PROPERTY("DevicePolicy", "s", property_get_cgroup_device_policy, offsetof(CGroupContext, device_policy), 0), + SD_BUS_PROPERTY("DeviceAllow", "a(ss)", property_get_device_allow, 0, 0), + SD_BUS_PROPERTY("TasksAccounting", "b", bus_property_get_bool, offsetof(CGroupContext, tasks_accounting), 0), + SD_BUS_PROPERTY("TasksMax", "t", NULL, offsetof(CGroupContext, tasks_max), 0), + SD_BUS_VTABLE_END +}; + +static int bus_cgroup_set_transient_property( + Unit *u, + CGroupContext *c, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(u); + assert(c); + assert(name); + assert(message); + + if (streq(name, "Delegate")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->delegate = b; + unit_write_drop_in_private(u, mode, name, b ? "Delegate=yes" : "Delegate=no"); + } + + return 1; + } + + return 0; +} + +int bus_cgroup_set_property( + Unit *u, + CGroupContext *c, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + CGroupIOLimitType iol_type; + int r; + + assert(u); + assert(c); + assert(name); + assert(message); + + if (streq(name, "CPUAccounting")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->cpu_accounting = b; + unit_invalidate_cgroup(u, CGROUP_MASK_CPUACCT|CGROUP_MASK_CPU); + unit_write_drop_in_private(u, mode, name, b ? "CPUAccounting=yes" : "CPUAccounting=no"); + } + + return 1; + + } else if (streq(name, "CPUWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "CPUWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->cpu_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (weight == CGROUP_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "CPUWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "CPUWeight=%" PRIu64, weight); + } + + return 1; + + } else if (streq(name, "StartupCPUWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->startup_cpu_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (weight == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupCPUWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupCPUWeight=%" PRIu64, weight); + } + + return 1; + + } else if (streq(name, "CPUShares")) { + uint64_t shares; + + r = sd_bus_message_read(message, "t", &shares); + if (r < 0) + return r; + + if (!CGROUP_CPU_SHARES_IS_OK(shares)) + return sd_bus_error_set_errnof(error, EINVAL, "CPUShares value out of range"); + + if (mode != UNIT_CHECK) { + c->cpu_shares = shares; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (shares == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "CPUShares="); + else + unit_write_drop_in_private_format(u, mode, name, "CPUShares=%" PRIu64, shares); + } + + return 1; + + } else if (streq(name, "StartupCPUShares")) { + uint64_t shares; + + r = sd_bus_message_read(message, "t", &shares); + if (r < 0) + return r; + + if (!CGROUP_CPU_SHARES_IS_OK(shares)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUShares value out of range"); + + if (mode != UNIT_CHECK) { + c->startup_cpu_shares = shares; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + + if (shares == CGROUP_CPU_SHARES_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupCPUShares="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupCPUShares=%" PRIu64, shares); + } + + return 1; + + } else if (streq(name, "CPUQuotaPerSecUSec")) { + uint64_t u64; + + r = sd_bus_message_read(message, "t", &u64); + if (r < 0) + return r; + + if (u64 <= 0) + return sd_bus_error_set_errnof(error, EINVAL, "CPUQuotaPerSecUSec value out of range"); + + if (mode != UNIT_CHECK) { + c->cpu_quota_per_sec_usec = u64; + unit_invalidate_cgroup(u, CGROUP_MASK_CPU); + unit_write_drop_in_private_format(u, mode, "CPUQuota", "CPUQuota=%0.f%%", (double) (c->cpu_quota_per_sec_usec / 10000)); + } + + return 1; + + } else if (streq(name, "IOAccounting")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->io_accounting = b; + unit_invalidate_cgroup(u, CGROUP_MASK_IO); + unit_write_drop_in_private(u, mode, name, b ? "IOAccounting=yes" : "IOAccounting=no"); + } + + return 1; + + } else if (streq(name, "IOWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "IOWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->io_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_IO); + + if (weight == CGROUP_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "IOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "IOWeight=%" PRIu64, weight); + } + + return 1; + + } else if (streq(name, "StartupIOWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (CGROUP_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupIOWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->startup_io_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_IO); + + if (weight == CGROUP_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupIOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupIOWeight=%" PRIu64, weight); + } + + return 1; + + } else if ((iol_type = cgroup_io_limit_type_from_string(name)) >= 0) { + const char *path; + unsigned n = 0; + uint64_t u64; + + r = sd_bus_message_enter_container(message, 'a', "(st)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "(st)", &path, &u64)) > 0) { + + if (mode != UNIT_CHECK) { + CGroupIODeviceLimit *a = NULL, *b; + + LIST_FOREACH(device_limits, b, c->io_device_limits) { + if (path_equal(path, b->path)) { + a = b; + break; + } + } + + if (!a) { + CGroupIOLimitType type; + + a = new0(CGroupIODeviceLimit, 1); + if (!a) + return -ENOMEM; + + a->path = strdup(path); + if (!a->path) { + free(a); + return -ENOMEM; + } + + for (type = 0; type < _CGROUP_IO_LIMIT_TYPE_MAX; type++) + a->limits[type] = cgroup_io_limit_defaults[type]; + + LIST_PREPEND(device_limits, c->io_device_limits, a); + } + + a->limits[iol_type] = u64; + } + + n++; + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + CGroupIODeviceLimit *a; + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + size_t size = 0; + + if (n == 0) { + LIST_FOREACH(device_limits, a, c->io_device_limits) + a->limits[iol_type] = cgroup_io_limit_defaults[iol_type]; + } + + unit_invalidate_cgroup(u, CGROUP_MASK_IO); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + fprintf(f, "%s=\n", name); + LIST_FOREACH(device_limits, a, c->io_device_limits) + if (a->limits[iol_type] != cgroup_io_limit_defaults[iol_type]) + fprintf(f, "%s=%s %" PRIu64 "\n", name, a->path, a->limits[iol_type]); + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(u, mode, name, buf); + } + + return 1; + + } else if (streq(name, "IODeviceWeight")) { + const char *path; + uint64_t weight; + unsigned n = 0; + + r = sd_bus_message_enter_container(message, 'a', "(st)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "(st)", &path, &weight)) > 0) { + + if (!CGROUP_WEIGHT_IS_OK(weight) || weight == CGROUP_WEIGHT_INVALID) + return sd_bus_error_set_errnof(error, EINVAL, "IODeviceWeight out of range"); + + if (mode != UNIT_CHECK) { + CGroupIODeviceWeight *a = NULL, *b; + + LIST_FOREACH(device_weights, b, c->io_device_weights) { + if (path_equal(b->path, path)) { + a = b; + break; + } + } + + if (!a) { + a = new0(CGroupIODeviceWeight, 1); + if (!a) + return -ENOMEM; + + a->path = strdup(path); + if (!a->path) { + free(a); + return -ENOMEM; + } + LIST_PREPEND(device_weights,c->io_device_weights, a); + } + + a->weight = weight; + } + + n++; + } + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + CGroupIODeviceWeight *a; + size_t size = 0; + + if (n == 0) { + while (c->io_device_weights) + cgroup_context_free_io_device_weight(c, c->io_device_weights); + } + + unit_invalidate_cgroup(u, CGROUP_MASK_IO); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + fputs("IODeviceWeight=\n", f); + LIST_FOREACH(device_weights, a, c->io_device_weights) + fprintf(f, "IODeviceWeight=%s %" PRIu64 "\n", a->path, a->weight); + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(u, mode, name, buf); + } + + return 1; + + } else if (streq(name, "BlockIOAccounting")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->blockio_accounting = b; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + unit_write_drop_in_private(u, mode, name, b ? "BlockIOAccounting=yes" : "BlockIOAccounting=no"); + } + + return 1; + + } else if (streq(name, "BlockIOWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_BLKIO_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "BlockIOWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->blockio_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + if (weight == CGROUP_BLKIO_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "BlockIOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "BlockIOWeight=%" PRIu64, weight); + } + + return 1; + + } else if (streq(name, "StartupBlockIOWeight")) { + uint64_t weight; + + r = sd_bus_message_read(message, "t", &weight); + if (r < 0) + return r; + + if (!CGROUP_BLKIO_WEIGHT_IS_OK(weight)) + return sd_bus_error_set_errnof(error, EINVAL, "StartupBlockIOWeight value out of range"); + + if (mode != UNIT_CHECK) { + c->startup_blockio_weight = weight; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + if (weight == CGROUP_BLKIO_WEIGHT_INVALID) + unit_write_drop_in_private(u, mode, name, "StartupBlockIOWeight="); + else + unit_write_drop_in_private_format(u, mode, name, "StartupBlockIOWeight=%" PRIu64, weight); + } + + return 1; + + } else if (STR_IN_SET(name, "BlockIOReadBandwidth", "BlockIOWriteBandwidth")) { + const char *path; + bool read = true; + unsigned n = 0; + uint64_t u64; + + if (streq(name, "BlockIOWriteBandwidth")) + read = false; + + r = sd_bus_message_enter_container(message, 'a', "(st)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "(st)", &path, &u64)) > 0) { + + if (mode != UNIT_CHECK) { + CGroupBlockIODeviceBandwidth *a = NULL, *b; + + LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { + if (path_equal(path, b->path)) { + a = b; + break; + } + } + + if (!a) { + a = new0(CGroupBlockIODeviceBandwidth, 1); + if (!a) + return -ENOMEM; + + a->rbps = CGROUP_LIMIT_MAX; + a->wbps = CGROUP_LIMIT_MAX; + a->path = strdup(path); + if (!a->path) { + free(a); + return -ENOMEM; + } + + LIST_PREPEND(device_bandwidths, c->blockio_device_bandwidths, a); + } + + if (read) + a->rbps = u64; + else + a->wbps = u64; + } + + n++; + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + CGroupBlockIODeviceBandwidth *a; + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + size_t size = 0; + + if (n == 0) { + LIST_FOREACH(device_bandwidths, a, c->blockio_device_bandwidths) { + if (read) + a->rbps = CGROUP_LIMIT_MAX; + else + a->wbps = CGROUP_LIMIT_MAX; + } + } + + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + if (read) { + fputs("BlockIOReadBandwidth=\n", f); + LIST_FOREACH(device_bandwidths, a, c->blockio_device_bandwidths) + if (a->rbps != CGROUP_LIMIT_MAX) + fprintf(f, "BlockIOReadBandwidth=%s %" PRIu64 "\n", a->path, a->rbps); + } else { + fputs("BlockIOWriteBandwidth=\n", f); + LIST_FOREACH(device_bandwidths, a, c->blockio_device_bandwidths) + if (a->wbps != CGROUP_LIMIT_MAX) + fprintf(f, "BlockIOWriteBandwidth=%s %" PRIu64 "\n", a->path, a->wbps); + } + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(u, mode, name, buf); + } + + return 1; + + } else if (streq(name, "BlockIODeviceWeight")) { + const char *path; + uint64_t weight; + unsigned n = 0; + + r = sd_bus_message_enter_container(message, 'a', "(st)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "(st)", &path, &weight)) > 0) { + + if (!CGROUP_BLKIO_WEIGHT_IS_OK(weight) || weight == CGROUP_BLKIO_WEIGHT_INVALID) + return sd_bus_error_set_errnof(error, EINVAL, "BlockIODeviceWeight out of range"); + + if (mode != UNIT_CHECK) { + CGroupBlockIODeviceWeight *a = NULL, *b; + + LIST_FOREACH(device_weights, b, c->blockio_device_weights) { + if (path_equal(b->path, path)) { + a = b; + break; + } + } + + if (!a) { + a = new0(CGroupBlockIODeviceWeight, 1); + if (!a) + return -ENOMEM; + + a->path = strdup(path); + if (!a->path) { + free(a); + return -ENOMEM; + } + LIST_PREPEND(device_weights,c->blockio_device_weights, a); + } + + a->weight = weight; + } + + n++; + } + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + CGroupBlockIODeviceWeight *a; + size_t size = 0; + + if (n == 0) { + while (c->blockio_device_weights) + cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); + } + + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + fputs("BlockIODeviceWeight=\n", f); + LIST_FOREACH(device_weights, a, c->blockio_device_weights) + fprintf(f, "BlockIODeviceWeight=%s %" PRIu64 "\n", a->path, a->weight); + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(u, mode, name, buf); + } + + return 1; + + } else if (streq(name, "MemoryAccounting")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->memory_accounting = b; + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); + unit_write_drop_in_private(u, mode, name, b ? "MemoryAccounting=yes" : "MemoryAccounting=no"); + } + + return 1; + + } else if (STR_IN_SET(name, "MemoryLow", "MemoryHigh", "MemoryMax", "MemorySwapMax")) { + uint64_t v; + + r = sd_bus_message_read(message, "t", &v); + if (r < 0) + return r; + if (v <= 0) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is too small", name); + + if (mode != UNIT_CHECK) { + if (streq(name, "MemoryLow")) + c->memory_low = v; + else if (streq(name, "MemoryHigh")) + c->memory_high = v; + else if (streq(name, "MemorySwapMax")) + c->memory_swap_max = v; + else + c->memory_max = v; + + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); + + if (v == CGROUP_LIMIT_MAX) + unit_write_drop_in_private_format(u, mode, name, "%s=infinity", name); + else + unit_write_drop_in_private_format(u, mode, name, "%s=%" PRIu64, name, v); + } + + return 1; + + } else if (STR_IN_SET(name, "MemoryLowScale", "MemoryHighScale", "MemoryMaxScale")) { + uint32_t raw; + uint64_t v; + + r = sd_bus_message_read(message, "u", &raw); + if (r < 0) + return r; + + v = physical_memory_scale(raw, UINT32_MAX); + if (v <= 0 || v == UINT64_MAX) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is out of range", name); + + if (mode != UNIT_CHECK) { + const char *e; + + /* Chop off suffix */ + assert_se(e = endswith(name, "Scale")); + name = strndupa(name, e - name); + + if (streq(name, "MemoryLow")) + c->memory_low = v; + else if (streq(name, "MemoryHigh")) + c->memory_high = v; + else + c->memory_max = v; + + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); + unit_write_drop_in_private_format(u, mode, name, "%s=%" PRIu32 "%%", name, + (uint32_t) (DIV_ROUND_UP((uint64_t) raw * 100U, (uint64_t) UINT32_MAX))); + } + + return 1; + + } else if (streq(name, "MemoryLimit")) { + uint64_t limit; + + r = sd_bus_message_read(message, "t", &limit); + if (r < 0) + return r; + if (limit <= 0) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is too small", name); + + if (mode != UNIT_CHECK) { + c->memory_limit = limit; + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); + + if (limit == (uint64_t) -1) + unit_write_drop_in_private(u, mode, name, "MemoryLimit=infinity"); + else + unit_write_drop_in_private_format(u, mode, name, "MemoryLimit=%" PRIu64, limit); + } + + return 1; + + } else if (streq(name, "MemoryLimitScale")) { + uint64_t limit; + uint32_t raw; + + r = sd_bus_message_read(message, "u", &raw); + if (r < 0) + return r; + + limit = physical_memory_scale(raw, UINT32_MAX); + if (limit <= 0 || limit == UINT64_MAX) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is out of range", name); + + if (mode != UNIT_CHECK) { + c->memory_limit = limit; + unit_invalidate_cgroup(u, CGROUP_MASK_MEMORY); + unit_write_drop_in_private_format(u, mode, "MemoryLimit", "MemoryLimit=%" PRIu32 "%%", + (uint32_t) (DIV_ROUND_UP((uint64_t) raw * 100U, (uint64_t) UINT32_MAX))); + } + + return 1; + + } else if (streq(name, "DevicePolicy")) { + const char *policy; + CGroupDevicePolicy p; + + r = sd_bus_message_read(message, "s", &policy); + if (r < 0) + return r; + + p = cgroup_device_policy_from_string(policy); + if (p < 0) + return -EINVAL; + + if (mode != UNIT_CHECK) { + c->device_policy = p; + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); + unit_write_drop_in_private_format(u, mode, name, "DevicePolicy=%s", policy); + } + + return 1; + + } else if (streq(name, "DeviceAllow")) { + const char *path, *rwm; + unsigned n = 0; + + r = sd_bus_message_enter_container(message, 'a', "(ss)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "(ss)", &path, &rwm)) > 0) { + + if ((!startswith(path, "/dev/") && + !startswith(path, "/run/systemd/inaccessible/") && + !startswith(path, "block-") && + !startswith(path, "char-")) || + strpbrk(path, WHITESPACE)) + return sd_bus_error_set_errnof(error, EINVAL, "DeviceAllow= requires device node"); + + if (isempty(rwm)) + rwm = "rwm"; + + if (!in_charset(rwm, "rwm")) + return sd_bus_error_set_errnof(error, EINVAL, "DeviceAllow= requires combination of rwm flags"); + + if (mode != UNIT_CHECK) { + CGroupDeviceAllow *a = NULL, *b; + + LIST_FOREACH(device_allow, b, c->device_allow) { + if (path_equal(b->path, path)) { + a = b; + break; + } + } + + if (!a) { + a = new0(CGroupDeviceAllow, 1); + if (!a) + return -ENOMEM; + + a->path = strdup(path); + if (!a->path) { + free(a); + return -ENOMEM; + } + + LIST_PREPEND(device_allow, c->device_allow, a); + } + + a->r = !!strchr(rwm, 'r'); + a->w = !!strchr(rwm, 'w'); + a->m = !!strchr(rwm, 'm'); + } + + n++; + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + CGroupDeviceAllow *a; + size_t size = 0; + + if (n == 0) { + while (c->device_allow) + cgroup_context_free_device_allow(c, c->device_allow); + } + + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + fputs("DeviceAllow=\n", f); + LIST_FOREACH(device_allow, a, c->device_allow) + fprintf(f, "DeviceAllow=%s %s%s%s\n", a->path, a->r ? "r" : "", a->w ? "w" : "", a->m ? "m" : ""); + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(u, mode, name, buf); + } + + return 1; + + } else if (streq(name, "TasksAccounting")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->tasks_accounting = b; + unit_invalidate_cgroup(u, CGROUP_MASK_PIDS); + unit_write_drop_in_private(u, mode, name, b ? "TasksAccounting=yes" : "TasksAccounting=no"); + } + + return 1; + + } else if (streq(name, "TasksMax")) { + uint64_t limit; + + r = sd_bus_message_read(message, "t", &limit); + if (r < 0) + return r; + if (limit <= 0) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is too small", name); + + if (mode != UNIT_CHECK) { + c->tasks_max = limit; + unit_invalidate_cgroup(u, CGROUP_MASK_PIDS); + + if (limit == (uint64_t) -1) + unit_write_drop_in_private(u, mode, name, "TasksMax=infinity"); + else + unit_write_drop_in_private_format(u, mode, name, "TasksMax=%" PRIu64, limit); + } + + return 1; + } else if (streq(name, "TasksMaxScale")) { + uint64_t limit; + uint32_t raw; + + r = sd_bus_message_read(message, "u", &raw); + if (r < 0) + return r; + + limit = system_tasks_max_scale(raw, UINT32_MAX); + if (limit <= 0 || limit >= UINT64_MAX) + return sd_bus_error_set_errnof(error, EINVAL, "%s= is out of range", name); + + if (mode != UNIT_CHECK) { + c->tasks_max = limit; + unit_invalidate_cgroup(u, CGROUP_MASK_PIDS); + unit_write_drop_in_private_format(u, mode, name, "TasksMax=%" PRIu32 "%%", + (uint32_t) (DIV_ROUND_UP((uint64_t) raw * 100U, (uint64_t) UINT32_MAX))); + } + + return 1; + } + + if (u->transient && u->load_state == UNIT_STUB) { + r = bus_cgroup_set_transient_property(u, c, name, message, mode, error); + if (r != 0) + return r; + + } + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-cgroup.h b/src/grp-system/libcore/src/dbus-cgroup.h new file mode 100644 index 0000000000..b61ca2fcef --- /dev/null +++ b/src/grp-system/libcore/src/dbus-cgroup.h @@ -0,0 +1,28 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/cgroup.h" + +extern const sd_bus_vtable bus_cgroup_vtable[]; + +int bus_cgroup_set_property(Unit *u, CGroupContext *c, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); diff --git a/src/grp-system/libcore/src/dbus-device.c b/src/grp-system/libcore/src/dbus-device.c new file mode 100644 index 0000000000..75e9beb55e --- /dev/null +++ b/src/grp-system/libcore/src/dbus-device.c @@ -0,0 +1,29 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/device.h" +#include "core/unit.h" + +#include "dbus-device.h" + +const sd_bus_vtable bus_device_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("SysFSPath", "s", NULL, offsetof(Device, sysfs), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_VTABLE_END +}; diff --git a/src/grp-system/libcore/src/dbus-device.h b/src/grp-system/libcore/src/dbus-device.h new file mode 100644 index 0000000000..84b7ed56ec --- /dev/null +++ b/src/grp-system/libcore/src/dbus-device.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/unit.h" + +extern const sd_bus_vtable bus_device_vtable[]; diff --git a/src/grp-system/libcore/src/dbus-execute.c b/src/grp-system/libcore/src/dbus-execute.c new file mode 100644 index 0000000000..2e447e8043 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-execute.c @@ -0,0 +1,1666 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/prctl.h> + +#ifdef HAVE_SECCOMP +#include <seccomp.h> +#endif + +#include "core/execute.h" +#include "core/namespace.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/af-list.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/capability-util.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/ioprio.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/rlimit-util.h" + +#include "dbus-execute.h" +#ifdef HAVE_SECCOMP +#include "systemd-shared/seccomp-util.h" +#endif +#include "systemd-basic/strv.h" +#include "systemd-basic/syslog-util.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/utf8.h" + +BUS_DEFINE_PROPERTY_GET_ENUM(bus_property_get_exec_output, exec_output, ExecOutput); + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_exec_input, exec_input, ExecInput); + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_exec_utmp_mode, exec_utmp_mode, ExecUtmpMode); + +static BUS_DEFINE_PROPERTY_GET_ENUM(bus_property_get_protect_home, protect_home, ProtectHome); +static BUS_DEFINE_PROPERTY_GET_ENUM(bus_property_get_protect_system, protect_system, ProtectSystem); + +static int property_get_environment_files( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + char **j; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'a', "(sb)"); + if (r < 0) + return r; + + STRV_FOREACH(j, c->environment_files) { + const char *fn = *j; + + r = sd_bus_message_append(reply, "(sb)", fn[0] == '-' ? fn + 1 : fn, fn[0] == '-'); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_oom_score_adjust( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + + ExecContext *c = userdata; + int32_t n; + + assert(bus); + assert(reply); + assert(c); + + if (c->oom_score_adjust_set) + n = c->oom_score_adjust; + else { + _cleanup_free_ char *t = NULL; + + n = 0; + if (read_one_line_file("/proc/self/oom_score_adj", &t) >= 0) + safe_atoi32(t, &n); + } + + return sd_bus_message_append(reply, "i", n); +} + +static int property_get_nice( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + + ExecContext *c = userdata; + int32_t n; + + assert(bus); + assert(reply); + assert(c); + + if (c->nice_set) + n = c->nice; + else { + errno = 0; + n = getpriority(PRIO_PROCESS, 0); + if (errno > 0) + n = 0; + } + + return sd_bus_message_append(reply, "i", n); +} + +static int property_get_ioprio( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + + ExecContext *c = userdata; + int32_t n; + + assert(bus); + assert(reply); + assert(c); + + if (c->ioprio_set) + n = c->ioprio; + else { + n = ioprio_get(IOPRIO_WHO_PROCESS, 0); + if (n < 0) + n = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 4); + } + + return sd_bus_message_append(reply, "i", n); +} + +static int property_get_cpu_sched_policy( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + int32_t n; + + assert(bus); + assert(reply); + assert(c); + + if (c->cpu_sched_set) + n = c->cpu_sched_policy; + else { + n = sched_getscheduler(0); + if (n < 0) + n = SCHED_OTHER; + } + + return sd_bus_message_append(reply, "i", n); +} + +static int property_get_cpu_sched_priority( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + int32_t n; + + assert(bus); + assert(reply); + assert(c); + + if (c->cpu_sched_set) + n = c->cpu_sched_priority; + else { + struct sched_param p = {}; + + if (sched_getparam(0, &p) >= 0) + n = p.sched_priority; + else + n = 0; + } + + return sd_bus_message_append(reply, "i", n); +} + +static int property_get_cpu_affinity( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + if (c->cpuset) + return sd_bus_message_append_array(reply, 'y', c->cpuset, CPU_ALLOC_SIZE(c->cpuset_ncpus)); + else + return sd_bus_message_append_array(reply, 'y', NULL, 0); +} + +static int property_get_timer_slack_nsec( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + uint64_t u; + + assert(bus); + assert(reply); + assert(c); + + if (c->timer_slack_nsec != NSEC_INFINITY) + u = (uint64_t) c->timer_slack_nsec; + else + u = (uint64_t) prctl(PR_GET_TIMERSLACK); + + return sd_bus_message_append(reply, "t", u); +} + +static int property_get_capability_bounding_set( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "t", c->capability_bounding_set); +} + +static int property_get_ambient_capabilities( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "t", c->capability_ambient_set); +} + +static int property_get_empty_string( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "s", ""); +} + +static int property_get_syscall_filter( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + _cleanup_strv_free_ char **l = NULL; + int r; + +#ifdef HAVE_SECCOMP + Iterator i; + void *id; +#endif + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'r', "bas"); + if (r < 0) + return r; + + r = sd_bus_message_append(reply, "b", c->syscall_whitelist); + if (r < 0) + return r; + +#ifdef HAVE_SECCOMP + SET_FOREACH(id, c->syscall_filter, i) { + char *name; + + name = seccomp_syscall_resolve_num_arch(SCMP_ARCH_NATIVE, PTR_TO_INT(id) - 1); + if (!name) + continue; + + r = strv_consume(&l, name); + if (r < 0) + return r; + } +#endif + + strv_sort(l); + + r = sd_bus_message_append_strv(reply, l); + if (r < 0) + return r; + + return sd_bus_message_close_container(reply); +} + +static int property_get_syscall_archs( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + _cleanup_strv_free_ char **l = NULL; + int r; + +#ifdef HAVE_SECCOMP + Iterator i; + void *id; +#endif + + assert(bus); + assert(reply); + assert(c); + +#ifdef HAVE_SECCOMP + SET_FOREACH(id, c->syscall_archs, i) { + const char *name; + + name = seccomp_arch_to_string(PTR_TO_UINT32(id) - 1); + if (!name) + continue; + + r = strv_extend(&l, name); + if (r < 0) + return -ENOMEM; + } +#endif + + strv_sort(l); + + r = sd_bus_message_append_strv(reply, l); + if (r < 0) + return r; + + return 0; +} + +static int property_get_syscall_errno( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "i", (int32_t) c->syscall_errno); +} + +static int property_get_selinux_context( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "(bs)", c->selinux_context_ignore, c->selinux_context); +} + +static int property_get_apparmor_profile( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "(bs)", c->apparmor_profile_ignore, c->apparmor_profile); +} + +static int property_get_smack_process_label( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "(bs)", c->smack_process_label_ignore, c->smack_process_label); +} + +static int property_get_personality( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "s", personality_to_string(c->personality)); +} + +static int property_get_address_families( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + _cleanup_strv_free_ char **l = NULL; + Iterator i; + void *af; + int r; + + assert(bus); + assert(reply); + assert(c); + + r = sd_bus_message_open_container(reply, 'r', "bas"); + if (r < 0) + return r; + + r = sd_bus_message_append(reply, "b", c->address_families_whitelist); + if (r < 0) + return r; + + SET_FOREACH(af, c->address_families, i) { + const char *name; + + name = af_to_name(PTR_TO_INT(af)); + if (!name) + continue; + + r = strv_extend(&l, name); + if (r < 0) + return -ENOMEM; + } + + strv_sort(l); + + r = sd_bus_message_append_strv(reply, l); + if (r < 0) + return r; + + return sd_bus_message_close_container(reply); +} + +static int property_get_working_directory( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + const char *wd; + + assert(bus); + assert(reply); + assert(c); + + if (c->working_directory_home) + wd = "~"; + else + wd = c->working_directory; + + if (c->working_directory_missing_ok) + wd = strjoina("!", wd); + + return sd_bus_message_append(reply, "s", wd); +} + +static int property_get_syslog_level( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "i", LOG_PRI(c->syslog_priority)); +} + +static int property_get_syslog_facility( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + + assert(bus); + assert(reply); + assert(c); + + return sd_bus_message_append(reply, "i", LOG_FAC(c->syslog_priority)); +} + +static int property_get_input_fdname( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + const char *name; + + assert(bus); + assert(c); + assert(property); + assert(reply); + + name = exec_context_fdname(c, STDIN_FILENO); + + return sd_bus_message_append(reply, "s", name); +} + +static int property_get_output_fdname( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + ExecContext *c = userdata; + const char *name = NULL; + + assert(bus); + assert(c); + assert(property); + assert(reply); + + if (c->std_output == EXEC_OUTPUT_NAMED_FD && streq(property, "StandardOutputFileDescriptorName")) + name = exec_context_fdname(c, STDOUT_FILENO); + else if (c->std_error == EXEC_OUTPUT_NAMED_FD && streq(property, "StandardErrorFileDescriptorName")) + name = exec_context_fdname(c, STDERR_FILENO); + + return sd_bus_message_append(reply, "s", name); +} + +const sd_bus_vtable bus_exec_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Environment", "as", NULL, offsetof(ExecContext, environment), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("EnvironmentFiles", "a(sb)", property_get_environment_files, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PassEnvironment", "as", NULL, offsetof(ExecContext, pass_environment), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UMask", "u", bus_property_get_mode, offsetof(ExecContext, umask), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitCPU", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CPU]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitCPUSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CPU]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitFSIZE", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_FSIZE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitFSIZESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_FSIZE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitDATA", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_DATA]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitDATASoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_DATA]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitSTACK", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_STACK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitSTACKSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_STACK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitCORE", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CORE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitCORESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CORE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRSS", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RSS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRSSSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RSS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNOFILE", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NOFILE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNOFILESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NOFILE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitAS", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_AS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitASSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_AS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNPROC", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NPROC]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNPROCSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NPROC]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitMEMLOCK", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MEMLOCK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitMEMLOCKSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MEMLOCK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitLOCKS", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_LOCKS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitLOCKSSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_LOCKS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitSIGPENDING", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_SIGPENDING]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitSIGPENDINGSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_SIGPENDING]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitMSGQUEUE", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MSGQUEUE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitMSGQUEUESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MSGQUEUE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNICE", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NICE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitNICESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NICE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRTPRIO", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTPRIO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRTPRIOSoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTPRIO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRTTIME", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LimitRTTIMESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("WorkingDirectory", "s", property_get_working_directory, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RootDirectory", "s", NULL, offsetof(ExecContext, root_directory), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("OOMScoreAdjust", "i", property_get_oom_score_adjust, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Nice", "i", property_get_nice, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("IOScheduling", "i", property_get_ioprio, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CPUSchedulingPolicy", "i", property_get_cpu_sched_policy, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CPUSchedulingPriority", "i", property_get_cpu_sched_priority, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CPUAffinity", "ay", property_get_cpu_affinity, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimerSlackNSec", "t", property_get_timer_slack_nsec, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CPUSchedulingResetOnFork", "b", bus_property_get_bool, offsetof(ExecContext, cpu_sched_reset_on_fork), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NonBlocking", "b", bus_property_get_bool, offsetof(ExecContext, non_blocking), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardInput", "s", property_get_exec_input, offsetof(ExecContext, std_input), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardInputFileDescriptorName", "s", property_get_input_fdname, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardOutput", "s", bus_property_get_exec_output, offsetof(ExecContext, std_output), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardOutputFileDescriptorName", "s", property_get_output_fdname, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardError", "s", bus_property_get_exec_output, offsetof(ExecContext, std_error), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StandardErrorFileDescriptorName", "s", property_get_output_fdname, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TTYPath", "s", NULL, offsetof(ExecContext, tty_path), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TTYReset", "b", bus_property_get_bool, offsetof(ExecContext, tty_reset), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TTYVHangup", "b", bus_property_get_bool, offsetof(ExecContext, tty_vhangup), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TTYVTDisallocate", "b", bus_property_get_bool, offsetof(ExecContext, tty_vt_disallocate), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SyslogPriority", "i", bus_property_get_int, offsetof(ExecContext, syslog_priority), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SyslogIdentifier", "s", NULL, offsetof(ExecContext, syslog_identifier), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SyslogLevelPrefix", "b", bus_property_get_bool, offsetof(ExecContext, syslog_level_prefix), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SyslogLevel", "i", property_get_syslog_level, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SyslogFacility", "i", property_get_syslog_facility, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Capabilities", "s", property_get_empty_string, 0, SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("SecureBits", "i", bus_property_get_int, offsetof(ExecContext, secure_bits), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CapabilityBoundingSet", "t", property_get_capability_bounding_set, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("AmbientCapabilities", "t", property_get_ambient_capabilities, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("User", "s", NULL, offsetof(ExecContext, user), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Group", "s", NULL, offsetof(ExecContext, group), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DynamicUser", "b", bus_property_get_bool, offsetof(ExecContext, dynamic_user), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RemoveIPC", "b", bus_property_get_bool, offsetof(ExecContext, remove_ipc), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SupplementaryGroups", "as", NULL, offsetof(ExecContext, supplementary_groups), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PAMName", "s", NULL, offsetof(ExecContext, pam_name), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ReadWriteDirectories", "as", NULL, offsetof(ExecContext, read_write_paths), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("ReadOnlyDirectories", "as", NULL, offsetof(ExecContext, read_only_paths), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("InaccessibleDirectories", "as", NULL, offsetof(ExecContext, inaccessible_paths), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("ReadWritePaths", "as", NULL, offsetof(ExecContext, read_write_paths), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ReadOnlyPaths", "as", NULL, offsetof(ExecContext, read_only_paths), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("InaccessiblePaths", "as", NULL, offsetof(ExecContext, inaccessible_paths), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MountFlags", "t", bus_property_get_ulong, offsetof(ExecContext, mount_flags), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PrivateTmp", "b", bus_property_get_bool, offsetof(ExecContext, private_tmp), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PrivateDevices", "b", bus_property_get_bool, offsetof(ExecContext, private_devices), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ProtectKernelTunables", "b", bus_property_get_bool, offsetof(ExecContext, protect_kernel_tunables), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ProtectKernelModules", "b", bus_property_get_bool, offsetof(ExecContext, protect_kernel_modules), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ProtectControlGroups", "b", bus_property_get_bool, offsetof(ExecContext, protect_control_groups), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PrivateNetwork", "b", bus_property_get_bool, offsetof(ExecContext, private_network), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PrivateUsers", "b", bus_property_get_bool, offsetof(ExecContext, private_users), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ProtectHome", "s", bus_property_get_protect_home, offsetof(ExecContext, protect_home), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ProtectSystem", "s", bus_property_get_protect_system, offsetof(ExecContext, protect_system), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SameProcessGroup", "b", bus_property_get_bool, offsetof(ExecContext, same_pgrp), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UtmpIdentifier", "s", NULL, offsetof(ExecContext, utmp_id), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UtmpMode", "s", property_get_exec_utmp_mode, offsetof(ExecContext, utmp_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SELinuxContext", "(bs)", property_get_selinux_context, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("AppArmorProfile", "(bs)", property_get_apparmor_profile, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SmackProcessLabel", "(bs)", property_get_smack_process_label, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("IgnoreSIGPIPE", "b", bus_property_get_bool, offsetof(ExecContext, ignore_sigpipe), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NoNewPrivileges", "b", bus_property_get_bool, offsetof(ExecContext, no_new_privileges), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SystemCallFilter", "(bas)", property_get_syscall_filter, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SystemCallArchitectures", "as", property_get_syscall_archs, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SystemCallErrorNumber", "i", property_get_syscall_errno, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Personality", "s", property_get_personality, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RestrictAddressFamilies", "(bas)", property_get_address_families, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RuntimeDirectoryMode", "u", bus_property_get_mode, offsetof(ExecContext, runtime_directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RuntimeDirectory", "as", NULL, offsetof(ExecContext, runtime_directory), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MemoryDenyWriteExecute", "b", bus_property_get_bool, offsetof(ExecContext, memory_deny_write_execute), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RestrictRealtime", "b", bus_property_get_bool, offsetof(ExecContext, restrict_realtime), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_VTABLE_END +}; + +static int append_exec_command(sd_bus_message *reply, ExecCommand *c) { + int r; + + assert(reply); + assert(c); + + if (!c->path) + return 0; + + r = sd_bus_message_open_container(reply, 'r', "sasbttttuii"); + if (r < 0) + return r; + + r = sd_bus_message_append(reply, "s", c->path); + if (r < 0) + return r; + + r = sd_bus_message_append_strv(reply, c->argv); + if (r < 0) + return r; + + r = sd_bus_message_append(reply, "bttttuii", + c->ignore, + c->exec_status.start_timestamp.realtime, + c->exec_status.start_timestamp.monotonic, + c->exec_status.exit_timestamp.realtime, + c->exec_status.exit_timestamp.monotonic, + (uint32_t) c->exec_status.pid, + (int32_t) c->exec_status.code, + (int32_t) c->exec_status.status); + if (r < 0) + return r; + + return sd_bus_message_close_container(reply); +} + +int bus_property_get_exec_command( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *ret_error) { + + ExecCommand *c = (ExecCommand*) userdata; + int r; + + assert(bus); + assert(reply); + + r = sd_bus_message_open_container(reply, 'a', "(sasbttttuii)"); + if (r < 0) + return r; + + r = append_exec_command(reply, c); + if (r < 0) + return r; + + return sd_bus_message_close_container(reply); +} + +int bus_property_get_exec_command_list( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *ret_error) { + + ExecCommand *c = *(ExecCommand**) userdata; + int r; + + assert(bus); + assert(reply); + + r = sd_bus_message_open_container(reply, 'a', "(sasbttttuii)"); + if (r < 0) + return r; + + LIST_FOREACH(command, c, c) { + r = append_exec_command(reply, c); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +int bus_exec_context_set_transient_property( + Unit *u, + ExecContext *c, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + const char *soft = NULL; + int r, ri; + + assert(u); + assert(c); + assert(name); + assert(message); + + if (streq(name, "User")) { + const char *uu; + + r = sd_bus_message_read(message, "s", &uu); + if (r < 0) + return r; + + if (!isempty(uu) && !valid_user_group_name_or_id(uu)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid user name: %s", uu); + + if (mode != UNIT_CHECK) { + + if (isempty(uu)) + c->user = mfree(c->user); + else if (free_and_strdup(&c->user, uu) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "User=%s", uu); + } + + return 1; + + } else if (streq(name, "Group")) { + const char *gg; + + r = sd_bus_message_read(message, "s", &gg); + if (r < 0) + return r; + + if (!isempty(gg) && !valid_user_group_name_or_id(gg)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid group name: %s", gg); + + if (mode != UNIT_CHECK) { + + if (isempty(gg)) + c->group = mfree(c->group); + else if (free_and_strdup(&c->group, gg) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "Group=%s", gg); + } + + return 1; + } else if (streq(name, "SyslogIdentifier")) { + const char *id; + + r = sd_bus_message_read(message, "s", &id); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + + if (isempty(id)) + c->syslog_identifier = mfree(c->syslog_identifier); + else if (free_and_strdup(&c->syslog_identifier, id) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "SyslogIdentifier=%s", id); + } + + return 1; + } else if (streq(name, "SyslogLevel")) { + int level; + + r = sd_bus_message_read(message, "i", &level); + if (r < 0) + return r; + + if (!log_level_is_valid(level)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Log level value out of range"); + + if (mode != UNIT_CHECK) { + c->syslog_priority = (c->syslog_priority & LOG_FACMASK) | level; + unit_write_drop_in_private_format(u, mode, name, "SyslogLevel=%i", level); + } + + return 1; + } else if (streq(name, "SyslogFacility")) { + int facility; + + r = sd_bus_message_read(message, "i", &facility); + if (r < 0) + return r; + + if (!log_facility_unshifted_is_valid(facility)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Log facility value out of range"); + + if (mode != UNIT_CHECK) { + c->syslog_priority = (facility << 3) | LOG_PRI(c->syslog_priority); + unit_write_drop_in_private_format(u, mode, name, "SyslogFacility=%i", facility); + } + + return 1; + } else if (streq(name, "Nice")) { + int n; + + r = sd_bus_message_read(message, "i", &n); + if (r < 0) + return r; + + if (!nice_is_valid(n)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Nice value out of range"); + + if (mode != UNIT_CHECK) { + c->nice = n; + unit_write_drop_in_private_format(u, mode, name, "Nice=%i", n); + } + + return 1; + + } else if (STR_IN_SET(name, "TTYPath", "RootDirectory")) { + const char *s; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + if (!path_is_absolute(s)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "%s takes an absolute path", name); + + if (mode != UNIT_CHECK) { + if (streq(name, "TTYPath")) + r = free_and_strdup(&c->tty_path, s); + else { + assert(streq(name, "RootDirectory")); + r = free_and_strdup(&c->root_directory, s); + } + if (r < 0) + return r; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, s); + } + + return 1; + + } else if (streq(name, "WorkingDirectory")) { + const char *s; + bool missing_ok; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + if (s[0] == '-') { + missing_ok = true; + s++; + } else + missing_ok = false; + + if (!streq(s, "~") && !path_is_absolute(s)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "WorkingDirectory= expects an absolute path or '~'"); + + if (mode != UNIT_CHECK) { + if (streq(s, "~")) { + c->working_directory = mfree(c->working_directory); + c->working_directory_home = true; + } else { + r = free_and_strdup(&c->working_directory, s); + if (r < 0) + return r; + + c->working_directory_home = false; + } + + c->working_directory_missing_ok = missing_ok; + unit_write_drop_in_private_format(u, mode, name, "WorkingDirectory=%s%s", missing_ok ? "-" : "", s); + } + + return 1; + + } else if (streq(name, "StandardInput")) { + const char *s; + ExecInput p; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + p = exec_input_from_string(s); + if (p < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid standard input name"); + + if (mode != UNIT_CHECK) { + c->std_input = p; + + unit_write_drop_in_private_format(u, mode, name, "StandardInput=%s", exec_input_to_string(p)); + } + + return 1; + + } else if (streq(name, "StandardOutput")) { + const char *s; + ExecOutput p; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + p = exec_output_from_string(s); + if (p < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid standard output name"); + + if (mode != UNIT_CHECK) { + c->std_output = p; + + unit_write_drop_in_private_format(u, mode, name, "StandardOutput=%s", exec_output_to_string(p)); + } + + return 1; + + } else if (streq(name, "StandardError")) { + const char *s; + ExecOutput p; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + p = exec_output_from_string(s); + if (p < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid standard error name"); + + if (mode != UNIT_CHECK) { + c->std_error = p; + + unit_write_drop_in_private_format(u, mode, name, "StandardError=%s", exec_output_to_string(p)); + } + + return 1; + + } else if (STR_IN_SET(name, + "StandardInputFileDescriptorName", "StandardOutputFileDescriptorName", "StandardErrorFileDescriptorName")) { + const char *s; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + if (!fdname_is_valid(s)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid file descriptor name"); + + if (mode != UNIT_CHECK) { + if (streq(name, "StandardInputFileDescriptorName")) { + c->std_input = EXEC_INPUT_NAMED_FD; + r = free_and_strdup(&c->stdio_fdname[STDIN_FILENO], s); + if (r < 0) + return r; + unit_write_drop_in_private_format(u, mode, name, "StandardInput=fd:%s", s); + } else if (streq(name, "StandardOutputFileDescriptorName")) { + c->std_output = EXEC_OUTPUT_NAMED_FD; + r = free_and_strdup(&c->stdio_fdname[STDOUT_FILENO], s); + if (r < 0) + return r; + unit_write_drop_in_private_format(u, mode, name, "StandardOutput=fd:%s", s); + } else if (streq(name, "StandardErrorFileDescriptorName")) { + c->std_error = EXEC_OUTPUT_NAMED_FD; + r = free_and_strdup(&c->stdio_fdname[STDERR_FILENO], s); + if (r < 0) + return r; + unit_write_drop_in_private_format(u, mode, name, "StandardError=fd:%s", s); + } + } + + return 1; + + } else if (STR_IN_SET(name, + "IgnoreSIGPIPE", "TTYVHangup", "TTYReset", + "PrivateTmp", "PrivateDevices", "PrivateNetwork", "PrivateUsers", + "NoNewPrivileges", "SyslogLevelPrefix", "MemoryDenyWriteExecute", + "RestrictRealtime", "DynamicUser", "RemoveIPC", "ProtectKernelTunables", + "ProtectKernelModules", "ProtectControlGroups")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + if (streq(name, "IgnoreSIGPIPE")) + c->ignore_sigpipe = b; + else if (streq(name, "TTYVHangup")) + c->tty_vhangup = b; + else if (streq(name, "TTYReset")) + c->tty_reset = b; + else if (streq(name, "PrivateTmp")) + c->private_tmp = b; + else if (streq(name, "PrivateDevices")) + c->private_devices = b; + else if (streq(name, "PrivateNetwork")) + c->private_network = b; + else if (streq(name, "PrivateUsers")) + c->private_users = b; + else if (streq(name, "NoNewPrivileges")) + c->no_new_privileges = b; + else if (streq(name, "SyslogLevelPrefix")) + c->syslog_level_prefix = b; + else if (streq(name, "MemoryDenyWriteExecute")) + c->memory_deny_write_execute = b; + else if (streq(name, "RestrictRealtime")) + c->restrict_realtime = b; + else if (streq(name, "DynamicUser")) + c->dynamic_user = b; + else if (streq(name, "RemoveIPC")) + c->remove_ipc = b; + else if (streq(name, "ProtectKernelTunables")) + c->protect_kernel_tunables = b; + else if (streq(name, "ProtectKernelModules")) + c->protect_kernel_modules = b; + else if (streq(name, "ProtectControlGroups")) + c->protect_control_groups = b; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, yes_no(b)); + } + + return 1; + + } else if (streq(name, "UtmpIdentifier")) { + const char *id; + + r = sd_bus_message_read(message, "s", &id); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + if (isempty(id)) + c->utmp_id = mfree(c->utmp_id); + else if (free_and_strdup(&c->utmp_id, id) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "UtmpIdentifier=%s", strempty(id)); + } + + return 1; + + } else if (streq(name, "UtmpMode")) { + const char *s; + ExecUtmpMode m; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + m = exec_utmp_mode_from_string(s); + if (m < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid utmp mode"); + + if (mode != UNIT_CHECK) { + c->utmp_mode = m; + + unit_write_drop_in_private_format(u, mode, name, "UtmpMode=%s", exec_utmp_mode_to_string(m)); + } + + return 1; + + } else if (streq(name, "PAMName")) { + const char *n; + + r = sd_bus_message_read(message, "s", &n); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + if (isempty(n)) + c->pam_name = mfree(c->pam_name); + else if (free_and_strdup(&c->pam_name, n) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "PAMName=%s", strempty(n)); + } + + return 1; + + } else if (streq(name, "Environment")) { + + _cleanup_strv_free_ char **l = NULL; + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + if (!strv_env_is_valid(l)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid environment block."); + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *joined = NULL; + char **e; + + if (strv_length(l) == 0) { + c->environment = strv_free(c->environment); + unit_write_drop_in_private_format(u, mode, name, "Environment="); + } else { + e = strv_env_merge(2, c->environment, l); + if (!e) + return -ENOMEM; + + strv_free(c->environment); + c->environment = e; + + joined = strv_join_quoted(c->environment); + if (!joined) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "Environment=%s", joined); + } + } + + return 1; + + } else if (streq(name, "TimerSlackNSec")) { + + nsec_t n; + + r = sd_bus_message_read(message, "t", &n); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->timer_slack_nsec = n; + unit_write_drop_in_private_format(u, mode, name, "TimerSlackNSec=" NSEC_FMT, n); + } + + return 1; + + } else if (streq(name, "OOMScoreAdjust")) { + int oa; + + r = sd_bus_message_read(message, "i", &oa); + if (r < 0) + return r; + + if (!oom_score_adjust_is_valid(oa)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "OOM score adjust value out of range"); + + if (mode != UNIT_CHECK) { + c->oom_score_adjust = oa; + c->oom_score_adjust_set = true; + unit_write_drop_in_private_format(u, mode, name, "OOMScoreAdjust=%i", oa); + } + + return 1; + + } else if (streq(name, "EnvironmentFiles")) { + + _cleanup_free_ char *joined = NULL; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_free_ char **l = NULL; + size_t size = 0; + char **i; + + r = sd_bus_message_enter_container(message, 'a', "(sb)"); + if (r < 0) + return r; + + f = open_memstream(&joined, &size); + if (!f) + return -ENOMEM; + + STRV_FOREACH(i, c->environment_files) + fprintf(f, "EnvironmentFile=%s", *i); + + while ((r = sd_bus_message_enter_container(message, 'r', "sb")) > 0) { + const char *path; + int b; + + r = sd_bus_message_read(message, "sb", &path, &b); + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (!isempty(path) && !path_is_absolute(path)) + return sd_bus_error_set_errnof(error, EINVAL, "Path %s is not absolute.", path); + + if (mode != UNIT_CHECK) { + char *buf = NULL; + + buf = strjoin(b ? "-" : "", path, NULL); + if (!buf) + return -ENOMEM; + + fprintf(f, "EnvironmentFile=%s", buf); + + r = strv_consume(&l, buf); + if (r < 0) + return r; + } + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + r = fflush_and_check(f); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + if (strv_isempty(l)) { + c->environment_files = strv_free(c->environment_files); + unit_write_drop_in_private(u, mode, name, "EnvironmentFile="); + } else { + r = strv_extend_strv(&c->environment_files, l, true); + if (r < 0) + return r; + + unit_write_drop_in_private(u, mode, name, joined); + } + } + + return 1; + + } else if (streq(name, "PassEnvironment")) { + + _cleanup_strv_free_ char **l = NULL; + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + if (!strv_env_name_is_valid(l)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid PassEnvironment block."); + + if (mode != UNIT_CHECK) { + if (strv_isempty(l)) { + c->pass_environment = strv_free(c->pass_environment); + unit_write_drop_in_private_format(u, mode, name, "PassEnvironment="); + } else { + _cleanup_free_ char *joined = NULL; + + r = strv_extend_strv(&c->pass_environment, l, true); + if (r < 0) + return r; + + joined = strv_join_quoted(c->pass_environment); + if (!joined) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "PassEnvironment=%s", joined); + } + } + + return 1; + + } else if (STR_IN_SET(name, "ReadWriteDirectories", "ReadOnlyDirectories", "InaccessibleDirectories", + "ReadWritePaths", "ReadOnlyPaths", "InaccessiblePaths")) { + _cleanup_strv_free_ char **l = NULL; + char ***dirs; + char **p; + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + STRV_FOREACH(p, l) { + int offset; + if (!utf8_is_valid(*p)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid %s", name); + + offset = **p == '-'; + if (!path_is_absolute(*p + offset)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid %s", name); + } + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *joined = NULL; + + if (STR_IN_SET(name, "ReadWriteDirectories", "ReadWritePaths")) + dirs = &c->read_write_paths; + else if (STR_IN_SET(name, "ReadOnlyDirectories", "ReadOnlyPaths")) + dirs = &c->read_only_paths; + else /* "InaccessiblePaths" */ + dirs = &c->inaccessible_paths; + + if (strv_length(l) == 0) { + *dirs = strv_free(*dirs); + unit_write_drop_in_private_format(u, mode, name, "%s=", name); + } else { + r = strv_extend_strv(dirs, l, true); + + if (r < 0) + return -ENOMEM; + + joined = strv_join_quoted(*dirs); + if (!joined) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, joined); + } + + } + + return 1; + + } else if (streq(name, "ProtectSystem")) { + const char *s; + ProtectSystem ps; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + r = parse_boolean(s); + if (r > 0) + ps = PROTECT_SYSTEM_YES; + else if (r == 0) + ps = PROTECT_SYSTEM_NO; + else { + ps = protect_system_from_string(s); + if (ps < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Failed to parse protect system value"); + } + + if (mode != UNIT_CHECK) { + c->protect_system = ps; + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, s); + } + + return 1; + + } else if (streq(name, "ProtectHome")) { + const char *s; + ProtectHome ph; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + r = parse_boolean(s); + if (r > 0) + ph = PROTECT_HOME_YES; + else if (r == 0) + ph = PROTECT_HOME_NO; + else { + ph = protect_home_from_string(s); + if (ph < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Failed to parse protect home value"); + } + + if (mode != UNIT_CHECK) { + c->protect_home = ph; + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, s); + } + + return 1; + + } else if (streq(name, "RuntimeDirectory")) { + _cleanup_strv_free_ char **l = NULL; + char **p; + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + STRV_FOREACH(p, l) { + if (!filename_is_valid(*p)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Runtime directory is not valid %s", *p); + } + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *joined = NULL; + + if (strv_isempty(l)) { + c->runtime_directory = strv_free(c->runtime_directory); + unit_write_drop_in_private_format(u, mode, name, "%s=", name); + } else { + r = strv_extend_strv(&c->runtime_directory, l, true); + + if (r < 0) + return -ENOMEM; + + joined = strv_join_quoted(c->runtime_directory); + if (!joined) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, joined); + } + } + + return 1; + + } else if (streq(name, "SELinuxContext")) { + const char *s; + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + if (isempty(s)) + c->selinux_context = mfree(c->selinux_context); + else if (free_and_strdup(&c->selinux_context, s) < 0) + return -ENOMEM; + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, strempty(s)); + } + + return 1; + + } + + ri = rlimit_from_string(name); + if (ri < 0) { + soft = endswith(name, "Soft"); + if (soft) { + const char *n; + + n = strndupa(name, soft - name); + ri = rlimit_from_string(n); + if (ri >= 0) + name = n; + + } + } + + if (ri >= 0) { + uint64_t rl; + rlim_t x; + + r = sd_bus_message_read(message, "t", &rl); + if (r < 0) + return r; + + if (rl == (uint64_t) -1) + x = RLIM_INFINITY; + else { + x = (rlim_t) rl; + + if ((uint64_t) x != rl) + return -ERANGE; + } + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *f = NULL; + struct rlimit nl; + + if (c->rlimit[ri]) { + nl = *c->rlimit[ri]; + + if (soft) + nl.rlim_cur = x; + else + nl.rlim_max = x; + } else + /* When the resource limit is not initialized yet, then assign the value to both fields */ + nl = (struct rlimit) { + .rlim_cur = x, + .rlim_max = x, + }; + + r = rlimit_format(&nl, &f); + if (r < 0) + return r; + + if (c->rlimit[ri]) + *c->rlimit[ri] = nl; + else { + c->rlimit[ri] = newdup(struct rlimit, &nl, 1); + if (!c->rlimit[ri]) + return -ENOMEM; + } + + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, f); + } + + return 1; + } + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-execute.h b/src/grp-system/libcore/src/dbus-execute.h new file mode 100644 index 0000000000..25137d9516 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-execute.h @@ -0,0 +1,45 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/execute.h" + +#define BUS_EXEC_STATUS_VTABLE(prefix, offset, flags) \ + BUS_PROPERTY_DUAL_TIMESTAMP(prefix "StartTimestamp", (offset) + offsetof(ExecStatus, start_timestamp), flags), \ + BUS_PROPERTY_DUAL_TIMESTAMP(prefix "ExitTimestamp", (offset) + offsetof(ExecStatus, exit_timestamp), flags), \ + SD_BUS_PROPERTY(prefix "PID", "u", bus_property_get_pid, (offset) + offsetof(ExecStatus, pid), flags), \ + SD_BUS_PROPERTY(prefix "Code", "i", bus_property_get_int, (offset) + offsetof(ExecStatus, code), flags), \ + SD_BUS_PROPERTY(prefix "Status", "i", bus_property_get_int, (offset) + offsetof(ExecStatus, status), flags) + +#define BUS_EXEC_COMMAND_VTABLE(name, offset, flags) \ + SD_BUS_PROPERTY(name, "a(sasbttttuii)", bus_property_get_exec_command, offset, flags) + +#define BUS_EXEC_COMMAND_LIST_VTABLE(name, offset, flags) \ + SD_BUS_PROPERTY(name, "a(sasbttttuii)", bus_property_get_exec_command_list, offset, flags) + +extern const sd_bus_vtable bus_exec_vtable[]; + +int bus_property_get_exec_output(sd_bus *bus, const char *path, const char *interface, const char *property, sd_bus_message *reply, void *userdata, sd_bus_error *ret_error); +int bus_property_get_exec_command(sd_bus *bus, const char *path, const char *interface, const char *property, sd_bus_message *reply, void *userdata, sd_bus_error *ret_error); +int bus_property_get_exec_command_list(sd_bus *bus, const char *path, const char *interface, const char *property, sd_bus_message *reply, void *userdata, sd_bus_error *ret_error); + +int bus_exec_context_set_transient_property(Unit *u, ExecContext *c, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); diff --git a/src/grp-system/libcore/src/dbus-job.c b/src/grp-system/libcore/src/dbus-job.c new file mode 100644 index 0000000000..0eddfe60d7 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-job.c @@ -0,0 +1,194 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/job.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/string-util.h" + +#include "dbus-job.h" +#include "dbus.h" +#include "selinux-access.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_type, job_type, JobType); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_state, job_state, JobState); + +static int property_get_unit( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + _cleanup_free_ char *p = NULL; + Job *j = userdata; + + assert(bus); + assert(reply); + assert(j); + + p = unit_dbus_path(j->unit); + if (!p) + return -ENOMEM; + + return sd_bus_message_append(reply, "(so)", j->unit->id, p); +} + +int bus_job_method_cancel(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Job *j = userdata; + int r; + + assert(message); + assert(j); + + r = mac_selinux_unit_access_check(j->unit, message, "stop", error); + if (r < 0) + return r; + + /* Access is granted to the job owner */ + if (!sd_bus_track_contains(j->clients, sd_bus_message_get_sender(message))) { + + /* And for everybody else consult PolicyKit */ + r = bus_verify_manage_units_async(j->unit->manager, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + } + + job_finish_and_invalidate(j, JOB_CANCELED, true, false); + + return sd_bus_reply_method_return(message, NULL); +} + +const sd_bus_vtable bus_job_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_METHOD("Cancel", NULL, NULL, bus_job_method_cancel, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_PROPERTY("Id", "u", NULL, offsetof(Job, id), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Unit", "(so)", property_get_unit, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("JobType", "s", property_get_type, offsetof(Job, type), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("State", "s", property_get_state, offsetof(Job, state), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_VTABLE_END +}; + +static int send_new_signal(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_free_ char *p = NULL; + Job *j = userdata; + int r; + + assert(bus); + assert(j); + + p = job_dbus_path(j); + if (!p) + return -ENOMEM; + + r = sd_bus_message_new_signal( + bus, + &m, + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "JobNew"); + if (r < 0) + return r; + + r = sd_bus_message_append(m, "uos", j->id, p, j->unit->id); + if (r < 0) + return r; + + return sd_bus_send(bus, m, NULL); +} + +static int send_changed_signal(sd_bus *bus, void *userdata) { + _cleanup_free_ char *p = NULL; + Job *j = userdata; + + assert(bus); + assert(j); + + p = job_dbus_path(j); + if (!p) + return -ENOMEM; + + return sd_bus_emit_properties_changed(bus, p, "org.freedesktop.systemd1.Job", "State", NULL); +} + +void bus_job_send_change_signal(Job *j) { + int r; + + assert(j); + + if (j->in_dbus_queue) { + LIST_REMOVE(dbus_queue, j->manager->dbus_job_queue, j); + j->in_dbus_queue = false; + } + + r = bus_foreach_bus(j->manager, j->clients, j->sent_dbus_new_signal ? send_changed_signal : send_new_signal, j); + if (r < 0) + log_debug_errno(r, "Failed to send job change signal for %u: %m", j->id); + + j->sent_dbus_new_signal = true; +} + +static int send_removed_signal(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_free_ char *p = NULL; + Job *j = userdata; + int r; + + assert(bus); + assert(j); + + p = job_dbus_path(j); + if (!p) + return -ENOMEM; + + r = sd_bus_message_new_signal( + bus, + &m, + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "JobRemoved"); + if (r < 0) + return r; + + r = sd_bus_message_append(m, "uoss", j->id, p, j->unit->id, job_result_to_string(j->result)); + if (r < 0) + return r; + + return sd_bus_send(bus, m, NULL); +} + +void bus_job_send_removed_signal(Job *j) { + int r; + + assert(j); + + if (!j->sent_dbus_new_signal) + bus_job_send_change_signal(j); + + r = bus_foreach_bus(j->manager, j->clients, send_removed_signal, j); + if (r < 0) + log_debug_errno(r, "Failed to send job remove signal for %u: %m", j->id); +} diff --git a/src/grp-system/libcore/src/dbus-job.h b/src/grp-system/libcore/src/dbus-job.h new file mode 100644 index 0000000000..4e2d4f086d --- /dev/null +++ b/src/grp-system/libcore/src/dbus-job.h @@ -0,0 +1,31 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/job.h" + +extern const sd_bus_vtable bus_job_vtable[]; + +int bus_job_method_cancel(sd_bus_message *message, void *job, sd_bus_error *error); + +void bus_job_send_change_signal(Job *j); +void bus_job_send_removed_signal(Job *j); diff --git a/src/grp-system/libcore/src/dbus-kill.c b/src/grp-system/libcore/src/dbus-kill.c new file mode 100644 index 0000000000..71337306a6 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-kill.c @@ -0,0 +1,123 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/kill.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/signal-util.h" + +#include "dbus-kill.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_kill_mode, kill_mode, KillMode); + +const sd_bus_vtable bus_kill_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("KillMode", "s", property_get_kill_mode, offsetof(KillContext, kill_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("KillSignal", "i", bus_property_get_int, offsetof(KillContext, kill_signal), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SendSIGKILL", "b", bus_property_get_bool, offsetof(KillContext, send_sigkill), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SendSIGHUP", "b", bus_property_get_bool, offsetof(KillContext, send_sighup), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_VTABLE_END +}; + +int bus_kill_context_set_transient_property( + Unit *u, + KillContext *c, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(u); + assert(c); + assert(name); + assert(message); + + if (streq(name, "KillMode")) { + const char *m; + KillMode k; + + r = sd_bus_message_read(message, "s", &m); + if (r < 0) + return r; + + k = kill_mode_from_string(m); + if (k < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Kill mode '%s' not known.", m); + + if (mode != UNIT_CHECK) { + c->kill_mode = k; + + unit_write_drop_in_private_format(u, mode, name, "KillMode=%s", kill_mode_to_string(k)); + } + + return 1; + + } else if (streq(name, "KillSignal")) { + int sig; + + r = sd_bus_message_read(message, "i", &sig); + if (r < 0) + return r; + + if (!SIGNAL_VALID(sig)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Signal %i out of range", sig); + + if (mode != UNIT_CHECK) { + c->kill_signal = sig; + + unit_write_drop_in_private_format(u, mode, name, "KillSignal=%s", signal_to_string(sig)); + } + + return 1; + + } else if (streq(name, "SendSIGHUP")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->send_sighup = b; + + unit_write_drop_in_private_format(u, mode, name, "SendSIGHUP=%s", yes_no(b)); + } + + return 1; + + } else if (streq(name, "SendSIGKILL")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + c->send_sigkill = b; + + unit_write_drop_in_private_format(u, mode, name, "SendSIGKILL=%s", yes_no(b)); + } + + return 1; + + } + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-kill.h b/src/grp-system/libcore/src/dbus-kill.h new file mode 100644 index 0000000000..3f908c40d9 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-kill.h @@ -0,0 +1,29 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/kill.h" +#include "core/unit.h" + +extern const sd_bus_vtable bus_kill_vtable[]; + +int bus_kill_context_set_transient_property(Unit *u, KillContext *c, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); diff --git a/src/grp-system/libcore/src/dbus-manager.c b/src/grp-system/libcore/src/dbus-manager.c new file mode 100644 index 0000000000..b704dd02cb --- /dev/null +++ b/src/grp-system/libcore/src/dbus-manager.c @@ -0,0 +1,2542 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/prctl.h> +#include <unistd.h> + +#include "core/dbus-manager.h" +#include "sd-bus/bus-common-errors.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/architecture.h" +#include "systemd-basic/build.h" +#include "systemd-basic/clock-util.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/syslog-util.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/virt.h" +#include "systemd-shared/install.h" +#include "systemd-shared/watchdog.h" + +#include "dbus-execute.h" +#include "dbus-job.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "selinux-access.h" + +static UnitFileFlags unit_file_bools_to_flags(bool runtime, bool force) { + return (runtime ? UNIT_FILE_RUNTIME : 0) | + (force ? UNIT_FILE_FORCE : 0); +} + +static int property_get_version( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "s", PACKAGE_VERSION); +} + +static int property_get_features( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "s", SYSTEMD_FEATURES); +} + +static int property_get_virtualization( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + int v; + + assert(bus); + assert(reply); + + v = detect_virtualization(); + + /* Make sure to return the empty string when we detect no virtualization, as that is the API. + * + * https://github.com/systemd/systemd/issues/1423 + */ + + return sd_bus_message_append( + reply, "s", + v == VIRTUALIZATION_NONE ? "" : virtualization_to_string(v)); +} + +static int property_get_architecture( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "s", architecture_to_string(uname_architecture())); +} + +static int property_get_tainted( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + char buf[sizeof("split-usr:cgroups-missing:local-hwclock:")] = "", *e = buf; + Manager *m = userdata; + + assert(bus); + assert(reply); + assert(m); + + if (m->taint_usr) + e = stpcpy(e, "split-usr:"); + + if (access("/proc/cgroups", F_OK) < 0) + e = stpcpy(e, "cgroups-missing:"); + + if (clock_is_localtime(NULL) > 0) + e = stpcpy(e, "local-hwclock:"); + + /* remove the last ':' */ + if (e != buf) + e[-1] = 0; + + return sd_bus_message_append(reply, "s", buf); +} + +static int property_get_log_target( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "s", log_target_to_string(log_get_target())); +} + +static int property_set_log_target( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *value, + void *userdata, + sd_bus_error *error) { + + const char *t; + int r; + + assert(bus); + assert(value); + + r = sd_bus_message_read(value, "s", &t); + if (r < 0) + return r; + + return log_set_target_from_string(t); +} + +static int property_get_log_level( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + _cleanup_free_ char *t = NULL; + int r; + + assert(bus); + assert(reply); + + r = log_level_to_string_alloc(log_get_max_level(), &t); + if (r < 0) + return r; + + return sd_bus_message_append(reply, "s", t); +} + +static int property_set_log_level( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *value, + void *userdata, + sd_bus_error *error) { + + const char *t; + int r; + + assert(bus); + assert(value); + + r = sd_bus_message_read(value, "s", &t); + if (r < 0) + return r; + + r = log_set_max_level_from_string(t); + if (r == 0) + log_info("Setting log level to %s.", t); + return r; +} + +static int property_get_n_names( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Manager *m = userdata; + + assert(bus); + assert(reply); + assert(m); + + return sd_bus_message_append(reply, "u", (uint32_t) hashmap_size(m->units)); +} + +static int property_get_n_failed_units( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Manager *m = userdata; + + assert(bus); + assert(reply); + assert(m); + + return sd_bus_message_append(reply, "u", (uint32_t) set_size(m->failed_units)); +} + +static int property_get_n_jobs( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Manager *m = userdata; + + assert(bus); + assert(reply); + assert(m); + + return sd_bus_message_append(reply, "u", (uint32_t) hashmap_size(m->jobs)); +} + +static int property_get_progress( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Manager *m = userdata; + double d; + + assert(bus); + assert(reply); + assert(m); + + if (dual_timestamp_is_set(&m->finish_timestamp)) + d = 1.0; + else + d = 1.0 - ((double) hashmap_size(m->jobs) / (double) m->n_installed_jobs); + + return sd_bus_message_append(reply, "d", d); +} + +static int property_get_system_state( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Manager *m = userdata; + + assert(bus); + assert(reply); + assert(m); + + return sd_bus_message_append(reply, "s", manager_state_to_string(manager_state(m))); +} + +static int property_set_runtime_watchdog( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *value, + void *userdata, + sd_bus_error *error) { + + usec_t *t = userdata; + int r; + + assert(bus); + assert(value); + + assert_cc(sizeof(usec_t) == sizeof(uint64_t)); + + r = sd_bus_message_read(value, "t", t); + if (r < 0) + return r; + + return watchdog_set_timeout(t); +} + +static int property_get_timer_slack_nsec( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + return sd_bus_message_append(reply, "t", (uint64_t) prctl(PR_GET_TIMERSLACK)); +} + +static int method_get_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *path = NULL; + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + if (isempty(name)) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + pid_t pid; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return r; + + u = manager_get_unit_by_pid(m, pid); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Client not member of any unit."); + } else { + u = manager_get_unit(m, name); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s not loaded.", name); + } + + r = mac_selinux_unit_access_check(u, message, "status", error); + if (r < 0) + return r; + + path = unit_dbus_path(u); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int method_get_unit_by_pid(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *path = NULL; + Manager *m = userdata; + pid_t pid; + Unit *u; + int r; + + assert(message); + assert(m); + + assert_cc(sizeof(pid_t) == sizeof(uint32_t)); + + /* Anyone can call this method */ + + r = sd_bus_message_read(message, "u", &pid); + if (r < 0) + return r; + if (pid < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid PID " PID_FMT, pid); + + if (pid == 0) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return r; + } + + u = manager_get_unit_by_pid(m, pid); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_UNIT_FOR_PID, "PID "PID_FMT" does not belong to any loaded unit.", pid); + + r = mac_selinux_unit_access_check(u, message, "status", error); + if (r < 0) + return r; + + path = unit_dbus_path(u); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int method_get_unit_by_invocation_id(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *path = NULL; + Manager *m = userdata; + sd_id128_t id; + const void *a; + Unit *u; + size_t sz; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = sd_bus_message_read_array(message, 'y', &a, &sz); + if (r < 0) + return r; + if (sz == 0) + id = SD_ID128_NULL; + else if (sz == 16) + memcpy(&id, a, sz); + else + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid invocation ID"); + + if (sd_id128_is_null(id)) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + pid_t pid; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return r; + + u = manager_get_unit_by_pid(m, pid); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Client " PID_FMT " not member of any unit.", pid); + } else { + u = hashmap_get(m->units_by_invocation_id, &id); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_UNIT_FOR_INVOCATION_ID, "No unit with the specified invocation ID " SD_ID128_FORMAT_STR " known.", SD_ID128_FORMAT_VAL(id)); + } + + r = mac_selinux_unit_access_check(u, message, "status", error); + if (r < 0) + return r; + + /* So here's a special trick: the bus path we return actually references the unit by its invocation ID instead + * of the unit name. This means it stays valid only as long as the invocation ID stays the same. */ + path = unit_dbus_path_invocation_id(u); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int method_load_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *path = NULL; + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + if (isempty(name)) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + pid_t pid; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return r; + + u = manager_get_unit_by_pid(m, pid); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Client not member of any unit."); + } else { + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + } + + r = mac_selinux_unit_access_check(u, message, "status", error); + if (r < 0) + return r; + + path = unit_dbus_path(u); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int method_start_unit_generic(sd_bus_message *message, Manager *m, JobType job_type, bool reload_if_possible, sd_bus_error *error) { + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + return bus_unit_method_start_generic(message, u, job_type, reload_if_possible, error); +} + +static int method_start_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_START, false, error); +} + +static int method_stop_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_STOP, false, error); +} + +static int method_reload_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_RELOAD, false, error); +} + +static int method_restart_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_RESTART, false, error); +} + +static int method_try_restart_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_TRY_RESTART, false, error); +} + +static int method_reload_or_restart_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_RESTART, true, error); +} + +static int method_reload_or_try_restart_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_start_unit_generic(message, userdata, JOB_TRY_RESTART, true, error); +} + +static int method_start_unit_replace(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *old_name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &old_name); + if (r < 0) + return r; + + u = manager_get_unit(m, old_name); + if (!u || !u->job || u->job->type != JOB_START) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_JOB, "No job queued for unit %s", old_name); + + return method_start_unit_generic(message, m, JOB_START, false, error); +} + +static int method_kill_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + u = manager_get_unit(m, name); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + + return bus_unit_method_kill(message, u, error); +} + +static int method_reset_failed_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + u = manager_get_unit(m, name); + if (!u) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not loaded.", name); + + return bus_unit_method_reset_failed(message, u, error); +} + +static int method_set_unit_properties(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + r = bus_unit_check_load_state(u, error); + if (r < 0) + return r; + + return bus_unit_method_set_properties(message, u, error); +} + +static int method_ref_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + r = bus_unit_check_load_state(u, error); + if (r < 0) + return r; + + return bus_unit_method_ref(message, u, error); +} + +static int method_unref_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + r = bus_unit_check_load_state(u, error); + if (r < 0) + return r; + + return bus_unit_method_unref(message, u, error); +} + +static int reply_unit_info(sd_bus_message *reply, Unit *u) { + _cleanup_free_ char *unit_path = NULL, *job_path = NULL; + Unit *following; + + following = unit_following(u); + + unit_path = unit_dbus_path(u); + if (!unit_path) + return -ENOMEM; + + if (u->job) { + job_path = job_dbus_path(u->job); + if (!job_path) + return -ENOMEM; + } + + return sd_bus_message_append( + reply, "(ssssssouso)", + u->id, + unit_description(u), + unit_load_state_to_string(u->load_state), + unit_active_state_to_string(unit_active_state(u)), + unit_sub_state_to_string(u), + following ? following->id : "", + unit_path, + u->job ? u->job->id : 0, + u->job ? job_type_to_string(u->job->type) : "", + job_path ? job_path : "/"); +} + +static int method_list_units_by_names(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + Manager *m = userdata; + int r; + char **unit; + _cleanup_strv_free_ char **units = NULL; + + assert(message); + assert(m); + + r = sd_bus_message_read_strv(message, &units); + if (r < 0) + return r; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, 'a', "(ssssssouso)"); + if (r < 0) + return r; + + STRV_FOREACH(unit, units) { + Unit *u; + + if (!unit_name_is_valid(*unit, UNIT_NAME_ANY)) + continue; + + r = manager_load_unit(m, *unit, NULL, error, &u); + if (r < 0) + return r; + + r = reply_unit_info(reply, u); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); +} + +static int method_get_unit_processes(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + r = bus_unit_check_load_state(u, error); + if (r < 0) + return r; + + return bus_unit_method_get_processes(message, u, error); +} + +static int transient_unit_from_message( + Manager *m, + sd_bus_message *message, + const char *name, + Unit **unit, + sd_bus_error *error) { + + UnitType t; + Unit *u; + int r; + + assert(m); + assert(message); + assert(name); + + t = unit_name_to_type(name); + if (t < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid unit name or type."); + + if (!unit_vtable[t]->can_transient) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Unit type %s does not support transient units.", unit_type_to_string(t)); + + r = manager_load_unit(m, name, NULL, error, &u); + if (r < 0) + return r; + + if (!unit_is_pristine(u)) + return sd_bus_error_setf(error, BUS_ERROR_UNIT_EXISTS, "Unit %s already exists.", name); + + /* OK, the unit failed to load and is unreferenced, now let's + * fill in the transient data instead */ + r = unit_make_transient(u); + if (r < 0) + return r; + + /* Set our properties */ + r = bus_unit_set_properties(u, message, UNIT_RUNTIME, false, error); + if (r < 0) + return r; + + /* If the client asked for it, automatically add a reference to this unit. */ + if (u->bus_track_add) { + r = bus_unit_track_add_sender(u, message); + if (r < 0) + return log_error_errno(r, "Failed to watch sender: %m"); + } + + /* Now load the missing bits of the unit we just created */ + unit_add_to_load_queue(u); + manager_dispatch_load_queue(m); + + *unit = u; + + return 0; +} + +static int transient_aux_units_from_message( + Manager *m, + sd_bus_message *message, + sd_bus_error *error) { + + int r; + + assert(m); + assert(message); + + r = sd_bus_message_enter_container(message, 'a', "(sa(sv))"); + if (r < 0) + return r; + + while ((r = sd_bus_message_enter_container(message, 'r', "sa(sv)")) > 0) { + const char *name = NULL; + Unit *u; + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = transient_unit_from_message(m, message, name, &u, error); + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + return 0; +} + +static int method_start_transient_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + const char *name, *smode; + Manager *m = userdata; + JobMode mode; + Unit *u; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "start", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "ss", &name, &smode); + if (r < 0) + return r; + + mode = job_mode_from_string(smode); + if (mode < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Job mode %s is invalid.", smode); + + r = bus_verify_manage_units_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = transient_unit_from_message(m, message, name, &u, error); + if (r < 0) + return r; + + r = transient_aux_units_from_message(m, message, error); + if (r < 0) + return r; + + /* Finally, start it */ + return bus_unit_queue_job(message, u, JOB_START, mode, false, error); +} + +static int method_get_job(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *path = NULL; + Manager *m = userdata; + uint32_t id; + Job *j; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = sd_bus_message_read(message, "u", &id); + if (r < 0) + return r; + + j = manager_get_job(m, id); + if (!j) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_JOB, "Job %u does not exist.", (unsigned) id); + + r = mac_selinux_unit_access_check(j->unit, message, "status", error); + if (r < 0) + return r; + + path = job_dbus_path(j); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int method_cancel_job(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + uint32_t id; + Job *j; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "u", &id); + if (r < 0) + return r; + + j = manager_get_job(m, id); + if (!j) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_JOB, "Job %u does not exist.", (unsigned) id); + + return bus_job_method_cancel(message, j, error); +} + +static int method_clear_jobs(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = bus_verify_manage_units_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + manager_clear_jobs(m); + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_reset_failed(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = bus_verify_manage_units_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + manager_reset_failed(m); + + return sd_bus_reply_method_return(message, NULL); +} + +static int list_units_filtered(sd_bus_message *message, void *userdata, sd_bus_error *error, char **states, char **patterns) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + Manager *m = userdata; + const char *k; + Iterator i; + Unit *u; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, 'a', "(ssssssouso)"); + if (r < 0) + return r; + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + if (k != u->id) + continue; + + if (!strv_isempty(states) && + !strv_contains(states, unit_load_state_to_string(u->load_state)) && + !strv_contains(states, unit_active_state_to_string(unit_active_state(u))) && + !strv_contains(states, unit_sub_state_to_string(u))) + continue; + + if (!strv_isempty(patterns) && + !strv_fnmatch_or_empty(patterns, u->id, FNM_NOESCAPE)) + continue; + + r = reply_unit_info(reply, u); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); +} + +static int method_list_units(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return list_units_filtered(message, userdata, error, NULL, NULL); +} + +static int method_list_units_filtered(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **states = NULL; + int r; + + r = sd_bus_message_read_strv(message, &states); + if (r < 0) + return r; + + return list_units_filtered(message, userdata, error, states, NULL); +} + +static int method_list_units_by_patterns(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **states = NULL; + _cleanup_strv_free_ char **patterns = NULL; + int r; + + r = sd_bus_message_read_strv(message, &states); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &patterns); + if (r < 0) + return r; + + return list_units_filtered(message, userdata, error, states, patterns); +} + +static int method_list_jobs(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + Manager *m = userdata; + Iterator i; + Job *j; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, 'a', "(usssoo)"); + if (r < 0) + return r; + + HASHMAP_FOREACH(j, m->jobs, i) { + _cleanup_free_ char *unit_path = NULL, *job_path = NULL; + + job_path = job_dbus_path(j); + if (!job_path) + return -ENOMEM; + + unit_path = unit_dbus_path(j->unit); + if (!unit_path) + return -ENOMEM; + + r = sd_bus_message_append( + reply, "(usssoo)", + j->id, + j->unit->id, + job_type_to_string(j->type), + job_state_to_string(j->state), + job_path, + unit_path); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); +} + +static int method_subscribe(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + if (sd_bus_message_get_bus(message) == m->api_bus) { + + /* Note that direct bus connection subscribe by + * default, we only track peers on the API bus here */ + + if (!m->subscribed) { + r = sd_bus_track_new(sd_bus_message_get_bus(message), &m->subscribed, NULL, NULL); + if (r < 0) + return r; + } + + r = sd_bus_track_add_sender(m->subscribed, message); + if (r < 0) + return r; + if (r == 0) + return sd_bus_error_setf(error, BUS_ERROR_ALREADY_SUBSCRIBED, "Client is already subscribed."); + } + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_unsubscribe(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + if (sd_bus_message_get_bus(message) == m->api_bus) { + r = sd_bus_track_remove_sender(m->subscribed, message); + if (r < 0) + return r; + if (r == 0) + return sd_bus_error_setf(error, BUS_ERROR_NOT_SUBSCRIBED, "Client is not subscribed."); + } + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_dump(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *dump = NULL; + _cleanup_fclose_ FILE *f = NULL; + Manager *m = userdata; + size_t size; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + f = open_memstream(&dump, &size); + if (!f) + return -ENOMEM; + + manager_dump_units(m, f, NULL); + manager_dump_jobs(m, f, NULL); + + r = fflush_and_check(f); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, "s", dump); +} + +static int method_refuse_snapshot(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Support for snapshots has been removed."); +} + +static int method_reload(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = bus_verify_reload_daemon_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + /* Instead of sending the reply back right away, we just + * remember that we need to and then send it after the reload + * is finished. That way the caller knows when the reload + * finished. */ + + assert(!m->queued_message); + r = sd_bus_message_new_method_return(message, &m->queued_message); + if (r < 0) + return r; + + m->exit_code = MANAGER_RELOAD; + + return 1; +} + +static int method_reexecute(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = bus_verify_reload_daemon_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + /* We don't send a reply back here, the client should + * just wait for us disconnecting. */ + + m->exit_code = MANAGER_REEXECUTE; + return 1; +} + +static int method_exit(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "halt", error); + if (r < 0) + return r; + + /* Exit() (in contrast to SetExitCode()) is actually allowed even if + * we are running on the host. It will fall back on reboot() in + * systemd-shutdown if it cannot do the exit() because it isn't a + * container. */ + + m->exit_code = MANAGER_EXIT; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_reboot(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reboot", error); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Reboot is only supported for system managers."); + + m->exit_code = MANAGER_REBOOT; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_poweroff(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "halt", error); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Powering off is only supported for system managers."); + + m->exit_code = MANAGER_POWEROFF; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_halt(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "halt", error); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Halt is only supported for system managers."); + + m->exit_code = MANAGER_HALT; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_kexec(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reboot", error); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "KExec is only supported for system managers."); + + m->exit_code = MANAGER_KEXEC; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_switch_root(sd_bus_message *message, void *userdata, sd_bus_error *error) { + char *ri = NULL, *rt = NULL; + const char *root, *init; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reboot", error); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Root switching is only supported by system manager."); + + r = sd_bus_message_read(message, "ss", &root, &init); + if (r < 0) + return r; + + if (path_equal(root, "/") || !path_is_absolute(root)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid switch root path %s", root); + + /* Safety check */ + if (isempty(init)) { + if (!path_is_os_tree(root)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Specified switch root path %s does not seem to be an OS tree. os-release file is missing.", root); + } else { + _cleanup_free_ char *p = NULL; + + if (!path_is_absolute(init)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid init path %s", init); + + p = strappend(root, init); + if (!p) + return -ENOMEM; + + if (access(p, X_OK) < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Specified init binary %s does not exist.", p); + } + + rt = strdup(root); + if (!rt) + return -ENOMEM; + + if (!isempty(init)) { + ri = strdup(init); + if (!ri) { + free(rt); + return -ENOMEM; + } + } + + free(m->switch_root); + m->switch_root = rt; + + free(m->switch_root_init); + m->switch_root_init = ri; + + m->exit_code = MANAGER_SWITCH_ROOT; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_set_environment(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **plus = NULL; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &plus); + if (r < 0) + return r; + if (!strv_env_is_valid(plus)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid environment assignments"); + + r = bus_verify_set_environment_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = manager_environment_add(m, NULL, plus); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_unset_environment(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **minus = NULL; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &minus); + if (r < 0) + return r; + + if (!strv_env_name_or_assignment_is_valid(minus)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid environment variable names or assignments"); + + r = bus_verify_set_environment_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = manager_environment_add(m, minus, NULL); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_unset_and_set_environment(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **minus = NULL, **plus = NULL; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "reload", error); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &minus); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &plus); + if (r < 0) + return r; + + if (!strv_env_name_or_assignment_is_valid(minus)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid environment variable names or assignments"); + if (!strv_env_is_valid(plus)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid environment assignments"); + + r = bus_verify_set_environment_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = manager_environment_add(m, minus, plus); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_set_exit_code(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + uint8_t code; + int r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "exit", error); + if (r < 0) + return r; + + r = sd_bus_message_read_basic(message, 'y', &code); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(m) && detect_container() <= 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "ExitCode can only be set for user service managers or in containers."); + + m->return_value = code; + + return sd_bus_reply_method_return(message, NULL); +} + +static int method_lookup_dynamic_user_by_name(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + uid_t uid; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read_basic(message, 's', &name); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Dynamic users are only supported in the system instance."); + if (!valid_user_group_name(name)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "User name invalid: %s", name); + + r = dynamic_user_lookup_name(m, name, &uid); + if (r == -ESRCH) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_DYNAMIC_USER, "Dynamic user %s does not exist.", name); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, "u", (uint32_t) uid); +} + +static int method_lookup_dynamic_user_by_uid(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *name = NULL; + Manager *m = userdata; + uid_t uid; + int r; + + assert(message); + assert(m); + + assert_cc(sizeof(uid) == sizeof(uint32_t)); + r = sd_bus_message_read_basic(message, 'u', &uid); + if (r < 0) + return r; + + if (!MANAGER_IS_SYSTEM(m)) + return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Dynamic users are only supported in the system instance."); + if (!uid_is_valid(uid)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "User ID invalid: " UID_FMT, uid); + + r = dynamic_user_lookup_uid(m, uid, &name); + if (r == -ESRCH) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_DYNAMIC_USER, "Dynamic user ID " UID_FMT " does not exist.", uid); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, "s", name); +} + +static int list_unit_files_by_patterns(sd_bus_message *message, void *userdata, sd_bus_error *error, char **states, char **patterns) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + Manager *m = userdata; + UnitFileList *item; + Hashmap *h; + Iterator i; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + h = hashmap_new(&string_hash_ops); + if (!h) + return -ENOMEM; + + r = unit_file_get_list(m->unit_file_scope, NULL, h, states, patterns); + if (r < 0) + goto fail; + + r = sd_bus_message_open_container(reply, 'a', "(ss)"); + if (r < 0) + goto fail; + + HASHMAP_FOREACH(item, h, i) { + + r = sd_bus_message_append(reply, "(ss)", item->path, unit_file_state_to_string(item->state)); + if (r < 0) + goto fail; + } + + unit_file_list_free(h); + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); + +fail: + unit_file_list_free(h); + return r; +} + +static int method_list_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return list_unit_files_by_patterns(message, userdata, error, NULL, NULL); +} + +static int method_list_unit_files_by_patterns(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **states = NULL; + _cleanup_strv_free_ char **patterns = NULL; + int r; + + r = sd_bus_message_read_strv(message, &states); + if (r < 0) + return r; + + r = sd_bus_message_read_strv(message, &patterns); + if (r < 0) + return r; + + return list_unit_files_by_patterns(message, userdata, error, states, patterns); +} + +static int method_get_unit_file_state(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *name; + UnitFileState state; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + r = unit_file_get_state(m->unit_file_scope, NULL, name, &state); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, "s", unit_file_state_to_string(state)); +} + +static int method_get_default_target(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_free_ char *default_target = NULL; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + /* Anyone can call this method */ + + r = mac_selinux_access_check(message, "status", error); + if (r < 0) + return r; + + r = unit_file_get_default(m->unit_file_scope, NULL, &default_target); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, "s", default_target); +} + +static int send_unit_files_changed(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; + int r; + + assert(bus); + + r = sd_bus_message_new_signal(bus, &message, "/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "UnitFilesChanged"); + if (r < 0) + return r; + + return sd_bus_send(bus, message, NULL); +} + +static int reply_unit_file_changes_and_free( + Manager *m, + sd_bus_message *message, + int carries_install_info, + UnitFileChange *changes, + unsigned n_changes) { + + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + unsigned i; + int r; + + if (unit_file_changes_have_modification(changes, n_changes)) { + r = bus_foreach_bus(m, NULL, send_unit_files_changed, NULL); + if (r < 0) + log_debug_errno(r, "Failed to send UnitFilesChanged signal: %m"); + } + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + goto fail; + + if (carries_install_info >= 0) { + r = sd_bus_message_append(reply, "b", carries_install_info); + if (r < 0) + goto fail; + } + + r = sd_bus_message_open_container(reply, 'a', "(sss)"); + if (r < 0) + goto fail; + + for (i = 0; i < n_changes; i++) + if (changes[i].type >= 0) { + const char *change = unit_file_change_type_to_string(changes[i].type); + assert(change != NULL); + + r = sd_bus_message_append( + reply, "(sss)", + change, + changes[i].path, + changes[i].source); + if (r < 0) + goto fail; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + goto fail; + + unit_file_changes_free(changes, n_changes); + return sd_bus_send(NULL, reply, NULL); + +fail: + unit_file_changes_free(changes, n_changes); + return r; +} + +/* Create an error reply, using the error information from changes[] + * if possible, and fall back to generating an error from error code c. + * The error message only describes the first error. + * + * Coordinate with unit_file_dump_changes() in install.c. + */ +static int install_error( + sd_bus_error *error, + int c, + UnitFileChange *changes, + unsigned n_changes) { + int r; + unsigned i; + assert(c < 0); + + for (i = 0; i < n_changes; i++) + switch(changes[i].type) { + case 0 ... INT_MAX: + continue; + case -EEXIST: + if (changes[i].source) + r = sd_bus_error_setf(error, BUS_ERROR_UNIT_EXISTS, + "File %s already exists and is a symlink to %s.", + changes[i].path, changes[i].source); + else + r = sd_bus_error_setf(error, BUS_ERROR_UNIT_EXISTS, + "File %s already exists.", + changes[i].path); + goto found; + case -ERFKILL: + r = sd_bus_error_setf(error, BUS_ERROR_UNIT_MASKED, + "Unit file %s is masked.", changes[i].path); + goto found; + case -EADDRNOTAVAIL: + r = sd_bus_error_setf(error, BUS_ERROR_UNIT_GENERATED, + "Unit %s is transient or generated.", changes[i].path); + goto found; + case -ELOOP: + r = sd_bus_error_setf(error, BUS_ERROR_UNIT_LINKED, + "Refusing to operate on linked unit file %s", changes[i].path); + goto found; + default: + r = sd_bus_error_set_errnof(error, changes[i].type, "File %s: %m", changes[i].path); + goto found; + } + + r = c; + found: + unit_file_changes_free(changes, n_changes); + return r; +} + +static int method_enable_unit_files_generic( + sd_bus_message *message, + Manager *m, + int (*call)(UnitFileScope scope, UnitFileFlags flags, const char *root_dir, char *files[], UnitFileChange **changes, unsigned *n_changes), + bool carries_install_info, + sd_bus_error *error) { + + _cleanup_strv_free_ char **l = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + UnitFileFlags flags; + int runtime, force, r; + + assert(message); + assert(m); + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "bb", &runtime, &force); + if (r < 0) + return r; + + flags = unit_file_bools_to_flags(runtime, force); + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = call(m->unit_file_scope, flags, NULL, l, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, carries_install_info ? r : -1, changes, n_changes); +} + +static int method_enable_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_enable_unit_files_generic(message, userdata, unit_file_enable, true, error); +} + +static int method_reenable_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_enable_unit_files_generic(message, userdata, unit_file_reenable, true, error); +} + +static int method_link_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_enable_unit_files_generic(message, userdata, unit_file_link, false, error); +} + +static int unit_file_preset_without_mode(UnitFileScope scope, UnitFileFlags flags, const char *root_dir, char **files, UnitFileChange **changes, unsigned *n_changes) { + return unit_file_preset(scope, flags, root_dir, files, UNIT_FILE_PRESET_FULL, changes, n_changes); +} + +static int method_preset_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_enable_unit_files_generic(message, userdata, unit_file_preset_without_mode, true, error); +} + +static int method_mask_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_enable_unit_files_generic(message, userdata, unit_file_mask, false, error); +} + +static int method_preset_unit_files_with_mode(sd_bus_message *message, void *userdata, sd_bus_error *error) { + + _cleanup_strv_free_ char **l = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + Manager *m = userdata; + UnitFilePresetMode mm; + int runtime, force, r; + UnitFileFlags flags; + const char *mode; + + assert(message); + assert(m); + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "sbb", &mode, &runtime, &force); + if (r < 0) + return r; + + flags = unit_file_bools_to_flags(runtime, force); + + if (isempty(mode)) + mm = UNIT_FILE_PRESET_FULL; + else { + mm = unit_file_preset_mode_from_string(mode); + if (mm < 0) + return -EINVAL; + } + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = unit_file_preset(m->unit_file_scope, flags, NULL, l, mm, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, r, changes, n_changes); +} + +static int method_disable_unit_files_generic( + sd_bus_message *message, + Manager *m, + int (*call)(UnitFileScope scope, UnitFileFlags flags, const char *root_dir, char *files[], UnitFileChange **changes, unsigned *n_changes), + sd_bus_error *error) { + + _cleanup_strv_free_ char **l = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + int r, runtime; + + assert(message); + assert(m); + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "b", &runtime); + if (r < 0) + return r; + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = call(m->unit_file_scope, runtime ? UNIT_FILE_RUNTIME : 0, NULL, l, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); +} + +static int method_disable_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_disable_unit_files_generic(message, userdata, unit_file_disable, error); +} + +static int method_unmask_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return method_disable_unit_files_generic(message, userdata, unit_file_unmask, error); +} + +static int method_revert_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **l = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + Manager *m = userdata; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = unit_file_revert(m->unit_file_scope, NULL, l, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); +} + +static int method_set_default_target(sd_bus_message *message, void *userdata, sd_bus_error *error) { + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + Manager *m = userdata; + const char *name; + int force, r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "enable", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "sb", &name, &force); + if (r < 0) + return r; + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = unit_file_set_default(m->unit_file_scope, force ? UNIT_FILE_FORCE : 0, NULL, name, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); +} + +static int method_preset_all_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + Manager *m = userdata; + UnitFilePresetMode mm; + const char *mode; + UnitFileFlags flags; + int force, runtime, r; + + assert(message); + assert(m); + + r = mac_selinux_access_check(message, "enable", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "sbb", &mode, &runtime, &force); + if (r < 0) + return r; + + flags = unit_file_bools_to_flags(runtime, force); + + if (isempty(mode)) + mm = UNIT_FILE_PRESET_FULL; + else { + mm = unit_file_preset_mode_from_string(mode); + if (mm < 0) + return -EINVAL; + } + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = unit_file_preset_all(m->unit_file_scope, flags, NULL, mm, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); +} + +static int method_add_dependency_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_strv_free_ char **l = NULL; + Manager *m = userdata; + UnitFileChange *changes = NULL; + unsigned n_changes = 0; + int runtime, force, r; + char *target, *type; + UnitDependency dep; + UnitFileFlags flags; + + assert(message); + assert(m); + + r = bus_verify_manage_unit_files_async(m, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = sd_bus_message_read_strv(message, &l); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "ssbb", &target, &type, &runtime, &force); + if (r < 0) + return r; + + flags = unit_file_bools_to_flags(runtime, force); + + dep = unit_dependency_from_string(type); + if (dep < 0) + return -EINVAL; + + r = unit_file_add_dependency(m->unit_file_scope, flags, NULL, l, target, dep, &changes, &n_changes); + if (r < 0) + return install_error(error, r, changes, n_changes); + + return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); +} + +static int method_get_unit_file_links(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + UnitFileChange *changes = NULL; + unsigned n_changes = 0, i; + UnitFileFlags flags; + const char *name; + char **p; + int runtime, r; + + r = sd_bus_message_read(message, "sb", &name, &runtime); + if (r < 0) + return r; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, SD_BUS_TYPE_ARRAY, "s"); + if (r < 0) + return r; + + p = STRV_MAKE(name); + flags = UNIT_FILE_DRY_RUN | + (runtime ? UNIT_FILE_RUNTIME : 0); + + r = unit_file_disable(UNIT_FILE_SYSTEM, flags, NULL, p, &changes, &n_changes); + if (r < 0) + return log_error_errno(r, "Failed to get file links for %s: %m", name); + + for (i = 0; i < n_changes; i++) + if (changes[i].type == UNIT_FILE_UNLINK) { + r = sd_bus_message_append(reply, "s", changes[i].path); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); +} + +const sd_bus_vtable bus_manager_vtable[] = { + SD_BUS_VTABLE_START(0), + + SD_BUS_PROPERTY("Version", "s", property_get_version, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Features", "s", property_get_features, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Virtualization", "s", property_get_virtualization, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Architecture", "s", property_get_architecture, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Tainted", "s", property_get_tainted, 0, SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("FirmwareTimestamp", offsetof(Manager, firmware_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("LoaderTimestamp", offsetof(Manager, loader_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("KernelTimestamp", offsetof(Manager, kernel_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("InitRDTimestamp", offsetof(Manager, initrd_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("UserspaceTimestamp", offsetof(Manager, userspace_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("FinishTimestamp", offsetof(Manager, finish_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("SecurityStartTimestamp", offsetof(Manager, security_start_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("SecurityFinishTimestamp", offsetof(Manager, security_finish_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("GeneratorsStartTimestamp", offsetof(Manager, generators_start_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("GeneratorsFinishTimestamp", offsetof(Manager, generators_finish_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("UnitsLoadStartTimestamp", offsetof(Manager, units_load_start_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("UnitsLoadFinishTimestamp", offsetof(Manager, units_load_finish_timestamp), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_WRITABLE_PROPERTY("LogLevel", "s", property_get_log_level, property_set_log_level, 0, 0), + SD_BUS_WRITABLE_PROPERTY("LogTarget", "s", property_get_log_target, property_set_log_target, 0, 0), + SD_BUS_PROPERTY("NNames", "u", property_get_n_names, 0, 0), + SD_BUS_PROPERTY("NFailedUnits", "u", property_get_n_failed_units, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("NJobs", "u", property_get_n_jobs, 0, 0), + SD_BUS_PROPERTY("NInstalledJobs", "u", bus_property_get_unsigned, offsetof(Manager, n_installed_jobs), 0), + SD_BUS_PROPERTY("NFailedJobs", "u", bus_property_get_unsigned, offsetof(Manager, n_failed_jobs), 0), + SD_BUS_PROPERTY("Progress", "d", property_get_progress, 0, 0), + SD_BUS_PROPERTY("Environment", "as", NULL, offsetof(Manager, environment), 0), + SD_BUS_PROPERTY("ConfirmSpawn", "b", bus_property_get_bool, offsetof(Manager, confirm_spawn), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ShowStatus", "b", bus_property_get_bool, offsetof(Manager, show_status), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UnitPath", "as", NULL, offsetof(Manager, lookup_paths.search_path), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultStandardOutput", "s", bus_property_get_exec_output, offsetof(Manager, default_std_output), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultStandardError", "s", bus_property_get_exec_output, offsetof(Manager, default_std_output), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_WRITABLE_PROPERTY("RuntimeWatchdogUSec", "t", bus_property_get_usec, property_set_runtime_watchdog, offsetof(Manager, runtime_watchdog), 0), + SD_BUS_WRITABLE_PROPERTY("ShutdownWatchdogUSec", "t", bus_property_get_usec, bus_property_set_usec, offsetof(Manager, shutdown_watchdog), 0), + SD_BUS_PROPERTY("ControlGroup", "s", NULL, offsetof(Manager, cgroup_root), 0), + SD_BUS_PROPERTY("SystemState", "s", property_get_system_state, 0, 0), + SD_BUS_PROPERTY("ExitCode", "y", bus_property_get_unsigned, offsetof(Manager, return_value), 0), + SD_BUS_PROPERTY("DefaultTimerAccuracyUSec", "t", bus_property_get_usec, offsetof(Manager, default_timer_accuracy_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultTimeoutStartUSec", "t", bus_property_get_usec, offsetof(Manager, default_timeout_start_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultTimeoutStopUSec", "t", bus_property_get_usec, offsetof(Manager, default_timeout_stop_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultRestartUSec", "t", bus_property_get_usec, offsetof(Manager, default_restart_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultStartLimitIntervalSec", "t", bus_property_get_usec, offsetof(Manager, default_start_limit_interval), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultStartLimitInterval", "t", bus_property_get_usec, offsetof(Manager, default_start_limit_interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), /* obsolete alias name */ + SD_BUS_PROPERTY("DefaultStartLimitBurst", "u", bus_property_get_unsigned, offsetof(Manager, default_start_limit_burst), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultCPUAccounting", "b", bus_property_get_bool, offsetof(Manager, default_cpu_accounting), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultBlockIOAccounting", "b", bus_property_get_bool, offsetof(Manager, default_blockio_accounting), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultMemoryAccounting", "b", bus_property_get_bool, offsetof(Manager, default_memory_accounting), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultTasksAccounting", "b", bus_property_get_bool, offsetof(Manager, default_tasks_accounting), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitCPU", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_CPU]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitCPUSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_CPU]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitFSIZE", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_FSIZE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitFSIZESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_FSIZE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitDATA", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_DATA]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitDATASoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_DATA]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitSTACK", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_STACK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitSTACKSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_STACK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitCORE", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_CORE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitCORESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_CORE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRSS", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RSS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRSSSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RSS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNOFILE", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NOFILE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNOFILESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NOFILE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitAS", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_AS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitASSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_AS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNPROC", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NPROC]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNPROCSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NPROC]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitMEMLOCK", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_MEMLOCK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitMEMLOCKSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_MEMLOCK]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitLOCKS", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_LOCKS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitLOCKSSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_LOCKS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitSIGPENDING", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_SIGPENDING]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitSIGPENDINGSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_SIGPENDING]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitMSGQUEUE", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_MSGQUEUE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitMSGQUEUESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_MSGQUEUE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNICE", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NICE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitNICESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_NICE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRTPRIO", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RTPRIO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRTPRIOSoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RTPRIO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRTTIME", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultLimitRTTIMESoft", "t", bus_property_get_rlimit, offsetof(Manager, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultTasksMax", "t", NULL, offsetof(Manager, default_tasks_max), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimerSlackNSec", "t", property_get_timer_slack_nsec, 0, SD_BUS_VTABLE_PROPERTY_CONST), + + SD_BUS_METHOD("GetUnit", "s", "o", method_get_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetUnitByPID", "u", "o", method_get_unit_by_pid, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetUnitByInvocationID", "ay", "o", method_get_unit_by_invocation_id, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("LoadUnit", "s", "o", method_load_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("StartUnit", "ss", "o", method_start_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("StartUnitReplace", "sss", "o", method_start_unit_replace, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("StopUnit", "ss", "o", method_stop_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReloadUnit", "ss", "o", method_reload_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("RestartUnit", "ss", "o", method_restart_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("TryRestartUnit", "ss", "o", method_try_restart_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReloadOrRestartUnit", "ss", "o", method_reload_or_restart_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReloadOrTryRestartUnit", "ss", "o", method_reload_or_try_restart_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("KillUnit", "ssi", NULL, method_kill_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ResetFailedUnit", "s", NULL, method_reset_failed_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("SetUnitProperties", "sba(sv)", NULL, method_set_unit_properties, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("RefUnit", "s", NULL, method_ref_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("UnrefUnit", "s", NULL, method_unref_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("StartTransientUnit", "ssa(sv)a(sa(sv))", "o", method_start_transient_unit, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetUnitProcesses", "s", "a(sus)", method_get_unit_processes, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetJob", "u", "o", method_get_job, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("CancelJob", "u", NULL, method_cancel_job, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ClearJobs", NULL, NULL, method_clear_jobs, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ResetFailed", NULL, NULL, method_reset_failed, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnits", NULL, "a(ssssssouso)", method_list_units, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnitsFiltered", "as", "a(ssssssouso)", method_list_units_filtered, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnitsByPatterns", "asas", "a(ssssssouso)", method_list_units_by_patterns, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnitsByNames", "as", "a(ssssssouso)", method_list_units_by_names, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListJobs", NULL, "a(usssoo)", method_list_jobs, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Subscribe", NULL, NULL, method_subscribe, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Unsubscribe", NULL, NULL, method_unsubscribe, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Dump", NULL, "s", method_dump, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("CreateSnapshot", "sb", "o", method_refuse_snapshot, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("RemoveSnapshot", "s", NULL, method_refuse_snapshot, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Reload", NULL, NULL, method_reload, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Reexecute", NULL, NULL, method_reexecute, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Exit", NULL, NULL, method_exit, 0), + SD_BUS_METHOD("Reboot", NULL, NULL, method_reboot, SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)), + SD_BUS_METHOD("PowerOff", NULL, NULL, method_poweroff, SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)), + SD_BUS_METHOD("Halt", NULL, NULL, method_halt, SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)), + SD_BUS_METHOD("KExec", NULL, NULL, method_kexec, SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)), + SD_BUS_METHOD("SwitchRoot", "ss", NULL, method_switch_root, SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)), + SD_BUS_METHOD("SetEnvironment", "as", NULL, method_set_environment, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("UnsetEnvironment", "as", NULL, method_unset_environment, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("UnsetAndSetEnvironment", "asas", NULL, method_unset_and_set_environment, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnitFiles", NULL, "a(ss)", method_list_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ListUnitFilesByPatterns", "asas", "a(ss)", method_list_unit_files_by_patterns, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetUnitFileState", "s", "s", method_get_unit_file_state, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("EnableUnitFiles", "asbb", "ba(sss)", method_enable_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("DisableUnitFiles", "asb", "a(sss)", method_disable_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReenableUnitFiles", "asbb", "ba(sss)", method_reenable_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("LinkUnitFiles", "asbb", "a(sss)", method_link_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("PresetUnitFiles", "asbb", "ba(sss)", method_preset_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("PresetUnitFilesWithMode", "assbb", "ba(sss)", method_preset_unit_files_with_mode, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("MaskUnitFiles", "asbb", "a(sss)", method_mask_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("UnmaskUnitFiles", "asb", "a(sss)", method_unmask_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("RevertUnitFiles", "as", "a(sss)", method_revert_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("SetDefaultTarget", "sb", "a(sss)", method_set_default_target, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetDefaultTarget", NULL, "s", method_get_default_target, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("PresetAllUnitFiles", "sbb", "a(sss)", method_preset_all_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("AddDependencyUnitFiles", "asssbb", "a(sss)", method_add_dependency_unit_files, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("GetUnitFileLinks", "sb", "as", method_get_unit_file_links, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("SetExitCode", "y", NULL, method_set_exit_code, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("LookupDynamicUserByName", "s", "u", method_lookup_dynamic_user_by_name, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("LookupDynamicUserByUID", "u", "s", method_lookup_dynamic_user_by_uid, SD_BUS_VTABLE_UNPRIVILEGED), + + SD_BUS_SIGNAL("UnitNew", "so", 0), + SD_BUS_SIGNAL("UnitRemoved", "so", 0), + SD_BUS_SIGNAL("JobNew", "uos", 0), + SD_BUS_SIGNAL("JobRemoved", "uoss", 0), + SD_BUS_SIGNAL("StartupFinished", "tttttt", 0), + SD_BUS_SIGNAL("UnitFilesChanged", NULL, 0), + SD_BUS_SIGNAL("Reloading", "b", 0), + + SD_BUS_VTABLE_END +}; + +static int send_finished(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; + usec_t *times = userdata; + int r; + + assert(bus); + assert(times); + + r = sd_bus_message_new_signal(bus, &message, "/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "StartupFinished"); + if (r < 0) + return r; + + r = sd_bus_message_append(message, "tttttt", times[0], times[1], times[2], times[3], times[4], times[5]); + if (r < 0) + return r; + + return sd_bus_send(bus, message, NULL); +} + +void bus_manager_send_finished( + Manager *m, + usec_t firmware_usec, + usec_t loader_usec, + usec_t kernel_usec, + usec_t initrd_usec, + usec_t userspace_usec, + usec_t total_usec) { + + int r; + + assert(m); + + r = bus_foreach_bus( + m, + NULL, + send_finished, + (usec_t[6]) { + firmware_usec, + loader_usec, + kernel_usec, + initrd_usec, + userspace_usec, + total_usec + }); + if (r < 0) + log_debug_errno(r, "Failed to send finished signal: %m"); +} + +static int send_reloading(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; + int r; + + assert(bus); + + r = sd_bus_message_new_signal(bus, &message, "/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", "Reloading"); + if (r < 0) + return r; + + r = sd_bus_message_append(message, "b", PTR_TO_INT(userdata)); + if (r < 0) + return r; + + return sd_bus_send(bus, message, NULL); +} + +void bus_manager_send_reloading(Manager *m, bool active) { + int r; + + assert(m); + + r = bus_foreach_bus(m, NULL, send_reloading, INT_TO_PTR(active)); + if (r < 0) + log_debug_errno(r, "Failed to send reloading signal: %m"); +} + +static int send_changed_signal(sd_bus *bus, void *userdata) { + assert(bus); + + return sd_bus_emit_properties_changed_strv(bus, + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + NULL); +} + +void bus_manager_send_change_signal(Manager *m) { + int r; + + assert(m); + + r = bus_foreach_bus(m, NULL, send_changed_signal, NULL); + if (r < 0) + log_debug_errno(r, "Failed to send manager change signal: %m"); +} diff --git a/src/grp-system/libcore/src/dbus-mount.c b/src/grp-system/libcore/src/dbus-mount.c new file mode 100644 index 0000000000..f9f396ec70 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-mount.c @@ -0,0 +1,219 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/mount.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-cgroup.h" +#include "dbus-execute.h" +#include "dbus-kill.h" +#include "dbus-mount.h" + +static int property_get_what( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Mount *m = userdata; + const char *d; + + assert(bus); + assert(reply); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.what) + d = m->parameters_proc_self_mountinfo.what; + else if (m->from_fragment && m->parameters_fragment.what) + d = m->parameters_fragment.what; + else + d = ""; + + return sd_bus_message_append(reply, "s", d); +} + +static int property_get_options( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Mount *m = userdata; + const char *d; + + assert(bus); + assert(reply); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.options) + d = m->parameters_proc_self_mountinfo.options; + else if (m->from_fragment && m->parameters_fragment.options) + d = m->parameters_fragment.options; + else + d = ""; + + return sd_bus_message_append(reply, "s", d); +} + +static int property_get_type( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Mount *m = userdata; + const char *d; + + assert(bus); + assert(reply); + assert(m); + + if (m->from_proc_self_mountinfo && m->parameters_proc_self_mountinfo.fstype) + d = m->parameters_proc_self_mountinfo.fstype; + else if (m->from_fragment && m->parameters_fragment.fstype) + d = m->parameters_fragment.fstype; + else + d = ""; + + return sd_bus_message_append(reply, "s", d); +} + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, mount_result, MountResult); + +const sd_bus_vtable bus_mount_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Where", "s", NULL, offsetof(Mount, where), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("What", "s", property_get_what, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Options","s", property_get_options, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Type", "s", property_get_type, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("TimeoutUSec", "t", bus_property_get_usec, offsetof(Mount, timeout_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(Mount, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("DirectoryMode", "u", bus_property_get_mode, offsetof(Mount, directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SloppyOptions", "b", bus_property_get_bool, offsetof(Mount, sloppy_options), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LazyUnmount", "b", bus_property_get_bool, offsetof(Mount, lazy_unmount), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ForceUnmount", "b", bus_property_get_bool, offsetof(Mount, force_unmount), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Mount, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_EXEC_COMMAND_VTABLE("ExecMount", offsetof(Mount, exec_command[MOUNT_EXEC_MOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_VTABLE("ExecUnmount", offsetof(Mount, exec_command[MOUNT_EXEC_UNMOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_VTABLE("ExecRemount", offsetof(Mount, exec_command[MOUNT_EXEC_REMOUNT]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + SD_BUS_VTABLE_END +}; + +static int bus_mount_set_transient_property( + Mount *m, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + const char *new_property; + char **property; + char *p; + int r; + + assert(m); + assert(name); + assert(message); + + if (streq(name, "What")) + property = &m->parameters_fragment.what; + else if (streq(name, "Options")) + property = &m->parameters_fragment.options; + else if (streq(name, "Type")) + property = &m->parameters_fragment.fstype; + else + return 0; + + r = sd_bus_message_read(message, "s", &new_property); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + p = strdup(new_property); + if (!p) + return -ENOMEM; + + unit_write_drop_in_format(UNIT(m), mode, name, "[Mount]\n%s=%s\n", + name, new_property); + + free(*property); + *property = p; + } + + return 1; +} + +int bus_mount_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Mount *m = MOUNT(u); + int r; + + assert(m); + assert(name); + assert(message); + + r = bus_cgroup_set_property(u, &m->cgroup_context, name, message, mode, error); + if (r != 0) + return r; + + if (u->transient && u->load_state == UNIT_STUB) { + /* This is a transient unit, let's load a little more */ + + r = bus_mount_set_transient_property(m, name, message, mode, error); + if (r != 0) + return r; + + r = bus_exec_context_set_transient_property(u, &m->exec_context, name, message, mode, error); + if (r != 0) + return r; + + r = bus_kill_context_set_transient_property(u, &m->kill_context, name, message, mode, error); + if (r != 0) + return r; + } + + return 0; +} + +int bus_mount_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-mount.h b/src/grp-system/libcore/src/dbus-mount.h new file mode 100644 index 0000000000..81ef769a44 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-mount.h @@ -0,0 +1,29 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_mount_vtable[]; + +int bus_mount_set_property(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_mount_commit_properties(Unit *u); diff --git a/src/grp-system/libcore/src/dbus-path.c b/src/grp-system/libcore/src/dbus-path.c new file mode 100644 index 0000000000..ed57c2133a --- /dev/null +++ b/src/grp-system/libcore/src/dbus-path.c @@ -0,0 +1,87 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/path.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-path.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, path_result, PathResult); + +static int property_get_paths( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Path *p = userdata; + PathSpec *k; + int r; + + assert(bus); + assert(reply); + assert(p); + + r = sd_bus_message_open_container(reply, 'a', "(ss)"); + if (r < 0) + return r; + + LIST_FOREACH(spec, k, p->specs) { + r = sd_bus_message_append(reply, "(ss)", path_type_to_string(k->type), k->path); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_unit( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *p = userdata, *trigger; + + assert(bus); + assert(reply); + assert(p); + + trigger = UNIT_TRIGGER(p); + + return sd_bus_message_append(reply, "s", trigger ? trigger->id : ""); +} + +const sd_bus_vtable bus_path_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Unit", "s", property_get_unit, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Paths", "a(ss)", property_get_paths, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MakeDirectory", "b", bus_property_get_bool, offsetof(Path, make_directory), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DirectoryMode", "u", bus_property_get_mode, offsetof(Path, directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Path, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_VTABLE_END +}; diff --git a/src/grp-system/libcore/src/dbus-path.h b/src/grp-system/libcore/src/dbus-path.h new file mode 100644 index 0000000000..d3c19e0c2b --- /dev/null +++ b/src/grp-system/libcore/src/dbus-path.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + + +extern const sd_bus_vtable bus_path_vtable[]; diff --git a/src/grp-system/libcore/src/dbus-scope.c b/src/grp-system/libcore/src/dbus-scope.c new file mode 100644 index 0000000000..da930418d5 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-scope.c @@ -0,0 +1,230 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/scope.h" +#include "core/unit.h" +#include "sd-bus/bus-common-errors.h" +#include "sd-bus/bus-internal.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" + +#include "dbus-cgroup.h" +#include "dbus-kill.h" +#include "dbus-scope.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "selinux-access.h" + +static int bus_scope_abandon(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Scope *s = userdata; + int r; + + assert(message); + assert(s); + + r = mac_selinux_unit_access_check(UNIT(s), message, "stop", error); + if (r < 0) + return r; + + r = bus_verify_manage_units_async(UNIT(s)->manager, message, error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = scope_abandon(s); + if (r == -ESTALE) + return sd_bus_error_setf(error, BUS_ERROR_SCOPE_NOT_RUNNING, "Scope %s is not running, cannot abandon.", UNIT(s)->id); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, scope_result, ScopeResult); + +const sd_bus_vtable bus_scope_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Controller", "s", NULL, offsetof(Scope, controller), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimeoutStopUSec", "t", bus_property_get_usec, offsetof(Scope, timeout_stop_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Scope, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_SIGNAL("RequestStop", NULL, 0), + SD_BUS_METHOD("Abandon", NULL, NULL, bus_scope_abandon, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_VTABLE_END +}; + +static int bus_scope_set_transient_property( + Scope *s, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(s); + assert(name); + assert(message); + + if (streq(name, "PIDs")) { + unsigned n = 0; + uint32_t pid; + + r = sd_bus_message_enter_container(message, 'a', "u"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "u", &pid)) > 0) { + + if (pid <= 1) + return -EINVAL; + + if (mode != UNIT_CHECK) { + r = unit_watch_pid(UNIT(s), pid); + if (r < 0 && r != -EEXIST) + return r; + } + + n++; + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (n <= 0) + return -EINVAL; + + return 1; + + } else if (streq(name, "Controller")) { + const char *controller; + char *c; + + r = sd_bus_message_read(message, "s", &controller); + if (r < 0) + return r; + + if (!isempty(controller) && !service_name_is_valid(controller)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Controller '%s' is not a valid bus name.", controller); + + if (mode != UNIT_CHECK) { + if (isempty(controller)) + c = NULL; + else { + c = strdup(controller); + if (!c) + return -ENOMEM; + } + + free(s->controller); + s->controller = c; + } + + return 1; + + } else if (streq(name, "TimeoutStopUSec")) { + + if (mode != UNIT_CHECK) { + r = sd_bus_message_read(message, "t", &s->timeout_stop_usec); + if (r < 0) + return r; + + unit_write_drop_in_private_format(UNIT(s), mode, name, "TimeoutStopSec="USEC_FMT"us", s->timeout_stop_usec); + } else { + r = sd_bus_message_skip(message, "t"); + if (r < 0) + return r; + } + + return 1; + } + + return 0; +} + +int bus_scope_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Scope *s = SCOPE(u); + int r; + + assert(s); + assert(name); + assert(message); + + r = bus_cgroup_set_property(u, &s->cgroup_context, name, message, mode, error); + if (r != 0) + return r; + + if (u->load_state == UNIT_STUB) { + /* While we are created we still accept PIDs */ + + r = bus_scope_set_transient_property(s, name, message, mode, error); + if (r != 0) + return r; + + r = bus_kill_context_set_transient_property(u, &s->kill_context, name, message, mode, error); + if (r != 0) + return r; + } + + return 0; +} + +int bus_scope_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} + +int bus_scope_send_request_stop(Scope *s) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_free_ char *p = NULL; + int r; + + assert(s); + + if (!s->controller) + return 0; + + p = unit_dbus_path(UNIT(s)); + if (!p) + return -ENOMEM; + + r = sd_bus_message_new_signal( + UNIT(s)->manager->api_bus, + &m, + p, + "org.freedesktop.systemd1.Scope", + "RequestStop"); + if (r < 0) + return r; + + return sd_bus_send_to(UNIT(s)->manager->api_bus, m, s->controller, NULL); +} diff --git a/src/grp-system/libcore/src/dbus-scope.h b/src/grp-system/libcore/src/dbus-scope.h new file mode 100644 index 0000000000..ad1cb52adc --- /dev/null +++ b/src/grp-system/libcore/src/dbus-scope.h @@ -0,0 +1,31 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_scope_vtable[]; + +int bus_scope_set_property(Unit *u, const char *name, sd_bus_message *i, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_scope_commit_properties(Unit *u); + +int bus_scope_send_request_stop(Scope *s); diff --git a/src/grp-system/libcore/src/dbus-service.c b/src/grp-system/libcore/src/dbus-service.c new file mode 100644 index 0000000000..da8a2298a2 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-service.c @@ -0,0 +1,327 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/service.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/async.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" + +#include "dbus-cgroup.h" +#include "dbus-execute.h" +#include "dbus-kill.h" +#include "dbus-service.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_type, service_type, ServiceType); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, service_result, ServiceResult); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_restart, service_restart, ServiceRestart); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_notify_access, notify_access, NotifyAccess); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_emergency_action, emergency_action, EmergencyAction); + +const sd_bus_vtable bus_service_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Type", "s", property_get_type, offsetof(Service, type), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Restart", "s", property_get_restart, offsetof(Service, restart), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PIDFile", "s", NULL, offsetof(Service, pid_file), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NotifyAccess", "s", property_get_notify_access, offsetof(Service, notify_access), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RestartUSec", "t", bus_property_get_usec, offsetof(Service, restart_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimeoutStartUSec", "t", bus_property_get_usec, offsetof(Service, timeout_start_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimeoutStopUSec", "t", bus_property_get_usec, offsetof(Service, timeout_stop_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RuntimeMaxUSec", "t", bus_property_get_usec, offsetof(Service, runtime_max_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("WatchdogUSec", "t", bus_property_get_usec, offsetof(Service, watchdog_usec), SD_BUS_VTABLE_PROPERTY_CONST), + BUS_PROPERTY_DUAL_TIMESTAMP("WatchdogTimestamp", offsetof(Service, watchdog_timestamp), 0), + SD_BUS_PROPERTY("FailureAction", "s", property_get_emergency_action, offsetof(Service, emergency_action), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PermissionsStartOnly", "b", bus_property_get_bool, offsetof(Service, permissions_start_only), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RootDirectoryStartOnly", "b", bus_property_get_bool, offsetof(Service, root_directory_start_only), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RemainAfterExit", "b", bus_property_get_bool, offsetof(Service, remain_after_exit), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("GuessMainPID", "b", bus_property_get_bool, offsetof(Service, guess_main_pid), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MainPID", "u", bus_property_get_pid, offsetof(Service, main_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(Service, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("BusName", "s", NULL, offsetof(Service, bus_name), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("FileDescriptorStoreMax", "u", bus_property_get_unsigned, offsetof(Service, n_fd_store_max), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NFileDescriptorStore", "u", bus_property_get_unsigned, offsetof(Service, n_fd_store), 0), + SD_BUS_PROPERTY("StatusText", "s", NULL, offsetof(Service, status_text), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("StatusErrno", "i", NULL, offsetof(Service, status_errno), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Service, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("USBFunctionDescriptors", "s", NULL, offsetof(Service, usb_function_descriptors), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("USBFunctionStrings", "s", NULL, offsetof(Service, usb_function_strings), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + + BUS_EXEC_STATUS_VTABLE("ExecMain", offsetof(Service, main_exec_status), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPre", offsetof(Service, exec_command[SERVICE_EXEC_START_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStart", offsetof(Service, exec_command[SERVICE_EXEC_START]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPost", offsetof(Service, exec_command[SERVICE_EXEC_START_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecReload", offsetof(Service, exec_command[SERVICE_EXEC_RELOAD]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStop", offsetof(Service, exec_command[SERVICE_EXEC_STOP]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStopPost", offsetof(Service, exec_command[SERVICE_EXEC_STOP_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + + /* The following four are obsolete, and thus marked hidden here. They moved into the Unit interface */ + SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Unit, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitAction", "s", property_get_emergency_action, offsetof(Unit, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Unit, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_VTABLE_END +}; + +static int bus_service_set_transient_property( + Service *s, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(s); + assert(name); + assert(message); + + if (streq(name, "RemainAfterExit")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + s->remain_after_exit = b; + unit_write_drop_in_private_format(UNIT(s), mode, name, "RemainAfterExit=%s", yes_no(b)); + } + + return 1; + + } else if (streq(name, "Type")) { + const char *t; + ServiceType k; + + r = sd_bus_message_read(message, "s", &t); + if (r < 0) + return r; + + k = service_type_from_string(t); + if (k < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid service type %s", t); + + if (mode != UNIT_CHECK) { + s->type = k; + unit_write_drop_in_private_format(UNIT(s), mode, name, "Type=%s", service_type_to_string(s->type)); + } + + return 1; + } else if (streq(name, "RuntimeMaxUSec")) { + usec_t u; + + r = sd_bus_message_read(message, "t", &u); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + s->runtime_max_usec = u; + unit_write_drop_in_private_format(UNIT(s), mode, name, "RuntimeMaxSec=" USEC_FMT "us", u); + } + + return 1; + + } else if (STR_IN_SET(name, + "StandardInputFileDescriptor", + "StandardOutputFileDescriptor", + "StandardErrorFileDescriptor")) { + int fd; + + r = sd_bus_message_read(message, "h", &fd); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + int copy; + + copy = fcntl(fd, F_DUPFD_CLOEXEC, 3); + if (copy < 0) + return -errno; + + if (streq(name, "StandardInputFileDescriptor")) { + asynchronous_close(s->stdin_fd); + s->stdin_fd = copy; + } else if (streq(name, "StandardOutputFileDescriptor")) { + asynchronous_close(s->stdout_fd); + s->stdout_fd = copy; + } else { + asynchronous_close(s->stderr_fd); + s->stderr_fd = copy; + } + + s->exec_context.stdio_as_fds = true; + } + + return 1; + + } else if (streq(name, "ExecStart")) { + unsigned n = 0; + + r = sd_bus_message_enter_container(message, 'a', "(sasb)"); + if (r < 0) + return r; + + while ((r = sd_bus_message_enter_container(message, 'r', "sasb")) > 0) { + _cleanup_strv_free_ char **argv = NULL; + const char *path; + int b; + + r = sd_bus_message_read(message, "s", &path); + if (r < 0) + return r; + + if (!path_is_absolute(path)) + return sd_bus_error_set_errnof(error, EINVAL, "Path %s is not absolute.", path); + + r = sd_bus_message_read_strv(message, &argv); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + ExecCommand *c; + + c = new0(ExecCommand, 1); + if (!c) + return -ENOMEM; + + c->path = strdup(path); + if (!c->path) { + free(c); + return -ENOMEM; + } + + c->argv = argv; + argv = NULL; + + c->ignore = b; + + path_kill_slashes(c->path); + exec_command_append_list(&s->exec_command[SERVICE_EXEC_START], c); + } + + n++; + } + + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *buf = NULL; + _cleanup_fclose_ FILE *f = NULL; + ExecCommand *c; + size_t size = 0; + + if (n == 0) + s->exec_command[SERVICE_EXEC_START] = exec_command_free_list(s->exec_command[SERVICE_EXEC_START]); + + f = open_memstream(&buf, &size); + if (!f) + return -ENOMEM; + + fputs("ExecStart=\n", f); + + LIST_FOREACH(command, c, s->exec_command[SERVICE_EXEC_START]) { + _cleanup_free_ char *a; + + a = strv_join_quoted(c->argv); + if (!a) + return -ENOMEM; + + fprintf(f, "ExecStart=%s@%s %s\n", + c->ignore ? "-" : "", + c->path, + a); + } + + r = fflush_and_check(f); + if (r < 0) + return r; + unit_write_drop_in_private(UNIT(s), mode, name, buf); + } + + return 1; + } + + return 0; +} + +int bus_service_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Service *s = SERVICE(u); + int r; + + assert(s); + assert(name); + assert(message); + + r = bus_cgroup_set_property(u, &s->cgroup_context, name, message, mode, error); + if (r != 0) + return r; + + if (u->transient && u->load_state == UNIT_STUB) { + /* This is a transient unit, let's load a little more */ + + r = bus_service_set_transient_property(s, name, message, mode, error); + if (r != 0) + return r; + + r = bus_exec_context_set_transient_property(u, &s->exec_context, name, message, mode, error); + if (r != 0) + return r; + + r = bus_kill_context_set_transient_property(u, &s->kill_context, name, message, mode, error); + if (r != 0) + return r; + } + + return 0; +} + +int bus_service_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-service.h b/src/grp-system/libcore/src/dbus-service.h new file mode 100644 index 0000000000..1d3df67bf0 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-service.h @@ -0,0 +1,29 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_service_vtable[]; + +int bus_service_set_property(Unit *u, const char *name, sd_bus_message *i, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_service_commit_properties(Unit *u); diff --git a/src/grp-system/libcore/src/dbus-slice.c b/src/grp-system/libcore/src/dbus-slice.c new file mode 100644 index 0000000000..ce357cadae --- /dev/null +++ b/src/grp-system/libcore/src/dbus-slice.c @@ -0,0 +1,53 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/slice.h" +#include "core/unit.h" + +#include "dbus-cgroup.h" +#include "dbus-slice.h" + +const sd_bus_vtable bus_slice_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_VTABLE_END +}; + +int bus_slice_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Slice *s = SLICE(u); + + assert(name); + assert(u); + + return bus_cgroup_set_property(u, &s->cgroup_context, name, message, mode, error); +} + +int bus_slice_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-slice.h b/src/grp-system/libcore/src/dbus-slice.h new file mode 100644 index 0000000000..fa039b2c0b --- /dev/null +++ b/src/grp-system/libcore/src/dbus-slice.h @@ -0,0 +1,29 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_slice_vtable[]; + +int bus_slice_set_property(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_slice_commit_properties(Unit *u); diff --git a/src/grp-system/libcore/src/dbus-socket.c b/src/grp-system/libcore/src/dbus-socket.c new file mode 100644 index 0000000000..4bd973a882 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-socket.c @@ -0,0 +1,188 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/socket.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-cgroup.h" +#include "dbus-execute.h" +#include "dbus-socket.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, socket_result, SocketResult); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_bind_ipv6_only, socket_address_bind_ipv6_only, SocketAddressBindIPv6Only); + +static int property_get_listen( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + + Socket *s = SOCKET(userdata); + SocketPort *p; + int r; + + assert(bus); + assert(reply); + assert(s); + + r = sd_bus_message_open_container(reply, 'a', "(ss)"); + if (r < 0) + return r; + + LIST_FOREACH(port, p, s->ports) { + _cleanup_free_ char *address = NULL; + const char *a; + + switch (p->type) { + case SOCKET_SOCKET: { + r = socket_address_print(&p->address, &address); + if (r) + return r; + + a = address; + break; + } + + case SOCKET_SPECIAL: + case SOCKET_MQUEUE: + case SOCKET_FIFO: + case SOCKET_USB_FUNCTION: + a = p->path; + break; + + default: + assert_not_reached("Unknown socket type"); + } + + r = sd_bus_message_append(reply, "(ss)", socket_port_type_to_string(p), a); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + + +static int property_get_fdname( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Socket *s = SOCKET(userdata); + + assert(bus); + assert(reply); + assert(s); + + return sd_bus_message_append(reply, "s", socket_fdname(s)); +} + +const sd_bus_vtable bus_socket_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("BindIPv6Only", "s", property_get_bind_ipv6_only, offsetof(Socket, bind_ipv6_only), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Backlog", "u", bus_property_get_unsigned, offsetof(Socket, backlog), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimeoutUSec", "t", bus_property_get_usec, offsetof(Socket, timeout_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("BindToDevice", "s", NULL, offsetof(Socket, bind_to_device), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SocketUser", "s", NULL, offsetof(Socket, user), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SocketGroup", "s", NULL, offsetof(Socket, group), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SocketMode", "u", bus_property_get_mode, offsetof(Socket, socket_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DirectoryMode", "u", bus_property_get_mode, offsetof(Socket, directory_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Accept", "b", bus_property_get_bool, offsetof(Socket, accept), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Writable", "b", bus_property_get_bool, offsetof(Socket, writable), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("KeepAlive", "b", bus_property_get_bool, offsetof(Socket, keep_alive), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("KeepAliveTimeUSec", "t", bus_property_get_usec, offsetof(Socket, keep_alive_time), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("KeepAliveIntervalUSec", "t", bus_property_get_usec, offsetof(Socket, keep_alive_interval), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("KeepAliveProbes", "u", bus_property_get_unsigned, offsetof(Socket, keep_alive_cnt), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DeferAcceptUSec" , "t", bus_property_get_usec, offsetof(Socket, defer_accept), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NoDelay", "b", bus_property_get_bool, offsetof(Socket, no_delay), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Priority", "i", bus_property_get_int, offsetof(Socket, priority), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ReceiveBuffer", "t", bus_property_get_size, offsetof(Socket, receive_buffer), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SendBuffer", "t", bus_property_get_size, offsetof(Socket, send_buffer), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("IPTOS", "i", bus_property_get_int, offsetof(Socket, ip_tos), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("IPTTL", "i", bus_property_get_int, offsetof(Socket, ip_ttl), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PipeSize", "t", bus_property_get_size, offsetof(Socket, pipe_size), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("FreeBind", "b", bus_property_get_bool, offsetof(Socket, free_bind), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Transparent", "b", bus_property_get_bool, offsetof(Socket, transparent), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Broadcast", "b", bus_property_get_bool, offsetof(Socket, broadcast), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PassCredentials", "b", bus_property_get_bool, offsetof(Socket, pass_cred), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PassSecurity", "b", bus_property_get_bool, offsetof(Socket, pass_sec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RemoveOnStop", "b", bus_property_get_bool, offsetof(Socket, remove_on_stop), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Listen", "a(ss)", property_get_listen, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Symlinks", "as", NULL, offsetof(Socket, symlinks), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Mark", "i", bus_property_get_int, offsetof(Socket, mark), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MaxConnections", "u", bus_property_get_unsigned, offsetof(Socket, max_connections), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MaxConnectionsPerSource", "u", bus_property_get_unsigned, offsetof(Socket, max_connections_per_source), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MessageQueueMaxMessages", "x", bus_property_get_long, offsetof(Socket, mq_maxmsg), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("MessageQueueMessageSize", "x", bus_property_get_long, offsetof(Socket, mq_msgsize), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ReusePort", "b", bus_property_get_bool, offsetof(Socket, reuse_port), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SmackLabel", "s", NULL, offsetof(Socket, smack), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SmackLabelIPIn", "s", NULL, offsetof(Socket, smack_ip_in), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SmackLabelIPOut", "s", NULL, offsetof(Socket, smack_ip_out), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(Socket, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Socket, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("NConnections", "u", bus_property_get_unsigned, offsetof(Socket, n_connections), 0), + SD_BUS_PROPERTY("NAccepted", "u", bus_property_get_unsigned, offsetof(Socket, n_accepted), 0), + SD_BUS_PROPERTY("FileDescriptorName", "s", property_get_fdname, 0, 0), + SD_BUS_PROPERTY("SocketProtocol", "i", bus_property_get_int, offsetof(Socket, socket_protocol), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TriggerLimitIntervalUSec", "t", bus_property_get_usec, offsetof(Socket, trigger_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TriggerLimitBurst", "u", bus_property_get_unsigned, offsetof(Socket, trigger_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPre", offsetof(Socket, exec_command[SOCKET_EXEC_START_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStartPost", offsetof(Socket, exec_command[SOCKET_EXEC_START_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStopPre", offsetof(Socket, exec_command[SOCKET_EXEC_STOP_PRE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_LIST_VTABLE("ExecStopPost", offsetof(Socket, exec_command[SOCKET_EXEC_STOP_POST]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + SD_BUS_VTABLE_END +}; + +int bus_socket_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Socket *s = SOCKET(u); + + assert(s); + assert(name); + assert(message); + + return bus_cgroup_set_property(u, &s->cgroup_context, name, message, mode, error); +} + +int bus_socket_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-socket.h b/src/grp-system/libcore/src/dbus-socket.h new file mode 100644 index 0000000000..e68d33ace7 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-socket.h @@ -0,0 +1,29 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_socket_vtable[]; + +int bus_socket_set_property(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_socket_commit_properties(Unit *u); diff --git a/src/grp-system/libcore/src/dbus-swap.c b/src/grp-system/libcore/src/dbus-swap.c new file mode 100644 index 0000000000..6e40d59808 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-swap.c @@ -0,0 +1,118 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/swap.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/string-util.h" + +#include "dbus-cgroup.h" +#include "dbus-execute.h" +#include "dbus-swap.h" + +static int property_get_priority( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Swap *s = SWAP(userdata); + int p; + + assert(bus); + assert(reply); + assert(s); + + if (s->from_proc_swaps) + p = s->parameters_proc_swaps.priority; + else if (s->from_fragment) + p = s->parameters_fragment.priority; + else + p = -1; + + return sd_bus_message_append(reply, "i", p); +} + +static int property_get_options( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Swap *s = SWAP(userdata); + const char *options = NULL; + + assert(bus); + assert(reply); + assert(s); + + if (s->from_fragment) + options = s->parameters_fragment.options; + + return sd_bus_message_append(reply, "s", options); +} + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, swap_result, SwapResult); + +const sd_bus_vtable bus_swap_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("What", "s", NULL, offsetof(Swap, what), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Priority", "i", property_get_priority, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Options", "s", property_get_options, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("TimeoutUSec", "t", bus_property_get_usec, offsetof(Swap, timeout_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ControlPID", "u", bus_property_get_pid, offsetof(Swap, control_pid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Swap, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("UID", "u", NULL, offsetof(Unit, ref_uid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("GID", "u", NULL, offsetof(Unit, ref_gid), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_EXEC_COMMAND_VTABLE("ExecActivate", offsetof(Swap, exec_command[SWAP_EXEC_ACTIVATE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + BUS_EXEC_COMMAND_VTABLE("ExecDeactivate", offsetof(Swap, exec_command[SWAP_EXEC_DEACTIVATE]), SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + SD_BUS_VTABLE_END +}; + +int bus_swap_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Swap *s = SWAP(u); + + assert(s); + assert(name); + assert(message); + + return bus_cgroup_set_property(u, &s->cgroup_context, name, message, mode, error); +} + +int bus_swap_commit_properties(Unit *u) { + assert(u); + + unit_update_cgroup_members_masks(u); + unit_realize_cgroup(u); + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-swap.h b/src/grp-system/libcore/src/dbus-swap.h new file mode 100644 index 0000000000..6c1b862665 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-swap.h @@ -0,0 +1,30 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2010 Maarten Lankhorst + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_swap_vtable[]; + +int bus_swap_set_property(Unit *u, const char *name, sd_bus_message *message, UnitSetPropertiesMode mode, sd_bus_error *error); +int bus_swap_commit_properties(Unit *u); diff --git a/src/grp-system/libcore/src/dbus-target.c b/src/grp-system/libcore/src/dbus-target.c new file mode 100644 index 0000000000..5a846432d0 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-target.c @@ -0,0 +1,27 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/unit.h" + +#include "dbus-target.h" + +const sd_bus_vtable bus_target_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_VTABLE_END +}; diff --git a/src/grp-system/libcore/src/dbus-target.h b/src/grp-system/libcore/src/dbus-target.h new file mode 100644 index 0000000000..c97a9d626e --- /dev/null +++ b/src/grp-system/libcore/src/dbus-target.h @@ -0,0 +1,24 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +extern const sd_bus_vtable bus_target_vtable[]; diff --git a/src/grp-system/libcore/src/dbus-timer.c b/src/grp-system/libcore/src/dbus-timer.c new file mode 100644 index 0000000000..6b00168361 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-timer.c @@ -0,0 +1,353 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/timer.h" +#include "core/unit.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/strv.h" + +#include "dbus-timer.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, timer_result, TimerResult); + +static int property_get_monotonic_timers( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Timer *t = userdata; + TimerValue *v; + int r; + + assert(bus); + assert(reply); + assert(t); + + r = sd_bus_message_open_container(reply, 'a', "(stt)"); + if (r < 0) + return r; + + LIST_FOREACH(value, v, t->values) { + _cleanup_free_ char *buf = NULL; + const char *s; + size_t l; + + if (v->base == TIMER_CALENDAR) + continue; + + s = timer_base_to_string(v->base); + assert(endswith(s, "Sec")); + + /* s/Sec/USec/ */ + l = strlen(s); + buf = new(char, l+2); + if (!buf) + return -ENOMEM; + + memcpy(buf, s, l-3); + memcpy(buf+l-3, "USec", 5); + + r = sd_bus_message_append(reply, "(stt)", buf, v->value, v->next_elapse); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_calendar_timers( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Timer *t = userdata; + TimerValue *v; + int r; + + assert(bus); + assert(reply); + assert(t); + + r = sd_bus_message_open_container(reply, 'a', "(sst)"); + if (r < 0) + return r; + + LIST_FOREACH(value, v, t->values) { + _cleanup_free_ char *buf = NULL; + + if (v->base != TIMER_CALENDAR) + continue; + + r = calendar_spec_to_string(v->calendar_spec, &buf); + if (r < 0) + return r; + + r = sd_bus_message_append(reply, "(sst)", timer_base_to_string(v->base), buf, v->next_elapse); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_unit( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata, *trigger; + + assert(bus); + assert(reply); + assert(u); + + trigger = UNIT_TRIGGER(u); + + return sd_bus_message_append(reply, "s", trigger ? trigger->id : ""); +} + +static int property_get_next_elapse_monotonic( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Timer *t = userdata; + usec_t x; + + assert(bus); + assert(reply); + assert(t); + + if (t->next_elapse_monotonic_or_boottime <= 0) + x = 0; + else if (t->wake_system) { + usec_t a, b; + + a = now(CLOCK_MONOTONIC); + b = now(clock_boottime_or_monotonic()); + + if (t->next_elapse_monotonic_or_boottime + a > b) + x = t->next_elapse_monotonic_or_boottime + a - b; + else + x = 0; + } else + x = t->next_elapse_monotonic_or_boottime; + + return sd_bus_message_append(reply, "t", x); +} + +const sd_bus_vtable bus_timer_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Unit", "s", property_get_unit, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TimersMonotonic", "a(stt)", property_get_monotonic_timers, 0, SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + SD_BUS_PROPERTY("TimersCalendar", "a(sst)", property_get_calendar_timers, 0, SD_BUS_VTABLE_PROPERTY_EMITS_INVALIDATION), + SD_BUS_PROPERTY("NextElapseUSecRealtime", "t", bus_property_get_usec, offsetof(Timer, next_elapse_realtime), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("NextElapseUSecMonotonic", "t", property_get_next_elapse_monotonic, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("LastTriggerUSec", offsetof(Timer, last_trigger), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Result", "s", property_get_result, offsetof(Timer, result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("AccuracyUSec", "t", bus_property_get_usec, offsetof(Timer, accuracy_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RandomizedDelayUSec", "t", bus_property_get_usec, offsetof(Timer, random_usec), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Persistent", "b", bus_property_get_bool, offsetof(Timer, persistent), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("WakeSystem", "b", bus_property_get_bool, offsetof(Timer, wake_system), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RemainAfterElapse", "b", bus_property_get_bool, offsetof(Timer, remain_after_elapse), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_VTABLE_END +}; + +static int bus_timer_set_transient_property( + Timer *t, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(t); + assert(name); + assert(message); + + if (STR_IN_SET(name, + "OnActiveSec", + "OnBootSec", + "OnStartupSec", + "OnUnitActiveSec", + "OnUnitInactiveSec")) { + + TimerValue *v; + TimerBase b = _TIMER_BASE_INVALID; + usec_t u = 0; + + b = timer_base_from_string(name); + if (b < 0) + return -EINVAL; + + r = sd_bus_message_read(message, "t", &u); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + char time[FORMAT_TIMESPAN_MAX]; + + unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s", name, format_timespan(time, sizeof(time), u, USEC_PER_MSEC)); + + v = new0(TimerValue, 1); + if (!v) + return -ENOMEM; + + v->base = b; + v->value = u; + + LIST_PREPEND(value, t->values, v); + } + + return 1; + + } else if (streq(name, "OnCalendar")) { + + TimerValue *v; + CalendarSpec *c = NULL; + const char *str; + + r = sd_bus_message_read(message, "s", &str); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + r = calendar_spec_from_string(str, &c); + if (r < 0) + return r; + + unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s", name, str); + + v = new0(TimerValue, 1); + if (!v) { + calendar_spec_free(c); + return -ENOMEM; + } + + v->base = TIMER_CALENDAR; + v->calendar_spec = c; + + LIST_PREPEND(value, t->values, v); + } + + return 1; + + } else if (STR_IN_SET(name, "AccuracyUSec", "AccuracySec")) { + usec_t u = 0; + + if (streq(name, "AccuracySec")) + log_notice("Client is using obsolete AccuracySec= transient property, please use AccuracyUSec= instead."); + + r = sd_bus_message_read(message, "t", &u); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + t->accuracy_usec = u; + unit_write_drop_in_private_format(UNIT(t), mode, name, "AccuracySec=" USEC_FMT "us", u); + } + + return 1; + + } else if (streq(name, "RandomizedDelayUSec")) { + usec_t u = 0; + + r = sd_bus_message_read(message, "t", &u); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + t->random_usec = u; + unit_write_drop_in_private_format(UNIT(t), mode, name, "RandomizedDelaySec=" USEC_FMT "us", u); + } + + return 1; + + } else if (streq(name, "WakeSystem")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + t->wake_system = b; + unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s", name, yes_no(b)); + } + + return 1; + + } else if (streq(name, "RemainAfterElapse")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + t->remain_after_elapse = b; + unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s", name, yes_no(b)); + } + + return 1; + } + + return 0; +} + +int bus_timer_set_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + Timer *t = TIMER(u); + int r; + + assert(t); + assert(name); + assert(message); + + if (u->transient && u->load_state == UNIT_STUB) { + r = bus_timer_set_transient_property(t, name, message, mode, error); + if (r != 0) + return r; + } + + return 0; +} diff --git a/src/grp-system/libcore/src/dbus-timer.h b/src/grp-system/libcore/src/dbus-timer.h new file mode 100644 index 0000000000..5e366918f8 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-timer.h @@ -0,0 +1,28 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_timer_vtable[]; + +int bus_timer_set_property(Unit *u, const char *name, sd_bus_message *i, UnitSetPropertiesMode mode, sd_bus_error *error); diff --git a/src/grp-system/libcore/src/dbus-unit.c b/src/grp-system/libcore/src/dbus-unit.c new file mode 100644 index 0000000000..93391c2c6e --- /dev/null +++ b/src/grp-system/libcore/src/dbus-unit.c @@ -0,0 +1,1577 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "sd-bus/bus-common-errors.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/locale-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/user-util.h" + +#include "dbus-unit.h" +#include "dbus.h" +#include "selinux-access.h" + +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_load_state, unit_load_state, UnitLoadState); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_job_mode, job_mode, JobMode); +static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_emergency_action, emergency_action, EmergencyAction); + +static int property_get_names( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + Iterator i; + const char *t; + int r; + + assert(bus); + assert(reply); + assert(u); + + r = sd_bus_message_open_container(reply, 'a', "s"); + if (r < 0) + return r; + + SET_FOREACH(t, u->names, i) { + r = sd_bus_message_append(reply, "s", t); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_following( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata, *f; + + assert(bus); + assert(reply); + assert(u); + + f = unit_following(u); + return sd_bus_message_append(reply, "s", f ? f->id : ""); +} + +static int property_get_dependencies( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Set *s = *(Set**) userdata; + Iterator j; + Unit *u; + int r; + + assert(bus); + assert(reply); + + r = sd_bus_message_open_container(reply, 'a', "s"); + if (r < 0) + return r; + + SET_FOREACH(u, s, j) { + r = sd_bus_message_append(reply, "s", u->id); + if (r < 0) + return r; + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_obsolete_dependencies( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + assert(bus); + assert(reply); + + /* For dependency types we don't support anymore always return an empty array */ + return sd_bus_message_append(reply, "as", 0); +} + +static int property_get_description( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "s", unit_description(u)); +} + +static int property_get_active_state( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "s", unit_active_state_to_string(unit_active_state(u))); +} + +static int property_get_sub_state( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "s", unit_sub_state_to_string(u)); +} + +static int property_get_unit_file_preset( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + int r; + + assert(bus); + assert(reply); + assert(u); + + r = unit_get_unit_file_preset(u); + + return sd_bus_message_append(reply, "s", + r < 0 ? "": + r > 0 ? "enabled" : "disabled"); +} + +static int property_get_unit_file_state( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "s", unit_file_state_to_string(unit_get_unit_file_state(u))); +} + +static int property_get_can_start( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "b", unit_can_start(u) && !u->refuse_manual_start); +} + +static int property_get_can_stop( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "b", unit_can_stop(u) && !u->refuse_manual_stop); +} + +static int property_get_can_reload( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "b", unit_can_reload(u)); +} + +static int property_get_can_isolate( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "b", unit_can_isolate(u) && !u->refuse_manual_start); +} + +static int property_get_job( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + _cleanup_free_ char *p = NULL; + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + if (!u->job) + return sd_bus_message_append(reply, "(uo)", 0, "/"); + + p = job_dbus_path(u->job); + if (!p) + return -ENOMEM; + + return sd_bus_message_append(reply, "(uo)", u->job->id, p); +} + +static int property_get_need_daemon_reload( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "b", unit_need_daemon_reload(u)); +} + +static int property_get_conditions( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + const char *(*to_string)(ConditionType type) = NULL; + Condition **list = userdata, *c; + int r; + + assert(bus); + assert(reply); + assert(list); + + to_string = streq(property, "Asserts") ? assert_type_to_string : condition_type_to_string; + + r = sd_bus_message_open_container(reply, 'a', "(sbbsi)"); + if (r < 0) + return r; + + LIST_FOREACH(conditions, c, *list) { + int tristate; + + tristate = + c->result == CONDITION_UNTESTED ? 0 : + c->result == CONDITION_SUCCEEDED ? 1 : -1; + + r = sd_bus_message_append(reply, "(sbbsi)", + to_string(c->type), + c->trigger, c->negate, + c->parameter, tristate); + if (r < 0) + return r; + + } + + return sd_bus_message_close_container(reply); +} + +static int property_get_load_error( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + _cleanup_(sd_bus_error_free) sd_bus_error e = SD_BUS_ERROR_NULL; + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + if (u->load_error != 0) + sd_bus_error_set_errno(&e, u->load_error); + + return sd_bus_message_append(reply, "(ss)", e.name, e.message); +} + +static int bus_verify_manage_units_async_full( + Unit *u, + const char *verb, + int capability, + const char *polkit_message, + bool interactive, + sd_bus_message *call, + sd_bus_error *error) { + + const char *details[9] = { + "unit", u->id, + "verb", verb, + }; + + if (polkit_message) { + details[4] = "polkit.message"; + details[5] = polkit_message; + details[6] = "polkit.gettext_domain"; + details[7] = GETTEXT_PACKAGE; + } + + return bus_verify_polkit_async( + call, + capability, + "org.freedesktop.systemd1.manage-units", + details, + interactive, + UID_INVALID, + &u->manager->polkit_registry, + error); +} + +int bus_unit_method_start_generic( + sd_bus_message *message, + Unit *u, + JobType job_type, + bool reload_if_possible, + sd_bus_error *error) { + + const char *smode; + JobMode mode; + _cleanup_free_ char *verb = NULL; + static const char *const polkit_message_for_job[_JOB_TYPE_MAX] = { + [JOB_START] = N_("Authentication is required to start '$(unit)'."), + [JOB_STOP] = N_("Authentication is required to stop '$(unit)'."), + [JOB_RELOAD] = N_("Authentication is required to reload '$(unit)'."), + [JOB_RESTART] = N_("Authentication is required to restart '$(unit)'."), + [JOB_TRY_RESTART] = N_("Authentication is required to restart '$(unit)'."), + }; + int r; + + assert(message); + assert(u); + assert(job_type >= 0 && job_type < _JOB_TYPE_MAX); + + r = mac_selinux_unit_access_check( + u, message, + job_type_to_access_method(job_type), + error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "s", &smode); + if (r < 0) + return r; + + mode = job_mode_from_string(smode); + if (mode < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Job mode %s invalid", smode); + + if (reload_if_possible) + verb = strjoin("reload-or-", job_type_to_string(job_type), NULL); + else + verb = strdup(job_type_to_string(job_type)); + if (!verb) + return -ENOMEM; + + r = bus_verify_manage_units_async_full( + u, + verb, + CAP_SYS_ADMIN, + job_type < _JOB_TYPE_MAX ? polkit_message_for_job[job_type] : NULL, + true, + message, + error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + return bus_unit_queue_job(message, u, job_type, mode, reload_if_possible, error); +} + +static int method_start(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_START, false, error); +} + +static int method_stop(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_STOP, false, error); +} + +static int method_reload(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_RELOAD, false, error); +} + +static int method_restart(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_RESTART, false, error); +} + +static int method_try_restart(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_TRY_RESTART, false, error); +} + +static int method_reload_or_restart(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_RESTART, true, error); +} + +static int method_reload_or_try_restart(sd_bus_message *message, void *userdata, sd_bus_error *error) { + return bus_unit_method_start_generic(message, userdata, JOB_TRY_RESTART, true, error); +} + +int bus_unit_method_kill(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Unit *u = userdata; + const char *swho; + int32_t signo; + KillWho who; + int r; + + assert(message); + assert(u); + + r = mac_selinux_unit_access_check(u, message, "stop", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "si", &swho, &signo); + if (r < 0) + return r; + + if (isempty(swho)) + who = KILL_ALL; + else { + who = kill_who_from_string(swho); + if (who < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid who argument %s", swho); + } + + if (!SIGNAL_VALID(signo)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Signal number out of range."); + + r = bus_verify_manage_units_async_full( + u, + "kill", + CAP_KILL, + N_("Authentication is required to kill '$(unit)'."), + true, + message, + error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = unit_kill(u, who, signo, error); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +int bus_unit_method_reset_failed(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Unit *u = userdata; + int r; + + assert(message); + assert(u); + + r = mac_selinux_unit_access_check(u, message, "reload", error); + if (r < 0) + return r; + + r = bus_verify_manage_units_async_full( + u, + "reset-failed", + CAP_SYS_ADMIN, + N_("Authentication is required to reset the \"failed\" state of '$(unit)'."), + true, + message, + error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + unit_reset_failed(u); + + return sd_bus_reply_method_return(message, NULL); +} + +int bus_unit_method_set_properties(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Unit *u = userdata; + int runtime, r; + + assert(message); + assert(u); + + r = mac_selinux_unit_access_check(u, message, "start", error); + if (r < 0) + return r; + + r = sd_bus_message_read(message, "b", &runtime); + if (r < 0) + return r; + + r = bus_verify_manage_units_async_full( + u, + "set-property", + CAP_SYS_ADMIN, + N_("Authentication is required to set properties on '$(unit)'."), + true, + message, + error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = bus_unit_set_properties(u, message, runtime ? UNIT_RUNTIME : UNIT_PERSISTENT, true, error); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +int bus_unit_method_ref(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Unit *u = userdata; + int r; + + assert(message); + assert(u); + + r = mac_selinux_unit_access_check(u, message, "start", error); + if (r < 0) + return r; + + r = bus_verify_manage_units_async_full( + u, + "ref", + CAP_SYS_ADMIN, + NULL, + false, + message, + error); + if (r < 0) + return r; + if (r == 0) + return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ + + r = bus_unit_track_add_sender(u, message); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +int bus_unit_method_unref(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Unit *u = userdata; + int r; + + assert(message); + assert(u); + + r = bus_unit_track_remove_sender(u, message); + if (r == -EUNATCH) + return sd_bus_error_setf(error, BUS_ERROR_NOT_REFERENCED, "Unit has not been referenced yet."); + if (r < 0) + return r; + + return sd_bus_reply_method_return(message, NULL); +} + +const sd_bus_vtable bus_unit_vtable[] = { + SD_BUS_VTABLE_START(0), + + SD_BUS_PROPERTY("Id", "s", NULL, offsetof(Unit, id), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Names", "as", property_get_names, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Following", "s", property_get_following, 0, 0), + SD_BUS_PROPERTY("Requires", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUIRES]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Requisite", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUISITE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Wants", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_WANTS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("BindsTo", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_BINDS_TO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PartOf", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_PART_OF]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RequiredBy", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUIRED_BY]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RequisiteOf", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUISITE_OF]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("WantedBy", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_WANTED_BY]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("BoundBy", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_BOUND_BY]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ConsistsOf", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_CONSISTS_OF]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Conflicts", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_CONFLICTS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ConflictedBy", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_CONFLICTED_BY]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Before", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_BEFORE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("After", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_AFTER]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("OnFailure", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_ON_FAILURE]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Triggers", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_TRIGGERS]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("TriggeredBy", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_TRIGGERED_BY]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("PropagatesReloadTo", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_PROPAGATES_RELOAD_TO]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ReloadPropagatedFrom", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_RELOAD_PROPAGATED_FROM]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("JoinsNamespaceOf", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_JOINS_NAMESPACE_OF]), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RequiresMountsFor", "as", NULL, offsetof(Unit, requires_mounts_for), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Documentation", "as", NULL, offsetof(Unit, documentation), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Description", "s", property_get_description, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("LoadState", "s", property_get_load_state, offsetof(Unit, load_state), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ActiveState", "s", property_get_active_state, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("SubState", "s", property_get_sub_state, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("FragmentPath", "s", NULL, offsetof(Unit, fragment_path), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("SourcePath", "s", NULL, offsetof(Unit, source_path), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DropInPaths", "as", NULL, offsetof(Unit, dropin_paths), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("UnitFileState", "s", property_get_unit_file_state, 0, 0), + SD_BUS_PROPERTY("UnitFilePreset", "s", property_get_unit_file_preset, 0, 0), + BUS_PROPERTY_DUAL_TIMESTAMP("StateChangeTimestamp", offsetof(Unit, state_change_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("InactiveExitTimestamp", offsetof(Unit, inactive_exit_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("ActiveEnterTimestamp", offsetof(Unit, active_enter_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("ActiveExitTimestamp", offsetof(Unit, active_exit_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("InactiveEnterTimestamp", offsetof(Unit, inactive_enter_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("CanStart", "b", property_get_can_start, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CanStop", "b", property_get_can_stop, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CanReload", "b", property_get_can_reload, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("CanIsolate", "b", property_get_can_isolate, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Job", "(uo)", property_get_job, 0, SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("StopWhenUnneeded", "b", bus_property_get_bool, offsetof(Unit, stop_when_unneeded), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RefuseManualStart", "b", bus_property_get_bool, offsetof(Unit, refuse_manual_start), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RefuseManualStop", "b", bus_property_get_bool, offsetof(Unit, refuse_manual_stop), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("AllowIsolate", "b", bus_property_get_bool, offsetof(Unit, allow_isolate), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("DefaultDependencies", "b", bus_property_get_bool, offsetof(Unit, default_dependencies), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("OnFailureJobMode", "s", property_get_job_mode, offsetof(Unit, on_failure_job_mode), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("IgnoreOnIsolate", "b", bus_property_get_bool, offsetof(Unit, ignore_on_isolate), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("NeedDaemonReload", "b", property_get_need_daemon_reload, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("JobTimeoutUSec", "t", bus_property_get_usec, offsetof(Unit, job_timeout), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("JobTimeoutAction", "s", property_get_emergency_action, offsetof(Unit, job_timeout_action), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("JobTimeoutRebootArgument", "s", NULL, offsetof(Unit, job_timeout_reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("ConditionResult", "b", bus_property_get_bool, offsetof(Unit, condition_result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("AssertResult", "b", bus_property_get_bool, offsetof(Unit, assert_result), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("ConditionTimestamp", offsetof(Unit, condition_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + BUS_PROPERTY_DUAL_TIMESTAMP("AssertTimestamp", offsetof(Unit, assert_timestamp), SD_BUS_VTABLE_PROPERTY_EMITS_CHANGE), + SD_BUS_PROPERTY("Conditions", "a(sbbsi)", property_get_conditions, offsetof(Unit, conditions), 0), + SD_BUS_PROPERTY("Asserts", "a(sbbsi)", property_get_conditions, offsetof(Unit, asserts), 0), + SD_BUS_PROPERTY("LoadError", "(ss)", property_get_load_error, 0, SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Transient", "b", bus_property_get_bool, offsetof(Unit, transient), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("Perpetual", "b", bus_property_get_bool, offsetof(Unit, perpetual), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StartLimitIntervalSec", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Unit, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("StartLimitAction", "s", property_get_emergency_action, offsetof(Unit, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Unit, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST), + SD_BUS_PROPERTY("InvocationID", "ay", bus_property_get_id128, offsetof(Unit, invocation_id), 0), + + SD_BUS_METHOD("Start", "s", "o", method_start, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Stop", "s", "o", method_stop, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Reload", "s", "o", method_reload, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Restart", "s", "o", method_restart, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("TryRestart", "s", "o", method_try_restart, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReloadOrRestart", "s", "o", method_reload_or_restart, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ReloadOrTryRestart", "s", "o", method_reload_or_try_restart, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Kill", "si", NULL, bus_unit_method_kill, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("ResetFailed", NULL, NULL, bus_unit_method_reset_failed, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("SetProperties", "ba(sv)", NULL, bus_unit_method_set_properties, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Ref", NULL, NULL, bus_unit_method_ref, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_METHOD("Unref", NULL, NULL, bus_unit_method_unref, SD_BUS_VTABLE_UNPRIVILEGED), + + /* Obsolete properties or obsolete alias names */ + SD_BUS_PROPERTY("RequiresOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequisiteOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequiredByOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("RequisiteOfOverridable", "as", property_get_obsolete_dependencies, 0, SD_BUS_VTABLE_HIDDEN), + SD_BUS_PROPERTY("StartLimitInterval", "t", bus_property_get_usec, offsetof(Unit, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST|SD_BUS_VTABLE_HIDDEN), + SD_BUS_VTABLE_END +}; + +static int property_get_slice( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + + assert(bus); + assert(reply); + assert(u); + + return sd_bus_message_append(reply, "s", unit_slice_name(u)); +} + +static int property_get_current_memory( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + uint64_t sz = (uint64_t) -1; + Unit *u = userdata; + int r; + + assert(bus); + assert(reply); + assert(u); + + r = unit_get_memory_current(u, &sz); + if (r < 0 && r != -ENODATA) + log_unit_warning_errno(u, r, "Failed to get memory.usage_in_bytes attribute: %m"); + + return sd_bus_message_append(reply, "t", sz); +} + +static int property_get_current_tasks( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + uint64_t cn = (uint64_t) -1; + Unit *u = userdata; + int r; + + assert(bus); + assert(reply); + assert(u); + + r = unit_get_tasks_current(u, &cn); + if (r < 0 && r != -ENODATA) + log_unit_warning_errno(u, r, "Failed to get pids.current attribute: %m"); + + return sd_bus_message_append(reply, "t", cn); +} + +static int property_get_cpu_usage( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + nsec_t ns = (nsec_t) -1; + Unit *u = userdata; + int r; + + assert(bus); + assert(reply); + assert(u); + + r = unit_get_cpu_usage(u, &ns); + if (r < 0 && r != -ENODATA) + log_unit_warning_errno(u, r, "Failed to get cpuacct.usage attribute: %m"); + + return sd_bus_message_append(reply, "t", ns); +} + +static int property_get_cgroup( + sd_bus *bus, + const char *path, + const char *interface, + const char *property, + sd_bus_message *reply, + void *userdata, + sd_bus_error *error) { + + Unit *u = userdata; + const char *t; + + assert(bus); + assert(reply); + assert(u); + + /* Three cases: a) u->cgroup_path is NULL, in which case the + * unit has no control group, which we report as the empty + * string. b) u->cgroup_path is the empty string, which + * indicates the root cgroup, which we report as "/". c) all + * other cases we report as-is. */ + + if (u->cgroup_path) + t = isempty(u->cgroup_path) ? "/" : u->cgroup_path; + else + t = ""; + + return sd_bus_message_append(reply, "s", t); +} + +static int append_process(sd_bus_message *reply, const char *p, pid_t pid, Set *pids) { + _cleanup_free_ char *buf = NULL, *cmdline = NULL; + int r; + + assert(reply); + assert(pid > 0); + + r = set_put(pids, PID_TO_PTR(pid)); + if (r == -EEXIST || r == 0) + return 0; + if (r < 0) + return r; + + if (!p) { + r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &buf); + if (r == -ESRCH) + return 0; + if (r < 0) + return r; + + p = buf; + } + + (void) get_process_cmdline(pid, 0, true, &cmdline); + + return sd_bus_message_append(reply, + "(sus)", + p, + (uint32_t) pid, + cmdline); +} + +static int append_cgroup(sd_bus_message *reply, const char *p, Set *pids) { + _cleanup_closedir_ DIR *d = NULL; + _cleanup_fclose_ FILE *f = NULL; + int r; + + assert(reply); + assert(p); + + r = cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, p, &f); + if (r == ENOENT) + return 0; + if (r < 0) + return r; + + for (;;) { + pid_t pid; + + r = cg_read_pid(f, &pid); + if (r < 0) + return r; + if (r == 0) + break; + + if (is_kernel_thread(pid) > 0) + continue; + + r = append_process(reply, p, pid, pids); + if (r < 0) + return r; + } + + r = cg_enumerate_subgroups(SYSTEMD_CGROUP_CONTROLLER, p, &d); + if (r == -ENOENT) + return 0; + if (r < 0) + return r; + + for (;;) { + _cleanup_free_ char *g = NULL, *j = NULL; + + r = cg_read_subgroup(d, &g); + if (r < 0) + return r; + if (r == 0) + break; + + j = strjoin(p, "/", g, NULL); + if (!j) + return -ENOMEM; + + r = append_cgroup(reply, j, pids); + if (r < 0) + return r; + } + + return 0; +} + +int bus_unit_method_get_processes(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + _cleanup_(set_freep) Set *pids = NULL; + Unit *u = userdata; + pid_t pid; + int r; + + assert(message); + + pids = set_new(NULL); + if (!pids) + return -ENOMEM; + + r = sd_bus_message_new_method_return(message, &reply); + if (r < 0) + return r; + + r = sd_bus_message_open_container(reply, 'a', "(sus)"); + if (r < 0) + return r; + + if (u->cgroup_path) { + r = append_cgroup(reply, u->cgroup_path, pids); + if (r < 0) + return r; + } + + /* The main and control pids might live outside of the cgroup, hence fetch them separately */ + pid = unit_main_pid(u); + if (pid > 0) { + r = append_process(reply, NULL, pid, pids); + if (r < 0) + return r; + } + + pid = unit_control_pid(u); + if (pid > 0) { + r = append_process(reply, NULL, pid, pids); + if (r < 0) + return r; + } + + r = sd_bus_message_close_container(reply); + if (r < 0) + return r; + + return sd_bus_send(NULL, reply, NULL); +} + +const sd_bus_vtable bus_unit_cgroup_vtable[] = { + SD_BUS_VTABLE_START(0), + SD_BUS_PROPERTY("Slice", "s", property_get_slice, 0, 0), + SD_BUS_PROPERTY("ControlGroup", "s", property_get_cgroup, 0, 0), + SD_BUS_PROPERTY("MemoryCurrent", "t", property_get_current_memory, 0, 0), + SD_BUS_PROPERTY("CPUUsageNSec", "t", property_get_cpu_usage, 0, 0), + SD_BUS_PROPERTY("TasksCurrent", "t", property_get_current_tasks, 0, 0), + SD_BUS_METHOD("GetProcesses", NULL, "a(sus)", bus_unit_method_get_processes, SD_BUS_VTABLE_UNPRIVILEGED), + SD_BUS_VTABLE_END +}; + +static int send_new_signal(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_free_ char *p = NULL; + Unit *u = userdata; + int r; + + assert(bus); + assert(u); + + p = unit_dbus_path(u); + if (!p) + return -ENOMEM; + + r = sd_bus_message_new_signal( + bus, + &m, + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "UnitNew"); + if (r < 0) + return r; + + r = sd_bus_message_append(m, "so", u->id, p); + if (r < 0) + return r; + + return sd_bus_send(bus, m, NULL); +} + +static int send_changed_signal(sd_bus *bus, void *userdata) { + _cleanup_free_ char *p = NULL; + Unit *u = userdata; + int r; + + assert(bus); + assert(u); + + p = unit_dbus_path(u); + if (!p) + return -ENOMEM; + + /* Send a properties changed signal. First for the specific + * type, then for the generic unit. The clients may rely on + * this order to get atomic behavior if needed. */ + + r = sd_bus_emit_properties_changed_strv( + bus, p, + unit_dbus_interface_from_type(u->type), + NULL); + if (r < 0) + return r; + + return sd_bus_emit_properties_changed_strv( + bus, p, + "org.freedesktop.systemd1.Unit", + NULL); +} + +void bus_unit_send_change_signal(Unit *u) { + int r; + assert(u); + + if (u->in_dbus_queue) { + LIST_REMOVE(dbus_queue, u->manager->dbus_unit_queue, u); + u->in_dbus_queue = false; + } + + if (!u->id) + return; + + r = bus_foreach_bus(u->manager, NULL, u->sent_dbus_new_signal ? send_changed_signal : send_new_signal, u); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to send unit change signal for %s: %m", u->id); + + u->sent_dbus_new_signal = true; +} + +static int send_removed_signal(sd_bus *bus, void *userdata) { + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; + _cleanup_free_ char *p = NULL; + Unit *u = userdata; + int r; + + assert(bus); + assert(u); + + p = unit_dbus_path(u); + if (!p) + return -ENOMEM; + + r = sd_bus_message_new_signal( + bus, + &m, + "/org/freedesktop/systemd1", + "org.freedesktop.systemd1.Manager", + "UnitRemoved"); + if (r < 0) + return r; + + r = sd_bus_message_append(m, "so", u->id, p); + if (r < 0) + return r; + + return sd_bus_send(bus, m, NULL); +} + +void bus_unit_send_removed_signal(Unit *u) { + int r; + assert(u); + + if (!u->sent_dbus_new_signal || u->in_dbus_queue) + bus_unit_send_change_signal(u); + + if (!u->id) + return; + + r = bus_foreach_bus(u->manager, NULL, send_removed_signal, u); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to send unit remove signal for %s: %m", u->id); +} + +int bus_unit_queue_job( + sd_bus_message *message, + Unit *u, + JobType type, + JobMode mode, + bool reload_if_possible, + sd_bus_error *error) { + + _cleanup_free_ char *path = NULL; + Job *j; + int r; + + assert(message); + assert(u); + assert(type >= 0 && type < _JOB_TYPE_MAX); + assert(mode >= 0 && mode < _JOB_MODE_MAX); + + r = mac_selinux_unit_access_check( + u, message, + job_type_to_access_method(type), + error); + if (r < 0) + return r; + + if (reload_if_possible && unit_can_reload(u)) { + if (type == JOB_RESTART) + type = JOB_RELOAD_OR_START; + else if (type == JOB_TRY_RESTART) + type = JOB_TRY_RELOAD; + } + + if (type == JOB_STOP && + (u->load_state == UNIT_NOT_FOUND || u->load_state == UNIT_ERROR) && + unit_active_state(u) == UNIT_INACTIVE) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s not loaded.", u->id); + + if ((type == JOB_START && u->refuse_manual_start) || + (type == JOB_STOP && u->refuse_manual_stop) || + ((type == JOB_RESTART || type == JOB_TRY_RESTART) && (u->refuse_manual_start || u->refuse_manual_stop)) || + (type == JOB_RELOAD_OR_START && job_type_collapse(type, u) == JOB_START && u->refuse_manual_start)) + return sd_bus_error_setf(error, BUS_ERROR_ONLY_BY_DEPENDENCY, "Operation refused, unit %s may be requested by dependency only.", u->id); + + r = manager_add_job(u->manager, type, u, mode, error, &j); + if (r < 0) + return r; + + if (sd_bus_message_get_bus(message) == u->manager->api_bus) { + if (!j->clients) { + r = sd_bus_track_new(sd_bus_message_get_bus(message), &j->clients, NULL, NULL); + if (r < 0) + return r; + } + + r = sd_bus_track_add_sender(j->clients, message); + if (r < 0) + return r; + } + + path = job_dbus_path(j); + if (!path) + return -ENOMEM; + + return sd_bus_reply_method_return(message, "o", path); +} + +static int bus_unit_set_transient_property( + Unit *u, + const char *name, + sd_bus_message *message, + UnitSetPropertiesMode mode, + sd_bus_error *error) { + + int r; + + assert(u); + assert(name); + assert(message); + + if (streq(name, "Description")) { + const char *d; + + r = sd_bus_message_read(message, "s", &d); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + r = unit_set_description(u, d); + if (r < 0) + return r; + + unit_write_drop_in_format(u, mode, name, "[Unit]\nDescription=%s", d); + } + + return 1; + + } else if (streq(name, "DefaultDependencies")) { + int b; + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) { + u->default_dependencies = b; + unit_write_drop_in_format(u, mode, name, "[Unit]\nDefaultDependencies=%s", yes_no(b)); + } + + return 1; + + } else if (streq(name, "Slice")) { + Unit *slice; + const char *s; + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "The slice property is only available for units with control groups."); + if (u->type == UNIT_SLICE) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Slice may not be set for slice units."); + if (unit_has_name(u, SPECIAL_INIT_SCOPE)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Cannot set slice for init.scope"); + + r = sd_bus_message_read(message, "s", &s); + if (r < 0) + return r; + + if (!unit_name_is_valid(s, UNIT_NAME_PLAIN)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid unit name '%s'", s); + + /* Note that we do not dispatch the load queue here yet, as we don't want our own transient unit to be + * loaded while we are still setting it up. Or in other words, we use manager_load_unit_prepare() + * instead of manager_load_unit() on purpose, here. */ + r = manager_load_unit_prepare(u->manager, s, NULL, error, &slice); + if (r < 0) + return r; + + if (slice->type != UNIT_SLICE) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Unit name '%s' is not a slice", s); + + if (mode != UNIT_CHECK) { + r = unit_set_slice(u, slice); + if (r < 0) + return r; + + unit_write_drop_in_private_format(u, mode, name, "Slice=%s", s); + } + + return 1; + + } else if (STR_IN_SET(name, + "Requires", "RequiresOverridable", + "Requisite", "RequisiteOverridable", + "Wants", + "BindsTo", + "Conflicts", + "Before", "After", + "OnFailure", + "PropagatesReloadTo", "ReloadPropagatedFrom", + "PartOf")) { + + UnitDependency d; + const char *other; + + if (streq(name, "RequiresOverridable")) + d = UNIT_REQUIRES; /* redirect for obsolete unit dependency type */ + else if (streq(name, "RequisiteOverridable")) + d = UNIT_REQUISITE; /* same here */ + else { + d = unit_dependency_from_string(name); + if (d < 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid unit dependency: %s", name); + } + + r = sd_bus_message_enter_container(message, 'a', "s"); + if (r < 0) + return r; + + while ((r = sd_bus_message_read(message, "s", &other)) > 0) { + if (!unit_name_is_valid(other, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) + return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid unit name %s", other); + + if (mode != UNIT_CHECK) { + _cleanup_free_ char *label = NULL; + + r = unit_add_dependency_by_name(u, d, other, NULL, true); + if (r < 0) + return r; + + label = strjoin(name, "-", other, NULL); + if (!label) + return -ENOMEM; + + unit_write_drop_in_format(u, mode, label, "[Unit]\n%s=%s", name, other); + } + + } + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + return 1; + + } else if (streq(name, "AddRef")) { + + int b; + + /* Why is this called "AddRef" rather than just "Ref", or "Reference"? There's already a "Ref()" method + * on the Unit interface, and it's probably not a good idea to expose a property and a method on the + * same interface (well, strictly speaking AddRef isn't exposed as full property, we just read it for + * transient units, but still). And "References" and "ReferencedBy" is already used as unit reference + * dependency type, hence let's not confuse things with that. + * + * Note that we don't acually add the reference to the bus track. We do that only after the setup of + * the transient unit is complete, so that setting this property multiple times in the same transient + * unit creation call doesn't count as individual references. */ + + r = sd_bus_message_read(message, "b", &b); + if (r < 0) + return r; + + if (mode != UNIT_CHECK) + u->bus_track_add = b; + + return 1; + } + + return 0; +} + +int bus_unit_set_properties( + Unit *u, + sd_bus_message *message, + UnitSetPropertiesMode mode, + bool commit, + sd_bus_error *error) { + + bool for_real = false; + unsigned n = 0; + int r; + + assert(u); + assert(message); + + /* We iterate through the array twice. First run we just check + * if all passed data is valid, second run actually applies + * it. This is to implement transaction-like behaviour without + * actually providing full transactions. */ + + r = sd_bus_message_enter_container(message, 'a', "(sv)"); + if (r < 0) + return r; + + for (;;) { + const char *name; + + r = sd_bus_message_enter_container(message, 'r', "sv"); + if (r < 0) + return r; + if (r == 0) { + if (for_real || mode == UNIT_CHECK) + break; + + /* Reached EOF. Let's try again, and this time for realz... */ + r = sd_bus_message_rewind(message, false); + if (r < 0) + return r; + + for_real = true; + continue; + } + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) + return r; + + if (!UNIT_VTABLE(u)->bus_set_property) + return sd_bus_error_setf(error, SD_BUS_ERROR_PROPERTY_READ_ONLY, "Objects of this type do not support setting properties."); + + r = sd_bus_message_enter_container(message, 'v', NULL); + if (r < 0) + return r; + + r = UNIT_VTABLE(u)->bus_set_property(u, name, message, for_real ? mode : UNIT_CHECK, error); + if (r == 0 && u->transient && u->load_state == UNIT_STUB) + r = bus_unit_set_transient_property(u, name, message, for_real ? mode : UNIT_CHECK, error); + if (r < 0) + return r; + if (r == 0) + return sd_bus_error_setf(error, SD_BUS_ERROR_PROPERTY_READ_ONLY, "Cannot set property %s, or unknown property.", name); + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + n += for_real; + } + + r = sd_bus_message_exit_container(message); + if (r < 0) + return r; + + if (commit && n > 0 && UNIT_VTABLE(u)->bus_commit_properties) + UNIT_VTABLE(u)->bus_commit_properties(u); + + return n; +} + +int bus_unit_check_load_state(Unit *u, sd_bus_error *error) { + assert(u); + + if (u->load_state == UNIT_LOADED) + return 0; + + /* Give a better description of the unit error when + * possible. Note that in the case of UNIT_MASKED, load_error + * is not set. */ + if (u->load_state == UNIT_MASKED) + return sd_bus_error_setf(error, BUS_ERROR_UNIT_MASKED, "Unit %s is masked.", u->id); + + if (u->load_state == UNIT_NOT_FOUND) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s not found.", u->id); + + return sd_bus_error_set_errnof(error, u->load_error, "Unit %s is not loaded properly: %m.", u->id); +} + +static int bus_track_handler(sd_bus_track *t, void *userdata) { + Unit *u = userdata; + + assert(t); + assert(u); + + u->bus_track = sd_bus_track_unref(u->bus_track); /* make sure we aren't called again */ + + unit_add_to_gc_queue(u); + return 0; +} + +static int allocate_bus_track(Unit *u) { + int r; + + assert(u); + + if (u->bus_track) + return 0; + + r = sd_bus_track_new(u->manager->api_bus, &u->bus_track, bus_track_handler, u); + if (r < 0) + return r; + + r = sd_bus_track_set_recursive(u->bus_track, true); + if (r < 0) { + u->bus_track = sd_bus_track_unref(u->bus_track); + return r; + } + + return 0; +} + +int bus_unit_track_add_name(Unit *u, const char *name) { + int r; + + assert(u); + + r = allocate_bus_track(u); + if (r < 0) + return r; + + return sd_bus_track_add_name(u->bus_track, name); +} + +int bus_unit_track_add_sender(Unit *u, sd_bus_message *m) { + int r; + + assert(u); + + r = allocate_bus_track(u); + if (r < 0) + return r; + + return sd_bus_track_add_sender(u->bus_track, m); +} + +int bus_unit_track_remove_sender(Unit *u, sd_bus_message *m) { + assert(u); + + /* If we haven't allocated the bus track object yet, then there's definitely no reference taken yet, return an + * error */ + if (!u->bus_track) + return -EUNATCH; + + return sd_bus_track_remove_sender(u->bus_track, m); +} diff --git a/src/grp-system/libcore/src/dbus-unit.h b/src/grp-system/libcore/src/dbus-unit.h new file mode 100644 index 0000000000..5133bec287 --- /dev/null +++ b/src/grp-system/libcore/src/dbus-unit.h @@ -0,0 +1,47 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/unit.h" + +extern const sd_bus_vtable bus_unit_vtable[]; +extern const sd_bus_vtable bus_unit_cgroup_vtable[]; + +void bus_unit_send_change_signal(Unit *u); +void bus_unit_send_removed_signal(Unit *u); + +int bus_unit_method_start_generic(sd_bus_message *message, Unit *u, JobType job_type, bool reload_if_possible, sd_bus_error *error); +int bus_unit_method_kill(sd_bus_message *message, void *userdata, sd_bus_error *error); +int bus_unit_method_reset_failed(sd_bus_message *message, void *userdata, sd_bus_error *error); + +int bus_unit_set_properties(Unit *u, sd_bus_message *message, UnitSetPropertiesMode mode, bool commit, sd_bus_error *error); +int bus_unit_method_set_properties(sd_bus_message *message, void *userdata, sd_bus_error *error); +int bus_unit_method_get_processes(sd_bus_message *message, void *userdata, sd_bus_error *error); +int bus_unit_method_ref(sd_bus_message *message, void *userdata, sd_bus_error *error); +int bus_unit_method_unref(sd_bus_message *message, void *userdata, sd_bus_error *error); + +int bus_unit_queue_job(sd_bus_message *message, Unit *u, JobType type, JobMode mode, bool reload_if_possible, sd_bus_error *error); +int bus_unit_check_load_state(Unit *u, sd_bus_error *error); + +int bus_unit_track_add_name(Unit *u, const char *name); +int bus_unit_track_add_sender(Unit *u, sd_bus_message *m); +int bus_unit_track_remove_sender(Unit *u, sd_bus_message *m); diff --git a/src/grp-system/libcore/src/dbus.c b/src/grp-system/libcore/src/dbus.c new file mode 100644 index 0000000000..77b5dc81cb --- /dev/null +++ b/src/grp-system/libcore/src/dbus.c @@ -0,0 +1,1237 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/epoll.h> +#include <unistd.h> + +#include <systemd/sd-bus.h> + +#include "core/dbus-manager.h" +#include "sd-bus/bus-common-errors.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-internal.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/strxcpyx.h" +#include "systemd-basic/user-util.h" + +#include "dbus-cgroup.h" +#include "dbus-execute.h" +#include "dbus-job.h" +#include "dbus-kill.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "selinux-access.h" + +#define CONNECTIONS_MAX 4096 + +static void destroy_bus(Manager *m, sd_bus **bus); + +int bus_send_queued_message(Manager *m) { + int r; + + assert(m); + + if (!m->queued_message) + return 0; + + /* If we cannot get rid of this message we won't dispatch any + * D-Bus messages, so that we won't end up wanting to queue + * another message. */ + + r = sd_bus_send(NULL, m->queued_message, NULL); + if (r < 0) + log_warning_errno(r, "Failed to send queued message: %m"); + + m->queued_message = sd_bus_message_unref(m->queued_message); + + return 0; +} + +int bus_forward_agent_released(Manager *m, const char *path) { + int r; + + assert(m); + assert(path); + + if (!MANAGER_IS_SYSTEM(m)) + return 0; + + if (!m->system_bus) + return 0; + + /* If we are running a system instance we forward the agent message on the system bus, so that the user + * instances get notified about this, too */ + + r = sd_bus_emit_signal(m->system_bus, + "/org/freedesktop/systemd1/agent", + "org.freedesktop.systemd1.Agent", + "Released", + "s", path); + if (r < 0) + return log_warning_errno(r, "Failed to propagate agent release message: %m"); + + return 1; +} + +static int signal_agent_released(sd_bus_message *message, void *userdata, sd_bus_error *error) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + Manager *m = userdata; + const char *cgroup; + uid_t sender_uid; + int r; + + assert(message); + assert(m); + + /* only accept org.freedesktop.systemd1.Agent from UID=0 */ + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_EUID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_euid(creds, &sender_uid); + if (r < 0 || sender_uid != 0) + return 0; + + /* parse 'cgroup-empty' notification */ + r = sd_bus_message_read(message, "s", &cgroup); + if (r < 0) { + bus_log_parse_error(r); + return 0; + } + + manager_notify_cgroup_empty(m, cgroup); + return 0; +} + +static int signal_disconnected(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + sd_bus *bus; + + assert(message); + assert(m); + assert_se(bus = sd_bus_message_get_bus(message)); + + if (bus == m->api_bus) + destroy_bus(m, &m->api_bus); + if (bus == m->system_bus) + destroy_bus(m, &m->system_bus); + if (set_remove(m->private_buses, bus)) { + log_debug("Got disconnect on private connection."); + destroy_bus(m, &bus); + } + + return 0; +} + +static int signal_activation_request(sd_bus_message *message, void *userdata, sd_bus_error *ret_error) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; + Manager *m = userdata; + const char *name; + Unit *u; + int r; + + assert(message); + assert(m); + + r = sd_bus_message_read(message, "s", &name); + if (r < 0) { + bus_log_parse_error(r); + return 0; + } + + if (manager_unit_inactive_or_pending(m, SPECIAL_DBUS_SERVICE) || + manager_unit_inactive_or_pending(m, SPECIAL_DBUS_SOCKET)) { + r = sd_bus_error_setf(&error, BUS_ERROR_SHUTTING_DOWN, "Refusing activation, D-Bus is shutting down."); + goto failed; + } + + r = manager_load_unit(m, name, NULL, &error, &u); + if (r < 0) + goto failed; + + if (u->refuse_manual_start) { + r = sd_bus_error_setf(&error, BUS_ERROR_ONLY_BY_DEPENDENCY, "Operation refused, %s may be requested by dependency only.", u->id); + goto failed; + } + + r = manager_add_job(m, JOB_START, u, JOB_REPLACE, &error, NULL); + if (r < 0) + goto failed; + + /* Successfully queued, that's it for us */ + return 0; + +failed: + if (!sd_bus_error_is_set(&error)) + sd_bus_error_set_errno(&error, r); + + log_debug("D-Bus activation failed for %s: %s", name, bus_error_message(&error, r)); + + r = sd_bus_message_new_signal(sd_bus_message_get_bus(message), &reply, "/org/freedesktop/systemd1", "org.freedesktop.systemd1.Activator", "ActivationFailure"); + if (r < 0) { + bus_log_create_error(r); + return 0; + } + + r = sd_bus_message_append(reply, "sss", name, error.name, error.message); + if (r < 0) { + bus_log_create_error(r); + return 0; + } + + r = sd_bus_send_to(NULL, reply, "org.freedesktop.DBus", NULL); + if (r < 0) + return log_error_errno(r, "Failed to respond with to bus activation request: %m"); + + return 0; +} + +#ifdef HAVE_SELINUX +static int mac_selinux_filter(sd_bus_message *message, void *userdata, sd_bus_error *error) { + Manager *m = userdata; + const char *verb, *path; + Unit *u = NULL; + Job *j; + int r; + + assert(message); + + /* Our own method calls are all protected individually with + * selinux checks, but the built-in interfaces need to be + * protected too. */ + + if (sd_bus_message_is_method_call(message, "org.freedesktop.DBus.Properties", "Set")) + verb = "reload"; + else if (sd_bus_message_is_method_call(message, "org.freedesktop.DBus.Introspectable", NULL) || + sd_bus_message_is_method_call(message, "org.freedesktop.DBus.Properties", NULL) || + sd_bus_message_is_method_call(message, "org.freedesktop.DBus.ObjectManager", NULL) || + sd_bus_message_is_method_call(message, "org.freedesktop.DBus.Peer", NULL)) + verb = "status"; + else + return 0; + + path = sd_bus_message_get_path(message); + + if (object_path_startswith("/org/freedesktop/systemd1", path)) { + + r = mac_selinux_access_check(message, verb, error); + if (r < 0) + return r; + + return 0; + } + + if (streq_ptr(path, "/org/freedesktop/systemd1/unit/self")) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + pid_t pid; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return 0; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return 0; + + u = manager_get_unit_by_pid(m, pid); + } else { + r = manager_get_job_from_dbus_path(m, path, &j); + if (r >= 0) + u = j->unit; + else + manager_load_unit_from_dbus_path(m, path, NULL, &u); + } + + if (!u) + return 0; + + r = mac_selinux_unit_access_check(u, message, verb, error); + if (r < 0) + return r; + + return 0; +} +#endif + +static int bus_job_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + Job *j; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = manager_get_job_from_dbus_path(m, path, &j); + if (r < 0) + return 0; + + *found = j; + return 1; +} + +static int find_unit(Manager *m, sd_bus *bus, const char *path, Unit **unit, sd_bus_error *error) { + Unit *u; + int r; + + assert(m); + assert(bus); + assert(path); + + if (streq_ptr(path, "/org/freedesktop/systemd1/unit/self")) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + sd_bus_message *message; + pid_t pid; + + message = sd_bus_get_current_message(bus); + if (!message) + return 0; + + r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_PID, &creds); + if (r < 0) + return r; + + r = sd_bus_creds_get_pid(creds, &pid); + if (r < 0) + return r; + + u = manager_get_unit_by_pid(m, pid); + } else { + r = manager_load_unit_from_dbus_path(m, path, error, &u); + if (r < 0) + return 0; + } + + if (!u) + return 0; + + *unit = u; + return 1; +} + +static int bus_unit_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + return find_unit(m, bus, path, (Unit**) found, error); +} + +static int bus_unit_interface_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + Unit *u; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = find_unit(m, bus, path, &u, error); + if (r <= 0) + return r; + + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) + return 0; + + *found = u; + return 1; +} + +static int bus_unit_cgroup_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + Unit *u; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = find_unit(m, bus, path, &u, error); + if (r <= 0) + return r; + + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) + return 0; + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return 0; + + *found = u; + return 1; +} + +static int bus_cgroup_context_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + CGroupContext *c; + Unit *u; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = find_unit(m, bus, path, &u, error); + if (r <= 0) + return r; + + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) + return 0; + + c = unit_get_cgroup_context(u); + if (!c) + return 0; + + *found = c; + return 1; +} + +static int bus_exec_context_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + ExecContext *c; + Unit *u; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = find_unit(m, bus, path, &u, error); + if (r <= 0) + return r; + + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) + return 0; + + c = unit_get_exec_context(u); + if (!c) + return 0; + + *found = c; + return 1; +} + +static int bus_kill_context_find(sd_bus *bus, const char *path, const char *interface, void *userdata, void **found, sd_bus_error *error) { + Manager *m = userdata; + KillContext *c; + Unit *u; + int r; + + assert(bus); + assert(path); + assert(interface); + assert(found); + assert(m); + + r = find_unit(m, bus, path, &u, error); + if (r <= 0) + return r; + + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) + return 0; + + c = unit_get_kill_context(u); + if (!c) + return 0; + + *found = c; + return 1; +} + +static int bus_job_enumerate(sd_bus *bus, const char *path, void *userdata, char ***nodes, sd_bus_error *error) { + _cleanup_free_ char **l = NULL; + Manager *m = userdata; + unsigned k = 0; + Iterator i; + Job *j; + + l = new0(char*, hashmap_size(m->jobs)+1); + if (!l) + return -ENOMEM; + + HASHMAP_FOREACH(j, m->jobs, i) { + l[k] = job_dbus_path(j); + if (!l[k]) + return -ENOMEM; + + k++; + } + + assert(hashmap_size(m->jobs) == k); + + *nodes = l; + l = NULL; + + return k; +} + +static int bus_unit_enumerate(sd_bus *bus, const char *path, void *userdata, char ***nodes, sd_bus_error *error) { + _cleanup_free_ char **l = NULL; + Manager *m = userdata; + unsigned k = 0; + Iterator i; + Unit *u; + + l = new0(char*, hashmap_size(m->units)+1); + if (!l) + return -ENOMEM; + + HASHMAP_FOREACH(u, m->units, i) { + l[k] = unit_dbus_path(u); + if (!l[k]) + return -ENOMEM; + + k++; + } + + *nodes = l; + l = NULL; + + return k; +} + +static int bus_setup_api_vtables(Manager *m, sd_bus *bus) { + UnitType t; + int r; + + assert(m); + assert(bus); + +#ifdef HAVE_SELINUX + r = sd_bus_add_filter(bus, NULL, mac_selinux_filter, m); + if (r < 0) + return log_error_errno(r, "Failed to add SELinux access filter: %m"); +#endif + + r = sd_bus_add_object_vtable(bus, NULL, "/org/freedesktop/systemd1", "org.freedesktop.systemd1.Manager", bus_manager_vtable, m); + if (r < 0) + return log_error_errno(r, "Failed to register Manager vtable: %m"); + + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/job", "org.freedesktop.systemd1.Job", bus_job_vtable, bus_job_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register Job vtable: %m"); + + r = sd_bus_add_node_enumerator(bus, NULL, "/org/freedesktop/systemd1/job", bus_job_enumerate, m); + if (r < 0) + return log_error_errno(r, "Failed to add job enumerator: %m"); + + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", "org.freedesktop.systemd1.Unit", bus_unit_vtable, bus_unit_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register Unit vtable: %m"); + + r = sd_bus_add_node_enumerator(bus, NULL, "/org/freedesktop/systemd1/unit", bus_unit_enumerate, m); + if (r < 0) + return log_error_errno(r, "Failed to add job enumerator: %m"); + + for (t = 0; t < _UNIT_TYPE_MAX; t++) { + const char *interface; + + assert_se(interface = unit_dbus_interface_from_type(t)); + + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", interface, unit_vtable[t]->bus_vtable, bus_unit_interface_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register type specific vtable for %s: %m", interface); + + if (unit_vtable[t]->cgroup_context_offset > 0) { + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", interface, bus_unit_cgroup_vtable, bus_unit_cgroup_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register control group unit vtable for %s: %m", interface); + + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", interface, bus_cgroup_vtable, bus_cgroup_context_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register control group vtable for %s: %m", interface); + } + + if (unit_vtable[t]->exec_context_offset > 0) { + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", interface, bus_exec_vtable, bus_exec_context_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register execute vtable for %s: %m", interface); + } + + if (unit_vtable[t]->kill_context_offset > 0) { + r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", interface, bus_kill_vtable, bus_kill_context_find, m); + if (r < 0) + return log_error_errno(r, "Failed to register kill vtable for %s: %m", interface); + } + } + + return 0; +} + +static int bus_setup_disconnected_match(Manager *m, sd_bus *bus) { + int r; + + assert(m); + assert(bus); + + r = sd_bus_add_match( + bus, + NULL, + "sender='org.freedesktop.DBus.Local'," + "type='signal'," + "path='/org/freedesktop/DBus/Local'," + "interface='org.freedesktop.DBus.Local'," + "member='Disconnected'", + signal_disconnected, m); + + if (r < 0) + return log_error_errno(r, "Failed to register match for Disconnected message: %m"); + + return 0; +} + +static int bus_on_connection(sd_event_source *s, int fd, uint32_t revents, void *userdata) { + _cleanup_(sd_bus_unrefp) sd_bus *bus = NULL; + _cleanup_close_ int nfd = -1; + Manager *m = userdata; + sd_id128_t id; + int r; + + assert(s); + assert(m); + + nfd = accept4(fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC); + if (nfd < 0) { + log_warning_errno(errno, "Failed to accept private connection, ignoring: %m"); + return 0; + } + + if (set_size(m->private_buses) >= CONNECTIONS_MAX) { + log_warning("Too many concurrent connections, refusing"); + return 0; + } + + r = set_ensure_allocated(&m->private_buses, NULL); + if (r < 0) { + log_oom(); + return 0; + } + + r = sd_bus_new(&bus); + if (r < 0) { + log_warning_errno(r, "Failed to allocate new private connection bus: %m"); + return 0; + } + + r = sd_bus_set_fd(bus, nfd, nfd); + if (r < 0) { + log_warning_errno(r, "Failed to set fd on new connection bus: %m"); + return 0; + } + + nfd = -1; + + r = bus_check_peercred(bus); + if (r < 0) { + log_warning_errno(r, "Incoming private connection from unprivileged client, refusing: %m"); + return 0; + } + + assert_se(sd_id128_randomize(&id) >= 0); + + r = sd_bus_set_server(bus, 1, id); + if (r < 0) { + log_warning_errno(r, "Failed to enable server support for new connection bus: %m"); + return 0; + } + + r = sd_bus_negotiate_creds(bus, 1, + SD_BUS_CREDS_PID|SD_BUS_CREDS_UID| + SD_BUS_CREDS_EUID|SD_BUS_CREDS_EFFECTIVE_CAPS| + SD_BUS_CREDS_SELINUX_CONTEXT); + if (r < 0) { + log_warning_errno(r, "Failed to enable credentials for new connection: %m"); + return 0; + } + + r = sd_bus_start(bus); + if (r < 0) { + log_warning_errno(r, "Failed to start new connection bus: %m"); + return 0; + } + + r = sd_bus_attach_event(bus, m->event, SD_EVENT_PRIORITY_NORMAL); + if (r < 0) { + log_warning_errno(r, "Failed to attach new connection bus to event loop: %m"); + return 0; + } + + r = bus_setup_disconnected_match(m, bus); + if (r < 0) + return 0; + + r = bus_setup_api_vtables(m, bus); + if (r < 0) { + log_warning_errno(r, "Failed to set up API vtables on new connection bus: %m"); + return 0; + } + + r = set_put(m->private_buses, bus); + if (r < 0) { + log_warning_errno(r, "Failed to add new connection bus to set: %m"); + return 0; + } + + bus = NULL; + + log_debug("Accepted new private connection."); + + return 0; +} + +int manager_sync_bus_names(Manager *m, sd_bus *bus) { + _cleanup_strv_free_ char **names = NULL; + const char *name; + Iterator i; + Unit *u; + int r; + + assert(m); + assert(bus); + + r = sd_bus_list_names(bus, &names, NULL); + if (r < 0) + return log_error_errno(r, "Failed to get initial list of names: %m"); + + /* We have to synchronize the current bus names with the + * list of active services. To do this, walk the list of + * all units with bus names. */ + HASHMAP_FOREACH_KEY(u, name, m->watch_bus, i) { + Service *s = SERVICE(u); + + assert(s); + + if (!streq_ptr(s->bus_name, name)) { + log_unit_warning(u, "Bus name has changed from %s → %s, ignoring.", s->bus_name, name); + continue; + } + + /* Check if a service's bus name is in the list of currently + * active names */ + if (strv_contains(names, name)) { + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + const char *unique; + + /* If it is, determine its current owner */ + r = sd_bus_get_name_creds(bus, name, SD_BUS_CREDS_UNIQUE_NAME, &creds); + if (r < 0) { + log_error_errno(r, "Failed to get bus name owner %s: %m", name); + continue; + } + + r = sd_bus_creds_get_unique_name(creds, &unique); + if (r < 0) { + log_error_errno(r, "Failed to get unique name for %s: %m", name); + continue; + } + + /* Now, let's compare that to the previous bus owner, and + * if it's still the same, all is fine, so just don't + * bother the service. Otherwise, the name has apparently + * changed, so synthesize a name owner changed signal. */ + + if (!streq_ptr(unique, s->bus_name_owner)) + UNIT_VTABLE(u)->bus_name_owner_change(u, name, s->bus_name_owner, unique); + } else { + /* So, the name we're watching is not on the bus. + * This either means it simply hasn't appeared yet, + * or it was lost during the daemon reload. + * Check if the service has a stored name owner, + * and synthesize a name loss signal in this case. */ + + if (s->bus_name_owner) + UNIT_VTABLE(u)->bus_name_owner_change(u, name, s->bus_name_owner, NULL); + } + } + + return 0; +} + +static int bus_setup_api(Manager *m, sd_bus *bus) { + Iterator i; + char *name; + Unit *u; + int r; + + assert(m); + assert(bus); + + /* Let's make sure we have enough credential bits so that we can make security and selinux decisions */ + r = sd_bus_negotiate_creds(bus, 1, + SD_BUS_CREDS_PID|SD_BUS_CREDS_UID| + SD_BUS_CREDS_EUID|SD_BUS_CREDS_EFFECTIVE_CAPS| + SD_BUS_CREDS_SELINUX_CONTEXT); + if (r < 0) + log_warning_errno(r, "Failed to enable credential passing, ignoring: %m"); + + r = bus_setup_api_vtables(m, bus); + if (r < 0) + return r; + + HASHMAP_FOREACH_KEY(u, name, m->watch_bus, i) { + r = unit_install_bus_match(u, bus, name); + if (r < 0) + log_error_errno(r, "Failed to subscribe to NameOwnerChanged signal for '%s': %m", name); + } + + r = sd_bus_add_match( + bus, + NULL, + "type='signal'," + "sender='org.freedesktop.DBus'," + "path='/org/freedesktop/DBus'," + "interface='org.freedesktop.systemd1.Activator'," + "member='ActivationRequest'", + signal_activation_request, m); + if (r < 0) + log_warning_errno(r, "Failed to subscribe to activation signal: %m"); + + /* Allow replacing of our name, to ease implementation of + * reexecution, where we keep the old connection open until + * after the new connection is set up and the name installed + * to allow clients to synchronously wait for reexecution to + * finish */ + r = sd_bus_request_name(bus,"org.freedesktop.systemd1", SD_BUS_NAME_REPLACE_EXISTING|SD_BUS_NAME_ALLOW_REPLACEMENT); + if (r < 0) + return log_error_errno(r, "Failed to register name: %m"); + + r = manager_sync_bus_names(m, bus); + if (r < 0) + return r; + + log_debug("Successfully connected to API bus."); + return 0; +} + +static int bus_init_api(Manager *m) { + _cleanup_(sd_bus_unrefp) sd_bus *bus = NULL; + int r; + + if (m->api_bus) + return 0; + + /* The API and system bus is the same if we are running in system mode */ + if (MANAGER_IS_SYSTEM(m) && m->system_bus) + bus = sd_bus_ref(m->system_bus); + else { + if (MANAGER_IS_SYSTEM(m)) + r = sd_bus_open_system(&bus); + else + r = sd_bus_open_user(&bus); + + if (r < 0) { + log_debug("Failed to connect to API bus, retrying later..."); + return 0; + } + + r = sd_bus_attach_event(bus, m->event, SD_EVENT_PRIORITY_NORMAL); + if (r < 0) { + log_error_errno(r, "Failed to attach API bus to event loop: %m"); + return 0; + } + + r = bus_setup_disconnected_match(m, bus); + if (r < 0) + return 0; + } + + r = bus_setup_api(m, bus); + if (r < 0) { + log_error_errno(r, "Failed to set up API bus: %m"); + return 0; + } + + m->api_bus = bus; + bus = NULL; + + return 0; +} + +static int bus_setup_system(Manager *m, sd_bus *bus) { + int r; + + assert(m); + assert(bus); + + /* if we are a user instance we get the Released message via the system bus */ + if (MANAGER_IS_USER(m)) { + r = sd_bus_add_match( + bus, + NULL, + "type='signal'," + "interface='org.freedesktop.systemd1.Agent'," + "member='Released'," + "path='/org/freedesktop/systemd1/agent'", + signal_agent_released, m); + if (r < 0) + log_warning_errno(r, "Failed to register Released match on system bus: %m"); + } + + log_debug("Successfully connected to system bus."); + return 0; +} + +static int bus_init_system(Manager *m) { + _cleanup_(sd_bus_unrefp) sd_bus *bus = NULL; + int r; + + if (m->system_bus) + return 0; + + /* The API and system bus is the same if we are running in system mode */ + if (MANAGER_IS_SYSTEM(m) && m->api_bus) { + m->system_bus = sd_bus_ref(m->api_bus); + return 0; + } + + r = sd_bus_open_system(&bus); + if (r < 0) { + log_debug("Failed to connect to system bus, retrying later..."); + return 0; + } + + r = bus_setup_disconnected_match(m, bus); + if (r < 0) + return 0; + + r = sd_bus_attach_event(bus, m->event, SD_EVENT_PRIORITY_NORMAL); + if (r < 0) { + log_error_errno(r, "Failed to attach system bus to event loop: %m"); + return 0; + } + + r = bus_setup_system(m, bus); + if (r < 0) { + log_error_errno(r, "Failed to set up system bus: %m"); + return 0; + } + + m->system_bus = bus; + bus = NULL; + + return 0; +} + +static int bus_init_private(Manager *m) { + _cleanup_close_ int fd = -1; + union sockaddr_union sa = { + .un.sun_family = AF_UNIX + }; + sd_event_source *s; + socklen_t salen; + int r; + + assert(m); + + if (m->private_listen_fd >= 0) + return 0; + + if (MANAGER_IS_SYSTEM(m)) { + + /* We want the private bus only when running as init */ + if (getpid() != 1) + return 0; + + strcpy(sa.un.sun_path, "/run/systemd/private"); + salen = SOCKADDR_UN_LEN(sa.un); + } else { + size_t left = sizeof(sa.un.sun_path); + char *p = sa.un.sun_path; + const char *e; + + e = secure_getenv("XDG_RUNTIME_DIR"); + if (!e) { + log_error("Failed to determine XDG_RUNTIME_DIR"); + return -EHOSTDOWN; + } + + left = strpcpy(&p, left, e); + left = strpcpy(&p, left, "/systemd/private"); + + salen = sizeof(sa.un) - left; + } + + (void) mkdir_parents_label(sa.un.sun_path, 0755); + (void) unlink(sa.un.sun_path); + + fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) + return log_error_errno(errno, "Failed to allocate private socket: %m"); + + r = bind(fd, &sa.sa, salen); + if (r < 0) + return log_error_errno(errno, "Failed to bind private socket: %m"); + + r = listen(fd, SOMAXCONN); + if (r < 0) + return log_error_errno(errno, "Failed to make private socket listening: %m"); + + r = sd_event_add_io(m->event, &s, fd, EPOLLIN, bus_on_connection, m); + if (r < 0) + return log_error_errno(r, "Failed to allocate event source: %m"); + + (void) sd_event_source_set_description(s, "bus-connection"); + + m->private_listen_fd = fd; + m->private_listen_event_source = s; + fd = -1; + + log_debug("Successfully created private D-Bus server."); + + return 0; +} + +int bus_init(Manager *m, bool try_bus_connect) { + int r; + + if (try_bus_connect) { + r = bus_init_system(m); + if (r < 0) + return r; + + r = bus_init_api(m); + if (r < 0) + return r; + } + + r = bus_init_private(m); + if (r < 0) + return r; + + return 0; +} + +static void destroy_bus(Manager *m, sd_bus **bus) { + Iterator i; + Job *j; + + assert(m); + assert(bus); + + if (!*bus) + return; + + /* Get rid of tracked clients on this bus */ + if (m->subscribed && sd_bus_track_get_bus(m->subscribed) == *bus) + m->subscribed = sd_bus_track_unref(m->subscribed); + + HASHMAP_FOREACH(j, m->jobs, i) + if (j->clients && sd_bus_track_get_bus(j->clients) == *bus) + j->clients = sd_bus_track_unref(j->clients); + + /* Get rid of queued message on this bus */ + if (m->queued_message && sd_bus_message_get_bus(m->queued_message) == *bus) + m->queued_message = sd_bus_message_unref(m->queued_message); + + /* Possibly flush unwritten data, but only if we are + * unprivileged, since we don't want to sync here */ + if (!MANAGER_IS_SYSTEM(m)) + sd_bus_flush(*bus); + + /* And destroy the object */ + sd_bus_close(*bus); + *bus = sd_bus_unref(*bus); +} + +void bus_done(Manager *m) { + sd_bus *b; + + assert(m); + + if (m->api_bus) + destroy_bus(m, &m->api_bus); + if (m->system_bus) + destroy_bus(m, &m->system_bus); + while ((b = set_steal_first(m->private_buses))) + destroy_bus(m, &b); + + m->private_buses = set_free(m->private_buses); + + m->subscribed = sd_bus_track_unref(m->subscribed); + m->deserialized_subscribed = strv_free(m->deserialized_subscribed); + + if (m->private_listen_event_source) + m->private_listen_event_source = sd_event_source_unref(m->private_listen_event_source); + + m->private_listen_fd = safe_close(m->private_listen_fd); + + bus_verify_polkit_async_registry_free(m->polkit_registry); +} + +int bus_fdset_add_all(Manager *m, FDSet *fds) { + Iterator i; + sd_bus *b; + int fd; + + assert(m); + assert(fds); + + /* When we are about to reexecute we add all D-Bus fds to the + * set to pass over to the newly executed systemd. They won't + * be used there however, except thatt they are closed at the + * very end of deserialization, those making it possible for + * clients to synchronously wait for systemd to reexec by + * simply waiting for disconnection */ + + if (m->api_bus) { + fd = sd_bus_get_fd(m->api_bus); + if (fd >= 0) { + fd = fdset_put_dup(fds, fd); + if (fd < 0) + return fd; + } + } + + SET_FOREACH(b, m->private_buses, i) { + fd = sd_bus_get_fd(b); + if (fd >= 0) { + fd = fdset_put_dup(fds, fd); + if (fd < 0) + return fd; + } + } + + /* We don't offer any APIs on the system bus (well, unless it + * is the same as the API bus) hence we don't bother with it + * here */ + + return 0; +} + +int bus_foreach_bus( + Manager *m, + sd_bus_track *subscribed2, + int (*send_message)(sd_bus *bus, void *userdata), + void *userdata) { + + Iterator i; + sd_bus *b; + int r, ret = 0; + + /* Send to all direct buses, unconditionally */ + SET_FOREACH(b, m->private_buses, i) { + r = send_message(b, userdata); + if (r < 0) + ret = r; + } + + /* Send to API bus, but only if somebody is subscribed */ + if (sd_bus_track_count(m->subscribed) > 0 || + sd_bus_track_count(subscribed2) > 0) { + r = send_message(m->api_bus, userdata); + if (r < 0) + ret = r; + } + + return ret; +} + +void bus_track_serialize(sd_bus_track *t, FILE *f, const char *prefix) { + const char *n; + + assert(f); + assert(prefix); + + for (n = sd_bus_track_first(t); n; n = sd_bus_track_next(t)) { + int c, j; + + c = sd_bus_track_count_name(t, n); + + for (j = 0; j < c; j++) { + fputs(prefix, f); + fputc('=', f); + fputs(n, f); + fputc('\n', f); + } + } +} + +int bus_track_coldplug(Manager *m, sd_bus_track **t, bool recursive, char **l) { + char **i; + int r = 0; + + assert(m); + assert(t); + + if (strv_isempty(l)) + return 0; + + if (!m->api_bus) + return 0; + + if (!*t) { + r = sd_bus_track_new(m->api_bus, t, NULL, NULL); + if (r < 0) + return r; + } + + r = sd_bus_track_set_recursive(*t, recursive); + if (r < 0) + return r; + + r = 0; + STRV_FOREACH(i, l) { + int k; + + k = sd_bus_track_add_name(*t, *i); + if (k < 0) + r = k; + } + + return r; +} + +int bus_verify_manage_units_async(Manager *m, sd_bus_message *call, sd_bus_error *error) { + return bus_verify_polkit_async(call, CAP_SYS_ADMIN, "org.freedesktop.systemd1.manage-units", NULL, false, UID_INVALID, &m->polkit_registry, error); +} + +int bus_verify_manage_unit_files_async(Manager *m, sd_bus_message *call, sd_bus_error *error) { + return bus_verify_polkit_async(call, CAP_SYS_ADMIN, "org.freedesktop.systemd1.manage-unit-files", NULL, false, UID_INVALID, &m->polkit_registry, error); +} + +int bus_verify_reload_daemon_async(Manager *m, sd_bus_message *call, sd_bus_error *error) { + return bus_verify_polkit_async(call, CAP_SYS_ADMIN, "org.freedesktop.systemd1.reload-daemon", NULL, false, UID_INVALID, &m->polkit_registry, error); +} + +int bus_verify_set_environment_async(Manager *m, sd_bus_message *call, sd_bus_error *error) { + return bus_verify_polkit_async(call, CAP_SYS_ADMIN, "org.freedesktop.systemd1.set-environment", NULL, false, UID_INVALID, &m->polkit_registry, error); +} diff --git a/src/grp-system/libcore/src/dbus.h b/src/grp-system/libcore/src/dbus.h new file mode 100644 index 0000000000..9f892599cc --- /dev/null +++ b/src/grp-system/libcore/src/dbus.h @@ -0,0 +1,43 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/manager.h" + +int bus_send_queued_message(Manager *m); + +int bus_init(Manager *m, bool try_bus_connect); +void bus_done(Manager *m); + +int bus_fdset_add_all(Manager *m, FDSet *fds); + +void bus_track_serialize(sd_bus_track *t, FILE *f, const char *prefix); +int bus_track_coldplug(Manager *m, sd_bus_track **t, bool recursive, char **l); + +int manager_sync_bus_names(Manager *m, sd_bus *bus); + +int bus_foreach_bus(Manager *m, sd_bus_track *subscribed2, int (*send_message)(sd_bus *bus, void *userdata), void *userdata); + +int bus_verify_manage_units_async(Manager *m, sd_bus_message *call, sd_bus_error *error); +int bus_verify_manage_unit_files_async(Manager *m, sd_bus_message *call, sd_bus_error *error); +int bus_verify_reload_daemon_async(Manager *m, sd_bus_message *call, sd_bus_error *error); +int bus_verify_set_environment_async(Manager *m, sd_bus_message *call, sd_bus_error *error); + +int bus_forward_agent_released(Manager *m, const char *path); diff --git a/src/grp-system/libcore/src/device.c b/src/grp-system/libcore/src/device.c new file mode 100644 index 0000000000..f7865195d7 --- /dev/null +++ b/src/grp-system/libcore/src/device.c @@ -0,0 +1,877 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/epoll.h> + +#include <libudev.h> + +#include "core/device.h" +#include "core/swap.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" +#include "systemd-shared/udev-util.h" + +#include "dbus-device.h" + +static const UnitActiveState state_translation_table[_DEVICE_STATE_MAX] = { + [DEVICE_DEAD] = UNIT_INACTIVE, + [DEVICE_TENTATIVE] = UNIT_ACTIVATING, + [DEVICE_PLUGGED] = UNIT_ACTIVE, +}; + +static int device_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); + +static void device_unset_sysfs(Device *d) { + Hashmap *devices; + Device *first; + + assert(d); + + if (!d->sysfs) + return; + + /* Remove this unit from the chain of devices which share the + * same sysfs path. */ + devices = UNIT(d)->manager->devices_by_sysfs; + first = hashmap_get(devices, d->sysfs); + LIST_REMOVE(same_sysfs, first, d); + + if (first) + hashmap_remove_and_replace(devices, d->sysfs, first->sysfs, first); + else + hashmap_remove(devices, d->sysfs); + + d->sysfs = mfree(d->sysfs); +} + +static int device_set_sysfs(Device *d, const char *sysfs) { + Device *first; + char *copy; + int r; + + assert(d); + + if (streq_ptr(d->sysfs, sysfs)) + return 0; + + r = hashmap_ensure_allocated(&UNIT(d)->manager->devices_by_sysfs, &string_hash_ops); + if (r < 0) + return r; + + copy = strdup(sysfs); + if (!copy) + return -ENOMEM; + + device_unset_sysfs(d); + + first = hashmap_get(UNIT(d)->manager->devices_by_sysfs, sysfs); + LIST_PREPEND(same_sysfs, first, d); + + r = hashmap_replace(UNIT(d)->manager->devices_by_sysfs, copy, first); + if (r < 0) { + LIST_REMOVE(same_sysfs, first, d); + free(copy); + return r; + } + + d->sysfs = copy; + + return 0; +} + +static void device_init(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + assert(UNIT(d)->load_state == UNIT_STUB); + + /* In contrast to all other unit types we timeout jobs waiting + * for devices by default. This is because they otherwise wait + * indefinitely for plugged in devices, something which cannot + * happen for the other units since their operations time out + * anyway. */ + u->job_timeout = u->manager->default_timeout_start_usec; + + u->ignore_on_isolate = true; +} + +static void device_done(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + + device_unset_sysfs(d); +} + +static void device_set_state(Device *d, DeviceState state) { + DeviceState old_state; + assert(d); + + old_state = d->state; + d->state = state; + + if (state != old_state) + log_unit_debug(UNIT(d), "Changed %s -> %s", device_state_to_string(old_state), device_state_to_string(state)); + + unit_notify(UNIT(d), state_translation_table[old_state], state_translation_table[state], true); +} + +static int device_coldplug(Unit *u) { + Device *d = DEVICE(u); + + assert(d); + assert(d->state == DEVICE_DEAD); + + if (d->found & DEVICE_FOUND_UDEV) + /* If udev says the device is around, it's around */ + device_set_state(d, DEVICE_PLUGGED); + else if (d->found != DEVICE_NOT_FOUND && d->deserialized_state != DEVICE_PLUGGED) + /* If a device is found in /proc/self/mountinfo or + * /proc/swaps, and was not yet announced via udev, + * it's "tentatively" around. */ + device_set_state(d, DEVICE_TENTATIVE); + + return 0; +} + +static int device_serialize(Unit *u, FILE *f, FDSet *fds) { + Device *d = DEVICE(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", device_state_to_string(d->state)); + + return 0; +} + +static int device_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Device *d = DEVICE(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + DeviceState state; + + state = device_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + d->deserialized_state = state; + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +static void device_dump(Unit *u, FILE *f, const char *prefix) { + Device *d = DEVICE(u); + + assert(d); + + fprintf(f, + "%sDevice State: %s\n" + "%sSysfs Path: %s\n", + prefix, device_state_to_string(d->state), + prefix, strna(d->sysfs)); +} + +_pure_ static UnitActiveState device_active_state(Unit *u) { + assert(u); + + return state_translation_table[DEVICE(u)->state]; +} + +_pure_ static const char *device_sub_state_to_string(Unit *u) { + assert(u); + + return device_state_to_string(DEVICE(u)->state); +} + +static int device_update_description(Unit *u, struct udev_device *dev, const char *path) { + const char *model; + int r; + + assert(u); + assert(dev); + assert(path); + + model = udev_device_get_property_value(dev, "ID_MODEL_FROM_DATABASE"); + if (!model) + model = udev_device_get_property_value(dev, "ID_MODEL"); + + if (model) { + const char *label; + + /* Try to concatenate the device model string with a label, if there is one */ + label = udev_device_get_property_value(dev, "ID_FS_LABEL"); + if (!label) + label = udev_device_get_property_value(dev, "ID_PART_ENTRY_NAME"); + if (!label) + label = udev_device_get_property_value(dev, "ID_PART_ENTRY_NUMBER"); + + if (label) { + _cleanup_free_ char *j; + + j = strjoin(model, " ", label, NULL); + if (j) + r = unit_set_description(u, j); + else + r = -ENOMEM; + } else + r = unit_set_description(u, model); + } else + r = unit_set_description(u, path); + + if (r < 0) + log_unit_error_errno(u, r, "Failed to set device description: %m"); + + return r; +} + +static int device_add_udev_wants(Unit *u, struct udev_device *dev) { + const char *wants; + const char *word, *state; + size_t l; + int r; + const char *property; + + assert(u); + assert(dev); + + property = MANAGER_IS_USER(u->manager) ? "SYSTEMD_USER_WANTS" : "SYSTEMD_WANTS"; + wants = udev_device_get_property_value(dev, property); + if (!wants) + return 0; + + FOREACH_WORD_QUOTED(word, l, wants, state) { + _cleanup_free_ char *n = NULL; + char e[l+1]; + + memcpy(e, word, l); + e[l] = 0; + + r = unit_name_mangle(e, UNIT_NAME_NOGLOB, &n); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to mangle unit name: %m"); + + r = unit_add_dependency_by_name(u, UNIT_WANTS, n, NULL, true); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to add wants dependency: %m"); + } + if (!isempty(state)) + log_unit_warning(u, "Property %s on %s has trailing garbage, ignoring.", property, strna(udev_device_get_syspath(dev))); + + return 0; +} + +static int device_setup_unit(Manager *m, struct udev_device *dev, const char *path, bool main) { + _cleanup_free_ char *e = NULL; + const char *sysfs = NULL; + Unit *u = NULL; + bool delete; + int r; + + assert(m); + assert(path); + + if (dev) { + sysfs = udev_device_get_syspath(dev); + if (!sysfs) + return 0; + } + + r = unit_name_from_path(path, ".device", &e); + if (r < 0) + return log_error_errno(r, "Failed to generate unit name from device path: %m"); + + u = manager_get_unit(m, e); + + /* The device unit can still be present even if the device was + * unplugged: a mount unit can reference it hence preventing + * the GC to have garbaged it. That's desired since the device + * unit may have a dependency on the mount unit which was + * added during the loading of the later. */ + if (dev && u && DEVICE(u)->state == DEVICE_PLUGGED) { + /* This unit is in plugged state: we're sure it's + * attached to a device. */ + if (!path_equal(DEVICE(u)->sysfs, sysfs)) { + log_unit_debug(u, "Dev %s appeared twice with different sysfs paths %s and %s", + e, DEVICE(u)->sysfs, sysfs); + return -EEXIST; + } + } + + if (!u) { + delete = true; + + r = unit_new_for_name(m, sizeof(Device), e, &u); + if (r < 0) + goto fail; + + unit_add_to_load_queue(u); + } else + delete = false; + + /* If this was created via some dependency and has not + * actually been seen yet ->sysfs will not be + * initialized. Hence initialize it if necessary. */ + if (sysfs) { + r = device_set_sysfs(DEVICE(u), sysfs); + if (r < 0) + goto fail; + + (void) device_update_description(u, dev, path); + + /* The additional systemd udev properties we only interpret + * for the main object */ + if (main) + (void) device_add_udev_wants(u, dev); + } + + + /* Note that this won't dispatch the load queue, the caller + * has to do that if needed and appropriate */ + + unit_add_to_dbus_queue(u); + return 0; + +fail: + log_unit_warning_errno(u, r, "Failed to set up device unit: %m"); + + if (delete && u) + unit_free(u); + + return r; +} + +static int device_process_new(Manager *m, struct udev_device *dev) { + const char *sysfs, *dn, *alias; + struct udev_list_entry *item = NULL, *first = NULL; + int r; + + assert(m); + + sysfs = udev_device_get_syspath(dev); + if (!sysfs) + return 0; + + /* Add the main unit named after the sysfs path */ + r = device_setup_unit(m, dev, sysfs, true); + if (r < 0) + return r; + + /* Add an additional unit for the device node */ + dn = udev_device_get_devnode(dev); + if (dn) + (void) device_setup_unit(m, dev, dn, false); + + /* Add additional units for all symlinks */ + first = udev_device_get_devlinks_list_entry(dev); + udev_list_entry_foreach(item, first) { + const char *p; + struct stat st; + + /* Don't bother with the /dev/block links */ + p = udev_list_entry_get_name(item); + + if (path_startswith(p, "/dev/block/") || + path_startswith(p, "/dev/char/")) + continue; + + /* Verify that the symlink in the FS actually belongs + * to this device. This is useful to deal with + * conflicting devices, e.g. when two disks want the + * same /dev/disk/by-label/xxx link because they have + * the same label. We want to make sure that the same + * device that won the symlink wins in systemd, so we + * check the device node major/minor */ + if (stat(p, &st) >= 0) + if ((!S_ISBLK(st.st_mode) && !S_ISCHR(st.st_mode)) || + st.st_rdev != udev_device_get_devnum(dev)) + continue; + + (void) device_setup_unit(m, dev, p, false); + } + + /* Add additional units for all explicitly configured + * aliases */ + alias = udev_device_get_property_value(dev, "SYSTEMD_ALIAS"); + if (alias) { + const char *word, *state; + size_t l; + + FOREACH_WORD_QUOTED(word, l, alias, state) { + char e[l+1]; + + memcpy(e, word, l); + e[l] = 0; + + if (path_is_absolute(e)) + (void) device_setup_unit(m, dev, e, false); + else + log_warning("SYSTEMD_ALIAS for %s is not an absolute path, ignoring: %s", sysfs, e); + } + if (!isempty(state)) + log_warning("SYSTEMD_ALIAS for %s has trailing garbage, ignoring.", sysfs); + } + + return 0; +} + +static void device_update_found_one(Device *d, bool add, DeviceFound found, bool now) { + DeviceFound n, previous; + + assert(d); + + n = add ? (d->found | found) : (d->found & ~found); + if (n == d->found) + return; + + previous = d->found; + d->found = n; + + if (!now) + return; + + /* Didn't exist before, but does now? if so, generate a new invocation ID for it */ + if (previous == DEVICE_NOT_FOUND && d->found != DEVICE_NOT_FOUND) + (void) unit_acquire_invocation_id(UNIT(d)); + + if (d->found & DEVICE_FOUND_UDEV) + /* When the device is known to udev we consider it + * plugged. */ + device_set_state(d, DEVICE_PLUGGED); + else if (d->found != DEVICE_NOT_FOUND && (previous & DEVICE_FOUND_UDEV) == 0) + /* If the device has not been seen by udev yet, but is + * now referenced by the kernel, then we assume the + * kernel knows it now, and udev might soon too. */ + device_set_state(d, DEVICE_TENTATIVE); + else + /* If nobody sees the device, or if the device was + * previously seen by udev and now is only referenced + * from the kernel, then we consider the device is + * gone, the kernel just hasn't noticed it yet. */ + device_set_state(d, DEVICE_DEAD); +} + +static int device_update_found_by_sysfs(Manager *m, const char *sysfs, bool add, DeviceFound found, bool now) { + Device *d, *l; + + assert(m); + assert(sysfs); + + if (found == DEVICE_NOT_FOUND) + return 0; + + l = hashmap_get(m->devices_by_sysfs, sysfs); + LIST_FOREACH(same_sysfs, d, l) + device_update_found_one(d, add, found, now); + + return 0; +} + +static int device_update_found_by_name(Manager *m, const char *path, bool add, DeviceFound found, bool now) { + _cleanup_free_ char *e = NULL; + Unit *u; + int r; + + assert(m); + assert(path); + + if (found == DEVICE_NOT_FOUND) + return 0; + + r = unit_name_from_path(path, ".device", &e); + if (r < 0) + return log_error_errno(r, "Failed to generate unit name from device path: %m"); + + u = manager_get_unit(m, e); + if (!u) + return 0; + + device_update_found_one(DEVICE(u), add, found, now); + return 0; +} + +static bool device_is_ready(struct udev_device *dev) { + const char *ready; + + assert(dev); + + ready = udev_device_get_property_value(dev, "SYSTEMD_READY"); + if (!ready) + return true; + + return parse_boolean(ready) != 0; +} + +static Unit *device_following(Unit *u) { + Device *d = DEVICE(u); + Device *other, *first = NULL; + + assert(d); + + if (startswith(u->id, "sys-")) + return NULL; + + /* Make everybody follow the unit that's named after the sysfs path */ + for (other = d->same_sysfs_next; other; other = other->same_sysfs_next) + if (startswith(UNIT(other)->id, "sys-")) + return UNIT(other); + + for (other = d->same_sysfs_prev; other; other = other->same_sysfs_prev) { + if (startswith(UNIT(other)->id, "sys-")) + return UNIT(other); + + first = other; + } + + return UNIT(first); +} + +static int device_following_set(Unit *u, Set **_set) { + Device *d = DEVICE(u), *other; + Set *set; + int r; + + assert(d); + assert(_set); + + if (LIST_JUST_US(same_sysfs, d)) { + *_set = NULL; + return 0; + } + + set = set_new(NULL); + if (!set) + return -ENOMEM; + + LIST_FOREACH_AFTER(same_sysfs, other, d) { + r = set_put(set, other); + if (r < 0) + goto fail; + } + + LIST_FOREACH_BEFORE(same_sysfs, other, d) { + r = set_put(set, other); + if (r < 0) + goto fail; + } + + *_set = set; + return 1; + +fail: + set_free(set); + return r; +} + +static void device_shutdown(Manager *m) { + assert(m); + + m->udev_event_source = sd_event_source_unref(m->udev_event_source); + + if (m->udev_monitor) { + udev_monitor_unref(m->udev_monitor); + m->udev_monitor = NULL; + } + + m->devices_by_sysfs = hashmap_free(m->devices_by_sysfs); +} + +static void device_enumerate(Manager *m) { + _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; + struct udev_list_entry *item = NULL, *first = NULL; + int r; + + assert(m); + + if (!m->udev_monitor) { + m->udev_monitor = udev_monitor_new_from_netlink(m->udev, "udev"); + if (!m->udev_monitor) { + log_oom(); + goto fail; + } + + /* This will fail if we are unprivileged, but that + * should not matter much, as user instances won't run + * during boot. */ + (void) udev_monitor_set_receive_buffer_size(m->udev_monitor, 128*1024*1024); + + r = udev_monitor_filter_add_match_tag(m->udev_monitor, "systemd"); + if (r < 0) { + log_error_errno(r, "Failed to add udev tag match: %m"); + goto fail; + } + + r = udev_monitor_enable_receiving(m->udev_monitor); + if (r < 0) { + log_error_errno(r, "Failed to enable udev event reception: %m"); + goto fail; + } + + r = sd_event_add_io(m->event, &m->udev_event_source, udev_monitor_get_fd(m->udev_monitor), EPOLLIN, device_dispatch_io, m); + if (r < 0) { + log_error_errno(r, "Failed to watch udev file descriptor: %m"); + goto fail; + } + + (void) sd_event_source_set_description(m->udev_event_source, "device"); + } + + e = udev_enumerate_new(m->udev); + if (!e) { + log_oom(); + goto fail; + } + + r = udev_enumerate_add_match_tag(e, "systemd"); + if (r < 0) { + log_error_errno(r, "Failed to create udev tag enumeration: %m"); + goto fail; + } + + r = udev_enumerate_add_match_is_initialized(e); + if (r < 0) { + log_error_errno(r, "Failed to install initialization match into enumeration: %m"); + goto fail; + } + + r = udev_enumerate_scan_devices(e); + if (r < 0) { + log_error_errno(r, "Failed to enumerate devices: %m"); + goto fail; + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) { + _cleanup_udev_device_unref_ struct udev_device *dev = NULL; + const char *sysfs; + + sysfs = udev_list_entry_get_name(item); + + dev = udev_device_new_from_syspath(m->udev, sysfs); + if (!dev) { + log_oom(); + continue; + } + + if (!device_is_ready(dev)) + continue; + + (void) device_process_new(m, dev); + + device_update_found_by_sysfs(m, sysfs, true, DEVICE_FOUND_UDEV, false); + } + + return; + +fail: + device_shutdown(m); +} + +static int device_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + _cleanup_udev_device_unref_ struct udev_device *dev = NULL; + Manager *m = userdata; + const char *action, *sysfs; + int r; + + assert(m); + + if (revents != EPOLLIN) { + static RATELIMIT_DEFINE(limit, 10*USEC_PER_SEC, 5); + + if (!ratelimit_test(&limit)) + log_error_errno(errno, "Failed to get udev event: %m"); + if (!(revents & EPOLLIN)) + return 0; + } + + /* + * libudev might filter-out devices which pass the bloom + * filter, so getting NULL here is not necessarily an error. + */ + dev = udev_monitor_receive_device(m->udev_monitor); + if (!dev) + return 0; + + sysfs = udev_device_get_syspath(dev); + if (!sysfs) { + log_error("Failed to get udev sys path."); + return 0; + } + + action = udev_device_get_action(dev); + if (!action) { + log_error("Failed to get udev action string."); + return 0; + } + + if (streq(action, "remove")) { + r = swap_process_device_remove(m, dev); + if (r < 0) + log_error_errno(r, "Failed to process swap device remove event: %m"); + + /* If we get notified that a device was removed by + * udev, then it's completely gone, hence unset all + * found bits */ + device_update_found_by_sysfs(m, sysfs, false, DEVICE_FOUND_UDEV|DEVICE_FOUND_MOUNT|DEVICE_FOUND_SWAP, true); + + } else if (device_is_ready(dev)) { + + (void) device_process_new(m, dev); + + r = swap_process_device_new(m, dev); + if (r < 0) + log_error_errno(r, "Failed to process swap device new event: %m"); + + manager_dispatch_load_queue(m); + + /* The device is found now, set the udev found bit */ + device_update_found_by_sysfs(m, sysfs, true, DEVICE_FOUND_UDEV, true); + + } else { + /* The device is nominally around, but not ready for + * us. Hence unset the udev bit, but leave the rest + * around. */ + + device_update_found_by_sysfs(m, sysfs, false, DEVICE_FOUND_UDEV, true); + } + + return 0; +} + +static bool device_supported(void) { + static int read_only = -1; + + /* If /sys is read-only we don't support device units, and any + * attempts to start one should fail immediately. */ + + if (read_only < 0) + read_only = path_is_read_only_fs("/sys"); + + return read_only <= 0; +} + +int device_found_node(Manager *m, const char *node, bool add, DeviceFound found, bool now) { + _cleanup_udev_device_unref_ struct udev_device *dev = NULL; + struct stat st; + + assert(m); + assert(node); + + if (!device_supported()) + return 0; + + /* This is called whenever we find a device referenced in + * /proc/swaps or /proc/self/mounts. Such a device might be + * mounted/enabled at a time where udev has not finished + * probing it yet, and we thus haven't learned about it + * yet. In this case we will set the device unit to + * "tentative" state. */ + + if (add) { + if (!path_startswith(node, "/dev")) + return 0; + + /* We make an extra check here, if the device node + * actually exists. If it's missing, then this is an + * indication that device was unplugged but is still + * referenced in /proc/swaps or + * /proc/self/mountinfo. Note that this check doesn't + * really cover all cases where a device might be gone + * away, since drives that can have a medium inserted + * will still have a device node even when the medium + * is not there... */ + + if (stat(node, &st) >= 0) { + if (!S_ISBLK(st.st_mode) && !S_ISCHR(st.st_mode)) + return 0; + + dev = udev_device_new_from_devnum(m->udev, S_ISBLK(st.st_mode) ? 'b' : 'c', st.st_rdev); + if (!dev && errno != ENOENT) + return log_error_errno(errno, "Failed to get udev device from devnum %u:%u: %m", major(st.st_rdev), minor(st.st_rdev)); + + } else if (errno != ENOENT) + return log_error_errno(errno, "Failed to stat device node file %s: %m", node); + + /* If the device is known in the kernel and newly + * appeared, then we'll create a device unit for it, + * under the name referenced in /proc/swaps or + * /proc/self/mountinfo. */ + + (void) device_setup_unit(m, dev, node, false); + } + + /* Update the device unit's state, should it exist */ + return device_update_found_by_name(m, node, add, found, now); +} + +const UnitVTable device_vtable = { + .object_size = sizeof(Device), + .sections = + "Unit\0" + "Device\0" + "Install\0", + + .init = device_init, + .done = device_done, + .load = unit_load_fragment_and_dropin_optional, + + .coldplug = device_coldplug, + + .serialize = device_serialize, + .deserialize_item = device_deserialize_item, + + .dump = device_dump, + + .active_state = device_active_state, + .sub_state_to_string = device_sub_state_to_string, + + .bus_vtable = bus_device_vtable, + + .following = device_following, + .following_set = device_following_set, + + .enumerate = device_enumerate, + .shutdown = device_shutdown, + .supported = device_supported, + + .status_message_formats = { + .starting_stopping = { + [0] = "Expecting device %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Found device %s.", + [JOB_TIMEOUT] = "Timed out waiting for device %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/dynamic-user.c b/src/grp-system/libcore/src/dynamic-user.c new file mode 100644 index 0000000000..9a1ea09e03 --- /dev/null +++ b/src/grp-system/libcore/src/dynamic-user.c @@ -0,0 +1,794 @@ +/*** + This file is part of systemd. + + Copyright 2016 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <grp.h> +#include <pwd.h> +#include <sys/file.h> + +#include "core/dynamic-user.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/random-util.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/user-util.h" + +/* Takes a value generated randomly or by hashing and turns it into a UID in the right range */ +#define UID_CLAMP_INTO_RANGE(rnd) (((uid_t) (rnd) % (DYNAMIC_UID_MAX - DYNAMIC_UID_MIN + 1)) + DYNAMIC_UID_MIN) + +static DynamicUser* dynamic_user_free(DynamicUser *d) { + if (!d) + return NULL; + + if (d->manager) + (void) hashmap_remove(d->manager->dynamic_users, d->name); + + safe_close_pair(d->storage_socket); + return mfree(d); +} + +static int dynamic_user_add(Manager *m, const char *name, int storage_socket[2], DynamicUser **ret) { + DynamicUser *d = NULL; + int r; + + assert(m); + assert(name); + assert(storage_socket); + + r = hashmap_ensure_allocated(&m->dynamic_users, &string_hash_ops); + if (r < 0) + return r; + + d = malloc0(offsetof(DynamicUser, name) + strlen(name) + 1); + if (!d) + return -ENOMEM; + + strcpy(d->name, name); + + d->storage_socket[0] = storage_socket[0]; + d->storage_socket[1] = storage_socket[1]; + + r = hashmap_put(m->dynamic_users, d->name, d); + if (r < 0) { + free(d); + return r; + } + + d->manager = m; + + if (ret) + *ret = d; + + return 0; +} + +int dynamic_user_acquire(Manager *m, const char *name, DynamicUser** ret) { + _cleanup_close_pair_ int storage_socket[2] = { -1, -1 }; + DynamicUser *d; + int r; + + assert(m); + assert(name); + + /* Return the DynamicUser structure for a specific user name. Note that this won't actually allocate a UID for + * it, but just prepare the data structure for it. The UID is allocated only on demand, when it's really + * needed, and in the child process we fork off, since allocation involves NSS checks which are not OK to do + * from PID 1. To allow the children and PID 1 share information about allocated UIDs we use an anonymous + * AF_UNIX/SOCK_DGRAM socket (called the "storage socket") that contains at most one datagram with the + * allocated UID number, plus an fd referencing the lock file for the UID + * (i.e. /run/systemd/dynamic-uid/$UID). Why involve the socket pair? So that PID 1 and all its children can + * share the same storage for the UID and lock fd, simply by inheriting the storage socket fds. The socket pair + * may exist in three different states: + * + * a) no datagram stored. This is the initial state. In this case the dynamic user was never realized. + * + * b) a datagram containing a UID stored, but no lock fd attached to it. In this case there was already a + * statically assigned UID by the same name, which we are reusing. + * + * c) a datagram containing a UID stored, and a lock fd is attached to it. In this case we allocated a dynamic + * UID and locked it in the file system, using the lock fd. + * + * As PID 1 and various children might access the socket pair simultaneously, and pop the datagram or push it + * back in any time, we also maintain a lock on the socket pair. Note one peculiarity regarding locking here: + * the UID lock on disk is protected via a BSD file lock (i.e. an fd-bound lock), so that the lock is kept in + * place as long as there's a reference to the fd open. The lock on the storage socket pair however is a POSIX + * file lock (i.e. a process-bound lock), as all users share the same fd of this (after all it is anonymous, + * nobody else could get any access to it except via our own fd) and we want to synchronize access between all + * processes that have access to it. */ + + d = hashmap_get(m->dynamic_users, name); + if (d) { + /* We already have a structure for the dynamic user, let's increase the ref count and reuse it */ + d->n_ref++; + *ret = d; + return 0; + } + + if (!valid_user_group_name_or_id(name)) + return -EINVAL; + + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, storage_socket) < 0) + return -errno; + + r = dynamic_user_add(m, name, storage_socket, &d); + if (r < 0) + return r; + + storage_socket[0] = storage_socket[1] = -1; + + if (ret) { + d->n_ref++; + *ret = d; + } + + return 1; +} + +static int make_uid_symlinks(uid_t uid, const char *name, bool b) { + + char path1[strlen("/run/systemd/dynamic-uid/direct:") + DECIMAL_STR_MAX(uid_t) + 1]; + const char *path2; + int r = 0, k; + + /* Add direct additional symlinks for direct lookups of dynamic UIDs and their names by userspace code. The + * only reason we have this is because dbus-daemon cannot use D-Bus for resolving users and groups (since it + * would be its own client then). We hence keep these world-readable symlinks in place, so that the + * unprivileged dbus user can read the mappings when it needs them via these symlinks instead of having to go + * via the bus. Ideally, we'd use the lock files we keep for this anyway, but we can't since we use BSD locks + * on them and as those may be taken by any user with read access we can't make them world-readable. */ + + xsprintf(path1, "/run/systemd/dynamic-uid/direct:" UID_FMT, uid); + if (unlink(path1) < 0 && errno != ENOENT) + r = -errno; + + if (b && symlink(name, path1) < 0) { + k = log_warning_errno(errno, "Failed to symlink \"%s\": %m", path1); + if (r == 0) + r = k; + } + + path2 = strjoina("/run/systemd/dynamic-uid/direct:", name); + if (unlink(path2) < 0 && errno != ENOENT) { + k = -errno; + if (r == 0) + r = k; + } + + if (b && symlink(path1 + strlen("/run/systemd/dynamic-uid/direct:"), path2) < 0) { + k = log_warning_errno(errno, "Failed to symlink \"%s\": %m", path2); + if (r == 0) + r = k; + } + + return r; +} + +static int pick_uid(const char *name, uid_t *ret_uid) { + + static const uint8_t hash_key[] = { + 0x37, 0x53, 0x7e, 0x31, 0xcf, 0xce, 0x48, 0xf5, + 0x8a, 0xbb, 0x39, 0x57, 0x8d, 0xd9, 0xec, 0x59 + }; + + unsigned n_tries = 100; + uid_t candidate; + int r; + + /* A static user by this name does not exist yet. Let's find a free ID then, and use that. We start with a UID + * generated as hash from the user name. */ + candidate = UID_CLAMP_INTO_RANGE(siphash24(name, strlen(name), hash_key)); + + (void) mkdir("/run/systemd/dynamic-uid", 0755); + + for (;;) { + char lock_path[strlen("/run/systemd/dynamic-uid/") + DECIMAL_STR_MAX(uid_t) + 1]; + _cleanup_close_ int lock_fd = -1; + ssize_t l; + + if (--n_tries <= 0) /* Give up retrying eventually */ + return -EBUSY; + + if (!uid_is_dynamic(candidate)) + goto next; + + xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, candidate); + + for (;;) { + struct stat st; + + lock_fd = open(lock_path, O_CREAT|O_RDWR|O_NOFOLLOW|O_CLOEXEC|O_NOCTTY, 0600); + if (lock_fd < 0) + return -errno; + + r = flock(lock_fd, LOCK_EX|LOCK_NB); /* Try to get a BSD file lock on the UID lock file */ + if (r < 0) { + if (errno == EBUSY || errno == EAGAIN) + goto next; /* already in use */ + + return -errno; + } + + if (fstat(lock_fd, &st) < 0) + return -errno; + if (st.st_nlink > 0) + break; + + /* Oh, bummer, we got the lock, but the file was unlinked between the time we opened it and + * got the lock. Close it, and try again. */ + lock_fd = safe_close(lock_fd); + } + + /* Some superficial check whether this UID/GID might already be taken by some static user */ + if (getpwuid(candidate) || getgrgid((gid_t) candidate)) { + (void) unlink(lock_path); + goto next; + } + + /* Let's store the user name in the lock file, so that we can use it for looking up the username for a UID */ + l = pwritev(lock_fd, + (struct iovec[2]) { + { .iov_base = (char*) name, .iov_len = strlen(name) }, + { .iov_base = (char[1]) { '\n' }, .iov_len = 1 } + }, 2, 0); + if (l < 0) { + (void) unlink(lock_path); + return -errno; + } + + (void) ftruncate(lock_fd, l); + (void) make_uid_symlinks(candidate, name, true); /* also add direct lookup symlinks */ + + *ret_uid = candidate; + r = lock_fd; + lock_fd = -1; + + return r; + + next: + /* Pick another random UID, and see if that works for us. */ + random_bytes(&candidate, sizeof(candidate)); + candidate = UID_CLAMP_INTO_RANGE(candidate); + } +} + +static int dynamic_user_pop(DynamicUser *d, uid_t *ret_uid, int *ret_lock_fd) { + uid_t uid = UID_INVALID; + struct iovec iov = { + .iov_base = &uid, + .iov_len = sizeof(uid), + }; + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(int))]; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + .msg_iov = &iov, + .msg_iovlen = 1, + }; + struct cmsghdr *cmsg; + + ssize_t k; + int lock_fd = -1; + + assert(d); + assert(ret_uid); + assert(ret_lock_fd); + + /* Read the UID and lock fd that is stored in the storage AF_UNIX socket. This should be called with the lock + * on the socket taken. */ + + k = recvmsg(d->storage_socket[0], &mh, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC); + if (k < 0) + return -errno; + + cmsg = cmsg_find(&mh, SOL_SOCKET, SCM_RIGHTS, CMSG_LEN(sizeof(int))); + if (cmsg) + lock_fd = *(int*) CMSG_DATA(cmsg); + else + cmsg_close_all(&mh); /* just in case... */ + + *ret_uid = uid; + *ret_lock_fd = lock_fd; + + return 0; +} + +static int dynamic_user_push(DynamicUser *d, uid_t uid, int lock_fd) { + struct iovec iov = { + .iov_base = &uid, + .iov_len = sizeof(uid), + }; + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(int))]; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + .msg_iov = &iov, + .msg_iovlen = 1, + }; + ssize_t k; + + assert(d); + + /* Store the UID and lock_fd in the storage socket. This should be called with the socket pair lock taken. */ + + if (lock_fd >= 0) { + struct cmsghdr *cmsg; + + cmsg = CMSG_FIRSTHDR(&mh); + cmsg->cmsg_level = SOL_SOCKET; + cmsg->cmsg_type = SCM_RIGHTS; + cmsg->cmsg_len = CMSG_LEN(sizeof(int)); + memcpy(CMSG_DATA(cmsg), &lock_fd, sizeof(int)); + + mh.msg_controllen = CMSG_SPACE(sizeof(int)); + } else { + mh.msg_control = NULL; + mh.msg_controllen = 0; + } + + k = sendmsg(d->storage_socket[1], &mh, MSG_DONTWAIT|MSG_NOSIGNAL); + if (k < 0) + return -errno; + + return 0; +} + +static void unlink_uid_lock(int lock_fd, uid_t uid, const char *name) { + char lock_path[strlen("/run/systemd/dynamic-uid/") + DECIMAL_STR_MAX(uid_t) + 1]; + + if (lock_fd < 0) + return; + + xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, uid); + (void) unlink(lock_path); + + (void) make_uid_symlinks(uid, name, false); /* remove direct lookup symlinks */ +} + +int dynamic_user_realize(DynamicUser *d, uid_t *ret) { + + _cleanup_close_ int etc_passwd_lock_fd = -1, uid_lock_fd = -1; + uid_t uid = UID_INVALID; + int r; + + assert(d); + + /* Acquire a UID for the user name. This will allocate a UID for the user name if the user doesn't exist + * yet. If it already exists its existing UID/GID will be reused. */ + + if (lockf(d->storage_socket[0], F_LOCK, 0) < 0) + return -errno; + + r = dynamic_user_pop(d, &uid, &uid_lock_fd); + if (r < 0) { + int new_uid_lock_fd; + uid_t new_uid; + + if (r != -EAGAIN) + goto finish; + + /* OK, nothing stored yet, let's try to find something useful. While we are working on this release the + * lock however, so that nobody else blocks on our NSS lookups. */ + (void) lockf(d->storage_socket[0], F_ULOCK, 0); + + /* Let's see if a proper, static user or group by this name exists. Try to take the lock on + * /etc/passwd, if that fails with EROFS then /etc is read-only. In that case it's fine if we don't + * take the lock, given that users can't be added there anyway in this case. */ + etc_passwd_lock_fd = take_etc_passwd_lock(NULL); + if (etc_passwd_lock_fd < 0 && etc_passwd_lock_fd != -EROFS) + return etc_passwd_lock_fd; + + /* First, let's parse this as numeric UID */ + r = parse_uid(d->name, &uid); + if (r < 0) { + struct passwd *p; + struct group *g; + + /* OK, this is not a numeric UID. Let's see if there's a user by this name */ + p = getpwnam(d->name); + if (p) + uid = p->pw_uid; + + /* Let's see if there's a group by this name */ + g = getgrnam(d->name); + if (g) { + /* If the UID/GID of the user/group of the same don't match, refuse operation */ + if (uid != UID_INVALID && uid != (uid_t) g->gr_gid) + return -EILSEQ; + + uid = (uid_t) g->gr_gid; + } + } + + if (uid == UID_INVALID) { + /* No static UID assigned yet, excellent. Let's pick a new dynamic one, and lock it. */ + + uid_lock_fd = pick_uid(d->name, &uid); + if (uid_lock_fd < 0) + return uid_lock_fd; + } + + /* So, we found a working UID/lock combination. Let's see if we actually still need it. */ + if (lockf(d->storage_socket[0], F_LOCK, 0) < 0) { + unlink_uid_lock(uid_lock_fd, uid, d->name); + return -errno; + } + + r = dynamic_user_pop(d, &new_uid, &new_uid_lock_fd); + if (r < 0) { + if (r != -EAGAIN) { + /* OK, something bad happened, let's get rid of the bits we acquired. */ + unlink_uid_lock(uid_lock_fd, uid, d->name); + goto finish; + } + + /* Great! Nothing is stored here, still. Store our newly acquired data. */ + } else { + /* Hmm, so as it appears there's now something stored in the storage socket. Throw away what we + * acquired, and use what's stored now. */ + + unlink_uid_lock(uid_lock_fd, uid, d->name); + safe_close(uid_lock_fd); + + uid = new_uid; + uid_lock_fd = new_uid_lock_fd; + } + } + + /* If the UID/GID was already allocated dynamically, push the data we popped out back in. If it was already + * allocated statically, push the UID back too, but do not push the lock fd in. If we allocated the UID + * dynamically right here, push that in along with the lock fd for it. */ + r = dynamic_user_push(d, uid, uid_lock_fd); + if (r < 0) + goto finish; + + *ret = uid; + r = 0; + +finish: + (void) lockf(d->storage_socket[0], F_ULOCK, 0); + return r; +} + +int dynamic_user_current(DynamicUser *d, uid_t *ret) { + _cleanup_close_ int lock_fd = -1; + uid_t uid; + int r; + + assert(d); + assert(ret); + + /* Get the currently assigned UID for the user, if there's any. This simply pops the data from the storage socket, and pushes it back in right-away. */ + + if (lockf(d->storage_socket[0], F_LOCK, 0) < 0) + return -errno; + + r = dynamic_user_pop(d, &uid, &lock_fd); + if (r < 0) + goto finish; + + r = dynamic_user_push(d, uid, lock_fd); + if (r < 0) + goto finish; + + *ret = uid; + r = 0; + +finish: + (void) lockf(d->storage_socket[0], F_ULOCK, 0); + return r; +} + +DynamicUser* dynamic_user_ref(DynamicUser *d) { + if (!d) + return NULL; + + assert(d->n_ref > 0); + d->n_ref++; + + return d; +} + +DynamicUser* dynamic_user_unref(DynamicUser *d) { + if (!d) + return NULL; + + /* Note that this doesn't actually release any resources itself. If a dynamic user should be fully destroyed + * and its UID released, use dynamic_user_destroy() instead. NB: the dynamic user table may contain entries + * with no references, which is commonly the case right before a daemon reload. */ + + assert(d->n_ref > 0); + d->n_ref--; + + return NULL; +} + +static int dynamic_user_close(DynamicUser *d) { + _cleanup_close_ int lock_fd = -1; + uid_t uid; + int r; + + /* Release the user ID, by releasing the lock on it, and emptying the storage socket. After this the user is + * unrealized again, much like it was after it the DynamicUser object was first allocated. */ + + if (lockf(d->storage_socket[0], F_LOCK, 0) < 0) + return -errno; + + r = dynamic_user_pop(d, &uid, &lock_fd); + if (r == -EAGAIN) { + /* User wasn't realized yet, nothing to do. */ + r = 0; + goto finish; + } + if (r < 0) + goto finish; + + /* This dynamic user was realized and dynamically allocated. In this case, let's remove the lock file. */ + unlink_uid_lock(lock_fd, uid, d->name); + r = 1; + +finish: + (void) lockf(d->storage_socket[0], F_ULOCK, 0); + return r; +} + +DynamicUser* dynamic_user_destroy(DynamicUser *d) { + if (!d) + return NULL; + + /* Drop a reference to a DynamicUser object, and destroy the user completely if this was the last + * reference. This is called whenever a service is shut down and wants its dynamic UID gone. Note that + * dynamic_user_unref() is what is called whenever a service is simply freed, for example during a reload + * cycle, where the dynamic users should not be destroyed, but our datastructures should. */ + + dynamic_user_unref(d); + + if (d->n_ref > 0) + return NULL; + + (void) dynamic_user_close(d); + return dynamic_user_free(d); +} + +int dynamic_user_serialize(Manager *m, FILE *f, FDSet *fds) { + DynamicUser *d; + Iterator i; + + assert(m); + assert(f); + assert(fds); + + /* Dump the dynamic user database into the manager serialization, to deal with daemon reloads. */ + + HASHMAP_FOREACH(d, m->dynamic_users, i) { + int copy0, copy1; + + copy0 = fdset_put_dup(fds, d->storage_socket[0]); + if (copy0 < 0) + return copy0; + + copy1 = fdset_put_dup(fds, d->storage_socket[1]); + if (copy1 < 0) + return copy1; + + fprintf(f, "dynamic-user=%s %i %i\n", d->name, copy0, copy1); + } + + return 0; +} + +void dynamic_user_deserialize_one(Manager *m, const char *value, FDSet *fds) { + _cleanup_free_ char *name = NULL, *s0 = NULL, *s1 = NULL; + int r, fd0, fd1; + + assert(m); + assert(value); + assert(fds); + + /* Parse the serialization again, after a daemon reload */ + + r = extract_many_words(&value, NULL, 0, &name, &s0, &s1, NULL); + if (r != 3 || !isempty(value)) { + log_debug("Unable to parse dynamic user line."); + return; + } + + if (safe_atoi(s0, &fd0) < 0 || !fdset_contains(fds, fd0)) { + log_debug("Unable to process dynamic user fd specification."); + return; + } + + if (safe_atoi(s1, &fd1) < 0 || !fdset_contains(fds, fd1)) { + log_debug("Unable to process dynamic user fd specification."); + return; + } + + r = dynamic_user_add(m, name, (int[]) { fd0, fd1 }, NULL); + if (r < 0) { + log_debug_errno(r, "Failed to add dynamic user: %m"); + return; + } + + (void) fdset_remove(fds, fd0); + (void) fdset_remove(fds, fd1); +} + +void dynamic_user_vacuum(Manager *m, bool close_user) { + DynamicUser *d; + Iterator i; + + assert(m); + + /* Empty the dynamic user database, optionally cleaning up orphaned dynamic users, i.e. destroy and free users + * to which no reference exist. This is called after a daemon reload finished, in order to destroy users which + * might not be referenced anymore. */ + + HASHMAP_FOREACH(d, m->dynamic_users, i) { + if (d->n_ref > 0) + continue; + + if (close_user) { + log_debug("Removing orphaned dynamic user %s", d->name); + (void) dynamic_user_close(d); + } + + dynamic_user_free(d); + } +} + +int dynamic_user_lookup_uid(Manager *m, uid_t uid, char **ret) { + char lock_path[strlen("/run/systemd/dynamic-uid/") + DECIMAL_STR_MAX(uid_t) + 1]; + _cleanup_free_ char *user = NULL; + uid_t check_uid; + int r; + + assert(m); + assert(ret); + + /* A friendly way to translate a dynamic user's UID into a name. */ + if (!uid_is_dynamic(uid)) + return -ESRCH; + + xsprintf(lock_path, "/run/systemd/dynamic-uid/" UID_FMT, uid); + r = read_one_line_file(lock_path, &user); + if (r == -ENOENT) + return -ESRCH; + if (r < 0) + return r; + + /* The lock file might be stale, hence let's verify the data before we return it */ + r = dynamic_user_lookup_name(m, user, &check_uid); + if (r < 0) + return r; + if (check_uid != uid) /* lock file doesn't match our own idea */ + return -ESRCH; + + *ret = user; + user = NULL; + + return 0; +} + +int dynamic_user_lookup_name(Manager *m, const char *name, uid_t *ret) { + DynamicUser *d; + int r; + + assert(m); + assert(name); + assert(ret); + + /* A friendly call for translating a dynamic user's name into its UID */ + + d = hashmap_get(m->dynamic_users, name); + if (!d) + return -ESRCH; + + r = dynamic_user_current(d, ret); + if (r == -EAGAIN) /* not realized yet? */ + return -ESRCH; + + return r; +} + +int dynamic_creds_acquire(DynamicCreds *creds, Manager *m, const char *user, const char *group) { + bool acquired = false; + int r; + + assert(creds); + assert(m); + + /* A DynamicUser object encapsulates an allocation of both a UID and a GID for a specific name. However, some + * services use different user and groups. For cases like that there's DynamicCreds containing a pair of user + * and group. This call allocates a pair. */ + + if (!creds->user && user) { + r = dynamic_user_acquire(m, user, &creds->user); + if (r < 0) + return r; + + acquired = true; + } + + if (!creds->group) { + + if (creds->user && (!group || streq_ptr(user, group))) + creds->group = dynamic_user_ref(creds->user); + else { + r = dynamic_user_acquire(m, group, &creds->group); + if (r < 0) { + if (acquired) + creds->user = dynamic_user_unref(creds->user); + return r; + } + } + } + + return 0; +} + +int dynamic_creds_realize(DynamicCreds *creds, uid_t *uid, gid_t *gid) { + uid_t u = UID_INVALID; + gid_t g = GID_INVALID; + int r; + + assert(creds); + assert(uid); + assert(gid); + + /* Realize both the referenced user and group */ + + if (creds->user) { + r = dynamic_user_realize(creds->user, &u); + if (r < 0) + return r; + } + + if (creds->group && creds->group != creds->user) { + r = dynamic_user_realize(creds->group, &g); + if (r < 0) + return r; + } else + g = u; + + *uid = u; + *gid = g; + + return 0; +} + +void dynamic_creds_unref(DynamicCreds *creds) { + assert(creds); + + creds->user = dynamic_user_unref(creds->user); + creds->group = dynamic_user_unref(creds->group); +} + +void dynamic_creds_destroy(DynamicCreds *creds) { + assert(creds); + + creds->user = dynamic_user_destroy(creds->user); + creds->group = dynamic_user_destroy(creds->group); +} diff --git a/src/grp-system/libcore/src/emergency-action.c b/src/grp-system/libcore/src/emergency-action.c new file mode 100644 index 0000000000..2c85702970 --- /dev/null +++ b/src/grp-system/libcore/src/emergency-action.c @@ -0,0 +1,129 @@ +/*** + This file is part of systemd. + + Copyright 2014 Lennart Poettering + Copyright 2012 Michael Olbrich + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/reboot.h> + +#include <linux/reboot.h> + +#include "core/emergency-action.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/terminal-util.h" + +static void log_and_status(Manager *m, const char *message, const char *reason) { + log_warning("%s: %s", message, reason); + manager_status_printf(m, STATUS_TYPE_EMERGENCY, + ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, + "%s: %s", message, reason); +} + +int emergency_action( + Manager *m, + EmergencyAction action, + const char *reboot_arg, + const char *reason) { + + assert(m); + assert(action >= 0); + assert(action < _EMERGENCY_ACTION_MAX); + + if (action == EMERGENCY_ACTION_NONE) + return -ECANCELED; + + if (!MANAGER_IS_SYSTEM(m)) { + /* Downgrade all options to simply exiting if we run + * in user mode */ + + log_warning("Exiting: %s", reason); + m->exit_code = MANAGER_EXIT; + return -ECANCELED; + } + + switch (action) { + + case EMERGENCY_ACTION_REBOOT: + log_and_status(m, "Rebooting", reason); + + (void) update_reboot_parameter_and_warn(reboot_arg); + (void) manager_add_job_by_name_and_warn(m, JOB_START, SPECIAL_REBOOT_TARGET, JOB_REPLACE_IRREVERSIBLY, NULL); + + break; + + case EMERGENCY_ACTION_REBOOT_FORCE: + log_and_status(m, "Forcibly rebooting", reason); + + (void) update_reboot_parameter_and_warn(reboot_arg); + m->exit_code = MANAGER_REBOOT; + + break; + + case EMERGENCY_ACTION_REBOOT_IMMEDIATE: + log_and_status(m, "Rebooting immediately", reason); + + sync(); + + if (!isempty(reboot_arg)) { + log_info("Rebooting with argument '%s'.", reboot_arg); + syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_RESTART2, reboot_arg); + log_warning_errno(errno, "Failed to reboot with parameter, retrying without: %m"); + } + + log_info("Rebooting."); + reboot(RB_AUTOBOOT); + break; + + case EMERGENCY_ACTION_POWEROFF: + log_and_status(m, "Powering off", reason); + (void) manager_add_job_by_name_and_warn(m, JOB_START, SPECIAL_POWEROFF_TARGET, JOB_REPLACE_IRREVERSIBLY, NULL); + break; + + case EMERGENCY_ACTION_POWEROFF_FORCE: + log_and_status(m, "Forcibly powering off", reason); + m->exit_code = MANAGER_POWEROFF; + break; + + case EMERGENCY_ACTION_POWEROFF_IMMEDIATE: + log_and_status(m, "Powering off immediately", reason); + + sync(); + + log_info("Powering off."); + reboot(RB_POWER_OFF); + break; + + default: + assert_not_reached("Unknown emergency action"); + } + + return -ECANCELED; +} + +static const char* const emergency_action_table[_EMERGENCY_ACTION_MAX] = { + [EMERGENCY_ACTION_NONE] = "none", + [EMERGENCY_ACTION_REBOOT] = "reboot", + [EMERGENCY_ACTION_REBOOT_FORCE] = "reboot-force", + [EMERGENCY_ACTION_REBOOT_IMMEDIATE] = "reboot-immediate", + [EMERGENCY_ACTION_POWEROFF] = "poweroff", + [EMERGENCY_ACTION_POWEROFF_FORCE] = "poweroff-force", + [EMERGENCY_ACTION_POWEROFF_IMMEDIATE] = "poweroff-immediate" +}; +DEFINE_STRING_TABLE_LOOKUP(emergency_action, EmergencyAction); diff --git a/src/grp-system/libcore/src/execute.c b/src/grp-system/libcore/src/execute.c new file mode 100644 index 0000000000..cbe772bf9f --- /dev/null +++ b/src/grp-system/libcore/src/execute.c @@ -0,0 +1,4000 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> +#include <glob.h> +#include <grp.h> +#include <poll.h> +#include <signal.h> +#include <string.h> +#include <sys/capability.h> +#include <sys/eventfd.h> +#include <sys/mman.h> +#include <sys/personality.h> +#include <sys/prctl.h> +#include <sys/shm.h> +#include <sys/socket.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <sys/un.h> +#include <unistd.h> +#include <utmpx.h> + +#ifdef HAVE_PAM +#include <security/pam_appl.h> +#endif + +#ifdef HAVE_SELINUX +#include <selinux/selinux.h> +#endif + +#ifdef HAVE_SECCOMP +#include <seccomp.h> +#endif + +#ifdef HAVE_APPARMOR +#include <sys/apparmor.h> +#endif + +#include <systemd/sd-messages.h> + +#include "systemd-basic/af-list.h" +#include "systemd-basic/alloc-util.h" +#ifdef HAVE_APPARMOR +#include "systemd-shared/apparmor-util.h" +#endif +#include "core/execute.h" +#include "core/namespace.h" +#include "systemd-basic/async.h" +#include "systemd-basic/barrier.h" +#include "systemd-basic/cap-list.h" +#include "systemd-basic/capability-util.h" +#include "systemd-basic/def.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/errno-list.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/glob-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/ioprio.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/rlimit-util.h" +#include "systemd-basic/rm-rf.h" +#ifdef HAVE_SECCOMP +#include "systemd-shared/seccomp-util.h" +#endif +#include "core/unit.h" +#include "systemd-basic/securebits.h" +#include "systemd-basic/selinux-util.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/smack-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/syslog-util.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/util.h" +#include "systemd-shared/utmp-wtmp.h" + +#define IDLE_TIMEOUT_USEC (5*USEC_PER_SEC) +#define IDLE_TIMEOUT2_USEC (1*USEC_PER_SEC) + +/* This assumes there is a 'tty' group */ +#define TTY_MODE 0620 + +#define SNDBUF_SIZE (8*1024*1024) + +static int shift_fds(int fds[], unsigned n_fds) { + int start, restart_from; + + if (n_fds <= 0) + return 0; + + /* Modifies the fds array! (sorts it) */ + + assert(fds); + + start = 0; + for (;;) { + int i; + + restart_from = -1; + + for (i = start; i < (int) n_fds; i++) { + int nfd; + + /* Already at right index? */ + if (fds[i] == i+3) + continue; + + nfd = fcntl(fds[i], F_DUPFD, i + 3); + if (nfd < 0) + return -errno; + + safe_close(fds[i]); + fds[i] = nfd; + + /* Hmm, the fd we wanted isn't free? Then + * let's remember that and try again from here */ + if (nfd != i+3 && restart_from < 0) + restart_from = i; + } + + if (restart_from < 0) + break; + + start = restart_from; + } + + return 0; +} + +static int flags_fds(const int fds[], unsigned n_fds, bool nonblock) { + unsigned i; + int r; + + if (n_fds <= 0) + return 0; + + assert(fds); + + /* Drops/Sets O_NONBLOCK and FD_CLOEXEC from the file flags */ + + for (i = 0; i < n_fds; i++) { + + r = fd_nonblock(fds[i], nonblock); + if (r < 0) + return r; + + /* We unconditionally drop FD_CLOEXEC from the fds, + * since after all we want to pass these fds to our + * children */ + + r = fd_cloexec(fds[i], false); + if (r < 0) + return r; + } + + return 0; +} + +static const char *exec_context_tty_path(const ExecContext *context) { + assert(context); + + if (context->stdio_as_fds) + return NULL; + + if (context->tty_path) + return context->tty_path; + + return "/dev/console"; +} + +static void exec_context_tty_reset(const ExecContext *context, const ExecParameters *p) { + const char *path; + + assert(context); + + path = exec_context_tty_path(context); + + if (context->tty_vhangup) { + if (p && p->stdin_fd >= 0) + (void) terminal_vhangup_fd(p->stdin_fd); + else if (path) + (void) terminal_vhangup(path); + } + + if (context->tty_reset) { + if (p && p->stdin_fd >= 0) + (void) reset_terminal_fd(p->stdin_fd, true); + else if (path) + (void) reset_terminal(path); + } + + if (context->tty_vt_disallocate && path) + (void) vt_disallocate(path); +} + +static bool is_terminal_input(ExecInput i) { + return IN_SET(i, + EXEC_INPUT_TTY, + EXEC_INPUT_TTY_FORCE, + EXEC_INPUT_TTY_FAIL); +} + +static bool is_terminal_output(ExecOutput o) { + return IN_SET(o, + EXEC_OUTPUT_TTY, + EXEC_OUTPUT_SYSLOG_AND_CONSOLE, + EXEC_OUTPUT_KMSG_AND_CONSOLE, + EXEC_OUTPUT_JOURNAL_AND_CONSOLE); +} + +static bool exec_context_needs_term(const ExecContext *c) { + assert(c); + + /* Return true if the execution context suggests we should set $TERM to something useful. */ + + if (is_terminal_input(c->std_input)) + return true; + + if (is_terminal_output(c->std_output)) + return true; + + if (is_terminal_output(c->std_error)) + return true; + + return !!c->tty_path; +} + +static int open_null_as(int flags, int nfd) { + int fd, r; + + assert(nfd >= 0); + + fd = open("/dev/null", flags|O_NOCTTY); + if (fd < 0) + return -errno; + + if (fd != nfd) { + r = dup2(fd, nfd) < 0 ? -errno : nfd; + safe_close(fd); + } else + r = nfd; + + return r; +} + +static int connect_journal_socket(int fd, uid_t uid, gid_t gid) { + union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/journal/stdout", + }; + uid_t olduid = UID_INVALID; + gid_t oldgid = GID_INVALID; + int r; + + if (gid != GID_INVALID) { + oldgid = getgid(); + + r = setegid(gid); + if (r < 0) + return -errno; + } + + if (uid != UID_INVALID) { + olduid = getuid(); + + r = seteuid(uid); + if (r < 0) { + r = -errno; + goto restore_gid; + } + } + + r = connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); + if (r < 0) + r = -errno; + + /* If we fail to restore the uid or gid, things will likely + fail later on. This should only happen if an LSM interferes. */ + + if (uid != UID_INVALID) + (void) seteuid(olduid); + + restore_gid: + if (gid != GID_INVALID) + (void) setegid(oldgid); + + return r; +} + +static int connect_logger_as( + Unit *unit, + const ExecContext *context, + ExecOutput output, + const char *ident, + int nfd, + uid_t uid, + gid_t gid) { + + int fd, r; + + assert(context); + assert(output < _EXEC_OUTPUT_MAX); + assert(ident); + assert(nfd >= 0); + + fd = socket(AF_UNIX, SOCK_STREAM, 0); + if (fd < 0) + return -errno; + + r = connect_journal_socket(fd, uid, gid); + if (r < 0) + return r; + + if (shutdown(fd, SHUT_RD) < 0) { + safe_close(fd); + return -errno; + } + + (void) fd_inc_sndbuf(fd, SNDBUF_SIZE); + + dprintf(fd, + "%s\n" + "%s\n" + "%i\n" + "%i\n" + "%i\n" + "%i\n" + "%i\n", + context->syslog_identifier ? context->syslog_identifier : ident, + unit->id, + context->syslog_priority, + !!context->syslog_level_prefix, + output == EXEC_OUTPUT_SYSLOG || output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE, + output == EXEC_OUTPUT_KMSG || output == EXEC_OUTPUT_KMSG_AND_CONSOLE, + is_terminal_output(output)); + + if (fd == nfd) + return nfd; + + r = dup2(fd, nfd) < 0 ? -errno : nfd; + safe_close(fd); + + return r; +} +static int open_terminal_as(const char *path, mode_t mode, int nfd) { + int fd, r; + + assert(path); + assert(nfd >= 0); + + fd = open_terminal(path, mode | O_NOCTTY); + if (fd < 0) + return fd; + + if (fd != nfd) { + r = dup2(fd, nfd) < 0 ? -errno : nfd; + safe_close(fd); + } else + r = nfd; + + return r; +} + +static int fixup_input(ExecInput std_input, int socket_fd, bool apply_tty_stdin) { + + if (is_terminal_input(std_input) && !apply_tty_stdin) + return EXEC_INPUT_NULL; + + if (std_input == EXEC_INPUT_SOCKET && socket_fd < 0) + return EXEC_INPUT_NULL; + + return std_input; +} + +static int fixup_output(ExecOutput std_output, int socket_fd) { + + if (std_output == EXEC_OUTPUT_SOCKET && socket_fd < 0) + return EXEC_OUTPUT_INHERIT; + + return std_output; +} + +static int setup_input( + const ExecContext *context, + const ExecParameters *params, + int socket_fd, + int named_iofds[3]) { + + ExecInput i; + + assert(context); + assert(params); + + if (params->stdin_fd >= 0) { + if (dup2(params->stdin_fd, STDIN_FILENO) < 0) + return -errno; + + /* Try to make this the controlling tty, if it is a tty, and reset it */ + (void) ioctl(STDIN_FILENO, TIOCSCTTY, context->std_input == EXEC_INPUT_TTY_FORCE); + (void) reset_terminal_fd(STDIN_FILENO, true); + + return STDIN_FILENO; + } + + i = fixup_input(context->std_input, socket_fd, params->flags & EXEC_APPLY_TTY_STDIN); + + switch (i) { + + case EXEC_INPUT_NULL: + return open_null_as(O_RDONLY, STDIN_FILENO); + + case EXEC_INPUT_TTY: + case EXEC_INPUT_TTY_FORCE: + case EXEC_INPUT_TTY_FAIL: { + int fd, r; + + fd = acquire_terminal(exec_context_tty_path(context), + i == EXEC_INPUT_TTY_FAIL, + i == EXEC_INPUT_TTY_FORCE, + false, + USEC_INFINITY); + if (fd < 0) + return fd; + + if (fd != STDIN_FILENO) { + r = dup2(fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; + safe_close(fd); + } else + r = STDIN_FILENO; + + return r; + } + + case EXEC_INPUT_SOCKET: + return dup2(socket_fd, STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; + + case EXEC_INPUT_NAMED_FD: + (void) fd_nonblock(named_iofds[STDIN_FILENO], false); + return dup2(named_iofds[STDIN_FILENO], STDIN_FILENO) < 0 ? -errno : STDIN_FILENO; + + default: + assert_not_reached("Unknown input type"); + } +} + +static int setup_output( + Unit *unit, + const ExecContext *context, + const ExecParameters *params, + int fileno, + int socket_fd, + int named_iofds[3], + const char *ident, + uid_t uid, + gid_t gid, + dev_t *journal_stream_dev, + ino_t *journal_stream_ino) { + + ExecOutput o; + ExecInput i; + int r; + + assert(unit); + assert(context); + assert(params); + assert(ident); + assert(journal_stream_dev); + assert(journal_stream_ino); + + if (fileno == STDOUT_FILENO && params->stdout_fd >= 0) { + + if (dup2(params->stdout_fd, STDOUT_FILENO) < 0) + return -errno; + + return STDOUT_FILENO; + } + + if (fileno == STDERR_FILENO && params->stderr_fd >= 0) { + if (dup2(params->stderr_fd, STDERR_FILENO) < 0) + return -errno; + + return STDERR_FILENO; + } + + i = fixup_input(context->std_input, socket_fd, params->flags & EXEC_APPLY_TTY_STDIN); + o = fixup_output(context->std_output, socket_fd); + + if (fileno == STDERR_FILENO) { + ExecOutput e; + e = fixup_output(context->std_error, socket_fd); + + /* This expects the input and output are already set up */ + + /* Don't change the stderr file descriptor if we inherit all + * the way and are not on a tty */ + if (e == EXEC_OUTPUT_INHERIT && + o == EXEC_OUTPUT_INHERIT && + i == EXEC_INPUT_NULL && + !is_terminal_input(context->std_input) && + getppid () != 1) + return fileno; + + /* Duplicate from stdout if possible */ + if ((e == o && e != EXEC_OUTPUT_NAMED_FD) || e == EXEC_OUTPUT_INHERIT) + return dup2(STDOUT_FILENO, fileno) < 0 ? -errno : fileno; + + o = e; + + } else if (o == EXEC_OUTPUT_INHERIT) { + /* If input got downgraded, inherit the original value */ + if (i == EXEC_INPUT_NULL && is_terminal_input(context->std_input)) + return open_terminal_as(exec_context_tty_path(context), O_WRONLY, fileno); + + /* If the input is connected to anything that's not a /dev/null, inherit that... */ + if (i != EXEC_INPUT_NULL) + return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; + + /* If we are not started from PID 1 we just inherit STDOUT from our parent process. */ + if (getppid() != 1) + return fileno; + + /* We need to open /dev/null here anew, to get the right access mode. */ + return open_null_as(O_WRONLY, fileno); + } + + switch (o) { + + case EXEC_OUTPUT_NULL: + return open_null_as(O_WRONLY, fileno); + + case EXEC_OUTPUT_TTY: + if (is_terminal_input(i)) + return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; + + /* We don't reset the terminal if this is just about output */ + return open_terminal_as(exec_context_tty_path(context), O_WRONLY, fileno); + + case EXEC_OUTPUT_SYSLOG: + case EXEC_OUTPUT_SYSLOG_AND_CONSOLE: + case EXEC_OUTPUT_KMSG: + case EXEC_OUTPUT_KMSG_AND_CONSOLE: + case EXEC_OUTPUT_JOURNAL: + case EXEC_OUTPUT_JOURNAL_AND_CONSOLE: + r = connect_logger_as(unit, context, o, ident, fileno, uid, gid); + if (r < 0) { + log_unit_error_errno(unit, r, "Failed to connect %s to the journal socket, ignoring: %m", fileno == STDOUT_FILENO ? "stdout" : "stderr"); + r = open_null_as(O_WRONLY, fileno); + } else { + struct stat st; + + /* If we connected this fd to the journal via a stream, patch the device/inode into the passed + * parameters, but only then. This is useful so that we can set $JOURNAL_STREAM that permits + * services to detect whether they are connected to the journal or not. */ + + if (fstat(fileno, &st) >= 0) { + *journal_stream_dev = st.st_dev; + *journal_stream_ino = st.st_ino; + } + } + return r; + + case EXEC_OUTPUT_SOCKET: + assert(socket_fd >= 0); + return dup2(socket_fd, fileno) < 0 ? -errno : fileno; + + case EXEC_OUTPUT_NAMED_FD: + (void) fd_nonblock(named_iofds[fileno], false); + return dup2(named_iofds[fileno], fileno) < 0 ? -errno : fileno; + + default: + assert_not_reached("Unknown error type"); + } +} + +static int chown_terminal(int fd, uid_t uid) { + struct stat st; + + assert(fd >= 0); + + /* Before we chown/chmod the TTY, let's ensure this is actually a tty */ + if (isatty(fd) < 1) + return 0; + + /* This might fail. What matters are the results. */ + (void) fchown(fd, uid, -1); + (void) fchmod(fd, TTY_MODE); + + if (fstat(fd, &st) < 0) + return -errno; + + if (st.st_uid != uid || (st.st_mode & 0777) != TTY_MODE) + return -EPERM; + + return 0; +} + +static int setup_confirm_stdio(int *_saved_stdin, int *_saved_stdout) { + _cleanup_close_ int fd = -1, saved_stdin = -1, saved_stdout = -1; + int r; + + assert(_saved_stdin); + assert(_saved_stdout); + + saved_stdin = fcntl(STDIN_FILENO, F_DUPFD, 3); + if (saved_stdin < 0) + return -errno; + + saved_stdout = fcntl(STDOUT_FILENO, F_DUPFD, 3); + if (saved_stdout < 0) + return -errno; + + fd = acquire_terminal( + "/dev/console", + false, + false, + false, + DEFAULT_CONFIRM_USEC); + if (fd < 0) + return fd; + + r = chown_terminal(fd, getuid()); + if (r < 0) + return r; + + r = reset_terminal_fd(fd, true); + if (r < 0) + return r; + + if (dup2(fd, STDIN_FILENO) < 0) + return -errno; + + if (dup2(fd, STDOUT_FILENO) < 0) + return -errno; + + if (fd >= 2) + safe_close(fd); + fd = -1; + + *_saved_stdin = saved_stdin; + *_saved_stdout = saved_stdout; + + saved_stdin = saved_stdout = -1; + + return 0; +} + +_printf_(1, 2) static int write_confirm_message(const char *format, ...) { + _cleanup_close_ int fd = -1; + va_list ap; + + assert(format); + + fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (fd < 0) + return fd; + + va_start(ap, format); + vdprintf(fd, format, ap); + va_end(ap); + + return 0; +} + +static int restore_confirm_stdio(int *saved_stdin, int *saved_stdout) { + int r = 0; + + assert(saved_stdin); + assert(saved_stdout); + + release_terminal(); + + if (*saved_stdin >= 0) + if (dup2(*saved_stdin, STDIN_FILENO) < 0) + r = -errno; + + if (*saved_stdout >= 0) + if (dup2(*saved_stdout, STDOUT_FILENO) < 0) + r = -errno; + + *saved_stdin = safe_close(*saved_stdin); + *saved_stdout = safe_close(*saved_stdout); + + return r; +} + +static int ask_for_confirmation(char *response, char **argv) { + int saved_stdout = -1, saved_stdin = -1, r; + _cleanup_free_ char *line = NULL; + + r = setup_confirm_stdio(&saved_stdin, &saved_stdout); + if (r < 0) + return r; + + line = exec_command_line(argv); + if (!line) + return -ENOMEM; + + r = ask_char(response, "yns", "Execute %s? [Yes, No, Skip] ", line); + + restore_confirm_stdio(&saved_stdin, &saved_stdout); + + return r; +} + +static int get_fixed_user(const ExecContext *c, const char **user, + uid_t *uid, gid_t *gid, + const char **home, const char **shell) { + int r; + const char *name; + + assert(c); + + if (!c->user) + return 0; + + /* Note that we don't set $HOME or $SHELL if they are not particularly enlightening anyway + * (i.e. are "/" or "/bin/nologin"). */ + + name = c->user; + r = get_user_creds_clean(&name, uid, gid, home, shell); + if (r < 0) + return r; + + *user = name; + return 0; +} + +static int get_fixed_group(const ExecContext *c, const char **group, gid_t *gid) { + int r; + const char *name; + + assert(c); + + if (!c->group) + return 0; + + name = c->group; + r = get_group_creds(&name, gid); + if (r < 0) + return r; + + *group = name; + return 0; +} + +static int get_supplementary_groups(const ExecContext *c, const char *user, + const char *group, gid_t gid, + gid_t **supplementary_gids, int *ngids) { + char **i; + int r, k = 0; + int ngroups_max; + bool keep_groups = false; + gid_t *groups = NULL; + _cleanup_free_ gid_t *l_gids = NULL; + + assert(c); + + /* + * If user is given, then lookup GID and supplementary groups list. + * We avoid NSS lookups for gid=0. Also we have to initialize groups + * here and as early as possible so we keep the list of supplementary + * groups of the caller. + */ + if (user && gid_is_valid(gid) && gid != 0) { + /* First step, initialize groups from /etc/groups */ + if (initgroups(user, gid) < 0) + return -errno; + + keep_groups = true; + } + + if (!c->supplementary_groups) + return 0; + + /* + * If SupplementaryGroups= was passed then NGROUPS_MAX has to + * be positive, otherwise fail. + */ + errno = 0; + ngroups_max = (int) sysconf(_SC_NGROUPS_MAX); + if (ngroups_max <= 0) { + if (errno > 0) + return -errno; + else + return -EOPNOTSUPP; /* For all other values */ + } + + l_gids = new(gid_t, ngroups_max); + if (!l_gids) + return -ENOMEM; + + if (keep_groups) { + /* + * Lookup the list of groups that the user belongs to, we + * avoid NSS lookups here too for gid=0. + */ + k = ngroups_max; + if (getgrouplist(user, gid, l_gids, &k) < 0) + return -EINVAL; + } else + k = 0; + + STRV_FOREACH(i, c->supplementary_groups) { + const char *g; + + if (k >= ngroups_max) + return -E2BIG; + + g = *i; + r = get_group_creds(&g, l_gids+k); + if (r < 0) + return r; + + k++; + } + + /* + * Sets ngids to zero to drop all supplementary groups, happens + * when we are under root and SupplementaryGroups= is empty. + */ + if (k == 0) { + *ngids = 0; + return 0; + } + + /* Otherwise get the final list of supplementary groups */ + groups = memdup(l_gids, sizeof(gid_t) * k); + if (!groups) + return -ENOMEM; + + *supplementary_gids = groups; + *ngids = k; + + groups = NULL; + + return 0; +} + +static int enforce_groups(const ExecContext *context, gid_t gid, + gid_t *supplementary_gids, int ngids) { + int r; + + assert(context); + + /* Handle SupplementaryGroups= even if it is empty */ + if (context->supplementary_groups) { + r = maybe_setgroups(ngids, supplementary_gids); + if (r < 0) + return r; + } + + if (gid_is_valid(gid)) { + /* Then set our gids */ + if (setresgid(gid, gid, gid) < 0) + return -errno; + } + + return 0; +} + +static int enforce_user(const ExecContext *context, uid_t uid) { + assert(context); + + if (!uid_is_valid(uid)) + return 0; + + /* Sets (but doesn't look up) the uid and make sure we keep the + * capabilities while doing so. */ + + if (context->capability_ambient_set != 0) { + + /* First step: If we need to keep capabilities but + * drop privileges we need to make sure we keep our + * caps, while we drop privileges. */ + if (uid != 0) { + int sb = context->secure_bits | 1<<SECURE_KEEP_CAPS; + + if (prctl(PR_GET_SECUREBITS) != sb) + if (prctl(PR_SET_SECUREBITS, sb) < 0) + return -errno; + } + } + + /* Second step: actually set the uids */ + if (setresuid(uid, uid, uid) < 0) + return -errno; + + /* At this point we should have all necessary capabilities but + are otherwise a normal user. However, the caps might got + corrupted due to the setresuid() so we need clean them up + later. This is done outside of this call. */ + + return 0; +} + +#ifdef HAVE_PAM + +static int null_conv( + int num_msg, + const struct pam_message **msg, + struct pam_response **resp, + void *appdata_ptr) { + + /* We don't support conversations */ + + return PAM_CONV_ERR; +} + +#endif + +static int setup_pam( + const char *name, + const char *user, + uid_t uid, + gid_t gid, + const char *tty, + char ***env, + int fds[], unsigned n_fds) { + +#ifdef HAVE_PAM + + static const struct pam_conv conv = { + .conv = null_conv, + .appdata_ptr = NULL + }; + + _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; + pam_handle_t *handle = NULL; + sigset_t old_ss; + int pam_code = PAM_SUCCESS, r; + char **nv, **e = NULL; + bool close_session = false; + pid_t pam_pid = 0, parent_pid; + int flags = 0; + + assert(name); + assert(user); + assert(env); + + /* We set up PAM in the parent process, then fork. The child + * will then stay around until killed via PR_GET_PDEATHSIG or + * systemd via the cgroup logic. It will then remove the PAM + * session again. The parent process will exec() the actual + * daemon. We do things this way to ensure that the main PID + * of the daemon is the one we initially fork()ed. */ + + r = barrier_create(&barrier); + if (r < 0) + goto fail; + + if (log_get_max_level() < LOG_DEBUG) + flags |= PAM_SILENT; + + pam_code = pam_start(name, user, &conv, &handle); + if (pam_code != PAM_SUCCESS) { + handle = NULL; + goto fail; + } + + if (tty) { + pam_code = pam_set_item(handle, PAM_TTY, tty); + if (pam_code != PAM_SUCCESS) + goto fail; + } + + STRV_FOREACH(nv, *env) { + pam_code = pam_putenv(handle, *nv); + if (pam_code != PAM_SUCCESS) + goto fail; + } + + pam_code = pam_acct_mgmt(handle, flags); + if (pam_code != PAM_SUCCESS) + goto fail; + + pam_code = pam_open_session(handle, flags); + if (pam_code != PAM_SUCCESS) + goto fail; + + close_session = true; + + e = pam_getenvlist(handle); + if (!e) { + pam_code = PAM_BUF_ERR; + goto fail; + } + + /* Block SIGTERM, so that we know that it won't get lost in + * the child */ + + assert_se(sigprocmask_many(SIG_BLOCK, &old_ss, SIGTERM, -1) >= 0); + + parent_pid = getpid(); + + pam_pid = fork(); + if (pam_pid < 0) { + r = -errno; + goto fail; + } + + if (pam_pid == 0) { + int sig, ret = EXIT_PAM; + + /* The child's job is to reset the PAM session on + * termination */ + barrier_set_role(&barrier, BARRIER_CHILD); + + /* This string must fit in 10 chars (i.e. the length + * of "/sbin/init"), to look pretty in /bin/ps */ + rename_process("(sd-pam)"); + + /* Make sure we don't keep open the passed fds in this + child. We assume that otherwise only those fds are + open here that have been opened by PAM. */ + close_many(fds, n_fds); + + /* Drop privileges - we don't need any to pam_close_session + * and this will make PR_SET_PDEATHSIG work in most cases. + * If this fails, ignore the error - but expect sd-pam threads + * to fail to exit normally */ + + r = maybe_setgroups(0, NULL); + if (r < 0) + log_warning_errno(r, "Failed to setgroups() in sd-pam: %m"); + if (setresgid(gid, gid, gid) < 0) + log_warning_errno(errno, "Failed to setresgid() in sd-pam: %m"); + if (setresuid(uid, uid, uid) < 0) + log_warning_errno(errno, "Failed to setresuid() in sd-pam: %m"); + + (void) ignore_signals(SIGPIPE, -1); + + /* Wait until our parent died. This will only work if + * the above setresuid() succeeds, otherwise the kernel + * will not allow unprivileged parents kill their privileged + * children this way. We rely on the control groups kill logic + * to do the rest for us. */ + if (prctl(PR_SET_PDEATHSIG, SIGTERM) < 0) + goto child_finish; + + /* Tell the parent that our setup is done. This is especially + * important regarding dropping privileges. Otherwise, unit + * setup might race against our setresuid(2) call. */ + barrier_place(&barrier); + + /* Check if our parent process might already have + * died? */ + if (getppid() == parent_pid) { + sigset_t ss; + + assert_se(sigemptyset(&ss) >= 0); + assert_se(sigaddset(&ss, SIGTERM) >= 0); + + for (;;) { + if (sigwait(&ss, &sig) < 0) { + if (errno == EINTR) + continue; + + goto child_finish; + } + + assert(sig == SIGTERM); + break; + } + } + + /* If our parent died we'll end the session */ + if (getppid() != parent_pid) { + pam_code = pam_close_session(handle, flags); + if (pam_code != PAM_SUCCESS) + goto child_finish; + } + + ret = 0; + + child_finish: + pam_end(handle, pam_code | flags); + _exit(ret); + } + + barrier_set_role(&barrier, BARRIER_PARENT); + + /* If the child was forked off successfully it will do all the + * cleanups, so forget about the handle here. */ + handle = NULL; + + /* Unblock SIGTERM again in the parent */ + assert_se(sigprocmask(SIG_SETMASK, &old_ss, NULL) >= 0); + + /* We close the log explicitly here, since the PAM modules + * might have opened it, but we don't want this fd around. */ + closelog(); + + /* Synchronously wait for the child to initialize. We don't care for + * errors as we cannot recover. However, warn loudly if it happens. */ + if (!barrier_place_and_sync(&barrier)) + log_error("PAM initialization failed"); + + strv_free(*env); + *env = e; + + return 0; + +fail: + if (pam_code != PAM_SUCCESS) { + log_error("PAM failed: %s", pam_strerror(handle, pam_code)); + r = -EPERM; /* PAM errors do not map to errno */ + } else + log_error_errno(r, "PAM failed: %m"); + + if (handle) { + if (close_session) + pam_code = pam_close_session(handle, flags); + + pam_end(handle, pam_code | flags); + } + + strv_free(e); + closelog(); + + return r; +#else + return 0; +#endif +} + +static void rename_process_from_path(const char *path) { + char process_name[11]; + const char *p; + size_t l; + + /* This resulting string must fit in 10 chars (i.e. the length + * of "/sbin/init") to look pretty in /bin/ps */ + + p = basename(path); + if (isempty(p)) { + rename_process("(...)"); + return; + } + + l = strlen(p); + if (l > 8) { + /* The end of the process name is usually more + * interesting, since the first bit might just be + * "systemd-" */ + p = p + l - 8; + l = 8; + } + + process_name[0] = '('; + memcpy(process_name+1, p, l); + process_name[1+l] = ')'; + process_name[1+l+1] = 0; + + rename_process(process_name); +} + +#ifdef HAVE_SECCOMP + +static bool skip_seccomp_unavailable(const Unit* u, const char* msg) { + + if (is_seccomp_available()) + return false; + + log_open(); + log_unit_debug(u, "SECCOMP features not detected in the kernel, skipping %s", msg); + log_close(); + return true; +} + +static int apply_seccomp(const Unit* u, const ExecContext *c) { + uint32_t negative_action, action; + scmp_filter_ctx seccomp; + Iterator i; + void *id; + int r; + + assert(c); + + if (skip_seccomp_unavailable(u, "syscall filtering")) + return 0; + + negative_action = c->syscall_errno == 0 ? SCMP_ACT_KILL : SCMP_ACT_ERRNO(c->syscall_errno); + + seccomp = seccomp_init(c->syscall_whitelist ? negative_action : SCMP_ACT_ALLOW); + if (!seccomp) + return -ENOMEM; + + if (c->syscall_archs) { + + SET_FOREACH(id, c->syscall_archs, i) { + r = seccomp_arch_add(seccomp, PTR_TO_UINT32(id) - 1); + if (r == -EEXIST) + continue; + if (r < 0) + goto finish; + } + + } else { + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) + goto finish; + } + + action = c->syscall_whitelist ? SCMP_ACT_ALLOW : negative_action; + SET_FOREACH(id, c->syscall_filter, i) { + r = seccomp_rule_add(seccomp, action, PTR_TO_INT(id) - 1, 0); + if (r < 0) + goto finish; + } + + r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); + if (r < 0) + goto finish; + + r = seccomp_load(seccomp); + +finish: + seccomp_release(seccomp); + return r; +} + +static int apply_address_families(const Unit* u, const ExecContext *c) { + scmp_filter_ctx seccomp; + Iterator i; + int r; + +#if defined(__i386__) + return 0; +#endif + + assert(c); + + if (skip_seccomp_unavailable(u, "RestrictAddressFamilies=")) + return 0; + + r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW); + if (r < 0) + return r; + + if (c->address_families_whitelist) { + int af, first = 0, last = 0; + void *afp; + + /* If this is a whitelist, we first block the address + * families that are out of range and then everything + * that is not in the set. First, we find the lowest + * and highest address family in the set. */ + + SET_FOREACH(afp, c->address_families, i) { + af = PTR_TO_INT(afp); + + if (af <= 0 || af >= af_max()) + continue; + + if (first == 0 || af < first) + first = af; + + if (last == 0 || af > last) + last = af; + } + + assert((first == 0) == (last == 0)); + + if (first == 0) { + + /* No entries in the valid range, block everything */ + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPROTONOSUPPORT), + SCMP_SYS(socket), + 0); + if (r < 0) + goto finish; + + } else { + + /* Block everything below the first entry */ + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPROTONOSUPPORT), + SCMP_SYS(socket), + 1, + SCMP_A0(SCMP_CMP_LT, first)); + if (r < 0) + goto finish; + + /* Block everything above the last entry */ + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPROTONOSUPPORT), + SCMP_SYS(socket), + 1, + SCMP_A0(SCMP_CMP_GT, last)); + if (r < 0) + goto finish; + + /* Block everything between the first and last + * entry */ + for (af = 1; af < af_max(); af++) { + + if (set_contains(c->address_families, INT_TO_PTR(af))) + continue; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPROTONOSUPPORT), + SCMP_SYS(socket), + 1, + SCMP_A0(SCMP_CMP_EQ, af)); + if (r < 0) + goto finish; + } + } + + } else { + void *af; + + /* If this is a blacklist, then generate one rule for + * each address family that are then combined in OR + * checks. */ + + SET_FOREACH(af, c->address_families, i) { + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPROTONOSUPPORT), + SCMP_SYS(socket), + 1, + SCMP_A0(SCMP_CMP_EQ, PTR_TO_INT(af))); + if (r < 0) + goto finish; + } + } + + r = seccomp_load(seccomp); + +finish: + seccomp_release(seccomp); + return r; +} + +static int apply_memory_deny_write_execute(const Unit* u, const ExecContext *c) { + scmp_filter_ctx seccomp; + int r; + + assert(c); + + if (skip_seccomp_unavailable(u, "MemoryDenyWriteExecute=")) + return 0; + + r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW); + if (r < 0) + return r; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(mmap), + 1, + SCMP_A2(SCMP_CMP_MASKED_EQ, PROT_EXEC|PROT_WRITE, PROT_EXEC|PROT_WRITE)); + if (r < 0) + goto finish; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(mprotect), + 1, + SCMP_A2(SCMP_CMP_MASKED_EQ, PROT_EXEC, PROT_EXEC)); + if (r < 0) + goto finish; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(shmat), + 1, + SCMP_A2(SCMP_CMP_MASKED_EQ, SHM_EXEC, SHM_EXEC)); + if (r < 0) + goto finish; + + r = seccomp_load(seccomp); + +finish: + seccomp_release(seccomp); + return r; +} + +static int apply_restrict_realtime(const Unit* u, const ExecContext *c) { + static const int permitted_policies[] = { + SCHED_OTHER, + SCHED_BATCH, + SCHED_IDLE, + }; + + scmp_filter_ctx seccomp; + unsigned i; + int r, p, max_policy = 0; + + assert(c); + + if (skip_seccomp_unavailable(u, "RestrictRealtime=")) + return 0; + + r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW); + if (r < 0) + return r; + + /* Determine the highest policy constant we want to allow */ + for (i = 0; i < ELEMENTSOF(permitted_policies); i++) + if (permitted_policies[i] > max_policy) + max_policy = permitted_policies[i]; + + /* Go through all policies with lower values than that, and block them -- unless they appear in the + * whitelist. */ + for (p = 0; p < max_policy; p++) { + bool good = false; + + /* Check if this is in the whitelist. */ + for (i = 0; i < ELEMENTSOF(permitted_policies); i++) + if (permitted_policies[i] == p) { + good = true; + break; + } + + if (good) + continue; + + /* Deny this policy */ + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(sched_setscheduler), + 1, + SCMP_A1(SCMP_CMP_EQ, p)); + if (r < 0) + goto finish; + } + + /* Blacklist all other policies, i.e. the ones with higher values. Note that all comparisons are unsigned here, + * hence no need no check for < 0 values. */ + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(sched_setscheduler), + 1, + SCMP_A1(SCMP_CMP_GT, max_policy)); + if (r < 0) + goto finish; + + r = seccomp_load(seccomp); + +finish: + seccomp_release(seccomp); + return r; +} + +static int apply_protect_sysctl(const Unit *u, const ExecContext *c) { + scmp_filter_ctx seccomp; + int r; + + assert(c); + + /* Turn off the legacy sysctl() system call. Many distributions turn this off while building the kernel, but + * let's protect even those systems where this is left on in the kernel. */ + + if (skip_seccomp_unavailable(u, "ProtectKernelTunables=")) + return 0; + + r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW); + if (r < 0) + return r; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(_sysctl), + 0); + if (r < 0) + goto finish; + + r = seccomp_load(seccomp); + +finish: + seccomp_release(seccomp); + return r; +} + +static int apply_protect_kernel_modules(const Unit *u, const ExecContext *c) { + assert(c); + + /* Turn off module syscalls on ProtectKernelModules=yes */ + + if (skip_seccomp_unavailable(u, "ProtectKernelModules=")) + return 0; + + return seccomp_load_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_MODULE, SCMP_ACT_ERRNO(EPERM)); +} + +static int apply_private_devices(const Unit *u, const ExecContext *c) { + assert(c); + + /* If PrivateDevices= is set, also turn off iopl and all @raw-io syscalls. */ + + if (skip_seccomp_unavailable(u, "PrivateDevices=")) + return 0; + + return seccomp_load_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_RAW_IO, SCMP_ACT_ERRNO(EPERM)); +} + +#endif + +static void do_idle_pipe_dance(int idle_pipe[4]) { + assert(idle_pipe); + + idle_pipe[1] = safe_close(idle_pipe[1]); + idle_pipe[2] = safe_close(idle_pipe[2]); + + if (idle_pipe[0] >= 0) { + int r; + + r = fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT_USEC); + + if (idle_pipe[3] >= 0 && r == 0 /* timeout */) { + ssize_t n; + + /* Signal systemd that we are bored and want to continue. */ + n = write(idle_pipe[3], "x", 1); + if (n > 0) + /* Wait for systemd to react to the signal above. */ + fd_wait_for_event(idle_pipe[0], POLLHUP, IDLE_TIMEOUT2_USEC); + } + + idle_pipe[0] = safe_close(idle_pipe[0]); + + } + + idle_pipe[3] = safe_close(idle_pipe[3]); +} + +static int build_environment( + Unit *u, + const ExecContext *c, + const ExecParameters *p, + unsigned n_fds, + const char *home, + const char *username, + const char *shell, + dev_t journal_stream_dev, + ino_t journal_stream_ino, + char ***ret) { + + _cleanup_strv_free_ char **our_env = NULL; + unsigned n_env = 0; + char *x; + + assert(u); + assert(c); + assert(ret); + + our_env = new0(char*, 14); + if (!our_env) + return -ENOMEM; + + if (n_fds > 0) { + _cleanup_free_ char *joined = NULL; + + if (asprintf(&x, "LISTEN_PID="PID_FMT, getpid()) < 0) + return -ENOMEM; + our_env[n_env++] = x; + + if (asprintf(&x, "LISTEN_FDS=%u", n_fds) < 0) + return -ENOMEM; + our_env[n_env++] = x; + + joined = strv_join(p->fd_names, ":"); + if (!joined) + return -ENOMEM; + + x = strjoin("LISTEN_FDNAMES=", joined, NULL); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if ((p->flags & EXEC_SET_WATCHDOG) && p->watchdog_usec > 0) { + if (asprintf(&x, "WATCHDOG_PID="PID_FMT, getpid()) < 0) + return -ENOMEM; + our_env[n_env++] = x; + + if (asprintf(&x, "WATCHDOG_USEC="USEC_FMT, p->watchdog_usec) < 0) + return -ENOMEM; + our_env[n_env++] = x; + } + + /* If this is D-Bus, tell the nss-systemd module, since it relies on being able to use D-Bus look up dynamic + * users via PID 1, possibly dead-locking the dbus daemon. This way it will not use D-Bus to resolve names, but + * check the database directly. */ + if (unit_has_name(u, SPECIAL_DBUS_SERVICE)) { + x = strdup("SYSTEMD_NSS_BYPASS_BUS=1"); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if (home) { + x = strappend("HOME=", home); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if (username) { + x = strappend("LOGNAME=", username); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + + x = strappend("USER=", username); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if (shell) { + x = strappend("SHELL=", shell); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if (!sd_id128_is_null(u->invocation_id)) { + if (asprintf(&x, "INVOCATION_ID=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(u->invocation_id)) < 0) + return -ENOMEM; + + our_env[n_env++] = x; + } + + if (exec_context_needs_term(c)) { + const char *tty_path, *term = NULL; + + tty_path = exec_context_tty_path(c); + + /* If we are forked off PID 1 and we are supposed to operate on /dev/console, then let's try to inherit + * the $TERM set for PID 1. This is useful for containers so that the $TERM the container manager + * passes to PID 1 ends up all the way in the console login shown. */ + + if (path_equal(tty_path, "/dev/console") && getppid() == 1) + term = getenv("TERM"); + if (!term) + term = default_term_for_tty(tty_path); + + x = strappend("TERM=", term); + if (!x) + return -ENOMEM; + our_env[n_env++] = x; + } + + if (journal_stream_dev != 0 && journal_stream_ino != 0) { + if (asprintf(&x, "JOURNAL_STREAM=" DEV_FMT ":" INO_FMT, journal_stream_dev, journal_stream_ino) < 0) + return -ENOMEM; + + our_env[n_env++] = x; + } + + our_env[n_env++] = NULL; + assert(n_env <= 12); + + *ret = our_env; + our_env = NULL; + + return 0; +} + +static int build_pass_environment(const ExecContext *c, char ***ret) { + _cleanup_strv_free_ char **pass_env = NULL; + size_t n_env = 0, n_bufsize = 0; + char **i; + + STRV_FOREACH(i, c->pass_environment) { + _cleanup_free_ char *x = NULL; + char *v; + + v = getenv(*i); + if (!v) + continue; + x = strjoin(*i, "=", v, NULL); + if (!x) + return -ENOMEM; + if (!GREEDY_REALLOC(pass_env, n_bufsize, n_env + 2)) + return -ENOMEM; + pass_env[n_env++] = x; + pass_env[n_env] = NULL; + x = NULL; + } + + *ret = pass_env; + pass_env = NULL; + + return 0; +} + +static bool exec_needs_mount_namespace( + const ExecContext *context, + const ExecParameters *params, + ExecRuntime *runtime) { + + assert(context); + assert(params); + + if (!strv_isempty(context->read_write_paths) || + !strv_isempty(context->read_only_paths) || + !strv_isempty(context->inaccessible_paths)) + return true; + + if (context->mount_flags != 0) + return true; + + if (context->private_tmp && runtime && (runtime->tmp_dir || runtime->var_tmp_dir)) + return true; + + if (context->private_devices || + context->protect_system != PROTECT_SYSTEM_NO || + context->protect_home != PROTECT_HOME_NO || + context->protect_kernel_tunables || + context->protect_kernel_modules || + context->protect_control_groups) + return true; + + return false; +} + +static int setup_private_users(uid_t uid, gid_t gid) { + _cleanup_free_ char *uid_map = NULL, *gid_map = NULL; + _cleanup_close_pair_ int errno_pipe[2] = { -1, -1 }; + _cleanup_close_ int unshare_ready_fd = -1; + _cleanup_(sigkill_waitp) pid_t pid = 0; + uint64_t c = 1; + siginfo_t si; + ssize_t n; + int r; + + /* Set up a user namespace and map root to root, the selected UID/GID to itself, and everything else to + * nobody. In order to be able to write this mapping we need CAP_SETUID in the original user namespace, which + * we however lack after opening the user namespace. To work around this we fork() a temporary child process, + * which waits for the parent to create the new user namespace while staying in the original namespace. The + * child then writes the UID mapping, under full privileges. The parent waits for the child to finish and + * continues execution normally. */ + + if (uid != 0 && uid_is_valid(uid)) + asprintf(&uid_map, + "0 0 1\n" /* Map root → root */ + UID_FMT " " UID_FMT " 1\n", /* Map $UID → $UID */ + uid, uid); + else + uid_map = strdup("0 0 1\n"); /* The case where the above is the same */ + if (!uid_map) + return -ENOMEM; + + if (gid != 0 && gid_is_valid(gid)) + asprintf(&gid_map, + "0 0 1\n" /* Map root → root */ + GID_FMT " " GID_FMT " 1\n", /* Map $GID → $GID */ + gid, gid); + else + gid_map = strdup("0 0 1\n"); /* The case where the above is the same */ + if (!gid_map) + return -ENOMEM; + + /* Create a communication channel so that the parent can tell the child when it finished creating the user + * namespace. */ + unshare_ready_fd = eventfd(0, EFD_CLOEXEC); + if (unshare_ready_fd < 0) + return -errno; + + /* Create a communication channel so that the child can tell the parent a proper error code in case it + * failed. */ + if (pipe2(errno_pipe, O_CLOEXEC) < 0) + return -errno; + + pid = fork(); + if (pid < 0) + return -errno; + + if (pid == 0) { + _cleanup_close_ int fd = -1; + const char *a; + pid_t ppid; + + /* Child process, running in the original user namespace. Let's update the parent's UID/GID map from + * here, after the parent opened its own user namespace. */ + + ppid = getppid(); + errno_pipe[0] = safe_close(errno_pipe[0]); + + /* Wait until the parent unshared the user namespace */ + if (read(unshare_ready_fd, &c, sizeof(c)) < 0) { + r = -errno; + goto child_fail; + } + + /* Disable the setgroups() system call in the child user namespace, for good. */ + a = procfs_file_alloca(ppid, "setgroups"); + fd = open(a, O_WRONLY|O_CLOEXEC); + if (fd < 0) { + if (errno != ENOENT) { + r = -errno; + goto child_fail; + } + + /* If the file is missing the kernel is too old, let's continue anyway. */ + } else { + if (write(fd, "deny\n", 5) < 0) { + r = -errno; + goto child_fail; + } + + fd = safe_close(fd); + } + + /* First write the GID map */ + a = procfs_file_alloca(ppid, "gid_map"); + fd = open(a, O_WRONLY|O_CLOEXEC); + if (fd < 0) { + r = -errno; + goto child_fail; + } + if (write(fd, gid_map, strlen(gid_map)) < 0) { + r = -errno; + goto child_fail; + } + fd = safe_close(fd); + + /* The write the UID map */ + a = procfs_file_alloca(ppid, "uid_map"); + fd = open(a, O_WRONLY|O_CLOEXEC); + if (fd < 0) { + r = -errno; + goto child_fail; + } + if (write(fd, uid_map, strlen(uid_map)) < 0) { + r = -errno; + goto child_fail; + } + + _exit(EXIT_SUCCESS); + + child_fail: + (void) write(errno_pipe[1], &r, sizeof(r)); + _exit(EXIT_FAILURE); + } + + errno_pipe[1] = safe_close(errno_pipe[1]); + + if (unshare(CLONE_NEWUSER) < 0) + return -errno; + + /* Let the child know that the namespace is ready now */ + if (write(unshare_ready_fd, &c, sizeof(c)) < 0) + return -errno; + + /* Try to read an error code from the child */ + n = read(errno_pipe[0], &r, sizeof(r)); + if (n < 0) + return -errno; + if (n == sizeof(r)) { /* an error code was sent to us */ + if (r < 0) + return r; + return -EIO; + } + if (n != 0) /* on success we should have read 0 bytes */ + return -EIO; + + r = wait_for_terminate(pid, &si); + if (r < 0) + return r; + pid = 0; + + /* If something strange happened with the child, let's consider this fatal, too */ + if (si.si_code != CLD_EXITED || si.si_status != 0) + return -EIO; + + return 0; +} + +static int setup_runtime_directory( + const ExecContext *context, + const ExecParameters *params, + uid_t uid, + gid_t gid) { + + char **rt; + int r; + + assert(context); + assert(params); + + STRV_FOREACH(rt, context->runtime_directory) { + _cleanup_free_ char *p; + + p = strjoin(params->runtime_prefix, "/", *rt, NULL); + if (!p) + return -ENOMEM; + + r = mkdir_p_label(p, context->runtime_directory_mode); + if (r < 0) + return r; + + r = chmod_and_chown(p, context->runtime_directory_mode, uid, gid); + if (r < 0) + return r; + } + + return 0; +} + +static int setup_smack( + const ExecContext *context, + const ExecCommand *command) { + +#ifdef HAVE_SMACK + int r; + + assert(context); + assert(command); + + if (!mac_smack_use()) + return 0; + + if (context->smack_process_label) { + r = mac_smack_apply_pid(0, context->smack_process_label); + if (r < 0) + return r; + } +#ifdef SMACK_DEFAULT_PROCESS_LABEL + else { + _cleanup_free_ char *exec_label = NULL; + + r = mac_smack_read(command->path, SMACK_ATTR_EXEC, &exec_label); + if (r < 0 && r != -ENODATA && r != -EOPNOTSUPP) + return r; + + r = mac_smack_apply_pid(0, exec_label ? : SMACK_DEFAULT_PROCESS_LABEL); + if (r < 0) + return r; + } +#endif +#endif + + return 0; +} + +static int compile_read_write_paths( + const ExecContext *context, + const ExecParameters *params, + char ***ret) { + + _cleanup_strv_free_ char **l = NULL; + char **rt; + + /* Compile the list of writable paths. This is the combination of the explicitly configured paths, plus all + * runtime directories. */ + + if (strv_isempty(context->read_write_paths) && + strv_isempty(context->runtime_directory)) { + *ret = NULL; /* NOP if neither is set */ + return 0; + } + + l = strv_copy(context->read_write_paths); + if (!l) + return -ENOMEM; + + STRV_FOREACH(rt, context->runtime_directory) { + char *s; + + s = strjoin(params->runtime_prefix, "/", *rt, NULL); + if (!s) + return -ENOMEM; + + if (strv_consume(&l, s) < 0) + return -ENOMEM; + } + + *ret = l; + l = NULL; + + return 0; +} + +static int apply_mount_namespace(Unit *u, const ExecContext *context, + const ExecParameters *params, + ExecRuntime *runtime) { + int r; + _cleanup_free_ char **rw = NULL; + char *tmp = NULL, *var = NULL; + const char *root_dir = NULL; + NameSpaceInfo ns_info = { + .private_dev = context->private_devices, + .protect_control_groups = context->protect_control_groups, + .protect_kernel_tunables = context->protect_kernel_tunables, + .protect_kernel_modules = context->protect_kernel_modules, + }; + + assert(context); + + /* The runtime struct only contains the parent of the private /tmp, + * which is non-accessible to world users. Inside of it there's a /tmp + * that is sticky, and that's the one we want to use here. */ + + if (context->private_tmp && runtime) { + if (runtime->tmp_dir) + tmp = strjoina(runtime->tmp_dir, "/tmp"); + if (runtime->var_tmp_dir) + var = strjoina(runtime->var_tmp_dir, "/tmp"); + } + + r = compile_read_write_paths(context, params, &rw); + if (r < 0) + return r; + + if (params->flags & EXEC_APPLY_CHROOT) + root_dir = context->root_directory; + + r = setup_namespace(root_dir, &ns_info, rw, + context->read_only_paths, + context->inaccessible_paths, + tmp, + var, + context->protect_home, + context->protect_system, + context->mount_flags); + + /* If we couldn't set up the namespace this is probably due to a + * missing capability. In this case, silently proceeed. */ + if (IN_SET(r, -EPERM, -EACCES)) { + log_open(); + log_unit_debug_errno(u, r, "Failed to set up namespace, assuming containerized execution, ignoring: %m"); + log_close(); + r = 0; + } + + return r; +} + +static int apply_working_directory(const ExecContext *context, + const ExecParameters *params, + const char *home, + const bool needs_mount_ns) { + const char *d; + const char *wd; + + assert(context); + + if (context->working_directory_home) + wd = home; + else if (context->working_directory) + wd = context->working_directory; + else + wd = "/"; + + if (params->flags & EXEC_APPLY_CHROOT) { + if (!needs_mount_ns && context->root_directory) + if (chroot(context->root_directory) < 0) + return -errno; + + d = wd; + } else + d = strjoina(strempty(context->root_directory), "/", strempty(wd)); + + if (chdir(d) < 0 && !context->working_directory_missing_ok) + return -errno; + + return 0; +} + +static void append_socket_pair(int *array, unsigned *n, int pair[2]) { + assert(array); + assert(n); + + if (!pair) + return; + + if (pair[0] >= 0) + array[(*n)++] = pair[0]; + if (pair[1] >= 0) + array[(*n)++] = pair[1]; +} + +static int close_remaining_fds( + const ExecParameters *params, + ExecRuntime *runtime, + DynamicCreds *dcreds, + int user_lookup_fd, + int socket_fd, + int *fds, unsigned n_fds) { + + unsigned n_dont_close = 0; + int dont_close[n_fds + 12]; + + assert(params); + + if (params->stdin_fd >= 0) + dont_close[n_dont_close++] = params->stdin_fd; + if (params->stdout_fd >= 0) + dont_close[n_dont_close++] = params->stdout_fd; + if (params->stderr_fd >= 0) + dont_close[n_dont_close++] = params->stderr_fd; + + if (socket_fd >= 0) + dont_close[n_dont_close++] = socket_fd; + if (n_fds > 0) { + memcpy(dont_close + n_dont_close, fds, sizeof(int) * n_fds); + n_dont_close += n_fds; + } + + if (runtime) + append_socket_pair(dont_close, &n_dont_close, runtime->netns_storage_socket); + + if (dcreds) { + if (dcreds->user) + append_socket_pair(dont_close, &n_dont_close, dcreds->user->storage_socket); + if (dcreds->group) + append_socket_pair(dont_close, &n_dont_close, dcreds->group->storage_socket); + } + + if (user_lookup_fd >= 0) + dont_close[n_dont_close++] = user_lookup_fd; + + return close_all_fds(dont_close, n_dont_close); +} + +static bool context_has_address_families(const ExecContext *c) { + assert(c); + + return c->address_families_whitelist || + !set_isempty(c->address_families); +} + +static bool context_has_syscall_filters(const ExecContext *c) { + assert(c); + + return c->syscall_whitelist || + !set_isempty(c->syscall_filter) || + !set_isempty(c->syscall_archs); +} + +static bool context_has_no_new_privileges(const ExecContext *c) { + assert(c); + + if (c->no_new_privileges) + return true; + + if (have_effective_cap(CAP_SYS_ADMIN)) /* if we are privileged, we don't need NNP */ + return false; + + return context_has_address_families(c) || /* we need NNP if we have any form of seccomp and are unprivileged */ + c->memory_deny_write_execute || + c->restrict_realtime || + c->protect_kernel_tunables || + c->protect_kernel_modules || + c->private_devices || + context_has_syscall_filters(c); +} + +static int send_user_lookup( + Unit *unit, + int user_lookup_fd, + uid_t uid, + gid_t gid) { + + assert(unit); + + /* Send the resolved UID/GID to PID 1 after we learnt it. We send a single datagram, containing the UID/GID + * data as well as the unit name. Note that we suppress sending this if no user/group to resolve was + * specified. */ + + if (user_lookup_fd < 0) + return 0; + + if (!uid_is_valid(uid) && !gid_is_valid(gid)) + return 0; + + if (writev(user_lookup_fd, + (struct iovec[]) { + { .iov_base = &uid, .iov_len = sizeof(uid) }, + { .iov_base = &gid, .iov_len = sizeof(gid) }, + { .iov_base = unit->id, .iov_len = strlen(unit->id) }}, 3) < 0) + return -errno; + + return 0; +} + +static int exec_child( + Unit *unit, + ExecCommand *command, + const ExecContext *context, + const ExecParameters *params, + ExecRuntime *runtime, + DynamicCreds *dcreds, + char **argv, + int socket_fd, + int named_iofds[3], + int *fds, unsigned n_fds, + char **files_env, + int user_lookup_fd, + int *exit_status) { + + _cleanup_strv_free_ char **our_env = NULL, **pass_env = NULL, **accum_env = NULL, **final_argv = NULL; + _cleanup_free_ char *mac_selinux_context_net = NULL; + _cleanup_free_ gid_t *supplementary_gids = NULL; + const char *username = NULL, *groupname = NULL; + const char *home = NULL, *shell = NULL; + dev_t journal_stream_dev = 0; + ino_t journal_stream_ino = 0; + bool needs_mount_namespace; + uid_t uid = UID_INVALID; + gid_t gid = GID_INVALID; + int i, r, ngids = 0; + + assert(unit); + assert(command); + assert(context); + assert(params); + assert(exit_status); + + rename_process_from_path(command->path); + + /* We reset exactly these signals, since they are the + * only ones we set to SIG_IGN in the main daemon. All + * others we leave untouched because we set them to + * SIG_DFL or a valid handler initially, both of which + * will be demoted to SIG_DFL. */ + (void) default_signals(SIGNALS_CRASH_HANDLER, + SIGNALS_IGNORE, -1); + + if (context->ignore_sigpipe) + (void) ignore_signals(SIGPIPE, -1); + + r = reset_signal_mask(); + if (r < 0) { + *exit_status = EXIT_SIGNAL_MASK; + return r; + } + + if (params->idle_pipe) + do_idle_pipe_dance(params->idle_pipe); + + /* Close sockets very early to make sure we don't + * block init reexecution because it cannot bind its + * sockets */ + + log_forget_fds(); + + r = close_remaining_fds(params, runtime, dcreds, user_lookup_fd, socket_fd, fds, n_fds); + if (r < 0) { + *exit_status = EXIT_FDS; + return r; + } + + if (!context->same_pgrp) + if (setsid() < 0) { + *exit_status = EXIT_SETSID; + return -errno; + } + + exec_context_tty_reset(context, params); + + if (params->flags & EXEC_CONFIRM_SPAWN) { + char response; + + r = ask_for_confirmation(&response, argv); + if (r == -ETIMEDOUT) + write_confirm_message("Confirmation question timed out, assuming positive response.\n"); + else if (r < 0) + write_confirm_message("Couldn't ask confirmation question, assuming positive response: %s\n", strerror(-r)); + else if (response == 's') { + write_confirm_message("Skipping execution.\n"); + *exit_status = EXIT_CONFIRM; + return -ECANCELED; + } else if (response == 'n') { + write_confirm_message("Failing execution.\n"); + *exit_status = 0; + return 0; + } + } + + if (context->dynamic_user && dcreds) { + + /* Make sure we bypass our own NSS module for any NSS checks */ + if (putenv((char*) "SYSTEMD_NSS_DYNAMIC_BYPASS=1") != 0) { + *exit_status = EXIT_USER; + return -errno; + } + + r = dynamic_creds_realize(dcreds, &uid, &gid); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + + if (!uid_is_valid(uid) || !gid_is_valid(gid)) { + *exit_status = EXIT_USER; + return -ESRCH; + } + + if (dcreds->user) + username = dcreds->user->name; + + } else { + r = get_fixed_user(context, &username, &uid, &gid, &home, &shell); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + + r = get_fixed_group(context, &groupname, &gid); + if (r < 0) { + *exit_status = EXIT_GROUP; + return r; + } + } + + /* Initialize user supplementary groups and get SupplementaryGroups= ones */ + r = get_supplementary_groups(context, username, groupname, gid, + &supplementary_gids, &ngids); + if (r < 0) { + *exit_status = EXIT_GROUP; + return r; + } + + r = send_user_lookup(unit, user_lookup_fd, uid, gid); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + + user_lookup_fd = safe_close(user_lookup_fd); + + /* If a socket is connected to STDIN/STDOUT/STDERR, we + * must sure to drop O_NONBLOCK */ + if (socket_fd >= 0) + (void) fd_nonblock(socket_fd, false); + + r = setup_input(context, params, socket_fd, named_iofds); + if (r < 0) { + *exit_status = EXIT_STDIN; + return r; + } + + r = setup_output(unit, context, params, STDOUT_FILENO, socket_fd, named_iofds, basename(command->path), uid, gid, &journal_stream_dev, &journal_stream_ino); + if (r < 0) { + *exit_status = EXIT_STDOUT; + return r; + } + + r = setup_output(unit, context, params, STDERR_FILENO, socket_fd, named_iofds, basename(command->path), uid, gid, &journal_stream_dev, &journal_stream_ino); + if (r < 0) { + *exit_status = EXIT_STDERR; + return r; + } + + if (params->cgroup_path) { + r = cg_attach_everywhere(params->cgroup_supported, params->cgroup_path, 0, NULL, NULL); + if (r < 0) { + *exit_status = EXIT_CGROUP; + return r; + } + } + + if (context->oom_score_adjust_set) { + char t[DECIMAL_STR_MAX(context->oom_score_adjust)]; + + /* When we can't make this change due to EPERM, then + * let's silently skip over it. User namespaces + * prohibit write access to this file, and we + * shouldn't trip up over that. */ + + sprintf(t, "%i", context->oom_score_adjust); + r = write_string_file("/proc/self/oom_score_adj", t, 0); + if (r == -EPERM || r == -EACCES) { + log_open(); + log_unit_debug_errno(unit, r, "Failed to adjust OOM setting, assuming containerized execution, ignoring: %m"); + log_close(); + } else if (r < 0) { + *exit_status = EXIT_OOM_ADJUST; + return -errno; + } + } + + if (context->nice_set) + if (setpriority(PRIO_PROCESS, 0, context->nice) < 0) { + *exit_status = EXIT_NICE; + return -errno; + } + + if (context->cpu_sched_set) { + struct sched_param param = { + .sched_priority = context->cpu_sched_priority, + }; + + r = sched_setscheduler(0, + context->cpu_sched_policy | + (context->cpu_sched_reset_on_fork ? + SCHED_RESET_ON_FORK : 0), + ¶m); + if (r < 0) { + *exit_status = EXIT_SETSCHEDULER; + return -errno; + } + } + + if (context->cpuset) + if (sched_setaffinity(0, CPU_ALLOC_SIZE(context->cpuset_ncpus), context->cpuset) < 0) { + *exit_status = EXIT_CPUAFFINITY; + return -errno; + } + + if (context->ioprio_set) + if (ioprio_set(IOPRIO_WHO_PROCESS, 0, context->ioprio) < 0) { + *exit_status = EXIT_IOPRIO; + return -errno; + } + + if (context->timer_slack_nsec != NSEC_INFINITY) + if (prctl(PR_SET_TIMERSLACK, context->timer_slack_nsec) < 0) { + *exit_status = EXIT_TIMERSLACK; + return -errno; + } + + if (context->personality != PERSONALITY_INVALID) + if (personality(context->personality) < 0) { + *exit_status = EXIT_PERSONALITY; + return -errno; + } + + if (context->utmp_id) + utmp_put_init_process(context->utmp_id, getpid(), getsid(0), context->tty_path, + context->utmp_mode == EXEC_UTMP_INIT ? INIT_PROCESS : + context->utmp_mode == EXEC_UTMP_LOGIN ? LOGIN_PROCESS : + USER_PROCESS, + username ? "root" : context->user); + + if (context->user) { + r = chown_terminal(STDIN_FILENO, uid); + if (r < 0) { + *exit_status = EXIT_STDIN; + return r; + } + } + + /* If delegation is enabled we'll pass ownership of the cgroup + * (but only in systemd's own controller hierarchy!) to the + * user of the new process. */ + if (params->cgroup_path && context->user && params->cgroup_delegate) { + r = cg_set_task_access(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, 0644, uid, gid); + if (r < 0) { + *exit_status = EXIT_CGROUP; + return r; + } + + + r = cg_set_group_access(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, 0755, uid, gid); + if (r < 0) { + *exit_status = EXIT_CGROUP; + return r; + } + } + + if (!strv_isempty(context->runtime_directory) && params->runtime_prefix) { + r = setup_runtime_directory(context, params, uid, gid); + if (r < 0) { + *exit_status = EXIT_RUNTIME_DIRECTORY; + return r; + } + } + + r = build_environment( + unit, + context, + params, + n_fds, + home, + username, + shell, + journal_stream_dev, + journal_stream_ino, + &our_env); + if (r < 0) { + *exit_status = EXIT_MEMORY; + return r; + } + + r = build_pass_environment(context, &pass_env); + if (r < 0) { + *exit_status = EXIT_MEMORY; + return r; + } + + accum_env = strv_env_merge(5, + params->environment, + our_env, + pass_env, + context->environment, + files_env, + NULL); + if (!accum_env) { + *exit_status = EXIT_MEMORY; + return -ENOMEM; + } + accum_env = strv_env_clean(accum_env); + + (void) umask(context->umask); + + if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { + if (context->pam_name && username) { + r = setup_pam(context->pam_name, username, uid, gid, context->tty_path, &accum_env, fds, n_fds); + if (r < 0) { + *exit_status = EXIT_PAM; + return r; + } + } + } + + if (context->private_network && runtime && runtime->netns_storage_socket[0] >= 0) { + r = setup_netns(runtime->netns_storage_socket); + if (r < 0) { + *exit_status = EXIT_NETWORK; + return r; + } + } + + needs_mount_namespace = exec_needs_mount_namespace(context, params, runtime); + if (needs_mount_namespace) { + r = apply_mount_namespace(unit, context, params, runtime); + if (r < 0) { + *exit_status = EXIT_NAMESPACE; + return r; + } + } + + /* Apply just after mount namespace setup */ + r = apply_working_directory(context, params, home, needs_mount_namespace); + if (r < 0) { + *exit_status = EXIT_CHROOT; + return r; + } + + /* Drop groups as early as possbile */ + if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { + r = enforce_groups(context, gid, supplementary_gids, ngids); + if (r < 0) { + *exit_status = EXIT_GROUP; + return r; + } + } + +#ifdef HAVE_SELINUX + if ((params->flags & EXEC_APPLY_PERMISSIONS) && + mac_selinux_use() && + params->selinux_context_net && + socket_fd >= 0 && + !command->privileged) { + + r = mac_selinux_get_child_mls_label(socket_fd, command->path, context->selinux_context, &mac_selinux_context_net); + if (r < 0) { + *exit_status = EXIT_SELINUX_CONTEXT; + return r; + } + } +#endif + + if ((params->flags & EXEC_APPLY_PERMISSIONS) && context->private_users) { + r = setup_private_users(uid, gid); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + } + + /* We repeat the fd closing here, to make sure that + * nothing is leaked from the PAM modules. Note that + * we are more aggressive this time since socket_fd + * and the netns fds we don't need anymore. The custom + * endpoint fd was needed to upload the policy and can + * now be closed as well. */ + r = close_all_fds(fds, n_fds); + if (r >= 0) + r = shift_fds(fds, n_fds); + if (r >= 0) + r = flags_fds(fds, n_fds, context->non_blocking); + if (r < 0) { + *exit_status = EXIT_FDS; + return r; + } + + if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { + + int secure_bits = context->secure_bits; + + for (i = 0; i < _RLIMIT_MAX; i++) { + + if (!context->rlimit[i]) + continue; + + r = setrlimit_closest(i, context->rlimit[i]); + if (r < 0) { + *exit_status = EXIT_LIMITS; + return r; + } + } + + /* Set the RTPRIO resource limit to 0, but only if nothing else was explicitly requested. */ + if (context->restrict_realtime && !context->rlimit[RLIMIT_RTPRIO]) { + if (setrlimit(RLIMIT_RTPRIO, &RLIMIT_MAKE_CONST(0)) < 0) { + *exit_status = EXIT_LIMITS; + return -errno; + } + } + + if (!cap_test_all(context->capability_bounding_set)) { + r = capability_bounding_set_drop(context->capability_bounding_set, false); + if (r < 0) { + *exit_status = EXIT_CAPABILITIES; + return r; + } + } + + /* This is done before enforce_user, but ambient set + * does not survive over setresuid() if keep_caps is not set. */ + if (context->capability_ambient_set != 0) { + r = capability_ambient_set_apply(context->capability_ambient_set, true); + if (r < 0) { + *exit_status = EXIT_CAPABILITIES; + return r; + } + } + + if (context->user) { + r = enforce_user(context, uid); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + if (context->capability_ambient_set != 0) { + + /* Fix the ambient capabilities after user change. */ + r = capability_ambient_set_apply(context->capability_ambient_set, false); + if (r < 0) { + *exit_status = EXIT_CAPABILITIES; + return r; + } + + /* If we were asked to change user and ambient capabilities + * were requested, we had to add keep-caps to the securebits + * so that we would maintain the inherited capability set + * through the setresuid(). Make sure that the bit is added + * also to the context secure_bits so that we don't try to + * drop the bit away next. */ + + secure_bits |= 1<<SECURE_KEEP_CAPS; + } + } + + /* Apply the MAC contexts late, but before seccomp syscall filtering, as those should really be last to + * influence our own codepaths as little as possible. Moreover, applying MAC contexts usually requires + * syscalls that are subject to seccomp filtering, hence should probably be applied before the syscalls + * are restricted. */ + +#ifdef HAVE_SELINUX + if (mac_selinux_use()) { + char *exec_context = mac_selinux_context_net ?: context->selinux_context; + + if (exec_context) { + r = setexeccon(exec_context); + if (r < 0) { + *exit_status = EXIT_SELINUX_CONTEXT; + return r; + } + } + } +#endif + + r = setup_smack(context, command); + if (r < 0) { + *exit_status = EXIT_SMACK_PROCESS_LABEL; + return r; + } + +#ifdef HAVE_APPARMOR + if (context->apparmor_profile && mac_apparmor_use()) { + r = aa_change_onexec(context->apparmor_profile); + if (r < 0 && !context->apparmor_profile_ignore) { + *exit_status = EXIT_APPARMOR_PROFILE; + return -errno; + } + } +#endif + + /* PR_GET_SECUREBITS is not privileged, while + * PR_SET_SECUREBITS is. So to suppress + * potential EPERMs we'll try not to call + * PR_SET_SECUREBITS unless necessary. */ + if (prctl(PR_GET_SECUREBITS) != secure_bits) + if (prctl(PR_SET_SECUREBITS, secure_bits) < 0) { + *exit_status = EXIT_SECUREBITS; + return -errno; + } + + if (context_has_no_new_privileges(context)) + if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) { + *exit_status = EXIT_NO_NEW_PRIVILEGES; + return -errno; + } + +#ifdef HAVE_SECCOMP + if (context_has_address_families(context)) { + r = apply_address_families(unit, context); + if (r < 0) { + *exit_status = EXIT_ADDRESS_FAMILIES; + return r; + } + } + + if (context->memory_deny_write_execute) { + r = apply_memory_deny_write_execute(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + if (context->restrict_realtime) { + r = apply_restrict_realtime(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + if (context->protect_kernel_tunables) { + r = apply_protect_sysctl(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + if (context->protect_kernel_modules) { + r = apply_protect_kernel_modules(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + if (context->private_devices) { + r = apply_private_devices(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + /* This really should remain the last step before the execve(), to make sure our own code is unaffected + * by the filter as little as possible. */ + if (context_has_syscall_filters(context)) { + r = apply_seccomp(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } +#endif + } + + final_argv = replace_env_argv(argv, accum_env); + if (!final_argv) { + *exit_status = EXIT_MEMORY; + return -ENOMEM; + } + + if (_unlikely_(log_get_max_level() >= LOG_DEBUG)) { + _cleanup_free_ char *line; + + line = exec_command_line(final_argv); + if (line) { + log_open(); + log_struct(LOG_DEBUG, + LOG_UNIT_ID(unit), + "EXECUTABLE=%s", command->path, + LOG_UNIT_MESSAGE(unit, "Executing: %s", line), + NULL); + log_close(); + } + } + + execve(command->path, final_argv, accum_env); + *exit_status = EXIT_EXEC; + return -errno; +} + +int exec_spawn(Unit *unit, + ExecCommand *command, + const ExecContext *context, + const ExecParameters *params, + ExecRuntime *runtime, + DynamicCreds *dcreds, + pid_t *ret) { + + _cleanup_strv_free_ char **files_env = NULL; + int *fds = NULL; unsigned n_fds = 0; + _cleanup_free_ char *line = NULL; + int socket_fd, r; + int named_iofds[3] = { -1, -1, -1 }; + char **argv; + pid_t pid; + + assert(unit); + assert(command); + assert(context); + assert(ret); + assert(params); + assert(params->fds || params->n_fds <= 0); + + if (context->std_input == EXEC_INPUT_SOCKET || + context->std_output == EXEC_OUTPUT_SOCKET || + context->std_error == EXEC_OUTPUT_SOCKET) { + + if (params->n_fds != 1) { + log_unit_error(unit, "Got more than one socket."); + return -EINVAL; + } + + socket_fd = params->fds[0]; + } else { + socket_fd = -1; + fds = params->fds; + n_fds = params->n_fds; + } + + r = exec_context_named_iofds(unit, context, params, named_iofds); + if (r < 0) + return log_unit_error_errno(unit, r, "Failed to load a named file descriptor: %m"); + + r = exec_context_load_environment(unit, context, &files_env); + if (r < 0) + return log_unit_error_errno(unit, r, "Failed to load environment files: %m"); + + argv = params->argv ?: command->argv; + line = exec_command_line(argv); + if (!line) + return log_oom(); + + log_struct(LOG_DEBUG, + LOG_UNIT_ID(unit), + LOG_UNIT_MESSAGE(unit, "About to execute: %s", line), + "EXECUTABLE=%s", command->path, + NULL); + pid = fork(); + if (pid < 0) + return log_unit_error_errno(unit, errno, "Failed to fork: %m"); + + if (pid == 0) { + int exit_status; + + r = exec_child(unit, + command, + context, + params, + runtime, + dcreds, + argv, + socket_fd, + named_iofds, + fds, n_fds, + files_env, + unit->manager->user_lookup_fds[1], + &exit_status); + if (r < 0) { + log_open(); + log_struct_errno(LOG_ERR, r, + LOG_MESSAGE_ID(SD_MESSAGE_SPAWN_FAILED), + LOG_UNIT_ID(unit), + LOG_UNIT_MESSAGE(unit, "Failed at step %s spawning %s: %m", + exit_status_to_string(exit_status, EXIT_STATUS_SYSTEMD), + command->path), + "EXECUTABLE=%s", command->path, + NULL); + } + + _exit(exit_status); + } + + log_unit_debug(unit, "Forked %s as "PID_FMT, command->path, pid); + + /* We add the new process to the cgroup both in the child (so + * that we can be sure that no user code is ever executed + * outside of the cgroup) and in the parent (so that we can be + * sure that when we kill the cgroup the process will be + * killed too). */ + if (params->cgroup_path) + (void) cg_attach(SYSTEMD_CGROUP_CONTROLLER, params->cgroup_path, pid); + + exec_status_start(&command->exec_status, pid); + + *ret = pid; + return 0; +} + +void exec_context_init(ExecContext *c) { + assert(c); + + c->umask = 0022; + c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 0); + c->cpu_sched_policy = SCHED_OTHER; + c->syslog_priority = LOG_DAEMON|LOG_INFO; + c->syslog_level_prefix = true; + c->ignore_sigpipe = true; + c->timer_slack_nsec = NSEC_INFINITY; + c->personality = PERSONALITY_INVALID; + c->runtime_directory_mode = 0755; + c->capability_bounding_set = CAP_ALL; +} + +void exec_context_done(ExecContext *c) { + unsigned l; + + assert(c); + + c->environment = strv_free(c->environment); + c->environment_files = strv_free(c->environment_files); + c->pass_environment = strv_free(c->pass_environment); + + for (l = 0; l < ELEMENTSOF(c->rlimit); l++) + c->rlimit[l] = mfree(c->rlimit[l]); + + for (l = 0; l < 3; l++) + c->stdio_fdname[l] = mfree(c->stdio_fdname[l]); + + c->working_directory = mfree(c->working_directory); + c->root_directory = mfree(c->root_directory); + c->tty_path = mfree(c->tty_path); + c->syslog_identifier = mfree(c->syslog_identifier); + c->user = mfree(c->user); + c->group = mfree(c->group); + + c->supplementary_groups = strv_free(c->supplementary_groups); + + c->pam_name = mfree(c->pam_name); + + c->read_only_paths = strv_free(c->read_only_paths); + c->read_write_paths = strv_free(c->read_write_paths); + c->inaccessible_paths = strv_free(c->inaccessible_paths); + + if (c->cpuset) + CPU_FREE(c->cpuset); + + c->utmp_id = mfree(c->utmp_id); + c->selinux_context = mfree(c->selinux_context); + c->apparmor_profile = mfree(c->apparmor_profile); + + c->syscall_filter = set_free(c->syscall_filter); + c->syscall_archs = set_free(c->syscall_archs); + c->address_families = set_free(c->address_families); + + c->runtime_directory = strv_free(c->runtime_directory); +} + +int exec_context_destroy_runtime_directory(ExecContext *c, const char *runtime_prefix) { + char **i; + + assert(c); + + if (!runtime_prefix) + return 0; + + STRV_FOREACH(i, c->runtime_directory) { + _cleanup_free_ char *p; + + p = strjoin(runtime_prefix, "/", *i, NULL); + if (!p) + return -ENOMEM; + + /* We execute this synchronously, since we need to be + * sure this is gone when we start the service + * next. */ + (void) rm_rf(p, REMOVE_ROOT); + } + + return 0; +} + +void exec_command_done(ExecCommand *c) { + assert(c); + + c->path = mfree(c->path); + + c->argv = strv_free(c->argv); +} + +void exec_command_done_array(ExecCommand *c, unsigned n) { + unsigned i; + + for (i = 0; i < n; i++) + exec_command_done(c+i); +} + +ExecCommand* exec_command_free_list(ExecCommand *c) { + ExecCommand *i; + + while ((i = c)) { + LIST_REMOVE(command, c, i); + exec_command_done(i); + free(i); + } + + return NULL; +} + +void exec_command_free_array(ExecCommand **c, unsigned n) { + unsigned i; + + for (i = 0; i < n; i++) + c[i] = exec_command_free_list(c[i]); +} + +typedef struct InvalidEnvInfo { + Unit *unit; + const char *path; +} InvalidEnvInfo; + +static void invalid_env(const char *p, void *userdata) { + InvalidEnvInfo *info = userdata; + + log_unit_error(info->unit, "Ignoring invalid environment assignment '%s': %s", p, info->path); +} + +const char* exec_context_fdname(const ExecContext *c, int fd_index) { + assert(c); + + switch (fd_index) { + case STDIN_FILENO: + if (c->std_input != EXEC_INPUT_NAMED_FD) + return NULL; + return c->stdio_fdname[STDIN_FILENO] ?: "stdin"; + case STDOUT_FILENO: + if (c->std_output != EXEC_OUTPUT_NAMED_FD) + return NULL; + return c->stdio_fdname[STDOUT_FILENO] ?: "stdout"; + case STDERR_FILENO: + if (c->std_error != EXEC_OUTPUT_NAMED_FD) + return NULL; + return c->stdio_fdname[STDERR_FILENO] ?: "stderr"; + default: + return NULL; + } +} + +int exec_context_named_iofds(Unit *unit, const ExecContext *c, const ExecParameters *p, int named_iofds[3]) { + unsigned i, targets; + const char *stdio_fdname[3]; + + assert(c); + assert(p); + + targets = (c->std_input == EXEC_INPUT_NAMED_FD) + + (c->std_output == EXEC_OUTPUT_NAMED_FD) + + (c->std_error == EXEC_OUTPUT_NAMED_FD); + + for (i = 0; i < 3; i++) + stdio_fdname[i] = exec_context_fdname(c, i); + + for (i = 0; i < p->n_fds && targets > 0; i++) + if (named_iofds[STDIN_FILENO] < 0 && c->std_input == EXEC_INPUT_NAMED_FD && stdio_fdname[STDIN_FILENO] && streq(p->fd_names[i], stdio_fdname[STDIN_FILENO])) { + named_iofds[STDIN_FILENO] = p->fds[i]; + targets--; + } else if (named_iofds[STDOUT_FILENO] < 0 && c->std_output == EXEC_OUTPUT_NAMED_FD && stdio_fdname[STDOUT_FILENO] && streq(p->fd_names[i], stdio_fdname[STDOUT_FILENO])) { + named_iofds[STDOUT_FILENO] = p->fds[i]; + targets--; + } else if (named_iofds[STDERR_FILENO] < 0 && c->std_error == EXEC_OUTPUT_NAMED_FD && stdio_fdname[STDERR_FILENO] && streq(p->fd_names[i], stdio_fdname[STDERR_FILENO])) { + named_iofds[STDERR_FILENO] = p->fds[i]; + targets--; + } + + return (targets == 0 ? 0 : -ENOENT); +} + +int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l) { + char **i, **r = NULL; + + assert(c); + assert(l); + + STRV_FOREACH(i, c->environment_files) { + char *fn; + int k; + bool ignore = false; + char **p; + _cleanup_globfree_ glob_t pglob = {}; + int count, n; + + fn = *i; + + if (fn[0] == '-') { + ignore = true; + fn++; + } + + if (!path_is_absolute(fn)) { + if (ignore) + continue; + + strv_free(r); + return -EINVAL; + } + + /* Filename supports globbing, take all matching files */ + errno = 0; + if (glob(fn, 0, NULL, &pglob) != 0) { + if (ignore) + continue; + + strv_free(r); + return errno > 0 ? -errno : -EINVAL; + } + count = pglob.gl_pathc; + if (count == 0) { + if (ignore) + continue; + + strv_free(r); + return -EINVAL; + } + for (n = 0; n < count; n++) { + k = load_env_file(NULL, pglob.gl_pathv[n], NULL, &p); + if (k < 0) { + if (ignore) + continue; + + strv_free(r); + return k; + } + /* Log invalid environment variables with filename */ + if (p) { + InvalidEnvInfo info = { + .unit = unit, + .path = pglob.gl_pathv[n] + }; + + p = strv_env_clean_with_callback(p, invalid_env, &info); + } + + if (r == NULL) + r = p; + else { + char **m; + + m = strv_env_merge(2, r, p); + strv_free(r); + strv_free(p); + if (!m) + return -ENOMEM; + + r = m; + } + } + } + + *l = r; + + return 0; +} + +static bool tty_may_match_dev_console(const char *tty) { + _cleanup_free_ char *active = NULL; + char *console; + + if (!tty) + return true; + + if (startswith(tty, "/dev/")) + tty += 5; + + /* trivial identity? */ + if (streq(tty, "console")) + return true; + + console = resolve_dev_console(&active); + /* if we could not resolve, assume it may */ + if (!console) + return true; + + /* "tty0" means the active VC, so it may be the same sometimes */ + return streq(console, tty) || (streq(console, "tty0") && tty_is_vc(tty)); +} + +bool exec_context_may_touch_console(ExecContext *ec) { + + return (ec->tty_reset || + ec->tty_vhangup || + ec->tty_vt_disallocate || + is_terminal_input(ec->std_input) || + is_terminal_output(ec->std_output) || + is_terminal_output(ec->std_error)) && + tty_may_match_dev_console(exec_context_tty_path(ec)); +} + +static void strv_fprintf(FILE *f, char **l) { + char **g; + + assert(f); + + STRV_FOREACH(g, l) + fprintf(f, " %s", *g); +} + +void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { + char **e, **d; + unsigned i; + + assert(c); + assert(f); + + prefix = strempty(prefix); + + fprintf(f, + "%sUMask: %04o\n" + "%sWorkingDirectory: %s\n" + "%sRootDirectory: %s\n" + "%sNonBlocking: %s\n" + "%sPrivateTmp: %s\n" + "%sPrivateDevices: %s\n" + "%sProtectKernelTunables: %s\n" + "%sProtectKernelModules: %s\n" + "%sProtectControlGroups: %s\n" + "%sPrivateNetwork: %s\n" + "%sPrivateUsers: %s\n" + "%sProtectHome: %s\n" + "%sProtectSystem: %s\n" + "%sIgnoreSIGPIPE: %s\n" + "%sMemoryDenyWriteExecute: %s\n" + "%sRestrictRealtime: %s\n", + prefix, c->umask, + prefix, c->working_directory ? c->working_directory : "/", + prefix, c->root_directory ? c->root_directory : "/", + prefix, yes_no(c->non_blocking), + prefix, yes_no(c->private_tmp), + prefix, yes_no(c->private_devices), + prefix, yes_no(c->protect_kernel_tunables), + prefix, yes_no(c->protect_kernel_modules), + prefix, yes_no(c->protect_control_groups), + prefix, yes_no(c->private_network), + prefix, yes_no(c->private_users), + prefix, protect_home_to_string(c->protect_home), + prefix, protect_system_to_string(c->protect_system), + prefix, yes_no(c->ignore_sigpipe), + prefix, yes_no(c->memory_deny_write_execute), + prefix, yes_no(c->restrict_realtime)); + + STRV_FOREACH(e, c->environment) + fprintf(f, "%sEnvironment: %s\n", prefix, *e); + + STRV_FOREACH(e, c->environment_files) + fprintf(f, "%sEnvironmentFile: %s\n", prefix, *e); + + STRV_FOREACH(e, c->pass_environment) + fprintf(f, "%sPassEnvironment: %s\n", prefix, *e); + + fprintf(f, "%sRuntimeDirectoryMode: %04o\n", prefix, c->runtime_directory_mode); + + STRV_FOREACH(d, c->runtime_directory) + fprintf(f, "%sRuntimeDirectory: %s\n", prefix, *d); + + if (c->nice_set) + fprintf(f, + "%sNice: %i\n", + prefix, c->nice); + + if (c->oom_score_adjust_set) + fprintf(f, + "%sOOMScoreAdjust: %i\n", + prefix, c->oom_score_adjust); + + for (i = 0; i < RLIM_NLIMITS; i++) + if (c->rlimit[i]) { + fprintf(f, "%s%s: " RLIM_FMT "\n", + prefix, rlimit_to_string(i), c->rlimit[i]->rlim_max); + fprintf(f, "%s%sSoft: " RLIM_FMT "\n", + prefix, rlimit_to_string(i), c->rlimit[i]->rlim_cur); + } + + if (c->ioprio_set) { + _cleanup_free_ char *class_str = NULL; + + ioprio_class_to_string_alloc(IOPRIO_PRIO_CLASS(c->ioprio), &class_str); + fprintf(f, + "%sIOSchedulingClass: %s\n" + "%sIOPriority: %i\n", + prefix, strna(class_str), + prefix, (int) IOPRIO_PRIO_DATA(c->ioprio)); + } + + if (c->cpu_sched_set) { + _cleanup_free_ char *policy_str = NULL; + + sched_policy_to_string_alloc(c->cpu_sched_policy, &policy_str); + fprintf(f, + "%sCPUSchedulingPolicy: %s\n" + "%sCPUSchedulingPriority: %i\n" + "%sCPUSchedulingResetOnFork: %s\n", + prefix, strna(policy_str), + prefix, c->cpu_sched_priority, + prefix, yes_no(c->cpu_sched_reset_on_fork)); + } + + if (c->cpuset) { + fprintf(f, "%sCPUAffinity:", prefix); + for (i = 0; i < c->cpuset_ncpus; i++) + if (CPU_ISSET_S(i, CPU_ALLOC_SIZE(c->cpuset_ncpus), c->cpuset)) + fprintf(f, " %u", i); + fputs("\n", f); + } + + if (c->timer_slack_nsec != NSEC_INFINITY) + fprintf(f, "%sTimerSlackNSec: "NSEC_FMT "\n", prefix, c->timer_slack_nsec); + + fprintf(f, + "%sStandardInput: %s\n" + "%sStandardOutput: %s\n" + "%sStandardError: %s\n", + prefix, exec_input_to_string(c->std_input), + prefix, exec_output_to_string(c->std_output), + prefix, exec_output_to_string(c->std_error)); + + if (c->tty_path) + fprintf(f, + "%sTTYPath: %s\n" + "%sTTYReset: %s\n" + "%sTTYVHangup: %s\n" + "%sTTYVTDisallocate: %s\n", + prefix, c->tty_path, + prefix, yes_no(c->tty_reset), + prefix, yes_no(c->tty_vhangup), + prefix, yes_no(c->tty_vt_disallocate)); + + if (c->std_output == EXEC_OUTPUT_SYSLOG || + c->std_output == EXEC_OUTPUT_KMSG || + c->std_output == EXEC_OUTPUT_JOURNAL || + c->std_output == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || + c->std_output == EXEC_OUTPUT_KMSG_AND_CONSOLE || + c->std_output == EXEC_OUTPUT_JOURNAL_AND_CONSOLE || + c->std_error == EXEC_OUTPUT_SYSLOG || + c->std_error == EXEC_OUTPUT_KMSG || + c->std_error == EXEC_OUTPUT_JOURNAL || + c->std_error == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || + c->std_error == EXEC_OUTPUT_KMSG_AND_CONSOLE || + c->std_error == EXEC_OUTPUT_JOURNAL_AND_CONSOLE) { + + _cleanup_free_ char *fac_str = NULL, *lvl_str = NULL; + + log_facility_unshifted_to_string_alloc(c->syslog_priority >> 3, &fac_str); + log_level_to_string_alloc(LOG_PRI(c->syslog_priority), &lvl_str); + + fprintf(f, + "%sSyslogFacility: %s\n" + "%sSyslogLevel: %s\n", + prefix, strna(fac_str), + prefix, strna(lvl_str)); + } + + if (c->secure_bits) + fprintf(f, "%sSecure Bits:%s%s%s%s%s%s\n", + prefix, + (c->secure_bits & 1<<SECURE_KEEP_CAPS) ? " keep-caps" : "", + (c->secure_bits & 1<<SECURE_KEEP_CAPS_LOCKED) ? " keep-caps-locked" : "", + (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP) ? " no-setuid-fixup" : "", + (c->secure_bits & 1<<SECURE_NO_SETUID_FIXUP_LOCKED) ? " no-setuid-fixup-locked" : "", + (c->secure_bits & 1<<SECURE_NOROOT) ? " noroot" : "", + (c->secure_bits & 1<<SECURE_NOROOT_LOCKED) ? "noroot-locked" : ""); + + if (c->capability_bounding_set != CAP_ALL) { + unsigned long l; + fprintf(f, "%sCapabilityBoundingSet:", prefix); + + for (l = 0; l <= cap_last_cap(); l++) + if (c->capability_bounding_set & (UINT64_C(1) << l)) + fprintf(f, " %s", strna(capability_to_name(l))); + + fputs("\n", f); + } + + if (c->capability_ambient_set != 0) { + unsigned long l; + fprintf(f, "%sAmbientCapabilities:", prefix); + + for (l = 0; l <= cap_last_cap(); l++) + if (c->capability_ambient_set & (UINT64_C(1) << l)) + fprintf(f, " %s", strna(capability_to_name(l))); + + fputs("\n", f); + } + + if (c->user) + fprintf(f, "%sUser: %s\n", prefix, c->user); + if (c->group) + fprintf(f, "%sGroup: %s\n", prefix, c->group); + + fprintf(f, "%sDynamicUser: %s\n", prefix, yes_no(c->dynamic_user)); + + if (strv_length(c->supplementary_groups) > 0) { + fprintf(f, "%sSupplementaryGroups:", prefix); + strv_fprintf(f, c->supplementary_groups); + fputs("\n", f); + } + + if (c->pam_name) + fprintf(f, "%sPAMName: %s\n", prefix, c->pam_name); + + if (strv_length(c->read_write_paths) > 0) { + fprintf(f, "%sReadWritePaths:", prefix); + strv_fprintf(f, c->read_write_paths); + fputs("\n", f); + } + + if (strv_length(c->read_only_paths) > 0) { + fprintf(f, "%sReadOnlyPaths:", prefix); + strv_fprintf(f, c->read_only_paths); + fputs("\n", f); + } + + if (strv_length(c->inaccessible_paths) > 0) { + fprintf(f, "%sInaccessiblePaths:", prefix); + strv_fprintf(f, c->inaccessible_paths); + fputs("\n", f); + } + + if (c->utmp_id) + fprintf(f, + "%sUtmpIdentifier: %s\n", + prefix, c->utmp_id); + + if (c->selinux_context) + fprintf(f, + "%sSELinuxContext: %s%s\n", + prefix, c->selinux_context_ignore ? "-" : "", c->selinux_context); + + if (c->personality != PERSONALITY_INVALID) + fprintf(f, + "%sPersonality: %s\n", + prefix, strna(personality_to_string(c->personality))); + + if (c->syscall_filter) { +#ifdef HAVE_SECCOMP + Iterator j; + void *id; + bool first = true; +#endif + + fprintf(f, + "%sSystemCallFilter: ", + prefix); + + if (!c->syscall_whitelist) + fputc('~', f); + +#ifdef HAVE_SECCOMP + SET_FOREACH(id, c->syscall_filter, j) { + _cleanup_free_ char *name = NULL; + + if (first) + first = false; + else + fputc(' ', f); + + name = seccomp_syscall_resolve_num_arch(SCMP_ARCH_NATIVE, PTR_TO_INT(id) - 1); + fputs(strna(name), f); + } +#endif + + fputc('\n', f); + } + + if (c->syscall_archs) { +#ifdef HAVE_SECCOMP + Iterator j; + void *id; +#endif + + fprintf(f, + "%sSystemCallArchitectures:", + prefix); + +#ifdef HAVE_SECCOMP + SET_FOREACH(id, c->syscall_archs, j) + fprintf(f, " %s", strna(seccomp_arch_to_string(PTR_TO_UINT32(id) - 1))); +#endif + fputc('\n', f); + } + + if (c->syscall_errno > 0) + fprintf(f, + "%sSystemCallErrorNumber: %s\n", + prefix, strna(errno_to_name(c->syscall_errno))); + + if (c->apparmor_profile) + fprintf(f, + "%sAppArmorProfile: %s%s\n", + prefix, c->apparmor_profile_ignore ? "-" : "", c->apparmor_profile); +} + +bool exec_context_maintains_privileges(ExecContext *c) { + assert(c); + + /* Returns true if the process forked off would run under + * an unchanged UID or as root. */ + + if (!c->user) + return true; + + if (streq(c->user, "root") || streq(c->user, "0")) + return true; + + return false; +} + +void exec_status_start(ExecStatus *s, pid_t pid) { + assert(s); + + zero(*s); + s->pid = pid; + dual_timestamp_get(&s->start_timestamp); +} + +void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, int status) { + assert(s); + + if (s->pid && s->pid != pid) + zero(*s); + + s->pid = pid; + dual_timestamp_get(&s->exit_timestamp); + + s->code = code; + s->status = status; + + if (context) { + if (context->utmp_id) + utmp_put_dead_process(context->utmp_id, pid, code, status); + + exec_context_tty_reset(context, NULL); + } +} + +void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix) { + char buf[FORMAT_TIMESTAMP_MAX]; + + assert(s); + assert(f); + + if (s->pid <= 0) + return; + + prefix = strempty(prefix); + + fprintf(f, + "%sPID: "PID_FMT"\n", + prefix, s->pid); + + if (dual_timestamp_is_set(&s->start_timestamp)) + fprintf(f, + "%sStart Timestamp: %s\n", + prefix, format_timestamp(buf, sizeof(buf), s->start_timestamp.realtime)); + + if (dual_timestamp_is_set(&s->exit_timestamp)) + fprintf(f, + "%sExit Timestamp: %s\n" + "%sExit Code: %s\n" + "%sExit Status: %i\n", + prefix, format_timestamp(buf, sizeof(buf), s->exit_timestamp.realtime), + prefix, sigchld_code_to_string(s->code), + prefix, s->status); +} + +char *exec_command_line(char **argv) { + size_t k; + char *n, *p, **a; + bool first = true; + + assert(argv); + + k = 1; + STRV_FOREACH(a, argv) + k += strlen(*a)+3; + + n = new(char, k); + if (!n) + return NULL; + + p = n; + STRV_FOREACH(a, argv) { + + if (!first) + *(p++) = ' '; + else + first = false; + + if (strpbrk(*a, WHITESPACE)) { + *(p++) = '\''; + p = stpcpy(p, *a); + *(p++) = '\''; + } else + p = stpcpy(p, *a); + + } + + *p = 0; + + /* FIXME: this doesn't really handle arguments that have + * spaces and ticks in them */ + + return n; +} + +void exec_command_dump(ExecCommand *c, FILE *f, const char *prefix) { + _cleanup_free_ char *cmd = NULL; + const char *prefix2; + + assert(c); + assert(f); + + prefix = strempty(prefix); + prefix2 = strjoina(prefix, "\t"); + + cmd = exec_command_line(c->argv); + fprintf(f, + "%sCommand Line: %s\n", + prefix, cmd ? cmd : strerror(ENOMEM)); + + exec_status_dump(&c->exec_status, f, prefix2); +} + +void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix) { + assert(f); + + prefix = strempty(prefix); + + LIST_FOREACH(command, c, c) + exec_command_dump(c, f, prefix); +} + +void exec_command_append_list(ExecCommand **l, ExecCommand *e) { + ExecCommand *end; + + assert(l); + assert(e); + + if (*l) { + /* It's kind of important, that we keep the order here */ + LIST_FIND_TAIL(command, *l, end); + LIST_INSERT_AFTER(command, *l, end, e); + } else + *l = e; +} + +int exec_command_set(ExecCommand *c, const char *path, ...) { + va_list ap; + char **l, *p; + + assert(c); + assert(path); + + va_start(ap, path); + l = strv_new_ap(path, ap); + va_end(ap); + + if (!l) + return -ENOMEM; + + p = strdup(path); + if (!p) { + strv_free(l); + return -ENOMEM; + } + + free(c->path); + c->path = p; + + strv_free(c->argv); + c->argv = l; + + return 0; +} + +int exec_command_append(ExecCommand *c, const char *path, ...) { + _cleanup_strv_free_ char **l = NULL; + va_list ap; + int r; + + assert(c); + assert(path); + + va_start(ap, path); + l = strv_new_ap(path, ap); + va_end(ap); + + if (!l) + return -ENOMEM; + + r = strv_extend_strv(&c->argv, l, false); + if (r < 0) + return r; + + return 0; +} + + +static int exec_runtime_allocate(ExecRuntime **rt) { + + if (*rt) + return 0; + + *rt = new0(ExecRuntime, 1); + if (!*rt) + return -ENOMEM; + + (*rt)->n_ref = 1; + (*rt)->netns_storage_socket[0] = (*rt)->netns_storage_socket[1] = -1; + + return 0; +} + +int exec_runtime_make(ExecRuntime **rt, ExecContext *c, const char *id) { + int r; + + assert(rt); + assert(c); + assert(id); + + if (*rt) + return 1; + + if (!c->private_network && !c->private_tmp) + return 0; + + r = exec_runtime_allocate(rt); + if (r < 0) + return r; + + if (c->private_network && (*rt)->netns_storage_socket[0] < 0) { + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, (*rt)->netns_storage_socket) < 0) + return -errno; + } + + if (c->private_tmp && !(*rt)->tmp_dir) { + r = setup_tmp_dirs(id, &(*rt)->tmp_dir, &(*rt)->var_tmp_dir); + if (r < 0) + return r; + } + + return 1; +} + +ExecRuntime *exec_runtime_ref(ExecRuntime *r) { + assert(r); + assert(r->n_ref > 0); + + r->n_ref++; + return r; +} + +ExecRuntime *exec_runtime_unref(ExecRuntime *r) { + + if (!r) + return NULL; + + assert(r->n_ref > 0); + + r->n_ref--; + if (r->n_ref > 0) + return NULL; + + free(r->tmp_dir); + free(r->var_tmp_dir); + safe_close_pair(r->netns_storage_socket); + return mfree(r); +} + +int exec_runtime_serialize(Unit *u, ExecRuntime *rt, FILE *f, FDSet *fds) { + assert(u); + assert(f); + assert(fds); + + if (!rt) + return 0; + + if (rt->tmp_dir) + unit_serialize_item(u, f, "tmp-dir", rt->tmp_dir); + + if (rt->var_tmp_dir) + unit_serialize_item(u, f, "var-tmp-dir", rt->var_tmp_dir); + + if (rt->netns_storage_socket[0] >= 0) { + int copy; + + copy = fdset_put_dup(fds, rt->netns_storage_socket[0]); + if (copy < 0) + return copy; + + unit_serialize_item_format(u, f, "netns-socket-0", "%i", copy); + } + + if (rt->netns_storage_socket[1] >= 0) { + int copy; + + copy = fdset_put_dup(fds, rt->netns_storage_socket[1]); + if (copy < 0) + return copy; + + unit_serialize_item_format(u, f, "netns-socket-1", "%i", copy); + } + + return 0; +} + +int exec_runtime_deserialize_item(Unit *u, ExecRuntime **rt, const char *key, const char *value, FDSet *fds) { + int r; + + assert(rt); + assert(key); + assert(value); + + if (streq(key, "tmp-dir")) { + char *copy; + + r = exec_runtime_allocate(rt); + if (r < 0) + return log_oom(); + + copy = strdup(value); + if (!copy) + return log_oom(); + + free((*rt)->tmp_dir); + (*rt)->tmp_dir = copy; + + } else if (streq(key, "var-tmp-dir")) { + char *copy; + + r = exec_runtime_allocate(rt); + if (r < 0) + return log_oom(); + + copy = strdup(value); + if (!copy) + return log_oom(); + + free((*rt)->var_tmp_dir); + (*rt)->var_tmp_dir = copy; + + } else if (streq(key, "netns-socket-0")) { + int fd; + + r = exec_runtime_allocate(rt); + if (r < 0) + return log_oom(); + + if (safe_atoi(value, &fd) < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse netns socket value: %s", value); + else { + safe_close((*rt)->netns_storage_socket[0]); + (*rt)->netns_storage_socket[0] = fdset_remove(fds, fd); + } + } else if (streq(key, "netns-socket-1")) { + int fd; + + r = exec_runtime_allocate(rt); + if (r < 0) + return log_oom(); + + if (safe_atoi(value, &fd) < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse netns socket value: %s", value); + else { + safe_close((*rt)->netns_storage_socket[1]); + (*rt)->netns_storage_socket[1] = fdset_remove(fds, fd); + } + } else + return 0; + + return 1; +} + +static void *remove_tmpdir_thread(void *p) { + _cleanup_free_ char *path = p; + + (void) rm_rf(path, REMOVE_ROOT|REMOVE_PHYSICAL); + return NULL; +} + +void exec_runtime_destroy(ExecRuntime *rt) { + int r; + + if (!rt) + return; + + /* If there are multiple users of this, let's leave the stuff around */ + if (rt->n_ref > 1) + return; + + if (rt->tmp_dir) { + log_debug("Spawning thread to nuke %s", rt->tmp_dir); + + r = asynchronous_job(remove_tmpdir_thread, rt->tmp_dir); + if (r < 0) { + log_warning_errno(r, "Failed to nuke %s: %m", rt->tmp_dir); + free(rt->tmp_dir); + } + + rt->tmp_dir = NULL; + } + + if (rt->var_tmp_dir) { + log_debug("Spawning thread to nuke %s", rt->var_tmp_dir); + + r = asynchronous_job(remove_tmpdir_thread, rt->var_tmp_dir); + if (r < 0) { + log_warning_errno(r, "Failed to nuke %s: %m", rt->var_tmp_dir); + free(rt->var_tmp_dir); + } + + rt->var_tmp_dir = NULL; + } + + safe_close_pair(rt->netns_storage_socket); +} + +static const char* const exec_input_table[_EXEC_INPUT_MAX] = { + [EXEC_INPUT_NULL] = "null", + [EXEC_INPUT_TTY] = "tty", + [EXEC_INPUT_TTY_FORCE] = "tty-force", + [EXEC_INPUT_TTY_FAIL] = "tty-fail", + [EXEC_INPUT_SOCKET] = "socket", + [EXEC_INPUT_NAMED_FD] = "fd", +}; + +DEFINE_STRING_TABLE_LOOKUP(exec_input, ExecInput); + +static const char* const exec_output_table[_EXEC_OUTPUT_MAX] = { + [EXEC_OUTPUT_INHERIT] = "inherit", + [EXEC_OUTPUT_NULL] = "null", + [EXEC_OUTPUT_TTY] = "tty", + [EXEC_OUTPUT_SYSLOG] = "syslog", + [EXEC_OUTPUT_SYSLOG_AND_CONSOLE] = "syslog+console", + [EXEC_OUTPUT_KMSG] = "kmsg", + [EXEC_OUTPUT_KMSG_AND_CONSOLE] = "kmsg+console", + [EXEC_OUTPUT_JOURNAL] = "journal", + [EXEC_OUTPUT_JOURNAL_AND_CONSOLE] = "journal+console", + [EXEC_OUTPUT_SOCKET] = "socket", + [EXEC_OUTPUT_NAMED_FD] = "fd", +}; + +DEFINE_STRING_TABLE_LOOKUP(exec_output, ExecOutput); + +static const char* const exec_utmp_mode_table[_EXEC_UTMP_MODE_MAX] = { + [EXEC_UTMP_INIT] = "init", + [EXEC_UTMP_LOGIN] = "login", + [EXEC_UTMP_USER] = "user", +}; + +DEFINE_STRING_TABLE_LOOKUP(exec_utmp_mode, ExecUtmpMode); diff --git a/src/grp-system/libcore/src/hostname-setup.c b/src/grp-system/libcore/src/hostname-setup.c new file mode 100644 index 0000000000..6170d5e030 --- /dev/null +++ b/src/grp-system/libcore/src/hostname-setup.c @@ -0,0 +1,68 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdio.h> +#include <stdlib.h> + +#include "core/hostname-setup.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/hostname-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" + +int hostname_setup(void) { + int r; + _cleanup_free_ char *b = NULL; + const char *hn; + bool enoent = false; + + r = read_hostname_config("/etc/hostname", &b); + if (r < 0) { + if (r == -ENOENT) + enoent = true; + else + log_warning_errno(r, "Failed to read configured hostname: %m"); + + hn = NULL; + } else + hn = b; + + if (isempty(hn)) { + /* Don't override the hostname if it is already set + * and not explicitly configured */ + if (hostname_is_set()) + return 0; + + if (enoent) + log_info("No hostname configured."); + + hn = "localhost"; + } + + r = sethostname_idempotent(hn); + if (r < 0) + return log_warning_errno(r, "Failed to set hostname to <%s>: %m", hn); + + log_info("Set hostname to <%s>.", hn); + return 0; +} diff --git a/src/grp-system/libcore/src/ima-setup.c b/src/grp-system/libcore/src/ima-setup.c new file mode 100644 index 0000000000..d1ec4852c9 --- /dev/null +++ b/src/grp-system/libcore/src/ima-setup.c @@ -0,0 +1,80 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright (C) 2012 Roberto Sassu - Politecnico di Torino, Italy + TORSEC group — http://security.polito.it + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <unistd.h> + +#include "core/ima-setup.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/log.h" +#include "systemd-basic/util.h" + +#define IMA_SECFS_DIR "/sys/kernel/security/ima" +#define IMA_SECFS_POLICY IMA_SECFS_DIR "/policy" +#define IMA_POLICY_PATH "/etc/ima/ima-policy" + +int ima_setup(void) { +#ifdef HAVE_IMA + _cleanup_fclose_ FILE *input = NULL; + _cleanup_close_ int imafd = -1; + unsigned lineno = 0; + char line[page_size()]; + + if (access(IMA_SECFS_DIR, F_OK) < 0) { + log_debug("IMA support is disabled in the kernel, ignoring."); + return 0; + } + + input = fopen(IMA_POLICY_PATH, "re"); + if (!input) { + log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno, + "Failed to open the IMA custom policy file "IMA_POLICY_PATH", ignoring: %m"); + return 0; + } + + if (access(IMA_SECFS_POLICY, F_OK) < 0) { + log_warning("Another IMA custom policy has already been loaded, ignoring."); + return 0; + } + + imafd = open(IMA_SECFS_POLICY, O_WRONLY|O_CLOEXEC); + if (imafd < 0) { + log_error_errno(errno, "Failed to open the IMA kernel interface "IMA_SECFS_POLICY", ignoring: %m"); + return 0; + } + + FOREACH_LINE(line, input, + return log_error_errno(errno, "Failed to read the IMA custom policy file "IMA_POLICY_PATH": %m")) { + size_t len; + + len = strlen(line); + lineno++; + + if (len > 0 && write(imafd, line, len) < 0) + return log_error_errno(errno, "Failed to load the IMA custom policy file "IMA_POLICY_PATH"%u: %m", + lineno); + } + + log_info("Successfully loaded the IMA custom policy "IMA_POLICY_PATH"."); +#endif /* HAVE_IMA */ + return 0; +} diff --git a/src/grp-system/libcore/src/job.c b/src/grp-system/libcore/src/job.c new file mode 100644 index 0000000000..69f5ae5cfc --- /dev/null +++ b/src/grp-system/libcore/src/job.c @@ -0,0 +1,1264 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include <systemd/sd-id128.h> +#include <systemd/sd-messages.h> + +#include "core/job.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/async.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/set.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/virt.h" + +#include "dbus-job.h" +#include "dbus.h" + +Job* job_new_raw(Unit *unit) { + Job *j; + + /* used for deserialization */ + + assert(unit); + + j = new0(Job, 1); + if (!j) + return NULL; + + j->manager = unit->manager; + j->unit = unit; + j->type = _JOB_TYPE_INVALID; + + return j; +} + +Job* job_new(Unit *unit, JobType type) { + Job *j; + + assert(type < _JOB_TYPE_MAX); + + j = job_new_raw(unit); + if (!j) + return NULL; + + j->id = j->manager->current_job_id++; + j->type = type; + + /* We don't link it here, that's what job_dependency() is for */ + + return j; +} + +void job_free(Job *j) { + assert(j); + assert(!j->installed); + assert(!j->transaction_prev); + assert(!j->transaction_next); + assert(!j->subject_list); + assert(!j->object_list); + + if (j->in_run_queue) + LIST_REMOVE(run_queue, j->manager->run_queue, j); + + if (j->in_dbus_queue) + LIST_REMOVE(dbus_queue, j->manager->dbus_job_queue, j); + + sd_event_source_unref(j->timer_event_source); + + sd_bus_track_unref(j->clients); + strv_free(j->deserialized_clients); + + free(j); +} + +static void job_set_state(Job *j, JobState state) { + assert(j); + assert(state >= 0); + assert(state < _JOB_STATE_MAX); + + if (j->state == state) + return; + + j->state = state; + + if (!j->installed) + return; + + if (j->state == JOB_RUNNING) + j->unit->manager->n_running_jobs++; + else { + assert(j->state == JOB_WAITING); + assert(j->unit->manager->n_running_jobs > 0); + + j->unit->manager->n_running_jobs--; + + if (j->unit->manager->n_running_jobs <= 0) + j->unit->manager->jobs_in_progress_event_source = sd_event_source_unref(j->unit->manager->jobs_in_progress_event_source); + } +} + +void job_uninstall(Job *j) { + Job **pj; + + assert(j->installed); + + job_set_state(j, JOB_WAITING); + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + assert(*pj == j); + + /* Detach from next 'bigger' objects */ + + /* daemon-reload should be transparent to job observers */ + if (!MANAGER_IS_RELOADING(j->manager)) + bus_job_send_removed_signal(j); + + *pj = NULL; + + unit_add_to_gc_queue(j->unit); + + hashmap_remove(j->manager->jobs, UINT32_TO_PTR(j->id)); + j->installed = false; +} + +static bool job_type_allows_late_merge(JobType t) { + /* Tells whether it is OK to merge a job of type 't' with an already + * running job. + * Reloads cannot be merged this way. Think of the sequence: + * 1. Reload of a daemon is in progress; the daemon has already loaded + * its config file, but hasn't completed the reload operation yet. + * 2. Edit foo's config file. + * 3. Trigger another reload to have the daemon use the new config. + * Should the second reload job be merged into the first one, the daemon + * would not know about the new config. + * JOB_RESTART jobs on the other hand can be merged, because they get + * patched into JOB_START after stopping the unit. So if we see a + * JOB_RESTART running, it means the unit hasn't stopped yet and at + * this time the merge is still allowed. */ + return t != JOB_RELOAD; +} + +static void job_merge_into_installed(Job *j, Job *other) { + assert(j->installed); + assert(j->unit == other->unit); + + if (j->type != JOB_NOP) + job_type_merge_and_collapse(&j->type, other->type, j->unit); + else + assert(other->type == JOB_NOP); + + j->irreversible = j->irreversible || other->irreversible; + j->ignore_order = j->ignore_order || other->ignore_order; +} + +Job* job_install(Job *j) { + Job **pj; + Job *uj; + + assert(!j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + assert(j->state == JOB_WAITING); + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + uj = *pj; + + if (uj) { + if (job_type_is_conflicting(uj->type, j->type)) + job_finish_and_invalidate(uj, JOB_CANCELED, false, false); + else { + /* not conflicting, i.e. mergeable */ + + if (uj->state == JOB_WAITING || + (job_type_allows_late_merge(j->type) && job_type_is_superset(uj->type, j->type))) { + job_merge_into_installed(uj, j); + log_unit_debug(uj->unit, + "Merged into installed job %s/%s as %u", + uj->unit->id, job_type_to_string(uj->type), (unsigned) uj->id); + return uj; + } else { + /* already running and not safe to merge into */ + /* Patch uj to become a merged job and re-run it. */ + /* XXX It should be safer to queue j to run after uj finishes, but it is + * not currently possible to have more than one installed job per unit. */ + job_merge_into_installed(uj, j); + log_unit_debug(uj->unit, + "Merged into running job, re-running: %s/%s as %u", + uj->unit->id, job_type_to_string(uj->type), (unsigned) uj->id); + + job_set_state(uj, JOB_WAITING); + return uj; + } + } + } + + /* Install the job */ + *pj = j; + j->installed = true; + + j->manager->n_installed_jobs++; + log_unit_debug(j->unit, + "Installed new job %s/%s as %u", + j->unit->id, job_type_to_string(j->type), (unsigned) j->id); + return j; +} + +int job_install_deserialized(Job *j) { + Job **pj; + + assert(!j->installed); + + if (j->type < 0 || j->type >= _JOB_TYPE_MAX_IN_TRANSACTION) { + log_debug("Invalid job type %s in deserialization.", strna(job_type_to_string(j->type))); + return -EINVAL; + } + + pj = (j->type == JOB_NOP) ? &j->unit->nop_job : &j->unit->job; + if (*pj) { + log_unit_debug(j->unit, "Unit already has a job installed. Not installing deserialized job."); + return -EEXIST; + } + + *pj = j; + j->installed = true; + + if (j->state == JOB_RUNNING) + j->unit->manager->n_running_jobs++; + + log_unit_debug(j->unit, + "Reinstalled deserialized job %s/%s as %u", + j->unit->id, job_type_to_string(j->type), (unsigned) j->id); + return 0; +} + +JobDependency* job_dependency_new(Job *subject, Job *object, bool matters, bool conflicts) { + JobDependency *l; + + assert(object); + + /* Adds a new job link, which encodes that the 'subject' job + * needs the 'object' job in some way. If 'subject' is NULL + * this means the 'anchor' job (i.e. the one the user + * explicitly asked for) is the requester. */ + + if (!(l = new0(JobDependency, 1))) + return NULL; + + l->subject = subject; + l->object = object; + l->matters = matters; + l->conflicts = conflicts; + + if (subject) + LIST_PREPEND(subject, subject->subject_list, l); + + LIST_PREPEND(object, object->object_list, l); + + return l; +} + +void job_dependency_free(JobDependency *l) { + assert(l); + + if (l->subject) + LIST_REMOVE(subject, l->subject->subject_list, l); + + LIST_REMOVE(object, l->object->object_list, l); + + free(l); +} + +void job_dump(Job *j, FILE*f, const char *prefix) { + assert(j); + assert(f); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%s-> Job %u:\n" + "%s\tAction: %s -> %s\n" + "%s\tState: %s\n" + "%s\tIrreversible: %s\n", + prefix, j->id, + prefix, j->unit->id, job_type_to_string(j->type), + prefix, job_state_to_string(j->state), + prefix, yes_no(j->irreversible)); +} + +/* + * Merging is commutative, so imagine the matrix as symmetric. We store only + * its lower triangle to avoid duplication. We don't store the main diagonal, + * because A merged with A is simply A. + * + * If the resulting type is collapsed immediately afterwards (to get rid of + * the JOB_RELOAD_OR_START, which lies outside the lookup function's domain), + * the following properties hold: + * + * Merging is associative! A merged with B, and then merged with C is the same + * as A merged with the result of B merged with C. + * + * Mergeability is transitive! If A can be merged with B and B with C then + * A also with C. + * + * Also, if A merged with B cannot be merged with C, then either A or B cannot + * be merged with C either. + */ +static const JobType job_merging_table[] = { +/* What \ With * JOB_START JOB_VERIFY_ACTIVE JOB_STOP JOB_RELOAD */ +/*********************************************************************************/ +/*JOB_START */ +/*JOB_VERIFY_ACTIVE */ JOB_START, +/*JOB_STOP */ -1, -1, +/*JOB_RELOAD */ JOB_RELOAD_OR_START, JOB_RELOAD, -1, +/*JOB_RESTART */ JOB_RESTART, JOB_RESTART, -1, JOB_RESTART, +}; + +JobType job_type_lookup_merge(JobType a, JobType b) { + assert_cc(ELEMENTSOF(job_merging_table) == _JOB_TYPE_MAX_MERGING * (_JOB_TYPE_MAX_MERGING - 1) / 2); + assert(a >= 0 && a < _JOB_TYPE_MAX_MERGING); + assert(b >= 0 && b < _JOB_TYPE_MAX_MERGING); + + if (a == b) + return a; + + if (a < b) { + JobType tmp = a; + a = b; + b = tmp; + } + + return job_merging_table[(a - 1) * a / 2 + b]; +} + +bool job_type_is_redundant(JobType a, UnitActiveState b) { + switch (a) { + + case JOB_START: + return + b == UNIT_ACTIVE || + b == UNIT_RELOADING; + + case JOB_STOP: + return + b == UNIT_INACTIVE || + b == UNIT_FAILED; + + case JOB_VERIFY_ACTIVE: + return + b == UNIT_ACTIVE || + b == UNIT_RELOADING; + + case JOB_RELOAD: + return + b == UNIT_RELOADING; + + case JOB_RESTART: + return + b == UNIT_ACTIVATING; + + case JOB_NOP: + return true; + + default: + assert_not_reached("Invalid job type"); + } +} + +JobType job_type_collapse(JobType t, Unit *u) { + UnitActiveState s; + + switch (t) { + + case JOB_TRY_RESTART: + s = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s)) + return JOB_NOP; + + return JOB_RESTART; + + case JOB_TRY_RELOAD: + s = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s)) + return JOB_NOP; + + return JOB_RELOAD; + + case JOB_RELOAD_OR_START: + s = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s)) + return JOB_START; + + return JOB_RELOAD; + + default: + return t; + } +} + +int job_type_merge_and_collapse(JobType *a, JobType b, Unit *u) { + JobType t; + + t = job_type_lookup_merge(*a, b); + if (t < 0) + return -EEXIST; + + *a = job_type_collapse(t, u); + return 0; +} + +static bool job_is_runnable(Job *j) { + Iterator i; + Unit *other; + + assert(j); + assert(j->installed); + + /* Checks whether there is any job running for the units this + * job needs to be running after (in the case of a 'positive' + * job type) or before (in the case of a 'negative' job + * type. */ + + /* Note that unit types have a say in what is runnable, + * too. For example, if they return -EAGAIN from + * unit_start() they can indicate they are not + * runnable yet. */ + + /* First check if there is an override */ + if (j->ignore_order) + return true; + + if (j->type == JOB_NOP) + return true; + + if (j->type == JOB_START || + j->type == JOB_VERIFY_ACTIVE || + j->type == JOB_RELOAD) { + + /* Immediate result is that the job is or might be + * started. In this case let's wait for the + * dependencies, regardless whether they are + * starting or stopping something. */ + + SET_FOREACH(other, j->unit->dependencies[UNIT_AFTER], i) + if (other->job) + return false; + } + + /* Also, if something else is being stopped and we should + * change state after it, then let's wait. */ + + SET_FOREACH(other, j->unit->dependencies[UNIT_BEFORE], i) + if (other->job && + (other->job->type == JOB_STOP || + other->job->type == JOB_RESTART)) + return false; + + /* This means that for a service a and a service b where b + * shall be started after a: + * + * start a + start b → 1st step start a, 2nd step start b + * start a + stop b → 1st step stop b, 2nd step start a + * stop a + start b → 1st step stop a, 2nd step start b + * stop a + stop b → 1st step stop b, 2nd step stop a + * + * This has the side effect that restarts are properly + * synchronized too. */ + + return true; +} + +static void job_change_type(Job *j, JobType newtype) { + assert(j); + + log_unit_debug(j->unit, + "Converting job %s/%s -> %s/%s", + j->unit->id, job_type_to_string(j->type), + j->unit->id, job_type_to_string(newtype)); + + j->type = newtype; +} + +static int job_perform_on_unit(Job **j) { + uint32_t id; + Manager *m; + JobType t; + Unit *u; + int r; + + /* While we execute this operation the job might go away (for + * example: because it finishes immediately or is replaced by + * a new, conflicting job.) To make sure we don't access a + * freed job later on we store the id here, so that we can + * verify the job is still valid. */ + + assert(j); + assert(*j); + + m = (*j)->manager; + u = (*j)->unit; + t = (*j)->type; + id = (*j)->id; + + switch (t) { + case JOB_START: + r = unit_start(u); + break; + + case JOB_RESTART: + t = JOB_STOP; + /* fall through */ + case JOB_STOP: + r = unit_stop(u); + break; + + case JOB_RELOAD: + r = unit_reload(u); + break; + + default: + assert_not_reached("Invalid job type"); + } + + /* Log if the job still exists and the start/stop/reload function + * actually did something. */ + *j = manager_get_job(m, id); + if (*j && r > 0) + unit_status_emit_starting_stopping_reloading(u, t); + + return r; +} + +int job_run_and_invalidate(Job *j) { + int r; + + assert(j); + assert(j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + assert(j->in_run_queue); + + LIST_REMOVE(run_queue, j->manager->run_queue, j); + j->in_run_queue = false; + + if (j->state != JOB_WAITING) + return 0; + + if (!job_is_runnable(j)) + return -EAGAIN; + + job_set_state(j, JOB_RUNNING); + job_add_to_dbus_queue(j); + + + switch (j->type) { + + case JOB_VERIFY_ACTIVE: { + UnitActiveState t = unit_active_state(j->unit); + if (UNIT_IS_ACTIVE_OR_RELOADING(t)) + r = -EALREADY; + else if (t == UNIT_ACTIVATING) + r = -EAGAIN; + else + r = -EBADR; + break; + } + + case JOB_START: + case JOB_STOP: + case JOB_RESTART: + r = job_perform_on_unit(&j); + + /* If the unit type does not support starting/stopping, + * then simply wait. */ + if (r == -EBADR) + r = 0; + break; + + case JOB_RELOAD: + r = job_perform_on_unit(&j); + break; + + case JOB_NOP: + r = -EALREADY; + break; + + default: + assert_not_reached("Unknown job type"); + } + + if (j) { + if (r == -EALREADY) + r = job_finish_and_invalidate(j, JOB_DONE, true, true); + else if (r == -EBADR) + r = job_finish_and_invalidate(j, JOB_SKIPPED, true, false); + else if (r == -ENOEXEC) + r = job_finish_and_invalidate(j, JOB_INVALID, true, false); + else if (r == -EPROTO) + r = job_finish_and_invalidate(j, JOB_ASSERT, true, false); + else if (r == -EOPNOTSUPP) + r = job_finish_and_invalidate(j, JOB_UNSUPPORTED, true, false); + else if (r == -EAGAIN) + job_set_state(j, JOB_WAITING); + else if (r < 0) + r = job_finish_and_invalidate(j, JOB_FAILED, true, false); + } + + return r; +} + +_pure_ static const char *job_get_status_message_format(Unit *u, JobType t, JobResult result) { + + static const char *const generic_finished_start_job[_JOB_RESULT_MAX] = { + [JOB_DONE] = "Started %s.", + [JOB_TIMEOUT] = "Timed out starting %s.", + [JOB_FAILED] = "Failed to start %s.", + [JOB_DEPENDENCY] = "Dependency failed for %s.", + [JOB_ASSERT] = "Assertion failed for %s.", + [JOB_UNSUPPORTED] = "Starting of %s not supported.", + }; + static const char *const generic_finished_stop_job[_JOB_RESULT_MAX] = { + [JOB_DONE] = "Stopped %s.", + [JOB_FAILED] = "Stopped (with error) %s.", + [JOB_TIMEOUT] = "Timed out stopping %s.", + }; + static const char *const generic_finished_reload_job[_JOB_RESULT_MAX] = { + [JOB_DONE] = "Reloaded %s.", + [JOB_FAILED] = "Reload failed for %s.", + [JOB_TIMEOUT] = "Timed out reloading %s.", + }; + /* When verify-active detects the unit is inactive, report it. + * Most likely a DEPEND warning from a requisiting unit will + * occur next and it's nice to see what was requisited. */ + static const char *const generic_finished_verify_active_job[_JOB_RESULT_MAX] = { + [JOB_SKIPPED] = "%s is not active.", + }; + + const UnitStatusMessageFormats *format_table; + const char *format; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + if (IN_SET(t, JOB_START, JOB_STOP, JOB_RESTART)) { + format_table = &UNIT_VTABLE(u)->status_message_formats; + if (format_table) { + format = t == JOB_START ? format_table->finished_start_job[result] : + format_table->finished_stop_job[result]; + if (format) + return format; + } + } + + /* Return generic strings */ + if (t == JOB_START) + return generic_finished_start_job[result]; + else if (t == JOB_STOP || t == JOB_RESTART) + return generic_finished_stop_job[result]; + else if (t == JOB_RELOAD) + return generic_finished_reload_job[result]; + else if (t == JOB_VERIFY_ACTIVE) + return generic_finished_verify_active_job[result]; + + return NULL; +} + +static void job_print_status_message(Unit *u, JobType t, JobResult result) { + static const struct { + const char *color, *word; + } const statuses[_JOB_RESULT_MAX] = { + [JOB_DONE] = { ANSI_GREEN, " OK " }, + [JOB_TIMEOUT] = { ANSI_HIGHLIGHT_RED, " TIME " }, + [JOB_FAILED] = { ANSI_HIGHLIGHT_RED, "FAILED" }, + [JOB_DEPENDENCY] = { ANSI_HIGHLIGHT_YELLOW, "DEPEND" }, + [JOB_SKIPPED] = { ANSI_HIGHLIGHT, " INFO " }, + [JOB_ASSERT] = { ANSI_HIGHLIGHT_YELLOW, "ASSERT" }, + [JOB_UNSUPPORTED] = { ANSI_HIGHLIGHT_YELLOW, "UNSUPP" }, + }; + + const char *format; + const char *status; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + /* Reload status messages have traditionally not been printed to console. */ + if (t == JOB_RELOAD) + return; + + format = job_get_status_message_format(u, t, result); + if (!format) + return; + + if (log_get_show_color()) + status = strjoina(statuses[result].color, statuses[result].word, ANSI_NORMAL); + else + status = statuses[result].word; + + if (result != JOB_DONE) + manager_flip_auto_status(u->manager, true); + + DISABLE_WARNING_FORMAT_NONLITERAL; + unit_status_printf(u, status, format); + REENABLE_WARNING; + + if (t == JOB_START && result == JOB_FAILED) { + _cleanup_free_ char *quoted; + + quoted = shell_maybe_quote(u->id); + manager_status_printf(u->manager, STATUS_TYPE_NORMAL, NULL, "See 'systemctl status %s' for details.", strna(quoted)); + } +} + +static void job_log_status_message(Unit *u, JobType t, JobResult result) { + const char *format; + char buf[LINE_MAX]; + sd_id128_t mid; + static const int job_result_log_level[_JOB_RESULT_MAX] = { + [JOB_DONE] = LOG_INFO, + [JOB_CANCELED] = LOG_INFO, + [JOB_TIMEOUT] = LOG_ERR, + [JOB_FAILED] = LOG_ERR, + [JOB_DEPENDENCY] = LOG_WARNING, + [JOB_SKIPPED] = LOG_NOTICE, + [JOB_INVALID] = LOG_INFO, + [JOB_ASSERT] = LOG_WARNING, + [JOB_UNSUPPORTED] = LOG_WARNING, + }; + + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + /* Skip this if it goes to the console. since we already print + * to the console anyway... */ + + if (log_on_console()) + return; + + format = job_get_status_message_format(u, t, result); + if (!format) + return; + + /* The description might be longer than the buffer, but that's OK, we'll just truncate it here */ + DISABLE_WARNING_FORMAT_NONLITERAL; + snprintf(buf, sizeof(buf), format, unit_description(u)); + REENABLE_WARNING; + + switch (t) { + + case JOB_START: + mid = result == JOB_DONE ? SD_MESSAGE_UNIT_STARTED : SD_MESSAGE_UNIT_FAILED; + break; + + case JOB_RELOAD: + mid = SD_MESSAGE_UNIT_RELOADED; + break; + + case JOB_STOP: + case JOB_RESTART: + mid = SD_MESSAGE_UNIT_STOPPED; + break; + + default: + log_struct(job_result_log_level[result], + LOG_UNIT_ID(u), + LOG_MESSAGE("%s", buf), + "RESULT=%s", job_result_to_string(result), + NULL); + return; + } + + log_struct(job_result_log_level[result], + LOG_MESSAGE_ID(mid), + LOG_UNIT_ID(u), + LOG_MESSAGE("%s", buf), + "RESULT=%s", job_result_to_string(result), + NULL); +} + +static void job_emit_status_message(Unit *u, JobType t, JobResult result) { + + /* No message if the job did not actually do anything due to failed condition. */ + if (t == JOB_START && result == JOB_DONE && !u->condition_result) + return; + + job_log_status_message(u, t, result); + job_print_status_message(u, t, result); +} + +static void job_fail_dependencies(Unit *u, UnitDependency d) { + Unit *other; + Iterator i; + + assert(u); + + SET_FOREACH(other, u->dependencies[d], i) { + Job *j = other->job; + + if (!j) + continue; + if (!IN_SET(j->type, JOB_START, JOB_VERIFY_ACTIVE)) + continue; + + job_finish_and_invalidate(j, JOB_DEPENDENCY, true, false); + } +} + +int job_finish_and_invalidate(Job *j, JobResult result, bool recursive, bool already) { + Unit *u; + Unit *other; + JobType t; + Iterator i; + + assert(j); + assert(j->installed); + assert(j->type < _JOB_TYPE_MAX_IN_TRANSACTION); + + u = j->unit; + t = j->type; + + j->result = result; + + log_unit_debug(u, "Job %s/%s finished, result=%s", u->id, job_type_to_string(t), job_result_to_string(result)); + + /* If this job did nothing to respective unit we don't log the status message */ + if (!already) + job_emit_status_message(u, t, result); + + job_add_to_dbus_queue(j); + + /* Patch restart jobs so that they become normal start jobs */ + if (result == JOB_DONE && t == JOB_RESTART) { + + job_change_type(j, JOB_START); + job_set_state(j, JOB_WAITING); + + job_add_to_run_queue(j); + + goto finish; + } + + if (result == JOB_FAILED || result == JOB_INVALID) + j->manager->n_failed_jobs++; + + job_uninstall(j); + job_free(j); + + /* Fail depending jobs on failure */ + if (result != JOB_DONE && recursive) { + if (IN_SET(t, JOB_START, JOB_VERIFY_ACTIVE)) { + job_fail_dependencies(u, UNIT_REQUIRED_BY); + job_fail_dependencies(u, UNIT_REQUISITE_OF); + job_fail_dependencies(u, UNIT_BOUND_BY); + } else if (t == JOB_STOP) + job_fail_dependencies(u, UNIT_CONFLICTED_BY); + } + + /* Trigger OnFailure dependencies that are not generated by + * the unit itself. We don't treat JOB_CANCELED as failure in + * this context. And JOB_FAILURE is already handled by the + * unit itself. */ + if (result == JOB_TIMEOUT || result == JOB_DEPENDENCY) { + log_struct(LOG_NOTICE, + "JOB_TYPE=%s", job_type_to_string(t), + "JOB_RESULT=%s", job_result_to_string(result), + LOG_UNIT_ID(u), + LOG_UNIT_MESSAGE(u, "Job %s/%s failed with result '%s'.", + u->id, + job_type_to_string(t), + job_result_to_string(result)), + NULL); + + unit_start_on_failure(u); + } + + unit_trigger_notify(u); + +finish: + /* Try to start the next jobs that can be started */ + SET_FOREACH(other, u->dependencies[UNIT_AFTER], i) + if (other->job) + job_add_to_run_queue(other->job); + SET_FOREACH(other, u->dependencies[UNIT_BEFORE], i) + if (other->job) + job_add_to_run_queue(other->job); + + manager_check_finished(u->manager); + + return 0; +} + +static int job_dispatch_timer(sd_event_source *s, uint64_t monotonic, void *userdata) { + Job *j = userdata; + Unit *u; + + assert(j); + assert(s == j->timer_event_source); + + log_unit_warning(j->unit, "Job %s/%s timed out.", j->unit->id, job_type_to_string(j->type)); + + u = j->unit; + job_finish_and_invalidate(j, JOB_TIMEOUT, true, false); + + emergency_action(u->manager, u->job_timeout_action, u->job_timeout_reboot_arg, "job timed out"); + + return 0; +} + +int job_start_timer(Job *j) { + int r; + + if (j->timer_event_source) + return 0; + + j->begin_usec = now(CLOCK_MONOTONIC); + + if (j->unit->job_timeout == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + j->manager->event, + &j->timer_event_source, + CLOCK_MONOTONIC, + usec_add(j->begin_usec, j->unit->job_timeout), 0, + job_dispatch_timer, j); + if (r < 0) + return r; + + (void) sd_event_source_set_description(j->timer_event_source, "job-start"); + + return 0; +} + +void job_add_to_run_queue(Job *j) { + assert(j); + assert(j->installed); + + if (j->in_run_queue) + return; + + if (!j->manager->run_queue) + sd_event_source_set_enabled(j->manager->run_queue_event_source, SD_EVENT_ONESHOT); + + LIST_PREPEND(run_queue, j->manager->run_queue, j); + j->in_run_queue = true; +} + +void job_add_to_dbus_queue(Job *j) { + assert(j); + assert(j->installed); + + if (j->in_dbus_queue) + return; + + /* We don't check if anybody is subscribed here, since this + * job might just have been created and not yet assigned to a + * connection/client. */ + + LIST_PREPEND(dbus_queue, j->manager->dbus_job_queue, j); + j->in_dbus_queue = true; +} + +char *job_dbus_path(Job *j) { + char *p; + + assert(j); + + if (asprintf(&p, "/org/freedesktop/systemd1/job/%"PRIu32, j->id) < 0) + return NULL; + + return p; +} + +int job_serialize(Job *j, FILE *f) { + assert(j); + assert(f); + + fprintf(f, "job-id=%u\n", j->id); + fprintf(f, "job-type=%s\n", job_type_to_string(j->type)); + fprintf(f, "job-state=%s\n", job_state_to_string(j->state)); + fprintf(f, "job-irreversible=%s\n", yes_no(j->irreversible)); + fprintf(f, "job-sent-dbus-new-signal=%s\n", yes_no(j->sent_dbus_new_signal)); + fprintf(f, "job-ignore-order=%s\n", yes_no(j->ignore_order)); + + if (j->begin_usec > 0) + fprintf(f, "job-begin="USEC_FMT"\n", j->begin_usec); + + bus_track_serialize(j->clients, f, "subscribed"); + + /* End marker */ + fputc('\n', f); + return 0; +} + +int job_deserialize(Job *j, FILE *f) { + assert(j); + assert(f); + + for (;;) { + char line[LINE_MAX], *l, *v; + size_t k; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + return 0; + return -errno; + } + + char_array_0(line); + l = strstrip(line); + + /* End marker */ + if (l[0] == 0) + return 0; + + k = strcspn(l, "="); + + if (l[k] == '=') { + l[k] = 0; + v = l+k+1; + } else + v = l+k; + + if (streq(l, "job-id")) { + + if (safe_atou32(v, &j->id) < 0) + log_debug("Failed to parse job id value %s", v); + + } else if (streq(l, "job-type")) { + JobType t; + + t = job_type_from_string(v); + if (t < 0) + log_debug("Failed to parse job type %s", v); + else if (t >= _JOB_TYPE_MAX_IN_TRANSACTION) + log_debug("Cannot deserialize job of type %s", v); + else + j->type = t; + + } else if (streq(l, "job-state")) { + JobState s; + + s = job_state_from_string(v); + if (s < 0) + log_debug("Failed to parse job state %s", v); + else + job_set_state(j, s); + + } else if (streq(l, "job-irreversible")) { + int b; + + b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job irreversible flag %s", v); + else + j->irreversible = j->irreversible || b; + + } else if (streq(l, "job-sent-dbus-new-signal")) { + int b; + + b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job sent_dbus_new_signal flag %s", v); + else + j->sent_dbus_new_signal = j->sent_dbus_new_signal || b; + + } else if (streq(l, "job-ignore-order")) { + int b; + + b = parse_boolean(v); + if (b < 0) + log_debug("Failed to parse job ignore_order flag %s", v); + else + j->ignore_order = j->ignore_order || b; + + } else if (streq(l, "job-begin")) { + unsigned long long ull; + + if (sscanf(v, "%llu", &ull) != 1) + log_debug("Failed to parse job-begin value %s", v); + else + j->begin_usec = ull; + + } else if (streq(l, "subscribed")) { + + if (strv_extend(&j->deserialized_clients, v) < 0) + log_oom(); + } + } +} + +int job_coldplug(Job *j) { + int r; + + assert(j); + + /* After deserialization is complete and the bus connection + * set up again, let's start watching our subscribers again */ + (void) bus_track_coldplug(j->manager, &j->clients, false, j->deserialized_clients); + j->deserialized_clients = strv_free(j->deserialized_clients); + + if (j->state == JOB_WAITING) + job_add_to_run_queue(j); + + if (j->begin_usec == 0 || j->unit->job_timeout == USEC_INFINITY) + return 0; + + j->timer_event_source = sd_event_source_unref(j->timer_event_source); + + r = sd_event_add_time( + j->manager->event, + &j->timer_event_source, + CLOCK_MONOTONIC, + usec_add(j->begin_usec, j->unit->job_timeout), 0, + job_dispatch_timer, j); + if (r < 0) + log_debug_errno(r, "Failed to restart timeout for job: %m"); + + (void) sd_event_source_set_description(j->timer_event_source, "job-timeout"); + + return r; +} + +void job_shutdown_magic(Job *j) { + assert(j); + + /* The shutdown target gets some special treatment here: we + * tell the kernel to begin with flushing its disk caches, to + * optimize shutdown time a bit. Ideally we wouldn't hardcode + * this magic into PID 1. However all other processes aren't + * options either since they'd exit much sooner than PID 1 and + * asynchronous sync() would cause their exit to be + * delayed. */ + + if (j->type != JOB_START) + return; + + if (!MANAGER_IS_SYSTEM(j->unit->manager)) + return; + + if (!unit_has_name(j->unit, SPECIAL_SHUTDOWN_TARGET)) + return; + + /* In case messages on console has been disabled on boot */ + j->unit->manager->no_console_output = false; + + if (detect_container() > 0) + return; + + asynchronous_sync(); +} + +int job_get_timeout(Job *j, usec_t *timeout) { + usec_t x = USEC_INFINITY, y = USEC_INFINITY; + Unit *u = j->unit; + int r; + + assert(u); + + if (j->timer_event_source) { + r = sd_event_source_get_time(j->timer_event_source, &x); + if (r < 0) + return r; + } + + if (UNIT_VTABLE(u)->get_timeout) { + r = UNIT_VTABLE(u)->get_timeout(u, &y); + if (r < 0) + return r; + } + + if (x == USEC_INFINITY && y == USEC_INFINITY) + return 0; + + *timeout = MIN(x, y); + return 1; +} + +static const char* const job_state_table[_JOB_STATE_MAX] = { + [JOB_WAITING] = "waiting", + [JOB_RUNNING] = "running" +}; + +DEFINE_STRING_TABLE_LOOKUP(job_state, JobState); + +static const char* const job_type_table[_JOB_TYPE_MAX] = { + [JOB_START] = "start", + [JOB_VERIFY_ACTIVE] = "verify-active", + [JOB_STOP] = "stop", + [JOB_RELOAD] = "reload", + [JOB_RELOAD_OR_START] = "reload-or-start", + [JOB_RESTART] = "restart", + [JOB_TRY_RESTART] = "try-restart", + [JOB_TRY_RELOAD] = "try-reload", + [JOB_NOP] = "nop", +}; + +DEFINE_STRING_TABLE_LOOKUP(job_type, JobType); + +static const char* const job_mode_table[_JOB_MODE_MAX] = { + [JOB_FAIL] = "fail", + [JOB_REPLACE] = "replace", + [JOB_REPLACE_IRREVERSIBLY] = "replace-irreversibly", + [JOB_ISOLATE] = "isolate", + [JOB_FLUSH] = "flush", + [JOB_IGNORE_DEPENDENCIES] = "ignore-dependencies", + [JOB_IGNORE_REQUIREMENTS] = "ignore-requirements", +}; + +DEFINE_STRING_TABLE_LOOKUP(job_mode, JobMode); + +static const char* const job_result_table[_JOB_RESULT_MAX] = { + [JOB_DONE] = "done", + [JOB_CANCELED] = "canceled", + [JOB_TIMEOUT] = "timeout", + [JOB_FAILED] = "failed", + [JOB_DEPENDENCY] = "dependency", + [JOB_SKIPPED] = "skipped", + [JOB_INVALID] = "invalid", + [JOB_ASSERT] = "assert", + [JOB_UNSUPPORTED] = "unsupported", +}; + +DEFINE_STRING_TABLE_LOOKUP(job_result, JobResult); + +const char* job_type_to_access_method(JobType t) { + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + if (IN_SET(t, JOB_START, JOB_RESTART, JOB_TRY_RESTART)) + return "start"; + else if (t == JOB_STOP) + return "stop"; + else + return "reload"; +} diff --git a/src/grp-system/libcore/src/kill.c b/src/grp-system/libcore/src/kill.c new file mode 100644 index 0000000000..2f6d81f4eb --- /dev/null +++ b/src/grp-system/libcore/src/kill.c @@ -0,0 +1,68 @@ +/*** + This file is part of systemd. + + Copyright 2012 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/kill.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/util.h" + +void kill_context_init(KillContext *c) { + assert(c); + + c->kill_signal = SIGTERM; + c->send_sigkill = true; + c->send_sighup = false; +} + +void kill_context_dump(KillContext *c, FILE *f, const char *prefix) { + assert(c); + + if (!prefix) + prefix = ""; + + fprintf(f, + "%sKillMode: %s\n" + "%sKillSignal: SIG%s\n" + "%sSendSIGKILL: %s\n" + "%sSendSIGHUP: %s\n", + prefix, kill_mode_to_string(c->kill_mode), + prefix, signal_to_string(c->kill_signal), + prefix, yes_no(c->send_sigkill), + prefix, yes_no(c->send_sighup)); +} + +static const char* const kill_mode_table[_KILL_MODE_MAX] = { + [KILL_CONTROL_GROUP] = "control-group", + [KILL_PROCESS] = "process", + [KILL_MIXED] = "mixed", + [KILL_NONE] = "none" +}; + +DEFINE_STRING_TABLE_LOOKUP(kill_mode, KillMode); + +static const char* const kill_who_table[_KILL_WHO_MAX] = { + [KILL_MAIN] = "main", + [KILL_CONTROL] = "control", + [KILL_ALL] = "all", + [KILL_MAIN_FAIL] = "main-fail", + [KILL_CONTROL_FAIL] = "control-fail", + [KILL_ALL_FAIL] = "all-fail" +}; + +DEFINE_STRING_TABLE_LOOKUP(kill_who, KillWho); diff --git a/src/grp-system/libcore/src/killall.c b/src/grp-system/libcore/src/killall.c new file mode 100644 index 0000000000..3e4b5e5186 --- /dev/null +++ b/src/grp-system/libcore/src/killall.c @@ -0,0 +1,248 @@ +/*** + This file is part of systemd. + + Copyright 2010 ProFUSION embedded systems + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <signal.h> +#include <sys/wait.h> +#include <unistd.h> + +#include "core/killall.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/def.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/set.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/util.h" + +static bool ignore_proc(pid_t pid, bool warn_rootfs) { + _cleanup_fclose_ FILE *f = NULL; + char c; + const char *p; + size_t count; + uid_t uid; + int r; + + /* We are PID 1, let's not commit suicide */ + if (pid == 1) + return true; + + r = get_process_uid(pid, &uid); + if (r < 0) + return true; /* not really, but better safe than sorry */ + + /* Non-root processes otherwise are always subject to be killed */ + if (uid != 0) + return false; + + p = procfs_file_alloca(pid, "cmdline"); + f = fopen(p, "re"); + if (!f) + return true; /* not really, but has the desired effect */ + + count = fread(&c, 1, 1, f); + + /* Kernel threads have an empty cmdline */ + if (count <= 0) + return true; + + /* Processes with argv[0][0] = '@' we ignore from the killing + * spree. + * + * http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */ + if (c == '@' && warn_rootfs) { + _cleanup_free_ char *comm = NULL; + + r = pid_from_same_root_fs(pid); + if (r < 0) + return true; + + get_process_comm(pid, &comm); + + if (r) + log_notice("Process " PID_FMT " (%s) has been marked to be excluded from killing. It is " + "running from the root file system, and thus likely to block re-mounting of the " + "root file system to read-only. Please consider moving it into an initrd file " + "system instead.", pid, strna(comm)); + return true; + } else if (c == '@') + return true; + + return false; +} + +static void wait_for_children(Set *pids, sigset_t *mask) { + usec_t until; + + assert(mask); + + if (set_isempty(pids)) + return; + + until = now(CLOCK_MONOTONIC) + DEFAULT_TIMEOUT_USEC; + for (;;) { + struct timespec ts; + int k; + usec_t n; + void *p; + Iterator i; + + /* First, let the kernel inform us about killed + * children. Most processes will probably be our + * children, but some are not (might be our + * grandchildren instead...). */ + for (;;) { + pid_t pid; + + pid = waitpid(-1, NULL, WNOHANG); + if (pid == 0) + break; + if (pid < 0) { + if (errno == ECHILD) + break; + + log_error_errno(errno, "waitpid() failed: %m"); + return; + } + + (void) set_remove(pids, PID_TO_PTR(pid)); + } + + /* Now explicitly check who might be remaining, who + * might not be our child. */ + SET_FOREACH(p, pids, i) { + + /* We misuse getpgid as a check whether a + * process still exists. */ + if (getpgid(PTR_TO_PID(p)) >= 0) + continue; + + if (errno != ESRCH) + continue; + + set_remove(pids, p); + } + + if (set_isempty(pids)) + return; + + n = now(CLOCK_MONOTONIC); + if (n >= until) + return; + + timespec_store(&ts, until - n); + k = sigtimedwait(mask, NULL, &ts); + if (k != SIGCHLD) { + + if (k < 0 && errno != EAGAIN) { + log_error_errno(errno, "sigtimedwait() failed: %m"); + return; + } + + if (k >= 0) + log_warning("sigtimedwait() returned unexpected signal."); + } + } +} + +static int killall(int sig, Set *pids, bool send_sighup) { + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *d; + + dir = opendir("/proc"); + if (!dir) + return -errno; + + while ((d = readdir(dir))) { + pid_t pid; + int r; + + if (d->d_type != DT_DIR && + d->d_type != DT_UNKNOWN) + continue; + + if (parse_pid(d->d_name, &pid) < 0) + continue; + + if (ignore_proc(pid, sig == SIGKILL && !in_initrd())) + continue; + + if (sig == SIGKILL) { + _cleanup_free_ char *s = NULL; + + get_process_comm(pid, &s); + log_notice("Sending SIGKILL to PID "PID_FMT" (%s).", pid, strna(s)); + } + + if (kill(pid, sig) >= 0) { + if (pids) { + r = set_put(pids, PID_TO_PTR(pid)); + if (r < 0) + log_oom(); + } + } else if (errno != ENOENT) + log_warning_errno(errno, "Could not kill %d: %m", pid); + + if (send_sighup) { + /* Optionally, also send a SIGHUP signal, but + only if the process has a controlling + tty. This is useful to allow handling of + shells which ignore SIGTERM but react to + SIGHUP. We do not send this to processes that + have no controlling TTY since we don't want to + trigger reloads of daemon processes. Also we + make sure to only send this after SIGTERM so + that SIGTERM is always first in the queue. */ + + + if (get_ctty_devnr(pid, NULL) >= 0) + kill(pid, SIGHUP); + } + } + + return set_size(pids); +} + +void broadcast_signal(int sig, bool wait_for_exit, bool send_sighup) { + sigset_t mask, oldmask; + _cleanup_set_free_ Set *pids = NULL; + + if (wait_for_exit) + pids = set_new(NULL); + + assert_se(sigemptyset(&mask) == 0); + assert_se(sigaddset(&mask, SIGCHLD) == 0); + assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0); + + if (kill(-1, SIGSTOP) < 0 && errno != ESRCH) + log_warning_errno(errno, "kill(-1, SIGSTOP) failed: %m"); + + killall(sig, pids, send_sighup); + + if (kill(-1, SIGCONT) < 0 && errno != ESRCH) + log_warning_errno(errno, "kill(-1, SIGCONT) failed: %m"); + + if (wait_for_exit) + wait_for_children(pids, &mask); + + assert_se(sigprocmask(SIG_SETMASK, &oldmask, NULL) == 0); +} diff --git a/src/grp-system/libcore/src/kmod-setup.c b/src/grp-system/libcore/src/kmod-setup.c new file mode 100644 index 0000000000..0f935380b2 --- /dev/null +++ b/src/grp-system/libcore/src/kmod-setup.c @@ -0,0 +1,128 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <string.h> +#include <unistd.h> + +#ifdef HAVE_KMOD +#include <libkmod.h> +#endif + +#include "core/kmod-setup.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/capability-util.h" +#include "systemd-basic/macro.h" + +#ifdef HAVE_KMOD +static void systemd_kmod_log( + void *data, + int priority, + const char *file, int line, + const char *fn, + const char *format, + va_list args) { + + /* library logging is enabled at debug only */ + DISABLE_WARNING_FORMAT_NONLITERAL; + log_internalv(LOG_DEBUG, 0, file, line, fn, format, args); + REENABLE_WARNING; +} +#endif + +int kmod_setup(void) { +#ifdef HAVE_KMOD + + static const struct { + const char *module; + const char *path; + bool warn_if_unavailable:1; + bool warn_if_module:1; + bool (*condition_fn)(void); + } kmod_table[] = { + /* auto-loading on use doesn't work before udev is up */ + { "autofs4", "/sys/class/misc/autofs", true, false, NULL }, + + /* early configure of ::1 on the loopback device */ + { "ipv6", "/sys/module/ipv6", false, true, NULL }, + + /* this should never be a module */ + { "unix", "/proc/net/unix", true, true, NULL }, + +#ifdef HAVE_LIBIPTC + /* netfilter is needed by networkd, nspawn among others, and cannot be autoloaded */ + { "ip_tables", "/proc/net/ip_tables_names", false, false, NULL }, +#endif + }; + struct kmod_ctx *ctx = NULL; + unsigned int i; + int r; + + if (have_effective_cap(CAP_SYS_MODULE) == 0) + return 0; + + for (i = 0; i < ELEMENTSOF(kmod_table); i++) { + struct kmod_module *mod; + + if (kmod_table[i].path && access(kmod_table[i].path, F_OK) >= 0) + continue; + + if (kmod_table[i].condition_fn && !kmod_table[i].condition_fn()) + continue; + + if (kmod_table[i].warn_if_module) + log_debug("Your kernel apparently lacks built-in %s support. Might be " + "a good idea to compile it in. We'll now try to work around " + "this by loading the module...", kmod_table[i].module); + + if (!ctx) { + ctx = kmod_new(NULL, NULL); + if (!ctx) + return log_oom(); + + kmod_set_log_fn(ctx, systemd_kmod_log, NULL); + kmod_load_resources(ctx); + } + + r = kmod_module_new_from_name(ctx, kmod_table[i].module, &mod); + if (r < 0) { + log_error("Failed to lookup module '%s'", kmod_table[i].module); + continue; + } + + r = kmod_module_probe_insert_module(mod, KMOD_PROBE_APPLY_BLACKLIST, NULL, NULL, NULL, NULL); + if (r == 0) + log_debug("Inserted module '%s'", kmod_module_get_name(mod)); + else if (r == KMOD_PROBE_APPLY_BLACKLIST) + log_info("Module '%s' is blacklisted", kmod_module_get_name(mod)); + else { + bool print_warning = kmod_table[i].warn_if_unavailable || (r < 0 && r != -ENOENT); + + log_full_errno(print_warning ? LOG_WARNING : LOG_DEBUG, r, + "Failed to insert module '%s': %m", kmod_module_get_name(mod)); + } + + kmod_module_unref(mod); + } + + if (ctx) + kmod_unref(ctx); + +#endif + return 0; +} diff --git a/src/grp-system/libcore/src/linux/auto_dev-ioctl.h b/src/grp-system/libcore/src/linux/auto_dev-ioctl.h new file mode 100644 index 0000000000..aeaeb3ea7a --- /dev/null +++ b/src/grp-system/libcore/src/linux/auto_dev-ioctl.h @@ -0,0 +1,228 @@ +/* + * Copyright 2008 Red Hat, Inc. All rights reserved. + * Copyright 2008 Ian Kent <raven@themaw.net> + * + * This file is part of the Linux kernel and is made available under + * the terms of the GNU General Public License, version 2, or at your + * option, any later version, incorporated herein by reference. + */ + +#ifndef _LINUX_AUTO_DEV_IOCTL_H +#define _LINUX_AUTO_DEV_IOCTL_H + +#include <linux/auto_fs.h> + +#ifdef __KERNEL__ +#include <linux/string.h> +#else +#include <string.h> +#endif /* __KERNEL__ */ + +#define AUTOFS_DEVICE_NAME "autofs" + +#define AUTOFS_DEV_IOCTL_VERSION_MAJOR 1 +#define AUTOFS_DEV_IOCTL_VERSION_MINOR 0 + +#define AUTOFS_DEVID_LEN 16 + +#define AUTOFS_DEV_IOCTL_SIZE sizeof(struct autofs_dev_ioctl) + +/* + * An ioctl interface for autofs mount point control. + */ + +struct args_protover { + __u32 version; +}; + +struct args_protosubver { + __u32 sub_version; +}; + +struct args_openmount { + __u32 devid; +}; + +struct args_ready { + __u32 token; +}; + +struct args_fail { + __u32 token; + __s32 status; +}; + +struct args_setpipefd { + __s32 pipefd; +}; + +struct args_timeout { + __u64 timeout; +}; + +struct args_requester { + __u32 uid; + __u32 gid; +}; + +struct args_expire { + __u32 how; +}; + +struct args_askumount { + __u32 may_umount; +}; + +struct args_ismountpoint { + union { + struct args_in { + __u32 type; + } in; + struct args_out { + __u32 devid; + __u32 magic; + } out; + }; +}; + +/* + * All the ioctls use this structure. + * When sending a path size must account for the total length + * of the chunk of memory otherwise is is the size of the + * structure. + */ + +struct autofs_dev_ioctl { + __u32 ver_major; + __u32 ver_minor; + __u32 size; /* total size of data passed in + * including this struct */ + __s32 ioctlfd; /* automount command fd */ + + /* Command parameters */ + + union { + struct args_protover protover; + struct args_protosubver protosubver; + struct args_openmount openmount; + struct args_ready ready; + struct args_fail fail; + struct args_setpipefd setpipefd; + struct args_timeout timeout; + struct args_requester requester; + struct args_expire expire; + struct args_askumount askumount; + struct args_ismountpoint ismountpoint; + }; + + char path[0]; +}; + +static inline void init_autofs_dev_ioctl(struct autofs_dev_ioctl *in) { + memset(in, 0, sizeof(struct autofs_dev_ioctl)); + in->ver_major = AUTOFS_DEV_IOCTL_VERSION_MAJOR; + in->ver_minor = AUTOFS_DEV_IOCTL_VERSION_MINOR; + in->size = sizeof(struct autofs_dev_ioctl); + in->ioctlfd = -1; + return; +} + +/* + * If you change this make sure you make the corresponding change + * to autofs-dev-ioctl.c:lookup_ioctl() + */ +enum { + /* Get various version info */ + AUTOFS_DEV_IOCTL_VERSION_CMD = 0x71, + AUTOFS_DEV_IOCTL_PROTOVER_CMD, + AUTOFS_DEV_IOCTL_PROTOSUBVER_CMD, + + /* Open mount ioctl fd */ + AUTOFS_DEV_IOCTL_OPENMOUNT_CMD, + + /* Close mount ioctl fd */ + AUTOFS_DEV_IOCTL_CLOSEMOUNT_CMD, + + /* Mount/expire status returns */ + AUTOFS_DEV_IOCTL_READY_CMD, + AUTOFS_DEV_IOCTL_FAIL_CMD, + + /* Activate/deactivate autofs mount */ + AUTOFS_DEV_IOCTL_SETPIPEFD_CMD, + AUTOFS_DEV_IOCTL_CATATONIC_CMD, + + /* Expiry timeout */ + AUTOFS_DEV_IOCTL_TIMEOUT_CMD, + + /* Get mount last requesting uid and gid */ + AUTOFS_DEV_IOCTL_REQUESTER_CMD, + + /* Check for eligible expire candidates */ + AUTOFS_DEV_IOCTL_EXPIRE_CMD, + + /* Request busy status */ + AUTOFS_DEV_IOCTL_ASKUMOUNT_CMD, + + /* Check if path is a mountpoint */ + AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD, +}; + +#define AUTOFS_IOCTL 0x93 + +#define AUTOFS_DEV_IOCTL_VERSION \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_VERSION_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_PROTOVER \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_PROTOVER_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_PROTOSUBVER \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_PROTOSUBVER_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_OPENMOUNT \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_OPENMOUNT_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_CLOSEMOUNT \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_CLOSEMOUNT_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_READY \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_READY_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_FAIL \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_FAIL_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_SETPIPEFD \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_SETPIPEFD_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_CATATONIC \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_CATATONIC_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_TIMEOUT \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_TIMEOUT_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_REQUESTER \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_REQUESTER_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_EXPIRE \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_EXPIRE_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_ASKUMOUNT \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_ASKUMOUNT_CMD, struct autofs_dev_ioctl) + +#define AUTOFS_DEV_IOCTL_ISMOUNTPOINT \ + _IOWR(AUTOFS_IOCTL, \ + AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD, struct autofs_dev_ioctl) + +#endif /* _LINUX_AUTO_DEV_IOCTL_H */ diff --git a/src/grp-system/libcore/src/load-dropin.c b/src/grp-system/libcore/src/load-dropin.c new file mode 100644 index 0000000000..4374a1fc25 --- /dev/null +++ b/src/grp-system/libcore/src/load-dropin.c @@ -0,0 +1,91 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + + +#include "core/load-fragment.h" +#include "core/unit.h" +#include "systemd-basic/log.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-shared/conf-parser.h" + +#include "load-dropin.h" + +static int add_dependency_consumer( + UnitDependency dependency, + const char *entry, + const char* filepath, + void *arg) { + Unit *u = arg; + int r; + + assert(u); + + r = unit_add_dependency_by_name(u, dependency, entry, filepath, true); + if (r < 0) + log_error_errno(r, "Cannot add dependency %s to %s, ignoring: %m", entry, u->id); + + return 0; +} + +int unit_load_dropin(Unit *u) { + _cleanup_strv_free_ char **l = NULL; + Iterator i; + char *t, **f; + int r; + + assert(u); + + /* Load dependencies from supplementary drop-in directories */ + + SET_FOREACH(t, u->names, i) { + char **p; + + STRV_FOREACH(p, u->manager->lookup_paths.search_path) { + unit_file_process_dir(u->manager->unit_path_cache, *p, t, ".wants", UNIT_WANTS, + add_dependency_consumer, u, NULL); + unit_file_process_dir(u->manager->unit_path_cache, *p, t, ".requires", UNIT_REQUIRES, + add_dependency_consumer, u, NULL); + } + } + + r = unit_find_dropin_paths(u, &l); + if (r <= 0) + return 0; + + if (!u->dropin_paths) { + u->dropin_paths = l; + l = NULL; + } else { + r = strv_extend_strv(&u->dropin_paths, l, true); + if (r < 0) + return log_oom(); + } + + STRV_FOREACH(f, u->dropin_paths) { + config_parse(u->id, *f, NULL, + UNIT_VTABLE(u)->sections, + config_item_perf_lookup, load_fragment_gperf_lookup, + false, false, false, u); + } + + u->dropin_mtime = now(CLOCK_REALTIME); + + return 0; +} diff --git a/src/grp-system/libcore/src/load-dropin.h b/src/grp-system/libcore/src/load-dropin.h new file mode 100644 index 0000000000..aa0fa024de --- /dev/null +++ b/src/grp-system/libcore/src/load-dropin.h @@ -0,0 +1,34 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/unit.h" +#include "systemd-shared/dropin.h" + +/* Read service data supplementary drop-in directories */ + +static inline int unit_find_dropin_paths(Unit *u, char ***paths) { + return unit_file_find_dropin_paths(u->manager->lookup_paths.search_path, + u->manager->unit_path_cache, + u->names, + paths); +} + +int unit_load_dropin(Unit *u); diff --git a/src/grp-system/libcore/src/load-fragment-gperf.gperf.m4 b/src/grp-system/libcore/src/load-fragment-gperf.gperf.m4 new file mode 100644 index 0000000000..11b5dd5dc2 --- /dev/null +++ b/src/grp-system/libcore/src/load-fragment-gperf.gperf.m4 @@ -0,0 +1,413 @@ +%{ +#include <stddef.h> + +#include "core/load-fragment.h" +#include "systemd-basic/missing.h" +#include "systemd-shared/conf-parser.h" +%} +struct ConfigPerfItem; +%null_strings +%language=ANSI-C +%define slot-name section_and_lvalue +%define hash-function-name load_fragment_gperf_hash +%define lookup-function-name load_fragment_gperf_lookup +%readonly-tables +%omit-struct-type +%struct-type +%includes +%% +m4_dnl Define the context options only once +m4_define(`EXEC_CONTEXT_CONFIG_ITEMS', +`$1.WorkingDirectory, config_parse_working_directory, 0, offsetof($1, exec_context) +$1.RootDirectory, config_parse_unit_path_printf, 0, offsetof($1, exec_context.root_directory) +$1.User, config_parse_user_group, 0, offsetof($1, exec_context.user) +$1.Group, config_parse_user_group, 0, offsetof($1, exec_context.group) +$1.SupplementaryGroups, config_parse_user_group_strv, 0, offsetof($1, exec_context.supplementary_groups) +$1.Nice, config_parse_exec_nice, 0, offsetof($1, exec_context) +$1.OOMScoreAdjust, config_parse_exec_oom_score_adjust, 0, offsetof($1, exec_context) +$1.IOSchedulingClass, config_parse_exec_io_class, 0, offsetof($1, exec_context) +$1.IOSchedulingPriority, config_parse_exec_io_priority, 0, offsetof($1, exec_context) +$1.CPUSchedulingPolicy, config_parse_exec_cpu_sched_policy, 0, offsetof($1, exec_context) +$1.CPUSchedulingPriority, config_parse_exec_cpu_sched_prio, 0, offsetof($1, exec_context) +$1.CPUSchedulingResetOnFork, config_parse_bool, 0, offsetof($1, exec_context.cpu_sched_reset_on_fork) +$1.CPUAffinity, config_parse_exec_cpu_affinity, 0, offsetof($1, exec_context) +$1.UMask, config_parse_mode, 0, offsetof($1, exec_context.umask) +$1.Environment, config_parse_environ, 0, offsetof($1, exec_context.environment) +$1.EnvironmentFile, config_parse_unit_env_file, 0, offsetof($1, exec_context.environment_files) +$1.PassEnvironment, config_parse_pass_environ, 0, offsetof($1, exec_context.pass_environment) +$1.DynamicUser, config_parse_bool, 0, offsetof($1, exec_context.dynamic_user) +$1.StandardInput, config_parse_exec_input, 0, offsetof($1, exec_context) +$1.StandardOutput, config_parse_exec_output, 0, offsetof($1, exec_context) +$1.StandardError, config_parse_exec_output, 0, offsetof($1, exec_context) +$1.TTYPath, config_parse_unit_path_printf, 0, offsetof($1, exec_context.tty_path) +$1.TTYReset, config_parse_bool, 0, offsetof($1, exec_context.tty_reset) +$1.TTYVHangup, config_parse_bool, 0, offsetof($1, exec_context.tty_vhangup) +$1.TTYVTDisallocate, config_parse_bool, 0, offsetof($1, exec_context.tty_vt_disallocate) +$1.SyslogIdentifier, config_parse_unit_string_printf, 0, offsetof($1, exec_context.syslog_identifier) +$1.SyslogFacility, config_parse_log_facility, 0, offsetof($1, exec_context.syslog_priority) +$1.SyslogLevel, config_parse_log_level, 0, offsetof($1, exec_context.syslog_priority) +$1.SyslogLevelPrefix, config_parse_bool, 0, offsetof($1, exec_context.syslog_level_prefix) +$1.Capabilities, config_parse_warn_compat, DISABLED_LEGACY, offsetof($1, exec_context) +$1.SecureBits, config_parse_exec_secure_bits, 0, offsetof($1, exec_context) +$1.CapabilityBoundingSet, config_parse_capability_set, 0, offsetof($1, exec_context.capability_bounding_set) +$1.AmbientCapabilities, config_parse_capability_set, 0, offsetof($1, exec_context.capability_ambient_set) +$1.TimerSlackNSec, config_parse_nsec, 0, offsetof($1, exec_context.timer_slack_nsec) +$1.NoNewPrivileges, config_parse_no_new_privileges, 0, offsetof($1, exec_context) +m4_ifdef(`HAVE_SECCOMP', +`$1.SystemCallFilter, config_parse_syscall_filter, 0, offsetof($1, exec_context) +$1.SystemCallArchitectures, config_parse_syscall_archs, 0, offsetof($1, exec_context.syscall_archs) +$1.SystemCallErrorNumber, config_parse_syscall_errno, 0, offsetof($1, exec_context) +$1.MemoryDenyWriteExecute, config_parse_bool, 0, offsetof($1, exec_context.memory_deny_write_execute) +$1.RestrictRealtime, config_parse_bool, 0, offsetof($1, exec_context.restrict_realtime) +$1.RestrictAddressFamilies, config_parse_address_families, 0, offsetof($1, exec_context)', +`$1.SystemCallFilter, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +$1.SystemCallArchitectures, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +$1.SystemCallErrorNumber, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +$1.MemoryDenyWriteExecute, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +$1.RestrictRealtime, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +$1.RestrictAddressFamilies, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +$1.LimitCPU, config_parse_limit, RLIMIT_CPU, offsetof($1, exec_context.rlimit) +$1.LimitFSIZE, config_parse_limit, RLIMIT_FSIZE, offsetof($1, exec_context.rlimit) +$1.LimitDATA, config_parse_limit, RLIMIT_DATA, offsetof($1, exec_context.rlimit) +$1.LimitSTACK, config_parse_limit, RLIMIT_STACK, offsetof($1, exec_context.rlimit) +$1.LimitCORE, config_parse_limit, RLIMIT_CORE, offsetof($1, exec_context.rlimit) +$1.LimitRSS, config_parse_limit, RLIMIT_RSS, offsetof($1, exec_context.rlimit) +$1.LimitNOFILE, config_parse_limit, RLIMIT_NOFILE, offsetof($1, exec_context.rlimit) +$1.LimitAS, config_parse_limit, RLIMIT_AS, offsetof($1, exec_context.rlimit) +$1.LimitNPROC, config_parse_limit, RLIMIT_NPROC, offsetof($1, exec_context.rlimit) +$1.LimitMEMLOCK, config_parse_limit, RLIMIT_MEMLOCK, offsetof($1, exec_context.rlimit) +$1.LimitLOCKS, config_parse_limit, RLIMIT_LOCKS, offsetof($1, exec_context.rlimit) +$1.LimitSIGPENDING, config_parse_limit, RLIMIT_SIGPENDING, offsetof($1, exec_context.rlimit) +$1.LimitMSGQUEUE, config_parse_limit, RLIMIT_MSGQUEUE, offsetof($1, exec_context.rlimit) +$1.LimitNICE, config_parse_limit, RLIMIT_NICE, offsetof($1, exec_context.rlimit) +$1.LimitRTPRIO, config_parse_limit, RLIMIT_RTPRIO, offsetof($1, exec_context.rlimit) +$1.LimitRTTIME, config_parse_limit, RLIMIT_RTTIME, offsetof($1, exec_context.rlimit) +$1.ReadWriteDirectories, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.read_write_paths) +$1.ReadOnlyDirectories, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.read_only_paths) +$1.InaccessibleDirectories, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.inaccessible_paths) +$1.ReadWritePaths, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.read_write_paths) +$1.ReadOnlyPaths, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.read_only_paths) +$1.InaccessiblePaths, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.inaccessible_paths) +$1.PrivateTmp, config_parse_bool, 0, offsetof($1, exec_context.private_tmp) +$1.PrivateDevices, config_parse_bool, 0, offsetof($1, exec_context.private_devices) +$1.ProtectKernelTunables, config_parse_bool, 0, offsetof($1, exec_context.protect_kernel_tunables) +$1.ProtectKernelModules, config_parse_bool, 0, offsetof($1, exec_context.protect_kernel_modules) +$1.ProtectControlGroups, config_parse_bool, 0, offsetof($1, exec_context.protect_control_groups) +$1.PrivateNetwork, config_parse_bool, 0, offsetof($1, exec_context.private_network) +$1.PrivateUsers, config_parse_bool, 0, offsetof($1, exec_context.private_users) +$1.ProtectSystem, config_parse_protect_system, 0, offsetof($1, exec_context) +$1.ProtectHome, config_parse_protect_home, 0, offsetof($1, exec_context) +$1.MountFlags, config_parse_exec_mount_flags, 0, offsetof($1, exec_context) +$1.Personality, config_parse_personality, 0, offsetof($1, exec_context.personality) +$1.RuntimeDirectoryMode, config_parse_mode, 0, offsetof($1, exec_context.runtime_directory_mode) +$1.RuntimeDirectory, config_parse_runtime_directory, 0, offsetof($1, exec_context.runtime_directory) +m4_ifdef(`HAVE_PAM', +`$1.PAMName, config_parse_unit_string_printf, 0, offsetof($1, exec_context.pam_name)', +`$1.PAMName, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +$1.IgnoreSIGPIPE, config_parse_bool, 0, offsetof($1, exec_context.ignore_sigpipe) +$1.UtmpIdentifier, config_parse_unit_string_printf, 0, offsetof($1, exec_context.utmp_id) +$1.UtmpMode, config_parse_exec_utmp_mode, 0, offsetof($1, exec_context.utmp_mode) +m4_ifdef(`HAVE_SELINUX', +`$1.SELinuxContext, config_parse_exec_selinux_context, 0, offsetof($1, exec_context)', +`$1.SELinuxContext, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +m4_ifdef(`HAVE_APPARMOR', +`$1.AppArmorProfile, config_parse_exec_apparmor_profile, 0, offsetof($1, exec_context)', +`$1.AppArmorProfile, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +m4_ifdef(`HAVE_SMACK', +`$1.SmackProcessLabel, config_parse_exec_smack_process_label, 0, offsetof($1, exec_context)', +`$1.SmackProcessLabel, config_parse_warn_compat, DISABLED_CONFIGURATION, 0')' +)m4_dnl +m4_define(`KILL_CONTEXT_CONFIG_ITEMS', +`$1.SendSIGKILL, config_parse_bool, 0, offsetof($1, kill_context.send_sigkill) +$1.SendSIGHUP, config_parse_bool, 0, offsetof($1, kill_context.send_sighup) +$1.KillMode, config_parse_kill_mode, 0, offsetof($1, kill_context.kill_mode) +$1.KillSignal, config_parse_signal, 0, offsetof($1, kill_context.kill_signal)' +)m4_dnl +m4_define(`CGROUP_CONTEXT_CONFIG_ITEMS', +`$1.Slice, config_parse_unit_slice, 0, 0 +$1.CPUAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.cpu_accounting) +$1.CPUWeight, config_parse_cpu_weight, 0, offsetof($1, cgroup_context.cpu_weight) +$1.StartupCPUWeight, config_parse_cpu_weight, 0, offsetof($1, cgroup_context.startup_cpu_weight) +$1.CPUShares, config_parse_cpu_shares, 0, offsetof($1, cgroup_context.cpu_shares) +$1.StartupCPUShares, config_parse_cpu_shares, 0, offsetof($1, cgroup_context.startup_cpu_shares) +$1.CPUQuota, config_parse_cpu_quota, 0, offsetof($1, cgroup_context) +$1.MemoryAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.memory_accounting) +$1.MemoryLow, config_parse_memory_limit, 0, offsetof($1, cgroup_context) +$1.MemoryHigh, config_parse_memory_limit, 0, offsetof($1, cgroup_context) +$1.MemoryMax, config_parse_memory_limit, 0, offsetof($1, cgroup_context) +$1.MemorySwapMax, config_parse_memory_limit, 0, offsetof($1, cgroup_context) +$1.MemoryLimit, config_parse_memory_limit, 0, offsetof($1, cgroup_context) +$1.DeviceAllow, config_parse_device_allow, 0, offsetof($1, cgroup_context) +$1.DevicePolicy, config_parse_device_policy, 0, offsetof($1, cgroup_context.device_policy) +$1.IOAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.io_accounting) +$1.IOWeight, config_parse_io_weight, 0, offsetof($1, cgroup_context.io_weight) +$1.StartupIOWeight, config_parse_io_weight, 0, offsetof($1, cgroup_context.startup_io_weight) +$1.IODeviceWeight, config_parse_io_device_weight, 0, offsetof($1, cgroup_context) +$1.IOReadBandwidthMax, config_parse_io_limit, 0, offsetof($1, cgroup_context) +$1.IOWriteBandwidthMax, config_parse_io_limit, 0, offsetof($1, cgroup_context) +$1.IOReadIOPSMax, config_parse_io_limit, 0, offsetof($1, cgroup_context) +$1.IOWriteIOPSMax, config_parse_io_limit, 0, offsetof($1, cgroup_context) +$1.BlockIOAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.blockio_accounting) +$1.BlockIOWeight, config_parse_blockio_weight, 0, offsetof($1, cgroup_context.blockio_weight) +$1.StartupBlockIOWeight, config_parse_blockio_weight, 0, offsetof($1, cgroup_context.startup_blockio_weight) +$1.BlockIODeviceWeight, config_parse_blockio_device_weight, 0, offsetof($1, cgroup_context) +$1.BlockIOReadBandwidth, config_parse_blockio_bandwidth, 0, offsetof($1, cgroup_context) +$1.BlockIOWriteBandwidth, config_parse_blockio_bandwidth, 0, offsetof($1, cgroup_context) +$1.TasksAccounting, config_parse_bool, 0, offsetof($1, cgroup_context.tasks_accounting) +$1.TasksMax, config_parse_tasks_max, 0, offsetof($1, cgroup_context.tasks_max) +$1.Delegate, config_parse_bool, 0, offsetof($1, cgroup_context.delegate) +$1.NetClass, config_parse_warn_compat, DISABLED_LEGACY, 0' +)m4_dnl +Unit.Description, config_parse_unit_string_printf, 0, offsetof(Unit, description) +Unit.Documentation, config_parse_documentation, 0, offsetof(Unit, documentation) +Unit.SourcePath, config_parse_path, 0, offsetof(Unit, source_path) +Unit.Requires, config_parse_unit_deps, UNIT_REQUIRES, 0 +Unit.Requisite, config_parse_unit_deps, UNIT_REQUISITE, 0 +Unit.Wants, config_parse_unit_deps, UNIT_WANTS, 0 +Unit.BindsTo, config_parse_unit_deps, UNIT_BINDS_TO, 0 +Unit.BindTo, config_parse_unit_deps, UNIT_BINDS_TO, 0 +Unit.Conflicts, config_parse_unit_deps, UNIT_CONFLICTS, 0 +Unit.Before, config_parse_unit_deps, UNIT_BEFORE, 0 +Unit.After, config_parse_unit_deps, UNIT_AFTER, 0 +Unit.OnFailure, config_parse_unit_deps, UNIT_ON_FAILURE, 0 +Unit.PropagatesReloadTo, config_parse_unit_deps, UNIT_PROPAGATES_RELOAD_TO, 0 +Unit.PropagateReloadTo, config_parse_unit_deps, UNIT_PROPAGATES_RELOAD_TO, 0 +Unit.ReloadPropagatedFrom, config_parse_unit_deps, UNIT_RELOAD_PROPAGATED_FROM, 0 +Unit.PropagateReloadFrom, config_parse_unit_deps, UNIT_RELOAD_PROPAGATED_FROM, 0 +Unit.PartOf, config_parse_unit_deps, UNIT_PART_OF, 0 +Unit.JoinsNamespaceOf, config_parse_unit_deps, UNIT_JOINS_NAMESPACE_OF, 0 +Unit.RequiresOverridable, config_parse_obsolete_unit_deps, UNIT_REQUIRES, 0 +Unit.RequisiteOverridable, config_parse_obsolete_unit_deps, UNIT_REQUISITE, 0 +Unit.RequiresMountsFor, config_parse_unit_requires_mounts_for, 0, 0 +Unit.StopWhenUnneeded, config_parse_bool, 0, offsetof(Unit, stop_when_unneeded) +Unit.RefuseManualStart, config_parse_bool, 0, offsetof(Unit, refuse_manual_start) +Unit.RefuseManualStop, config_parse_bool, 0, offsetof(Unit, refuse_manual_stop) +Unit.AllowIsolate, config_parse_bool, 0, offsetof(Unit, allow_isolate) +Unit.DefaultDependencies, config_parse_bool, 0, offsetof(Unit, default_dependencies) +Unit.OnFailureJobMode, config_parse_job_mode, 0, offsetof(Unit, on_failure_job_mode) +Unit.OnFailureIsolate, config_parse_job_mode_isolate, 0, offsetof(Unit, on_failure_job_mode) +Unit.IgnoreOnIsolate, config_parse_bool, 0, offsetof(Unit, ignore_on_isolate) +Unit.IgnoreOnSnapshot, config_parse_warn_compat, DISABLED_LEGACY, 0 +Unit.JobTimeoutSec, config_parse_sec_fix_0, 0, offsetof(Unit, job_timeout) +Unit.JobTimeoutAction, config_parse_emergency_action, 0, offsetof(Unit, job_timeout_action) +Unit.JobTimeoutRebootArgument, config_parse_string, 0, offsetof(Unit, job_timeout_reboot_arg) +Unit.StartLimitIntervalSec, config_parse_sec, 0, offsetof(Unit, start_limit.interval) +m4_dnl The following is a legacy alias name for compatibility +Unit.StartLimitInterval, config_parse_sec, 0, offsetof(Unit, start_limit.interval) +Unit.StartLimitBurst, config_parse_unsigned, 0, offsetof(Unit, start_limit.burst) +Unit.StartLimitAction, config_parse_emergency_action, 0, offsetof(Unit, start_limit_action) +Unit.RebootArgument, config_parse_string, 0, offsetof(Unit, reboot_arg) +Unit.ConditionPathExists, config_parse_unit_condition_path, CONDITION_PATH_EXISTS, offsetof(Unit, conditions) +Unit.ConditionPathExistsGlob, config_parse_unit_condition_path, CONDITION_PATH_EXISTS_GLOB, offsetof(Unit, conditions) +Unit.ConditionPathIsDirectory, config_parse_unit_condition_path, CONDITION_PATH_IS_DIRECTORY, offsetof(Unit, conditions) +Unit.ConditionPathIsSymbolicLink,config_parse_unit_condition_path, CONDITION_PATH_IS_SYMBOLIC_LINK,offsetof(Unit, conditions) +Unit.ConditionPathIsMountPoint, config_parse_unit_condition_path, CONDITION_PATH_IS_MOUNT_POINT, offsetof(Unit, conditions) +Unit.ConditionPathIsReadWrite, config_parse_unit_condition_path, CONDITION_PATH_IS_READ_WRITE, offsetof(Unit, conditions) +Unit.ConditionDirectoryNotEmpty, config_parse_unit_condition_path, CONDITION_DIRECTORY_NOT_EMPTY, offsetof(Unit, conditions) +Unit.ConditionFileNotEmpty, config_parse_unit_condition_path, CONDITION_FILE_NOT_EMPTY, offsetof(Unit, conditions) +Unit.ConditionFileIsExecutable, config_parse_unit_condition_path, CONDITION_FILE_IS_EXECUTABLE, offsetof(Unit, conditions) +Unit.ConditionNeedsUpdate, config_parse_unit_condition_path, CONDITION_NEEDS_UPDATE, offsetof(Unit, conditions) +Unit.ConditionFirstBoot, config_parse_unit_condition_string, CONDITION_FIRST_BOOT, offsetof(Unit, conditions) +Unit.ConditionKernelCommandLine, config_parse_unit_condition_string, CONDITION_KERNEL_COMMAND_LINE, offsetof(Unit, conditions) +Unit.ConditionArchitecture, config_parse_unit_condition_string, CONDITION_ARCHITECTURE, offsetof(Unit, conditions) +Unit.ConditionVirtualization, config_parse_unit_condition_string, CONDITION_VIRTUALIZATION, offsetof(Unit, conditions) +Unit.ConditionSecurity, config_parse_unit_condition_string, CONDITION_SECURITY, offsetof(Unit, conditions) +Unit.ConditionCapability, config_parse_unit_condition_string, CONDITION_CAPABILITY, offsetof(Unit, conditions) +Unit.ConditionHost, config_parse_unit_condition_string, CONDITION_HOST, offsetof(Unit, conditions) +Unit.ConditionACPower, config_parse_unit_condition_string, CONDITION_AC_POWER, offsetof(Unit, conditions) +Unit.ConditionNull, config_parse_unit_condition_null, 0, offsetof(Unit, conditions) +Unit.AssertPathExists, config_parse_unit_condition_path, CONDITION_PATH_EXISTS, offsetof(Unit, asserts) +Unit.AssertPathExistsGlob, config_parse_unit_condition_path, CONDITION_PATH_EXISTS_GLOB, offsetof(Unit, asserts) +Unit.AssertPathIsDirectory, config_parse_unit_condition_path, CONDITION_PATH_IS_DIRECTORY, offsetof(Unit, asserts) +Unit.AssertPathIsSymbolicLink, config_parse_unit_condition_path, CONDITION_PATH_IS_SYMBOLIC_LINK,offsetof(Unit, asserts) +Unit.AssertPathIsMountPoint, config_parse_unit_condition_path, CONDITION_PATH_IS_MOUNT_POINT, offsetof(Unit, asserts) +Unit.AssertPathIsReadWrite, config_parse_unit_condition_path, CONDITION_PATH_IS_READ_WRITE, offsetof(Unit, asserts) +Unit.AssertDirectoryNotEmpty, config_parse_unit_condition_path, CONDITION_DIRECTORY_NOT_EMPTY, offsetof(Unit, asserts) +Unit.AssertFileNotEmpty, config_parse_unit_condition_path, CONDITION_FILE_NOT_EMPTY, offsetof(Unit, asserts) +Unit.AssertFileIsExecutable, config_parse_unit_condition_path, CONDITION_FILE_IS_EXECUTABLE, offsetof(Unit, asserts) +Unit.AssertNeedsUpdate, config_parse_unit_condition_path, CONDITION_NEEDS_UPDATE, offsetof(Unit, asserts) +Unit.AssertFirstBoot, config_parse_unit_condition_string, CONDITION_FIRST_BOOT, offsetof(Unit, asserts) +Unit.AssertKernelCommandLine, config_parse_unit_condition_string, CONDITION_KERNEL_COMMAND_LINE, offsetof(Unit, asserts) +Unit.AssertArchitecture, config_parse_unit_condition_string, CONDITION_ARCHITECTURE, offsetof(Unit, asserts) +Unit.AssertVirtualization, config_parse_unit_condition_string, CONDITION_VIRTUALIZATION, offsetof(Unit, asserts) +Unit.AssertSecurity, config_parse_unit_condition_string, CONDITION_SECURITY, offsetof(Unit, asserts) +Unit.AssertCapability, config_parse_unit_condition_string, CONDITION_CAPABILITY, offsetof(Unit, asserts) +Unit.AssertHost, config_parse_unit_condition_string, CONDITION_HOST, offsetof(Unit, asserts) +Unit.AssertACPower, config_parse_unit_condition_string, CONDITION_AC_POWER, offsetof(Unit, asserts) +Unit.AssertNull, config_parse_unit_condition_null, 0, offsetof(Unit, asserts) +m4_dnl +Service.PIDFile, config_parse_unit_path_printf, 0, offsetof(Service, pid_file) +Service.ExecStartPre, config_parse_exec, SERVICE_EXEC_START_PRE, offsetof(Service, exec_command) +Service.ExecStart, config_parse_exec, SERVICE_EXEC_START, offsetof(Service, exec_command) +Service.ExecStartPost, config_parse_exec, SERVICE_EXEC_START_POST, offsetof(Service, exec_command) +Service.ExecReload, config_parse_exec, SERVICE_EXEC_RELOAD, offsetof(Service, exec_command) +Service.ExecStop, config_parse_exec, SERVICE_EXEC_STOP, offsetof(Service, exec_command) +Service.ExecStopPost, config_parse_exec, SERVICE_EXEC_STOP_POST, offsetof(Service, exec_command) +Service.RestartSec, config_parse_sec, 0, offsetof(Service, restart_usec) +Service.TimeoutSec, config_parse_service_timeout, 0, 0 +Service.TimeoutStartSec, config_parse_service_timeout, 0, 0 +Service.TimeoutStopSec, config_parse_service_timeout, 0, 0 +Service.RuntimeMaxSec, config_parse_sec, 0, offsetof(Service, runtime_max_usec) +Service.WatchdogSec, config_parse_sec, 0, offsetof(Service, watchdog_usec) +m4_dnl The following three only exist for compatibility, they moved into Unit, see above +Service.StartLimitInterval, config_parse_sec, 0, offsetof(Unit, start_limit.interval) +Service.StartLimitBurst, config_parse_unsigned, 0, offsetof(Unit, start_limit.burst) +Service.StartLimitAction, config_parse_emergency_action, 0, offsetof(Unit, start_limit_action) +Service.RebootArgument, config_parse_string, 0, offsetof(Unit, reboot_arg) +Service.FailureAction, config_parse_emergency_action, 0, offsetof(Service, emergency_action) +Service.Type, config_parse_service_type, 0, offsetof(Service, type) +Service.Restart, config_parse_service_restart, 0, offsetof(Service, restart) +Service.PermissionsStartOnly, config_parse_bool, 0, offsetof(Service, permissions_start_only) +Service.RootDirectoryStartOnly, config_parse_bool, 0, offsetof(Service, root_directory_start_only) +Service.RemainAfterExit, config_parse_bool, 0, offsetof(Service, remain_after_exit) +Service.GuessMainPID, config_parse_bool, 0, offsetof(Service, guess_main_pid) +Service.RestartPreventExitStatus, config_parse_set_status, 0, offsetof(Service, restart_prevent_status) +Service.RestartForceExitStatus, config_parse_set_status, 0, offsetof(Service, restart_force_status) +Service.SuccessExitStatus, config_parse_set_status, 0, offsetof(Service, success_status) +Service.SysVStartPriority, config_parse_warn_compat, DISABLED_LEGACY, 0 +Service.NonBlocking, config_parse_bool, 0, offsetof(Service, exec_context.non_blocking) +Service.BusName, config_parse_bus_name, 0, offsetof(Service, bus_name) +Service.FileDescriptorStoreMax, config_parse_unsigned, 0, offsetof(Service, n_fd_store_max) +Service.NotifyAccess, config_parse_notify_access, 0, offsetof(Service, notify_access) +Service.Sockets, config_parse_service_sockets, 0, 0 +Service.BusPolicy, config_parse_warn_compat, DISABLED_LEGACY, 0 +Service.USBFunctionDescriptors, config_parse_path, 0, offsetof(Service, usb_function_descriptors) +Service.USBFunctionStrings, config_parse_path, 0, offsetof(Service, usb_function_strings) +EXEC_CONTEXT_CONFIG_ITEMS(Service)m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Service)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Service)m4_dnl +m4_dnl +Socket.ListenStream, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenDatagram, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenSequentialPacket, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenFIFO, config_parse_socket_listen, SOCKET_FIFO, 0 +Socket.ListenNetlink, config_parse_socket_listen, SOCKET_SOCKET, 0 +Socket.ListenSpecial, config_parse_socket_listen, SOCKET_SPECIAL, 0 +Socket.ListenMessageQueue, config_parse_socket_listen, SOCKET_MQUEUE, 0 +Socket.ListenUSBFunction, config_parse_socket_listen, SOCKET_USB_FUNCTION, 0 +Socket.SocketProtocol, config_parse_socket_protocol, 0, 0 +Socket.BindIPv6Only, config_parse_socket_bind, 0, 0, +Socket.Backlog, config_parse_unsigned, 0, offsetof(Socket, backlog) +Socket.BindToDevice, config_parse_socket_bindtodevice, 0, 0 +Socket.ExecStartPre, config_parse_exec, SOCKET_EXEC_START_PRE, offsetof(Socket, exec_command) +Socket.ExecStartPost, config_parse_exec, SOCKET_EXEC_START_POST, offsetof(Socket, exec_command) +Socket.ExecStopPre, config_parse_exec, SOCKET_EXEC_STOP_PRE, offsetof(Socket, exec_command) +Socket.ExecStopPost, config_parse_exec, SOCKET_EXEC_STOP_POST, offsetof(Socket, exec_command) +Socket.TimeoutSec, config_parse_sec, 0, offsetof(Socket, timeout_usec) +Socket.SocketUser, config_parse_user_group, 0, offsetof(Socket, user) +Socket.SocketGroup, config_parse_user_group, 0, offsetof(Socket, group) +Socket.SocketMode, config_parse_mode, 0, offsetof(Socket, socket_mode) +Socket.DirectoryMode, config_parse_mode, 0, offsetof(Socket, directory_mode) +Socket.Accept, config_parse_bool, 0, offsetof(Socket, accept) +Socket.Writable, config_parse_bool, 0, offsetof(Socket, writable) +Socket.MaxConnections, config_parse_unsigned, 0, offsetof(Socket, max_connections) +Socket.MaxConnectionsPerSource, config_parse_unsigned, 0, offsetof(Socket, max_connections_per_source) +Socket.KeepAlive, config_parse_bool, 0, offsetof(Socket, keep_alive) +Socket.KeepAliveTimeSec, config_parse_sec, 0, offsetof(Socket, keep_alive_time) +Socket.KeepAliveIntervalSec, config_parse_sec, 0, offsetof(Socket, keep_alive_interval) +Socket.KeepAliveProbes, config_parse_unsigned, 0, offsetof(Socket, keep_alive_cnt) +Socket.DeferAcceptSec, config_parse_sec, 0, offsetof(Socket, defer_accept) +Socket.NoDelay, config_parse_bool, 0, offsetof(Socket, no_delay) +Socket.Priority, config_parse_int, 0, offsetof(Socket, priority) +Socket.ReceiveBuffer, config_parse_iec_size, 0, offsetof(Socket, receive_buffer) +Socket.SendBuffer, config_parse_iec_size, 0, offsetof(Socket, send_buffer) +Socket.IPTOS, config_parse_ip_tos, 0, offsetof(Socket, ip_tos) +Socket.IPTTL, config_parse_int, 0, offsetof(Socket, ip_ttl) +Socket.Mark, config_parse_int, 0, offsetof(Socket, mark) +Socket.PipeSize, config_parse_iec_size, 0, offsetof(Socket, pipe_size) +Socket.FreeBind, config_parse_bool, 0, offsetof(Socket, free_bind) +Socket.Transparent, config_parse_bool, 0, offsetof(Socket, transparent) +Socket.Broadcast, config_parse_bool, 0, offsetof(Socket, broadcast) +Socket.PassCredentials, config_parse_bool, 0, offsetof(Socket, pass_cred) +Socket.PassSecurity, config_parse_bool, 0, offsetof(Socket, pass_sec) +Socket.TCPCongestion, config_parse_string, 0, offsetof(Socket, tcp_congestion) +Socket.ReusePort, config_parse_bool, 0, offsetof(Socket, reuse_port) +Socket.MessageQueueMaxMessages, config_parse_long, 0, offsetof(Socket, mq_maxmsg) +Socket.MessageQueueMessageSize, config_parse_long, 0, offsetof(Socket, mq_msgsize) +Socket.RemoveOnStop, config_parse_bool, 0, offsetof(Socket, remove_on_stop) +Socket.Symlinks, config_parse_unit_path_strv_printf, 0, offsetof(Socket, symlinks) +Socket.FileDescriptorName, config_parse_fdname, 0, 0 +Socket.Service, config_parse_socket_service, 0, 0 +Socket.TriggerLimitIntervalSec, config_parse_sec, 0, offsetof(Socket, trigger_limit.interval) +Socket.TriggerLimitBurst, config_parse_unsigned, 0, offsetof(Socket, trigger_limit.burst) +m4_ifdef(`HAVE_SMACK', +`Socket.SmackLabel, config_parse_string, 0, offsetof(Socket, smack) +Socket.SmackLabelIPIn, config_parse_string, 0, offsetof(Socket, smack_ip_in) +Socket.SmackLabelIPOut, config_parse_string, 0, offsetof(Socket, smack_ip_out)', +`Socket.SmackLabel, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +Socket.SmackLabelIPIn, config_parse_warn_compat, DISABLED_CONFIGURATION, 0 +Socket.SmackLabelIPOut, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +m4_ifdef(`HAVE_SELINUX', +`Socket.SELinuxContextFromNet, config_parse_bool, 0, offsetof(Socket, selinux_context_from_net)', +`Socket.SELinuxContextFromNet, config_parse_warn_compat, DISABLED_CONFIGURATION, 0') +EXEC_CONTEXT_CONFIG_ITEMS(Socket)m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Socket)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Socket)m4_dnl +m4_dnl +BusName.Name, config_parse_string, 0, offsetof(BusName, name) +BusName.Activating, config_parse_bool, 0, offsetof(BusName, activating) +BusName.Service, config_parse_busname_service, 0, 0 +BusName.AllowUser, config_parse_bus_policy, 0, 0 +BusName.AllowGroup, config_parse_bus_policy, 0, 0 +BusName.AllowWorld, config_parse_bus_policy_world, 0, offsetof(BusName, policy_world) +BusName.SELinuxContext, config_parse_exec_selinux_context, 0, 0 +BusName.AcceptFileDescriptors, config_parse_bool, 0, offsetof(BusName, accept_fd) +m4_dnl +Mount.What, config_parse_string, 0, offsetof(Mount, parameters_fragment.what) +Mount.Where, config_parse_path, 0, offsetof(Mount, where) +Mount.Options, config_parse_string, 0, offsetof(Mount, parameters_fragment.options) +Mount.Type, config_parse_string, 0, offsetof(Mount, parameters_fragment.fstype) +Mount.TimeoutSec, config_parse_sec, 0, offsetof(Mount, timeout_usec) +Mount.DirectoryMode, config_parse_mode, 0, offsetof(Mount, directory_mode) +Mount.SloppyOptions, config_parse_bool, 0, offsetof(Mount, sloppy_options) +Mount.LazyUnmount, config_parse_bool, 0, offsetof(Mount, lazy_unmount) +Mount.ForceUnmount, config_parse_bool, 0, offsetof(Mount, force_unmount) +EXEC_CONTEXT_CONFIG_ITEMS(Mount)m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Mount)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Mount)m4_dnl +m4_dnl +Automount.Where, config_parse_path, 0, offsetof(Automount, where) +Automount.DirectoryMode, config_parse_mode, 0, offsetof(Automount, directory_mode) +Automount.TimeoutIdleSec, config_parse_sec, 0, offsetof(Automount, timeout_idle_usec) +m4_dnl +Swap.What, config_parse_path, 0, offsetof(Swap, parameters_fragment.what) +Swap.Priority, config_parse_int, 0, offsetof(Swap, parameters_fragment.priority) +Swap.Options, config_parse_string, 0, offsetof(Swap, parameters_fragment.options) +Swap.TimeoutSec, config_parse_sec, 0, offsetof(Swap, timeout_usec) +EXEC_CONTEXT_CONFIG_ITEMS(Swap)m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Swap)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Swap)m4_dnl +m4_dnl +Timer.OnCalendar, config_parse_timer, 0, 0 +Timer.OnActiveSec, config_parse_timer, 0, 0 +Timer.OnBootSec, config_parse_timer, 0, 0 +Timer.OnStartupSec, config_parse_timer, 0, 0 +Timer.OnUnitActiveSec, config_parse_timer, 0, 0 +Timer.OnUnitInactiveSec, config_parse_timer, 0, 0 +Timer.Persistent, config_parse_bool, 0, offsetof(Timer, persistent) +Timer.WakeSystem, config_parse_bool, 0, offsetof(Timer, wake_system) +Timer.RemainAfterElapse, config_parse_bool, 0, offsetof(Timer, remain_after_elapse) +Timer.AccuracySec, config_parse_sec, 0, offsetof(Timer, accuracy_usec) +Timer.RandomizedDelaySec, config_parse_sec, 0, offsetof(Timer, random_usec) +Timer.Unit, config_parse_trigger_unit, 0, 0 +m4_dnl +Path.PathExists, config_parse_path_spec, 0, 0 +Path.PathExistsGlob, config_parse_path_spec, 0, 0 +Path.PathChanged, config_parse_path_spec, 0, 0 +Path.PathModified, config_parse_path_spec, 0, 0 +Path.DirectoryNotEmpty, config_parse_path_spec, 0, 0 +Path.Unit, config_parse_trigger_unit, 0, 0 +Path.MakeDirectory, config_parse_bool, 0, offsetof(Path, make_directory) +Path.DirectoryMode, config_parse_mode, 0, offsetof(Path, directory_mode) +m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Slice)m4_dnl +m4_dnl +CGROUP_CONTEXT_CONFIG_ITEMS(Scope)m4_dnl +KILL_CONTEXT_CONFIG_ITEMS(Scope)m4_dnl +Scope.TimeoutStopSec, config_parse_sec, 0, offsetof(Scope, timeout_stop_usec) +m4_dnl The [Install] section is ignored here. +Install.Alias, NULL, 0, 0 +Install.WantedBy, NULL, 0, 0 +Install.RequiredBy, NULL, 0, 0 +Install.Also, NULL, 0, 0 +Install.DefaultInstance, NULL, 0, 0 diff --git a/src/grp-system/libcore/src/load-fragment.c b/src/grp-system/libcore/src/load-fragment.c new file mode 100644 index 0000000000..927615360e --- /dev/null +++ b/src/grp-system/libcore/src/load-fragment.c @@ -0,0 +1,4389 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + Copyright 2012 Holger Hans Peter Freyther + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <fcntl.h> + +#include <linux/fs.h> +#include <linux/oom.h> +#ifdef HAVE_SECCOMP +#include <seccomp.h> +#endif +#include <sched.h> +#include <string.h> +#include <sys/resource.h> +#include <sys/stat.h> + +#include "core/cgroup.h" +#include "core/load-fragment.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-internal.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/af-list.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cap-list.h" +#include "systemd-basic/capability-util.h" +#include "systemd-basic/cpu-set-util.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/errno-list.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/ioprio.h" +#include "systemd-basic/log.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/rlimit-util.h" +#include "systemd-shared/conf-parser.h" +#ifdef HAVE_SECCOMP +#include "systemd-shared/seccomp-util.h" +#endif +#include "core/unit.h" +#include "systemd-basic/securebits.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/utf8.h" +#include "systemd-basic/web-util.h" + +#include "unit-printf.h" + +int config_parse_warn_compat( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Disabled reason = ltype; + + switch(reason) { + case DISABLED_CONFIGURATION: + log_syntax(unit, LOG_DEBUG, filename, line, 0, + "Support for option %s= has been disabled at compile time and it is ignored", lvalue); + break; + case DISABLED_LEGACY: + log_syntax(unit, LOG_INFO, filename, line, 0, + "Support for option %s= has been removed and it is ignored", lvalue); + break; + case DISABLED_EXPERIMENTAL: + log_syntax(unit, LOG_INFO, filename, line, 0, + "Support for option %s= has not yet been enabled and it is ignored", lvalue); + break; + }; + + return 0; +} + +int config_parse_unit_deps( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + UnitDependency d = ltype; + Unit *u = userdata; + const char *p; + + assert(filename); + assert(lvalue); + assert(rvalue); + + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL, *k = NULL; + int r; + + r = extract_first_word(&p, &word, NULL, EXTRACT_RETAIN_ESCAPE); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Invalid syntax, ignoring: %s", rvalue); + break; + } + + r = unit_name_printf(u, word, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + continue; + } + + r = unit_add_dependency_by_name(u, d, k, NULL, true); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add dependency on %s, ignoring: %m", k); + } + + return 0; +} + +int config_parse_obsolete_unit_deps( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + log_syntax(unit, LOG_WARNING, filename, line, 0, + "Unit dependency type %s= is obsolete, replacing by %s=, please update your unit file", lvalue, unit_dependency_to_string(ltype)); + + return config_parse_unit_deps(unit, filename, line, section, section_line, lvalue, ltype, rvalue, data, userdata); +} + +int config_parse_unit_string_printf( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *k = NULL; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + return config_parse_string(unit, filename, line, section, section_line, lvalue, ltype, k, data, userdata); +} + +int config_parse_unit_strv_printf( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + _cleanup_free_ char *k = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + return config_parse_strv(unit, filename, line, section, section_line, lvalue, ltype, k, data, userdata); +} + +int config_parse_unit_path_printf( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *k = NULL; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + return config_parse_path(unit, filename, line, section, section_line, lvalue, ltype, k, data, userdata); +} + +int config_parse_unit_path_strv_printf( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char ***x = data; + const char *word, *state; + Unit *u = userdata; + size_t l; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + _cleanup_free_ char *k = NULL; + char t[l+1]; + + memcpy(t, word, l); + t[l] = 0; + + r = unit_full_printf(u, t, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", t); + return 0; + } + + if (!utf8_is_valid(k)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, rvalue); + return 0; + } + + if (!path_is_absolute(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Symlink path %s is not absolute, ignoring: %m", k); + return 0; + } + + path_kill_slashes(k); + + r = strv_push(x, k); + if (r < 0) + return log_oom(); + + k = NULL; + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid syntax, ignoring."); + + return 0; +} + +int config_parse_socket_listen(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ SocketPort *p = NULL; + SocketPort *tail; + Socket *s; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + s = SOCKET(data); + + if (isempty(rvalue)) { + /* An empty assignment removes all ports */ + socket_free_ports(s); + return 0; + } + + p = new0(SocketPort, 1); + if (!p) + return log_oom(); + + if (ltype != SOCKET_SOCKET) { + + p->type = ltype; + r = unit_full_printf(UNIT(s), rvalue, &p->path); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + path_kill_slashes(p->path); + + } else if (streq(lvalue, "ListenNetlink")) { + _cleanup_free_ char *k = NULL; + + p->type = SOCKET_SOCKET; + r = unit_full_printf(UNIT(s), rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + r = socket_address_parse_netlink(&p->address, k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse address value, ignoring: %s", rvalue); + return 0; + } + + } else { + _cleanup_free_ char *k = NULL; + + p->type = SOCKET_SOCKET; + r = unit_full_printf(UNIT(s), rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r,"Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + r = socket_address_parse_and_warn(&p->address, k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse address value, ignoring: %s", rvalue); + return 0; + } + + if (streq(lvalue, "ListenStream")) + p->address.type = SOCK_STREAM; + else if (streq(lvalue, "ListenDatagram")) + p->address.type = SOCK_DGRAM; + else { + assert(streq(lvalue, "ListenSequentialPacket")); + p->address.type = SOCK_SEQPACKET; + } + + if (socket_address_family(&p->address) != AF_LOCAL && p->address.type == SOCK_SEQPACKET) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Address family not supported, ignoring: %s", rvalue); + return 0; + } + } + + p->fd = -1; + p->auxiliary_fds = NULL; + p->n_auxiliary_fds = 0; + p->socket = s; + + if (s->ports) { + LIST_FIND_TAIL(port, s->ports, tail); + LIST_INSERT_AFTER(port, s->ports, tail, p); + } else + LIST_PREPEND(port, s->ports, p); + p = NULL; + + return 0; +} + +int config_parse_socket_protocol(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Socket *s; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + s = SOCKET(data); + + if (streq(rvalue, "udplite")) + s->socket_protocol = IPPROTO_UDPLITE; + else if (streq(rvalue, "sctp")) + s->socket_protocol = IPPROTO_SCTP; + else { + log_syntax(unit, LOG_ERR, filename, line, 0, "Socket protocol not supported, ignoring: %s", rvalue); + return 0; + } + + return 0; +} + +int config_parse_socket_bind(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Socket *s; + SocketAddressBindIPv6Only b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + s = SOCKET(data); + + b = socket_address_bind_ipv6_only_from_string(rvalue); + if (b < 0) { + int r; + + r = parse_boolean(rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse bind IPv6 only value, ignoring: %s", rvalue); + return 0; + } + + s->bind_ipv6_only = r ? SOCKET_ADDRESS_IPV6_ONLY : SOCKET_ADDRESS_BOTH; + } else + s->bind_ipv6_only = b; + + return 0; +} + +int config_parse_exec_nice( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int priority, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = parse_nice(rvalue, &priority); + if (r < 0) { + if (r == -ERANGE) + log_syntax(unit, LOG_ERR, filename, line, r, "Nice priority out of range, ignoring: %s", rvalue); + else + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse nice priority, ignoring: %s", rvalue); + + return 0; + } + + c->nice = priority; + c->nice_set = true; + + return 0; +} + +int config_parse_exec_oom_score_adjust(const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int oa, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &oa); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse the OOM score adjust value, ignoring: %s", rvalue); + return 0; + } + + if (oa < OOM_SCORE_ADJ_MIN || oa > OOM_SCORE_ADJ_MAX) { + log_syntax(unit, LOG_ERR, filename, line, 0, "OOM score adjust value out of range, ignoring: %s", rvalue); + return 0; + } + + c->oom_score_adjust = oa; + c->oom_score_adjust_set = true; + + return 0; +} + +int config_parse_exec( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecCommand **e = data; + const char *p; + bool semicolon; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(e); + + e += ltype; + rvalue += strspn(rvalue, WHITESPACE); + + if (isempty(rvalue)) { + /* An empty assignment resets the list */ + *e = exec_command_free_list(*e); + return 0; + } + + p = rvalue; + do { + _cleanup_free_ char *path = NULL, *firstword = NULL; + bool separate_argv0 = false, ignore = false, privileged = false; + _cleanup_free_ ExecCommand *nce = NULL; + _cleanup_strv_free_ char **n = NULL; + size_t nlen = 0, nbufsize = 0; + char *f; + int i; + + semicolon = false; + + r = extract_first_word_and_warn(&p, &firstword, WHITESPACE, EXTRACT_QUOTES|EXTRACT_CUNESCAPE, unit, filename, line, rvalue); + if (r <= 0) + return 0; + + f = firstword; + for (i = 0; i < 3; i++) { + /* We accept an absolute path as first argument. + * If it's prefixed with - and the path doesn't exist, + * we ignore it instead of erroring out; + * if it's prefixed with @, we allow overriding of argv[0]; + * and if it's prefixed with !, it will be run with full privileges */ + if (*f == '-' && !ignore) + ignore = true; + else if (*f == '@' && !separate_argv0) + separate_argv0 = true; + else if (*f == '+' && !privileged) + privileged = true; + else + break; + f++; + } + + if (isempty(f)) { + /* First word is either "-" or "@" with no command. */ + log_syntax(unit, LOG_ERR, filename, line, 0, "Empty path in command line, ignoring: \"%s\"", rvalue); + return 0; + } + if (!string_is_safe(f)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Executable path contains special characters, ignoring: %s", rvalue); + return 0; + } + if (!path_is_absolute(f)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Executable path is not absolute, ignoring: %s", rvalue); + return 0; + } + if (endswith(f, "/")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Executable path specifies a directory, ignoring: %s", rvalue); + return 0; + } + + if (f == firstword) { + path = firstword; + firstword = NULL; + } else { + path = strdup(f); + if (!path) + return log_oom(); + } + + if (!separate_argv0) { + if (!GREEDY_REALLOC(n, nbufsize, nlen + 2)) + return log_oom(); + f = strdup(path); + if (!f) + return log_oom(); + n[nlen++] = f; + n[nlen] = NULL; + } + + path_kill_slashes(path); + + while (!isempty(p)) { + _cleanup_free_ char *word = NULL; + + /* Check explicitly for an unquoted semicolon as + * command separator token. */ + if (p[0] == ';' && (!p[1] || strchr(WHITESPACE, p[1]))) { + p++; + p += strspn(p, WHITESPACE); + semicolon = true; + break; + } + + /* Check for \; explicitly, to not confuse it with \\; + * or "\;" or "\\;" etc. extract_first_word would + * return the same for all of those. */ + if (p[0] == '\\' && p[1] == ';' && (!p[2] || strchr(WHITESPACE, p[2]))) { + p += 2; + p += strspn(p, WHITESPACE); + if (!GREEDY_REALLOC(n, nbufsize, nlen + 2)) + return log_oom(); + f = strdup(";"); + if (!f) + return log_oom(); + n[nlen++] = f; + n[nlen] = NULL; + continue; + } + + r = extract_first_word_and_warn(&p, &word, WHITESPACE, EXTRACT_QUOTES|EXTRACT_CUNESCAPE, unit, filename, line, rvalue); + if (r == 0) + break; + else if (r < 0) + return 0; + + if (!GREEDY_REALLOC(n, nbufsize, nlen + 2)) + return log_oom(); + n[nlen++] = word; + n[nlen] = NULL; + word = NULL; + } + + if (!n || !n[0]) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Empty executable name or zeroeth argument, ignoring: %s", rvalue); + return 0; + } + + nce = new0(ExecCommand, 1); + if (!nce) + return log_oom(); + + nce->argv = n; + nce->path = path; + nce->ignore = ignore; + nce->privileged = privileged; + + exec_command_append_list(e, nce); + + /* Do not _cleanup_free_ these. */ + n = NULL; + path = NULL; + nce = NULL; + + rvalue = p; + } while (semicolon); + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_service_type, service_type, ServiceType, "Failed to parse service type"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_service_restart, service_restart, ServiceRestart, "Failed to parse service restart specifier"); + +int config_parse_socket_bindtodevice( + const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Socket *s = data; + char *n; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (rvalue[0] && !streq(rvalue, "*")) { + if (!ifname_valid(rvalue)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Interface name is invalid, ignoring: %s", rvalue); + return 0; + } + + n = strdup(rvalue); + if (!n) + return log_oom(); + } else + n = NULL; + + free(s->bind_to_device); + s->bind_to_device = n; + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_input, exec_input, ExecInput, "Failed to parse input literal specifier"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_output, exec_output, ExecOutput, "Failed to parse output literal specifier"); + +int config_parse_exec_input(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + ExecContext *c = data; + const char *name; + int r; + + assert(data); + assert(filename); + assert(line); + assert(rvalue); + + name = startswith(rvalue, "fd:"); + if (name) { + /* Strip prefix and validate fd name */ + if (!fdname_is_valid(name)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid file descriptor name, ignoring: %s", name); + return 0; + } + c->std_input = EXEC_INPUT_NAMED_FD; + r = free_and_strdup(&c->stdio_fdname[STDIN_FILENO], name); + if (r < 0) + log_oom(); + return r; + } else { + ExecInput ei = exec_input_from_string(rvalue); + if (ei == _EXEC_INPUT_INVALID) + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse input specifier, ignoring: %s", rvalue); + else + c->std_input = ei; + return 0; + } +} + +int config_parse_exec_output(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + ExecContext *c = data; + ExecOutput eo; + const char *name; + int r; + + assert(data); + assert(filename); + assert(line); + assert(lvalue); + assert(rvalue); + + name = startswith(rvalue, "fd:"); + if (name) { + /* Strip prefix and validate fd name */ + if (!fdname_is_valid(name)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid file descriptor name, ignoring: %s", name); + return 0; + } + eo = EXEC_OUTPUT_NAMED_FD; + } else { + eo = exec_output_from_string(rvalue); + if (eo == _EXEC_OUTPUT_INVALID) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output specifier, ignoring: %s", rvalue); + return 0; + } + } + + if (streq(lvalue, "StandardOutput")) { + c->std_output = eo; + r = free_and_strdup(&c->stdio_fdname[STDOUT_FILENO], name); + if (r < 0) + log_oom(); + return r; + } else if (streq(lvalue, "StandardError")) { + c->std_error = eo; + r = free_and_strdup(&c->stdio_fdname[STDERR_FILENO], name); + if (r < 0) + log_oom(); + return r; + } else { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output property, ignoring: %s", lvalue); + return 0; + } +} + +int config_parse_exec_io_class(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = ioprio_class_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse IO scheduling class, ignoring: %s", rvalue); + return 0; + } + + c->ioprio = IOPRIO_PRIO_VALUE(x, IOPRIO_PRIO_DATA(c->ioprio)); + c->ioprio_set = true; + + return 0; +} + +int config_parse_exec_io_priority(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int i, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0 || i < 0 || i >= IOPRIO_BE_NR) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse IO priority, ignoring: %s", rvalue); + return 0; + } + + c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_PRIO_CLASS(c->ioprio), i); + c->ioprio_set = true; + + return 0; +} + +int config_parse_exec_cpu_sched_policy(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + + ExecContext *c = data; + int x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = sched_policy_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); + return 0; + } + + c->cpu_sched_policy = x; + /* Moving to or from real-time policy? We need to adjust the priority */ + c->cpu_sched_priority = CLAMP(c->cpu_sched_priority, sched_get_priority_min(x), sched_get_priority_max(x)); + c->cpu_sched_set = true; + + return 0; +} + +int config_parse_exec_cpu_sched_prio(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int i, min, max, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); + return 0; + } + + /* On Linux RR/FIFO range from 1 to 99 and OTHER/BATCH may only be 0 */ + min = sched_get_priority_min(c->cpu_sched_policy); + max = sched_get_priority_max(c->cpu_sched_policy); + + if (i < min || i > max) { + log_syntax(unit, LOG_ERR, filename, line, 0, "CPU scheduling priority is out of range, ignoring: %s", rvalue); + return 0; + } + + c->cpu_sched_priority = i; + c->cpu_sched_set = true; + + return 0; +} + +int config_parse_exec_cpu_affinity(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + _cleanup_cpu_free_ cpu_set_t *cpuset = NULL; + int ncpus; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + ncpus = parse_cpu_set_and_warn(rvalue, &cpuset, unit, filename, line, lvalue); + if (ncpus < 0) + return ncpus; + + if (c->cpuset) + CPU_FREE(c->cpuset); + + if (ncpus == 0) + /* An empty assignment resets the CPU list */ + c->cpuset = NULL; + else { + c->cpuset = cpuset; + cpuset = NULL; + } + c->cpuset_ncpus = ncpus; + + return 0; +} + +int config_parse_exec_secure_bits(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + size_t l; + const char *word, *state; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* An empty assignment resets the field */ + c->secure_bits = 0; + return 0; + } + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + if (first_word(word, "keep-caps")) + c->secure_bits |= 1<<SECURE_KEEP_CAPS; + else if (first_word(word, "keep-caps-locked")) + c->secure_bits |= 1<<SECURE_KEEP_CAPS_LOCKED; + else if (first_word(word, "no-setuid-fixup")) + c->secure_bits |= 1<<SECURE_NO_SETUID_FIXUP; + else if (first_word(word, "no-setuid-fixup-locked")) + c->secure_bits |= 1<<SECURE_NO_SETUID_FIXUP_LOCKED; + else if (first_word(word, "noroot")) + c->secure_bits |= 1<<SECURE_NOROOT; + else if (first_word(word, "noroot-locked")) + c->secure_bits |= 1<<SECURE_NOROOT_LOCKED; + else { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse secure bits, ignoring: %s", rvalue); + return 0; + } + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid syntax, garbage at the end, ignoring."); + + return 0; +} + +int config_parse_capability_set( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *capability_set = data; + uint64_t sum = 0, initial = 0; + bool invert = false; + const char *p; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (rvalue[0] == '~') { + invert = true; + rvalue++; + } + + if (strcmp(lvalue, "CapabilityBoundingSet") == 0) + initial = CAP_ALL; /* initialized to all bits on */ + /* else "AmbientCapabilities" initialized to all bits off */ + + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL; + int cap, r; + + r = extract_first_word(&p, &word, NULL, EXTRACT_QUOTES); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse word, ignoring: %s", rvalue); + break; + } + + cap = capability_from_name(word); + if (cap < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse capability in bounding/ambient set, ignoring: %s", word); + continue; + } + + sum |= ((uint64_t) UINT64_C(1)) << (uint64_t) cap; + } + + sum = invert ? ~sum : sum; + + if (sum == 0 || *capability_set == initial) + /* "" or uninitialized data -> replace */ + *capability_set = sum; + else + /* previous data -> merge */ + *capability_set |= sum; + + return 0; +} + +int config_parse_limit( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + struct rlimit **rl = data, d = {}; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = rlimit_parse(ltype, rvalue, &d); + if (r == -EILSEQ) { + log_syntax(unit, LOG_WARNING, filename, line, r, "Soft resource limit chosen higher than hard limit, ignoring: %s", rvalue); + return 0; + } + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse resource value, ignoring: %s", rvalue); + return 0; + } + + if (rl[ltype]) + *rl[ltype] = d; + else { + rl[ltype] = newdup(struct rlimit, &d, 1); + if (!rl[ltype]) + return log_oom(); + } + + return 0; +} + +#ifdef HAVE_SYSV_COMPAT +int config_parse_sysv_priority(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *priority = data; + int i, r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atoi(rvalue, &i); + if (r < 0 || i < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse SysV start priority, ignoring: %s", rvalue); + return 0; + } + + *priority = (int) i; + return 0; +} +#endif + +DEFINE_CONFIG_PARSE_ENUM(config_parse_exec_utmp_mode, exec_utmp_mode, ExecUtmpMode, "Failed to parse utmp mode"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_kill_mode, kill_mode, KillMode, "Failed to parse kill mode"); + +int config_parse_exec_mount_flags(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + + unsigned long flags = 0; + ExecContext *c = data; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (streq(rvalue, "shared")) + flags = MS_SHARED; + else if (streq(rvalue, "slave")) + flags = MS_SLAVE; + else if (streq(rvalue, "private")) + flags = MS_PRIVATE; + else { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse mount flag %s, ignoring.", rvalue); + return 0; + } + + c->mount_flags = flags; + + return 0; +} + +int config_parse_exec_selinux_context( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool ignore; + char *k; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + c->selinux_context = mfree(c->selinux_context); + c->selinux_context_ignore = false; + return 0; + } + + if (rvalue[0] == '-') { + ignore = true; + rvalue++; + } else + ignore = false; + + r = unit_name_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + return 0; + } + + free(c->selinux_context); + c->selinux_context = k; + c->selinux_context_ignore = ignore; + + return 0; +} + +int config_parse_exec_apparmor_profile( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool ignore; + char *k; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + c->apparmor_profile = mfree(c->apparmor_profile); + c->apparmor_profile_ignore = false; + return 0; + } + + if (rvalue[0] == '-') { + ignore = true; + rvalue++; + } else + ignore = false; + + r = unit_name_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + return 0; + } + + free(c->apparmor_profile); + c->apparmor_profile = k; + c->apparmor_profile_ignore = ignore; + + return 0; +} + +int config_parse_exec_smack_process_label( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool ignore; + char *k; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + c->smack_process_label = mfree(c->smack_process_label); + c->smack_process_label_ignore = false; + return 0; + } + + if (rvalue[0] == '-') { + ignore = true; + rvalue++; + } else + ignore = false; + + r = unit_name_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + return 0; + } + + free(c->smack_process_label); + c->smack_process_label = k; + c->smack_process_label_ignore = ignore; + + return 0; +} + +int config_parse_timer(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Timer *t = data; + usec_t usec = 0; + TimerValue *v; + TimerBase b; + CalendarSpec *c = NULL; + Unit *u = userdata; + _cleanup_free_ char *k = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets list */ + timer_free_values(t); + return 0; + } + + b = timer_base_from_string(lvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse timer base, ignoring: %s", lvalue); + return 0; + } + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers in %s, ignoring: %m", rvalue); + return 0; + } + + if (b == TIMER_CALENDAR) { + if (calendar_spec_from_string(k, &c) < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse calendar specification, ignoring: %s", k); + return 0; + } + } else { + if (parse_sec(k, &usec) < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse timer value, ignoring: %s", k); + return 0; + } + } + + v = new0(TimerValue, 1); + if (!v) { + calendar_spec_free(c); + return log_oom(); + } + + v->base = b; + v->value = usec; + v->calendar_spec = c; + + LIST_PREPEND(value, t->values, v); + + return 0; +} + +int config_parse_trigger_unit( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *p = NULL; + Unit *u = data; + UnitType type; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (!set_isempty(u->dependencies[UNIT_TRIGGERS])) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Multiple units to trigger specified, ignoring: %s", rvalue); + return 0; + } + + r = unit_name_printf(u, rvalue, &p); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + return 0; + } + + type = unit_name_to_type(p); + if (type < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unit type not valid, ignoring: %s", rvalue); + return 0; + } + + if (type == u->type) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Trigger cannot be of same type, ignoring: %s", rvalue); + return 0; + } + + r = unit_add_two_dependencies_by_name(u, UNIT_BEFORE, UNIT_TRIGGERS, p, NULL, true); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add trigger on %s, ignoring: %m", p); + return 0; + } + + return 0; +} + +int config_parse_path_spec(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Path *p = data; + PathSpec *s; + PathType b; + _cleanup_free_ char *k = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment clears list */ + path_free_specs(p); + return 0; + } + + b = path_type_from_string(lvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse path type, ignoring: %s", lvalue); + return 0; + } + + r = unit_full_printf(UNIT(p), rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + return 0; + } + + if (!path_is_absolute(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Path is not absolute, ignoring: %s", k); + return 0; + } + + s = new0(PathSpec, 1); + if (!s) + return log_oom(); + + s->unit = UNIT(p); + s->path = path_kill_slashes(k); + k = NULL; + s->type = b; + s->inotify_fd = -1; + + LIST_PREPEND(spec, p->specs, s); + + return 0; +} + +int config_parse_socket_service( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *p = NULL; + Socket *s = data; + Unit *x; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = unit_name_printf(UNIT(s), rvalue, &p); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + if (!endswith(p, ".service")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unit must be of type service, ignoring: %s", rvalue); + return 0; + } + + r = manager_load_unit(UNIT(s)->manager, p, NULL, &error, &x); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load unit %s, ignoring: %s", rvalue, bus_error_message(&error, r)); + return 0; + } + + unit_ref_set(&s->service, x); + + return 0; +} + +int config_parse_fdname( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *p = NULL; + Socket *s = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + s->fdname = mfree(s->fdname); + return 0; + } + + r = unit_name_printf(UNIT(s), rvalue, &p); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + if (!fdname_is_valid(p)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid file descriptor name, ignoring: %s", p); + return 0; + } + + return free_and_replace(s->fdname, p); +} + +int config_parse_service_sockets( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Service *s = data; + const char *p; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL, *k = NULL; + + r = extract_first_word(&p, &word, NULL, 0); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Trailing garbage in sockets, ignoring: %s", rvalue); + break; + } + + r = unit_name_printf(UNIT(s), word, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + continue; + } + + if (!endswith(k, ".socket")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unit must be of type socket, ignoring: %s", k); + continue; + } + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_WANTS, UNIT_AFTER, k, NULL, true); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add dependency on %s, ignoring: %m", k); + + r = unit_add_dependency_by_name(UNIT(s), UNIT_TRIGGERED_BY, k, NULL, true); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add dependency on %s, ignoring: %m", k); + } + + return 0; +} + +int config_parse_bus_name( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *k = NULL; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); + return 0; + } + + if (!service_name_is_valid(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid bus name %s, ignoring.", k); + return 0; + } + + return config_parse_string(unit, filename, line, section, section_line, lvalue, ltype, k, data, userdata); +} + +int config_parse_service_timeout( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Service *s = userdata; + usec_t usec; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(s); + + /* This is called for three cases: TimeoutSec=, TimeoutStopSec= and TimeoutStartSec=. */ + + r = parse_sec(rvalue, &usec); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse %s= parameter, ignoring: %s", lvalue, rvalue); + return 0; + } + + /* Traditionally, these options accepted 0 to disable the timeouts. However, a timeout of 0 suggests it happens + * immediately, hence fix this to become USEC_INFINITY instead. This is in-line with how we internally handle + * all other timeouts. */ + if (usec <= 0) + usec = USEC_INFINITY; + + if (!streq(lvalue, "TimeoutStopSec")) { + s->start_timeout_defined = true; + s->timeout_start_usec = usec; + } + + if (!streq(lvalue, "TimeoutStartSec")) + s->timeout_stop_usec = usec; + + return 0; +} + +int config_parse_sec_fix_0( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + usec_t *usec = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(usec); + + /* This is pretty much like config_parse_sec(), except that this treats a time of 0 as infinity, for + * compatibility with older versions of systemd where 0 instead of infinity was used as indicator to turn off a + * timeout. */ + + r = parse_sec(rvalue, usec); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse %s= parameter, ignoring: %s", lvalue, rvalue); + return 0; + } + + if (*usec <= 0) + *usec = USEC_INFINITY; + + return 0; +} + +int config_parse_user_group( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char **user = data, *n; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) + n = NULL; + else { + _cleanup_free_ char *k = NULL; + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers in %s, ignoring: %m", rvalue); + return 0; + } + + if (!valid_user_group_name_or_id(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid user/group name or numeric ID, ignoring: %s", k); + return 0; + } + + n = k; + k = NULL; + } + + free(*user); + *user = n; + + return 0; +} + +int config_parse_user_group_strv( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char ***users = data; + Unit *u = userdata; + const char *p; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + char **empty; + + empty = new0(char*, 1); + if (!empty) + return log_oom(); + + strv_free(*users); + *users = empty; + + return 0; + } + + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL, *k = NULL; + + r = extract_first_word(&p, &word, WHITESPACE, 0); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Invalid syntax, ignoring: %s", rvalue); + break; + } + + r = unit_full_printf(u, word, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers in %s, ignoring: %m", word); + continue; + } + + if (!valid_user_group_name_or_id(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid user/group name or numeric ID, ignoring: %s", k); + continue; + } + + r = strv_push(users, k); + if (r < 0) + return log_oom(); + + k = NULL; + } + + return 0; +} + +int config_parse_busname_service( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + BusName *n = data; + int r; + Unit *x; + _cleanup_free_ char *p = NULL; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = unit_name_printf(UNIT(n), rvalue, &p); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + if (!endswith(p, ".service")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unit must be of type service, ignoring: %s", rvalue); + return 0; + } + + r = manager_load_unit(UNIT(n)->manager, p, NULL, &error, &x); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load unit %s, ignoring: %s", rvalue, bus_error_message(&error, r)); + return 0; + } + + unit_ref_set(&n->service, x); + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_bus_policy_world, bus_policy_access, BusPolicyAccess, "Failed to parse bus name policy access"); + +int config_parse_bus_policy( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ BusNamePolicy *p = NULL; + _cleanup_free_ char *id_str = NULL; + BusName *busname = data; + char *access_str; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + p = new0(BusNamePolicy, 1); + if (!p) + return log_oom(); + + if (streq(lvalue, "AllowUser")) + p->type = BUSNAME_POLICY_TYPE_USER; + else if (streq(lvalue, "AllowGroup")) + p->type = BUSNAME_POLICY_TYPE_GROUP; + else + assert_not_reached("Unknown lvalue"); + + id_str = strdup(rvalue); + if (!id_str) + return log_oom(); + + access_str = strpbrk(id_str, WHITESPACE); + if (!access_str) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid busname policy value '%s'", rvalue); + return 0; + } + + *access_str = '\0'; + access_str++; + access_str += strspn(access_str, WHITESPACE); + + p->access = bus_policy_access_from_string(access_str); + if (p->access < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid busname policy access type '%s'", access_str); + return 0; + } + + p->name = id_str; + id_str = NULL; + + LIST_PREPEND(policy, busname->policy, p); + p = NULL; + + return 0; +} + +int config_parse_working_directory( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool missing_ok; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(c); + assert(u); + + if (rvalue[0] == '-') { + missing_ok = true; + rvalue++; + } else + missing_ok = false; + + if (streq(rvalue, "~")) { + c->working_directory_home = true; + c->working_directory = mfree(c->working_directory); + } else { + _cleanup_free_ char *k = NULL; + + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers in working directory path '%s', ignoring: %m", rvalue); + return 0; + } + + path_kill_slashes(k); + + if (!utf8_is_valid(k)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, rvalue); + return 0; + } + + if (!path_is_absolute(k)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Working directory path '%s' is not absolute, ignoring.", rvalue); + return 0; + } + + free_and_replace(c->working_directory, k); + + c->working_directory_home = false; + } + + c->working_directory_missing_ok = missing_ok; + return 0; +} + +int config_parse_unit_env_file(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char ***env = data; + Unit *u = userdata; + _cleanup_free_ char *n = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment frees the list */ + *env = strv_free(*env); + return 0; + } + + r = unit_full_printf(u, rvalue, &n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + if (!path_is_absolute(n[0] == '-' ? n + 1 : n)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Path '%s' is not absolute, ignoring.", n); + return 0; + } + + r = strv_extend(env, n); + if (r < 0) + return log_oom(); + + return 0; +} + +int config_parse_environ(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + char*** env = data; + const char *word, *state; + size_t l; + _cleanup_free_ char *k = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *env = strv_free(*env); + return 0; + } + + if (u) { + r = unit_full_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + } + + if (!k) { + k = strdup(rvalue); + if (!k) + return log_oom(); + } + + FOREACH_WORD_QUOTED(word, l, k, state) { + _cleanup_free_ char *n = NULL; + char **x; + + r = cunescape_length(word, l, 0, &n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Couldn't unescape assignment, ignoring: %s", rvalue); + continue; + } + + if (!env_assignment_is_valid(n)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid environment assignment, ignoring: %s", rvalue); + continue; + } + + x = strv_env_set(*env, n); + if (!x) + return log_oom(); + + strv_free(*env); + *env = x; + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} + +int config_parse_pass_environ(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + const char *whole_rvalue = rvalue; + char*** passenv = data; + _cleanup_strv_free_ char **n = NULL; + size_t nlen = 0, nbufsize = 0; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *passenv = strv_free(*passenv); + return 0; + } + + for (;;) { + _cleanup_free_ char *word = NULL; + + r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, + "Trailing garbage in %s, ignoring: %s", lvalue, whole_rvalue); + break; + } + + if (!env_name_is_valid(word)) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Invalid environment name for %s, ignoring: %s", lvalue, word); + continue; + } + + if (!GREEDY_REALLOC(n, nbufsize, nlen + 2)) + return log_oom(); + n[nlen++] = word; + n[nlen] = NULL; + word = NULL; + } + + if (n) { + r = strv_extend_strv(passenv, n, true); + if (r < 0) + return r; + } + + return 0; +} + +int config_parse_ip_tos(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + int *ip_tos = data, x; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + x = ip_tos_from_string(rvalue); + if (x < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse IP TOS value, ignoring: %s", rvalue); + return 0; + } + + *ip_tos = x; + return 0; +} + +int config_parse_unit_condition_path( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *p = NULL; + Condition **list = data, *c; + ConditionType t = ltype; + bool trigger, negate; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *list = condition_free_list(*list); + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + r = unit_full_printf(u, rvalue, &p); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + if (!path_is_absolute(p)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Path in condition not absolute, ignoring: %s", p); + return 0; + } + + c = condition_new(t, p, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(conditions, *list, c); + return 0; +} + +int config_parse_unit_condition_string( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *s = NULL; + Condition **list = data, *c; + ConditionType t = ltype; + bool trigger, negate; + Unit *u = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *list = condition_free_list(*list); + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + r = unit_full_printf(u, rvalue, &s); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } + + c = condition_new(t, s, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(conditions, *list, c); + return 0; +} + +int config_parse_unit_condition_null( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Condition **list = data, *c; + bool trigger, negate; + int b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *list = condition_free_list(*list); + return 0; + } + + trigger = rvalue[0] == '|'; + if (trigger) + rvalue++; + + negate = rvalue[0] == '!'; + if (negate) + rvalue++; + + b = parse_boolean(rvalue); + if (b < 0) { + log_syntax(unit, LOG_ERR, filename, line, b, "Failed to parse boolean value in condition, ignoring: %s", rvalue); + return 0; + } + + if (!b) + negate = !negate; + + c = condition_new(CONDITION_NULL, NULL, trigger, negate); + if (!c) + return log_oom(); + + LIST_PREPEND(conditions, *list, c); + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_notify_access, notify_access, NotifyAccess, "Failed to parse notify access specifier"); +DEFINE_CONFIG_PARSE_ENUM(config_parse_emergency_action, emergency_action, EmergencyAction, "Failed to parse failure action specifier"); + +int config_parse_unit_requires_mounts_for( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + const char *word, *state; + size_t l; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + int r; + _cleanup_free_ char *n; + + n = strndup(word, l); + if (!n) + return log_oom(); + + if (!utf8_is_valid(n)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, rvalue); + continue; + } + + r = unit_require_mounts_for(u, n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add required mount for, ignoring: %s", rvalue); + continue; + } + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} + +int config_parse_documentation(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Unit *u = userdata; + int r; + char **a, **b; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + u->documentation = strv_free(u->documentation); + return 0; + } + + r = config_parse_unit_strv_printf(unit, filename, line, section, section_line, lvalue, ltype, + rvalue, data, userdata); + if (r < 0) + return r; + + for (a = b = u->documentation; a && *a; a++) { + + if (documentation_url_is_valid(*a)) + *(b++) = *a; + else { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid URL, ignoring: %s", *a); + free(*a); + } + } + if (b) + *b = NULL; + + return r; +} + +#ifdef HAVE_SECCOMP + +static int syscall_filter_parse_one( + const char *unit, + const char *filename, + unsigned line, + ExecContext *c, + bool invert, + const char *t, + bool warn) { + int r; + + if (t[0] == '@') { + const SyscallFilterSet *set; + const char *i; + + set = syscall_filter_set_find(t); + if (!set) { + if (warn) + log_syntax(unit, LOG_WARNING, filename, line, 0, "Don't know system call group, ignoring: %s", t); + return 0; + } + + NULSTR_FOREACH(i, set->value) { + r = syscall_filter_parse_one(unit, filename, line, c, invert, i, false); + if (r < 0) + return r; + } + } else { + int id; + + id = seccomp_syscall_resolve_name(t); + if (id == __NR_SCMP_ERROR) { + if (warn) + log_syntax(unit, LOG_WARNING, filename, line, 0, "Failed to parse system call, ignoring: %s", t); + return 0; + } + + /* If we previously wanted to forbid a syscall and now + * we want to allow it, then remove it from the list + */ + if (!invert == c->syscall_whitelist) { + r = set_put(c->syscall_filter, INT_TO_PTR(id + 1)); + if (r == 0) + return 0; + if (r < 0) + return log_oom(); + } else + (void) set_remove(c->syscall_filter, INT_TO_PTR(id + 1)); + } + + return 0; +} + +int config_parse_syscall_filter( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + Unit *u = userdata; + bool invert = false; + const char *p; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + c->syscall_filter = set_free(c->syscall_filter); + c->syscall_whitelist = false; + return 0; + } + + if (rvalue[0] == '~') { + invert = true; + rvalue++; + } + + if (!c->syscall_filter) { + c->syscall_filter = set_new(NULL); + if (!c->syscall_filter) + return log_oom(); + + if (invert) + /* Allow everything but the ones listed */ + c->syscall_whitelist = false; + else { + /* Allow nothing but the ones listed */ + c->syscall_whitelist = true; + + /* Accept default syscalls if we are on a whitelist */ + r = syscall_filter_parse_one(unit, filename, line, c, false, "@default", false); + if (r < 0) + return r; + } + } + + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL; + + r = extract_first_word(&p, &word, NULL, 0); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_WARNING, filename, line, r, "Invalid syntax, ignoring: %s", rvalue); + break; + } + + r = syscall_filter_parse_one(unit, filename, line, c, invert, word, true); + if (r < 0) + return r; + } + + return 0; +} + +int config_parse_syscall_archs( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + Set **archs = data; + const char *word, *state; + size_t l; + int r; + + if (isempty(rvalue)) { + *archs = set_free(*archs); + return 0; + } + + r = set_ensure_allocated(archs, NULL); + if (r < 0) + return log_oom(); + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + _cleanup_free_ char *t = NULL; + uint32_t a; + + t = strndup(word, l); + if (!t) + return log_oom(); + + r = seccomp_arch_from_string(t, &a); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse system call architecture, ignoring: %s", t); + continue; + } + + r = set_put(*archs, UINT32_TO_PTR(a + 1)); + if (r == 0) + continue; + if (r < 0) + return log_oom(); + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} + +int config_parse_syscall_errno( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int e; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (isempty(rvalue)) { + /* Empty assignment resets to KILL */ + c->syscall_errno = 0; + return 0; + } + + e = errno_from_name(rvalue); + if (e < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse error number, ignoring: %s", rvalue); + return 0; + } + + c->syscall_errno = e; + return 0; +} + +int config_parse_address_families( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + bool invert = false; + const char *word, *state; + size_t l; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + c->address_families = set_free(c->address_families); + c->address_families_whitelist = false; + return 0; + } + + if (rvalue[0] == '~') { + invert = true; + rvalue++; + } + + if (!c->address_families) { + c->address_families = set_new(NULL); + if (!c->address_families) + return log_oom(); + + c->address_families_whitelist = !invert; + } + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + _cleanup_free_ char *t = NULL; + int af; + + t = strndup(word, l); + if (!t) + return log_oom(); + + af = af_from_name(t); + if (af <= 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse address family, ignoring: %s", t); + continue; + } + + /* If we previously wanted to forbid an address family and now + * we want to allow it, then remove it from the list + */ + if (!invert == c->address_families_whitelist) { + r = set_put(c->address_families, INT_TO_PTR(af)); + if (r == 0) + continue; + if (r < 0) + return log_oom(); + } else + set_remove(c->address_families, INT_TO_PTR(af)); + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} +#endif + +int config_parse_unit_slice( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *k = NULL; + Unit *u = userdata, *slice = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(u); + + r = unit_name_printf(u, rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s. Ignoring.", rvalue); + return 0; + } + + r = manager_load_unit(u->manager, k, NULL, NULL, &slice); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load slice unit %s. Ignoring.", k); + return 0; + } + + r = unit_set_slice(u, slice); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to assign slice %s to unit %s. Ignoring.", slice->id, u->id); + return 0; + } + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_device_policy, cgroup_device_policy, CGroupDevicePolicy, "Failed to parse device policy"); + +int config_parse_cpu_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *weight = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = cg_weight_parse(rvalue, weight); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU weight '%s' invalid. Ignoring.", rvalue); + return 0; + } + + return 0; +} + +int config_parse_cpu_shares( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *shares = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = cg_cpu_shares_parse(rvalue, shares); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU shares '%s' invalid. Ignoring.", rvalue); + return 0; + } + + return 0; +} + +int config_parse_cpu_quota( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + CGroupContext *c = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (isempty(rvalue)) { + c->cpu_quota_per_sec_usec = USEC_INFINITY; + return 0; + } + + r = parse_percent_unbounded(rvalue); + if (r <= 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU quota '%s' invalid. Ignoring.", rvalue); + return 0; + } + + c->cpu_quota_per_sec_usec = ((usec_t) r * USEC_PER_SEC) / 100U; + return 0; +} + +int config_parse_memory_limit( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + CGroupContext *c = data; + uint64_t bytes = CGROUP_LIMIT_MAX; + int r; + + if (!isempty(rvalue) && !streq(rvalue, "infinity")) { + + r = parse_percent(rvalue); + if (r < 0) { + r = parse_size(rvalue, 1024, &bytes); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Memory limit '%s' invalid. Ignoring.", rvalue); + return 0; + } + } else + bytes = physical_memory_scale(r, 100U); + + if (bytes <= 0 || bytes >= UINT64_MAX) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Memory limit '%s' out of range. Ignoring.", rvalue); + return 0; + } + } + + if (streq(lvalue, "MemoryLow")) + c->memory_low = bytes; + else if (streq(lvalue, "MemoryHigh")) + c->memory_high = bytes; + else if (streq(lvalue, "MemoryMax")) + c->memory_max = bytes; + else if (streq(lvalue, "MemorySwapMax")) + c->memory_swap_max = bytes; + else if (streq(lvalue, "MemoryLimit")) + c->memory_limit = bytes; + else + return -EINVAL; + + return 0; +} + +int config_parse_tasks_max( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *tasks_max = data, v; + Unit *u = userdata; + int r; + + if (isempty(rvalue)) { + *tasks_max = u->manager->default_tasks_max; + return 0; + } + + if (streq(rvalue, "infinity")) { + *tasks_max = CGROUP_LIMIT_MAX; + return 0; + } + + r = parse_percent(rvalue); + if (r < 0) { + r = safe_atou64(rvalue, &v); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Maximum tasks value '%s' invalid. Ignoring.", rvalue); + return 0; + } + } else + v = system_tasks_max_scale(r, 100U); + + if (v <= 0 || v >= UINT64_MAX) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Maximum tasks value '%s' out of range. Ignoring.", rvalue); + return 0; + } + + *tasks_max = v; + return 0; +} + +int config_parse_device_allow( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *path = NULL, *t = NULL; + CGroupContext *c = data; + CGroupDeviceAllow *a; + const char *m = NULL; + size_t n; + int r; + + if (isempty(rvalue)) { + while (c->device_allow) + cgroup_context_free_device_allow(c, c->device_allow); + + return 0; + } + + r = unit_full_printf(userdata, rvalue, &t); + if(r < 0) { + log_syntax(unit, LOG_WARNING, filename, line, r, + "Failed to resolve specifiers in %s, ignoring: %m", + rvalue); + } + + n = strcspn(t, WHITESPACE); + + path = strndup(t, n); + if (!path) + return log_oom(); + + if (!is_deviceallow_pattern(path)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); + return 0; + } + + m = t + n + strspn(t + n, WHITESPACE); + if (isempty(m)) + m = "rwm"; + + if (!in_charset(m, "rwm")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device rights '%s'. Ignoring.", m); + return 0; + } + + a = new0(CGroupDeviceAllow, 1); + if (!a) + return log_oom(); + + a->path = path; + path = NULL; + a->r = !!strchr(m, 'r'); + a->w = !!strchr(m, 'w'); + a->m = !!strchr(m, 'm'); + + LIST_PREPEND(device_allow, c->device_allow, a); + return 0; +} + +int config_parse_io_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *weight = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = cg_weight_parse(rvalue, weight); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "IO weight '%s' invalid. Ignoring.", rvalue); + return 0; + } + + return 0; +} + +int config_parse_io_device_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *path = NULL; + CGroupIODeviceWeight *w; + CGroupContext *c = data; + const char *weight; + uint64_t u; + size_t n; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (isempty(rvalue)) { + while (c->io_device_weights) + cgroup_context_free_io_device_weight(c, c->io_device_weights); + + return 0; + } + + n = strcspn(rvalue, WHITESPACE); + weight = rvalue + n; + weight += strspn(weight, WHITESPACE); + + if (isempty(weight)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected block device and device weight. Ignoring."); + return 0; + } + + path = strndup(rvalue, n); + if (!path) + return log_oom(); + + if (!path_startswith(path, "/dev")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); + return 0; + } + + r = cg_weight_parse(weight, &u); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "IO weight '%s' invalid. Ignoring.", weight); + return 0; + } + + assert(u != CGROUP_WEIGHT_INVALID); + + w = new0(CGroupIODeviceWeight, 1); + if (!w) + return log_oom(); + + w->path = path; + path = NULL; + + w->weight = u; + + LIST_PREPEND(device_weights, c->io_device_weights, w); + return 0; +} + +int config_parse_io_limit( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *path = NULL; + CGroupIODeviceLimit *l = NULL, *t; + CGroupContext *c = data; + CGroupIOLimitType type; + const char *limit; + uint64_t num; + size_t n; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + type = cgroup_io_limit_type_from_string(lvalue); + assert(type >= 0); + + if (isempty(rvalue)) { + LIST_FOREACH(device_limits, l, c->io_device_limits) + l->limits[type] = cgroup_io_limit_defaults[type]; + return 0; + } + + n = strcspn(rvalue, WHITESPACE); + limit = rvalue + n; + limit += strspn(limit, WHITESPACE); + + if (!*limit) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected space separated pair of device node and bandwidth. Ignoring."); + return 0; + } + + path = strndup(rvalue, n); + if (!path) + return log_oom(); + + if (!path_startswith(path, "/dev")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); + return 0; + } + + if (streq("infinity", limit)) { + num = CGROUP_LIMIT_MAX; + } else { + r = parse_size(limit, 1000, &num); + if (r < 0 || num <= 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "IO Limit '%s' invalid. Ignoring.", rvalue); + return 0; + } + } + + LIST_FOREACH(device_limits, t, c->io_device_limits) { + if (path_equal(path, t->path)) { + l = t; + break; + } + } + + if (!l) { + CGroupIOLimitType ttype; + + l = new0(CGroupIODeviceLimit, 1); + if (!l) + return log_oom(); + + l->path = path; + path = NULL; + for (ttype = 0; ttype < _CGROUP_IO_LIMIT_TYPE_MAX; ttype++) + l->limits[ttype] = cgroup_io_limit_defaults[ttype]; + + LIST_PREPEND(device_limits, c->io_device_limits, l); + } + + l->limits[type] = num; + + return 0; +} + +int config_parse_blockio_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + uint64_t *weight = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = cg_blkio_weight_parse(rvalue, weight); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Block IO weight '%s' invalid. Ignoring.", rvalue); + return 0; + } + + return 0; +} + +int config_parse_blockio_device_weight( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *path = NULL; + CGroupBlockIODeviceWeight *w; + CGroupContext *c = data; + const char *weight; + uint64_t u; + size_t n; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (isempty(rvalue)) { + while (c->blockio_device_weights) + cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); + + return 0; + } + + n = strcspn(rvalue, WHITESPACE); + weight = rvalue + n; + weight += strspn(weight, WHITESPACE); + + if (isempty(weight)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected block device and device weight. Ignoring."); + return 0; + } + + path = strndup(rvalue, n); + if (!path) + return log_oom(); + + if (!path_startswith(path, "/dev")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); + return 0; + } + + r = cg_blkio_weight_parse(weight, &u); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Block IO weight '%s' invalid. Ignoring.", weight); + return 0; + } + + assert(u != CGROUP_BLKIO_WEIGHT_INVALID); + + w = new0(CGroupBlockIODeviceWeight, 1); + if (!w) + return log_oom(); + + w->path = path; + path = NULL; + + w->weight = u; + + LIST_PREPEND(device_weights, c->blockio_device_weights, w); + return 0; +} + +int config_parse_blockio_bandwidth( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + _cleanup_free_ char *path = NULL; + CGroupBlockIODeviceBandwidth *b = NULL, *t; + CGroupContext *c = data; + const char *bandwidth; + uint64_t bytes; + bool read; + size_t n; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + read = streq("BlockIOReadBandwidth", lvalue); + + if (isempty(rvalue)) { + LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { + b->rbps = CGROUP_LIMIT_MAX; + b->wbps = CGROUP_LIMIT_MAX; + } + return 0; + } + + n = strcspn(rvalue, WHITESPACE); + bandwidth = rvalue + n; + bandwidth += strspn(bandwidth, WHITESPACE); + + if (!*bandwidth) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected space separated pair of device node and bandwidth. Ignoring."); + return 0; + } + + path = strndup(rvalue, n); + if (!path) + return log_oom(); + + if (!path_startswith(path, "/dev")) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); + return 0; + } + + r = parse_size(bandwidth, 1000, &bytes); + if (r < 0 || bytes <= 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Block IO Bandwidth '%s' invalid. Ignoring.", rvalue); + return 0; + } + + LIST_FOREACH(device_bandwidths, t, c->blockio_device_bandwidths) { + if (path_equal(path, t->path)) { + b = t; + break; + } + } + + if (!t) { + b = new0(CGroupBlockIODeviceBandwidth, 1); + if (!b) + return log_oom(); + + b->path = path; + path = NULL; + b->rbps = CGROUP_LIMIT_MAX; + b->wbps = CGROUP_LIMIT_MAX; + + LIST_PREPEND(device_bandwidths, c->blockio_device_bandwidths, b); + } + + if (read) + b->rbps = bytes; + else + b->wbps = bytes; + + return 0; +} + +DEFINE_CONFIG_PARSE_ENUM(config_parse_job_mode, job_mode, JobMode, "Failed to parse job mode"); + +int config_parse_job_mode_isolate( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + JobMode *m = data; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + r = parse_boolean(rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse boolean, ignoring: %s", rvalue); + return 0; + } + + *m = r ? JOB_ISOLATE : JOB_REPLACE; + return 0; +} + +int config_parse_runtime_directory( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char***rt = data; + Unit *u = userdata; + const char *word, *state; + size_t l; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *rt = strv_free(*rt); + return 0; + } + + FOREACH_WORD_QUOTED(word, l, rvalue, state) { + _cleanup_free_ char *t = NULL, *n = NULL; + + t = strndup(word, l); + if (!t) + return log_oom(); + + r = unit_name_printf(u, t, &n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); + continue; + } + + if (!filename_is_valid(n)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Runtime directory is not valid, ignoring assignment: %s", rvalue); + continue; + } + + r = strv_push(rt, n); + if (r < 0) + return log_oom(); + + n = NULL; + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} + +int config_parse_set_status( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + size_t l; + const char *word, *state; + int r; + ExitStatusSet *status_set = data; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + /* Empty assignment resets the list */ + if (isempty(rvalue)) { + exit_status_set_free(status_set); + return 0; + } + + FOREACH_WORD(word, l, rvalue, state) { + _cleanup_free_ char *temp; + int val; + Set **set; + + temp = strndup(word, l); + if (!temp) + return log_oom(); + + r = safe_atoi(temp, &val); + if (r < 0) { + val = signal_from_string_try_harder(temp); + + if (val <= 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse value, ignoring: %s", word); + continue; + } + set = &status_set->signal; + } else { + if (val < 0 || val > 255) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Value %d is outside range 0-255, ignoring", val); + continue; + } + set = &status_set->status; + } + + r = set_ensure_allocated(set, NULL); + if (r < 0) + return log_oom(); + + r = set_put(*set, INT_TO_PTR(val)); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Unable to store: %s", word); + return r; + } + } + if (!isempty(state)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); + + return 0; +} + +int config_parse_namespace_path_strv( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + char*** sv = data; + const char *prev; + const char *cur; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + if (isempty(rvalue)) { + /* Empty assignment resets the list */ + *sv = strv_free(*sv); + return 0; + } + + prev = cur = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL; + int offset; + + r = extract_first_word(&cur, &word, NULL, EXTRACT_QUOTES); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Trailing garbage, ignoring: %s", prev); + return 0; + } + + if (!utf8_is_valid(word)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, word); + prev = cur; + continue; + } + + offset = word[0] == '-'; + if (!path_is_absolute(word + offset)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Not an absolute path, ignoring: %s", word); + prev = cur; + continue; + } + + path_kill_slashes(word + offset); + + r = strv_push(sv, word); + if (r < 0) + return log_oom(); + + prev = cur; + word = NULL; + } + + return 0; +} + +int config_parse_no_new_privileges( + const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int k; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + k = parse_boolean(rvalue); + if (k < 0) { + log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse boolean value, ignoring: %s", rvalue); + return 0; + } + + c->no_new_privileges = k; + c->no_new_privileges_set = true; + + return 0; +} + +int config_parse_protect_home( + const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int k; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + /* Our enum shall be a superset of booleans, hence first try + * to parse as boolean, and then as enum */ + + k = parse_boolean(rvalue); + if (k > 0) + c->protect_home = PROTECT_HOME_YES; + else if (k == 0) + c->protect_home = PROTECT_HOME_NO; + else { + ProtectHome h; + + h = protect_home_from_string(rvalue); + if (h < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse protect home value, ignoring: %s", rvalue); + return 0; + } + + c->protect_home = h; + } + + return 0; +} + +int config_parse_protect_system( + const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + ExecContext *c = data; + int k; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + /* Our enum shall be a superset of booleans, hence first try + * to parse as boolean, and then as enum */ + + k = parse_boolean(rvalue); + if (k > 0) + c->protect_system = PROTECT_SYSTEM_YES; + else if (k == 0) + c->protect_system = PROTECT_SYSTEM_NO; + else { + ProtectSystem s; + + s = protect_system_from_string(rvalue); + if (s < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse protect system value, ignoring: %s", rvalue); + return 0; + } + + c->protect_system = s; + } + + return 0; +} + +#define FOLLOW_MAX 8 + +static int open_follow(char **filename, FILE **_f, Set *names, char **_final) { + char *id = NULL; + unsigned c = 0; + int fd, r; + FILE *f; + + assert(filename); + assert(*filename); + assert(_f); + assert(names); + + /* This will update the filename pointer if the loaded file is + * reached by a symlink. The old string will be freed. */ + + for (;;) { + char *target, *name; + + if (c++ >= FOLLOW_MAX) + return -ELOOP; + + path_kill_slashes(*filename); + + /* Add the file name we are currently looking at to + * the names of this unit, but only if it is a valid + * unit name. */ + name = basename(*filename); + if (unit_name_is_valid(name, UNIT_NAME_ANY)) { + + id = set_get(names, name); + if (!id) { + id = strdup(name); + if (!id) + return -ENOMEM; + + r = set_consume(names, id); + if (r < 0) + return r; + } + } + + /* Try to open the file name, but don't if its a symlink */ + fd = open(*filename, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); + if (fd >= 0) + break; + + if (errno != ELOOP) + return -errno; + + /* Hmm, so this is a symlink. Let's read the name, and follow it manually */ + r = readlink_and_make_absolute(*filename, &target); + if (r < 0) + return r; + + free(*filename); + *filename = target; + } + + f = fdopen(fd, "re"); + if (!f) { + safe_close(fd); + return -errno; + } + + *_f = f; + *_final = id; + + return 0; +} + +static int merge_by_names(Unit **u, Set *names, const char *id) { + char *k; + int r; + + assert(u); + assert(*u); + assert(names); + + /* Let's try to add in all symlink names we found */ + while ((k = set_steal_first(names))) { + + /* First try to merge in the other name into our + * unit */ + r = unit_merge_by_name(*u, k); + if (r < 0) { + Unit *other; + + /* Hmm, we couldn't merge the other unit into + * ours? Then let's try it the other way + * round */ + + /* If the symlink name we are looking at is unit template, then + we must search for instance of this template */ + if (unit_name_is_valid(k, UNIT_NAME_TEMPLATE) && (*u)->instance) { + _cleanup_free_ char *instance = NULL; + + r = unit_name_replace_instance(k, (*u)->instance, &instance); + if (r < 0) + return r; + + other = manager_get_unit((*u)->manager, instance); + } else + other = manager_get_unit((*u)->manager, k); + + free(k); + + if (other) { + r = unit_merge(other, *u); + if (r >= 0) { + *u = other; + return merge_by_names(u, names, NULL); + } + } + + return r; + } + + if (id == k) + unit_choose_id(*u, id); + + free(k); + } + + return 0; +} + +static int load_from_path(Unit *u, const char *path) { + _cleanup_set_free_free_ Set *symlink_names = NULL; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_free_ char *filename = NULL; + char *id = NULL; + Unit *merged; + struct stat st; + int r; + + assert(u); + assert(path); + + symlink_names = set_new(&string_hash_ops); + if (!symlink_names) + return -ENOMEM; + + if (path_is_absolute(path)) { + + filename = strdup(path); + if (!filename) + return -ENOMEM; + + r = open_follow(&filename, &f, symlink_names, &id); + if (r < 0) { + filename = mfree(filename); + if (r != -ENOENT) + return r; + } + + } else { + char **p; + + STRV_FOREACH(p, u->manager->lookup_paths.search_path) { + + /* Instead of opening the path right away, we manually + * follow all symlinks and add their name to our unit + * name set while doing so */ + filename = path_make_absolute(path, *p); + if (!filename) + return -ENOMEM; + + if (u->manager->unit_path_cache && + !set_get(u->manager->unit_path_cache, filename)) + r = -ENOENT; + else + r = open_follow(&filename, &f, symlink_names, &id); + if (r >= 0) + break; + filename = mfree(filename); + + /* ENOENT means that the file is missing or is a dangling symlink. + * ENOTDIR means that one of paths we expect to be is a directory + * is not a directory, we should just ignore that. + * EACCES means that the directory or file permissions are wrong. + */ + if (r == -EACCES) + log_debug_errno(r, "Cannot access \"%s\": %m", filename); + else if (!IN_SET(r, -ENOENT, -ENOTDIR)) + return r; + + /* Empty the symlink names for the next run */ + set_clear_free(symlink_names); + } + } + + if (!filename) + /* Hmm, no suitable file found? */ + return 0; + + if (!unit_type_may_alias(u->type) && set_size(symlink_names) > 1) { + log_unit_warning(u, "Unit type of %s does not support alias names, refusing loading via symlink.", u->id); + return -ELOOP; + } + + merged = u; + r = merge_by_names(&merged, symlink_names, id); + if (r < 0) + return r; + + if (merged != u) { + u->load_state = UNIT_MERGED; + return 0; + } + + if (fstat(fileno(f), &st) < 0) + return -errno; + + if (null_or_empty(&st)) { + u->load_state = UNIT_MASKED; + u->fragment_mtime = 0; + } else { + u->load_state = UNIT_LOADED; + u->fragment_mtime = timespec_load(&st.st_mtim); + + /* Now, parse the file contents */ + r = config_parse(u->id, filename, f, + UNIT_VTABLE(u)->sections, + config_item_perf_lookup, load_fragment_gperf_lookup, + false, true, false, u); + if (r < 0) + return r; + } + + free(u->fragment_path); + u->fragment_path = filename; + filename = NULL; + + if (u->source_path) { + if (stat(u->source_path, &st) >= 0) + u->source_mtime = timespec_load(&st.st_mtim); + else + u->source_mtime = 0; + } + + return 0; +} + +int unit_load_fragment(Unit *u) { + int r; + Iterator i; + const char *t; + + assert(u); + assert(u->load_state == UNIT_STUB); + assert(u->id); + + if (u->transient) { + u->load_state = UNIT_LOADED; + return 0; + } + + /* First, try to find the unit under its id. We always look + * for unit files in the default directories, to make it easy + * to override things by placing things in /etc/systemd/system */ + r = load_from_path(u, u->id); + if (r < 0) + return r; + + /* Try to find an alias we can load this with */ + if (u->load_state == UNIT_STUB) { + SET_FOREACH(t, u->names, i) { + + if (t == u->id) + continue; + + r = load_from_path(u, t); + if (r < 0) + return r; + + if (u->load_state != UNIT_STUB) + break; + } + } + + /* And now, try looking for it under the suggested (originally linked) path */ + if (u->load_state == UNIT_STUB && u->fragment_path) { + + r = load_from_path(u, u->fragment_path); + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) + /* Hmm, this didn't work? Then let's get rid + * of the fragment path stored for us, so that + * we don't point to an invalid location. */ + u->fragment_path = mfree(u->fragment_path); + } + + /* Look for a template */ + if (u->load_state == UNIT_STUB && u->instance) { + _cleanup_free_ char *k = NULL; + + r = unit_name_template(u->id, &k); + if (r < 0) + return r; + + r = load_from_path(u, k); + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) { + SET_FOREACH(t, u->names, i) { + _cleanup_free_ char *z = NULL; + + if (t == u->id) + continue; + + r = unit_name_template(t, &z); + if (r < 0) + return r; + + r = load_from_path(u, z); + if (r < 0) + return r; + + if (u->load_state != UNIT_STUB) + break; + } + } + } + + return 0; +} + +void unit_dump_config_items(FILE *f) { + static const struct { + const ConfigParserCallback callback; + const char *rvalue; + } table[] = { +#if !defined(HAVE_SYSV_COMPAT) || !defined(HAVE_SECCOMP) || !defined(HAVE_PAM) || !defined(HAVE_SELINUX) || !defined(HAVE_SMACK) || !defined(HAVE_APPARMOR) + { config_parse_warn_compat, "NOTSUPPORTED" }, +#endif + { config_parse_int, "INTEGER" }, + { config_parse_unsigned, "UNSIGNED" }, + { config_parse_iec_size, "SIZE" }, + { config_parse_iec_uint64, "SIZE" }, + { config_parse_si_size, "SIZE" }, + { config_parse_bool, "BOOLEAN" }, + { config_parse_string, "STRING" }, + { config_parse_path, "PATH" }, + { config_parse_unit_path_printf, "PATH" }, + { config_parse_strv, "STRING [...]" }, + { config_parse_exec_nice, "NICE" }, + { config_parse_exec_oom_score_adjust, "OOMSCOREADJUST" }, + { config_parse_exec_io_class, "IOCLASS" }, + { config_parse_exec_io_priority, "IOPRIORITY" }, + { config_parse_exec_cpu_sched_policy, "CPUSCHEDPOLICY" }, + { config_parse_exec_cpu_sched_prio, "CPUSCHEDPRIO" }, + { config_parse_exec_cpu_affinity, "CPUAFFINITY" }, + { config_parse_mode, "MODE" }, + { config_parse_unit_env_file, "FILE" }, + { config_parse_exec_output, "OUTPUT" }, + { config_parse_exec_input, "INPUT" }, + { config_parse_log_facility, "FACILITY" }, + { config_parse_log_level, "LEVEL" }, + { config_parse_exec_secure_bits, "SECUREBITS" }, + { config_parse_capability_set, "BOUNDINGSET" }, + { config_parse_limit, "LIMIT" }, + { config_parse_unit_deps, "UNIT [...]" }, + { config_parse_exec, "PATH [ARGUMENT [...]]" }, + { config_parse_service_type, "SERVICETYPE" }, + { config_parse_service_restart, "SERVICERESTART" }, +#ifdef HAVE_SYSV_COMPAT + { config_parse_sysv_priority, "SYSVPRIORITY" }, +#endif + { config_parse_kill_mode, "KILLMODE" }, + { config_parse_signal, "SIGNAL" }, + { config_parse_socket_listen, "SOCKET [...]" }, + { config_parse_socket_bind, "SOCKETBIND" }, + { config_parse_socket_bindtodevice, "NETWORKINTERFACE" }, + { config_parse_sec, "SECONDS" }, + { config_parse_nsec, "NANOSECONDS" }, + { config_parse_namespace_path_strv, "PATH [...]" }, + { config_parse_unit_requires_mounts_for, "PATH [...]" }, + { config_parse_exec_mount_flags, "MOUNTFLAG [...]" }, + { config_parse_unit_string_printf, "STRING" }, + { config_parse_trigger_unit, "UNIT" }, + { config_parse_timer, "TIMER" }, + { config_parse_path_spec, "PATH" }, + { config_parse_notify_access, "ACCESS" }, + { config_parse_ip_tos, "TOS" }, + { config_parse_unit_condition_path, "CONDITION" }, + { config_parse_unit_condition_string, "CONDITION" }, + { config_parse_unit_condition_null, "CONDITION" }, + { config_parse_unit_slice, "SLICE" }, + { config_parse_documentation, "URL" }, + { config_parse_service_timeout, "SECONDS" }, + { config_parse_emergency_action, "ACTION" }, + { config_parse_set_status, "STATUS" }, + { config_parse_service_sockets, "SOCKETS" }, + { config_parse_environ, "ENVIRON" }, +#ifdef HAVE_SECCOMP + { config_parse_syscall_filter, "SYSCALLS" }, + { config_parse_syscall_archs, "ARCHS" }, + { config_parse_syscall_errno, "ERRNO" }, + { config_parse_address_families, "FAMILIES" }, +#endif + { config_parse_cpu_shares, "SHARES" }, + { config_parse_cpu_weight, "WEIGHT" }, + { config_parse_memory_limit, "LIMIT" }, + { config_parse_device_allow, "DEVICE" }, + { config_parse_device_policy, "POLICY" }, + { config_parse_io_limit, "LIMIT" }, + { config_parse_io_weight, "WEIGHT" }, + { config_parse_io_device_weight, "DEVICEWEIGHT" }, + { config_parse_blockio_bandwidth, "BANDWIDTH" }, + { config_parse_blockio_weight, "WEIGHT" }, + { config_parse_blockio_device_weight, "DEVICEWEIGHT" }, + { config_parse_long, "LONG" }, + { config_parse_socket_service, "SERVICE" }, +#ifdef HAVE_SELINUX + { config_parse_exec_selinux_context, "LABEL" }, +#endif + { config_parse_job_mode, "MODE" }, + { config_parse_job_mode_isolate, "BOOLEAN" }, + { config_parse_personality, "PERSONALITY" }, + }; + + const char *prev = NULL; + const char *i; + + assert(f); + + NULSTR_FOREACH(i, load_fragment_gperf_nulstr) { + const char *rvalue = "OTHER", *lvalue; + unsigned j; + size_t prefix_len; + const char *dot; + const ConfigPerfItem *p; + + assert_se(p = load_fragment_gperf_lookup(i, strlen(i))); + + dot = strchr(i, '.'); + lvalue = dot ? dot + 1 : i; + prefix_len = dot-i; + + if (dot) + if (!prev || !strneq(prev, i, prefix_len+1)) { + if (prev) + fputc('\n', f); + + fprintf(f, "[%.*s]\n", (int) prefix_len, i); + } + + for (j = 0; j < ELEMENTSOF(table); j++) + if (p->parse == table[j].callback) { + rvalue = table[j].rvalue; + break; + } + + fprintf(f, "%s=%s\n", lvalue, rvalue); + prev = i; + } +} diff --git a/src/grp-system/libcore/src/locale-setup.c b/src/grp-system/libcore/src/locale-setup.c new file mode 100644 index 0000000000..1794b474b8 --- /dev/null +++ b/src/grp-system/libcore/src/locale-setup.c @@ -0,0 +1,125 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdlib.h> + +#include "systemd-basic/env-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/locale-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/util.h" +#include "systemd-basic/virt.h" + +#include "locale-setup.h" + +int locale_setup(char ***environment) { + char **add; + char *variables[_VARIABLE_LC_MAX] = {}; + int r = 0, i; + + if (detect_container() <= 0) { + r = parse_env_file("/proc/cmdline", WHITESPACE, + "locale.LANG", &variables[VARIABLE_LANG], + "locale.LANGUAGE", &variables[VARIABLE_LANGUAGE], + "locale.LC_CTYPE", &variables[VARIABLE_LC_CTYPE], + "locale.LC_NUMERIC", &variables[VARIABLE_LC_NUMERIC], + "locale.LC_TIME", &variables[VARIABLE_LC_TIME], + "locale.LC_COLLATE", &variables[VARIABLE_LC_COLLATE], + "locale.LC_MONETARY", &variables[VARIABLE_LC_MONETARY], + "locale.LC_MESSAGES", &variables[VARIABLE_LC_MESSAGES], + "locale.LC_PAPER", &variables[VARIABLE_LC_PAPER], + "locale.LC_NAME", &variables[VARIABLE_LC_NAME], + "locale.LC_ADDRESS", &variables[VARIABLE_LC_ADDRESS], + "locale.LC_TELEPHONE", &variables[VARIABLE_LC_TELEPHONE], + "locale.LC_MEASUREMENT", &variables[VARIABLE_LC_MEASUREMENT], + "locale.LC_IDENTIFICATION", &variables[VARIABLE_LC_IDENTIFICATION], + NULL); + + if (r < 0 && r != -ENOENT) + log_warning_errno(r, "Failed to read /proc/cmdline: %m"); + } + + /* Hmm, nothing set on the kernel cmd line? Then let's + * try /etc/locale.conf */ + if (r <= 0) { + r = parse_env_file("/etc/locale.conf", NEWLINE, + "LANG", &variables[VARIABLE_LANG], + "LANGUAGE", &variables[VARIABLE_LANGUAGE], + "LC_CTYPE", &variables[VARIABLE_LC_CTYPE], + "LC_NUMERIC", &variables[VARIABLE_LC_NUMERIC], + "LC_TIME", &variables[VARIABLE_LC_TIME], + "LC_COLLATE", &variables[VARIABLE_LC_COLLATE], + "LC_MONETARY", &variables[VARIABLE_LC_MONETARY], + "LC_MESSAGES", &variables[VARIABLE_LC_MESSAGES], + "LC_PAPER", &variables[VARIABLE_LC_PAPER], + "LC_NAME", &variables[VARIABLE_LC_NAME], + "LC_ADDRESS", &variables[VARIABLE_LC_ADDRESS], + "LC_TELEPHONE", &variables[VARIABLE_LC_TELEPHONE], + "LC_MEASUREMENT", &variables[VARIABLE_LC_MEASUREMENT], + "LC_IDENTIFICATION", &variables[VARIABLE_LC_IDENTIFICATION], + NULL); + + if (r < 0 && r != -ENOENT) + log_warning_errno(r, "Failed to read /etc/locale.conf: %m"); + } + + add = NULL; + for (i = 0; i < _VARIABLE_LC_MAX; i++) { + char *s; + + if (!variables[i]) + continue; + + s = strjoin(locale_variable_to_string(i), "=", variables[i], NULL); + if (!s) { + r = -ENOMEM; + goto finish; + } + + if (strv_consume(&add, s) < 0) { + r = -ENOMEM; + goto finish; + } + } + + if (!strv_isempty(add)) { + char **e; + + e = strv_env_merge(2, *environment, add); + if (!e) { + r = -ENOMEM; + goto finish; + } + + strv_free(*environment); + *environment = e; + } + + r = 0; + +finish: + strv_free(add); + + for (i = 0; i < _VARIABLE_LC_MAX; i++) + free(variables[i]); + + return r; +} diff --git a/src/grp-system/libcore/src/locale-setup.h b/src/grp-system/libcore/src/locale-setup.h new file mode 100644 index 0000000000..3b97497afe --- /dev/null +++ b/src/grp-system/libcore/src/locale-setup.h @@ -0,0 +1,22 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +int locale_setup(char ***environment); diff --git a/src/grp-system/libcore/src/loopback-setup.c b/src/grp-system/libcore/src/loopback-setup.c new file mode 100644 index 0000000000..8eb6251bd7 --- /dev/null +++ b/src/grp-system/libcore/src/loopback-setup.c @@ -0,0 +1,89 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <net/if.h> +#include <stdlib.h> + +#include "core/loopback-setup.h" +#include "sd-netlink/netlink-util.h" +#include "systemd-basic/missing.h" +#include "systemd-staging/sd-netlink.h" + +static int start_loopback(sd_netlink *rtnl) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + r = sd_rtnl_message_new_link(rtnl, &req, RTM_SETLINK, LOOPBACK_IFINDEX); + if (r < 0) + return r; + + r = sd_rtnl_message_link_set_flags(req, IFF_UP, IFF_UP); + if (r < 0) + return r; + + r = sd_netlink_call(rtnl, req, 0, NULL); + if (r < 0) + return r; + + return 0; +} + +static bool check_loopback(sd_netlink *rtnl) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL, *reply = NULL; + unsigned flags; + int r; + + r = sd_rtnl_message_new_link(rtnl, &req, RTM_GETLINK, LOOPBACK_IFINDEX); + if (r < 0) + return false; + + r = sd_netlink_call(rtnl, req, 0, &reply); + if (r < 0) + return false; + + r = sd_rtnl_message_link_get_flags(reply, &flags); + if (r < 0) + return false; + + return flags & IFF_UP; +} + +int loopback_setup(void) { + _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL; + int r; + + r = sd_netlink_open(&rtnl); + if (r < 0) + return r; + + r = start_loopback(rtnl); + if (r < 0) { + + /* If we lack the permissions to configure the + * loopback device, but we find it to be already + * configured, let's exit cleanly, in order to + * supported unprivileged containers. */ + if (r == -EPERM && check_loopback(rtnl)) + return 0; + + return log_warning_errno(r, "Failed to configure loopback device: %m"); + } + + return 0; +} diff --git a/src/grp-system/libcore/src/manager.c b/src/grp-system/libcore/src/manager.c new file mode 100644 index 0000000000..bad8cf0dbb --- /dev/null +++ b/src/grp-system/libcore/src/manager.c @@ -0,0 +1,3577 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dirent.h> +#include <errno.h> +#include <fcntl.h> +#include <signal.h> +#include <string.h> +#include <sys/epoll.h> +#include <sys/inotify.h> +#include <sys/ioctl.h> +#include <sys/reboot.h> +#include <sys/timerfd.h> +#include <sys/wait.h> +#include <unistd.h> + +#include <linux/kd.h> + +#ifdef HAVE_AUDIT +#include <libaudit.h> +#endif + +#include <systemd/sd-daemon.h> +#include <systemd/sd-messages.h> + +#include "core/dbus-manager.h" +#include "core/manager.h" +#include "sd-bus/bus-common-errors.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-kernel.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/dirent-util.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/hashmap.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/ratelimit.h" +#include "systemd-basic/rm-rf.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/time-util.h" +#include "systemd-basic/umask-util.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/util.h" +#include "systemd-basic/virt.h" +#include "systemd-shared/boot-timestamps.h" +#include "systemd-shared/clean-ipc.h" +#include "systemd-shared/path-lookup.h" +#include "systemd-shared/watchdog.h" + +#include "audit-fd.h" +#include "dbus-job.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "locale-setup.h" +#include "transaction.h" + +#define NOTIFY_RCVBUF_SIZE (8*1024*1024) +#define CGROUPS_AGENT_RCVBUF_SIZE (8*1024*1024) + +/* Initial delay and the interval for printing status messages about running jobs */ +#define JOBS_IN_PROGRESS_WAIT_USEC (5*USEC_PER_SEC) +#define JOBS_IN_PROGRESS_PERIOD_USEC (USEC_PER_SEC / 3) +#define JOBS_IN_PROGRESS_PERIOD_DIVISOR 3 + +static int manager_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_cgroups_agent_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_time_change_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_idle_pipe_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_user_lookup_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int manager_dispatch_jobs_in_progress(sd_event_source *source, usec_t usec, void *userdata); +static int manager_dispatch_run_queue(sd_event_source *source, void *userdata); +static int manager_run_generators(Manager *m); + +static void manager_watch_jobs_in_progress(Manager *m) { + usec_t next; + int r; + + assert(m); + + if (m->jobs_in_progress_event_source) + return; + + next = now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_WAIT_USEC; + r = sd_event_add_time( + m->event, + &m->jobs_in_progress_event_source, + CLOCK_MONOTONIC, + next, 0, + manager_dispatch_jobs_in_progress, m); + if (r < 0) + return; + + (void) sd_event_source_set_description(m->jobs_in_progress_event_source, "manager-jobs-in-progress"); +} + +#define CYLON_BUFFER_EXTRA (2*(sizeof(ANSI_RED)-1) + sizeof(ANSI_HIGHLIGHT_RED)-1 + 2*(sizeof(ANSI_NORMAL)-1)) + +static void draw_cylon(char buffer[], size_t buflen, unsigned width, unsigned pos) { + char *p = buffer; + + assert(buflen >= CYLON_BUFFER_EXTRA + width + 1); + assert(pos <= width+1); /* 0 or width+1 mean that the center light is behind the corner */ + + if (pos > 1) { + if (pos > 2) + p = mempset(p, ' ', pos-2); + if (log_get_show_color()) + p = stpcpy(p, ANSI_RED); + *p++ = '*'; + } + + if (pos > 0 && pos <= width) { + if (log_get_show_color()) + p = stpcpy(p, ANSI_HIGHLIGHT_RED); + *p++ = '*'; + } + + if (log_get_show_color()) + p = stpcpy(p, ANSI_NORMAL); + + if (pos < width) { + if (log_get_show_color()) + p = stpcpy(p, ANSI_RED); + *p++ = '*'; + if (pos < width-1) + p = mempset(p, ' ', width-1-pos); + if (log_get_show_color()) + strcpy(p, ANSI_NORMAL); + } +} + +void manager_flip_auto_status(Manager *m, bool enable) { + assert(m); + + if (enable) { + if (m->show_status == SHOW_STATUS_AUTO) + manager_set_show_status(m, SHOW_STATUS_TEMPORARY); + } else { + if (m->show_status == SHOW_STATUS_TEMPORARY) + manager_set_show_status(m, SHOW_STATUS_AUTO); + } +} + +static void manager_print_jobs_in_progress(Manager *m) { + _cleanup_free_ char *job_of_n = NULL; + Iterator i; + Job *j; + unsigned counter = 0, print_nr; + char cylon[6 + CYLON_BUFFER_EXTRA + 1]; + unsigned cylon_pos; + char time[FORMAT_TIMESPAN_MAX], limit[FORMAT_TIMESPAN_MAX] = "no limit"; + uint64_t x; + + assert(m); + assert(m->n_running_jobs > 0); + + manager_flip_auto_status(m, true); + + print_nr = (m->jobs_in_progress_iteration / JOBS_IN_PROGRESS_PERIOD_DIVISOR) % m->n_running_jobs; + + HASHMAP_FOREACH(j, m->jobs, i) + if (j->state == JOB_RUNNING && counter++ == print_nr) + break; + + /* m->n_running_jobs must be consistent with the contents of m->jobs, + * so the above loop must have succeeded in finding j. */ + assert(counter == print_nr + 1); + assert(j); + + cylon_pos = m->jobs_in_progress_iteration % 14; + if (cylon_pos >= 8) + cylon_pos = 14 - cylon_pos; + draw_cylon(cylon, sizeof(cylon), 6, cylon_pos); + + m->jobs_in_progress_iteration++; + + if (m->n_running_jobs > 1) { + if (asprintf(&job_of_n, "(%u of %u) ", counter, m->n_running_jobs) < 0) + job_of_n = NULL; + } + + format_timespan(time, sizeof(time), now(CLOCK_MONOTONIC) - j->begin_usec, 1*USEC_PER_SEC); + if (job_get_timeout(j, &x) > 0) + format_timespan(limit, sizeof(limit), x - j->begin_usec, 1*USEC_PER_SEC); + + manager_status_printf(m, STATUS_TYPE_EPHEMERAL, cylon, + "%sA %s job is running for %s (%s / %s)", + strempty(job_of_n), + job_type_to_string(j->type), + unit_description(j->unit), + time, limit); +} + +static int have_ask_password(void) { + _cleanup_closedir_ DIR *dir; + + dir = opendir("/run/systemd/ask-password"); + if (!dir) { + if (errno == ENOENT) + return false; + else + return -errno; + } + + for (;;) { + struct dirent *de; + + errno = 0; + de = readdir(dir); + if (!de && errno > 0) + return -errno; + if (!de) + return false; + + if (startswith(de->d_name, "ask.")) + return true; + } +} + +static int manager_dispatch_ask_password_fd(sd_event_source *source, + int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + + assert(m); + + flush_fd(fd); + + m->have_ask_password = have_ask_password(); + if (m->have_ask_password < 0) + /* Log error but continue. Negative have_ask_password + * is treated as unknown status. */ + log_error_errno(m->have_ask_password, "Failed to list /run/systemd/ask-password: %m"); + + return 0; +} + +static void manager_close_ask_password(Manager *m) { + assert(m); + + m->ask_password_event_source = sd_event_source_unref(m->ask_password_event_source); + m->ask_password_inotify_fd = safe_close(m->ask_password_inotify_fd); + m->have_ask_password = -EINVAL; +} + +static int manager_check_ask_password(Manager *m) { + int r; + + assert(m); + + if (!m->ask_password_event_source) { + assert(m->ask_password_inotify_fd < 0); + + mkdir_p_label("/run/systemd/ask-password", 0755); + + m->ask_password_inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (m->ask_password_inotify_fd < 0) + return log_error_errno(errno, "inotify_init1() failed: %m"); + + if (inotify_add_watch(m->ask_password_inotify_fd, "/run/systemd/ask-password", IN_CREATE|IN_DELETE|IN_MOVE) < 0) { + log_error_errno(errno, "Failed to add watch on /run/systemd/ask-password: %m"); + manager_close_ask_password(m); + return -errno; + } + + r = sd_event_add_io(m->event, &m->ask_password_event_source, + m->ask_password_inotify_fd, EPOLLIN, + manager_dispatch_ask_password_fd, m); + if (r < 0) { + log_error_errno(errno, "Failed to add event source for /run/systemd/ask-password: %m"); + manager_close_ask_password(m); + return -errno; + } + + (void) sd_event_source_set_description(m->ask_password_event_source, "manager-ask-password"); + + /* Queries might have been added meanwhile... */ + manager_dispatch_ask_password_fd(m->ask_password_event_source, + m->ask_password_inotify_fd, EPOLLIN, m); + } + + return m->have_ask_password; +} + +static int manager_watch_idle_pipe(Manager *m) { + int r; + + assert(m); + + if (m->idle_pipe_event_source) + return 0; + + if (m->idle_pipe[2] < 0) + return 0; + + r = sd_event_add_io(m->event, &m->idle_pipe_event_source, m->idle_pipe[2], EPOLLIN, manager_dispatch_idle_pipe_fd, m); + if (r < 0) + return log_error_errno(r, "Failed to watch idle pipe: %m"); + + (void) sd_event_source_set_description(m->idle_pipe_event_source, "manager-idle-pipe"); + + return 0; +} + +static void manager_close_idle_pipe(Manager *m) { + assert(m); + + m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); + + safe_close_pair(m->idle_pipe); + safe_close_pair(m->idle_pipe + 2); +} + +static int manager_setup_time_change(Manager *m) { + int r; + + /* We only care for the cancellation event, hence we set the + * timeout to the latest possible value. */ + struct itimerspec its = { + .it_value.tv_sec = TIME_T_MAX, + }; + + assert(m); + assert_cc(sizeof(time_t) == sizeof(TIME_T_MAX)); + + if (m->test_run) + return 0; + + /* Uses TFD_TIMER_CANCEL_ON_SET to get notifications whenever + * CLOCK_REALTIME makes a jump relative to CLOCK_MONOTONIC */ + + m->time_change_fd = timerfd_create(CLOCK_REALTIME, TFD_NONBLOCK|TFD_CLOEXEC); + if (m->time_change_fd < 0) + return log_error_errno(errno, "Failed to create timerfd: %m"); + + if (timerfd_settime(m->time_change_fd, TFD_TIMER_ABSTIME|TFD_TIMER_CANCEL_ON_SET, &its, NULL) < 0) { + log_debug_errno(errno, "Failed to set up TFD_TIMER_CANCEL_ON_SET, ignoring: %m"); + m->time_change_fd = safe_close(m->time_change_fd); + return 0; + } + + r = sd_event_add_io(m->event, &m->time_change_event_source, m->time_change_fd, EPOLLIN, manager_dispatch_time_change_fd, m); + if (r < 0) + return log_error_errno(r, "Failed to create time change event source: %m"); + + (void) sd_event_source_set_description(m->time_change_event_source, "manager-time-change"); + + log_debug("Set up TFD_TIMER_CANCEL_ON_SET timerfd."); + + return 0; +} + +static int enable_special_signals(Manager *m) { + _cleanup_close_ int fd = -1; + + assert(m); + + if (m->test_run) + return 0; + + /* Enable that we get SIGINT on control-alt-del. In containers + * this will fail with EPERM (older) or EINVAL (newer), so + * ignore that. */ + if (reboot(RB_DISABLE_CAD) < 0 && errno != EPERM && errno != EINVAL) + log_warning_errno(errno, "Failed to enable ctrl-alt-del handling: %m"); + + fd = open_terminal("/dev/tty0", O_RDWR|O_NOCTTY|O_CLOEXEC); + if (fd < 0) { + /* Support systems without virtual console */ + if (fd != -ENOENT) + log_warning_errno(errno, "Failed to open /dev/tty0: %m"); + } else { + /* Enable that we get SIGWINCH on kbrequest */ + if (ioctl(fd, KDSIGACCEPT, SIGWINCH) < 0) + log_warning_errno(errno, "Failed to enable kbrequest handling: %m"); + } + + return 0; +} + +static int manager_setup_signals(Manager *m) { + struct sigaction sa = { + .sa_handler = SIG_DFL, + .sa_flags = SA_NOCLDSTOP|SA_RESTART, + }; + sigset_t mask; + int r; + + assert(m); + + assert_se(sigaction(SIGCHLD, &sa, NULL) == 0); + + /* We make liberal use of realtime signals here. On + * Linux/glibc we have 30 of them (with the exception of Linux + * on hppa, see below), between SIGRTMIN+0 ... SIGRTMIN+30 + * (aka SIGRTMAX). */ + + assert_se(sigemptyset(&mask) == 0); + sigset_add_many(&mask, + SIGCHLD, /* Child died */ + SIGTERM, /* Reexecute daemon */ + SIGHUP, /* Reload configuration */ + SIGUSR1, /* systemd/upstart: reconnect to D-Bus */ + SIGUSR2, /* systemd: dump status */ + SIGINT, /* Kernel sends us this on control-alt-del */ + SIGWINCH, /* Kernel sends us this on kbrequest (alt-arrowup) */ + SIGPWR, /* Some kernel drivers and upsd send us this on power failure */ + + SIGRTMIN+0, /* systemd: start default.target */ + SIGRTMIN+1, /* systemd: isolate rescue.target */ + SIGRTMIN+2, /* systemd: isolate emergency.target */ + SIGRTMIN+3, /* systemd: start halt.target */ + SIGRTMIN+4, /* systemd: start poweroff.target */ + SIGRTMIN+5, /* systemd: start reboot.target */ + SIGRTMIN+6, /* systemd: start kexec.target */ + + /* ... space for more special targets ... */ + + SIGRTMIN+13, /* systemd: Immediate halt */ + SIGRTMIN+14, /* systemd: Immediate poweroff */ + SIGRTMIN+15, /* systemd: Immediate reboot */ + SIGRTMIN+16, /* systemd: Immediate kexec */ + + /* ... space for more immediate system state changes ... */ + + SIGRTMIN+20, /* systemd: enable status messages */ + SIGRTMIN+21, /* systemd: disable status messages */ + SIGRTMIN+22, /* systemd: set log level to LOG_DEBUG */ + SIGRTMIN+23, /* systemd: set log level to LOG_INFO */ + SIGRTMIN+24, /* systemd: Immediate exit (--user only) */ + + /* .. one free signal here ... */ + +#if !defined(__hppa64__) && !defined(__hppa__) + /* Apparently Linux on hppa has fewer RT + * signals (SIGRTMAX is SIGRTMIN+25 there), + * hence let's not try to make use of them + * here. Since these commands are accessible + * by different means and only really a safety + * net, the missing functionality on hppa + * shouldn't matter. */ + + SIGRTMIN+26, /* systemd: set log target to journal-or-kmsg */ + SIGRTMIN+27, /* systemd: set log target to console */ + SIGRTMIN+28, /* systemd: set log target to kmsg */ + SIGRTMIN+29, /* systemd: set log target to syslog-or-kmsg (obsolete) */ + + /* ... one free signal here SIGRTMIN+30 ... */ +#endif + -1); + assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); + + m->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC); + if (m->signal_fd < 0) + return -errno; + + r = sd_event_add_io(m->event, &m->signal_event_source, m->signal_fd, EPOLLIN, manager_dispatch_signal_fd, m); + if (r < 0) + return r; + + (void) sd_event_source_set_description(m->signal_event_source, "manager-signal"); + + /* Process signals a bit earlier than the rest of things, but later than notify_fd processing, so that the + * notify processing can still figure out to which process/service a message belongs, before we reap the + * process. Also, process this before handling cgroup notifications, so that we always collect child exit + * status information before detecting that there's no process in a cgroup. */ + r = sd_event_source_set_priority(m->signal_event_source, SD_EVENT_PRIORITY_NORMAL-6); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(m)) + return enable_special_signals(m); + + return 0; +} + +static void manager_clean_environment(Manager *m) { + assert(m); + + /* Let's remove some environment variables that we + * need ourselves to communicate with our clients */ + strv_env_unset_many( + m->environment, + "NOTIFY_SOCKET", + "MAINPID", + "MANAGERPID", + "LISTEN_PID", + "LISTEN_FDS", + "LISTEN_FDNAMES", + "WATCHDOG_PID", + "WATCHDOG_USEC", + "INVOCATION_ID", + NULL); +} + +static int manager_default_environment(Manager *m) { + assert(m); + + if (MANAGER_IS_SYSTEM(m)) { + /* The system manager always starts with a clean + * environment for its children. It does not import + * the kernel or the parents exported variables. + * + * The initial passed environ is untouched to keep + * /proc/self/environ valid; it is used for tagging + * the init process inside containers. */ + m->environment = strv_new("PATH=" DEFAULT_PATH, + NULL); + + /* Import locale variables LC_*= from configuration */ + locale_setup(&m->environment); + } else { + /* The user manager passes its own environment + * along to its children. */ + m->environment = strv_copy(environ); + } + + if (!m->environment) + return -ENOMEM; + + manager_clean_environment(m); + strv_sort(m->environment); + + return 0; +} + +int manager_new(UnitFileScope scope, bool test_run, Manager **_m) { + Manager *m; + int r; + + assert(_m); + assert(IN_SET(scope, UNIT_FILE_SYSTEM, UNIT_FILE_USER)); + + m = new0(Manager, 1); + if (!m) + return -ENOMEM; + + m->unit_file_scope = scope; + m->exit_code = _MANAGER_EXIT_CODE_INVALID; + m->default_timer_accuracy_usec = USEC_PER_MINUTE; + m->default_tasks_accounting = true; + m->default_tasks_max = UINT64_MAX; + +#ifdef ENABLE_EFI + if (MANAGER_IS_SYSTEM(m) && detect_container() <= 0) + boot_timestamps(&m->userspace_timestamp, &m->firmware_timestamp, &m->loader_timestamp); +#endif + + /* Prepare log fields we can use for structured logging */ + if (MANAGER_IS_SYSTEM(m)) { + m->unit_log_field = "UNIT="; + m->unit_log_format_string = "UNIT=%s"; + + m->invocation_log_field = "INVOCATION_ID="; + m->invocation_log_format_string = "INVOCATION_ID=" SD_ID128_FORMAT_STR; + } else { + m->unit_log_field = "USER_UNIT="; + m->unit_log_format_string = "USER_UNIT=%s"; + + m->invocation_log_field = "USER_INVOCATION_ID="; + m->invocation_log_format_string = "USER_INVOCATION_ID=" SD_ID128_FORMAT_STR; + } + + m->idle_pipe[0] = m->idle_pipe[1] = m->idle_pipe[2] = m->idle_pipe[3] = -1; + + m->pin_cgroupfs_fd = m->notify_fd = m->cgroups_agent_fd = m->signal_fd = m->time_change_fd = + m->dev_autofs_fd = m->private_listen_fd = m->cgroup_inotify_fd = + m->ask_password_inotify_fd = -1; + + m->user_lookup_fds[0] = m->user_lookup_fds[1] = -1; + + m->current_job_id = 1; /* start as id #1, so that we can leave #0 around as "null-like" value */ + + m->have_ask_password = -EINVAL; /* we don't know */ + m->first_boot = -1; + + m->test_run = test_run; + + /* Reboot immediately if the user hits C-A-D more often than 7x per 2s */ + RATELIMIT_INIT(m->ctrl_alt_del_ratelimit, 2 * USEC_PER_SEC, 7); + + r = manager_default_environment(m); + if (r < 0) + goto fail; + + r = hashmap_ensure_allocated(&m->units, &string_hash_ops); + if (r < 0) + goto fail; + + r = hashmap_ensure_allocated(&m->jobs, NULL); + if (r < 0) + goto fail; + + r = hashmap_ensure_allocated(&m->cgroup_unit, &string_hash_ops); + if (r < 0) + goto fail; + + r = hashmap_ensure_allocated(&m->watch_bus, &string_hash_ops); + if (r < 0) + goto fail; + + r = sd_event_default(&m->event); + if (r < 0) + goto fail; + + r = sd_event_add_defer(m->event, &m->run_queue_event_source, manager_dispatch_run_queue, m); + if (r < 0) + goto fail; + + r = sd_event_source_set_priority(m->run_queue_event_source, SD_EVENT_PRIORITY_IDLE); + if (r < 0) + goto fail; + + r = sd_event_source_set_enabled(m->run_queue_event_source, SD_EVENT_OFF); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(m->run_queue_event_source, "manager-run-queue"); + + r = manager_setup_signals(m); + if (r < 0) + goto fail; + + r = manager_setup_cgroup(m); + if (r < 0) + goto fail; + + r = manager_setup_time_change(m); + if (r < 0) + goto fail; + + m->udev = udev_new(); + if (!m->udev) { + r = -ENOMEM; + goto fail; + } + + /* Note that we do not set up the notify fd here. We do that after deserialization, + * since they might have gotten serialized across the reexec. */ + + m->taint_usr = dir_is_empty("/usr") > 0; + + *_m = m; + return 0; + +fail: + manager_free(m); + return r; +} + +static int manager_setup_notify(Manager *m) { + int r; + + if (m->test_run) + return 0; + + if (m->notify_fd < 0) { + _cleanup_close_ int fd = -1; + union sockaddr_union sa = { + .sa.sa_family = AF_UNIX, + }; + static const int one = 1; + const char *e; + + /* First free all secondary fields */ + m->notify_socket = mfree(m->notify_socket); + m->notify_event_source = sd_event_source_unref(m->notify_event_source); + + fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) + return log_error_errno(errno, "Failed to allocate notification socket: %m"); + + fd_inc_rcvbuf(fd, NOTIFY_RCVBUF_SIZE); + + e = manager_get_runtime_prefix(m); + if (!e) { + log_error("Failed to determine runtime prefix."); + return -EINVAL; + } + + m->notify_socket = strappend(e, "/systemd/notify"); + if (!m->notify_socket) + return log_oom(); + + (void) mkdir_parents_label(m->notify_socket, 0755); + (void) unlink(m->notify_socket); + + strncpy(sa.un.sun_path, m->notify_socket, sizeof(sa.un.sun_path)-1); + r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); + if (r < 0) + return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path); + + r = setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one)); + if (r < 0) + return log_error_errno(errno, "SO_PASSCRED failed: %m"); + + m->notify_fd = fd; + fd = -1; + + log_debug("Using notification socket %s", m->notify_socket); + } + + if (!m->notify_event_source) { + r = sd_event_add_io(m->event, &m->notify_event_source, m->notify_fd, EPOLLIN, manager_dispatch_notify_fd, m); + if (r < 0) + return log_error_errno(r, "Failed to allocate notify event source: %m"); + + /* Process notification messages a bit earlier than SIGCHLD, so that we can still identify to which + * service an exit message belongs. */ + r = sd_event_source_set_priority(m->notify_event_source, SD_EVENT_PRIORITY_NORMAL-7); + if (r < 0) + return log_error_errno(r, "Failed to set priority of notify event source: %m"); + + (void) sd_event_source_set_description(m->notify_event_source, "manager-notify"); + } + + return 0; +} + +static int manager_setup_cgroups_agent(Manager *m) { + + static const union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/cgroups-agent", + }; + int r; + + /* This creates a listening socket we receive cgroups agent messages on. We do not use D-Bus for delivering + * these messages from the cgroups agent binary to PID 1, as the cgroups agent binary is very short-living, and + * each instance of it needs a new D-Bus connection. Since D-Bus connections are SOCK_STREAM/AF_UNIX, on + * overloaded systems the backlog of the D-Bus socket becomes relevant, as not more than the configured number + * of D-Bus connections may be queued until the kernel will start dropping further incoming connections, + * possibly resulting in lost cgroups agent messages. To avoid this, we'll use a private SOCK_DGRAM/AF_UNIX + * socket, where no backlog is relevant as communication may take place without an actual connect() cycle, and + * we thus won't lose messages. + * + * Note that PID 1 will forward the agent message to system bus, so that the user systemd instance may listen + * to it. The system instance hence listens on this special socket, but the user instances listen on the system + * bus for these messages. */ + + if (m->test_run) + return 0; + + if (!MANAGER_IS_SYSTEM(m)) + return 0; + + if (cg_unified(SYSTEMD_CGROUP_CONTROLLER) > 0) /* We don't need this anymore on the unified hierarchy */ + return 0; + + if (m->cgroups_agent_fd < 0) { + _cleanup_close_ int fd = -1; + + /* First free all secondary fields */ + m->cgroups_agent_event_source = sd_event_source_unref(m->cgroups_agent_event_source); + + fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) + return log_error_errno(errno, "Failed to allocate cgroups agent socket: %m"); + + fd_inc_rcvbuf(fd, CGROUPS_AGENT_RCVBUF_SIZE); + + (void) unlink(sa.un.sun_path); + + /* Only allow root to connect to this socket */ + RUN_WITH_UMASK(0077) + r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); + if (r < 0) + return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path); + + m->cgroups_agent_fd = fd; + fd = -1; + } + + if (!m->cgroups_agent_event_source) { + r = sd_event_add_io(m->event, &m->cgroups_agent_event_source, m->cgroups_agent_fd, EPOLLIN, manager_dispatch_cgroups_agent_fd, m); + if (r < 0) + return log_error_errno(r, "Failed to allocate cgroups agent event source: %m"); + + /* Process cgroups notifications early, but after having processed service notification messages or + * SIGCHLD signals, so that a cgroup running empty is always just the last safety net of notification, + * and we collected the metadata the notification and SIGCHLD stuff offers first. Also see handling of + * cgroup inotify for the unified cgroup stuff. */ + r = sd_event_source_set_priority(m->cgroups_agent_event_source, SD_EVENT_PRIORITY_NORMAL-5); + if (r < 0) + return log_error_errno(r, "Failed to set priority of cgroups agent event source: %m"); + + (void) sd_event_source_set_description(m->cgroups_agent_event_source, "manager-cgroups-agent"); + } + + return 0; +} + +static int manager_setup_user_lookup_fd(Manager *m) { + int r; + + assert(m); + + /* Set up the socket pair used for passing UID/GID resolution results from forked off processes to PID + * 1. Background: we can't do name lookups (NSS) from PID 1, since it might involve IPC and thus activation, + * and we might hence deadlock on ourselves. Hence we do all user/group lookups asynchronously from the forked + * off processes right before executing the binaries to start. In order to be able to clean up any IPC objects + * created by a unit (see RemoveIPC=) we need to know in PID 1 the used UID/GID of the executed processes, + * hence we establish this communication channel so that forked off processes can pass their UID/GID + * information back to PID 1. The forked off processes send their resolved UID/GID to PID 1 in a simple + * datagram, along with their unit name, so that we can share one communication socket pair among all units for + * this purpose. + * + * You might wonder why we need a communication channel for this that is independent of the usual notification + * socket scheme (i.e. $NOTIFY_SOCKET). The primary difference is about trust: data sent via the $NOTIFY_SOCKET + * channel is only accepted if it originates from the right unit and if reception was enabled for it. The user + * lookup socket OTOH is only accessible by PID 1 and its children until they exec(), and always available. + * + * Note that this function is called under two circumstances: when we first initialize (in which case we + * allocate both the socket pair and the event source to listen on it), and when we deserialize after a reload + * (in which case the socket pair already exists but we still need to allocate the event source for it). */ + + if (m->user_lookup_fds[0] < 0) { + + /* Free all secondary fields */ + safe_close_pair(m->user_lookup_fds); + m->user_lookup_event_source = sd_event_source_unref(m->user_lookup_event_source); + + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, m->user_lookup_fds) < 0) + return log_error_errno(errno, "Failed to allocate user lookup socket: %m"); + + (void) fd_inc_rcvbuf(m->user_lookup_fds[0], NOTIFY_RCVBUF_SIZE); + } + + if (!m->user_lookup_event_source) { + r = sd_event_add_io(m->event, &m->user_lookup_event_source, m->user_lookup_fds[0], EPOLLIN, manager_dispatch_user_lookup_fd, m); + if (r < 0) + return log_error_errno(errno, "Failed to allocate user lookup event source: %m"); + + /* Process even earlier than the notify event source, so that we always know first about valid UID/GID + * resolutions */ + r = sd_event_source_set_priority(m->user_lookup_event_source, SD_EVENT_PRIORITY_NORMAL-8); + if (r < 0) + return log_error_errno(errno, "Failed to set priority ot user lookup event source: %m"); + + (void) sd_event_source_set_description(m->user_lookup_event_source, "user-lookup"); + } + + return 0; +} + +static int manager_connect_bus(Manager *m, bool reexecuting) { + bool try_bus_connect; + + assert(m); + + if (m->test_run) + return 0; + + try_bus_connect = + reexecuting || + (MANAGER_IS_USER(m) && getenv("DBUS_SESSION_BUS_ADDRESS")); + + /* Try to connect to the buses, if possible. */ + return bus_init(m, try_bus_connect); +} + +static unsigned manager_dispatch_cleanup_queue(Manager *m) { + Unit *u; + unsigned n = 0; + + assert(m); + + while ((u = m->cleanup_queue)) { + assert(u->in_cleanup_queue); + + unit_free(u); + n++; + } + + return n; +} + +enum { + GC_OFFSET_IN_PATH, /* This one is on the path we were traveling */ + GC_OFFSET_UNSURE, /* No clue */ + GC_OFFSET_GOOD, /* We still need this unit */ + GC_OFFSET_BAD, /* We don't need this unit anymore */ + _GC_OFFSET_MAX +}; + +static void unit_gc_mark_good(Unit *u, unsigned gc_marker) { + Iterator i; + Unit *other; + + u->gc_marker = gc_marker + GC_OFFSET_GOOD; + + /* Recursively mark referenced units as GOOD as well */ + SET_FOREACH(other, u->dependencies[UNIT_REFERENCES], i) + if (other->gc_marker == gc_marker + GC_OFFSET_UNSURE) + unit_gc_mark_good(other, gc_marker); +} + +static void unit_gc_sweep(Unit *u, unsigned gc_marker) { + Iterator i; + Unit *other; + bool is_bad; + + assert(u); + + if (u->gc_marker == gc_marker + GC_OFFSET_GOOD || + u->gc_marker == gc_marker + GC_OFFSET_BAD || + u->gc_marker == gc_marker + GC_OFFSET_UNSURE || + u->gc_marker == gc_marker + GC_OFFSET_IN_PATH) + return; + + if (u->in_cleanup_queue) + goto bad; + + if (unit_check_gc(u)) + goto good; + + u->gc_marker = gc_marker + GC_OFFSET_IN_PATH; + + is_bad = true; + + SET_FOREACH(other, u->dependencies[UNIT_REFERENCED_BY], i) { + unit_gc_sweep(other, gc_marker); + + if (other->gc_marker == gc_marker + GC_OFFSET_GOOD) + goto good; + + if (other->gc_marker != gc_marker + GC_OFFSET_BAD) + is_bad = false; + } + + if (is_bad) + goto bad; + + /* We were unable to find anything out about this entry, so + * let's investigate it later */ + u->gc_marker = gc_marker + GC_OFFSET_UNSURE; + unit_add_to_gc_queue(u); + return; + +bad: + /* We definitely know that this one is not useful anymore, so + * let's mark it for deletion */ + u->gc_marker = gc_marker + GC_OFFSET_BAD; + unit_add_to_cleanup_queue(u); + return; + +good: + unit_gc_mark_good(u, gc_marker); +} + +static unsigned manager_dispatch_gc_queue(Manager *m) { + Unit *u; + unsigned n = 0; + unsigned gc_marker; + + assert(m); + + /* log_debug("Running GC..."); */ + + m->gc_marker += _GC_OFFSET_MAX; + if (m->gc_marker + _GC_OFFSET_MAX <= _GC_OFFSET_MAX) + m->gc_marker = 1; + + gc_marker = m->gc_marker; + + while ((u = m->gc_queue)) { + assert(u->in_gc_queue); + + unit_gc_sweep(u, gc_marker); + + LIST_REMOVE(gc_queue, m->gc_queue, u); + u->in_gc_queue = false; + + n++; + + if (u->gc_marker == gc_marker + GC_OFFSET_BAD || + u->gc_marker == gc_marker + GC_OFFSET_UNSURE) { + if (u->id) + log_unit_debug(u, "Collecting."); + u->gc_marker = gc_marker + GC_OFFSET_BAD; + unit_add_to_cleanup_queue(u); + } + } + + m->n_in_gc_queue = 0; + + return n; +} + +static void manager_clear_jobs_and_units(Manager *m) { + Unit *u; + + assert(m); + + while ((u = hashmap_first(m->units))) + unit_free(u); + + manager_dispatch_cleanup_queue(m); + + assert(!m->load_queue); + assert(!m->run_queue); + assert(!m->dbus_unit_queue); + assert(!m->dbus_job_queue); + assert(!m->cleanup_queue); + assert(!m->gc_queue); + + assert(hashmap_isempty(m->jobs)); + assert(hashmap_isempty(m->units)); + + m->n_on_console = 0; + m->n_running_jobs = 0; +} + +Manager* manager_free(Manager *m) { + UnitType c; + int i; + + if (!m) + return NULL; + + manager_clear_jobs_and_units(m); + + for (c = 0; c < _UNIT_TYPE_MAX; c++) + if (unit_vtable[c]->shutdown) + unit_vtable[c]->shutdown(m); + + /* If we reexecute ourselves, we keep the root cgroup + * around */ + manager_shutdown_cgroup(m, m->exit_code != MANAGER_REEXECUTE); + + lookup_paths_flush_generator(&m->lookup_paths); + + bus_done(m); + + dynamic_user_vacuum(m, false); + hashmap_free(m->dynamic_users); + + hashmap_free(m->units); + hashmap_free(m->units_by_invocation_id); + hashmap_free(m->jobs); + hashmap_free(m->watch_pids1); + hashmap_free(m->watch_pids2); + hashmap_free(m->watch_bus); + + set_free(m->startup_units); + set_free(m->failed_units); + + sd_event_source_unref(m->signal_event_source); + sd_event_source_unref(m->notify_event_source); + sd_event_source_unref(m->cgroups_agent_event_source); + sd_event_source_unref(m->time_change_event_source); + sd_event_source_unref(m->jobs_in_progress_event_source); + sd_event_source_unref(m->run_queue_event_source); + sd_event_source_unref(m->user_lookup_event_source); + + safe_close(m->signal_fd); + safe_close(m->notify_fd); + safe_close(m->cgroups_agent_fd); + safe_close(m->time_change_fd); + safe_close_pair(m->user_lookup_fds); + + manager_close_ask_password(m); + + manager_close_idle_pipe(m); + + udev_unref(m->udev); + sd_event_unref(m->event); + + free(m->notify_socket); + + lookup_paths_free(&m->lookup_paths); + strv_free(m->environment); + + hashmap_free(m->cgroup_unit); + set_free_free(m->unit_path_cache); + + free(m->switch_root); + free(m->switch_root_init); + + for (i = 0; i < _RLIMIT_MAX; i++) + m->rlimit[i] = mfree(m->rlimit[i]); + + assert(hashmap_isempty(m->units_requiring_mounts_for)); + hashmap_free(m->units_requiring_mounts_for); + + hashmap_free(m->uid_refs); + hashmap_free(m->gid_refs); + + return mfree(m); +} + +void manager_enumerate(Manager *m) { + UnitType c; + + assert(m); + + /* Let's ask every type to load all units from disk/kernel + * that it might know */ + for (c = 0; c < _UNIT_TYPE_MAX; c++) { + if (!unit_type_supported(c)) { + log_debug("Unit type .%s is not supported on this system.", unit_type_to_string(c)); + continue; + } + + if (!unit_vtable[c]->enumerate) + continue; + + unit_vtable[c]->enumerate(m); + } + + manager_dispatch_load_queue(m); +} + +static void manager_coldplug(Manager *m) { + Iterator i; + Unit *u; + char *k; + int r; + + assert(m); + + /* Then, let's set up their initial state. */ + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + + /* ignore aliases */ + if (u->id != k) + continue; + + r = unit_coldplug(u); + if (r < 0) + log_warning_errno(r, "We couldn't coldplug %s, proceeding anyway: %m", u->id); + } +} + +static void manager_build_unit_path_cache(Manager *m) { + char **i; + int r; + + assert(m); + + set_free_free(m->unit_path_cache); + + m->unit_path_cache = set_new(&string_hash_ops); + if (!m->unit_path_cache) { + r = -ENOMEM; + goto fail; + } + + /* This simply builds a list of files we know exist, so that + * we don't always have to go to disk */ + + STRV_FOREACH(i, m->lookup_paths.search_path) { + _cleanup_closedir_ DIR *d = NULL; + struct dirent *de; + + d = opendir(*i); + if (!d) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open directory %s, ignoring: %m", *i); + continue; + } + + FOREACH_DIRENT(de, d, r = -errno; goto fail) { + char *p; + + p = strjoin(streq(*i, "/") ? "" : *i, "/", de->d_name, NULL); + if (!p) { + r = -ENOMEM; + goto fail; + } + + r = set_consume(m->unit_path_cache, p); + if (r < 0) + goto fail; + } + } + + return; + +fail: + log_warning_errno(r, "Failed to build unit path cache, proceeding without: %m"); + m->unit_path_cache = set_free_free(m->unit_path_cache); +} + +static void manager_distribute_fds(Manager *m, FDSet *fds) { + Iterator i; + Unit *u; + + assert(m); + + HASHMAP_FOREACH(u, m->units, i) { + + if (fdset_size(fds) <= 0) + break; + + if (!UNIT_VTABLE(u)->distribute_fds) + continue; + + UNIT_VTABLE(u)->distribute_fds(u, fds); + } +} + +int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { + int r, q; + + assert(m); + + r = lookup_paths_init(&m->lookup_paths, m->unit_file_scope, 0, NULL); + if (r < 0) + return r; + + /* Make sure the transient directory always exists, so that it remains in the search path */ + if (!m->test_run) { + r = mkdir_p_label(m->lookup_paths.transient, 0755); + if (r < 0) + return r; + } + + dual_timestamp_get(&m->generators_start_timestamp); + r = manager_run_generators(m); + dual_timestamp_get(&m->generators_finish_timestamp); + if (r < 0) + return r; + + lookup_paths_reduce(&m->lookup_paths); + manager_build_unit_path_cache(m); + + /* If we will deserialize make sure that during enumeration + * this is already known, so we increase the counter here + * already */ + if (serialization) + m->n_reloading++; + + /* First, enumerate what we can from all config files */ + dual_timestamp_get(&m->units_load_start_timestamp); + manager_enumerate(m); + dual_timestamp_get(&m->units_load_finish_timestamp); + + /* Second, deserialize if there is something to deserialize */ + if (serialization) + r = manager_deserialize(m, serialization, fds); + + /* Any fds left? Find some unit which wants them. This is + * useful to allow container managers to pass some file + * descriptors to us pre-initialized. This enables + * socket-based activation of entire containers. */ + manager_distribute_fds(m, fds); + + /* We might have deserialized the notify fd, but if we didn't + * then let's create the bus now */ + q = manager_setup_notify(m); + if (q < 0 && r == 0) + r = q; + + q = manager_setup_cgroups_agent(m); + if (q < 0 && r == 0) + r = q; + + q = manager_setup_user_lookup_fd(m); + if (q < 0 && r == 0) + r = q; + + /* Let's connect to the bus now. */ + (void) manager_connect_bus(m, !!serialization); + + (void) bus_track_coldplug(m, &m->subscribed, false, m->deserialized_subscribed); + m->deserialized_subscribed = strv_free(m->deserialized_subscribed); + + /* Third, fire things up! */ + manager_coldplug(m); + + /* Release any dynamic users no longer referenced */ + dynamic_user_vacuum(m, true); + + /* Release any references to UIDs/GIDs no longer referenced, and destroy any IPC owned by them */ + manager_vacuum_uid_refs(m); + manager_vacuum_gid_refs(m); + + if (serialization) { + assert(m->n_reloading > 0); + m->n_reloading--; + + /* Let's wait for the UnitNew/JobNew messages being + * sent, before we notify that the reload is + * finished */ + m->send_reloading_done = true; + } + + return r; +} + +int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, sd_bus_error *e, Job **_ret) { + int r; + Transaction *tr; + + assert(m); + assert(type < _JOB_TYPE_MAX); + assert(unit); + assert(mode < _JOB_MODE_MAX); + + if (mode == JOB_ISOLATE && type != JOB_START) + return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Isolate is only valid for start."); + + if (mode == JOB_ISOLATE && !unit->allow_isolate) + return sd_bus_error_setf(e, BUS_ERROR_NO_ISOLATION, "Operation refused, unit may not be isolated."); + + log_unit_debug(unit, "Trying to enqueue job %s/%s/%s", unit->id, job_type_to_string(type), job_mode_to_string(mode)); + + type = job_type_collapse(type, unit); + + tr = transaction_new(mode == JOB_REPLACE_IRREVERSIBLY); + if (!tr) + return -ENOMEM; + + r = transaction_add_job_and_dependencies(tr, type, unit, NULL, true, false, + mode == JOB_IGNORE_DEPENDENCIES || mode == JOB_IGNORE_REQUIREMENTS, + mode == JOB_IGNORE_DEPENDENCIES, e); + if (r < 0) + goto tr_abort; + + if (mode == JOB_ISOLATE) { + r = transaction_add_isolate_jobs(tr, m); + if (r < 0) + goto tr_abort; + } + + r = transaction_activate(tr, m, mode, e); + if (r < 0) + goto tr_abort; + + log_unit_debug(unit, + "Enqueued job %s/%s as %u", unit->id, + job_type_to_string(type), (unsigned) tr->anchor_job->id); + + if (_ret) + *_ret = tr->anchor_job; + + transaction_free(tr); + return 0; + +tr_abort: + transaction_abort(tr); + transaction_free(tr); + return r; +} + +int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, sd_bus_error *e, Job **ret) { + Unit *unit; + int r; + + assert(m); + assert(type < _JOB_TYPE_MAX); + assert(name); + assert(mode < _JOB_MODE_MAX); + + r = manager_load_unit(m, name, NULL, NULL, &unit); + if (r < 0) + return r; + + return manager_add_job(m, type, unit, mode, e, ret); +} + +int manager_add_job_by_name_and_warn(Manager *m, JobType type, const char *name, JobMode mode, Job **ret) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + assert(m); + assert(type < _JOB_TYPE_MAX); + assert(name); + assert(mode < _JOB_MODE_MAX); + + r = manager_add_job_by_name(m, type, name, mode, &error, ret); + if (r < 0) + return log_warning_errno(r, "Failed to enqueue %s job for %s: %s", job_mode_to_string(mode), name, bus_error_message(&error, r)); + + return r; +} + +Job *manager_get_job(Manager *m, uint32_t id) { + assert(m); + + return hashmap_get(m->jobs, UINT32_TO_PTR(id)); +} + +Unit *manager_get_unit(Manager *m, const char *name) { + assert(m); + assert(name); + + return hashmap_get(m->units, name); +} + +unsigned manager_dispatch_load_queue(Manager *m) { + Unit *u; + unsigned n = 0; + + assert(m); + + /* Make sure we are not run recursively */ + if (m->dispatching_load_queue) + return 0; + + m->dispatching_load_queue = true; + + /* Dispatches the load queue. Takes a unit from the queue and + * tries to load its data until the queue is empty */ + + while ((u = m->load_queue)) { + assert(u->in_load_queue); + + unit_load(u); + n++; + } + + m->dispatching_load_queue = false; + return n; +} + +int manager_load_unit_prepare( + Manager *m, + const char *name, + const char *path, + sd_bus_error *e, + Unit **_ret) { + + Unit *ret; + UnitType t; + int r; + + assert(m); + assert(name || path); + + /* This will prepare the unit for loading, but not actually + * load anything from disk. */ + + if (path && !is_path(path)) + return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Path %s is not absolute.", path); + + if (!name) + name = basename(path); + + t = unit_name_to_type(name); + + if (t == _UNIT_TYPE_INVALID || !unit_name_is_valid(name, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) { + if (unit_name_is_valid(name, UNIT_NAME_TEMPLATE)) + return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Unit name %s is missing the instance name.", name); + + return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Unit name %s is not valid.", name); + } + + ret = manager_get_unit(m, name); + if (ret) { + *_ret = ret; + return 1; + } + + ret = unit_new(m, unit_vtable[t]->object_size); + if (!ret) + return -ENOMEM; + + if (path) { + ret->fragment_path = strdup(path); + if (!ret->fragment_path) { + unit_free(ret); + return -ENOMEM; + } + } + + r = unit_add_name(ret, name); + if (r < 0) { + unit_free(ret); + return r; + } + + unit_add_to_load_queue(ret); + unit_add_to_dbus_queue(ret); + unit_add_to_gc_queue(ret); + + if (_ret) + *_ret = ret; + + return 0; +} + +int manager_load_unit( + Manager *m, + const char *name, + const char *path, + sd_bus_error *e, + Unit **_ret) { + + int r; + + assert(m); + + /* This will load the service information files, but not actually + * start any services or anything. */ + + r = manager_load_unit_prepare(m, name, path, e, _ret); + if (r != 0) + return r; + + manager_dispatch_load_queue(m); + + if (_ret) + *_ret = unit_follow_merge(*_ret); + + return 0; +} + +void manager_dump_jobs(Manager *s, FILE *f, const char *prefix) { + Iterator i; + Job *j; + + assert(s); + assert(f); + + HASHMAP_FOREACH(j, s->jobs, i) + job_dump(j, f, prefix); +} + +void manager_dump_units(Manager *s, FILE *f, const char *prefix) { + Iterator i; + Unit *u; + const char *t; + + assert(s); + assert(f); + + HASHMAP_FOREACH_KEY(u, t, s->units, i) + if (u->id == t) + unit_dump(u, f, prefix); +} + +void manager_clear_jobs(Manager *m) { + Job *j; + + assert(m); + + while ((j = hashmap_first(m->jobs))) + /* No need to recurse. We're cancelling all jobs. */ + job_finish_and_invalidate(j, JOB_CANCELED, false, false); +} + +static int manager_dispatch_run_queue(sd_event_source *source, void *userdata) { + Manager *m = userdata; + Job *j; + + assert(source); + assert(m); + + while ((j = m->run_queue)) { + assert(j->installed); + assert(j->in_run_queue); + + job_run_and_invalidate(j); + } + + if (m->n_running_jobs > 0) + manager_watch_jobs_in_progress(m); + + if (m->n_on_console > 0) + manager_watch_idle_pipe(m); + + return 1; +} + +static unsigned manager_dispatch_dbus_queue(Manager *m) { + Job *j; + Unit *u; + unsigned n = 0; + + assert(m); + + if (m->dispatching_dbus_queue) + return 0; + + m->dispatching_dbus_queue = true; + + while ((u = m->dbus_unit_queue)) { + assert(u->in_dbus_queue); + + bus_unit_send_change_signal(u); + n++; + } + + while ((j = m->dbus_job_queue)) { + assert(j->in_dbus_queue); + + bus_job_send_change_signal(j); + n++; + } + + m->dispatching_dbus_queue = false; + + if (m->send_reloading_done) { + m->send_reloading_done = false; + + bus_manager_send_reloading(m, false); + } + + if (m->queued_message) + bus_send_queued_message(m); + + return n; +} + +static int manager_dispatch_cgroups_agent_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + char buf[PATH_MAX+1]; + ssize_t n; + + n = recv(fd, buf, sizeof(buf), 0); + if (n < 0) + return log_error_errno(errno, "Failed to read cgroups agent message: %m"); + if (n == 0) { + log_error("Got zero-length cgroups agent message, ignoring."); + return 0; + } + if ((size_t) n >= sizeof(buf)) { + log_error("Got overly long cgroups agent message, ignoring."); + return 0; + } + + if (memchr(buf, 0, n)) { + log_error("Got cgroups agent message with embedded NUL byte, ignoring."); + return 0; + } + buf[n] = 0; + + manager_notify_cgroup_empty(m, buf); + bus_forward_agent_released(m, buf); + + return 0; +} + +static void manager_invoke_notify_message(Manager *m, Unit *u, pid_t pid, const char *buf, FDSet *fds) { + _cleanup_strv_free_ char **tags = NULL; + + assert(m); + assert(u); + assert(buf); + + tags = strv_split(buf, "\n\r"); + if (!tags) { + log_oom(); + return; + } + + if (UNIT_VTABLE(u)->notify_message) + UNIT_VTABLE(u)->notify_message(u, pid, tags, fds); + else if (_unlikely_(log_get_max_level() >= LOG_DEBUG)) { + _cleanup_free_ char *x = NULL, *y = NULL; + + x = cescape(buf); + if (x) + y = ellipsize(x, 20, 90); + log_unit_debug(u, "Got notification message \"%s\", ignoring.", strnull(y)); + } +} + +static int manager_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + + _cleanup_fdset_free_ FDSet *fds = NULL; + Manager *m = userdata; + char buf[NOTIFY_BUFFER_MAX+1]; + struct iovec iovec = { + .iov_base = buf, + .iov_len = sizeof(buf)-1, + }; + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) + + CMSG_SPACE(sizeof(int) * NOTIFY_FD_MAX)]; + } control = {}; + struct msghdr msghdr = { + .msg_iov = &iovec, + .msg_iovlen = 1, + .msg_control = &control, + .msg_controllen = sizeof(control), + }; + + struct cmsghdr *cmsg; + struct ucred *ucred = NULL; + Unit *u1, *u2, *u3; + int r, *fd_array = NULL; + unsigned n_fds = 0; + ssize_t n; + + assert(m); + assert(m->notify_fd == fd); + + if (revents != EPOLLIN) { + log_warning("Got unexpected poll event for notify fd."); + return 0; + } + + n = recvmsg(m->notify_fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC|MSG_TRUNC); + if (n < 0) { + if (IN_SET(errno, EAGAIN, EINTR)) + return 0; /* Spurious wakeup, try again */ + + /* If this is any other, real error, then let's stop processing this socket. This of course means we + * won't take notification messages anymore, but that's still better than busy looping around this: + * being woken up over and over again but being unable to actually read the message off the socket. */ + return log_error_errno(errno, "Failed to receive notification message: %m"); + } + + CMSG_FOREACH(cmsg, &msghdr) { + if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SCM_RIGHTS) { + + fd_array = (int*) CMSG_DATA(cmsg); + n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int); + + } else if (cmsg->cmsg_level == SOL_SOCKET && + cmsg->cmsg_type == SCM_CREDENTIALS && + cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) { + + ucred = (struct ucred*) CMSG_DATA(cmsg); + } + } + + if (n_fds > 0) { + assert(fd_array); + + r = fdset_new_array(&fds, fd_array, n_fds); + if (r < 0) { + close_many(fd_array, n_fds); + log_oom(); + return 0; + } + } + + if (!ucred || ucred->pid <= 0) { + log_warning("Received notify message without valid credentials. Ignoring."); + return 0; + } + + if ((size_t) n >= sizeof(buf) || (msghdr.msg_flags & MSG_TRUNC)) { + log_warning("Received notify message exceeded maximum size. Ignoring."); + return 0; + } + + /* As extra safety check, let's make sure the string we get doesn't contain embedded NUL bytes. We permit one + * trailing NUL byte in the message, but don't expect it. */ + if (n > 1 && memchr(buf, 0, n-1)) { + log_warning("Received notify message with embedded NUL bytes. Ignoring."); + return 0; + } + + /* Make sure it's NUL-terminated. */ + buf[n] = 0; + + /* Notify every unit that might be interested, but try + * to avoid notifying the same one multiple times. */ + u1 = manager_get_unit_by_pid_cgroup(m, ucred->pid); + if (u1) + manager_invoke_notify_message(m, u1, ucred->pid, buf, fds); + + u2 = hashmap_get(m->watch_pids1, PID_TO_PTR(ucred->pid)); + if (u2 && u2 != u1) + manager_invoke_notify_message(m, u2, ucred->pid, buf, fds); + + u3 = hashmap_get(m->watch_pids2, PID_TO_PTR(ucred->pid)); + if (u3 && u3 != u2 && u3 != u1) + manager_invoke_notify_message(m, u3, ucred->pid, buf, fds); + + if (!u1 && !u2 && !u3) + log_warning("Cannot find unit for notify message of PID "PID_FMT".", ucred->pid); + + if (fdset_size(fds) > 0) + log_warning("Got extra auxiliary fds with notification message, closing them."); + + return 0; +} + +static void invoke_sigchld_event(Manager *m, Unit *u, const siginfo_t *si) { + uint64_t iteration; + + assert(m); + assert(u); + assert(si); + + sd_event_get_iteration(m->event, &iteration); + + log_unit_debug(u, "Child "PID_FMT" belongs to %s", si->si_pid, u->id); + + unit_unwatch_pid(u, si->si_pid); + + if (UNIT_VTABLE(u)->sigchld_event) { + if (set_size(u->pids) <= 1 || + iteration != u->sigchldgen || + unit_main_pid(u) == si->si_pid || + unit_control_pid(u) == si->si_pid) { + UNIT_VTABLE(u)->sigchld_event(u, si->si_pid, si->si_code, si->si_status); + u->sigchldgen = iteration; + } else + log_debug("%s already issued a sigchld this iteration %" PRIu64 ", skipping. Pids still being watched %d", u->id, iteration, set_size(u->pids)); + } +} + +static int manager_dispatch_sigchld(Manager *m) { + assert(m); + + for (;;) { + siginfo_t si = {}; + + /* First we call waitd() for a PID and do not reap the + * zombie. That way we can still access /proc/$PID for + * it while it is a zombie. */ + if (waitid(P_ALL, 0, &si, WEXITED|WNOHANG|WNOWAIT) < 0) { + + if (errno == ECHILD) + break; + + if (errno == EINTR) + continue; + + return -errno; + } + + if (si.si_pid <= 0) + break; + + if (si.si_code == CLD_EXITED || si.si_code == CLD_KILLED || si.si_code == CLD_DUMPED) { + _cleanup_free_ char *name = NULL; + Unit *u1, *u2, *u3; + + get_process_comm(si.si_pid, &name); + + log_debug("Child "PID_FMT" (%s) died (code=%s, status=%i/%s)", + si.si_pid, strna(name), + sigchld_code_to_string(si.si_code), + si.si_status, + strna(si.si_code == CLD_EXITED + ? exit_status_to_string(si.si_status, EXIT_STATUS_FULL) + : signal_to_string(si.si_status))); + + /* And now figure out the unit this belongs + * to, it might be multiple... */ + u1 = manager_get_unit_by_pid_cgroup(m, si.si_pid); + if (u1) + invoke_sigchld_event(m, u1, &si); + u2 = hashmap_get(m->watch_pids1, PID_TO_PTR(si.si_pid)); + if (u2 && u2 != u1) + invoke_sigchld_event(m, u2, &si); + u3 = hashmap_get(m->watch_pids2, PID_TO_PTR(si.si_pid)); + if (u3 && u3 != u2 && u3 != u1) + invoke_sigchld_event(m, u3, &si); + } + + /* And now, we actually reap the zombie. */ + if (waitid(P_PID, si.si_pid, &si, WEXITED) < 0) { + if (errno == EINTR) + continue; + + return -errno; + } + } + + return 0; +} + +static int manager_start_target(Manager *m, const char *name, JobMode mode) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + log_debug("Activating special unit %s", name); + + r = manager_add_job_by_name(m, JOB_START, name, mode, &error, NULL); + if (r < 0) + log_error("Failed to enqueue %s job: %s", name, bus_error_message(&error, r)); + + return r; +} + +static void manager_handle_ctrl_alt_del(Manager *m) { + /* If the user presses C-A-D more than + * 7 times within 2s, we reboot/shutdown immediately, + * unless it was disabled in system.conf */ + + if (ratelimit_test(&m->ctrl_alt_del_ratelimit) || m->cad_burst_action == EMERGENCY_ACTION_NONE) + manager_start_target(m, SPECIAL_CTRL_ALT_DEL_TARGET, JOB_REPLACE_IRREVERSIBLY); + else + emergency_action(m, m->cad_burst_action, NULL, + "Ctrl-Alt-Del was pressed more than 7 times within 2s"); +} + +static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + ssize_t n; + struct signalfd_siginfo sfsi; + bool sigchld = false; + int r; + + assert(m); + assert(m->signal_fd == fd); + + if (revents != EPOLLIN) { + log_warning("Got unexpected events from signal file descriptor."); + return 0; + } + + for (;;) { + n = read(m->signal_fd, &sfsi, sizeof(sfsi)); + if (n != sizeof(sfsi)) { + if (n >= 0) { + log_warning("Truncated read from signal fd (%zu bytes)!", n); + return 0; + } + + if (IN_SET(errno, EINTR, EAGAIN)) + break; + + /* We return an error here, which will kill this handler, + * to avoid a busy loop on read error. */ + return log_error_errno(errno, "Reading from signal fd failed: %m"); + } + + log_received_signal(sfsi.ssi_signo == SIGCHLD || + (sfsi.ssi_signo == SIGTERM && MANAGER_IS_USER(m)) + ? LOG_DEBUG : LOG_INFO, + &sfsi); + + switch (sfsi.ssi_signo) { + + case SIGCHLD: + sigchld = true; + break; + + case SIGTERM: + if (MANAGER_IS_SYSTEM(m)) { + /* This is for compatibility with the + * original sysvinit */ + m->exit_code = MANAGER_REEXECUTE; + break; + } + + /* Fall through */ + + case SIGINT: + if (MANAGER_IS_SYSTEM(m)) { + manager_handle_ctrl_alt_del(m); + break; + } + + /* Run the exit target if there is one, if not, just exit. */ + if (manager_start_target(m, SPECIAL_EXIT_TARGET, JOB_REPLACE) < 0) { + m->exit_code = MANAGER_EXIT; + return 0; + } + + break; + + case SIGWINCH: + if (MANAGER_IS_SYSTEM(m)) + manager_start_target(m, SPECIAL_KBREQUEST_TARGET, JOB_REPLACE); + + /* This is a nop on non-init */ + break; + + case SIGPWR: + if (MANAGER_IS_SYSTEM(m)) + manager_start_target(m, SPECIAL_SIGPWR_TARGET, JOB_REPLACE); + + /* This is a nop on non-init */ + break; + + case SIGUSR1: { + Unit *u; + + u = manager_get_unit(m, SPECIAL_DBUS_SERVICE); + + if (!u || UNIT_IS_ACTIVE_OR_RELOADING(unit_active_state(u))) { + log_info("Trying to reconnect to bus..."); + bus_init(m, true); + } + + if (!u || !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) { + log_info("Loading D-Bus service..."); + manager_start_target(m, SPECIAL_DBUS_SERVICE, JOB_REPLACE); + } + + break; + } + + case SIGUSR2: { + _cleanup_free_ char *dump = NULL; + _cleanup_fclose_ FILE *f = NULL; + size_t size; + + f = open_memstream(&dump, &size); + if (!f) { + log_warning_errno(errno, "Failed to allocate memory stream: %m"); + break; + } + + manager_dump_units(m, f, "\t"); + manager_dump_jobs(m, f, "\t"); + + r = fflush_and_check(f); + if (r < 0) { + log_warning_errno(r, "Failed to write status stream: %m"); + break; + } + + log_dump(LOG_INFO, dump); + break; + } + + case SIGHUP: + m->exit_code = MANAGER_RELOAD; + break; + + default: { + + /* Starting SIGRTMIN+0 */ + static const char * const target_table[] = { + [0] = SPECIAL_DEFAULT_TARGET, + [1] = SPECIAL_RESCUE_TARGET, + [2] = SPECIAL_EMERGENCY_TARGET, + [3] = SPECIAL_HALT_TARGET, + [4] = SPECIAL_POWEROFF_TARGET, + [5] = SPECIAL_REBOOT_TARGET, + [6] = SPECIAL_KEXEC_TARGET + }; + + /* Starting SIGRTMIN+13, so that target halt and system halt are 10 apart */ + static const ManagerExitCode code_table[] = { + [0] = MANAGER_HALT, + [1] = MANAGER_POWEROFF, + [2] = MANAGER_REBOOT, + [3] = MANAGER_KEXEC + }; + + if ((int) sfsi.ssi_signo >= SIGRTMIN+0 && + (int) sfsi.ssi_signo < SIGRTMIN+(int) ELEMENTSOF(target_table)) { + int idx = (int) sfsi.ssi_signo - SIGRTMIN; + manager_start_target(m, target_table[idx], + (idx == 1 || idx == 2) ? JOB_ISOLATE : JOB_REPLACE); + break; + } + + if ((int) sfsi.ssi_signo >= SIGRTMIN+13 && + (int) sfsi.ssi_signo < SIGRTMIN+13+(int) ELEMENTSOF(code_table)) { + m->exit_code = code_table[sfsi.ssi_signo - SIGRTMIN - 13]; + break; + } + + switch (sfsi.ssi_signo - SIGRTMIN) { + + case 20: + manager_set_show_status(m, SHOW_STATUS_YES); + break; + + case 21: + manager_set_show_status(m, SHOW_STATUS_NO); + break; + + case 22: + log_set_max_level(LOG_DEBUG); + log_info("Setting log level to debug."); + break; + + case 23: + log_set_max_level(LOG_INFO); + log_info("Setting log level to info."); + break; + + case 24: + if (MANAGER_IS_USER(m)) { + m->exit_code = MANAGER_EXIT; + return 0; + } + + /* This is a nop on init */ + break; + + case 26: + case 29: /* compatibility: used to be mapped to LOG_TARGET_SYSLOG_OR_KMSG */ + log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); + log_notice("Setting log target to journal-or-kmsg."); + break; + + case 27: + log_set_target(LOG_TARGET_CONSOLE); + log_notice("Setting log target to console."); + break; + + case 28: + log_set_target(LOG_TARGET_KMSG); + log_notice("Setting log target to kmsg."); + break; + + default: + log_warning("Got unhandled signal <%s>.", signal_to_string(sfsi.ssi_signo)); + } + } + } + } + + if (sigchld) + manager_dispatch_sigchld(m); + + return 0; +} + +static int manager_dispatch_time_change_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + Iterator i; + Unit *u; + + assert(m); + assert(m->time_change_fd == fd); + + log_struct(LOG_INFO, + LOG_MESSAGE_ID(SD_MESSAGE_TIME_CHANGE), + LOG_MESSAGE("Time has been changed"), + NULL); + + /* Restart the watch */ + m->time_change_event_source = sd_event_source_unref(m->time_change_event_source); + m->time_change_fd = safe_close(m->time_change_fd); + + manager_setup_time_change(m); + + HASHMAP_FOREACH(u, m->units, i) + if (UNIT_VTABLE(u)->time_change) + UNIT_VTABLE(u)->time_change(u); + + return 0; +} + +static int manager_dispatch_idle_pipe_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + + assert(m); + assert(m->idle_pipe[2] == fd); + + m->no_console_output = m->n_on_console > 0; + + manager_close_idle_pipe(m); + + return 0; +} + +static int manager_dispatch_jobs_in_progress(sd_event_source *source, usec_t usec, void *userdata) { + Manager *m = userdata; + int r; + uint64_t next; + + assert(m); + assert(source); + + manager_print_jobs_in_progress(m); + + next = now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_PERIOD_USEC; + r = sd_event_source_set_time(source, next); + if (r < 0) + return r; + + return sd_event_source_set_enabled(source, SD_EVENT_ONESHOT); +} + +int manager_loop(Manager *m) { + int r; + + RATELIMIT_DEFINE(rl, 1*USEC_PER_SEC, 50000); + + assert(m); + m->exit_code = MANAGER_OK; + + /* Release the path cache */ + m->unit_path_cache = set_free_free(m->unit_path_cache); + + manager_check_finished(m); + + /* There might still be some zombies hanging around from + * before we were exec()'ed. Let's reap them. */ + r = manager_dispatch_sigchld(m); + if (r < 0) + return r; + + while (m->exit_code == MANAGER_OK) { + usec_t wait_usec; + + if (m->runtime_watchdog > 0 && m->runtime_watchdog != USEC_INFINITY && MANAGER_IS_SYSTEM(m)) + watchdog_ping(); + + if (!ratelimit_test(&rl)) { + /* Yay, something is going seriously wrong, pause a little */ + log_warning("Looping too fast. Throttling execution a little."); + sleep(1); + } + + if (manager_dispatch_load_queue(m) > 0) + continue; + + if (manager_dispatch_gc_queue(m) > 0) + continue; + + if (manager_dispatch_cleanup_queue(m) > 0) + continue; + + if (manager_dispatch_cgroup_queue(m) > 0) + continue; + + if (manager_dispatch_dbus_queue(m) > 0) + continue; + + /* Sleep for half the watchdog time */ + if (m->runtime_watchdog > 0 && m->runtime_watchdog != USEC_INFINITY && MANAGER_IS_SYSTEM(m)) { + wait_usec = m->runtime_watchdog / 2; + if (wait_usec <= 0) + wait_usec = 1; + } else + wait_usec = USEC_INFINITY; + + r = sd_event_run(m->event, wait_usec); + if (r < 0) + return log_error_errno(r, "Failed to run event loop: %m"); + } + + return m->exit_code; +} + +int manager_load_unit_from_dbus_path(Manager *m, const char *s, sd_bus_error *e, Unit **_u) { + _cleanup_free_ char *n = NULL; + sd_id128_t invocation_id; + Unit *u; + int r; + + assert(m); + assert(s); + assert(_u); + + r = unit_name_from_dbus_path(s, &n); + if (r < 0) + return r; + + /* Permit addressing units by invocation ID: if the passed bus path is suffixed by a 128bit ID then we use it + * as invocation ID. */ + r = sd_id128_from_string(n, &invocation_id); + if (r >= 0) { + u = hashmap_get(m->units_by_invocation_id, &invocation_id); + if (u) { + *_u = u; + return 0; + } + + return sd_bus_error_setf(e, BUS_ERROR_NO_UNIT_FOR_INVOCATION_ID, "No unit with the specified invocation ID " SD_ID128_FORMAT_STR " known.", SD_ID128_FORMAT_VAL(invocation_id)); + } + + /* If this didn't work, we use the suffix as unit name. */ + r = manager_load_unit(m, n, NULL, e, &u); + if (r < 0) + return r; + + *_u = u; + return 0; +} + +int manager_get_job_from_dbus_path(Manager *m, const char *s, Job **_j) { + const char *p; + unsigned id; + Job *j; + int r; + + assert(m); + assert(s); + assert(_j); + + p = startswith(s, "/org/freedesktop/systemd1/job/"); + if (!p) + return -EINVAL; + + r = safe_atou(p, &id); + if (r < 0) + return r; + + j = manager_get_job(m, id); + if (!j) + return -ENOENT; + + *_j = j; + + return 0; +} + +void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success) { + +#ifdef HAVE_AUDIT + _cleanup_free_ char *p = NULL; + const char *msg; + int audit_fd, r; + + if (!MANAGER_IS_SYSTEM(m)) + return; + + audit_fd = get_audit_fd(); + if (audit_fd < 0) + return; + + /* Don't generate audit events if the service was already + * started and we're just deserializing */ + if (MANAGER_IS_RELOADING(m)) + return; + + if (u->type != UNIT_SERVICE) + return; + + r = unit_name_to_prefix_and_instance(u->id, &p); + if (r < 0) { + log_error_errno(r, "Failed to extract prefix and instance of unit name: %m"); + return; + } + + msg = strjoina("unit=", p); + if (audit_log_user_comm_message(audit_fd, type, msg, "systemd", NULL, NULL, NULL, success) < 0) { + if (errno == EPERM) + /* We aren't allowed to send audit messages? + * Then let's not retry again. */ + close_audit_fd(); + else + log_warning_errno(errno, "Failed to send audit message: %m"); + } +#endif + +} + +void manager_send_unit_plymouth(Manager *m, Unit *u) { + static const union sockaddr_union sa = PLYMOUTH_SOCKET; + _cleanup_free_ char *message = NULL; + _cleanup_close_ int fd = -1; + int n = 0; + + /* Don't generate plymouth events if the service was already + * started and we're just deserializing */ + if (MANAGER_IS_RELOADING(m)) + return; + + if (!MANAGER_IS_SYSTEM(m)) + return; + + if (detect_container() > 0) + return; + + if (u->type != UNIT_SERVICE && + u->type != UNIT_MOUNT && + u->type != UNIT_SWAP) + return; + + /* We set SOCK_NONBLOCK here so that we rather drop the + * message then wait for plymouth */ + fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); + if (fd < 0) { + log_error_errno(errno, "socket() failed: %m"); + return; + } + + if (connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)) < 0) { + + if (!IN_SET(errno, EPIPE, EAGAIN, ENOENT, ECONNREFUSED, ECONNRESET, ECONNABORTED)) + log_error_errno(errno, "connect() failed: %m"); + return; + } + + if (asprintf(&message, "U\002%c%s%n", (int) (strlen(u->id) + 1), u->id, &n) < 0) { + log_oom(); + return; + } + + errno = 0; + if (write(fd, message, n + 1) != n + 1) + if (!IN_SET(errno, EPIPE, EAGAIN, ENOENT, ECONNREFUSED, ECONNRESET, ECONNABORTED)) + log_error_errno(errno, "Failed to write Plymouth message: %m"); +} + +int manager_open_serialization(Manager *m, FILE **_f) { + const char *path; + int fd = -1; + FILE *f; + + assert(_f); + + path = MANAGER_IS_SYSTEM(m) ? "/run/systemd" : "/tmp"; + fd = open_tmpfile_unlinkable(path, O_RDWR|O_CLOEXEC); + if (fd < 0) + return -errno; + + log_debug("Serializing state to %s", path); + + f = fdopen(fd, "w+"); + if (!f) { + safe_close(fd); + return -errno; + } + + *_f = f; + + return 0; +} + +int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { + Iterator i; + Unit *u; + const char *t; + char **e; + int r; + + assert(m); + assert(f); + assert(fds); + + m->n_reloading++; + + fprintf(f, "current-job-id=%"PRIu32"\n", m->current_job_id); + fprintf(f, "taint-usr=%s\n", yes_no(m->taint_usr)); + fprintf(f, "n-installed-jobs=%u\n", m->n_installed_jobs); + fprintf(f, "n-failed-jobs=%u\n", m->n_failed_jobs); + + dual_timestamp_serialize(f, "firmware-timestamp", &m->firmware_timestamp); + dual_timestamp_serialize(f, "loader-timestamp", &m->loader_timestamp); + dual_timestamp_serialize(f, "kernel-timestamp", &m->kernel_timestamp); + dual_timestamp_serialize(f, "initrd-timestamp", &m->initrd_timestamp); + + if (!in_initrd()) { + dual_timestamp_serialize(f, "userspace-timestamp", &m->userspace_timestamp); + dual_timestamp_serialize(f, "finish-timestamp", &m->finish_timestamp); + dual_timestamp_serialize(f, "security-start-timestamp", &m->security_start_timestamp); + dual_timestamp_serialize(f, "security-finish-timestamp", &m->security_finish_timestamp); + dual_timestamp_serialize(f, "generators-start-timestamp", &m->generators_start_timestamp); + dual_timestamp_serialize(f, "generators-finish-timestamp", &m->generators_finish_timestamp); + dual_timestamp_serialize(f, "units-load-start-timestamp", &m->units_load_start_timestamp); + dual_timestamp_serialize(f, "units-load-finish-timestamp", &m->units_load_finish_timestamp); + } + + if (!switching_root) { + STRV_FOREACH(e, m->environment) { + _cleanup_free_ char *ce; + + ce = cescape(*e); + if (!ce) + return -ENOMEM; + + fprintf(f, "env=%s\n", *e); + } + } + + if (m->notify_fd >= 0) { + int copy; + + copy = fdset_put_dup(fds, m->notify_fd); + if (copy < 0) + return copy; + + fprintf(f, "notify-fd=%i\n", copy); + fprintf(f, "notify-socket=%s\n", m->notify_socket); + } + + if (m->cgroups_agent_fd >= 0) { + int copy; + + copy = fdset_put_dup(fds, m->cgroups_agent_fd); + if (copy < 0) + return copy; + + fprintf(f, "cgroups-agent-fd=%i\n", copy); + } + + if (m->user_lookup_fds[0] >= 0) { + int copy0, copy1; + + copy0 = fdset_put_dup(fds, m->user_lookup_fds[0]); + if (copy0 < 0) + return copy0; + + copy1 = fdset_put_dup(fds, m->user_lookup_fds[1]); + if (copy1 < 0) + return copy1; + + fprintf(f, "user-lookup=%i %i\n", copy0, copy1); + } + + bus_track_serialize(m->subscribed, f, "subscribed"); + + r = dynamic_user_serialize(m, f, fds); + if (r < 0) + return r; + + manager_serialize_uid_refs(m, f); + manager_serialize_gid_refs(m, f); + + fputc('\n', f); + + HASHMAP_FOREACH_KEY(u, t, m->units, i) { + if (u->id != t) + continue; + + /* Start marker */ + fputs(u->id, f); + fputc('\n', f); + + r = unit_serialize(u, f, fds, !switching_root); + if (r < 0) { + m->n_reloading--; + return r; + } + } + + assert(m->n_reloading > 0); + m->n_reloading--; + + if (ferror(f)) + return -EIO; + + r = bus_fdset_add_all(m, fds); + if (r < 0) + return r; + + return 0; +} + +int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { + int r = 0; + + assert(m); + assert(f); + + log_debug("Deserializing state..."); + + m->n_reloading++; + + for (;;) { + char line[LINE_MAX], *l; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + r = 0; + else + r = -errno; + + goto finish; + } + + char_array_0(line); + l = strstrip(line); + + if (l[0] == 0) + break; + + if (startswith(l, "current-job-id=")) { + uint32_t id; + + if (safe_atou32(l+15, &id) < 0) + log_debug("Failed to parse current job id value %s", l+15); + else + m->current_job_id = MAX(m->current_job_id, id); + + } else if (startswith(l, "n-installed-jobs=")) { + uint32_t n; + + if (safe_atou32(l+17, &n) < 0) + log_debug("Failed to parse installed jobs counter %s", l+17); + else + m->n_installed_jobs += n; + + } else if (startswith(l, "n-failed-jobs=")) { + uint32_t n; + + if (safe_atou32(l+14, &n) < 0) + log_debug("Failed to parse failed jobs counter %s", l+14); + else + m->n_failed_jobs += n; + + } else if (startswith(l, "taint-usr=")) { + int b; + + b = parse_boolean(l+10); + if (b < 0) + log_debug("Failed to parse taint /usr flag %s", l+10); + else + m->taint_usr = m->taint_usr || b; + + } else if (startswith(l, "firmware-timestamp=")) + dual_timestamp_deserialize(l+19, &m->firmware_timestamp); + else if (startswith(l, "loader-timestamp=")) + dual_timestamp_deserialize(l+17, &m->loader_timestamp); + else if (startswith(l, "kernel-timestamp=")) + dual_timestamp_deserialize(l+17, &m->kernel_timestamp); + else if (startswith(l, "initrd-timestamp=")) + dual_timestamp_deserialize(l+17, &m->initrd_timestamp); + else if (startswith(l, "userspace-timestamp=")) + dual_timestamp_deserialize(l+20, &m->userspace_timestamp); + else if (startswith(l, "finish-timestamp=")) + dual_timestamp_deserialize(l+17, &m->finish_timestamp); + else if (startswith(l, "security-start-timestamp=")) + dual_timestamp_deserialize(l+25, &m->security_start_timestamp); + else if (startswith(l, "security-finish-timestamp=")) + dual_timestamp_deserialize(l+26, &m->security_finish_timestamp); + else if (startswith(l, "generators-start-timestamp=")) + dual_timestamp_deserialize(l+27, &m->generators_start_timestamp); + else if (startswith(l, "generators-finish-timestamp=")) + dual_timestamp_deserialize(l+28, &m->generators_finish_timestamp); + else if (startswith(l, "units-load-start-timestamp=")) + dual_timestamp_deserialize(l+27, &m->units_load_start_timestamp); + else if (startswith(l, "units-load-finish-timestamp=")) + dual_timestamp_deserialize(l+28, &m->units_load_finish_timestamp); + else if (startswith(l, "env=")) { + _cleanup_free_ char *uce = NULL; + char **e; + + r = cunescape(l + 4, UNESCAPE_RELAX, &uce); + if (r < 0) + goto finish; + + e = strv_env_set(m->environment, uce); + if (!e) { + r = -ENOMEM; + goto finish; + } + + strv_free(m->environment); + m->environment = e; + + } else if (startswith(l, "notify-fd=")) { + int fd; + + if (safe_atoi(l + 10, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_debug("Failed to parse notify fd: %s", l + 10); + else { + m->notify_event_source = sd_event_source_unref(m->notify_event_source); + safe_close(m->notify_fd); + m->notify_fd = fdset_remove(fds, fd); + } + + } else if (startswith(l, "notify-socket=")) { + char *n; + + n = strdup(l+14); + if (!n) { + r = -ENOMEM; + goto finish; + } + + free(m->notify_socket); + m->notify_socket = n; + + } else if (startswith(l, "cgroups-agent-fd=")) { + int fd; + + if (safe_atoi(l + 17, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_debug("Failed to parse cgroups agent fd: %s", l + 10); + else { + m->cgroups_agent_event_source = sd_event_source_unref(m->cgroups_agent_event_source); + safe_close(m->cgroups_agent_fd); + m->cgroups_agent_fd = fdset_remove(fds, fd); + } + + } else if (startswith(l, "user-lookup=")) { + int fd0, fd1; + + if (sscanf(l + 12, "%i %i", &fd0, &fd1) != 2 || fd0 < 0 || fd1 < 0 || fd0 == fd1 || !fdset_contains(fds, fd0) || !fdset_contains(fds, fd1)) + log_debug("Failed to parse user lookup fd: %s", l + 12); + else { + m->user_lookup_event_source = sd_event_source_unref(m->user_lookup_event_source); + safe_close_pair(m->user_lookup_fds); + m->user_lookup_fds[0] = fdset_remove(fds, fd0); + m->user_lookup_fds[1] = fdset_remove(fds, fd1); + } + + } else if (startswith(l, "dynamic-user=")) + dynamic_user_deserialize_one(m, l + 13, fds); + else if (startswith(l, "destroy-ipc-uid=")) + manager_deserialize_uid_refs_one(m, l + 16); + else if (startswith(l, "destroy-ipc-gid=")) + manager_deserialize_gid_refs_one(m, l + 16); + else if (startswith(l, "subscribed=")) { + + if (strv_extend(&m->deserialized_subscribed, l+11) < 0) + log_oom(); + + } else if (!startswith(l, "kdbus-fd=")) /* ignore this one */ + log_debug("Unknown serialization item '%s'", l); + } + + for (;;) { + Unit *u; + char name[UNIT_NAME_MAX+2]; + + /* Start marker */ + if (!fgets(name, sizeof(name), f)) { + if (feof(f)) + r = 0; + else + r = -errno; + + goto finish; + } + + char_array_0(name); + + r = manager_load_unit(m, strstrip(name), NULL, NULL, &u); + if (r < 0) + goto finish; + + r = unit_deserialize(u, f, fds); + if (r < 0) + goto finish; + } + +finish: + if (ferror(f)) + r = -EIO; + + assert(m->n_reloading > 0); + m->n_reloading--; + + return r; +} + +int manager_reload(Manager *m) { + int r, q; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_fdset_free_ FDSet *fds = NULL; + + assert(m); + + r = manager_open_serialization(m, &f); + if (r < 0) + return r; + + m->n_reloading++; + bus_manager_send_reloading(m, true); + + fds = fdset_new(); + if (!fds) { + m->n_reloading--; + return -ENOMEM; + } + + r = manager_serialize(m, f, fds, false); + if (r < 0) { + m->n_reloading--; + return r; + } + + if (fseeko(f, 0, SEEK_SET) < 0) { + m->n_reloading--; + return -errno; + } + + /* From here on there is no way back. */ + manager_clear_jobs_and_units(m); + lookup_paths_flush_generator(&m->lookup_paths); + lookup_paths_free(&m->lookup_paths); + dynamic_user_vacuum(m, false); + m->uid_refs = hashmap_free(m->uid_refs); + m->gid_refs = hashmap_free(m->gid_refs); + + q = lookup_paths_init(&m->lookup_paths, m->unit_file_scope, 0, NULL); + if (q < 0 && r >= 0) + r = q; + + /* Find new unit paths */ + q = manager_run_generators(m); + if (q < 0 && r >= 0) + r = q; + + lookup_paths_reduce(&m->lookup_paths); + manager_build_unit_path_cache(m); + + /* First, enumerate what we can from all config files */ + manager_enumerate(m); + + /* Second, deserialize our stored data */ + q = manager_deserialize(m, f, fds); + if (q < 0 && r >= 0) + r = q; + + fclose(f); + f = NULL; + + /* Re-register notify_fd as event source */ + q = manager_setup_notify(m); + if (q < 0 && r >= 0) + r = q; + + q = manager_setup_cgroups_agent(m); + if (q < 0 && r >= 0) + r = q; + + q = manager_setup_user_lookup_fd(m); + if (q < 0 && r >= 0) + r = q; + + /* Third, fire things up! */ + manager_coldplug(m); + + /* Release any dynamic users no longer referenced */ + dynamic_user_vacuum(m, true); + + /* Release any references to UIDs/GIDs no longer referenced, and destroy any IPC owned by them */ + manager_vacuum_uid_refs(m); + manager_vacuum_gid_refs(m); + + /* Sync current state of bus names with our set of listening units */ + if (m->api_bus) + manager_sync_bus_names(m, m->api_bus); + + assert(m->n_reloading > 0); + m->n_reloading--; + + m->send_reloading_done = true; + + return r; +} + +void manager_reset_failed(Manager *m) { + Unit *u; + Iterator i; + + assert(m); + + HASHMAP_FOREACH(u, m->units, i) + unit_reset_failed(u); +} + +bool manager_unit_inactive_or_pending(Manager *m, const char *name) { + Unit *u; + + assert(m); + assert(name); + + /* Returns true if the unit is inactive or going down */ + u = manager_get_unit(m, name); + if (!u) + return true; + + return unit_inactive_or_pending(u); +} + +static void manager_notify_finished(Manager *m) { + char userspace[FORMAT_TIMESPAN_MAX], initrd[FORMAT_TIMESPAN_MAX], kernel[FORMAT_TIMESPAN_MAX], sum[FORMAT_TIMESPAN_MAX]; + usec_t firmware_usec, loader_usec, kernel_usec, initrd_usec, userspace_usec, total_usec; + + if (m->test_run) + return; + + if (MANAGER_IS_SYSTEM(m) && detect_container() <= 0) { + + /* Note that m->kernel_usec.monotonic is always at 0, + * and m->firmware_usec.monotonic and + * m->loader_usec.monotonic should be considered + * negative values. */ + + firmware_usec = m->firmware_timestamp.monotonic - m->loader_timestamp.monotonic; + loader_usec = m->loader_timestamp.monotonic - m->kernel_timestamp.monotonic; + userspace_usec = m->finish_timestamp.monotonic - m->userspace_timestamp.monotonic; + total_usec = m->firmware_timestamp.monotonic + m->finish_timestamp.monotonic; + + if (dual_timestamp_is_set(&m->initrd_timestamp)) { + + kernel_usec = m->initrd_timestamp.monotonic - m->kernel_timestamp.monotonic; + initrd_usec = m->userspace_timestamp.monotonic - m->initrd_timestamp.monotonic; + + log_struct(LOG_INFO, + LOG_MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "KERNEL_USEC="USEC_FMT, kernel_usec, + "INITRD_USEC="USEC_FMT, initrd_usec, + "USERSPACE_USEC="USEC_FMT, userspace_usec, + LOG_MESSAGE("Startup finished in %s (kernel) + %s (initrd) + %s (userspace) = %s.", + format_timespan(kernel, sizeof(kernel), kernel_usec, USEC_PER_MSEC), + format_timespan(initrd, sizeof(initrd), initrd_usec, USEC_PER_MSEC), + format_timespan(userspace, sizeof(userspace), userspace_usec, USEC_PER_MSEC), + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC)), + NULL); + } else { + kernel_usec = m->userspace_timestamp.monotonic - m->kernel_timestamp.monotonic; + initrd_usec = 0; + + log_struct(LOG_INFO, + LOG_MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "KERNEL_USEC="USEC_FMT, kernel_usec, + "USERSPACE_USEC="USEC_FMT, userspace_usec, + LOG_MESSAGE("Startup finished in %s (kernel) + %s (userspace) = %s.", + format_timespan(kernel, sizeof(kernel), kernel_usec, USEC_PER_MSEC), + format_timespan(userspace, sizeof(userspace), userspace_usec, USEC_PER_MSEC), + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC)), + NULL); + } + } else { + firmware_usec = loader_usec = initrd_usec = kernel_usec = 0; + total_usec = userspace_usec = m->finish_timestamp.monotonic - m->userspace_timestamp.monotonic; + + log_struct(LOG_INFO, + LOG_MESSAGE_ID(SD_MESSAGE_STARTUP_FINISHED), + "USERSPACE_USEC="USEC_FMT, userspace_usec, + LOG_MESSAGE("Startup finished in %s.", + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC)), + NULL); + } + + bus_manager_send_finished(m, firmware_usec, loader_usec, kernel_usec, initrd_usec, userspace_usec, total_usec); + + sd_notifyf(false, + "READY=1\n" + "STATUS=Startup finished in %s.", + format_timespan(sum, sizeof(sum), total_usec, USEC_PER_MSEC)); +} + +void manager_check_finished(Manager *m) { + assert(m); + + if (MANAGER_IS_RELOADING(m)) + return; + + /* Verify that we are actually running currently. Initially + * the exit code is set to invalid, and during operation it is + * then set to MANAGER_OK */ + if (m->exit_code != MANAGER_OK) + return; + + if (hashmap_size(m->jobs) > 0) { + if (m->jobs_in_progress_event_source) + /* Ignore any failure, this is only for feedback */ + (void) sd_event_source_set_time(m->jobs_in_progress_event_source, now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_WAIT_USEC); + + return; + } + + manager_flip_auto_status(m, false); + + /* Notify Type=idle units that we are done now */ + manager_close_idle_pipe(m); + + /* Turn off confirm spawn now */ + m->confirm_spawn = false; + + /* No need to update ask password status when we're going non-interactive */ + manager_close_ask_password(m); + + /* This is no longer the first boot */ + manager_set_first_boot(m, false); + + if (dual_timestamp_is_set(&m->finish_timestamp)) + return; + + dual_timestamp_get(&m->finish_timestamp); + + manager_notify_finished(m); + + manager_invalidate_startup_units(m); +} + +static int manager_run_generators(Manager *m) { + _cleanup_strv_free_ char **paths = NULL; + const char *argv[5]; + char **path; + int r; + + assert(m); + + if (m->test_run) + return 0; + + paths = generator_binary_paths(m->unit_file_scope); + if (!paths) + return log_oom(); + + /* Optimize by skipping the whole process by not creating output directories + * if no generators are found. */ + STRV_FOREACH(path, paths) { + if (access(*path, F_OK) >= 0) + goto found; + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open generator directory %s: %m", *path); + } + + return 0; + + found: + r = lookup_paths_mkdir_generator(&m->lookup_paths); + if (r < 0) + goto finish; + + argv[0] = NULL; /* Leave this empty, execute_directory() will fill something in */ + argv[1] = m->lookup_paths.generator; + argv[2] = m->lookup_paths.generator_early; + argv[3] = m->lookup_paths.generator_late; + argv[4] = NULL; + + RUN_WITH_UMASK(0022) + execute_directories((const char* const*) paths, DEFAULT_TIMEOUT_USEC, (char**) argv); + +finish: + lookup_paths_trim_generator(&m->lookup_paths); + return r; +} + +int manager_environment_add(Manager *m, char **minus, char **plus) { + char **a = NULL, **b = NULL, **l; + assert(m); + + l = m->environment; + + if (!strv_isempty(minus)) { + a = strv_env_delete(l, 1, minus); + if (!a) + return -ENOMEM; + + l = a; + } + + if (!strv_isempty(plus)) { + b = strv_env_merge(2, l, plus); + if (!b) { + strv_free(a); + return -ENOMEM; + } + + l = b; + } + + if (m->environment != l) + strv_free(m->environment); + if (a != l) + strv_free(a); + if (b != l) + strv_free(b); + + m->environment = l; + manager_clean_environment(m); + strv_sort(m->environment); + + return 0; +} + +int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit) { + int i; + + assert(m); + + for (i = 0; i < _RLIMIT_MAX; i++) { + m->rlimit[i] = mfree(m->rlimit[i]); + + if (!default_rlimit[i]) + continue; + + m->rlimit[i] = newdup(struct rlimit, default_rlimit[i], 1); + if (!m->rlimit[i]) + return log_oom(); + } + + return 0; +} + +void manager_recheck_journal(Manager *m) { + Unit *u; + + assert(m); + + if (!MANAGER_IS_SYSTEM(m)) + return; + + u = manager_get_unit(m, SPECIAL_JOURNALD_SOCKET); + if (u && SOCKET(u)->state != SOCKET_RUNNING) { + log_close_journal(); + return; + } + + u = manager_get_unit(m, SPECIAL_JOURNALD_SERVICE); + if (u && SERVICE(u)->state != SERVICE_RUNNING) { + log_close_journal(); + return; + } + + /* Hmm, OK, so the socket is fully up and the service is up + * too, then let's make use of the thing. */ + log_open(); +} + +void manager_set_show_status(Manager *m, ShowStatus mode) { + assert(m); + assert(IN_SET(mode, SHOW_STATUS_AUTO, SHOW_STATUS_NO, SHOW_STATUS_YES, SHOW_STATUS_TEMPORARY)); + + if (!MANAGER_IS_SYSTEM(m)) + return; + + if (m->show_status != mode) + log_debug("%s showing of status.", + mode == SHOW_STATUS_NO ? "Disabling" : "Enabling"); + m->show_status = mode; + + if (mode > 0) + (void) touch("/run/systemd/show-status"); + else + (void) unlink("/run/systemd/show-status"); +} + +static bool manager_get_show_status(Manager *m, StatusType type) { + assert(m); + + if (!MANAGER_IS_SYSTEM(m)) + return false; + + if (m->no_console_output) + return false; + + if (!IN_SET(manager_state(m), MANAGER_INITIALIZING, MANAGER_STARTING, MANAGER_STOPPING)) + return false; + + /* If we cannot find out the status properly, just proceed. */ + if (type != STATUS_TYPE_EMERGENCY && manager_check_ask_password(m) > 0) + return false; + + if (m->show_status > 0) + return true; + + return false; +} + +void manager_set_first_boot(Manager *m, bool b) { + assert(m); + + if (!MANAGER_IS_SYSTEM(m)) + return; + + if (m->first_boot != (int) b) { + if (b) + (void) touch("/run/systemd/first-boot"); + else + (void) unlink("/run/systemd/first-boot"); + } + + m->first_boot = b; +} + +void manager_status_printf(Manager *m, StatusType type, const char *status, const char *format, ...) { + va_list ap; + + /* If m is NULL, assume we're after shutdown and let the messages through. */ + + if (m && !manager_get_show_status(m, type)) + return; + + /* XXX We should totally drop the check for ephemeral here + * and thus effectively make 'Type=idle' pointless. */ + if (type == STATUS_TYPE_EPHEMERAL && m && m->n_on_console > 0) + return; + + va_start(ap, format); + status_vprintf(status, true, type == STATUS_TYPE_EPHEMERAL, format, ap); + va_end(ap); +} + +Set *manager_get_units_requiring_mounts_for(Manager *m, const char *path) { + char p[strlen(path)+1]; + + assert(m); + assert(path); + + strcpy(p, path); + path_kill_slashes(p); + + return hashmap_get(m->units_requiring_mounts_for, streq(p, "/") ? "" : p); +} + +const char *manager_get_runtime_prefix(Manager *m) { + assert(m); + + return MANAGER_IS_SYSTEM(m) ? + "/run" : + getenv("XDG_RUNTIME_DIR"); +} + +int manager_update_failed_units(Manager *m, Unit *u, bool failed) { + unsigned size; + int r; + + assert(m); + assert(u->manager == m); + + size = set_size(m->failed_units); + + if (failed) { + r = set_ensure_allocated(&m->failed_units, NULL); + if (r < 0) + return log_oom(); + + if (set_put(m->failed_units, u) < 0) + return log_oom(); + } else + (void) set_remove(m->failed_units, u); + + if (set_size(m->failed_units) != size) + bus_manager_send_change_signal(m); + + return 0; +} + +ManagerState manager_state(Manager *m) { + Unit *u; + + assert(m); + + /* Did we ever finish booting? If not then we are still starting up */ + if (!dual_timestamp_is_set(&m->finish_timestamp)) { + + u = manager_get_unit(m, SPECIAL_BASIC_TARGET); + if (!u || !UNIT_IS_ACTIVE_OR_RELOADING(unit_active_state(u))) + return MANAGER_INITIALIZING; + + return MANAGER_STARTING; + } + + /* Is the special shutdown target queued? If so, we are in shutdown state */ + u = manager_get_unit(m, SPECIAL_SHUTDOWN_TARGET); + if (u && u->job && IN_SET(u->job->type, JOB_START, JOB_RESTART, JOB_RELOAD_OR_START)) + return MANAGER_STOPPING; + + /* Are the rescue or emergency targets active or queued? If so we are in maintenance state */ + u = manager_get_unit(m, SPECIAL_RESCUE_TARGET); + if (u && (UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u)) || + (u->job && IN_SET(u->job->type, JOB_START, JOB_RESTART, JOB_RELOAD_OR_START)))) + return MANAGER_MAINTENANCE; + + u = manager_get_unit(m, SPECIAL_EMERGENCY_TARGET); + if (u && (UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u)) || + (u->job && IN_SET(u->job->type, JOB_START, JOB_RESTART, JOB_RELOAD_OR_START)))) + return MANAGER_MAINTENANCE; + + /* Are there any failed units? If so, we are in degraded mode */ + if (set_size(m->failed_units) > 0) + return MANAGER_DEGRADED; + + return MANAGER_RUNNING; +} + +#define DESTROY_IPC_FLAG (UINT32_C(1) << 31) + +static void manager_unref_uid_internal( + Manager *m, + Hashmap **uid_refs, + uid_t uid, + bool destroy_now, + int (*_clean_ipc)(uid_t uid)) { + + uint32_t c, n; + + assert(m); + assert(uid_refs); + assert(uid_is_valid(uid)); + assert(_clean_ipc); + + /* A generic implementation, covering both manager_unref_uid() and manager_unref_gid(), under the assumption + * that uid_t and gid_t are actually defined the same way, with the same validity rules. + * + * We store a hashmap where the UID/GID is they key and the value is a 32bit reference counter, whose highest + * bit is used as flag for marking UIDs/GIDs whose IPC objects to remove when the last reference to the UID/GID + * is dropped. The flag is set to on, once at least one reference from a unit where RemoveIPC= is set is added + * on a UID/GID. It is reset when the UID's/GID's reference counter drops to 0 again. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (uid == 0) /* We don't keep track of root, and will never destroy it */ + return; + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + + n = c & ~DESTROY_IPC_FLAG; + assert(n > 0); + n--; + + if (destroy_now && n == 0) { + hashmap_remove(*uid_refs, UID_TO_PTR(uid)); + + if (c & DESTROY_IPC_FLAG) { + log_debug("%s " UID_FMT " is no longer referenced, cleaning up its IPC.", + _clean_ipc == clean_ipc_by_uid ? "UID" : "GID", + uid); + (void) _clean_ipc(uid); + } + } else { + c = n | (c & DESTROY_IPC_FLAG); + assert_se(hashmap_update(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)) >= 0); + } +} + +void manager_unref_uid(Manager *m, uid_t uid, bool destroy_now) { + manager_unref_uid_internal(m, &m->uid_refs, uid, destroy_now, clean_ipc_by_uid); +} + +void manager_unref_gid(Manager *m, gid_t gid, bool destroy_now) { + manager_unref_uid_internal(m, &m->gid_refs, (uid_t) gid, destroy_now, clean_ipc_by_gid); +} + +static int manager_ref_uid_internal( + Manager *m, + Hashmap **uid_refs, + uid_t uid, + bool clean_ipc) { + + uint32_t c, n; + int r; + + assert(m); + assert(uid_refs); + assert(uid_is_valid(uid)); + + /* A generic implementation, covering both manager_ref_uid() and manager_ref_gid(), under the assumption + * that uid_t and gid_t are actually defined the same way, with the same validity rules. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (uid == 0) /* We don't keep track of root, and will never destroy it */ + return 0; + + r = hashmap_ensure_allocated(uid_refs, &trivial_hash_ops); + if (r < 0) + return r; + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + + n = c & ~DESTROY_IPC_FLAG; + n++; + + if (n & DESTROY_IPC_FLAG) /* check for overflow */ + return -EOVERFLOW; + + c = n | (c & DESTROY_IPC_FLAG) | (clean_ipc ? DESTROY_IPC_FLAG : 0); + + return hashmap_replace(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)); +} + +int manager_ref_uid(Manager *m, uid_t uid, bool clean_ipc) { + return manager_ref_uid_internal(m, &m->uid_refs, uid, clean_ipc); +} + +int manager_ref_gid(Manager *m, gid_t gid, bool clean_ipc) { + return manager_ref_uid_internal(m, &m->gid_refs, (uid_t) gid, clean_ipc); +} + +static void manager_vacuum_uid_refs_internal( + Manager *m, + Hashmap **uid_refs, + int (*_clean_ipc)(uid_t uid)) { + + Iterator i; + void *p, *k; + + assert(m); + assert(uid_refs); + assert(_clean_ipc); + + HASHMAP_FOREACH_KEY(p, k, *uid_refs, i) { + uint32_t c, n; + uid_t uid; + + uid = PTR_TO_UID(k); + c = PTR_TO_UINT32(p); + + n = c & ~DESTROY_IPC_FLAG; + if (n > 0) + continue; + + if (c & DESTROY_IPC_FLAG) { + log_debug("Found unreferenced %s " UID_FMT " after reload/reexec. Cleaning up.", + _clean_ipc == clean_ipc_by_uid ? "UID" : "GID", + uid); + (void) _clean_ipc(uid); + } + + assert_se(hashmap_remove(*uid_refs, k) == p); + } +} + +void manager_vacuum_uid_refs(Manager *m) { + manager_vacuum_uid_refs_internal(m, &m->uid_refs, clean_ipc_by_uid); +} + +void manager_vacuum_gid_refs(Manager *m) { + manager_vacuum_uid_refs_internal(m, &m->gid_refs, clean_ipc_by_gid); +} + +static void manager_serialize_uid_refs_internal( + Manager *m, + FILE *f, + Hashmap **uid_refs, + const char *field_name) { + + Iterator i; + void *p, *k; + + assert(m); + assert(f); + assert(uid_refs); + assert(field_name); + + /* Serialize the UID reference table. Or actually, just the IPC destruction flag of it, as the actual counter + * of it is better rebuild after a reload/reexec. */ + + HASHMAP_FOREACH_KEY(p, k, *uid_refs, i) { + uint32_t c; + uid_t uid; + + uid = PTR_TO_UID(k); + c = PTR_TO_UINT32(p); + + if (!(c & DESTROY_IPC_FLAG)) + continue; + + fprintf(f, "%s=" UID_FMT "\n", field_name, uid); + } +} + +void manager_serialize_uid_refs(Manager *m, FILE *f) { + manager_serialize_uid_refs_internal(m, f, &m->uid_refs, "destroy-ipc-uid"); +} + +void manager_serialize_gid_refs(Manager *m, FILE *f) { + manager_serialize_uid_refs_internal(m, f, &m->gid_refs, "destroy-ipc-gid"); +} + +static void manager_deserialize_uid_refs_one_internal( + Manager *m, + Hashmap** uid_refs, + const char *value) { + + uid_t uid; + uint32_t c; + int r; + + assert(m); + assert(uid_refs); + assert(value); + + r = parse_uid(value, &uid); + if (r < 0 || uid == 0) { + log_debug("Unable to parse UID reference serialization"); + return; + } + + r = hashmap_ensure_allocated(uid_refs, &trivial_hash_ops); + if (r < 0) { + log_oom(); + return; + } + + c = PTR_TO_UINT32(hashmap_get(*uid_refs, UID_TO_PTR(uid))); + if (c & DESTROY_IPC_FLAG) + return; + + c |= DESTROY_IPC_FLAG; + + r = hashmap_replace(*uid_refs, UID_TO_PTR(uid), UINT32_TO_PTR(c)); + if (r < 0) { + log_debug("Failed to add UID reference entry"); + return; + } +} + +void manager_deserialize_uid_refs_one(Manager *m, const char *value) { + manager_deserialize_uid_refs_one_internal(m, &m->uid_refs, value); +} + +void manager_deserialize_gid_refs_one(Manager *m, const char *value) { + manager_deserialize_uid_refs_one_internal(m, &m->gid_refs, value); +} + +int manager_dispatch_user_lookup_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + struct buffer { + uid_t uid; + gid_t gid; + char unit_name[UNIT_NAME_MAX+1]; + } _packed_ buffer; + + Manager *m = userdata; + ssize_t l; + size_t n; + Unit *u; + + assert_se(source); + assert_se(m); + + /* Invoked whenever a child process succeeded resolving its user/group to use and sent us the resulting UID/GID + * in a datagram. We parse the datagram here and pass it off to the unit, so that it can add a reference to the + * UID/GID so that it can destroy the UID/GID's IPC objects when the reference counter drops to 0. */ + + l = recv(fd, &buffer, sizeof(buffer), MSG_DONTWAIT); + if (l < 0) { + if (errno == EINTR || errno == EAGAIN) + return 0; + + return log_error_errno(errno, "Failed to read from user lookup fd: %m"); + } + + if ((size_t) l <= offsetof(struct buffer, unit_name)) { + log_warning("Received too short user lookup message, ignoring."); + return 0; + } + + if ((size_t) l > offsetof(struct buffer, unit_name) + UNIT_NAME_MAX) { + log_warning("Received too long user lookup message, ignoring."); + return 0; + } + + if (!uid_is_valid(buffer.uid) && !gid_is_valid(buffer.gid)) { + log_warning("Got user lookup message with invalid UID/GID pair, ignoring."); + return 0; + } + + n = (size_t) l - offsetof(struct buffer, unit_name); + if (memchr(buffer.unit_name, 0, n)) { + log_warning("Received lookup message with embedded NUL character, ignoring."); + return 0; + } + + buffer.unit_name[n] = 0; + u = manager_get_unit(m, buffer.unit_name); + if (!u) { + log_debug("Got user lookup message but unit doesn't exist, ignoring."); + return 0; + } + + log_unit_debug(u, "User lookup succeeded: uid=" UID_FMT " gid=" GID_FMT, buffer.uid, buffer.gid); + + unit_notify_user_lookup(u, buffer.uid, buffer.gid); + return 0; +} + +static const char *const manager_state_table[_MANAGER_STATE_MAX] = { + [MANAGER_INITIALIZING] = "initializing", + [MANAGER_STARTING] = "starting", + [MANAGER_RUNNING] = "running", + [MANAGER_DEGRADED] = "degraded", + [MANAGER_MAINTENANCE] = "maintenance", + [MANAGER_STOPPING] = "stopping", +}; + +DEFINE_STRING_TABLE_LOOKUP(manager_state, ManagerState); diff --git a/src/grp-system/libcore/src/mount-setup.c b/src/grp-system/libcore/src/mount-setup.c new file mode 100644 index 0000000000..46e6f71425 --- /dev/null +++ b/src/grp-system/libcore/src/mount-setup.c @@ -0,0 +1,421 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <ftw.h> +#include <stdlib.h> +#include <sys/mount.h> +#include <unistd.h> + +#include "core/mount-setup.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/label.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/mount-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/set.h" +#include "systemd-basic/smack-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/util.h" +#include "systemd-basic/virt.h" +#include "systemd-shared/dev-setup.h" +#include "systemd-shared/efivars.h" + +typedef enum MountMode { + MNT_NONE = 0, + MNT_FATAL = 1 << 0, + MNT_IN_CONTAINER = 1 << 1, +} MountMode; + +typedef struct MountPoint { + const char *what; + const char *where; + const char *type; + const char *options; + unsigned long flags; + bool (*condition_fn)(void); + MountMode mode; +} MountPoint; + +/* The first three entries we might need before SELinux is up. The + * fourth (securityfs) is needed by IMA to load a custom policy. The + * other ones we can delay until SELinux and IMA are loaded. When + * SMACK is enabled we need smackfs, too, so it's a fifth one. */ +#ifdef HAVE_SMACK +#define N_EARLY_MOUNT 5 +#else +#define N_EARLY_MOUNT 4 +#endif + +static const MountPoint mount_table[] = { + { "sysfs", "/sys", "sysfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "devtmpfs", "/dev", "devtmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "securityfs", "/sys/kernel/security", "securityfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_NONE }, +#ifdef HAVE_SMACK + { "smackfs", "/sys/fs/smackfs", "smackfs", "smackfsdef=*", MS_NOSUID|MS_NOEXEC|MS_NODEV, + mac_smack_use, MNT_FATAL }, + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + mac_smack_use, MNT_FATAL }, +#endif + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "devpts", "/dev/pts", "devpts", "mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, + NULL, MNT_IN_CONTAINER }, +#ifdef HAVE_SMACK + { "tmpfs", "/run", "tmpfs", "mode=755,smackfsroot=*", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + mac_smack_use, MNT_FATAL }, +#endif + { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, + NULL, MNT_FATAL|MNT_IN_CONTAINER }, + { "cgroup", "/sys/fs/cgroup", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER }, + { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, + cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER }, + { "cgroup", "/sys/fs/cgroup/systemd", "cgroup2", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + cg_is_unified_systemd_controller_wanted, MNT_IN_CONTAINER }, + { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd,xattr", MS_NOSUID|MS_NOEXEC|MS_NODEV, + cg_is_legacy_systemd_controller_wanted, MNT_IN_CONTAINER }, + { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, + cg_is_legacy_systemd_controller_wanted, MNT_FATAL|MNT_IN_CONTAINER }, + { "pstore", "/sys/fs/pstore", "pstore", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + NULL, MNT_NONE }, +#ifdef ENABLE_EFI + { "efivarfs", "/sys/firmware/efi/efivars", "efivarfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, + is_efi_boot, MNT_NONE }, +#endif +}; + +/* These are API file systems that might be mounted by other software, + * we just list them here so that we know that we should ignore them */ + +static const char ignore_paths[] = + /* SELinux file systems */ + "/sys/fs/selinux\0" + /* Container bind mounts */ + "/proc/sys\0" + "/dev/console\0" + "/proc/kmsg\0"; + +bool mount_point_is_api(const char *path) { + unsigned i; + + /* Checks if this mount point is considered "API", and hence + * should be ignored */ + + for (i = 0; i < ELEMENTSOF(mount_table); i ++) + if (path_equal(path, mount_table[i].where)) + return true; + + return path_startswith(path, "/sys/fs/cgroup/"); +} + +bool mount_point_ignore(const char *path) { + const char *i; + + NULSTR_FOREACH(i, ignore_paths) + if (path_equal(path, i)) + return true; + + return false; +} + +static int mount_one(const MountPoint *p, bool relabel) { + int r; + + assert(p); + + if (p->condition_fn && !p->condition_fn()) + return 0; + + /* Relabel first, just in case */ + if (relabel) + (void) label_fix(p->where, true, true); + + r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW); + if (r < 0 && r != -ENOENT) { + log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, r, "Failed to determine whether %s is a mount point: %m", p->where); + return (p->mode & MNT_FATAL) ? r : 0; + } + if (r > 0) + return 0; + + /* Skip securityfs in a container */ + if (!(p->mode & MNT_IN_CONTAINER) && detect_container() > 0) + return 0; + + /* The access mode here doesn't really matter too much, since + * the mounted file system will take precedence anyway. */ + if (relabel) + (void) mkdir_p_label(p->where, 0755); + else + (void) mkdir_p(p->where, 0755); + + log_debug("Mounting %s to %s of type %s with options %s.", + p->what, + p->where, + p->type, + strna(p->options)); + + if (mount(p->what, + p->where, + p->type, + p->flags, + p->options) < 0) { + log_full_errno((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, errno, "Failed to mount %s at %s: %m", p->type, p->where); + return (p->mode & MNT_FATAL) ? -errno : 0; + } + + /* Relabel again, since we now mounted something fresh here */ + if (relabel) + (void) label_fix(p->where, false, false); + + return 1; +} + +static int mount_points_setup(unsigned n, bool loaded_policy) { + unsigned i; + int r = 0; + + for (i = 0; i < n; i ++) { + int j; + + j = mount_one(mount_table + i, loaded_policy); + if (j != 0 && r >= 0) + r = j; + } + + return r; +} + +int mount_setup_early(void) { + assert_cc(N_EARLY_MOUNT <= ELEMENTSOF(mount_table)); + + /* Do a minimal mount of /proc and friends to enable the most + * basic stuff, such as SELinux */ + return mount_points_setup(N_EARLY_MOUNT, false); +} + +int mount_cgroup_controllers(char ***join_controllers) { + _cleanup_set_free_free_ Set *controllers = NULL; + int r; + + if (!cg_is_legacy_wanted()) + return 0; + + /* Mount all available cgroup controllers that are built into the kernel. */ + + controllers = set_new(&string_hash_ops); + if (!controllers) + return log_oom(); + + r = cg_kernel_controllers(controllers); + if (r < 0) + return log_error_errno(r, "Failed to enumerate cgroup controllers: %m"); + + for (;;) { + _cleanup_free_ char *options = NULL, *controller = NULL, *where = NULL; + MountPoint p = { + .what = "cgroup", + .type = "cgroup", + .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV, + .mode = MNT_IN_CONTAINER, + }; + char ***k = NULL; + + controller = set_steal_first(controllers); + if (!controller) + break; + + if (join_controllers) + for (k = join_controllers; *k; k++) + if (strv_find(*k, controller)) + break; + + if (k && *k) { + char **i, **j; + + for (i = *k, j = *k; *i; i++) { + + if (!streq(*i, controller)) { + _cleanup_free_ char *t; + + t = set_remove(controllers, *i); + if (!t) { + free(*i); + continue; + } + } + + *(j++) = *i; + } + + *j = NULL; + + options = strv_join(*k, ","); + if (!options) + return log_oom(); + } else { + options = controller; + controller = NULL; + } + + where = strappend("/sys/fs/cgroup/", options); + if (!where) + return log_oom(); + + p.where = where; + p.options = options; + + r = mount_one(&p, true); + if (r < 0) + return r; + + if (r > 0 && k && *k) { + char **i; + + for (i = *k; *i; i++) { + _cleanup_free_ char *t = NULL; + + t = strappend("/sys/fs/cgroup/", *i); + if (!t) + return log_oom(); + + r = symlink(options, t); + if (r >= 0) { +#ifdef SMACK_RUN_LABEL + _cleanup_free_ char *src; + src = strappend("/sys/fs/cgroup/", options); + if (!src) + return log_oom(); + r = mac_smack_copy(t, src); + if (r < 0 && r != -EOPNOTSUPP) + return log_error_errno(r, "Failed to copy smack label from %s to %s: %m", src, t); +#endif + } else if (errno != EEXIST) + return log_error_errno(errno, "Failed to create symlink %s: %m", t); + } + } + } + + /* Now that we mounted everything, let's make the tmpfs the + * cgroup file systems are mounted into read-only. */ + (void) mount("tmpfs", "/sys/fs/cgroup", "tmpfs", MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755"); + + return 0; +} + +#if defined(HAVE_SELINUX) || defined(HAVE_SMACK) +static int nftw_cb( + const char *fpath, + const struct stat *sb, + int tflag, + struct FTW *ftwbuf) { + + /* No need to label /dev twice in a row... */ + if (_unlikely_(ftwbuf->level == 0)) + return FTW_CONTINUE; + + label_fix(fpath, false, false); + + /* /run/initramfs is static data and big, no need to + * dynamically relabel its contents at boot... */ + if (_unlikely_(ftwbuf->level == 1 && + tflag == FTW_D && + streq(fpath, "/run/initramfs"))) + return FTW_SKIP_SUBTREE; + + return FTW_CONTINUE; +}; +#endif + +int mount_setup(bool loaded_policy) { + int r = 0; + + r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy); + + if (r < 0) + return r; + +#if defined(HAVE_SELINUX) || defined(HAVE_SMACK) + /* Nodes in devtmpfs and /run need to be manually updated for + * the appropriate labels, after mounting. The other virtual + * API file systems like /sys and /proc do not need that, they + * use the same label for all their files. */ + if (loaded_policy) { + usec_t before_relabel, after_relabel; + char timespan[FORMAT_TIMESPAN_MAX]; + + before_relabel = now(CLOCK_MONOTONIC); + + nftw("/dev", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL); + nftw("/dev/shm", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL); + nftw("/run", nftw_cb, 64, FTW_MOUNT|FTW_PHYS|FTW_ACTIONRETVAL); + + after_relabel = now(CLOCK_MONOTONIC); + + log_info("Relabelled /dev and /run in %s.", + format_timespan(timespan, sizeof(timespan), after_relabel - before_relabel, 0)); + } +#endif + + /* Create a few default symlinks, which are normally created + * by udevd, but some scripts might need them before we start + * udevd. */ + dev_setup(NULL, UID_INVALID, GID_INVALID); + + /* Mark the root directory as shared in regards to mount + * propagation. The kernel defaults to "private", but we think + * it makes more sense to have a default of "shared" so that + * nspawn and the container tools work out of the box. If + * specific setups need other settings they can reset the + * propagation mode to private if needed. */ + if (detect_container() <= 0) + if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0) + log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m"); + + /* Create a few directories we always want around, Note that + * sd_booted() checks for /run/systemd/system, so this mkdir + * really needs to stay for good, otherwise software that + * copied sd-daemon.c into their sources will misdetect + * systemd. */ + (void) mkdir_label("/run/systemd", 0755); + (void) mkdir_label("/run/systemd/system", 0755); + (void) mkdir_label("/run/systemd/inaccessible", 0000); + /* Set up inaccessible items */ + (void) mknod("/run/systemd/inaccessible/reg", S_IFREG | 0000, 0); + (void) mkdir_label("/run/systemd/inaccessible/dir", 0000); + (void) mknod("/run/systemd/inaccessible/chr", S_IFCHR | 0000, makedev(0, 0)); + (void) mknod("/run/systemd/inaccessible/blk", S_IFBLK | 0000, makedev(0, 0)); + (void) mkfifo("/run/systemd/inaccessible/fifo", 0000); + (void) mknod("/run/systemd/inaccessible/sock", S_IFSOCK | 0000, 0); + + return 0; +} diff --git a/src/grp-system/libcore/src/mount.c b/src/grp-system/libcore/src/mount.c new file mode 100644 index 0000000000..06e8313e3d --- /dev/null +++ b/src/grp-system/libcore/src/mount.c @@ -0,0 +1,1947 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <signal.h> +#include <stdio.h> +#include <sys/epoll.h> + +#include <systemd/sd-messages.h> + +#include "core/manager.h" +#include "core/mount-setup.h" +#include "core/mount.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/mount-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-shared/fstab-util.h" + +#include "dbus-mount.h" + +#define RETRY_UMOUNT_MAX 32 + +DEFINE_TRIVIAL_CLEANUP_FUNC(struct libmnt_table*, mnt_free_table); +DEFINE_TRIVIAL_CLEANUP_FUNC(struct libmnt_iter*, mnt_free_iter); + +static const UnitActiveState state_translation_table[_MOUNT_STATE_MAX] = { + [MOUNT_DEAD] = UNIT_INACTIVE, + [MOUNT_MOUNTING] = UNIT_ACTIVATING, + [MOUNT_MOUNTING_DONE] = UNIT_ACTIVE, + [MOUNT_MOUNTED] = UNIT_ACTIVE, + [MOUNT_REMOUNTING] = UNIT_RELOADING, + [MOUNT_UNMOUNTING] = UNIT_DEACTIVATING, + [MOUNT_MOUNTING_SIGTERM] = UNIT_DEACTIVATING, + [MOUNT_MOUNTING_SIGKILL] = UNIT_DEACTIVATING, + [MOUNT_REMOUNTING_SIGTERM] = UNIT_RELOADING, + [MOUNT_REMOUNTING_SIGKILL] = UNIT_RELOADING, + [MOUNT_UNMOUNTING_SIGTERM] = UNIT_DEACTIVATING, + [MOUNT_UNMOUNTING_SIGKILL] = UNIT_DEACTIVATING, + [MOUNT_FAILED] = UNIT_FAILED +}; + +static int mount_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); +static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); + +static bool mount_needs_network(const char *options, const char *fstype) { + if (fstab_test_option(options, "_netdev\0")) + return true; + + if (fstype && fstype_is_network(fstype)) + return true; + + return false; +} + +static bool mount_is_network(const MountParameters *p) { + assert(p); + + return mount_needs_network(p->options, p->fstype); +} + +static bool mount_is_loop(const MountParameters *p) { + assert(p); + + if (fstab_test_option(p->options, "loop\0")) + return true; + + return false; +} + +static bool mount_is_bind(const MountParameters *p) { + assert(p); + + if (fstab_test_option(p->options, "bind\0" "rbind\0")) + return true; + + if (p->fstype && STR_IN_SET(p->fstype, "bind", "rbind")) + return true; + + return false; +} + +static bool mount_is_auto(const MountParameters *p) { + assert(p); + + return !fstab_test_option(p->options, "noauto\0"); +} + +static bool mount_is_automount(const MountParameters *p) { + assert(p); + + return fstab_test_option(p->options, + "comment=systemd.automount\0" + "x-systemd.automount\0"); +} + +static bool mount_state_active(MountState state) { + return IN_SET(state, + MOUNT_MOUNTING, + MOUNT_MOUNTING_DONE, + MOUNT_REMOUNTING, + MOUNT_UNMOUNTING, + MOUNT_MOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_UNMOUNTING_SIGKILL, + MOUNT_REMOUNTING_SIGTERM, + MOUNT_REMOUNTING_SIGKILL); +} + +static bool needs_quota(const MountParameters *p) { + assert(p); + + /* Quotas are not enabled on network filesystems, + * but we want them, for example, on storage connected via iscsi */ + if (p->fstype && fstype_is_network(p->fstype)) + return false; + + if (mount_is_bind(p)) + return false; + + return fstab_test_option(p->options, + "usrquota\0" "grpquota\0" "quota\0" "usrjquota\0" "grpjquota\0"); +} + +static void mount_init(Unit *u) { + Mount *m = MOUNT(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + m->timeout_usec = u->manager->default_timeout_start_usec; + m->directory_mode = 0755; + + /* We need to make sure that /usr/bin/mount is always called + * in the same process group as us, so that the autofs kernel + * side doesn't send us another mount request while we are + * already trying to comply its last one. */ + m->exec_context.same_pgrp = true; + + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + + u->ignore_on_isolate = true; +} + +static int mount_arm_timer(Mount *m, usec_t usec) { + int r; + + assert(m); + + if (m->timer_event_source) { + r = sd_event_source_set_time(m->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(m->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(m)->manager->event, + &m->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + mount_dispatch_timer, m); + if (r < 0) + return r; + + (void) sd_event_source_set_description(m->timer_event_source, "mount-timer"); + + return 0; +} + +static void mount_unwatch_control_pid(Mount *m) { + assert(m); + + if (m->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(m), m->control_pid); + m->control_pid = 0; +} + +static void mount_parameters_done(MountParameters *p) { + assert(p); + + free(p->what); + free(p->options); + free(p->fstype); + + p->what = p->options = p->fstype = NULL; +} + +static void mount_done(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + m->where = mfree(m->where); + + mount_parameters_done(&m->parameters_proc_self_mountinfo); + mount_parameters_done(&m->parameters_fragment); + + m->exec_runtime = exec_runtime_unref(m->exec_runtime); + exec_command_done_array(m->exec_command, _MOUNT_EXEC_COMMAND_MAX); + m->control_command = NULL; + + dynamic_creds_unref(&m->dynamic_creds); + + mount_unwatch_control_pid(m); + + m->timer_event_source = sd_event_source_unref(m->timer_event_source); +} + +_pure_ static MountParameters* get_mount_parameters_fragment(Mount *m) { + assert(m); + + if (m->from_fragment) + return &m->parameters_fragment; + + return NULL; +} + +_pure_ static MountParameters* get_mount_parameters(Mount *m) { + assert(m); + + if (m->from_proc_self_mountinfo) + return &m->parameters_proc_self_mountinfo; + + return get_mount_parameters_fragment(m); +} + +static int mount_add_mount_links(Mount *m) { + _cleanup_free_ char *parent = NULL; + MountParameters *pm; + Unit *other; + Iterator i; + Set *s; + int r; + + assert(m); + + if (!path_equal(m->where, "/")) { + /* Adds in links to other mount points that might lie further + * up in the hierarchy */ + + parent = dirname_malloc(m->where); + if (!parent) + return -ENOMEM; + + r = unit_require_mounts_for(UNIT(m), parent); + if (r < 0) + return r; + } + + /* Adds in links to other mount points that might be needed + * for the source path (if this is a bind mount or a loop mount) to be + * available. */ + pm = get_mount_parameters_fragment(m); + if (pm && pm->what && + path_is_absolute(pm->what) && + (mount_is_bind(pm) || mount_is_loop(pm) || !mount_is_network(pm))) { + + r = unit_require_mounts_for(UNIT(m), pm->what); + if (r < 0) + return r; + } + + /* Adds in links to other units that use this path or paths + * further down in the hierarchy */ + s = manager_get_units_requiring_mounts_for(UNIT(m)->manager, m->where); + SET_FOREACH(other, s, i) { + + if (other->load_state != UNIT_LOADED) + continue; + + if (other == UNIT(m)) + continue; + + r = unit_add_dependency(other, UNIT_AFTER, UNIT(m), true); + if (r < 0) + return r; + + if (UNIT(m)->fragment_path) { + /* If we have fragment configuration, then make this dependency required */ + r = unit_add_dependency(other, UNIT_REQUIRES, UNIT(m), true); + if (r < 0) + return r; + } + } + + return 0; +} + +static int mount_add_device_links(Mount *m) { + MountParameters *p; + bool device_wants_mount = false; + int r; + + assert(m); + + p = get_mount_parameters(m); + if (!p) + return 0; + + if (!p->what) + return 0; + + if (mount_is_bind(p)) + return 0; + + if (!is_device_path(p->what)) + return 0; + + /* /dev/root is a really weird thing, it's not a real device, + * but just a path the kernel exports for the root file system + * specified on the kernel command line. Ignore it here. */ + if (path_equal(p->what, "/dev/root")) + return 0; + + if (path_equal(m->where, "/")) + return 0; + + if (mount_is_auto(p) && !mount_is_automount(p) && MANAGER_IS_SYSTEM(UNIT(m)->manager)) + device_wants_mount = true; + + r = unit_add_node_link(UNIT(m), p->what, device_wants_mount, m->from_fragment ? UNIT_BINDS_TO : UNIT_REQUIRES); + if (r < 0) + return r; + + return 0; +} + +static int mount_add_quota_links(Mount *m) { + int r; + MountParameters *p; + + assert(m); + + if (!MANAGER_IS_SYSTEM(UNIT(m)->manager)) + return 0; + + p = get_mount_parameters_fragment(m); + if (!p) + return 0; + + if (!needs_quota(p)) + return 0; + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTACHECK_SERVICE, NULL, true); + if (r < 0) + return r; + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_WANTS, SPECIAL_QUOTAON_SERVICE, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static bool should_umount(Mount *m) { + MountParameters *p; + + if (PATH_IN_SET(m->where, "/", "/usr") || + path_startswith(m->where, "/run/initramfs")) + return false; + + p = get_mount_parameters(m); + if (p && fstab_test_option(p->options, "x-initrd.mount\0") && + !in_initrd()) + return false; + + return true; +} + +static int mount_add_default_dependencies(Mount *m) { + MountParameters *p; + const char *after; + int r; + + assert(m); + + if (!UNIT(m)->default_dependencies) + return 0; + + if (!MANAGER_IS_SYSTEM(UNIT(m)->manager)) + return 0; + + /* We do not add any default dependencies to /, /usr or + * /run/initramfs/, since they are guaranteed to stay + * mounted the whole time, since our system is on it. + * Also, don't bother with anything mounted below virtual + * file systems, it's also going to be virtual, and hence + * not worth the effort. */ + if (PATH_IN_SET(m->where, "/", "/usr") || + path_startswith(m->where, "/run/initramfs") || + path_startswith(m->where, "/proc") || + path_startswith(m->where, "/sys") || + path_startswith(m->where, "/dev")) + return 0; + + p = get_mount_parameters(m); + if (!p) + return 0; + + if (mount_is_network(p)) { + /* We order ourselves after network.target. This is + * primarily useful at shutdown: services that take + * down the network should order themselves before + * network.target, so that they are shut down only + * after this mount unit is stopped. */ + + r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, SPECIAL_NETWORK_TARGET, NULL, true); + if (r < 0) + return r; + + /* We pull in network-online.target, and order + * ourselves after it. This is useful at start-up to + * actively pull in tools that want to be started + * before we start mounting network file systems, and + * whose purpose it is to delay this until the network + * is "up". */ + + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_WANTS, UNIT_AFTER, SPECIAL_NETWORK_ONLINE_TARGET, NULL, true); + if (r < 0) + return r; + + after = SPECIAL_REMOTE_FS_PRE_TARGET; + } else + after = SPECIAL_LOCAL_FS_PRE_TARGET; + + r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after, NULL, true); + if (r < 0) + return r; + + if (should_umount(m)) { + r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return 0; +} + +static int mount_verify(Mount *m) { + _cleanup_free_ char *e = NULL; + MountParameters *p; + int r; + + assert(m); + + if (UNIT(m)->load_state != UNIT_LOADED) + return 0; + + if (!m->from_fragment && !m->from_proc_self_mountinfo) + return -ENOENT; + + r = unit_name_from_path(m->where, ".mount", &e); + if (r < 0) + return log_unit_error_errno(UNIT(m), r, "Failed to generate unit name from mount path: %m"); + + if (!unit_has_name(UNIT(m), e)) { + log_unit_error(UNIT(m), "Where= setting doesn't match unit name. Refusing."); + return -EINVAL; + } + + if (mount_point_is_api(m->where) || mount_point_ignore(m->where)) { + log_unit_error(UNIT(m), "Cannot create mount unit for API file system %s. Refusing.", m->where); + return -EINVAL; + } + + p = get_mount_parameters_fragment(m); + if (p && !p->what) { + log_unit_error(UNIT(m), "What= setting is missing. Refusing."); + return -EBADMSG; + } + + if (m->exec_context.pam_name && m->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_unit_error(UNIT(m), "Unit has PAM enabled. Kill mode must be set to control-group'. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int mount_add_extras(Mount *m) { + Unit *u = UNIT(m); + int r; + + assert(m); + + if (u->fragment_path) + m->from_fragment = true; + + if (!m->where) { + r = unit_name_to_path(u->id, &m->where); + if (r < 0) + return r; + } + + path_kill_slashes(m->where); + + if (!u->description) { + r = unit_set_description(u, m->where); + if (r < 0) + return r; + } + + r = mount_add_device_links(m); + if (r < 0) + return r; + + r = mount_add_mount_links(m); + if (r < 0) + return r; + + r = mount_add_quota_links(m); + if (r < 0) + return r; + + r = unit_patch_contexts(u); + if (r < 0) + return r; + + r = unit_add_exec_dependencies(u, &m->exec_context); + if (r < 0) + return r; + + r = unit_set_default_slice(u); + if (r < 0) + return r; + + r = mount_add_default_dependencies(m); + if (r < 0) + return r; + + return 0; +} + +static int mount_load_root_mount(Unit *u) { + assert(u); + + if (!unit_has_name(u, SPECIAL_ROOT_MOUNT)) + return 0; + + u->perpetual = true; + u->default_dependencies = false; + + /* The stdio/kmsg bridge socket is on /, in order to avoid a dep loop, don't use kmsg logging for -.mount */ + MOUNT(u)->exec_context.std_output = EXEC_OUTPUT_NULL; + MOUNT(u)->exec_context.std_input = EXEC_INPUT_NULL; + + if (!u->description) + u->description = strdup("Root Mount"); + + return 1; +} + +static int mount_load(Unit *u) { + Mount *m = MOUNT(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = mount_load_root_mount(u); + if (r < 0) + return r; + + if (m->from_proc_self_mountinfo || u->perpetual) + r = unit_load_fragment_and_dropin_optional(u); + else + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + r = mount_add_extras(m); + if (r < 0) + return r; + } + + return mount_verify(m); +} + +static void mount_set_state(Mount *m, MountState state) { + MountState old_state; + assert(m); + + old_state = m->state; + m->state = state; + + if (!mount_state_active(state)) { + m->timer_event_source = sd_event_source_unref(m->timer_event_source); + mount_unwatch_control_pid(m); + m->control_command = NULL; + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + } + + if (state != old_state) + log_unit_debug(UNIT(m), "Changed %s -> %s", mount_state_to_string(old_state), mount_state_to_string(state)); + + unit_notify(UNIT(m), state_translation_table[old_state], state_translation_table[state], m->reload_result == MOUNT_SUCCESS); + m->reload_result = MOUNT_SUCCESS; +} + +static int mount_coldplug(Unit *u) { + Mount *m = MOUNT(u); + MountState new_state = MOUNT_DEAD; + int r; + + assert(m); + assert(m->state == MOUNT_DEAD); + + if (m->deserialized_state != m->state) + new_state = m->deserialized_state; + else if (m->from_proc_self_mountinfo) + new_state = MOUNT_MOUNTED; + + if (new_state == m->state) + return 0; + + if (m->control_pid > 0 && + pid_is_unwaited(m->control_pid) && + mount_state_active(new_state)) { + + r = unit_watch_pid(UNIT(m), m->control_pid); + if (r < 0) + return r; + + r = mount_arm_timer(m, usec_add(u->state_change_timestamp.monotonic, m->timeout_usec)); + if (r < 0) + return r; + } + + if (!IN_SET(new_state, MOUNT_DEAD, MOUNT_FAILED)) + (void) unit_setup_dynamic_creds(u); + + mount_set_state(m, new_state); + return 0; +} + +static void mount_dump(Unit *u, FILE *f, const char *prefix) { + Mount *m = MOUNT(u); + MountParameters *p; + + assert(m); + assert(f); + + p = get_mount_parameters(m); + + fprintf(f, + "%sMount State: %s\n" + "%sResult: %s\n" + "%sWhere: %s\n" + "%sWhat: %s\n" + "%sFile System Type: %s\n" + "%sOptions: %s\n" + "%sFrom /proc/self/mountinfo: %s\n" + "%sFrom fragment: %s\n" + "%sDirectoryMode: %04o\n" + "%sSloppyOptions: %s\n" + "%sLazyUnmount: %s\n" + "%sForceUnmount: %s\n", + prefix, mount_state_to_string(m->state), + prefix, mount_result_to_string(m->result), + prefix, m->where, + prefix, p ? strna(p->what) : "n/a", + prefix, p ? strna(p->fstype) : "n/a", + prefix, p ? strna(p->options) : "n/a", + prefix, yes_no(m->from_proc_self_mountinfo), + prefix, yes_no(m->from_fragment), + prefix, m->directory_mode, + prefix, yes_no(m->sloppy_options), + prefix, yes_no(m->lazy_unmount), + prefix, yes_no(m->force_unmount)); + + if (m->control_pid > 0) + fprintf(f, + "%sControl PID: "PID_FMT"\n", + prefix, m->control_pid); + + exec_context_dump(&m->exec_context, f, prefix); + kill_context_dump(&m->kill_context, f, prefix); +} + +static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) { + pid_t pid; + int r; + ExecParameters exec_params = { + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, + }; + + assert(m); + assert(c); + assert(_pid); + + (void) unit_realize_cgroup(UNIT(m)); + if (m->reset_cpu_usage) { + (void) unit_reset_cpu_usage(UNIT(m)); + m->reset_cpu_usage = false; + } + + r = unit_setup_exec_runtime(UNIT(m)); + if (r < 0) + return r; + + r = unit_setup_dynamic_creds(UNIT(m)); + if (r < 0) + return r; + + r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec)); + if (r < 0) + return r; + + exec_params.environment = UNIT(m)->manager->environment; + exec_params.flags |= UNIT(m)->manager->confirm_spawn ? EXEC_CONFIRM_SPAWN : 0; + exec_params.cgroup_supported = UNIT(m)->manager->cgroup_supported; + exec_params.cgroup_path = UNIT(m)->cgroup_path; + exec_params.cgroup_delegate = m->cgroup_context.delegate; + exec_params.runtime_prefix = manager_get_runtime_prefix(UNIT(m)->manager); + + r = exec_spawn(UNIT(m), + c, + &m->exec_context, + &exec_params, + m->exec_runtime, + &m->dynamic_creds, + &pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(m), pid); + if (r < 0) + /* FIXME: we need to do something here */ + return r; + + *_pid = pid; + + return 0; +} + +static void mount_enter_dead(Mount *m, MountResult f) { + assert(m); + + if (m->result == MOUNT_SUCCESS) + m->result = f; + + mount_set_state(m, m->result != MOUNT_SUCCESS ? MOUNT_FAILED : MOUNT_DEAD); + + exec_runtime_destroy(m->exec_runtime); + m->exec_runtime = exec_runtime_unref(m->exec_runtime); + + exec_context_destroy_runtime_directory(&m->exec_context, manager_get_runtime_prefix(UNIT(m)->manager)); + + unit_unref_uid_gid(UNIT(m), true); + + dynamic_creds_destroy(&m->dynamic_creds); +} + +static void mount_enter_mounted(Mount *m, MountResult f) { + assert(m); + + if (m->result == MOUNT_SUCCESS) + m->result = f; + + mount_set_state(m, MOUNT_MOUNTED); +} + +static void mount_enter_signal(Mount *m, MountState state, MountResult f) { + int r; + + assert(m); + + if (m->result == MOUNT_SUCCESS) + m->result = f; + + r = unit_kill_context( + UNIT(m), + &m->kill_context, + (state != MOUNT_MOUNTING_SIGTERM && state != MOUNT_UNMOUNTING_SIGTERM && state != MOUNT_REMOUNTING_SIGTERM) ? + KILL_KILL : KILL_TERMINATE, + -1, + m->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec)); + if (r < 0) + goto fail; + + mount_set_state(m, state); + } else if (state == MOUNT_REMOUNTING_SIGTERM) + mount_enter_signal(m, MOUNT_REMOUNTING_SIGKILL, MOUNT_SUCCESS); + else if (state == MOUNT_REMOUNTING_SIGKILL) + mount_enter_mounted(m, MOUNT_SUCCESS); + else if (state == MOUNT_MOUNTING_SIGTERM) + mount_enter_signal(m, MOUNT_MOUNTING_SIGKILL, MOUNT_SUCCESS); + else if (state == MOUNT_UNMOUNTING_SIGTERM) + mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_SUCCESS); + else + mount_enter_dead(m, MOUNT_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(m), r, "Failed to kill processes: %m"); + + if (IN_SET(state, MOUNT_REMOUNTING_SIGTERM, MOUNT_REMOUNTING_SIGKILL)) + mount_enter_mounted(m, MOUNT_FAILURE_RESOURCES); + else + mount_enter_dead(m, MOUNT_FAILURE_RESOURCES); +} + +static void mount_enter_unmounting(Mount *m) { + int r; + + assert(m); + + /* Start counting our attempts */ + if (!IN_SET(m->state, + MOUNT_UNMOUNTING, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_UNMOUNTING_SIGKILL)) + m->n_retry_umount = 0; + + m->control_command_id = MOUNT_EXEC_UNMOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_UNMOUNT; + + r = exec_command_set(m->control_command, UMOUNT_PATH, m->where, NULL); + if (r >= 0 && m->lazy_unmount) + r = exec_command_append(m->control_command, "-l", NULL); + if (r >= 0 && m->force_unmount) + r = exec_command_append(m->control_command, "-f", NULL); + if (r < 0) + goto fail; + + mount_unwatch_control_pid(m); + + r = mount_spawn(m, m->control_command, &m->control_pid); + if (r < 0) + goto fail; + + mount_set_state(m, MOUNT_UNMOUNTING); + + return; + +fail: + log_unit_warning_errno(UNIT(m), r, "Failed to run 'umount' task: %m"); + mount_enter_mounted(m, MOUNT_FAILURE_RESOURCES); +} + +static void mount_enter_mounting(Mount *m) { + int r; + MountParameters *p; + + assert(m); + + m->control_command_id = MOUNT_EXEC_MOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_MOUNT; + + r = unit_fail_if_symlink(UNIT(m), m->where); + if (r < 0) + goto fail; + + (void) mkdir_p_label(m->where, m->directory_mode); + + unit_warn_if_dir_nonempty(UNIT(m), m->where); + + /* Create the source directory for bind-mounts if needed */ + p = get_mount_parameters_fragment(m); + if (p && mount_is_bind(p)) + (void) mkdir_p_label(p->what, m->directory_mode); + + if (p) { + _cleanup_free_ char *opts = NULL; + + r = fstab_filter_options(p->options, "nofail\0" "noauto\0" "auto\0", NULL, NULL, &opts); + if (r < 0) + goto fail; + + r = exec_command_set(m->control_command, MOUNT_PATH, p->what, m->where, NULL); + if (r >= 0 && m->sloppy_options) + r = exec_command_append(m->control_command, "-s", NULL); + if (r >= 0 && p->fstype) + r = exec_command_append(m->control_command, "-t", p->fstype, NULL); + if (r >= 0 && !isempty(opts)) + r = exec_command_append(m->control_command, "-o", opts, NULL); + } else + r = -ENOENT; + + if (r < 0) + goto fail; + + mount_unwatch_control_pid(m); + + r = mount_spawn(m, m->control_command, &m->control_pid); + if (r < 0) + goto fail; + + mount_set_state(m, MOUNT_MOUNTING); + + return; + +fail: + log_unit_warning_errno(UNIT(m), r, "Failed to run 'mount' task: %m"); + mount_enter_dead(m, MOUNT_FAILURE_RESOURCES); +} + +static void mount_enter_remounting(Mount *m) { + int r; + MountParameters *p; + + assert(m); + + m->control_command_id = MOUNT_EXEC_REMOUNT; + m->control_command = m->exec_command + MOUNT_EXEC_REMOUNT; + + p = get_mount_parameters_fragment(m); + if (p) { + const char *o; + + if (p->options) + o = strjoina("remount,", p->options); + else + o = "remount"; + + r = exec_command_set(m->control_command, MOUNT_PATH, + p->what, m->where, + "-o", o, NULL); + if (r >= 0 && m->sloppy_options) + r = exec_command_append(m->control_command, "-s", NULL); + if (r >= 0 && p->fstype) + r = exec_command_append(m->control_command, "-t", p->fstype, NULL); + } else + r = -ENOENT; + + if (r < 0) + goto fail; + + mount_unwatch_control_pid(m); + + r = mount_spawn(m, m->control_command, &m->control_pid); + if (r < 0) + goto fail; + + mount_set_state(m, MOUNT_REMOUNTING); + + return; + +fail: + log_unit_warning_errno(UNIT(m), r, "Failed to run 'remount' task: %m"); + m->reload_result = MOUNT_FAILURE_RESOURCES; + mount_enter_mounted(m, MOUNT_SUCCESS); +} + +static int mount_start(Unit *u) { + Mount *m = MOUNT(u); + int r; + + assert(m); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (IN_SET(m->state, + MOUNT_UNMOUNTING, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_UNMOUNTING_SIGKILL, + MOUNT_MOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGKILL)) + return -EAGAIN; + + /* Already on it! */ + if (m->state == MOUNT_MOUNTING) + return 0; + + assert(IN_SET(m->state, MOUNT_DEAD, MOUNT_FAILED)); + + r = unit_start_limit_test(u); + if (r < 0) { + mount_enter_dead(m, MOUNT_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + m->result = MOUNT_SUCCESS; + m->reload_result = MOUNT_SUCCESS; + m->reset_cpu_usage = true; + + mount_enter_mounting(m); + return 1; +} + +static int mount_stop(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + /* Already on it */ + if (IN_SET(m->state, + MOUNT_UNMOUNTING, + MOUNT_UNMOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGKILL)) + return 0; + + assert(IN_SET(m->state, + MOUNT_MOUNTING, + MOUNT_MOUNTING_DONE, + MOUNT_MOUNTED, + MOUNT_REMOUNTING, + MOUNT_REMOUNTING_SIGTERM, + MOUNT_REMOUNTING_SIGKILL)); + + mount_enter_unmounting(m); + return 1; +} + +static int mount_reload(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + if (m->state == MOUNT_MOUNTING_DONE) + return -EAGAIN; + + assert(m->state == MOUNT_MOUNTED); + + mount_enter_remounting(m); + return 1; +} + +static int mount_serialize(Unit *u, FILE *f, FDSet *fds) { + Mount *m = MOUNT(u); + + assert(m); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", mount_state_to_string(m->state)); + unit_serialize_item(u, f, "result", mount_result_to_string(m->result)); + unit_serialize_item(u, f, "reload-result", mount_result_to_string(m->reload_result)); + + if (m->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", PID_FMT, m->control_pid); + + if (m->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", mount_exec_command_to_string(m->control_command_id)); + + return 0; +} + +static int mount_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Mount *m = MOUNT(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + MountState state; + + if ((state = mount_state_from_string(value)) < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + m->deserialized_state = state; + } else if (streq(key, "result")) { + MountResult f; + + f = mount_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != MOUNT_SUCCESS) + m->result = f; + + } else if (streq(key, "reload-result")) { + MountResult f; + + f = mount_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse reload result value: %s", value); + else if (f != MOUNT_SUCCESS) + m->reload_result = f; + + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse control-pid value: %s", value); + else + m->control_pid = pid; + } else if (streq(key, "control-command")) { + MountExecCommand id; + + id = mount_exec_command_from_string(value); + if (id < 0) + log_unit_debug(u, "Failed to parse exec-command value: %s", value); + else { + m->control_command_id = id; + m->control_command = m->exec_command + id; + } + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState mount_active_state(Unit *u) { + assert(u); + + return state_translation_table[MOUNT(u)->state]; +} + +_pure_ static const char *mount_sub_state_to_string(Unit *u) { + assert(u); + + return mount_state_to_string(MOUNT(u)->state); +} + +_pure_ static bool mount_check_gc(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + return m->from_proc_self_mountinfo; +} + +static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Mount *m = MOUNT(u); + MountResult f; + + assert(m); + assert(pid >= 0); + + if (pid != m->control_pid) + return; + + m->control_pid = 0; + + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) + f = MOUNT_SUCCESS; + else if (code == CLD_EXITED) + f = MOUNT_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = MOUNT_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = MOUNT_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (m->result == MOUNT_SUCCESS) + m->result = f; + + if (m->control_command) { + exec_status_exit(&m->control_command->exec_status, &m->exec_context, pid, code, status); + + m->control_command = NULL; + m->control_command_id = _MOUNT_EXEC_COMMAND_INVALID; + } + + log_unit_full(u, f == MOUNT_SUCCESS ? LOG_DEBUG : LOG_NOTICE, 0, + "Mount process exited, code=%s status=%i", sigchld_code_to_string(code), status); + + /* Note that mount(8) returning and the kernel sending us a + * mount table change event might happen out-of-order. If an + * operation succeed we assume the kernel will follow soon too + * and already change into the resulting state. If it fails + * we check if the kernel still knows about the mount. and + * change state accordingly. */ + + switch (m->state) { + + case MOUNT_MOUNTING: + case MOUNT_MOUNTING_DONE: + case MOUNT_MOUNTING_SIGKILL: + case MOUNT_MOUNTING_SIGTERM: + + if (f == MOUNT_SUCCESS || m->from_proc_self_mountinfo) + /* If /bin/mount returned success, or if we see the mount point in /proc/self/mountinfo we are + * happy. If we see the first condition first, we should see the the second condition + * immediately after – or /bin/mount lies to us and is broken. */ + mount_enter_mounted(m, f); + else + mount_enter_dead(m, f); + break; + + case MOUNT_REMOUNTING: + case MOUNT_REMOUNTING_SIGKILL: + case MOUNT_REMOUNTING_SIGTERM: + + m->reload_result = f; + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_SUCCESS); + else + mount_enter_dead(m, MOUNT_SUCCESS); + + break; + + case MOUNT_UNMOUNTING: + case MOUNT_UNMOUNTING_SIGKILL: + case MOUNT_UNMOUNTING_SIGTERM: + + if (f == MOUNT_SUCCESS) { + + if (m->from_proc_self_mountinfo) { + + /* Still a mount point? If so, let's + * try again. Most likely there were + * multiple mount points stacked on + * top of each other. Note that due to + * the io event priority logic we can + * be sure the new mountinfo is loaded + * before we process the SIGCHLD for + * the mount command. */ + + if (m->n_retry_umount < RETRY_UMOUNT_MAX) { + log_unit_debug(u, "Mount still present, trying again."); + m->n_retry_umount++; + mount_enter_unmounting(m); + } else { + log_unit_debug(u, "Mount still present after %u attempts to unmount, giving up.", m->n_retry_umount); + mount_enter_mounted(m, f); + } + } else + mount_enter_dead(m, f); + + } else if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, f); + else + mount_enter_dead(m, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static int mount_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + Mount *m = MOUNT(userdata); + + assert(m); + assert(m->timer_event_source == source); + + switch (m->state) { + + case MOUNT_MOUNTING: + case MOUNT_MOUNTING_DONE: + log_unit_warning(UNIT(m), "Mounting timed out. Stopping."); + mount_enter_signal(m, MOUNT_MOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT); + break; + + case MOUNT_REMOUNTING: + log_unit_warning(UNIT(m), "Remounting timed out. Stopping."); + m->reload_result = MOUNT_FAILURE_TIMEOUT; + mount_enter_mounted(m, MOUNT_SUCCESS); + break; + + case MOUNT_UNMOUNTING: + log_unit_warning(UNIT(m), "Unmounting timed out. Stopping."); + mount_enter_signal(m, MOUNT_UNMOUNTING_SIGTERM, MOUNT_FAILURE_TIMEOUT); + break; + + case MOUNT_MOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_unit_warning(UNIT(m), "Mounting timed out. Killing."); + mount_enter_signal(m, MOUNT_MOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(m), "Mounting timed out. Skipping SIGKILL. Ignoring."); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_REMOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_unit_warning(UNIT(m), "Remounting timed out. Killing."); + mount_enter_signal(m, MOUNT_REMOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(m), "Remounting timed out. Skipping SIGKILL. Ignoring."); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_UNMOUNTING_SIGTERM: + if (m->kill_context.send_sigkill) { + log_unit_warning(UNIT(m), "Unmounting timed out. Killing."); + mount_enter_signal(m, MOUNT_UNMOUNTING_SIGKILL, MOUNT_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(m), "Unmounting timed out. Skipping SIGKILL. Ignoring."); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + } + break; + + case MOUNT_MOUNTING_SIGKILL: + case MOUNT_REMOUNTING_SIGKILL: + case MOUNT_UNMOUNTING_SIGKILL: + log_unit_warning(UNIT(m),"Mount process still around after SIGKILL. Ignoring."); + + if (m->from_proc_self_mountinfo) + mount_enter_mounted(m, MOUNT_FAILURE_TIMEOUT); + else + mount_enter_dead(m, MOUNT_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +static int mount_setup_unit( + Manager *m, + const char *what, + const char *where, + const char *options, + const char *fstype, + bool set_flags) { + + _cleanup_free_ char *e = NULL, *w = NULL, *o = NULL, *f = NULL; + bool load_extras = false; + MountParameters *p; + bool delete, changed = false; + Unit *u; + int r; + + assert(m); + assert(what); + assert(where); + assert(options); + assert(fstype); + + /* Ignore API mount points. They should never be referenced in + * dependencies ever. */ + if (mount_point_is_api(where) || mount_point_ignore(where)) + return 0; + + if (streq(fstype, "autofs")) + return 0; + + /* probably some kind of swap, ignore */ + if (!is_path(where)) + return 0; + + r = unit_name_from_path(where, ".mount", &e); + if (r < 0) + return r; + + u = manager_get_unit(m, e); + if (!u) { + delete = true; + + r = unit_new_for_name(m, sizeof(Mount), e, &u); + if (r < 0) + goto fail; + + MOUNT(u)->where = strdup(where); + if (!MOUNT(u)->where) { + r = -ENOMEM; + goto fail; + } + + u->source_path = strdup("/proc/self/mountinfo"); + if (!u->source_path) { + r = -ENOMEM; + goto fail; + } + + if (MANAGER_IS_SYSTEM(m)) { + const char* target; + + target = mount_needs_network(options, fstype) ? SPECIAL_REMOTE_FS_TARGET : SPECIAL_LOCAL_FS_TARGET; + r = unit_add_dependency_by_name(u, UNIT_BEFORE, target, NULL, true); + if (r < 0) + goto fail; + + if (should_umount(MOUNT(u))) { + r = unit_add_dependency_by_name(u, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); + if (r < 0) + goto fail; + } + } + + unit_add_to_load_queue(u); + changed = true; + } else { + delete = false; + + if (!MOUNT(u)->where) { + MOUNT(u)->where = strdup(where); + if (!MOUNT(u)->where) { + r = -ENOMEM; + goto fail; + } + } + + if (MANAGER_IS_SYSTEM(m) && + mount_needs_network(options, fstype)) { + /* _netdev option may have shown up late, or on a + * remount. Add remote-fs dependencies, even though + * local-fs ones may already be there. */ + unit_add_dependency_by_name(u, UNIT_BEFORE, SPECIAL_REMOTE_FS_TARGET, NULL, true); + load_extras = true; + } + + if (u->load_state == UNIT_NOT_FOUND) { + u->load_state = UNIT_LOADED; + u->load_error = 0; + + /* Load in the extras later on, after we + * finished initialization of the unit */ + load_extras = true; + changed = true; + } + } + + w = strdup(what); + o = strdup(options); + f = strdup(fstype); + if (!w || !o || !f) { + r = -ENOMEM; + goto fail; + } + + p = &MOUNT(u)->parameters_proc_self_mountinfo; + + changed = changed || + !streq_ptr(p->options, options) || + !streq_ptr(p->what, what) || + !streq_ptr(p->fstype, fstype); + + if (set_flags) { + MOUNT(u)->is_mounted = true; + MOUNT(u)->just_mounted = !MOUNT(u)->from_proc_self_mountinfo; + MOUNT(u)->just_changed = changed; + } + + MOUNT(u)->from_proc_self_mountinfo = true; + + free_and_replace(p->what, w); + free_and_replace(p->options, o); + free_and_replace(p->fstype, f); + + if (load_extras) { + r = mount_add_extras(MOUNT(u)); + if (r < 0) + goto fail; + } + + if (changed) + unit_add_to_dbus_queue(u); + + return 0; + +fail: + log_warning_errno(r, "Failed to set up mount unit: %m"); + + if (delete && u) + unit_free(u); + + return r; +} + +static int mount_load_proc_self_mountinfo(Manager *m, bool set_flags) { + _cleanup_(mnt_free_tablep) struct libmnt_table *t = NULL; + _cleanup_(mnt_free_iterp) struct libmnt_iter *i = NULL; + int r = 0; + + assert(m); + + t = mnt_new_table(); + if (!t) + return log_oom(); + + i = mnt_new_iter(MNT_ITER_FORWARD); + if (!i) + return log_oom(); + + r = mnt_table_parse_mtab(t, NULL); + if (r < 0) + return log_error_errno(r, "Failed to parse /proc/self/mountinfo: %m"); + + r = 0; + for (;;) { + const char *device, *path, *options, *fstype; + _cleanup_free_ char *d = NULL, *p = NULL; + struct libmnt_fs *fs; + int k; + + k = mnt_table_next_fs(t, i, &fs); + if (k == 1) + break; + if (k < 0) + return log_error_errno(k, "Failed to get next entry from /proc/self/mountinfo: %m"); + + device = mnt_fs_get_source(fs); + path = mnt_fs_get_target(fs); + options = mnt_fs_get_options(fs); + fstype = mnt_fs_get_fstype(fs); + + if (!device || !path) + continue; + + if (cunescape(device, UNESCAPE_RELAX, &d) < 0) + return log_oom(); + + if (cunescape(path, UNESCAPE_RELAX, &p) < 0) + return log_oom(); + + (void) device_found_node(m, d, true, DEVICE_FOUND_MOUNT, set_flags); + + k = mount_setup_unit(m, d, p, options, fstype, set_flags); + if (r == 0 && k < 0) + r = k; + } + + return r; +} + +static void mount_shutdown(Manager *m) { + + assert(m); + + m->mount_event_source = sd_event_source_unref(m->mount_event_source); + + mnt_unref_monitor(m->mount_monitor); + m->mount_monitor = NULL; +} + +static int mount_get_timeout(Unit *u, usec_t *timeout) { + Mount *m = MOUNT(u); + usec_t t; + int r; + + if (!m->timer_event_source) + return 0; + + r = sd_event_source_get_time(m->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +static int synthesize_root_mount(Manager *m) { + Unit *u; + int r; + + assert(m); + + /* Whatever happens, we know for sure that the root directory is around, and cannot go away. Let's + * unconditionally synthesize it here and mark it as perpetual. */ + + u = manager_get_unit(m, SPECIAL_ROOT_MOUNT); + if (!u) { + r = unit_new_for_name(m, sizeof(Mount), SPECIAL_ROOT_MOUNT, &u); + if (r < 0) + return log_error_errno(r, "Failed to allocate the special " SPECIAL_ROOT_MOUNT " unit: %m"); + } + + u->perpetual = true; + MOUNT(u)->deserialized_state = MOUNT_MOUNTED; + + unit_add_to_load_queue(u); + unit_add_to_dbus_queue(u); + + return 0; +} + +static bool mount_is_mounted(Mount *m) { + assert(m); + + return UNIT(m)->perpetual || m->is_mounted; +} + +static void mount_enumerate(Manager *m) { + int r; + + assert(m); + + r = synthesize_root_mount(m); + if (r < 0) + goto fail; + + mnt_init_debug(0); + + if (!m->mount_monitor) { + int fd; + + m->mount_monitor = mnt_new_monitor(); + if (!m->mount_monitor) { + log_oom(); + goto fail; + } + + r = mnt_monitor_enable_kernel(m->mount_monitor, 1); + if (r < 0) { + log_error_errno(r, "Failed to enable watching of kernel mount events: %m"); + goto fail; + } + + r = mnt_monitor_enable_userspace(m->mount_monitor, 1, NULL); + if (r < 0) { + log_error_errno(r, "Failed to enable watching of userspace mount events: %m"); + goto fail; + } + + /* mnt_unref_monitor() will close the fd */ + fd = r = mnt_monitor_get_fd(m->mount_monitor); + if (r < 0) { + log_error_errno(r, "Failed to acquire watch file descriptor: %m"); + goto fail; + } + + r = sd_event_add_io(m->event, &m->mount_event_source, fd, EPOLLIN, mount_dispatch_io, m); + if (r < 0) { + log_error_errno(r, "Failed to watch mount file descriptor: %m"); + goto fail; + } + + r = sd_event_source_set_priority(m->mount_event_source, -10); + if (r < 0) { + log_error_errno(r, "Failed to adjust mount watch priority: %m"); + goto fail; + } + + (void) sd_event_source_set_description(m->mount_event_source, "mount-monitor-dispatch"); + } + + r = mount_load_proc_self_mountinfo(m, false); + if (r < 0) + goto fail; + + return; + +fail: + mount_shutdown(m); +} + +static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + _cleanup_set_free_ Set *around = NULL, *gone = NULL; + Manager *m = userdata; + const char *what; + Iterator i; + Unit *u; + int r; + + assert(m); + assert(revents & EPOLLIN); + + if (fd == mnt_monitor_get_fd(m->mount_monitor)) { + bool rescan = false; + + /* Drain all events and verify that the event is valid. + * + * Note that libmount also monitors /run/mount mkdir if the + * directory does not exist yet. The mkdir may generate event + * which is irrelevant for us. + * + * error: r < 0; valid: r == 0, false positive: rc == 1 */ + do { + r = mnt_monitor_next_change(m->mount_monitor, NULL, NULL); + if (r == 0) + rescan = true; + else if (r < 0) + return log_error_errno(r, "Failed to drain libmount events"); + } while (r == 0); + + log_debug("libmount event [rescan: %s]", yes_no(rescan)); + if (!rescan) + return 0; + } + + r = mount_load_proc_self_mountinfo(m, true); + if (r < 0) { + /* Reset flags, just in case, for later calls */ + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT]) { + Mount *mount = MOUNT(u); + + mount->is_mounted = mount->just_mounted = mount->just_changed = false; + } + + return 0; + } + + manager_dispatch_load_queue(m); + + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_MOUNT]) { + Mount *mount = MOUNT(u); + + if (!mount_is_mounted(mount)) { + + /* A mount point is not around right now. It + * might be gone, or might never have + * existed. */ + + if (mount->from_proc_self_mountinfo && + mount->parameters_proc_self_mountinfo.what) { + + /* Remember that this device might just have disappeared */ + if (set_ensure_allocated(&gone, &string_hash_ops) < 0 || + set_put(gone, mount->parameters_proc_self_mountinfo.what) < 0) + log_oom(); /* we don't care too much about OOM here... */ + } + + mount->from_proc_self_mountinfo = false; + + switch (mount->state) { + + case MOUNT_MOUNTED: + /* This has just been unmounted by + * somebody else, follow the state + * change. */ + mount->result = MOUNT_SUCCESS; /* make sure we forget any earlier umount failures */ + mount_enter_dead(mount, MOUNT_SUCCESS); + break; + + default: + break; + } + + } else if (mount->just_mounted || mount->just_changed) { + + /* A mount point was added or changed */ + + switch (mount->state) { + + case MOUNT_DEAD: + case MOUNT_FAILED: + + /* This has just been mounted by somebody else, follow the state change, but let's + * generate a new invocation ID for this implicitly and automatically. */ + (void) unit_acquire_invocation_id(UNIT(mount)); + mount_enter_mounted(mount, MOUNT_SUCCESS); + break; + + case MOUNT_MOUNTING: + mount_set_state(mount, MOUNT_MOUNTING_DONE); + break; + + default: + /* Nothing really changed, but let's + * issue an notification call + * nonetheless, in case somebody is + * waiting for this. (e.g. file system + * ro/rw remounts.) */ + mount_set_state(mount, mount->state); + break; + } + } + + if (mount_is_mounted(mount) && + mount->from_proc_self_mountinfo && + mount->parameters_proc_self_mountinfo.what) { + + if (set_ensure_allocated(&around, &string_hash_ops) < 0 || + set_put(around, mount->parameters_proc_self_mountinfo.what) < 0) + log_oom(); + } + + /* Reset the flags for later calls */ + mount->is_mounted = mount->just_mounted = mount->just_changed = false; + } + + SET_FOREACH(what, gone, i) { + if (set_contains(around, what)) + continue; + + /* Let the device units know that the device is no longer mounted */ + (void) device_found_node(m, what, false, DEVICE_FOUND_MOUNT, true); + } + + return 0; +} + +static void mount_reset_failed(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + if (m->state == MOUNT_FAILED) + mount_set_state(m, MOUNT_DEAD); + + m->result = MOUNT_SUCCESS; + m->reload_result = MOUNT_SUCCESS; +} + +static int mount_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, MOUNT(u)->control_pid, error); +} + +static int mount_control_pid(Unit *u) { + Mount *m = MOUNT(u); + + assert(m); + + return m->control_pid; +} + +static const char* const mount_exec_command_table[_MOUNT_EXEC_COMMAND_MAX] = { + [MOUNT_EXEC_MOUNT] = "ExecMount", + [MOUNT_EXEC_UNMOUNT] = "ExecUnmount", + [MOUNT_EXEC_REMOUNT] = "ExecRemount", +}; + +DEFINE_STRING_TABLE_LOOKUP(mount_exec_command, MountExecCommand); + +static const char* const mount_result_table[_MOUNT_RESULT_MAX] = { + [MOUNT_SUCCESS] = "success", + [MOUNT_FAILURE_RESOURCES] = "resources", + [MOUNT_FAILURE_TIMEOUT] = "timeout", + [MOUNT_FAILURE_EXIT_CODE] = "exit-code", + [MOUNT_FAILURE_SIGNAL] = "signal", + [MOUNT_FAILURE_CORE_DUMP] = "core-dump", + [MOUNT_FAILURE_START_LIMIT_HIT] = "start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(mount_result, MountResult); + +const UnitVTable mount_vtable = { + .object_size = sizeof(Mount), + .exec_context_offset = offsetof(Mount, exec_context), + .cgroup_context_offset = offsetof(Mount, cgroup_context), + .kill_context_offset = offsetof(Mount, kill_context), + .exec_runtime_offset = offsetof(Mount, exec_runtime), + .dynamic_creds_offset = offsetof(Mount, dynamic_creds), + + .sections = + "Unit\0" + "Mount\0" + "Install\0", + .private_section = "Mount", + + .init = mount_init, + .load = mount_load, + .done = mount_done, + + .coldplug = mount_coldplug, + + .dump = mount_dump, + + .start = mount_start, + .stop = mount_stop, + .reload = mount_reload, + + .kill = mount_kill, + + .serialize = mount_serialize, + .deserialize_item = mount_deserialize_item, + + .active_state = mount_active_state, + .sub_state_to_string = mount_sub_state_to_string, + + .check_gc = mount_check_gc, + + .sigchld_event = mount_sigchld_event, + + .reset_failed = mount_reset_failed, + + .control_pid = mount_control_pid, + + .bus_vtable = bus_mount_vtable, + .bus_set_property = bus_mount_set_property, + .bus_commit_properties = bus_mount_commit_properties, + + .get_timeout = mount_get_timeout, + + .can_transient = true, + + .enumerate = mount_enumerate, + .shutdown = mount_shutdown, + + .status_message_formats = { + .starting_stopping = { + [0] = "Mounting %s...", + [1] = "Unmounting %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Mounted %s.", + [JOB_FAILED] = "Failed to mount %s.", + [JOB_TIMEOUT] = "Timed out mounting %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Unmounted %s.", + [JOB_FAILED] = "Failed unmounting %s.", + [JOB_TIMEOUT] = "Timed out unmounting %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/namespace.c b/src/grp-system/libcore/src/namespace.c new file mode 100644 index 0000000000..988516d775 --- /dev/null +++ b/src/grp-system/libcore/src/namespace.c @@ -0,0 +1,1044 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sched.h> +#include <stdio.h> +#include <string.h> +#include <sys/mount.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <linux/fs.h> + +#include "core/loopback-setup.h" +#include "core/namespace.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/mount-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/selinux-util.h" +#include "systemd-basic/socket-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/umask-util.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/util.h" +#include "systemd-shared/dev-setup.h" + +#define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) + +typedef enum MountMode { + /* This is ordered by priority! */ + INACCESSIBLE, + READONLY, + PRIVATE_TMP, + PRIVATE_VAR_TMP, + PRIVATE_DEV, + READWRITE, +} MountMode; + +typedef struct BindMount { + const char *path; /* stack memory, doesn't need to be freed explicitly */ + char *chased; /* malloc()ed memory, needs to be freed */ + MountMode mode; + bool ignore; /* Ignore if path does not exist */ +} BindMount; + +typedef struct TargetMount { + const char *path; + MountMode mode; + bool ignore; /* Ignore if path does not exist */ +} TargetMount; + +/* + * The following Protect tables are to protect paths and mark some of them + * READONLY, in case a path is covered by an option from another table, then + * it is marked READWRITE in the current one, and the more restrictive mode is + * applied from that other table. This way all options can be combined in a + * safe and comprehensible way for users. + */ + +/* ProtectKernelTunables= option and the related filesystem APIs */ +static const TargetMount protect_kernel_tunables_table[] = { + { "/proc/sys", READONLY, false }, + { "/proc/sysrq-trigger", READONLY, true }, + { "/proc/latency_stats", READONLY, true }, + { "/proc/mtrr", READONLY, true }, + { "/proc/apm", READONLY, true }, + { "/proc/acpi", READONLY, true }, + { "/proc/timer_stats", READONLY, true }, + { "/proc/asound", READONLY, true }, + { "/proc/bus", READONLY, true }, + { "/proc/fs", READONLY, true }, + { "/proc/irq", READONLY, true }, + { "/sys", READONLY, false }, + { "/sys/kernel/debug", READONLY, true }, + { "/sys/kernel/tracing", READONLY, true }, + { "/sys/fs/cgroup", READWRITE, false }, /* READONLY is set by ProtectControlGroups= option */ +}; + +/* ProtectKernelModules= option */ +static const TargetMount protect_kernel_modules_table[] = { +#ifdef HAVE_SPLIT_USR + { "/lib/modules", INACCESSIBLE, true }, +#endif + { "/usr/lib/modules", INACCESSIBLE, true }, +}; + +/* + * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of + * system should be protected by ProtectSystem= + */ +static const TargetMount protect_home_read_only_table[] = { + { "/home", READONLY, true }, + { "/run/user", READONLY, true }, + { "/root", READONLY, true }, +}; + +/* ProtectHome=yes table */ +static const TargetMount protect_home_yes_table[] = { + { "/home", INACCESSIBLE, true }, + { "/run/user", INACCESSIBLE, true }, + { "/root", INACCESSIBLE, true }, +}; + +/* ProtectSystem=yes table */ +static const TargetMount protect_system_yes_table[] = { + { "/usr", READONLY, false }, + { "/boot", READONLY, true }, + { "/efi", READONLY, true }, +}; + +/* ProtectSystem=full includes ProtectSystem=yes */ +static const TargetMount protect_system_full_table[] = { + { "/usr", READONLY, false }, + { "/boot", READONLY, true }, + { "/efi", READONLY, true }, + { "/etc", READONLY, false }, +}; + +/* + * ProtectSystem=strict table. In this strict mode, we mount everything + * read-only, except for /proc, /dev, /sys which are the kernel API VFS, + * which are left writable, but PrivateDevices= + ProtectKernelTunables= + * protect those, and these options should be fully orthogonal. + * (And of course /home and friends are also left writable, as ProtectHome= + * shall manage those, orthogonally). + */ +static const TargetMount protect_system_strict_table[] = { + { "/", READONLY, false }, + { "/proc", READWRITE, false }, /* ProtectKernelTunables= */ + { "/sys", READWRITE, false }, /* ProtectKernelTunables= */ + { "/dev", READWRITE, false }, /* PrivateDevices= */ + { "/home", READWRITE, true }, /* ProtectHome= */ + { "/run/user", READWRITE, true }, /* ProtectHome= */ + { "/root", READWRITE, true }, /* ProtectHome= */ +}; + +static void set_bind_mount(BindMount **p, const char *path, MountMode mode, bool ignore) { + (*p)->path = path; + (*p)->mode = mode; + (*p)->ignore = ignore; +} + +static int append_mounts(BindMount **p, char **strv, MountMode mode) { + char **i; + + assert(p); + + STRV_FOREACH(i, strv) { + bool ignore = false; + + if (IN_SET(mode, INACCESSIBLE, READONLY, READWRITE) && startswith(*i, "-")) { + (*i)++; + ignore = true; + } + + if (!path_is_absolute(*i)) + return -EINVAL; + + set_bind_mount(p, *i, mode, ignore); + (*p)++; + } + + return 0; +} + +static int append_target_mounts(BindMount **p, const char *root_directory, const TargetMount *mounts, const size_t size) { + unsigned i; + + assert(p); + assert(mounts); + + for (i = 0; i < size; i++) { + /* + * Here we assume that the ignore field is set during + * declaration we do not support "-" at the beginning. + */ + const TargetMount *m = &mounts[i]; + const char *path = prefix_roota(root_directory, m->path); + + if (!path_is_absolute(path)) + return -EINVAL; + + set_bind_mount(p, path, m->mode, m->ignore); + (*p)++; + } + + return 0; +} + +static int append_protect_kernel_tunables(BindMount **p, const char *root_directory) { + assert(p); + + return append_target_mounts(p, root_directory, protect_kernel_tunables_table, + ELEMENTSOF(protect_kernel_tunables_table)); +} + +static int append_protect_kernel_modules(BindMount **p, const char *root_directory) { + assert(p); + + return append_target_mounts(p, root_directory, protect_kernel_modules_table, + ELEMENTSOF(protect_kernel_modules_table)); +} + +static int append_protect_home(BindMount **p, const char *root_directory, ProtectHome protect_home) { + int r = 0; + + assert(p); + + if (protect_home == PROTECT_HOME_NO) + return 0; + + switch (protect_home) { + case PROTECT_HOME_READ_ONLY: + r = append_target_mounts(p, root_directory, protect_home_read_only_table, + ELEMENTSOF(protect_home_read_only_table)); + break; + case PROTECT_HOME_YES: + r = append_target_mounts(p, root_directory, protect_home_yes_table, + ELEMENTSOF(protect_home_yes_table)); + break; + default: + r = -EINVAL; + break; + } + + return r; +} + +static int append_protect_system(BindMount **p, const char *root_directory, ProtectSystem protect_system) { + int r = 0; + + assert(p); + + if (protect_system == PROTECT_SYSTEM_NO) + return 0; + + switch (protect_system) { + case PROTECT_SYSTEM_STRICT: + r = append_target_mounts(p, root_directory, protect_system_strict_table, + ELEMENTSOF(protect_system_strict_table)); + break; + case PROTECT_SYSTEM_YES: + r = append_target_mounts(p, root_directory, protect_system_yes_table, + ELEMENTSOF(protect_system_yes_table)); + break; + case PROTECT_SYSTEM_FULL: + r = append_target_mounts(p, root_directory, protect_system_full_table, + ELEMENTSOF(protect_system_full_table)); + break; + default: + r = -EINVAL; + break; + } + + return r; +} + +static int mount_path_compare(const void *a, const void *b) { + const BindMount *p = a, *q = b; + int d; + + /* If the paths are not equal, then order prefixes first */ + d = path_compare(p->path, q->path); + if (d != 0) + return d; + + /* If the paths are equal, check the mode */ + if (p->mode < q->mode) + return -1; + + if (p->mode > q->mode) + return 1; + + return 0; +} + +static void drop_duplicates(BindMount *m, unsigned *n) { + BindMount *f, *t, *previous; + + assert(m); + assert(n); + + /* Drops duplicate entries. Expects that the array is properly ordered already. */ + + for (f = m, t = m, previous = NULL; f < m+*n; f++) { + + /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() + * above. */ + if (previous && path_equal(f->path, previous->path)) { + log_debug("%s is duplicate.", f->path); + continue; + } + + *t = *f; + previous = t; + t++; + } + + *n = t - m; +} + +static void drop_inaccessible(BindMount *m, unsigned *n) { + BindMount *f, *t; + const char *clear = NULL; + + assert(m); + assert(n); + + /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly + * ordered already. */ + + for (f = m, t = m; f < m+*n; f++) { + + /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop + * it, as inaccessible paths really should drop the entire subtree. */ + if (clear && path_startswith(f->path, clear)) { + log_debug("%s is masked by %s.", f->path, clear); + continue; + } + + clear = f->mode == INACCESSIBLE ? f->path : NULL; + + *t = *f; + t++; + } + + *n = t - m; +} + +static void drop_nop(BindMount *m, unsigned *n) { + BindMount *f, *t; + + assert(m); + assert(n); + + /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the + * list is ordered by prefixes. */ + + for (f = m, t = m; f < m+*n; f++) { + + /* Only suppress such subtrees for READONLY and READWRITE entries */ + if (IN_SET(f->mode, READONLY, READWRITE)) { + BindMount *p; + bool found = false; + + /* Now let's find the first parent of the entry we are looking at. */ + for (p = t-1; p >= m; p--) { + if (path_startswith(f->path, p->path)) { + found = true; + break; + } + } + + /* We found it, let's see if it's the same mode, if so, we can drop this entry */ + if (found && p->mode == f->mode) { + log_debug("%s is redundant by %s", f->path, p->path); + continue; + } + } + + *t = *f; + t++; + } + + *n = t - m; +} + +static void drop_outside_root(const char *root_directory, BindMount *m, unsigned *n) { + BindMount *f, *t; + + assert(m); + assert(n); + + if (!root_directory) + return; + + /* Drops all mounts that are outside of the root directory. */ + + for (f = m, t = m; f < m+*n; f++) { + + if (!path_startswith(f->path, root_directory)) { + log_debug("%s is outside of root directory.", f->path); + continue; + } + + *t = *f; + t++; + } + + *n = t - m; +} + +static int mount_dev(BindMount *m) { + static const char devnodes[] = + "/dev/null\0" + "/dev/zero\0" + "/dev/full\0" + "/dev/random\0" + "/dev/urandom\0" + "/dev/tty\0"; + + char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; + const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; + _cleanup_umask_ mode_t u; + int r; + + assert(m); + + u = umask(0000); + + if (!mkdtemp(temporary_mount)) + return -errno; + + dev = strjoina(temporary_mount, "/dev"); + (void) mkdir(dev, 0755); + if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { + r = -errno; + goto fail; + } + + devpts = strjoina(temporary_mount, "/dev/pts"); + (void) mkdir(devpts, 0755); + if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { + r = -errno; + goto fail; + } + + devptmx = strjoina(temporary_mount, "/dev/ptmx"); + if (symlink("pts/ptmx", devptmx) < 0) { + r = -errno; + goto fail; + } + + devshm = strjoina(temporary_mount, "/dev/shm"); + (void) mkdir(devshm, 01777); + r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); + if (r < 0) { + r = -errno; + goto fail; + } + + devmqueue = strjoina(temporary_mount, "/dev/mqueue"); + (void) mkdir(devmqueue, 0755); + (void) mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); + + devhugepages = strjoina(temporary_mount, "/dev/hugepages"); + (void) mkdir(devhugepages, 0755); + (void) mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); + + devlog = strjoina(temporary_mount, "/dev/log"); + (void) symlink("/run/systemd/journal/dev-log", devlog); + + NULSTR_FOREACH(d, devnodes) { + _cleanup_free_ char *dn = NULL; + struct stat st; + + r = stat(d, &st); + if (r < 0) { + + if (errno == ENOENT) + continue; + + r = -errno; + goto fail; + } + + if (!S_ISBLK(st.st_mode) && + !S_ISCHR(st.st_mode)) { + r = -EINVAL; + goto fail; + } + + if (st.st_rdev == 0) + continue; + + dn = strappend(temporary_mount, d); + if (!dn) { + r = -ENOMEM; + goto fail; + } + + mac_selinux_create_file_prepare(d, st.st_mode); + r = mknod(dn, st.st_mode, st.st_rdev); + mac_selinux_create_file_clear(); + + if (r < 0) { + r = -errno; + goto fail; + } + } + + dev_setup(temporary_mount, UID_INVALID, GID_INVALID); + + /* Create the /dev directory if missing. It is more likely to be + * missing when the service is started with RootDirectory. This is + * consistent with mount units creating the mount points when missing. + */ + (void) mkdir_p_label(m->path, 0755); + + /* Unmount everything in old /dev */ + umount_recursive(m->path, 0); + if (mount(dev, m->path, NULL, MS_MOVE, NULL) < 0) { + r = -errno; + goto fail; + } + + rmdir(dev); + rmdir(temporary_mount); + + return 0; + +fail: + if (devpts) + umount(devpts); + + if (devshm) + umount(devshm); + + if (devhugepages) + umount(devhugepages); + + if (devmqueue) + umount(devmqueue); + + umount(dev); + rmdir(dev); + rmdir(temporary_mount); + + return r; +} + +static int apply_mount( + BindMount *m, + const char *tmp_dir, + const char *var_tmp_dir) { + + const char *what; + int r; + + assert(m); + + log_debug("Applying namespace mount on %s", m->path); + + switch (m->mode) { + + case INACCESSIBLE: { + struct stat target; + + /* First, get rid of everything that is below if there + * is anything... Then, overmount it with an + * inaccessible path. */ + (void) umount_recursive(m->path, 0); + + if (lstat(m->path, &target) < 0) + return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", m->path); + + what = mode_to_inaccessible_node(target.st_mode); + if (!what) { + log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); + return -ELOOP; + } + break; + } + + case READONLY: + case READWRITE: + + r = path_is_mount_point(m->path, 0); + if (r < 0) + return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", m->path); + if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY bit for the mount point if needed. */ + return 0; + + /* This isn't a mount point yet, let's make it one. */ + what = m->path; + break; + + case PRIVATE_TMP: + what = tmp_dir; + break; + + case PRIVATE_VAR_TMP: + what = var_tmp_dir; + break; + + case PRIVATE_DEV: + return mount_dev(m); + + default: + assert_not_reached("Unknown mode"); + } + + assert(what); + + if (mount(what, m->path, NULL, MS_BIND|MS_REC, NULL) < 0) + return log_debug_errno(errno, "Failed to mount %s to %s: %m", what, m->path); + + log_debug("Successfully mounted %s to %s", what, m->path); + return 0; +} + +static int make_read_only(BindMount *m, char **blacklist) { + int r = 0; + + assert(m); + + if (IN_SET(m->mode, INACCESSIBLE, READONLY)) + r = bind_remount_recursive(m->path, true, blacklist); + else if (m->mode == PRIVATE_DEV) { /* Can be readonly but the submounts can't*/ + if (mount(NULL, m->path, NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0) + r = -errno; + } else + return 0; + + /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only + * already stays this way. This improves compatibility with container managers, where we won't attempt to undo + * read-only mounts already applied. */ + + return r; +} + +static int chase_all_symlinks(const char *root_directory, BindMount *m, unsigned *n) { + BindMount *f, *t; + int r; + + assert(m); + assert(n); + + /* Since mount() will always follow symlinks and we need to take the different root directory into account we + * chase the symlinks on our own first. This call wil do so for all entries and remove all entries where we + * can't resolve the path, and which have been marked for such removal. */ + + for (f = m, t = m; f < m+*n; f++) { + + r = chase_symlinks(f->path, root_directory, &f->chased); + if (r == -ENOENT && f->ignore) /* Doesn't exist? Then remove it! */ + continue; + if (r < 0) + return log_debug_errno(r, "Failed to chase symlinks for %s: %m", f->path); + + if (path_equal(f->path, f->chased)) + f->chased = mfree(f->chased); + else { + log_debug("Chased %s → %s", f->path, f->chased); + f->path = f->chased; + } + + *t = *f; + t++; + } + + *n = t - m; + return 0; +} + +static unsigned namespace_calculate_mounts( + const NameSpaceInfo *ns_info, + char** read_write_paths, + char** read_only_paths, + char** inaccessible_paths, + const char* tmp_dir, + const char* var_tmp_dir, + ProtectHome protect_home, + ProtectSystem protect_system) { + + unsigned protect_home_cnt; + unsigned protect_system_cnt = + (protect_system == PROTECT_SYSTEM_STRICT ? + ELEMENTSOF(protect_system_strict_table) : + ((protect_system == PROTECT_SYSTEM_FULL) ? + ELEMENTSOF(protect_system_full_table) : + ((protect_system == PROTECT_SYSTEM_YES) ? + ELEMENTSOF(protect_system_yes_table) : 0))); + + protect_home_cnt = + (protect_home == PROTECT_HOME_YES ? + ELEMENTSOF(protect_home_yes_table) : + ((protect_home == PROTECT_HOME_READ_ONLY) ? + ELEMENTSOF(protect_home_read_only_table) : 0)); + + return !!tmp_dir + !!var_tmp_dir + + strv_length(read_write_paths) + + strv_length(read_only_paths) + + strv_length(inaccessible_paths) + + ns_info->private_dev + + (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + + (ns_info->protect_control_groups ? 1 : 0) + + (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + + protect_home_cnt + protect_system_cnt; +} + +int setup_namespace( + const char* root_directory, + const NameSpaceInfo *ns_info, + char** read_write_paths, + char** read_only_paths, + char** inaccessible_paths, + const char* tmp_dir, + const char* var_tmp_dir, + ProtectHome protect_home, + ProtectSystem protect_system, + unsigned long mount_flags) { + + BindMount *m, *mounts = NULL; + bool make_slave = false; + unsigned n; + int r = 0; + + if (mount_flags == 0) + mount_flags = MS_SHARED; + + n = namespace_calculate_mounts(ns_info, + read_write_paths, + read_only_paths, + inaccessible_paths, + tmp_dir, var_tmp_dir, + protect_home, protect_system); + + /* Set mount slave mode */ + if (root_directory || n > 0) + make_slave = true; + + if (n > 0) { + m = mounts = (BindMount *) alloca0(n * sizeof(BindMount)); + r = append_mounts(&m, read_write_paths, READWRITE); + if (r < 0) + return r; + + r = append_mounts(&m, read_only_paths, READONLY); + if (r < 0) + return r; + + r = append_mounts(&m, inaccessible_paths, INACCESSIBLE); + if (r < 0) + return r; + + if (tmp_dir) { + m->path = prefix_roota(root_directory, "/tmp"); + m->mode = PRIVATE_TMP; + m++; + } + + if (var_tmp_dir) { + m->path = prefix_roota(root_directory, "/var/tmp"); + m->mode = PRIVATE_VAR_TMP; + m++; + } + + if (ns_info->private_dev) { + m->path = prefix_roota(root_directory, "/dev"); + m->mode = PRIVATE_DEV; + m++; + } + + if (ns_info->protect_kernel_tunables) { + r = append_protect_kernel_tunables(&m, root_directory); + if (r < 0) + return r; + } + + if (ns_info->protect_kernel_modules) { + r = append_protect_kernel_modules(&m, root_directory); + if (r < 0) + return r; + } + + if (ns_info->protect_control_groups) { + m->path = prefix_roota(root_directory, "/sys/fs/cgroup"); + m->mode = READONLY; + m++; + } + + r = append_protect_home(&m, root_directory, protect_home); + if (r < 0) + return r; + + r = append_protect_system(&m, root_directory, protect_system); + if (r < 0) + return r; + + assert(mounts + n == m); + + /* Resolve symlinks manually first, as mount() will always follow them relative to the host's + * root. Moreover we want to suppress duplicates based on the resolved paths. This of course is a bit + * racy. */ + r = chase_all_symlinks(root_directory, mounts, &n); + if (r < 0) + goto finish; + + qsort(mounts, n, sizeof(BindMount), mount_path_compare); + + drop_duplicates(mounts, &n); + drop_outside_root(root_directory, mounts, &n); + drop_inaccessible(mounts, &n); + drop_nop(mounts, &n); + } + + if (unshare(CLONE_NEWNS) < 0) { + r = -errno; + goto finish; + } + + if (make_slave) { + /* Remount / as SLAVE so that nothing now mounted in the namespace + shows up in the parent */ + if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { + r = -errno; + goto finish; + } + } + + if (root_directory) { + /* Turn directory into bind mount, if it isn't one yet */ + r = path_is_mount_point(root_directory, AT_SYMLINK_FOLLOW); + if (r < 0) + goto finish; + if (r == 0) { + if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0) { + r = -errno; + goto finish; + } + } + } + + if (n > 0) { + char **blacklist; + unsigned j; + + /* First round, add in all special mounts we need */ + for (m = mounts; m < mounts + n; ++m) { + r = apply_mount(m, tmp_dir, var_tmp_dir); + if (r < 0) + goto finish; + } + + /* Create a blacklist we can pass to bind_mount_recursive() */ + blacklist = newa(char*, n+1); + for (j = 0; j < n; j++) + blacklist[j] = (char*) mounts[j].path; + blacklist[j] = NULL; + + /* Second round, flip the ro bits if necessary. */ + for (m = mounts; m < mounts + n; ++m) { + r = make_read_only(m, blacklist); + if (r < 0) + goto finish; + } + } + + if (root_directory) { + /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ + r = mount_move_root(root_directory); + if (r < 0) + goto finish; + } + + /* Remount / as the desired mode. Not that this will not + * reestablish propagation from our side to the host, since + * what's disconnected is disconnected. */ + if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { + r = -errno; + goto finish; + } + + r = 0; + +finish: + for (m = mounts; m < mounts + n; m++) + free(m->chased); + + return r; +} + +static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { + _cleanup_free_ char *x = NULL; + char bid[SD_ID128_STRING_MAX]; + sd_id128_t boot_id; + int r; + + assert(id); + assert(prefix); + assert(path); + + /* We include the boot id in the directory so that after a + * reboot we can easily identify obsolete directories. */ + + r = sd_id128_get_boot(&boot_id); + if (r < 0) + return r; + + x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX", NULL); + if (!x) + return -ENOMEM; + + RUN_WITH_UMASK(0077) + if (!mkdtemp(x)) + return -errno; + + RUN_WITH_UMASK(0000) { + char *y; + + y = strjoina(x, "/tmp"); + + if (mkdir(y, 0777 | S_ISVTX) < 0) + return -errno; + } + + *path = x; + x = NULL; + + return 0; +} + +int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { + char *a, *b; + int r; + + assert(id); + assert(tmp_dir); + assert(var_tmp_dir); + + r = setup_one_tmp_dir(id, "/tmp", &a); + if (r < 0) + return r; + + r = setup_one_tmp_dir(id, "/var/tmp", &b); + if (r < 0) { + char *t; + + t = strjoina(a, "/tmp"); + rmdir(t); + rmdir(a); + + free(a); + return r; + } + + *tmp_dir = a; + *var_tmp_dir = b; + + return 0; +} + +int setup_netns(int netns_storage_socket[2]) { + _cleanup_close_ int netns = -1; + int r, q; + + assert(netns_storage_socket); + assert(netns_storage_socket[0] >= 0); + assert(netns_storage_socket[1] >= 0); + + /* We use the passed socketpair as a storage buffer for our + * namespace reference fd. Whatever process runs this first + * shall create a new namespace, all others should just join + * it. To serialize that we use a file lock on the socket + * pair. + * + * It's a bit crazy, but hey, works great! */ + + if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) + return -errno; + + netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); + if (netns == -EAGAIN) { + /* Nothing stored yet, so let's create a new namespace */ + + if (unshare(CLONE_NEWNET) < 0) { + r = -errno; + goto fail; + } + + loopback_setup(); + + netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); + if (netns < 0) { + r = -errno; + goto fail; + } + + r = 1; + + } else if (netns < 0) { + r = netns; + goto fail; + + } else { + /* Yay, found something, so let's join the namespace */ + if (setns(netns, CLONE_NEWNET) < 0) { + r = -errno; + goto fail; + } + + r = 0; + } + + q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); + if (q < 0) { + r = q; + goto fail; + } + +fail: + (void) lockf(netns_storage_socket[0], F_ULOCK, 0); + return r; +} + +static const char *const protect_home_table[_PROTECT_HOME_MAX] = { + [PROTECT_HOME_NO] = "no", + [PROTECT_HOME_YES] = "yes", + [PROTECT_HOME_READ_ONLY] = "read-only", +}; + +DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome); + +static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { + [PROTECT_SYSTEM_NO] = "no", + [PROTECT_SYSTEM_YES] = "yes", + [PROTECT_SYSTEM_FULL] = "full", + [PROTECT_SYSTEM_STRICT] = "strict", +}; + +DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem); diff --git a/src/grp-system/libcore/src/path.c b/src/grp-system/libcore/src/path.c new file mode 100644 index 0000000000..a1b0bdd042 --- /dev/null +++ b/src/grp-system/libcore/src/path.c @@ -0,0 +1,789 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/epoll.h> +#include <sys/inotify.h> +#include <unistd.h> + +#include "core/path.h" +#include "core/unit.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/glob-util.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" + +#include "dbus-path.h" + +static const UnitActiveState state_translation_table[_PATH_STATE_MAX] = { + [PATH_DEAD] = UNIT_INACTIVE, + [PATH_WAITING] = UNIT_ACTIVE, + [PATH_RUNNING] = UNIT_ACTIVE, + [PATH_FAILED] = UNIT_FAILED +}; + +static int path_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); + +int path_spec_watch(PathSpec *s, sd_event_io_handler_t handler) { + + static const int flags_table[_PATH_TYPE_MAX] = { + [PATH_EXISTS] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB, + [PATH_EXISTS_GLOB] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB, + [PATH_CHANGED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO, + [PATH_MODIFIED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO|IN_MODIFY, + [PATH_DIRECTORY_NOT_EMPTY] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CREATE|IN_MOVED_TO + }; + + bool exists = false; + char *slash, *oldslash = NULL; + int r; + + assert(s); + assert(s->unit); + assert(handler); + + path_spec_unwatch(s); + + s->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (s->inotify_fd < 0) { + r = -errno; + goto fail; + } + + r = sd_event_add_io(s->unit->manager->event, &s->event_source, s->inotify_fd, EPOLLIN, handler, s); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(s->event_source, "path"); + + /* This assumes the path was passed through path_kill_slashes()! */ + + for (slash = strchr(s->path, '/'); ; slash = strchr(slash+1, '/')) { + char *cut = NULL; + int flags; + char tmp; + + if (slash) { + cut = slash + (slash == s->path); + tmp = *cut; + *cut = '\0'; + + flags = IN_MOVE_SELF | IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO; + } else + flags = flags_table[s->type]; + + r = inotify_add_watch(s->inotify_fd, s->path, flags); + if (r < 0) { + if (errno == EACCES || errno == ENOENT) { + if (cut) + *cut = tmp; + break; + } + + r = log_warning_errno(errno, "Failed to add watch on %s: %s", s->path, errno == ENOSPC ? "too many watches" : strerror(-r)); + if (cut) + *cut = tmp; + goto fail; + } else { + exists = true; + + /* Path exists, we don't need to watch parent too closely. */ + if (oldslash) { + char *cut2 = oldslash + (oldslash == s->path); + char tmp2 = *cut2; + *cut2 = '\0'; + + (void) inotify_add_watch(s->inotify_fd, s->path, IN_MOVE_SELF); + /* Error is ignored, the worst can happen is we get spurious events. */ + + *cut2 = tmp2; + } + } + + if (cut) + *cut = tmp; + + if (slash) + oldslash = slash; + else { + /* whole path has been iterated over */ + s->primary_wd = r; + break; + } + } + + if (!exists) { + r = log_error_errno(errno, "Failed to add watch on any of the components of %s: %m", s->path); + /* either EACCESS or ENOENT */ + goto fail; + } + + return 0; + +fail: + path_spec_unwatch(s); + return r; +} + +void path_spec_unwatch(PathSpec *s) { + assert(s); + + s->event_source = sd_event_source_unref(s->event_source); + s->inotify_fd = safe_close(s->inotify_fd); +} + +int path_spec_fd_event(PathSpec *s, uint32_t revents) { + union inotify_event_buffer buffer; + struct inotify_event *e; + ssize_t l; + int r = 0; + + if (revents != EPOLLIN) { + log_error("Got invalid poll event on inotify."); + return -EINVAL; + } + + l = read(s->inotify_fd, &buffer, sizeof(buffer)); + if (l < 0) { + if (errno == EAGAIN || errno == EINTR) + return 0; + + return log_error_errno(errno, "Failed to read inotify event: %m"); + } + + FOREACH_INOTIFY_EVENT(e, buffer, l) { + if ((s->type == PATH_CHANGED || s->type == PATH_MODIFIED) && + s->primary_wd == e->wd) + r = 1; + } + + return r; +} + +static bool path_spec_check_good(PathSpec *s, bool initial) { + bool good = false; + + switch (s->type) { + + case PATH_EXISTS: + good = access(s->path, F_OK) >= 0; + break; + + case PATH_EXISTS_GLOB: + good = glob_exists(s->path) > 0; + break; + + case PATH_DIRECTORY_NOT_EMPTY: { + int k; + + k = dir_is_empty(s->path); + good = !(k == -ENOENT || k > 0); + break; + } + + case PATH_CHANGED: + case PATH_MODIFIED: { + bool b; + + b = access(s->path, F_OK) >= 0; + good = !initial && b != s->previous_exists; + s->previous_exists = b; + break; + } + + default: + ; + } + + return good; +} + +static void path_spec_mkdir(PathSpec *s, mode_t mode) { + int r; + + if (s->type == PATH_EXISTS || s->type == PATH_EXISTS_GLOB) + return; + + r = mkdir_p_label(s->path, mode); + if (r < 0) + log_warning_errno(r, "mkdir(%s) failed: %m", s->path); +} + +static void path_spec_dump(PathSpec *s, FILE *f, const char *prefix) { + fprintf(f, + "%s%s: %s\n", + prefix, + path_type_to_string(s->type), + s->path); +} + +void path_spec_done(PathSpec *s) { + assert(s); + assert(s->inotify_fd == -1); + + free(s->path); +} + +static void path_init(Unit *u) { + Path *p = PATH(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + p->directory_mode = 0755; +} + +void path_free_specs(Path *p) { + PathSpec *s; + + assert(p); + + while ((s = p->specs)) { + path_spec_unwatch(s); + LIST_REMOVE(spec, p->specs, s); + path_spec_done(s); + free(s); + } +} + +static void path_done(Unit *u) { + Path *p = PATH(u); + + assert(p); + + path_free_specs(p); +} + +static int path_add_mount_links(Path *p) { + PathSpec *s; + int r; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) { + r = unit_require_mounts_for(UNIT(p), s->path); + if (r < 0) + return r; + } + + return 0; +} + +static int path_verify(Path *p) { + assert(p); + + if (UNIT(p)->load_state != UNIT_LOADED) + return 0; + + if (!p->specs) { + log_unit_error(UNIT(p), "Path unit lacks path setting. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int path_add_default_dependencies(Path *p) { + int r; + + assert(p); + + if (!UNIT(p)->default_dependencies) + return 0; + + r = unit_add_dependency_by_name(UNIT(p), UNIT_BEFORE, SPECIAL_PATHS_TARGET, NULL, true); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(UNIT(p)->manager)) { + r = unit_add_two_dependencies_by_name(UNIT(p), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(p), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int path_load(Unit *u) { + Path *p = PATH(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + + if (set_isempty(u->dependencies[UNIT_TRIGGERS])) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + } + + r = path_add_mount_links(p); + if (r < 0) + return r; + + r = path_add_default_dependencies(p); + if (r < 0) + return r; + } + + return path_verify(p); +} + +static void path_dump(Unit *u, FILE *f, const char *prefix) { + Path *p = PATH(u); + Unit *trigger; + PathSpec *s; + + assert(p); + assert(f); + + trigger = UNIT_TRIGGER(u); + + fprintf(f, + "%sPath State: %s\n" + "%sResult: %s\n" + "%sUnit: %s\n" + "%sMakeDirectory: %s\n" + "%sDirectoryMode: %04o\n", + prefix, path_state_to_string(p->state), + prefix, path_result_to_string(p->result), + prefix, trigger ? trigger->id : "n/a", + prefix, yes_no(p->make_directory), + prefix, p->directory_mode); + + LIST_FOREACH(spec, s, p->specs) + path_spec_dump(s, f, prefix); +} + +static void path_unwatch(Path *p) { + PathSpec *s; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) + path_spec_unwatch(s); +} + +static int path_watch(Path *p) { + int r; + PathSpec *s; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) { + r = path_spec_watch(s, path_dispatch_io); + if (r < 0) + return r; + } + + return 0; +} + +static void path_set_state(Path *p, PathState state) { + PathState old_state; + assert(p); + + old_state = p->state; + p->state = state; + + if (state != PATH_WAITING && + (state != PATH_RUNNING || p->inotify_triggered)) + path_unwatch(p); + + if (state != old_state) + log_unit_debug(UNIT(p), "Changed %s -> %s", path_state_to_string(old_state), path_state_to_string(state)); + + unit_notify(UNIT(p), state_translation_table[old_state], state_translation_table[state], true); +} + +static void path_enter_waiting(Path *p, bool initial, bool recheck); + +static int path_coldplug(Unit *u) { + Path *p = PATH(u); + + assert(p); + assert(p->state == PATH_DEAD); + + if (p->deserialized_state != p->state) { + + if (p->deserialized_state == PATH_WAITING || + p->deserialized_state == PATH_RUNNING) + path_enter_waiting(p, true, true); + else + path_set_state(p, p->deserialized_state); + } + + return 0; +} + +static void path_enter_dead(Path *p, PathResult f) { + assert(p); + + if (p->result == PATH_SUCCESS) + p->result = f; + + path_set_state(p, p->result != PATH_SUCCESS ? PATH_FAILED : PATH_DEAD); +} + +static void path_enter_running(Path *p) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + Unit *trigger; + int r; + + assert(p); + + /* Don't start job if we are supposed to go down */ + if (unit_stop_pending(UNIT(p))) + return; + + trigger = UNIT_TRIGGER(UNIT(p)); + if (!trigger) { + log_unit_error(UNIT(p), "Unit to trigger vanished."); + path_enter_dead(p, PATH_FAILURE_RESOURCES); + return; + } + + r = manager_add_job(UNIT(p)->manager, JOB_START, trigger, JOB_REPLACE, &error, NULL); + if (r < 0) + goto fail; + + p->inotify_triggered = false; + + r = path_watch(p); + if (r < 0) + goto fail; + + path_set_state(p, PATH_RUNNING); + return; + +fail: + log_unit_warning(UNIT(p), "Failed to queue unit startup job: %s", bus_error_message(&error, r)); + path_enter_dead(p, PATH_FAILURE_RESOURCES); +} + +static bool path_check_good(Path *p, bool initial) { + PathSpec *s; + bool good = false; + + assert(p); + + LIST_FOREACH(spec, s, p->specs) { + good = path_spec_check_good(s, initial); + + if (good) + break; + } + + return good; +} + +static void path_enter_waiting(Path *p, bool initial, bool recheck) { + int r; + + if (recheck) + if (path_check_good(p, initial)) { + log_unit_debug(UNIT(p), "Got triggered."); + path_enter_running(p); + return; + } + + r = path_watch(p); + if (r < 0) + goto fail; + + /* Hmm, so now we have created inotify watches, but the file + * might have appeared/been removed by now, so we must + * recheck */ + + if (recheck) + if (path_check_good(p, false)) { + log_unit_debug(UNIT(p), "Got triggered."); + path_enter_running(p); + return; + } + + path_set_state(p, PATH_WAITING); + return; + +fail: + log_unit_warning_errno(UNIT(p), r, "Failed to enter waiting state: %m"); + path_enter_dead(p, PATH_FAILURE_RESOURCES); +} + +static void path_mkdir(Path *p) { + PathSpec *s; + + assert(p); + + if (!p->make_directory) + return; + + LIST_FOREACH(spec, s, p->specs) + path_spec_mkdir(s, p->directory_mode); +} + +static int path_start(Unit *u) { + Path *p = PATH(u); + Unit *trigger; + int r; + + assert(p); + assert(p->state == PATH_DEAD || p->state == PATH_FAILED); + + trigger = UNIT_TRIGGER(u); + if (!trigger || trigger->load_state != UNIT_LOADED) { + log_unit_error(u, "Refusing to start, unit to trigger not loaded."); + return -ENOENT; + } + + r = unit_start_limit_test(u); + if (r < 0) { + path_enter_dead(p, PATH_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + path_mkdir(p); + + p->result = PATH_SUCCESS; + path_enter_waiting(p, true, true); + + return 1; +} + +static int path_stop(Unit *u) { + Path *p = PATH(u); + + assert(p); + assert(p->state == PATH_WAITING || p->state == PATH_RUNNING); + + path_enter_dead(p, PATH_SUCCESS); + return 1; +} + +static int path_serialize(Unit *u, FILE *f, FDSet *fds) { + Path *p = PATH(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", path_state_to_string(p->state)); + unit_serialize_item(u, f, "result", path_result_to_string(p->result)); + + return 0; +} + +static int path_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Path *p = PATH(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + PathState state; + + state = path_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + p->deserialized_state = state; + + } else if (streq(key, "result")) { + PathResult f; + + f = path_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != PATH_SUCCESS) + p->result = f; + + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState path_active_state(Unit *u) { + assert(u); + + return state_translation_table[PATH(u)->state]; +} + +_pure_ static const char *path_sub_state_to_string(Unit *u) { + assert(u); + + return path_state_to_string(PATH(u)->state); +} + +static int path_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + PathSpec *s = userdata; + Path *p; + int changed; + + assert(s); + assert(s->unit); + assert(fd >= 0); + + p = PATH(s->unit); + + if (p->state != PATH_WAITING && + p->state != PATH_RUNNING) + return 0; + + /* log_debug("inotify wakeup on %s.", u->id); */ + + LIST_FOREACH(spec, s, p->specs) + if (path_spec_owns_inotify_fd(s, fd)) + break; + + if (!s) { + log_error("Got event on unknown fd."); + goto fail; + } + + changed = path_spec_fd_event(s, revents); + if (changed < 0) + goto fail; + + /* If we are already running, then remember that one event was + * dispatched so that we restart the service only if something + * actually changed on disk */ + p->inotify_triggered = true; + + if (changed) + path_enter_running(p); + else + path_enter_waiting(p, false, true); + + return 0; + +fail: + path_enter_dead(p, PATH_FAILURE_RESOURCES); + return 0; +} + +static void path_trigger_notify(Unit *u, Unit *other) { + Path *p = PATH(u); + + assert(u); + assert(other); + + /* Invoked whenever the unit we trigger changes state or gains + * or loses a job */ + + if (other->load_state != UNIT_LOADED) + return; + + if (p->state == PATH_RUNNING && + UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) { + log_unit_debug(UNIT(p), "Got notified about unit deactivation."); + + /* Hmm, so inotify was triggered since the + * last activation, so I guess we need to + * recheck what is going on. */ + path_enter_waiting(p, false, p->inotify_triggered); + } +} + +static void path_reset_failed(Unit *u) { + Path *p = PATH(u); + + assert(p); + + if (p->state == PATH_FAILED) + path_set_state(p, PATH_DEAD); + + p->result = PATH_SUCCESS; +} + +static const char* const path_type_table[_PATH_TYPE_MAX] = { + [PATH_EXISTS] = "PathExists", + [PATH_EXISTS_GLOB] = "PathExistsGlob", + [PATH_DIRECTORY_NOT_EMPTY] = "DirectoryNotEmpty", + [PATH_CHANGED] = "PathChanged", + [PATH_MODIFIED] = "PathModified", +}; + +DEFINE_STRING_TABLE_LOOKUP(path_type, PathType); + +static const char* const path_result_table[_PATH_RESULT_MAX] = { + [PATH_SUCCESS] = "success", + [PATH_FAILURE_RESOURCES] = "resources", + [PATH_FAILURE_START_LIMIT_HIT] = "start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(path_result, PathResult); + +const UnitVTable path_vtable = { + .object_size = sizeof(Path), + + .sections = + "Unit\0" + "Path\0" + "Install\0", + + .init = path_init, + .done = path_done, + .load = path_load, + + .coldplug = path_coldplug, + + .dump = path_dump, + + .start = path_start, + .stop = path_stop, + + .serialize = path_serialize, + .deserialize_item = path_deserialize_item, + + .active_state = path_active_state, + .sub_state_to_string = path_sub_state_to_string, + + .trigger_notify = path_trigger_notify, + + .reset_failed = path_reset_failed, + + .bus_vtable = bus_path_vtable +}; diff --git a/src/grp-system/libcore/src/scope.c b/src/grp-system/libcore/src/scope.c new file mode 100644 index 0000000000..da1aad4a32 --- /dev/null +++ b/src/grp-system/libcore/src/scope.c @@ -0,0 +1,637 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <unistd.h> + +#include "core/scope.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" + +#include "dbus-scope.h" +#include "load-dropin.h" + +static const UnitActiveState state_translation_table[_SCOPE_STATE_MAX] = { + [SCOPE_DEAD] = UNIT_INACTIVE, + [SCOPE_RUNNING] = UNIT_ACTIVE, + [SCOPE_ABANDONED] = UNIT_ACTIVE, + [SCOPE_STOP_SIGTERM] = UNIT_DEACTIVATING, + [SCOPE_STOP_SIGKILL] = UNIT_DEACTIVATING, + [SCOPE_FAILED] = UNIT_FAILED +}; + +static int scope_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); + +static void scope_init(Unit *u) { + Scope *s = SCOPE(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + s->timeout_stop_usec = u->manager->default_timeout_stop_usec; + u->ignore_on_isolate = true; +} + +static void scope_done(Unit *u) { + Scope *s = SCOPE(u); + + assert(u); + + free(s->controller); + + s->timer_event_source = sd_event_source_unref(s->timer_event_source); +} + +static int scope_arm_timer(Scope *s, usec_t usec) { + int r; + + assert(s); + + if (s->timer_event_source) { + r = sd_event_source_set_time(s->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(s->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + scope_dispatch_timer, s); + if (r < 0) + return r; + + (void) sd_event_source_set_description(s->timer_event_source, "scope-timer"); + + return 0; +} + +static void scope_set_state(Scope *s, ScopeState state) { + ScopeState old_state; + assert(s); + + old_state = s->state; + s->state = state; + + if (!IN_SET(state, SCOPE_STOP_SIGTERM, SCOPE_STOP_SIGKILL)) + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + + if (IN_SET(state, SCOPE_DEAD, SCOPE_FAILED)) + unit_unwatch_all_pids(UNIT(s)); + + if (state != old_state) + log_debug("%s changed %s -> %s", UNIT(s)->id, scope_state_to_string(old_state), scope_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); +} + +static int scope_add_default_dependencies(Scope *s) { + int r; + + assert(s); + + if (!UNIT(s)->default_dependencies) + return 0; + + /* Make sure scopes are unloaded on shutdown */ + r = unit_add_two_dependencies_by_name( + UNIT(s), + UNIT_BEFORE, UNIT_CONFLICTS, + SPECIAL_SHUTDOWN_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int scope_verify(Scope *s) { + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (set_isempty(UNIT(s)->pids) && + !MANAGER_IS_RELOADING(UNIT(s)->manager) && + !unit_has_name(UNIT(s), SPECIAL_INIT_SCOPE)) { + log_unit_error(UNIT(s), "Scope has no PIDs. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int scope_load_init_scope(Unit *u) { + assert(u); + + if (!unit_has_name(u, SPECIAL_INIT_SCOPE)) + return 0; + + u->transient = true; + u->perpetual = true; + + /* init.scope is a bit special, as it has to stick around forever. Because of its special semantics we + * synthesize it here, instead of relying on the unit file on disk. */ + + u->default_dependencies = false; + u->ignore_on_isolate = true; + + SCOPE(u)->kill_context.kill_signal = SIGRTMIN+14; + + /* Prettify things, if we can. */ + if (!u->description) + u->description = strdup("System and Service Manager"); + if (!u->documentation) + (void) strv_extend(&u->documentation, "man:systemd(1)"); + + return 1; +} + +static int scope_load(Unit *u) { + Scope *s = SCOPE(u); + int r; + + assert(s); + assert(u->load_state == UNIT_STUB); + + if (!u->transient && !MANAGER_IS_RELOADING(u->manager)) + /* Refuse to load non-transient scope units, but allow them while reloading. */ + return -ENOENT; + + r = scope_load_init_scope(u); + if (r < 0) + return r; + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + r = unit_patch_contexts(u); + if (r < 0) + return r; + + r = unit_set_default_slice(u); + if (r < 0) + return r; + + r = scope_add_default_dependencies(s); + if (r < 0) + return r; + } + + return scope_verify(s); +} + +static int scope_coldplug(Unit *u) { + Scope *s = SCOPE(u); + int r; + + assert(s); + assert(s->state == SCOPE_DEAD); + + if (s->deserialized_state == s->state) + return 0; + + if (IN_SET(s->deserialized_state, SCOPE_STOP_SIGKILL, SCOPE_STOP_SIGTERM)) { + r = scope_arm_timer(s, usec_add(u->state_change_timestamp.monotonic, s->timeout_stop_usec)); + if (r < 0) + return r; + } + + if (!IN_SET(s->deserialized_state, SCOPE_DEAD, SCOPE_FAILED)) + unit_watch_all_pids(UNIT(s)); + + scope_set_state(s, s->deserialized_state); + return 0; +} + +static void scope_dump(Unit *u, FILE *f, const char *prefix) { + Scope *s = SCOPE(u); + + assert(s); + assert(f); + + fprintf(f, + "%sScope State: %s\n" + "%sResult: %s\n", + prefix, scope_state_to_string(s->state), + prefix, scope_result_to_string(s->result)); + + cgroup_context_dump(&s->cgroup_context, f, prefix); + kill_context_dump(&s->kill_context, f, prefix); +} + +static void scope_enter_dead(Scope *s, ScopeResult f) { + assert(s); + + if (s->result == SCOPE_SUCCESS) + s->result = f; + + scope_set_state(s, s->result != SCOPE_SUCCESS ? SCOPE_FAILED : SCOPE_DEAD); +} + +static void scope_enter_signal(Scope *s, ScopeState state, ScopeResult f) { + bool skip_signal = false; + int r; + + assert(s); + + if (s->result == SCOPE_SUCCESS) + s->result = f; + + unit_watch_all_pids(UNIT(s)); + + /* If we have a controller set let's ask the controller nicely + * to terminate the scope, instead of us going directly into + * SIGTERM berserk mode */ + if (state == SCOPE_STOP_SIGTERM) + skip_signal = bus_scope_send_request_stop(s) > 0; + + if (!skip_signal) { + r = unit_kill_context( + UNIT(s), + &s->kill_context, + state != SCOPE_STOP_SIGTERM ? KILL_KILL : + s->was_abandoned ? KILL_TERMINATE_AND_LOG : + KILL_TERMINATE, + -1, -1, false); + if (r < 0) + goto fail; + } else + r = 1; + + if (r > 0) { + r = scope_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_stop_usec)); + if (r < 0) + goto fail; + + scope_set_state(s, state); + } else if (state == SCOPE_STOP_SIGTERM) + scope_enter_signal(s, SCOPE_STOP_SIGKILL, SCOPE_SUCCESS); + else + scope_enter_dead(s, SCOPE_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to kill processes: %m"); + + scope_enter_dead(s, SCOPE_FAILURE_RESOURCES); +} + +static int scope_start(Unit *u) { + Scope *s = SCOPE(u); + int r; + + assert(s); + + if (unit_has_name(u, SPECIAL_INIT_SCOPE)) + return -EPERM; + + if (s->state == SCOPE_FAILED) + return -EPERM; + + /* We can't fulfill this right now, please try again later */ + if (s->state == SCOPE_STOP_SIGTERM || + s->state == SCOPE_STOP_SIGKILL) + return -EAGAIN; + + assert(s->state == SCOPE_DEAD); + + if (!u->transient && !MANAGER_IS_RELOADING(u->manager)) + return -ENOENT; + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + (void) unit_realize_cgroup(u); + (void) unit_reset_cpu_usage(u); + + r = unit_attach_pids_to_cgroup(u); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to add PIDs to scope's control group: %m"); + scope_enter_dead(s, SCOPE_FAILURE_RESOURCES); + return r; + } + + s->result = SCOPE_SUCCESS; + + scope_set_state(s, SCOPE_RUNNING); + return 1; +} + +static int scope_stop(Unit *u) { + Scope *s = SCOPE(u); + + assert(s); + + if (s->state == SCOPE_STOP_SIGTERM || + s->state == SCOPE_STOP_SIGKILL) + return 0; + + assert(s->state == SCOPE_RUNNING || + s->state == SCOPE_ABANDONED); + + scope_enter_signal(s, SCOPE_STOP_SIGTERM, SCOPE_SUCCESS); + return 1; +} + +static void scope_reset_failed(Unit *u) { + Scope *s = SCOPE(u); + + assert(s); + + if (s->state == SCOPE_FAILED) + scope_set_state(s, SCOPE_DEAD); + + s->result = SCOPE_SUCCESS; +} + +static int scope_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, -1, error); +} + +static int scope_get_timeout(Unit *u, usec_t *timeout) { + Scope *s = SCOPE(u); + usec_t t; + int r; + + if (!s->timer_event_source) + return 0; + + r = sd_event_source_get_time(s->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +static int scope_serialize(Unit *u, FILE *f, FDSet *fds) { + Scope *s = SCOPE(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", scope_state_to_string(s->state)); + unit_serialize_item(u, f, "was-abandoned", yes_no(s->was_abandoned)); + return 0; +} + +static int scope_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Scope *s = SCOPE(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + ScopeState state; + + state = scope_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + s->deserialized_state = state; + + } else if (streq(key, "was-abandoned")) { + int k; + + k = parse_boolean(value); + if (k < 0) + log_unit_debug(u, "Failed to parse boolean value: %s", value); + else + s->was_abandoned = k; + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +static bool scope_check_gc(Unit *u) { + assert(u); + + /* Never clean up scopes that still have a process around, + * even if the scope is formally dead. */ + + if (!u->cgroup_path) + return false; + + return cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path) <= 0; +} + +static void scope_notify_cgroup_empty_event(Unit *u) { + Scope *s = SCOPE(u); + assert(u); + + log_unit_debug(u, "cgroup is empty"); + + if (IN_SET(s->state, SCOPE_RUNNING, SCOPE_ABANDONED, SCOPE_STOP_SIGTERM, SCOPE_STOP_SIGKILL)) + scope_enter_dead(s, SCOPE_SUCCESS); +} + +static void scope_sigchld_event(Unit *u, pid_t pid, int code, int status) { + + /* If we get a SIGCHLD event for one of the processes we were + interested in, then we look for others to watch, under the + assumption that we'll sooner or later get a SIGCHLD for + them, as the original process we watched was probably the + parent of them, and they are hence now our children. */ + + unit_tidy_watch_pids(u, 0, 0); + unit_watch_all_pids(u); + + /* If the PID set is empty now, then let's finish this off + (On unified we use proper notifications) */ + if (cg_unified(SYSTEMD_CGROUP_CONTROLLER) <= 0 && set_isempty(u->pids)) + scope_notify_cgroup_empty_event(u); +} + +static int scope_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + Scope *s = SCOPE(userdata); + + assert(s); + assert(s->timer_event_source == source); + + switch (s->state) { + + case SCOPE_STOP_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "Stopping timed out. Killing."); + scope_enter_signal(s, SCOPE_STOP_SIGKILL, SCOPE_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "Stopping timed out. Skipping SIGKILL."); + scope_enter_dead(s, SCOPE_FAILURE_TIMEOUT); + } + + break; + + case SCOPE_STOP_SIGKILL: + log_unit_warning(UNIT(s), "Still around after SIGKILL. Ignoring."); + scope_enter_dead(s, SCOPE_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +int scope_abandon(Scope *s) { + assert(s); + + if (unit_has_name(UNIT(s), SPECIAL_INIT_SCOPE)) + return -EPERM; + + if (!IN_SET(s->state, SCOPE_RUNNING, SCOPE_ABANDONED)) + return -ESTALE; + + s->was_abandoned = true; + s->controller = mfree(s->controller); + + /* The client is no longer watching the remaining processes, + * so let's step in here, under the assumption that the + * remaining processes will be sooner or later reassigned to + * us as parent. */ + + unit_tidy_watch_pids(UNIT(s), 0, 0); + unit_watch_all_pids(UNIT(s)); + + /* If the PID set is empty now, then let's finish this off */ + if (set_isempty(UNIT(s)->pids)) + scope_notify_cgroup_empty_event(UNIT(s)); + else + scope_set_state(s, SCOPE_ABANDONED); + + return 0; +} + +_pure_ static UnitActiveState scope_active_state(Unit *u) { + assert(u); + + return state_translation_table[SCOPE(u)->state]; +} + +_pure_ static const char *scope_sub_state_to_string(Unit *u) { + assert(u); + + return scope_state_to_string(SCOPE(u)->state); +} + +static void scope_enumerate(Manager *m) { + Unit *u; + int r; + + assert(m); + + /* Let's unconditionally add the "init.scope" special unit + * that encapsulates PID 1. Note that PID 1 already is in the + * cgroup for this, we hence just need to allocate the object + * for it and that's it. */ + + u = manager_get_unit(m, SPECIAL_INIT_SCOPE); + if (!u) { + r = unit_new_for_name(m, sizeof(Scope), SPECIAL_INIT_SCOPE, &u); + if (r < 0) { + log_error_errno(r, "Failed to allocate the special " SPECIAL_INIT_SCOPE " unit: %m"); + return; + } + } + + u->transient = true; + u->perpetual = true; + SCOPE(u)->deserialized_state = SCOPE_RUNNING; + + unit_add_to_load_queue(u); + unit_add_to_dbus_queue(u); +} + +static const char* const scope_result_table[_SCOPE_RESULT_MAX] = { + [SCOPE_SUCCESS] = "success", + [SCOPE_FAILURE_RESOURCES] = "resources", + [SCOPE_FAILURE_TIMEOUT] = "timeout", +}; + +DEFINE_STRING_TABLE_LOOKUP(scope_result, ScopeResult); + +const UnitVTable scope_vtable = { + .object_size = sizeof(Scope), + .cgroup_context_offset = offsetof(Scope, cgroup_context), + .kill_context_offset = offsetof(Scope, kill_context), + + .sections = + "Unit\0" + "Scope\0" + "Install\0", + .private_section = "Scope", + + .can_transient = true, + + .init = scope_init, + .load = scope_load, + .done = scope_done, + + .coldplug = scope_coldplug, + + .dump = scope_dump, + + .start = scope_start, + .stop = scope_stop, + + .kill = scope_kill, + + .get_timeout = scope_get_timeout, + + .serialize = scope_serialize, + .deserialize_item = scope_deserialize_item, + + .active_state = scope_active_state, + .sub_state_to_string = scope_sub_state_to_string, + + .check_gc = scope_check_gc, + + .sigchld_event = scope_sigchld_event, + + .reset_failed = scope_reset_failed, + + .notify_cgroup_empty = scope_notify_cgroup_empty_event, + + .bus_vtable = bus_scope_vtable, + .bus_set_property = bus_scope_set_property, + .bus_commit_properties = bus_scope_commit_properties, + + .enumerate = scope_enumerate, +}; diff --git a/src/grp-system/libcore/src/selinux-access.c b/src/grp-system/libcore/src/selinux-access.c new file mode 100644 index 0000000000..6807af86c1 --- /dev/null +++ b/src/grp-system/libcore/src/selinux-access.c @@ -0,0 +1,284 @@ +/*** + This file is part of systemd. + + Copyright 2012 Dan Walsh + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "selinux-access.h" + +#ifdef HAVE_SELINUX + +#include <errno.h> +#include <selinux/avc.h> +#include <selinux/selinux.h> +#include <stdio.h> +#ifdef HAVE_AUDIT +#include <libaudit.h> +#endif + +#include <systemd/sd-bus.h> + +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/selinux-util.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/util.h" + +#include "audit-fd.h" + +static bool initialized = false; + +struct audit_info { + sd_bus_creds *creds; + const char *path; + const char *cmdline; +}; + +/* + Any time an access gets denied this callback will be called + with the audit data. We then need to just copy the audit data into the msgbuf. +*/ +static int audit_callback( + void *auditdata, + security_class_t cls, + char *msgbuf, + size_t msgbufsize) { + + const struct audit_info *audit = auditdata; + uid_t uid = 0, login_uid = 0; + gid_t gid = 0; + char login_uid_buf[DECIMAL_STR_MAX(uid_t) + 1] = "n/a"; + char uid_buf[DECIMAL_STR_MAX(uid_t) + 1] = "n/a"; + char gid_buf[DECIMAL_STR_MAX(gid_t) + 1] = "n/a"; + + if (sd_bus_creds_get_audit_login_uid(audit->creds, &login_uid) >= 0) + xsprintf(login_uid_buf, UID_FMT, login_uid); + if (sd_bus_creds_get_euid(audit->creds, &uid) >= 0) + xsprintf(uid_buf, UID_FMT, uid); + if (sd_bus_creds_get_egid(audit->creds, &gid) >= 0) + xsprintf(gid_buf, GID_FMT, gid); + + snprintf(msgbuf, msgbufsize, + "auid=%s uid=%s gid=%s%s%s%s%s%s%s", + login_uid_buf, uid_buf, gid_buf, + audit->path ? " path=\"" : "", strempty(audit->path), audit->path ? "\"" : "", + audit->cmdline ? " cmdline=\"" : "", strempty(audit->cmdline), audit->cmdline ? "\"" : ""); + + return 0; +} + +static int callback_type_to_priority(int type) { + switch(type) { + + case SELINUX_ERROR: + return LOG_ERR; + + case SELINUX_WARNING: + return LOG_WARNING; + + case SELINUX_INFO: + return LOG_INFO; + + case SELINUX_AVC: + default: + return LOG_NOTICE; + } +} + +/* + libselinux uses this callback when access gets denied or other + events happen. If audit is turned on, messages will be reported + using audit netlink, otherwise they will be logged using the usual + channels. + + Code copied from dbus and modified. +*/ +_printf_(2, 3) static int log_callback(int type, const char *fmt, ...) { + va_list ap; + const char *fmt2; + +#ifdef HAVE_AUDIT + int fd; + + fd = get_audit_fd(); + + if (fd >= 0) { + _cleanup_free_ char *buf = NULL; + int r; + + va_start(ap, fmt); + r = vasprintf(&buf, fmt, ap); + va_end(ap); + + if (r >= 0) { + audit_log_user_avc_message(fd, AUDIT_USER_AVC, buf, NULL, NULL, NULL, 0); + return 0; + } + } +#endif + + fmt2 = strjoina("selinux: ", fmt); + + va_start(ap, fmt); + log_internalv(LOG_AUTH | callback_type_to_priority(type), 0, __FILE__, __LINE__, __FUNCTION__, fmt2, ap); + va_end(ap); + + return 0; +} + +static int access_init(sd_bus_error *error) { + + if (!mac_selinux_use()) + return 0; + + if (initialized) + return 1; + + if (avc_open(NULL, 0) != 0) { + int enforce, saved_errno = errno; + + enforce = security_getenforce(); + log_full_errno(enforce != 0 ? LOG_ERR : LOG_WARNING, saved_errno, "Failed to open the SELinux AVC: %m"); + + /* If enforcement isn't on, then let's suppress this + * error, and just don't do any AVC checks. The + * warning we printed is hence all the admin will + * see. */ + if (enforce == 0) + return 0; + + /* Return an access denied error, if we couldn't load + * the AVC but enforcing mode was on, or we couldn't + * determine whether it is one. */ + return sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "Failed to open the SELinux AVC: %s", strerror(saved_errno)); + } + + selinux_set_callback(SELINUX_CB_AUDIT, (union selinux_callback) audit_callback); + selinux_set_callback(SELINUX_CB_LOG, (union selinux_callback) log_callback); + + initialized = true; + return 1; +} + +/* + This function communicates with the kernel to check whether or not it should + allow the access. + If the machine is in permissive mode it will return ok. Audit messages will + still be generated if the access would be denied in enforcing mode. +*/ +int mac_selinux_generic_access_check( + sd_bus_message *message, + const char *path, + const char *permission, + sd_bus_error *error) { + + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + const char *tclass = NULL, *scon = NULL; + struct audit_info audit_info = {}; + _cleanup_free_ char *cl = NULL; + char *fcon = NULL; + char **cmdline = NULL; + int r = 0; + + assert(message); + assert(permission); + assert(error); + + r = access_init(error); + if (r <= 0) + return r; + + r = sd_bus_query_sender_creds( + message, + SD_BUS_CREDS_PID|SD_BUS_CREDS_EUID|SD_BUS_CREDS_EGID| + SD_BUS_CREDS_CMDLINE|SD_BUS_CREDS_AUDIT_LOGIN_UID| + SD_BUS_CREDS_SELINUX_CONTEXT| + SD_BUS_CREDS_AUGMENT /* get more bits from /proc */, + &creds); + if (r < 0) + goto finish; + + /* The SELinux context is something we really should have + * gotten directly from the message or sender, and not be an + * augmented field. If it was augmented we cannot use it for + * authorization, since this is racy and vulnerable. Let's add + * an extra check, just in case, even though this really + * shouldn't be possible. */ + assert_return((sd_bus_creds_get_augmented_mask(creds) & SD_BUS_CREDS_SELINUX_CONTEXT) == 0, -EPERM); + + r = sd_bus_creds_get_selinux_context(creds, &scon); + if (r < 0) + goto finish; + + if (path) { + /* Get the file context of the unit file */ + + r = getfilecon_raw(path, &fcon); + if (r < 0) { + r = sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "Failed to get file context on %s.", path); + goto finish; + } + + tclass = "service"; + } else { + r = getcon_raw(&fcon); + if (r < 0) { + r = sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "Failed to get current context."); + goto finish; + } + + tclass = "system"; + } + + sd_bus_creds_get_cmdline(creds, &cmdline); + cl = strv_join(cmdline, " "); + + audit_info.creds = creds; + audit_info.path = path; + audit_info.cmdline = cl; + + r = selinux_check_access(scon, fcon, tclass, permission, &audit_info); + if (r < 0) + r = sd_bus_error_setf(error, SD_BUS_ERROR_ACCESS_DENIED, "SELinux policy denies access."); + + log_debug("SELinux access check scon=%s tcon=%s tclass=%s perm=%s path=%s cmdline=%s: %i", scon, fcon, tclass, permission, path, cl, r); + +finish: + freecon(fcon); + + if (r < 0 && security_getenforce() != 1) { + sd_bus_error_free(error); + r = 0; + } + + return r; +} + +#else + +int mac_selinux_generic_access_check( + sd_bus_message *message, + const char *path, + const char *permission, + sd_bus_error *error) { + + return 0; +} + +#endif diff --git a/src/grp-system/libcore/src/selinux-access.h b/src/grp-system/libcore/src/selinux-access.h new file mode 100644 index 0000000000..6cd37bed3d --- /dev/null +++ b/src/grp-system/libcore/src/selinux-access.h @@ -0,0 +1,45 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2012 Dan Walsh + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <systemd/sd-bus.h> + +#include "core/manager.h" +#include "sd-bus/bus-util.h" + +int mac_selinux_generic_access_check(sd_bus_message *message, const char *path, const char *permission, sd_bus_error *error); + +#ifdef HAVE_SELINUX + +#define mac_selinux_access_check(message, permission, error) \ + mac_selinux_generic_access_check((message), NULL, (permission), (error)) + +#define mac_selinux_unit_access_check(unit, message, permission, error) \ + ({ \ + const Unit *_unit = (unit); \ + mac_selinux_generic_access_check((message), _unit->source_path ?: _unit->fragment_path, (permission), (error)); \ + }) + +#else + +#define mac_selinux_access_check(message, permission, error) 0 +#define mac_selinux_unit_access_check(unit, message, permission, error) 0 + +#endif diff --git a/src/grp-system/libcore/src/selinux-setup.c b/src/grp-system/libcore/src/selinux-setup.c new file mode 100644 index 0000000000..d81a8e7fa1 --- /dev/null +++ b/src/grp-system/libcore/src/selinux-setup.c @@ -0,0 +1,121 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdio.h> +#include <unistd.h> + +#ifdef HAVE_SELINUX +#include <selinux/selinux.h> +#endif + +#include "core/selinux-setup.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/selinux-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" + +#ifdef HAVE_SELINUX +_printf_(2,3) +static int null_log(int type, const char *fmt, ...) { + return 0; +} +#endif + +int mac_selinux_setup(bool *loaded_policy) { + +#ifdef HAVE_SELINUX + int enforce = 0; + usec_t before_load, after_load; + char *con; + int r; + union selinux_callback cb; + bool initialized = false; + + assert(loaded_policy); + + /* Turn off all of SELinux' own logging, we want to do that */ + cb.func_log = null_log; + selinux_set_callback(SELINUX_CB_LOG, cb); + + /* Don't load policy in the initrd if we don't appear to have + * it. For the real root, we check below if we've already + * loaded policy, and return gracefully. + */ + if (in_initrd() && access(selinux_path(), F_OK) < 0) + return 0; + + /* Already initialized by somebody else? */ + r = getcon_raw(&con); + if (r == 0) { + initialized = !streq(con, "kernel"); + freecon(con); + } + + /* Make sure we have no fds open while loading the policy and + * transitioning */ + log_close(); + + /* Now load the policy */ + before_load = now(CLOCK_MONOTONIC); + r = selinux_init_load_policy(&enforce); + if (r == 0) { + _cleanup_(mac_selinux_freep) char *label = NULL; + char timespan[FORMAT_TIMESPAN_MAX]; + + mac_selinux_retest(); + + /* Transition to the new context */ + r = mac_selinux_get_create_label_from_exe(SYSTEMD_BINARY_PATH, &label); + if (r < 0 || !label) { + log_open(); + log_error("Failed to compute init label, ignoring."); + } else { + r = setcon_raw(label); + + log_open(); + if (r < 0) + log_error("Failed to transition into init label '%s', ignoring.", label); + } + + after_load = now(CLOCK_MONOTONIC); + + log_info("Successfully loaded SELinux policy in %s.", + format_timespan(timespan, sizeof(timespan), after_load - before_load, 0)); + + *loaded_policy = true; + + } else { + log_open(); + + if (enforce > 0) { + if (!initialized) { + log_emergency("Failed to load SELinux policy."); + return -EIO; + } + + log_warning("Failed to load new SELinux policy. Continuing with old policy."); + } else + log_debug("Unable to load SELinux policy. Ignoring."); + } +#endif + + return 0; +} diff --git a/src/grp-system/libcore/src/service.c b/src/grp-system/libcore/src/service.c new file mode 100644 index 0000000000..0defd0cc35 --- /dev/null +++ b/src/grp-system/libcore/src/service.c @@ -0,0 +1,3470 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <signal.h> +#include <unistd.h> + +#include "core/load-fragment.h" +#include "core/manager.h" +#include "core/service.h" +#include "core/unit.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-kernel.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/async.h" +#include "systemd-basic/def.h" +#include "systemd-basic/env-util.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/utf8.h" +#include "systemd-basic/util.h" + +#include "dbus-service.h" +#include "load-dropin.h" +#include "unit-printf.h" + +static const UnitActiveState state_translation_table[_SERVICE_STATE_MAX] = { + [SERVICE_DEAD] = UNIT_INACTIVE, + [SERVICE_START_PRE] = UNIT_ACTIVATING, + [SERVICE_START] = UNIT_ACTIVATING, + [SERVICE_START_POST] = UNIT_ACTIVATING, + [SERVICE_RUNNING] = UNIT_ACTIVE, + [SERVICE_EXITED] = UNIT_ACTIVE, + [SERVICE_RELOAD] = UNIT_RELOADING, + [SERVICE_STOP] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGABRT] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_STOP_POST] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_FAILED] = UNIT_FAILED, + [SERVICE_AUTO_RESTART] = UNIT_ACTIVATING +}; + +/* For Type=idle we never want to delay any other jobs, hence we + * consider idle jobs active as soon as we start working on them */ +static const UnitActiveState state_translation_table_idle[_SERVICE_STATE_MAX] = { + [SERVICE_DEAD] = UNIT_INACTIVE, + [SERVICE_START_PRE] = UNIT_ACTIVE, + [SERVICE_START] = UNIT_ACTIVE, + [SERVICE_START_POST] = UNIT_ACTIVE, + [SERVICE_RUNNING] = UNIT_ACTIVE, + [SERVICE_EXITED] = UNIT_ACTIVE, + [SERVICE_RELOAD] = UNIT_RELOADING, + [SERVICE_STOP] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGABRT] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_STOP_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_STOP_POST] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SERVICE_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SERVICE_FAILED] = UNIT_FAILED, + [SERVICE_AUTO_RESTART] = UNIT_ACTIVATING +}; + +static int service_dispatch_io(sd_event_source *source, int fd, uint32_t events, void *userdata); +static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); +static int service_dispatch_watchdog(sd_event_source *source, usec_t usec, void *userdata); + +static void service_enter_signal(Service *s, ServiceState state, ServiceResult f); +static void service_enter_reload_by_notify(Service *s); + +static void service_init(Unit *u) { + Service *s = SERVICE(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + s->timeout_start_usec = u->manager->default_timeout_start_usec; + s->timeout_stop_usec = u->manager->default_timeout_stop_usec; + s->restart_usec = u->manager->default_restart_usec; + s->runtime_max_usec = USEC_INFINITY; + s->type = _SERVICE_TYPE_INVALID; + s->socket_fd = -1; + s->stdin_fd = s->stdout_fd = s->stderr_fd = -1; + s->guess_main_pid = true; + + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; +} + +static void service_unwatch_control_pid(Service *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +static void service_unwatch_main_pid(Service *s) { + assert(s); + + if (s->main_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->main_pid); + s->main_pid = 0; +} + +static void service_unwatch_pid_file(Service *s) { + if (!s->pid_file_pathspec) + return; + + log_unit_debug(UNIT(s), "Stopping watch for PID file %s", s->pid_file_pathspec->path); + path_spec_unwatch(s->pid_file_pathspec); + path_spec_done(s->pid_file_pathspec); + s->pid_file_pathspec = mfree(s->pid_file_pathspec); +} + +static int service_set_main_pid(Service *s, pid_t pid) { + pid_t ppid; + + assert(s); + + if (pid <= 1) + return -EINVAL; + + if (pid == getpid()) + return -EINVAL; + + if (s->main_pid == pid && s->main_pid_known) + return 0; + + if (s->main_pid != pid) { + service_unwatch_main_pid(s); + exec_status_start(&s->main_exec_status, pid); + } + + s->main_pid = pid; + s->main_pid_known = true; + + if (get_process_ppid(pid, &ppid) >= 0 && ppid != getpid()) { + log_unit_warning(UNIT(s), "Supervising process "PID_FMT" which is not our child. We'll most likely not notice when it exits.", pid); + s->main_pid_alien = true; + } else + s->main_pid_alien = false; + + return 0; +} + +void service_close_socket_fd(Service *s) { + assert(s); + + /* Undo the effect of service_set_socket_fd(). */ + + s->socket_fd = asynchronous_close(s->socket_fd); + + if (UNIT_ISSET(s->accept_socket)) { + socket_connection_unref(SOCKET(UNIT_DEREF(s->accept_socket))); + unit_ref_unset(&s->accept_socket); + } +} + +static void service_stop_watchdog(Service *s) { + assert(s); + + s->watchdog_event_source = sd_event_source_unref(s->watchdog_event_source); + s->watchdog_timestamp = DUAL_TIMESTAMP_NULL; +} + +static usec_t service_get_watchdog_usec(Service *s) { + assert(s); + + if (s->watchdog_override_enable) + return s->watchdog_override_usec; + else + return s->watchdog_usec; +} + +static void service_start_watchdog(Service *s) { + int r; + usec_t watchdog_usec; + + assert(s); + + watchdog_usec = service_get_watchdog_usec(s); + if (watchdog_usec == 0 || watchdog_usec == USEC_INFINITY) + return; + + if (s->watchdog_event_source) { + r = sd_event_source_set_time(s->watchdog_event_source, usec_add(s->watchdog_timestamp.monotonic, watchdog_usec)); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to reset watchdog timer: %m"); + return; + } + + r = sd_event_source_set_enabled(s->watchdog_event_source, SD_EVENT_ONESHOT); + } else { + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->watchdog_event_source, + CLOCK_MONOTONIC, + usec_add(s->watchdog_timestamp.monotonic, watchdog_usec), 0, + service_dispatch_watchdog, s); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to add watchdog timer: %m"); + return; + } + + (void) sd_event_source_set_description(s->watchdog_event_source, "service-watchdog"); + + /* Let's process everything else which might be a sign + * of living before we consider a service died. */ + r = sd_event_source_set_priority(s->watchdog_event_source, SD_EVENT_PRIORITY_IDLE); + } + + if (r < 0) + log_unit_warning_errno(UNIT(s), r, "Failed to install watchdog timer: %m"); +} + +static void service_reset_watchdog(Service *s) { + assert(s); + + dual_timestamp_get(&s->watchdog_timestamp); + service_start_watchdog(s); +} + +static void service_reset_watchdog_timeout(Service *s, usec_t watchdog_override_usec) { + assert(s); + + s->watchdog_override_enable = true; + s->watchdog_override_usec = watchdog_override_usec; + service_reset_watchdog(s); + + log_unit_debug(UNIT(s), "watchdog_usec="USEC_FMT, s->watchdog_usec); + log_unit_debug(UNIT(s), "watchdog_override_usec="USEC_FMT, s->watchdog_override_usec); +} + +static void service_fd_store_unlink(ServiceFDStore *fs) { + + if (!fs) + return; + + if (fs->service) { + assert(fs->service->n_fd_store > 0); + LIST_REMOVE(fd_store, fs->service->fd_store, fs); + fs->service->n_fd_store--; + } + + if (fs->event_source) { + sd_event_source_set_enabled(fs->event_source, SD_EVENT_OFF); + sd_event_source_unref(fs->event_source); + } + + free(fs->fdname); + safe_close(fs->fd); + free(fs); +} + +static void service_release_fd_store(Service *s) { + assert(s); + + log_unit_debug(UNIT(s), "Releasing all stored fds"); + while (s->fd_store) + service_fd_store_unlink(s->fd_store); + + assert(s->n_fd_store == 0); +} + +static void service_release_resources(Unit *u, bool inactive) { + Service *s = SERVICE(u); + + assert(s); + + if (!s->fd_store && s->stdin_fd < 0 && s->stdout_fd < 0 && s->stderr_fd < 0) + return; + + log_unit_debug(u, "Releasing resources."); + + s->stdin_fd = safe_close(s->stdin_fd); + s->stdout_fd = safe_close(s->stdout_fd); + s->stderr_fd = safe_close(s->stderr_fd); + + if (inactive) + service_release_fd_store(s); +} + +static void service_done(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + s->pid_file = mfree(s->pid_file); + s->status_text = mfree(s->status_text); + + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + exec_command_free_array(s->exec_command, _SERVICE_EXEC_COMMAND_MAX); + s->control_command = NULL; + s->main_command = NULL; + + dynamic_creds_unref(&s->dynamic_creds); + + exit_status_set_free(&s->restart_prevent_status); + exit_status_set_free(&s->restart_force_status); + exit_status_set_free(&s->success_status); + + /* This will leak a process, but at least no memory or any of + * our resources */ + service_unwatch_main_pid(s); + service_unwatch_control_pid(s); + service_unwatch_pid_file(s); + + if (s->bus_name) { + unit_unwatch_bus_name(u, s->bus_name); + s->bus_name = mfree(s->bus_name); + } + + s->bus_name_owner = mfree(s->bus_name_owner); + + service_close_socket_fd(s); + s->peer = socket_peer_unref(s->peer); + + unit_ref_unset(&s->accept_socket); + + service_stop_watchdog(s); + + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + + service_release_resources(u, true); +} + +static int on_fd_store_io(sd_event_source *e, int fd, uint32_t revents, void *userdata) { + ServiceFDStore *fs = userdata; + + assert(e); + assert(fs); + + /* If we get either EPOLLHUP or EPOLLERR, it's time to remove this entry from the fd store */ + log_unit_debug(UNIT(fs->service), + "Received %s on stored fd %d (%s), closing.", + revents & EPOLLERR ? "EPOLLERR" : "EPOLLHUP", + fs->fd, strna(fs->fdname)); + service_fd_store_unlink(fs); + return 0; +} + +static int service_add_fd_store(Service *s, int fd, const char *name) { + ServiceFDStore *fs; + int r; + + /* fd is always consumed if we return >= 0 */ + + assert(s); + assert(fd >= 0); + + if (s->n_fd_store >= s->n_fd_store_max) + return -EXFULL; /* Our store is full. + * Use this errno rather than E[NM]FILE to distinguish from + * the case where systemd itself hits the file limit. */ + + LIST_FOREACH(fd_store, fs, s->fd_store) { + r = same_fd(fs->fd, fd); + if (r < 0) + return r; + if (r > 0) { + safe_close(fd); + return 0; /* fd already included */ + } + } + + fs = new0(ServiceFDStore, 1); + if (!fs) + return -ENOMEM; + + fs->fd = fd; + fs->service = s; + fs->fdname = strdup(name ?: "stored"); + if (!fs->fdname) { + free(fs); + return -ENOMEM; + } + + r = sd_event_add_io(UNIT(s)->manager->event, &fs->event_source, fd, 0, on_fd_store_io, fs); + if (r < 0) { + free(fs->fdname); + free(fs); + return r; + } + + (void) sd_event_source_set_description(fs->event_source, "service-fd-store"); + + LIST_PREPEND(fd_store, s->fd_store, fs); + s->n_fd_store++; + + return 1; /* fd newly stored */ +} + +static int service_add_fd_store_set(Service *s, FDSet *fds, const char *name) { + int r; + + assert(s); + + while (fdset_size(fds) > 0) { + _cleanup_close_ int fd = -1; + + fd = fdset_steal_first(fds); + if (fd < 0) + break; + + r = service_add_fd_store(s, fd, name); + if (r == -EXFULL) + return log_unit_warning_errno(UNIT(s), r, + "Cannot store more fds than FileDescriptorStoreMax=%u, closing remaining.", + s->n_fd_store_max); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Failed to add fd to store: %m"); + if (r > 0) + log_unit_debug(UNIT(s), "Added fd %u (%s) to fd store.", fd, strna(name)); + fd = -1; + } + + return 0; +} + +static int service_arm_timer(Service *s, usec_t usec) { + int r; + + assert(s); + + if (s->timer_event_source) { + r = sd_event_source_set_time(s->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(s->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + service_dispatch_timer, s); + if (r < 0) + return r; + + (void) sd_event_source_set_description(s->timer_event_source, "service-timer"); + + return 0; +} + +static int service_verify(Service *s) { + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (!s->exec_command[SERVICE_EXEC_START] && !s->exec_command[SERVICE_EXEC_STOP]) { + log_unit_error(UNIT(s), "Service lacks both ExecStart= and ExecStop= setting. Refusing."); + return -EINVAL; + } + + if (s->type != SERVICE_ONESHOT && !s->exec_command[SERVICE_EXEC_START]) { + log_unit_error(UNIT(s), "Service has no ExecStart= setting, which is only allowed for Type=oneshot services. Refusing."); + return -EINVAL; + } + + if (!s->remain_after_exit && !s->exec_command[SERVICE_EXEC_START]) { + log_unit_error(UNIT(s), "Service has no ExecStart= setting, which is only allowed for RemainAfterExit=yes services. Refusing."); + return -EINVAL; + } + + if (s->type != SERVICE_ONESHOT && s->exec_command[SERVICE_EXEC_START]->command_next) { + log_unit_error(UNIT(s), "Service has more than one ExecStart= setting, which is only allowed for Type=oneshot services. Refusing."); + return -EINVAL; + } + + if (s->type == SERVICE_ONESHOT && s->restart != SERVICE_RESTART_NO) { + log_unit_error(UNIT(s), "Service has Restart= setting other than no, which isn't allowed for Type=oneshot services. Refusing."); + return -EINVAL; + } + + if (s->type == SERVICE_ONESHOT && !exit_status_set_is_empty(&s->restart_force_status)) { + log_unit_error(UNIT(s), "Service has RestartForceStatus= set, which isn't allowed for Type=oneshot services. Refusing."); + return -EINVAL; + } + + if (s->type == SERVICE_DBUS && !s->bus_name) { + log_unit_error(UNIT(s), "Service is of type D-Bus but no D-Bus service name has been specified. Refusing."); + return -EINVAL; + } + + if (s->bus_name && s->type != SERVICE_DBUS) + log_unit_warning(UNIT(s), "Service has a D-Bus service name specified, but is not of type dbus. Ignoring."); + + if (s->exec_context.pam_name && !(s->kill_context.kill_mode == KILL_CONTROL_GROUP || s->kill_context.kill_mode == KILL_MIXED)) { + log_unit_error(UNIT(s), "Service has PAM enabled. Kill mode must be set to 'control-group' or 'mixed'. Refusing."); + return -EINVAL; + } + + if (s->usb_function_descriptors && !s->usb_function_strings) + log_unit_warning(UNIT(s), "Service has USBFunctionDescriptors= setting, but no USBFunctionStrings=. Ignoring."); + + if (!s->usb_function_descriptors && s->usb_function_strings) + log_unit_warning(UNIT(s), "Service has USBFunctionStrings= setting, but no USBFunctionDescriptors=. Ignoring."); + + if (s->runtime_max_usec != USEC_INFINITY && s->type == SERVICE_ONESHOT) + log_unit_warning(UNIT(s), "MaxRuntimeSec= has no effect in combination with Type=oneshot. Ignoring."); + + return 0; +} + +static int service_add_default_dependencies(Service *s) { + int r; + + assert(s); + + if (!UNIT(s)->default_dependencies) + return 0; + + /* Add a number of automatic dependencies useful for the + * majority of services. */ + + if (MANAGER_IS_SYSTEM(UNIT(s)->manager)) { + /* First, pull in the really early boot stuff, and + * require it, so that we fail if we can't acquire + * it. */ + + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } else { + + /* In the --user instance there's no sysinit.target, + * in that case require basic.target instead. */ + + r = unit_add_dependency_by_name(UNIT(s), UNIT_REQUIRES, SPECIAL_BASIC_TARGET, NULL, true); + if (r < 0) + return r; + } + + /* Second, if the rest of the base system is in the same + * transaction, order us after it, but do not pull it in or + * even require it. */ + r = unit_add_dependency_by_name(UNIT(s), UNIT_AFTER, SPECIAL_BASIC_TARGET, NULL, true); + if (r < 0) + return r; + + /* Third, add us in for normal shutdown. */ + return unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static void service_fix_output(Service *s) { + assert(s); + + /* If nothing has been explicitly configured, patch default + * output in. If input is socket/tty we avoid this however, + * since in that case we want output to default to the same + * place as we read input from. */ + + if (s->exec_context.std_error == EXEC_OUTPUT_INHERIT && + s->exec_context.std_output == EXEC_OUTPUT_INHERIT && + s->exec_context.std_input == EXEC_INPUT_NULL) + s->exec_context.std_error = UNIT(s)->manager->default_std_error; + + if (s->exec_context.std_output == EXEC_OUTPUT_INHERIT && + s->exec_context.std_input == EXEC_INPUT_NULL) + s->exec_context.std_output = UNIT(s)->manager->default_std_output; +} + +static int service_setup_bus_name(Service *s) { + int r; + + assert(s); + + if (!s->bus_name) + return 0; + + r = unit_add_dependency_by_name(UNIT(s), UNIT_REQUIRES, SPECIAL_DBUS_SOCKET, NULL, true); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Failed to add dependency on " SPECIAL_DBUS_SOCKET ": %m"); + + /* Regardless if kdbus is used or not, we always want to be ordered against dbus.socket if both are in the transaction. */ + r = unit_add_dependency_by_name(UNIT(s), UNIT_AFTER, SPECIAL_DBUS_SOCKET, NULL, true); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Failed to add dependency on " SPECIAL_DBUS_SOCKET ": %m"); + + r = unit_watch_bus_name(UNIT(s), s->bus_name); + if (r == -EEXIST) + return log_unit_error_errno(UNIT(s), r, "Two services allocated for the same bus name %s, refusing operation.", s->bus_name); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Cannot watch bus name %s: %m", s->bus_name); + + return 0; +} + +static int service_add_extras(Service *s) { + int r; + + assert(s); + + if (s->type == _SERVICE_TYPE_INVALID) { + /* Figure out a type automatically */ + if (s->bus_name) + s->type = SERVICE_DBUS; + else if (s->exec_command[SERVICE_EXEC_START]) + s->type = SERVICE_SIMPLE; + else + s->type = SERVICE_ONESHOT; + } + + /* Oneshot services have disabled start timeout by default */ + if (s->type == SERVICE_ONESHOT && !s->start_timeout_defined) + s->timeout_start_usec = USEC_INFINITY; + + service_fix_output(s); + + r = unit_patch_contexts(UNIT(s)); + if (r < 0) + return r; + + r = unit_add_exec_dependencies(UNIT(s), &s->exec_context); + if (r < 0) + return r; + + r = unit_set_default_slice(UNIT(s)); + if (r < 0) + return r; + + if (s->type == SERVICE_NOTIFY && s->notify_access == NOTIFY_NONE) + s->notify_access = NOTIFY_MAIN; + + if (s->watchdog_usec > 0 && s->notify_access == NOTIFY_NONE) + s->notify_access = NOTIFY_MAIN; + + r = service_add_default_dependencies(s); + if (r < 0) + return r; + + r = service_setup_bus_name(s); + if (r < 0) + return r; + + return 0; +} + +static int service_load(Unit *u) { + Service *s = SERVICE(u); + int r; + + assert(s); + + /* Load a .service file */ + r = unit_load_fragment(u); + if (r < 0) + return r; + + /* Still nothing found? Then let's give up */ + if (u->load_state == UNIT_STUB) + return -ENOENT; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + + /* We were able to load something, then let's add in + * the dropin directories. */ + r = unit_load_dropin(u); + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some + * extras */ + r = service_add_extras(s); + if (r < 0) + return r; + } + + return service_verify(s); +} + +static void service_dump(Unit *u, FILE *f, const char *prefix) { + ServiceExecCommand c; + Service *s = SERVICE(u); + const char *prefix2; + + assert(s); + + prefix = strempty(prefix); + prefix2 = strjoina(prefix, "\t"); + + fprintf(f, + "%sService State: %s\n" + "%sResult: %s\n" + "%sReload Result: %s\n" + "%sPermissionsStartOnly: %s\n" + "%sRootDirectoryStartOnly: %s\n" + "%sRemainAfterExit: %s\n" + "%sGuessMainPID: %s\n" + "%sType: %s\n" + "%sRestart: %s\n" + "%sNotifyAccess: %s\n" + "%sNotifyState: %s\n", + prefix, service_state_to_string(s->state), + prefix, service_result_to_string(s->result), + prefix, service_result_to_string(s->reload_result), + prefix, yes_no(s->permissions_start_only), + prefix, yes_no(s->root_directory_start_only), + prefix, yes_no(s->remain_after_exit), + prefix, yes_no(s->guess_main_pid), + prefix, service_type_to_string(s->type), + prefix, service_restart_to_string(s->restart), + prefix, notify_access_to_string(s->notify_access), + prefix, notify_state_to_string(s->notify_state)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: "PID_FMT"\n", + prefix, s->control_pid); + + if (s->main_pid > 0) + fprintf(f, + "%sMain PID: "PID_FMT"\n" + "%sMain PID Known: %s\n" + "%sMain PID Alien: %s\n", + prefix, s->main_pid, + prefix, yes_no(s->main_pid_known), + prefix, yes_no(s->main_pid_alien)); + + if (s->pid_file) + fprintf(f, + "%sPIDFile: %s\n", + prefix, s->pid_file); + + if (s->bus_name) + fprintf(f, + "%sBusName: %s\n" + "%sBus Name Good: %s\n", + prefix, s->bus_name, + prefix, yes_no(s->bus_name_good)); + + if (UNIT_ISSET(s->accept_socket)) + fprintf(f, + "%sAccept Socket: %s\n", + prefix, UNIT_DEREF(s->accept_socket)->id); + + kill_context_dump(&s->kill_context, f, prefix); + exec_context_dump(&s->exec_context, f, prefix); + + for (c = 0; c < _SERVICE_EXEC_COMMAND_MAX; c++) { + + if (!s->exec_command[c]) + continue; + + fprintf(f, "%s-> %s:\n", + prefix, service_exec_command_to_string(c)); + + exec_command_dump_list(s->exec_command[c], f, prefix2); + } + + if (s->status_text) + fprintf(f, "%sStatus Text: %s\n", + prefix, s->status_text); + + if (s->n_fd_store_max > 0) + fprintf(f, + "%sFile Descriptor Store Max: %u\n" + "%sFile Descriptor Store Current: %u\n", + prefix, s->n_fd_store_max, + prefix, s->n_fd_store); +} + +static int service_load_pid_file(Service *s, bool may_warn) { + _cleanup_free_ char *k = NULL; + int r; + pid_t pid; + + assert(s); + + if (!s->pid_file) + return -ENOENT; + + r = read_one_line_file(s->pid_file, &k); + if (r < 0) { + if (may_warn) + log_unit_info_errno(UNIT(s), r, "PID file %s not readable (yet?) after %s: %m", s->pid_file, service_state_to_string(s->state)); + return r; + } + + r = parse_pid(k, &pid); + if (r < 0) { + if (may_warn) + log_unit_info_errno(UNIT(s), r, "Failed to read PID from file %s: %m", s->pid_file); + return r; + } + + if (!pid_is_alive(pid)) { + if (may_warn) + log_unit_info(UNIT(s), "PID "PID_FMT" read from file %s does not exist or is a zombie.", pid, s->pid_file); + return -ESRCH; + } + + if (s->main_pid_known) { + if (pid == s->main_pid) + return 0; + + log_unit_debug(UNIT(s), "Main PID changing: "PID_FMT" -> "PID_FMT, s->main_pid, pid); + + service_unwatch_main_pid(s); + s->main_pid_known = false; + } else + log_unit_debug(UNIT(s), "Main PID loaded: "PID_FMT, pid); + + r = service_set_main_pid(s, pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) { + /* FIXME: we need to do something here */ + log_unit_warning_errno(UNIT(s), r, "Failed to watch PID "PID_FMT" for service: %m", pid); + return r; + } + + return 0; +} + +static void service_search_main_pid(Service *s) { + pid_t pid = 0; + int r; + + assert(s); + + /* If we know it anyway, don't ever fallback to unreliable + * heuristics */ + if (s->main_pid_known) + return; + + if (!s->guess_main_pid) + return; + + assert(s->main_pid <= 0); + + if (unit_search_main_pid(UNIT(s), &pid) < 0) + return; + + log_unit_debug(UNIT(s), "Main PID guessed: "PID_FMT, pid); + if (service_set_main_pid(s, pid) < 0) + return; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + log_unit_warning_errno(UNIT(s), r, "Failed to watch PID "PID_FMT" from: %m", pid); +} + +static void service_set_state(Service *s, ServiceState state) { + ServiceState old_state; + const UnitActiveState *table; + + assert(s); + + table = s->type == SERVICE_IDLE ? state_translation_table_idle : state_translation_table; + + old_state = s->state; + s->state = state; + + service_unwatch_pid_file(s); + + if (!IN_SET(state, + SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, + SERVICE_RUNNING, + SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL, + SERVICE_AUTO_RESTART)) + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + + if (!IN_SET(state, + SERVICE_START, SERVICE_START_POST, + SERVICE_RUNNING, SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) { + service_unwatch_main_pid(s); + s->main_command = NULL; + } + + if (!IN_SET(state, + SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, + SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) { + service_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + } + + if (IN_SET(state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) + unit_unwatch_all_pids(UNIT(s)); + + if (!IN_SET(state, + SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, + SERVICE_RUNNING, SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL) && + !(state == SERVICE_DEAD && UNIT(s)->job)) + service_close_socket_fd(s); + + if (!IN_SET(state, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) + service_stop_watchdog(s); + + /* For the inactive states unit_notify() will trim the cgroup, + * but for exit we have to do that ourselves... */ + if (state == SERVICE_EXITED && !MANAGER_IS_RELOADING(UNIT(s)->manager)) + unit_prune_cgroup(UNIT(s)); + + /* For remain_after_exit services, let's see if we can "release" the + * hold on the console, since unit_notify() only does that in case of + * change of state */ + if (state == SERVICE_EXITED && + s->remain_after_exit && + UNIT(s)->manager->n_on_console > 0) { + + ExecContext *ec; + + ec = unit_get_exec_context(UNIT(s)); + if (ec && exec_context_may_touch_console(ec)) { + Manager *m = UNIT(s)->manager; + + m->n_on_console--; + if (m->n_on_console == 0) + /* unset no_console_output flag, since the console is free */ + m->no_console_output = false; + } + } + + if (old_state != state) + log_unit_debug(UNIT(s), "Changed %s -> %s", service_state_to_string(old_state), service_state_to_string(state)); + + unit_notify(UNIT(s), table[old_state], table[state], s->reload_result == SERVICE_SUCCESS); +} + +static usec_t service_coldplug_timeout(Service *s) { + assert(s); + + switch (s->deserialized_state) { + + case SERVICE_START_PRE: + case SERVICE_START: + case SERVICE_START_POST: + case SERVICE_RELOAD: + return usec_add(UNIT(s)->state_change_timestamp.monotonic, s->timeout_start_usec); + + case SERVICE_RUNNING: + return usec_add(UNIT(s)->active_enter_timestamp.monotonic, s->runtime_max_usec); + + case SERVICE_STOP: + case SERVICE_STOP_SIGABRT: + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + return usec_add(UNIT(s)->state_change_timestamp.monotonic, s->timeout_stop_usec); + + case SERVICE_AUTO_RESTART: + return usec_add(UNIT(s)->inactive_enter_timestamp.monotonic, s->restart_usec); + + default: + return USEC_INFINITY; + } +} + +static int service_coldplug(Unit *u) { + Service *s = SERVICE(u); + int r; + + assert(s); + assert(s->state == SERVICE_DEAD); + + if (s->deserialized_state == s->state) + return 0; + + r = service_arm_timer(s, service_coldplug_timeout(s)); + if (r < 0) + return r; + + if (s->main_pid > 0 && + pid_is_unwaited(s->main_pid) && + ((s->deserialized_state == SERVICE_START && IN_SET(s->type, SERVICE_FORKING, SERVICE_DBUS, SERVICE_ONESHOT, SERVICE_NOTIFY)) || + IN_SET(s->deserialized_state, + SERVICE_START, SERVICE_START_POST, + SERVICE_RUNNING, SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL))) { + r = unit_watch_pid(UNIT(s), s->main_pid); + if (r < 0) + return r; + } + + if (s->control_pid > 0 && + pid_is_unwaited(s->control_pid) && + IN_SET(s->deserialized_state, + SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, + SERVICE_RELOAD, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) { + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + } + + if (!IN_SET(s->deserialized_state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) + unit_watch_all_pids(UNIT(s)); + + if (IN_SET(s->deserialized_state, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) + service_start_watchdog(s); + + if (!IN_SET(s->deserialized_state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) + (void) unit_setup_dynamic_creds(u); + + if (UNIT_ISSET(s->accept_socket)) { + Socket* socket = SOCKET(UNIT_DEREF(s->accept_socket)); + + if (socket->max_connections_per_source > 0) { + SocketPeer *peer; + + /* Make a best-effort attempt at bumping the connection count */ + if (socket_acquire_peer(socket, s->socket_fd, &peer) > 0) { + socket_peer_unref(s->peer); + s->peer = peer; + } + } + } + + service_set_state(s, s->deserialized_state); + return 0; +} + +static int service_collect_fds(Service *s, int **fds, char ***fd_names) { + _cleanup_strv_free_ char **rfd_names = NULL; + _cleanup_free_ int *rfds = NULL; + int rn_fds = 0, r; + + assert(s); + assert(fds); + assert(fd_names); + + if (s->socket_fd >= 0) { + + /* Pass the per-connection socket */ + + rfds = new(int, 1); + if (!rfds) + return -ENOMEM; + rfds[0] = s->socket_fd; + + rfd_names = strv_new("connection", NULL); + if (!rfd_names) + return -ENOMEM; + + rn_fds = 1; + } else { + Iterator i; + Unit *u; + + /* Pass all our configured sockets for singleton services */ + + SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERED_BY], i) { + _cleanup_free_ int *cfds = NULL; + Socket *sock; + int cn_fds; + + if (u->type != UNIT_SOCKET) + continue; + + sock = SOCKET(u); + + cn_fds = socket_collect_fds(sock, &cfds); + if (cn_fds < 0) + return cn_fds; + + if (cn_fds <= 0) + continue; + + if (!rfds) { + rfds = cfds; + rn_fds = cn_fds; + + cfds = NULL; + } else { + int *t; + + t = realloc(rfds, (rn_fds + cn_fds) * sizeof(int)); + if (!t) + return -ENOMEM; + + memcpy(t + rn_fds, cfds, cn_fds * sizeof(int)); + + rfds = t; + rn_fds += cn_fds; + } + + r = strv_extend_n(&rfd_names, socket_fdname(sock), cn_fds); + if (r < 0) + return r; + } + } + + if (s->n_fd_store > 0) { + ServiceFDStore *fs; + char **nl; + int *t; + + t = realloc(rfds, (rn_fds + s->n_fd_store) * sizeof(int)); + if (!t) + return -ENOMEM; + + rfds = t; + + nl = realloc(rfd_names, (rn_fds + s->n_fd_store + 1) * sizeof(char*)); + if (!nl) + return -ENOMEM; + + rfd_names = nl; + + LIST_FOREACH(fd_store, fs, s->fd_store) { + rfds[rn_fds] = fs->fd; + rfd_names[rn_fds] = strdup(strempty(fs->fdname)); + if (!rfd_names[rn_fds]) + return -ENOMEM; + + rn_fds++; + } + + rfd_names[rn_fds] = NULL; + } + + *fds = rfds; + *fd_names = rfd_names; + + rfds = NULL; + rfd_names = NULL; + + return rn_fds; +} + +static int service_spawn( + Service *s, + ExecCommand *c, + usec_t timeout, + ExecFlags flags, + pid_t *_pid) { + + _cleanup_strv_free_ char **argv = NULL, **final_env = NULL, **our_env = NULL, **fd_names = NULL; + _cleanup_free_ int *fds = NULL; + unsigned n_fds = 0, n_env = 0; + const char *path; + pid_t pid; + + ExecParameters exec_params = { + .flags = flags, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, + }; + + int r; + + assert(s); + assert(c); + assert(_pid); + + if (flags & EXEC_IS_CONTROL) { + /* If this is a control process, mask the permissions/chroot application if this is requested. */ + if (s->permissions_start_only) + exec_params.flags &= ~EXEC_APPLY_PERMISSIONS; + if (s->root_directory_start_only) + exec_params.flags &= ~EXEC_APPLY_CHROOT; + } + + (void) unit_realize_cgroup(UNIT(s)); + if (s->reset_cpu_usage) { + (void) unit_reset_cpu_usage(UNIT(s)); + s->reset_cpu_usage = false; + } + + r = unit_setup_exec_runtime(UNIT(s)); + if (r < 0) + return r; + + r = unit_setup_dynamic_creds(UNIT(s)); + if (r < 0) + return r; + + if ((flags & EXEC_PASS_FDS) || + s->exec_context.std_input == EXEC_INPUT_SOCKET || + s->exec_context.std_output == EXEC_OUTPUT_SOCKET || + s->exec_context.std_error == EXEC_OUTPUT_SOCKET) { + + r = service_collect_fds(s, &fds, &fd_names); + if (r < 0) + return r; + + n_fds = r; + log_unit_debug(UNIT(s), "Passing %i fds to service", n_fds); + } + + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), timeout)); + if (r < 0) + return r; + + r = unit_full_printf_strv(UNIT(s), c->argv, &argv); + if (r < 0) + return r; + + our_env = new0(char*, 9); + if (!our_env) + return -ENOMEM; + + if ((flags & EXEC_IS_CONTROL) ? s->notify_access == NOTIFY_ALL : s->notify_access != NOTIFY_NONE) + if (asprintf(our_env + n_env++, "NOTIFY_SOCKET=%s", UNIT(s)->manager->notify_socket) < 0) + return -ENOMEM; + + if (s->main_pid > 0) + if (asprintf(our_env + n_env++, "MAINPID="PID_FMT, s->main_pid) < 0) + return -ENOMEM; + + if (MANAGER_IS_USER(UNIT(s)->manager)) + if (asprintf(our_env + n_env++, "MANAGERPID="PID_FMT, getpid()) < 0) + return -ENOMEM; + + if (s->socket_fd >= 0) { + union sockaddr_union sa; + socklen_t salen = sizeof(sa); + + r = getpeername(s->socket_fd, &sa.sa, &salen); + if (r < 0) { + r = -errno; + + /* ENOTCONN is legitimate if the endpoint disappeared on shutdown. + * This connection is over, but the socket unit lives on. */ + if (r != -ENOTCONN || !IN_SET(s->control_command_id, SERVICE_EXEC_STOP, SERVICE_EXEC_STOP_POST)) + return r; + } + + if (r == 0 && IN_SET(sa.sa.sa_family, AF_INET, AF_INET6)) { + _cleanup_free_ char *addr = NULL; + char *t; + int port; + + r = sockaddr_pretty(&sa.sa, salen, true, false, &addr); + if (r < 0) + return r; + + t = strappend("REMOTE_ADDR=", addr); + if (!t) + return -ENOMEM; + our_env[n_env++] = t; + + port = sockaddr_port(&sa.sa); + if (port < 0) + return port; + + if (asprintf(&t, "REMOTE_PORT=%u", port) < 0) + return -ENOMEM; + our_env[n_env++] = t; + } + } + + if (flags & EXEC_SETENV_RESULT) { + if (asprintf(our_env + n_env++, "SERVICE_RESULT=%s", service_result_to_string(s->result)) < 0) + return -ENOMEM; + + if (s->main_exec_status.pid > 0 && + dual_timestamp_is_set(&s->main_exec_status.exit_timestamp)) { + if (asprintf(our_env + n_env++, "EXIT_CODE=%s", sigchld_code_to_string(s->main_exec_status.code)) < 0) + return -ENOMEM; + + if (s->main_exec_status.code == CLD_EXITED) + r = asprintf(our_env + n_env++, "EXIT_STATUS=%i", s->main_exec_status.status); + else + r = asprintf(our_env + n_env++, "EXIT_STATUS=%s", signal_to_string(s->main_exec_status.status)); + if (r < 0) + return -ENOMEM; + } + } + + final_env = strv_env_merge(2, UNIT(s)->manager->environment, our_env, NULL); + if (!final_env) + return -ENOMEM; + + if ((flags & EXEC_IS_CONTROL) && UNIT(s)->cgroup_path) { + path = strjoina(UNIT(s)->cgroup_path, "/control"); + (void) cg_create(SYSTEMD_CGROUP_CONTROLLER, path); + } else + path = UNIT(s)->cgroup_path; + + exec_params.argv = argv; + exec_params.environment = final_env; + exec_params.fds = fds; + exec_params.fd_names = fd_names; + exec_params.n_fds = n_fds; + exec_params.flags |= UNIT(s)->manager->confirm_spawn ? EXEC_CONFIRM_SPAWN : 0; + exec_params.cgroup_supported = UNIT(s)->manager->cgroup_supported; + exec_params.cgroup_path = path; + exec_params.cgroup_delegate = s->cgroup_context.delegate; + exec_params.runtime_prefix = manager_get_runtime_prefix(UNIT(s)->manager); + exec_params.watchdog_usec = s->watchdog_usec; + exec_params.selinux_context_net = s->socket_fd_selinux_context_net; + if (s->type == SERVICE_IDLE) + exec_params.idle_pipe = UNIT(s)->manager->idle_pipe; + exec_params.stdin_fd = s->stdin_fd; + exec_params.stdout_fd = s->stdout_fd; + exec_params.stderr_fd = s->stderr_fd; + + r = exec_spawn(UNIT(s), + c, + &s->exec_context, + &exec_params, + s->exec_runtime, + &s->dynamic_creds, + &pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + return r; + + *_pid = pid; + + return 0; +} + +static int main_pid_good(Service *s) { + assert(s); + + /* Returns 0 if the pid is dead, 1 if it is good, -1 if we + * don't know */ + + /* If we know the pid file, then let's just check if it is + * still valid */ + if (s->main_pid_known) { + + /* If it's an alien child let's check if it is still + * alive ... */ + if (s->main_pid_alien && s->main_pid > 0) + return pid_is_alive(s->main_pid); + + /* .. otherwise assume we'll get a SIGCHLD for it, + * which we really should wait for to collect exit + * status and code */ + return s->main_pid > 0; + } + + /* We don't know the pid */ + return -EAGAIN; +} + +_pure_ static int control_pid_good(Service *s) { + assert(s); + + return s->control_pid > 0; +} + +static int cgroup_good(Service *s) { + int r; + + assert(s); + + if (!UNIT(s)->cgroup_path) + return 0; + + r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, UNIT(s)->cgroup_path); + if (r < 0) + return r; + + return !r; +} + +static bool service_shall_restart(Service *s) { + assert(s); + + /* Don't restart after manual stops */ + if (s->forbid_restart) + return false; + + /* Never restart if this is configured as special exception */ + if (exit_status_set_test(&s->restart_prevent_status, s->main_exec_status.code, s->main_exec_status.status)) + return false; + + /* Restart if the exit code/status are configured as restart triggers */ + if (exit_status_set_test(&s->restart_force_status, s->main_exec_status.code, s->main_exec_status.status)) + return true; + + switch (s->restart) { + + case SERVICE_RESTART_NO: + return false; + + case SERVICE_RESTART_ALWAYS: + return true; + + case SERVICE_RESTART_ON_SUCCESS: + return s->result == SERVICE_SUCCESS; + + case SERVICE_RESTART_ON_FAILURE: + return s->result != SERVICE_SUCCESS; + + case SERVICE_RESTART_ON_ABNORMAL: + return !IN_SET(s->result, SERVICE_SUCCESS, SERVICE_FAILURE_EXIT_CODE); + + case SERVICE_RESTART_ON_WATCHDOG: + return s->result == SERVICE_FAILURE_WATCHDOG; + + case SERVICE_RESTART_ON_ABORT: + return IN_SET(s->result, SERVICE_FAILURE_SIGNAL, SERVICE_FAILURE_CORE_DUMP); + + default: + assert_not_reached("unknown restart setting"); + } +} + +static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) { + int r; + assert(s); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + service_set_state(s, s->result != SERVICE_SUCCESS ? SERVICE_FAILED : SERVICE_DEAD); + + if (s->result != SERVICE_SUCCESS) { + log_unit_warning(UNIT(s), "Failed with result '%s'.", service_result_to_string(s->result)); + emergency_action(UNIT(s)->manager, s->emergency_action, UNIT(s)->reboot_arg, "service failed"); + } + + if (allow_restart && service_shall_restart(s)) { + + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->restart_usec)); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_AUTO_RESTART); + } + + /* The next restart might not be a manual stop, hence reset the flag indicating manual stops */ + s->forbid_restart = false; + + /* We want fresh tmpdirs in case service is started again immediately */ + exec_runtime_destroy(s->exec_runtime); + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + + /* Also, remove the runtime directory */ + exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + + /* Get rid of the IPC bits of the user */ + unit_unref_uid_gid(UNIT(s), true); + + /* Release the user, and destroy it if we are the only remaining owner */ + dynamic_creds_destroy(&s->dynamic_creds); + + /* Try to delete the pid file. At this point it will be + * out-of-date, and some software might be confused by it, so + * let's remove it. */ + if (s->pid_file) + (void) unlink(s->pid_file); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run install restart timer: %m"); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, false); +} + +static void service_enter_stop_post(Service *s, ServiceResult f) { + int r; + assert(s); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + service_unwatch_control_pid(s); + unit_watch_all_pids(UNIT(s)); + + s->control_command = s->exec_command[SERVICE_EXEC_STOP_POST]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_STOP_POST; + + r = service_spawn(s, + s->control_command, + s->timeout_stop_usec, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_IS_CONTROL|EXEC_SETENV_RESULT, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_STOP_POST); + } else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'stop-post' task: %m"); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static int state_to_kill_operation(ServiceState state) { + switch (state) { + + case SERVICE_STOP_SIGABRT: + return KILL_ABORT; + + case SERVICE_STOP_SIGTERM: + case SERVICE_FINAL_SIGTERM: + return KILL_TERMINATE; + + case SERVICE_STOP_SIGKILL: + case SERVICE_FINAL_SIGKILL: + return KILL_KILL; + + default: + return _KILL_OPERATION_INVALID; + } +} + +static void service_enter_signal(Service *s, ServiceState state, ServiceResult f) { + int r; + + assert(s); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + unit_watch_all_pids(UNIT(s)); + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + state_to_kill_operation(state), + s->main_pid, + s->control_pid, + s->main_pid_alien); + + if (r < 0) + goto fail; + + if (r > 0) { + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_stop_usec)); + if (r < 0) + goto fail; + + service_set_state(s, state); + } else if (IN_SET(state, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM) && s->kill_context.send_sigkill) + service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_SUCCESS); + else if (IN_SET(state, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL)) + service_enter_stop_post(s, SERVICE_SUCCESS); + else if (state == SERVICE_FINAL_SIGTERM && s->kill_context.send_sigkill) + service_enter_signal(s, SERVICE_FINAL_SIGKILL, SERVICE_SUCCESS); + else + service_enter_dead(s, SERVICE_SUCCESS, true); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to kill processes: %m"); + + if (IN_SET(state, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL)) + service_enter_stop_post(s, SERVICE_FAILURE_RESOURCES); + else + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); +} + +static void service_enter_stop_by_notify(Service *s) { + assert(s); + + unit_watch_all_pids(UNIT(s)); + + service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_stop_usec)); + + /* The service told us it's stopping, so it's as if we SIGTERM'd it. */ + service_set_state(s, SERVICE_STOP_SIGTERM); +} + +static void service_enter_stop(Service *s, ServiceResult f) { + int r; + + assert(s); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + service_unwatch_control_pid(s); + unit_watch_all_pids(UNIT(s)); + + s->control_command = s->exec_command[SERVICE_EXEC_STOP]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_STOP; + + r = service_spawn(s, + s->control_command, + s->timeout_stop_usec, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL|EXEC_SETENV_RESULT, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_STOP); + } else + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'stop' task: %m"); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static bool service_good(Service *s) { + int main_pid_ok; + assert(s); + + if (s->type == SERVICE_DBUS && !s->bus_name_good) + return false; + + main_pid_ok = main_pid_good(s); + if (main_pid_ok > 0) /* It's alive */ + return true; + if (main_pid_ok == 0) /* It's dead */ + return false; + + /* OK, we don't know anything about the main PID, maybe + * because there is none. Let's check the control group + * instead. */ + + return cgroup_good(s) != 0; +} + +static void service_enter_running(Service *s, ServiceResult f) { + assert(s); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + service_unwatch_control_pid(s); + + if (service_good(s)) { + + /* If there are any queued up sd_notify() + * notifications, process them now */ + if (s->notify_state == NOTIFY_RELOADING) + service_enter_reload_by_notify(s); + else if (s->notify_state == NOTIFY_STOPPING) + service_enter_stop_by_notify(s); + else { + service_set_state(s, SERVICE_RUNNING); + service_arm_timer(s, usec_add(UNIT(s)->active_enter_timestamp.monotonic, s->runtime_max_usec)); + } + + } else if (s->remain_after_exit) + service_set_state(s, SERVICE_EXITED); + else + service_enter_stop(s, SERVICE_SUCCESS); +} + +static void service_enter_start_post(Service *s) { + int r; + assert(s); + + service_unwatch_control_pid(s); + service_reset_watchdog(s); + + s->control_command = s->exec_command[SERVICE_EXEC_START_POST]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_START_POST; + + r = service_spawn(s, + s->control_command, + s->timeout_start_usec, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_START_POST); + } else + service_enter_running(s, SERVICE_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'start-post' task: %m"); + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static void service_kill_control_processes(Service *s) { + char *p; + + if (!UNIT(s)->cgroup_path) + return; + + p = strjoina(UNIT(s)->cgroup_path, "/control"); + cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, p, SIGKILL, CGROUP_SIGCONT|CGROUP_IGNORE_SELF|CGROUP_REMOVE, NULL, NULL, NULL); +} + +static void service_enter_start(Service *s) { + ExecCommand *c; + usec_t timeout; + pid_t pid; + int r; + + assert(s); + + service_unwatch_control_pid(s); + service_unwatch_main_pid(s); + + /* We want to ensure that nobody leaks processes from + * START_PRE here, so let's go on a killing spree, People + * should not spawn long running processes from START_PRE. */ + service_kill_control_processes(s); + + if (s->type == SERVICE_FORKING) { + s->control_command_id = SERVICE_EXEC_START; + c = s->control_command = s->exec_command[SERVICE_EXEC_START]; + + s->main_command = NULL; + } else { + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + s->control_command = NULL; + + c = s->main_command = s->exec_command[SERVICE_EXEC_START]; + } + + if (!c) { + if (s->type != SERVICE_ONESHOT) { + /* There's no command line configured for the main command? Hmm, that is strange. This can only + * happen if the configuration changes at runtime. In this case, let's enter a failure + * state. */ + log_unit_error(UNIT(s), "There's no 'start' task anymore we could start: %m"); + r = -ENXIO; + goto fail; + } + + service_enter_start_post(s); + return; + } + + if (IN_SET(s->type, SERVICE_SIMPLE, SERVICE_IDLE)) + /* For simple + idle this is the main process. We don't apply any timeout here, but + * service_enter_running() will later apply the .runtime_max_usec timeout. */ + timeout = USEC_INFINITY; + else + timeout = s->timeout_start_usec; + + r = service_spawn(s, + c, + timeout, + EXEC_PASS_FDS|EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_SET_WATCHDOG, + &pid); + if (r < 0) + goto fail; + + if (IN_SET(s->type, SERVICE_SIMPLE, SERVICE_IDLE)) { + /* For simple services we immediately start + * the START_POST binaries. */ + + service_set_main_pid(s, pid); + service_enter_start_post(s); + + } else if (s->type == SERVICE_FORKING) { + + /* For forking services we wait until the start + * process exited. */ + + s->control_pid = pid; + service_set_state(s, SERVICE_START); + + } else if (IN_SET(s->type, SERVICE_ONESHOT, SERVICE_DBUS, SERVICE_NOTIFY)) { + + /* For oneshot services we wait until the start + * process exited, too, but it is our main process. */ + + /* For D-Bus services we know the main pid right away, + * but wait for the bus name to appear on the + * bus. Notify services are similar. */ + + service_set_main_pid(s, pid); + service_set_state(s, SERVICE_START); + } else + assert_not_reached("Unknown service type"); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'start' task: %m"); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); +} + +static void service_enter_start_pre(Service *s) { + int r; + + assert(s); + + service_unwatch_control_pid(s); + + s->control_command = s->exec_command[SERVICE_EXEC_START_PRE]; + if (s->control_command) { + /* Before we start anything, let's clear up what might + * be left from previous runs. */ + service_kill_control_processes(s); + + s->control_command_id = SERVICE_EXEC_START_PRE; + + r = service_spawn(s, + s->control_command, + s->timeout_start_usec, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL|EXEC_APPLY_TTY_STDIN, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_START_PRE); + } else + service_enter_start(s); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'start-pre' task: %m"); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); +} + +static void service_enter_restart(Service *s) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + assert(s); + + if (UNIT(s)->job && UNIT(s)->job->type == JOB_STOP) { + /* Don't restart things if we are going down anyway */ + log_unit_info(UNIT(s), "Stop job pending for unit, delaying automatic restart."); + + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->restart_usec)); + if (r < 0) + goto fail; + + return; + } + + /* Any units that are bound to this service must also be + * restarted. We use JOB_RESTART (instead of the more obvious + * JOB_START) here so that those dependency jobs will be added + * as well. */ + r = manager_add_job(UNIT(s)->manager, JOB_RESTART, UNIT(s), JOB_FAIL, &error, NULL); + if (r < 0) + goto fail; + + /* Note that we stay in the SERVICE_AUTO_RESTART state here, + * it will be canceled as part of the service_stop() call that + * is executed as part of JOB_RESTART. */ + + log_unit_debug(UNIT(s), "Scheduled restart job."); + return; + +fail: + log_unit_warning(UNIT(s), "Failed to schedule restart job: %s", bus_error_message(&error, -r)); + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, false); +} + +static void service_enter_reload_by_notify(Service *s) { + assert(s); + + service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_start_usec)); + service_set_state(s, SERVICE_RELOAD); +} + +static void service_enter_reload(Service *s) { + int r; + + assert(s); + + service_unwatch_control_pid(s); + s->reload_result = SERVICE_SUCCESS; + + s->control_command = s->exec_command[SERVICE_EXEC_RELOAD]; + if (s->control_command) { + s->control_command_id = SERVICE_EXEC_RELOAD; + + r = service_spawn(s, + s->control_command, + s->timeout_start_usec, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL, + &s->control_pid); + if (r < 0) + goto fail; + + service_set_state(s, SERVICE_RELOAD); + } else + service_enter_running(s, SERVICE_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'reload' task: %m"); + s->reload_result = SERVICE_FAILURE_RESOURCES; + service_enter_running(s, SERVICE_SUCCESS); +} + +static void service_run_next_control(Service *s) { + usec_t timeout; + int r; + + assert(s); + assert(s->control_command); + assert(s->control_command->command_next); + + assert(s->control_command_id != SERVICE_EXEC_START); + + s->control_command = s->control_command->command_next; + service_unwatch_control_pid(s); + + if (IN_SET(s->state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) + timeout = s->timeout_start_usec; + else + timeout = s->timeout_stop_usec; + + r = service_spawn(s, + s->control_command, + timeout, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL| + (IN_SET(s->control_command_id, SERVICE_EXEC_START_PRE, SERVICE_EXEC_STOP_POST) ? EXEC_APPLY_TTY_STDIN : 0)| + (IN_SET(s->control_command_id, SERVICE_EXEC_STOP, SERVICE_EXEC_STOP_POST) ? EXEC_SETENV_RESULT : 0), + &s->control_pid); + if (r < 0) + goto fail; + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run next control task: %m"); + + if (s->state == SERVICE_START_PRE) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + else if (s->state == SERVICE_STOP) + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); + else if (s->state == SERVICE_STOP_POST) + service_enter_dead(s, SERVICE_FAILURE_RESOURCES, true); + else if (s->state == SERVICE_RELOAD) { + s->reload_result = SERVICE_FAILURE_RESOURCES; + service_enter_running(s, SERVICE_SUCCESS); + } else + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static void service_run_next_main(Service *s) { + pid_t pid; + int r; + + assert(s); + assert(s->main_command); + assert(s->main_command->command_next); + assert(s->type == SERVICE_ONESHOT); + + s->main_command = s->main_command->command_next; + service_unwatch_main_pid(s); + + r = service_spawn(s, + s->main_command, + s->timeout_start_usec, + EXEC_PASS_FDS|EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_SET_WATCHDOG, + &pid); + if (r < 0) + goto fail; + + service_set_main_pid(s, pid); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run next main task: %m"); + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); +} + +static int service_start(Unit *u) { + Service *s = SERVICE(u); + int r; + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (IN_SET(s->state, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) + return -EAGAIN; + + /* Already on it! */ + if (IN_SET(s->state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST)) + return 0; + + /* A service that will be restarted must be stopped first to + * trigger BindsTo and/or OnFailure dependencies. If a user + * does not want to wait for the holdoff time to elapse, the + * service should be manually restarted, not started. We + * simply return EAGAIN here, so that any start jobs stay + * queued, and assume that the auto restart timer will + * eventually trigger the restart. */ + if (s->state == SERVICE_AUTO_RESTART) + return -EAGAIN; + + assert(IN_SET(s->state, SERVICE_DEAD, SERVICE_FAILED)); + + /* Make sure we don't enter a busy loop of some kind. */ + r = unit_start_limit_test(u); + if (r < 0) { + service_enter_dead(s, SERVICE_FAILURE_START_LIMIT_HIT, false); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + s->result = SERVICE_SUCCESS; + s->reload_result = SERVICE_SUCCESS; + s->main_pid_known = false; + s->main_pid_alien = false; + s->forbid_restart = false; + s->reset_cpu_usage = true; + + s->status_text = mfree(s->status_text); + s->status_errno = 0; + + s->notify_state = NOTIFY_UNKNOWN; + + s->watchdog_override_enable = false; + s->watchdog_override_usec = 0; + + service_enter_start_pre(s); + return 1; +} + +static int service_stop(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + /* Don't create restart jobs from manual stops. */ + s->forbid_restart = true; + + /* Already on it */ + if (IN_SET(s->state, + SERVICE_STOP, SERVICE_STOP_SIGABRT, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, SERVICE_STOP_POST, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL)) + return 0; + + /* A restart will be scheduled or is in progress. */ + if (s->state == SERVICE_AUTO_RESTART) { + service_set_state(s, SERVICE_DEAD); + return 0; + } + + /* If there's already something running we go directly into + * kill mode. */ + if (IN_SET(s->state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RELOAD)) { + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_SUCCESS); + return 0; + } + + assert(IN_SET(s->state, SERVICE_RUNNING, SERVICE_EXITED)); + + service_enter_stop(s, SERVICE_SUCCESS); + return 1; +} + +static int service_reload(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + assert(s->state == SERVICE_RUNNING || s->state == SERVICE_EXITED); + + service_enter_reload(s); + return 1; +} + +_pure_ static bool service_can_reload(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + return !!s->exec_command[SERVICE_EXEC_RELOAD]; +} + +static int service_serialize(Unit *u, FILE *f, FDSet *fds) { + Service *s = SERVICE(u); + ServiceFDStore *fs; + int r; + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", service_state_to_string(s->state)); + unit_serialize_item(u, f, "result", service_result_to_string(s->result)); + unit_serialize_item(u, f, "reload-result", service_result_to_string(s->reload_result)); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", PID_FMT, s->control_pid); + + if (s->main_pid_known && s->main_pid > 0) + unit_serialize_item_format(u, f, "main-pid", PID_FMT, s->main_pid); + + unit_serialize_item(u, f, "main-pid-known", yes_no(s->main_pid_known)); + unit_serialize_item(u, f, "bus-name-good", yes_no(s->bus_name_good)); + unit_serialize_item(u, f, "bus-name-owner", s->bus_name_owner); + + r = unit_serialize_item_escaped(u, f, "status-text", s->status_text); + if (r < 0) + return r; + + /* FIXME: There's a minor uncleanliness here: if there are + * multiple commands attached here, we will start from the + * first one again */ + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", service_exec_command_to_string(s->control_command_id)); + + r = unit_serialize_item_fd(u, f, fds, "stdin-fd", s->stdin_fd); + if (r < 0) + return r; + r = unit_serialize_item_fd(u, f, fds, "stdout-fd", s->stdout_fd); + if (r < 0) + return r; + r = unit_serialize_item_fd(u, f, fds, "stderr-fd", s->stderr_fd); + if (r < 0) + return r; + + if (UNIT_ISSET(s->accept_socket)) { + r = unit_serialize_item(u, f, "accept-socket", UNIT_DEREF(s->accept_socket)->id); + if (r < 0) + return r; + } + + r = unit_serialize_item_fd(u, f, fds, "socket-fd", s->socket_fd); + if (r < 0) + return r; + + LIST_FOREACH(fd_store, fs, s->fd_store) { + _cleanup_free_ char *c = NULL; + int copy; + + copy = fdset_put_dup(fds, fs->fd); + if (copy < 0) + return copy; + + c = cescape(fs->fdname); + + unit_serialize_item_format(u, f, "fd-store-fd", "%i %s", copy, strempty(c)); + } + + if (s->main_exec_status.pid > 0) { + unit_serialize_item_format(u, f, "main-exec-status-pid", PID_FMT, s->main_exec_status.pid); + dual_timestamp_serialize(f, "main-exec-status-start", &s->main_exec_status.start_timestamp); + dual_timestamp_serialize(f, "main-exec-status-exit", &s->main_exec_status.exit_timestamp); + + if (dual_timestamp_is_set(&s->main_exec_status.exit_timestamp)) { + unit_serialize_item_format(u, f, "main-exec-status-code", "%i", s->main_exec_status.code); + unit_serialize_item_format(u, f, "main-exec-status-status", "%i", s->main_exec_status.status); + } + } + + dual_timestamp_serialize(f, "watchdog-timestamp", &s->watchdog_timestamp); + + unit_serialize_item(u, f, "forbid-restart", yes_no(s->forbid_restart)); + + if (s->watchdog_override_enable) + unit_serialize_item_format(u, f, "watchdog-override-usec", USEC_FMT, s->watchdog_override_usec); + + return 0; +} + +static int service_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Service *s = SERVICE(u); + int r; + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + ServiceState state; + + state = service_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + ServiceResult f; + + f = service_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != SERVICE_SUCCESS) + s->result = f; + + } else if (streq(key, "reload-result")) { + ServiceResult f; + + f = service_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse reload result value: %s", value); + else if (f != SERVICE_SUCCESS) + s->reload_result = f; + + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse control-pid value: %s", value); + else + s->control_pid = pid; + } else if (streq(key, "main-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse main-pid value: %s", value); + else { + service_set_main_pid(s, pid); + unit_watch_pid(UNIT(s), pid); + } + } else if (streq(key, "main-pid-known")) { + int b; + + b = parse_boolean(value); + if (b < 0) + log_unit_debug(u, "Failed to parse main-pid-known value: %s", value); + else + s->main_pid_known = b; + } else if (streq(key, "bus-name-good")) { + int b; + + b = parse_boolean(value); + if (b < 0) + log_unit_debug(u, "Failed to parse bus-name-good value: %s", value); + else + s->bus_name_good = b; + } else if (streq(key, "bus-name-owner")) { + r = free_and_strdup(&s->bus_name_owner, value); + if (r < 0) + log_unit_error_errno(u, r, "Unable to deserialize current bus owner %s: %m", value); + } else if (streq(key, "status-text")) { + char *t; + + r = cunescape(value, 0, &t); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to unescape status text: %s", value); + else { + free(s->status_text); + s->status_text = t; + } + + } else if (streq(key, "control-command")) { + ServiceExecCommand id; + + id = service_exec_command_from_string(value); + if (id < 0) + log_unit_debug(u, "Failed to parse exec-command value: %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command[id]; + } + } else if (streq(key, "accept-socket")) { + Unit *socket; + + r = manager_load_unit(u->manager, value, NULL, NULL, &socket); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to load accept-socket unit: %s", value); + else { + unit_ref_set(&s->accept_socket, socket); + SOCKET(socket)->n_connections++; + } + + } else if (streq(key, "socket-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse socket-fd value: %s", value); + else { + asynchronous_close(s->socket_fd); + s->socket_fd = fdset_remove(fds, fd); + } + } else if (streq(key, "fd-store-fd")) { + const char *fdv; + size_t pf; + int fd; + + pf = strcspn(value, WHITESPACE); + fdv = strndupa(value, pf); + + if (safe_atoi(fdv, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse fd-store-fd value: %s", value); + else { + _cleanup_free_ char *t = NULL; + const char *fdn; + + fdn = value + pf; + fdn += strspn(fdn, WHITESPACE); + (void) cunescape(fdn, 0, &t); + + r = service_add_fd_store(s, fd, t); + if (r < 0) + log_unit_error_errno(u, r, "Failed to add fd to store: %m"); + else + fdset_remove(fds, fd); + } + + } else if (streq(key, "main-exec-status-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse main-exec-status-pid value: %s", value); + else + s->main_exec_status.pid = pid; + } else if (streq(key, "main-exec-status-code")) { + int i; + + if (safe_atoi(value, &i) < 0) + log_unit_debug(u, "Failed to parse main-exec-status-code value: %s", value); + else + s->main_exec_status.code = i; + } else if (streq(key, "main-exec-status-status")) { + int i; + + if (safe_atoi(value, &i) < 0) + log_unit_debug(u, "Failed to parse main-exec-status-status value: %s", value); + else + s->main_exec_status.status = i; + } else if (streq(key, "main-exec-status-start")) + dual_timestamp_deserialize(value, &s->main_exec_status.start_timestamp); + else if (streq(key, "main-exec-status-exit")) + dual_timestamp_deserialize(value, &s->main_exec_status.exit_timestamp); + else if (streq(key, "watchdog-timestamp")) + dual_timestamp_deserialize(value, &s->watchdog_timestamp); + else if (streq(key, "forbid-restart")) { + int b; + + b = parse_boolean(value); + if (b < 0) + log_unit_debug(u, "Failed to parse forbid-restart value: %s", value); + else + s->forbid_restart = b; + } else if (streq(key, "stdin-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse stdin-fd value: %s", value); + else { + asynchronous_close(s->stdin_fd); + s->stdin_fd = fdset_remove(fds, fd); + s->exec_context.stdio_as_fds = true; + } + } else if (streq(key, "stdout-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse stdout-fd value: %s", value); + else { + asynchronous_close(s->stdout_fd); + s->stdout_fd = fdset_remove(fds, fd); + s->exec_context.stdio_as_fds = true; + } + } else if (streq(key, "stderr-fd")) { + int fd; + + if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse stderr-fd value: %s", value); + else { + asynchronous_close(s->stderr_fd); + s->stderr_fd = fdset_remove(fds, fd); + s->exec_context.stdio_as_fds = true; + } + } else if (streq(key, "watchdog-override-usec")) { + usec_t watchdog_override_usec; + if (timestamp_deserialize(value, &watchdog_override_usec) < 0) + log_unit_debug(u, "Failed to parse watchdog_override_usec value: %s", value); + else { + s->watchdog_override_enable = true; + s->watchdog_override_usec = watchdog_override_usec; + } + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState service_active_state(Unit *u) { + const UnitActiveState *table; + + assert(u); + + table = SERVICE(u)->type == SERVICE_IDLE ? state_translation_table_idle : state_translation_table; + + return table[SERVICE(u)->state]; +} + +static const char *service_sub_state_to_string(Unit *u) { + assert(u); + + return service_state_to_string(SERVICE(u)->state); +} + +static bool service_check_gc(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + /* Never clean up services that still have a process around, + * even if the service is formally dead. */ + if (cgroup_good(s) > 0 || + main_pid_good(s) > 0 || + control_pid_good(s) > 0) + return true; + + return false; +} + +static int service_retry_pid_file(Service *s) { + int r; + + assert(s->pid_file); + assert(s->state == SERVICE_START || s->state == SERVICE_START_POST); + + r = service_load_pid_file(s, false); + if (r < 0) + return r; + + service_unwatch_pid_file(s); + + service_enter_running(s, SERVICE_SUCCESS); + return 0; +} + +static int service_watch_pid_file(Service *s) { + int r; + + log_unit_debug(UNIT(s), "Setting watch for PID file %s", s->pid_file_pathspec->path); + + r = path_spec_watch(s->pid_file_pathspec, service_dispatch_io); + if (r < 0) + goto fail; + + /* the pidfile might have appeared just before we set the watch */ + log_unit_debug(UNIT(s), "Trying to read PID file %s in case it changed", s->pid_file_pathspec->path); + service_retry_pid_file(s); + + return 0; +fail: + log_unit_error_errno(UNIT(s), r, "Failed to set a watch for PID file %s: %m", s->pid_file_pathspec->path); + service_unwatch_pid_file(s); + return r; +} + +static int service_demand_pid_file(Service *s) { + PathSpec *ps; + + assert(s->pid_file); + assert(!s->pid_file_pathspec); + + ps = new0(PathSpec, 1); + if (!ps) + return -ENOMEM; + + ps->unit = UNIT(s); + ps->path = strdup(s->pid_file); + if (!ps->path) { + free(ps); + return -ENOMEM; + } + + path_kill_slashes(ps->path); + + /* PATH_CHANGED would not be enough. There are daemons (sendmail) that + * keep their PID file open all the time. */ + ps->type = PATH_MODIFIED; + ps->inotify_fd = -1; + + s->pid_file_pathspec = ps; + + return service_watch_pid_file(s); +} + +static int service_dispatch_io(sd_event_source *source, int fd, uint32_t events, void *userdata) { + PathSpec *p = userdata; + Service *s; + + assert(p); + + s = SERVICE(p->unit); + + assert(s); + assert(fd >= 0); + assert(s->state == SERVICE_START || s->state == SERVICE_START_POST); + assert(s->pid_file_pathspec); + assert(path_spec_owns_inotify_fd(s->pid_file_pathspec, fd)); + + log_unit_debug(UNIT(s), "inotify event"); + + if (path_spec_fd_event(p, events) < 0) + goto fail; + + if (service_retry_pid_file(s) == 0) + return 0; + + if (service_watch_pid_file(s) < 0) + goto fail; + + return 0; + +fail: + service_unwatch_pid_file(s); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_RESOURCES); + return 0; +} + +static void service_notify_cgroup_empty_event(Unit *u) { + Service *s = SERVICE(u); + + assert(u); + + log_unit_debug(u, "cgroup is empty"); + + switch (s->state) { + + /* Waiting for SIGCHLD is usually more interesting, + * because it includes return codes/signals. Which is + * why we ignore the cgroup events for most cases, + * except when we don't know pid which to expect the + * SIGCHLD for. */ + + case SERVICE_START: + case SERVICE_START_POST: + /* If we were hoping for the daemon to write its PID file, + * we can give up now. */ + if (s->pid_file_pathspec) { + log_unit_warning(u, "Daemon never wrote its PID file. Failing."); + + service_unwatch_pid_file(s); + if (s->state == SERVICE_START) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + else + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); + } + break; + + case SERVICE_RUNNING: + /* service_enter_running() will figure out what to do */ + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP_SIGABRT: + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_stop_post(s, SERVICE_SUCCESS); + + break; + + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + if (main_pid_good(s) <= 0 && !control_pid_good(s)) + service_enter_dead(s, SERVICE_SUCCESS, true); + + break; + + default: + ; + } +} + +static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Service *s = SERVICE(u); + ServiceResult f; + + assert(s); + assert(pid >= 0); + + if (is_clean_exit(code, status, s->type == SERVICE_ONESHOT ? EXIT_CLEAN_COMMAND : EXIT_CLEAN_DAEMON, &s->success_status)) + f = SERVICE_SUCCESS; + else if (code == CLD_EXITED) + f = SERVICE_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SERVICE_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SERVICE_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (s->main_pid == pid) { + /* Forking services may occasionally move to a new PID. + * As long as they update the PID file before exiting the old + * PID, they're fine. */ + if (service_load_pid_file(s, false) == 0) + return; + + s->main_pid = 0; + exec_status_exit(&s->main_exec_status, &s->exec_context, pid, code, status); + + if (s->main_command) { + /* If this is not a forking service than the + * main process got started and hence we copy + * the exit status so that it is recorded both + * as main and as control process exit + * status */ + + s->main_command->exec_status = s->main_exec_status; + + if (s->main_command->ignore) + f = SERVICE_SUCCESS; + } else if (s->exec_command[SERVICE_EXEC_START]) { + + /* If this is a forked process, then we should + * ignore the return value if this was + * configured for the starter process */ + + if (s->exec_command[SERVICE_EXEC_START]->ignore) + f = SERVICE_SUCCESS; + } + + /* When this is a successful exit, let's log about the exit code on DEBUG level. If this is a failure + * and the process exited on its own via exit(), then let's make this a NOTICE, under the assumption + * that the service already logged the reason at a higher log level on its own. However, if the service + * died due to a signal, then it most likely didn't say anything about any reason, hence let's raise + * our log level to WARNING then. */ + + log_struct(f == SERVICE_SUCCESS ? LOG_DEBUG : + (code == CLD_EXITED ? LOG_NOTICE : LOG_WARNING), + LOG_UNIT_ID(u), + LOG_UNIT_MESSAGE(u, "Main process exited, code=%s, status=%i/%s", + sigchld_code_to_string(code), status, + strna(code == CLD_EXITED + ? exit_status_to_string(status, EXIT_STATUS_FULL) + : signal_to_string(status))), + "EXIT_CODE=%s", sigchld_code_to_string(code), + "EXIT_STATUS=%i", status, + NULL); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + if (s->main_command && + s->main_command->command_next && + f == SERVICE_SUCCESS) { + + /* There is another command to * + * execute, so let's do that. */ + + log_unit_debug(u, "Running next main command for state %s.", service_state_to_string(s->state)); + service_run_next_main(s); + + } else { + + /* The service exited, so the service is officially + * gone. */ + s->main_command = NULL; + + switch (s->state) { + + case SERVICE_START_POST: + case SERVICE_RELOAD: + case SERVICE_STOP: + /* Need to wait until the operation is + * done */ + break; + + case SERVICE_START: + if (s->type == SERVICE_ONESHOT) { + /* This was our main goal, so let's go on */ + if (f == SERVICE_SUCCESS) + service_enter_start_post(s); + else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + } + + /* Fall through */ + + case SERVICE_RUNNING: + service_enter_running(s, f); + break; + + case SERVICE_STOP_SIGABRT: + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + + if (!control_pid_good(s)) + service_enter_stop_post(s, f); + + /* If there is still a control process, wait for that first */ + break; + + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + + if (!control_pid_good(s)) + service_enter_dead(s, f, true); + break; + + default: + assert_not_reached("Uh, main process died at wrong time."); + } + } + + } else if (s->control_pid == pid) { + s->control_pid = 0; + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, &s->exec_context, pid, code, status); + + if (s->control_command->ignore) + f = SERVICE_SUCCESS; + } + + log_unit_full(u, f == SERVICE_SUCCESS ? LOG_DEBUG : LOG_NOTICE, 0, + "Control process exited, code=%s status=%i", + sigchld_code_to_string(code), status); + + if (s->result == SERVICE_SUCCESS) + s->result = f; + + /* Immediately get rid of the cgroup, so that the + * kernel doesn't delay the cgroup empty messages for + * the service cgroup any longer than necessary */ + service_kill_control_processes(s); + + if (s->control_command && + s->control_command->command_next && + f == SERVICE_SUCCESS) { + + /* There is another command to * + * execute, so let's do that. */ + + log_unit_debug(u, "Running next control command for state %s.", service_state_to_string(s->state)); + service_run_next_control(s); + + } else { + /* No further commands for this step, so let's + * figure out what to do next */ + + s->control_command = NULL; + s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; + + log_unit_debug(u, "Got final SIGCHLD for state %s.", service_state_to_string(s->state)); + + switch (s->state) { + + case SERVICE_START_PRE: + if (f == SERVICE_SUCCESS) + service_enter_start(s); + else + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + + case SERVICE_START: + if (s->type != SERVICE_FORKING) + /* Maybe spurious event due to a reload that changed the type? */ + break; + + if (f != SERVICE_SUCCESS) { + service_enter_signal(s, SERVICE_FINAL_SIGTERM, f); + break; + } + + if (s->pid_file) { + bool has_start_post; + int r; + + /* Let's try to load the pid file here if we can. + * The PID file might actually be created by a START_POST + * script. In that case don't worry if the loading fails. */ + + has_start_post = !!s->exec_command[SERVICE_EXEC_START_POST]; + r = service_load_pid_file(s, !has_start_post); + if (!has_start_post && r < 0) { + r = service_demand_pid_file(s); + if (r < 0 || !cgroup_good(s)) + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_RESOURCES); + break; + } + } else + service_search_main_pid(s); + + service_enter_start_post(s); + break; + + case SERVICE_START_POST: + if (f != SERVICE_SUCCESS) { + service_enter_signal(s, SERVICE_STOP_SIGTERM, f); + break; + } + + if (s->pid_file) { + int r; + + r = service_load_pid_file(s, true); + if (r < 0) { + r = service_demand_pid_file(s); + if (r < 0 || !cgroup_good(s)) + service_enter_stop(s, SERVICE_FAILURE_RESOURCES); + break; + } + } else + service_search_main_pid(s); + + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_RELOAD: + if (f == SERVICE_SUCCESS) + if (service_load_pid_file(s, true) < 0) + service_search_main_pid(s); + + s->reload_result = f; + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP: + service_enter_signal(s, SERVICE_STOP_SIGTERM, f); + break; + + case SERVICE_STOP_SIGABRT: + case SERVICE_STOP_SIGTERM: + case SERVICE_STOP_SIGKILL: + if (main_pid_good(s) <= 0) + service_enter_stop_post(s, f); + + /* If there is still a service + * process around, wait until + * that one quit, too */ + break; + + case SERVICE_STOP_POST: + case SERVICE_FINAL_SIGTERM: + case SERVICE_FINAL_SIGKILL: + if (main_pid_good(s) <= 0) + service_enter_dead(s, f, true); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + } + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); + + /* We got one SIGCHLD for the service, let's watch all + * processes that are now running of the service, and watch + * that. Among the PIDs we then watch will be children + * reassigned to us, which hopefully allows us to identify + * when all children are gone */ + unit_tidy_watch_pids(u, s->main_pid, s->control_pid); + unit_watch_all_pids(u); + + /* If the PID set is empty now, then let's finish this off + (On unified we use proper notifications) */ + if (cg_unified(SYSTEMD_CGROUP_CONTROLLER) <= 0 && set_isempty(u->pids)) + service_notify_cgroup_empty_event(u); +} + +static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + Service *s = SERVICE(userdata); + + assert(s); + assert(source == s->timer_event_source); + + switch (s->state) { + + case SERVICE_START_PRE: + case SERVICE_START: + log_unit_warning(UNIT(s), "%s operation timed out. Terminating.", s->state == SERVICE_START ? "Start" : "Start-pre"); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_START_POST: + log_unit_warning(UNIT(s), "Start-post operation timed out. Stopping."); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_RUNNING: + log_unit_warning(UNIT(s), "Service reached runtime time limit. Stopping."); + service_enter_stop(s, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_RELOAD: + log_unit_warning(UNIT(s), "Reload operation timed out. Killing reload process."); + service_kill_control_processes(s); + s->reload_result = SERVICE_FAILURE_TIMEOUT; + service_enter_running(s, SERVICE_SUCCESS); + break; + + case SERVICE_STOP: + log_unit_warning(UNIT(s), "Stopping timed out. Terminating."); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_STOP_SIGABRT: + log_unit_warning(UNIT(s), "State 'stop-sigabrt' timed out. Terminating."); + service_enter_signal(s, SERVICE_STOP_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_STOP_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "State 'stop-sigterm' timed out. Killing."); + service_enter_signal(s, SERVICE_STOP_SIGKILL, SERVICE_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "State 'stop-sigterm' timed out. Skipping SIGKILL."); + service_enter_stop_post(s, SERVICE_FAILURE_TIMEOUT); + } + + break; + + case SERVICE_STOP_SIGKILL: + /* Uh, we sent a SIGKILL and it is still not gone? + * Must be something we cannot kill, so let's just be + * weirded out and continue */ + + log_unit_warning(UNIT(s), "Processes still around after SIGKILL. Ignoring."); + service_enter_stop_post(s, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_STOP_POST: + log_unit_warning(UNIT(s), "State 'stop-post' timed out. Terminating."); + service_enter_signal(s, SERVICE_FINAL_SIGTERM, SERVICE_FAILURE_TIMEOUT); + break; + + case SERVICE_FINAL_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "State 'stop-final-sigterm' timed out. Killing."); + service_enter_signal(s, SERVICE_FINAL_SIGKILL, SERVICE_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "State 'stop-final-sigterm' timed out. Skipping SIGKILL. Entering failed mode."); + service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, false); + } + + break; + + case SERVICE_FINAL_SIGKILL: + log_unit_warning(UNIT(s), "Processes still around after final SIGKILL. Entering failed mode."); + service_enter_dead(s, SERVICE_FAILURE_TIMEOUT, true); + break; + + case SERVICE_AUTO_RESTART: + log_unit_info(UNIT(s), + s->restart_usec > 0 ? + "Service hold-off time over, scheduling restart." : + "Service has no hold-off time, scheduling restart."); + service_enter_restart(s); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +static int service_dispatch_watchdog(sd_event_source *source, usec_t usec, void *userdata) { + Service *s = SERVICE(userdata); + char t[FORMAT_TIMESPAN_MAX]; + usec_t watchdog_usec; + + assert(s); + assert(source == s->watchdog_event_source); + + watchdog_usec = service_get_watchdog_usec(s); + + log_unit_error(UNIT(s), "Watchdog timeout (limit %s)!", + format_timespan(t, sizeof(t), watchdog_usec, 1)); + + service_enter_signal(s, SERVICE_STOP_SIGABRT, SERVICE_FAILURE_WATCHDOG); + + return 0; +} + +static void service_notify_message(Unit *u, pid_t pid, char **tags, FDSet *fds) { + Service *s = SERVICE(u); + _cleanup_free_ char *cc = NULL; + bool notify_dbus = false; + const char *e; + + assert(u); + + cc = strv_join(tags, ", "); + + if (s->notify_access == NOTIFY_NONE) { + log_unit_warning(u, "Got notification message from PID "PID_FMT", but reception is disabled.", pid); + return; + } else if (s->notify_access == NOTIFY_MAIN && pid != s->main_pid) { + if (s->main_pid != 0) + log_unit_warning(u, "Got notification message from PID "PID_FMT", but reception only permitted for main PID "PID_FMT, pid, s->main_pid); + else + log_unit_debug(u, "Got notification message from PID "PID_FMT", but reception only permitted for main PID which is currently not known", pid); + return; + } else + log_unit_debug(u, "Got notification message from PID "PID_FMT" (%s)", pid, isempty(cc) ? "n/a" : cc); + + /* Interpret MAINPID= */ + e = strv_find_startswith(tags, "MAINPID="); + if (e && IN_SET(s->state, SERVICE_START, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) { + if (parse_pid(e, &pid) < 0) + log_unit_warning(u, "Failed to parse MAINPID= field in notification message: %s", e); + else { + service_set_main_pid(s, pid); + unit_watch_pid(UNIT(s), pid); + notify_dbus = true; + } + } + + /* Interpret RELOADING= */ + if (strv_find(tags, "RELOADING=1")) { + + s->notify_state = NOTIFY_RELOADING; + + if (s->state == SERVICE_RUNNING) + service_enter_reload_by_notify(s); + + notify_dbus = true; + } + + /* Interpret READY= */ + if (strv_find(tags, "READY=1")) { + + s->notify_state = NOTIFY_READY; + + /* Type=notify services inform us about completed + * initialization with READY=1 */ + if (s->type == SERVICE_NOTIFY && s->state == SERVICE_START) + service_enter_start_post(s); + + /* Sending READY=1 while we are reloading informs us + * that the reloading is complete */ + if (s->state == SERVICE_RELOAD && s->control_pid == 0) + service_enter_running(s, SERVICE_SUCCESS); + + notify_dbus = true; + } + + /* Interpret STOPPING= */ + if (strv_find(tags, "STOPPING=1")) { + + s->notify_state = NOTIFY_STOPPING; + + if (s->state == SERVICE_RUNNING) + service_enter_stop_by_notify(s); + + notify_dbus = true; + } + + /* Interpret STATUS= */ + e = strv_find_startswith(tags, "STATUS="); + if (e) { + _cleanup_free_ char *t = NULL; + + if (!isempty(e)) { + if (!utf8_is_valid(e)) + log_unit_warning(u, "Status message in notification message is not UTF-8 clean."); + else { + t = strdup(e); + if (!t) + log_oom(); + } + } + + if (!streq_ptr(s->status_text, t)) { + + free_and_replace(s->status_text, t); + + notify_dbus = true; + } + } + + /* Interpret ERRNO= */ + e = strv_find_startswith(tags, "ERRNO="); + if (e) { + int status_errno; + + if (safe_atoi(e, &status_errno) < 0 || status_errno < 0) + log_unit_warning(u, "Failed to parse ERRNO= field in notification message: %s", e); + else { + if (s->status_errno != status_errno) { + s->status_errno = status_errno; + notify_dbus = true; + } + } + } + + /* Interpret WATCHDOG= */ + if (strv_find(tags, "WATCHDOG=1")) + service_reset_watchdog(s); + + if (strv_find(tags, "FDSTORE=1")) { + const char *name; + + name = strv_find_startswith(tags, "FDNAME="); + if (name && !fdname_is_valid(name)) { + log_unit_warning(u, "Passed FDNAME= name is invalid, ignoring."); + name = NULL; + } + + service_add_fd_store_set(s, fds, name); + } + + e = strv_find_startswith(tags, "WATCHDOG_USEC="); + if (e) { + usec_t watchdog_override_usec; + if (safe_atou64(e, &watchdog_override_usec) < 0) + log_unit_warning(u, "Failed to parse WATCHDOG_USEC=%s", e); + else + service_reset_watchdog_timeout(s, watchdog_override_usec); + } + + /* Notify clients about changed status or main pid */ + if (notify_dbus) + unit_add_to_dbus_queue(u); +} + +static int service_get_timeout(Unit *u, usec_t *timeout) { + Service *s = SERVICE(u); + uint64_t t; + int r; + + if (!s->timer_event_source) + return 0; + + r = sd_event_source_get_time(s->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +static void service_bus_name_owner_change( + Unit *u, + const char *name, + const char *old_owner, + const char *new_owner) { + + Service *s = SERVICE(u); + int r; + + assert(s); + assert(name); + + assert(streq(s->bus_name, name)); + assert(old_owner || new_owner); + + if (old_owner && new_owner) + log_unit_debug(u, "D-Bus name %s changed owner from %s to %s", name, old_owner, new_owner); + else if (old_owner) + log_unit_debug(u, "D-Bus name %s no longer registered by %s", name, old_owner); + else + log_unit_debug(u, "D-Bus name %s now registered by %s", name, new_owner); + + s->bus_name_good = !!new_owner; + + /* Track the current owner, so we can reconstruct changes after a daemon reload */ + r = free_and_strdup(&s->bus_name_owner, new_owner); + if (r < 0) { + log_unit_error_errno(u, r, "Unable to set new bus name owner %s: %m", new_owner); + return; + } + + if (s->type == SERVICE_DBUS) { + + /* service_enter_running() will figure out what to + * do */ + if (s->state == SERVICE_RUNNING) + service_enter_running(s, SERVICE_SUCCESS); + else if (s->state == SERVICE_START && new_owner) + service_enter_start_post(s); + + } else if (new_owner && + s->main_pid <= 0 && + (s->state == SERVICE_START || + s->state == SERVICE_START_POST || + s->state == SERVICE_RUNNING || + s->state == SERVICE_RELOAD)) { + + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; + pid_t pid; + + /* Try to acquire PID from bus service */ + + r = sd_bus_get_name_creds(u->manager->api_bus, name, SD_BUS_CREDS_PID, &creds); + if (r >= 0) + r = sd_bus_creds_get_pid(creds, &pid); + if (r >= 0) { + log_unit_debug(u, "D-Bus name %s is now owned by process %u", name, (unsigned) pid); + + service_set_main_pid(s, pid); + unit_watch_pid(UNIT(s), pid); + } + } +} + +int service_set_socket_fd(Service *s, int fd, Socket *sock, bool selinux_context_net) { + _cleanup_free_ char *peer = NULL; + int r; + + assert(s); + assert(fd >= 0); + + /* This is called by the socket code when instantiating a new service for a stream socket and the socket needs + * to be configured. We take ownership of the passed fd on success. */ + + if (UNIT(s)->load_state != UNIT_LOADED) + return -EINVAL; + + if (s->socket_fd >= 0) + return -EBUSY; + + if (s->state != SERVICE_DEAD) + return -EAGAIN; + + if (getpeername_pretty(fd, true, &peer) >= 0) { + + if (UNIT(s)->description) { + _cleanup_free_ char *a; + + a = strjoin(UNIT(s)->description, " (", peer, ")", NULL); + if (!a) + return -ENOMEM; + + r = unit_set_description(UNIT(s), a); + } else + r = unit_set_description(UNIT(s), peer); + + if (r < 0) + return r; + } + + r = unit_add_two_dependencies(UNIT(sock), UNIT_BEFORE, UNIT_TRIGGERS, UNIT(s), false); + if (r < 0) + return r; + + s->socket_fd = fd; + s->socket_fd_selinux_context_net = selinux_context_net; + + unit_ref_set(&s->accept_socket, UNIT(sock)); + return 0; +} + +static void service_reset_failed(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + if (s->state == SERVICE_FAILED) + service_set_state(s, SERVICE_DEAD); + + s->result = SERVICE_SUCCESS; + s->reload_result = SERVICE_SUCCESS; +} + +static int service_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + Service *s = SERVICE(u); + + return unit_kill_common(u, who, signo, s->main_pid, s->control_pid, error); +} + +static int service_main_pid(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + return s->main_pid; +} + +static int service_control_pid(Unit *u) { + Service *s = SERVICE(u); + + assert(s); + + return s->control_pid; +} + +static const char* const service_restart_table[_SERVICE_RESTART_MAX] = { + [SERVICE_RESTART_NO] = "no", + [SERVICE_RESTART_ON_SUCCESS] = "on-success", + [SERVICE_RESTART_ON_FAILURE] = "on-failure", + [SERVICE_RESTART_ON_ABNORMAL] = "on-abnormal", + [SERVICE_RESTART_ON_WATCHDOG] = "on-watchdog", + [SERVICE_RESTART_ON_ABORT] = "on-abort", + [SERVICE_RESTART_ALWAYS] = "always", +}; + +DEFINE_STRING_TABLE_LOOKUP(service_restart, ServiceRestart); + +static const char* const service_type_table[_SERVICE_TYPE_MAX] = { + [SERVICE_SIMPLE] = "simple", + [SERVICE_FORKING] = "forking", + [SERVICE_ONESHOT] = "oneshot", + [SERVICE_DBUS] = "dbus", + [SERVICE_NOTIFY] = "notify", + [SERVICE_IDLE] = "idle" +}; + +DEFINE_STRING_TABLE_LOOKUP(service_type, ServiceType); + +static const char* const service_exec_command_table[_SERVICE_EXEC_COMMAND_MAX] = { + [SERVICE_EXEC_START_PRE] = "ExecStartPre", + [SERVICE_EXEC_START] = "ExecStart", + [SERVICE_EXEC_START_POST] = "ExecStartPost", + [SERVICE_EXEC_RELOAD] = "ExecReload", + [SERVICE_EXEC_STOP] = "ExecStop", + [SERVICE_EXEC_STOP_POST] = "ExecStopPost", +}; + +DEFINE_STRING_TABLE_LOOKUP(service_exec_command, ServiceExecCommand); + +static const char* const notify_access_table[_NOTIFY_ACCESS_MAX] = { + [NOTIFY_NONE] = "none", + [NOTIFY_MAIN] = "main", + [NOTIFY_ALL] = "all" +}; + +DEFINE_STRING_TABLE_LOOKUP(notify_access, NotifyAccess); + +static const char* const notify_state_table[_NOTIFY_STATE_MAX] = { + [NOTIFY_UNKNOWN] = "unknown", + [NOTIFY_READY] = "ready", + [NOTIFY_RELOADING] = "reloading", + [NOTIFY_STOPPING] = "stopping", +}; + +DEFINE_STRING_TABLE_LOOKUP(notify_state, NotifyState); + +static const char* const service_result_table[_SERVICE_RESULT_MAX] = { + [SERVICE_SUCCESS] = "success", + [SERVICE_FAILURE_RESOURCES] = "resources", + [SERVICE_FAILURE_TIMEOUT] = "timeout", + [SERVICE_FAILURE_EXIT_CODE] = "exit-code", + [SERVICE_FAILURE_SIGNAL] = "signal", + [SERVICE_FAILURE_CORE_DUMP] = "core-dump", + [SERVICE_FAILURE_WATCHDOG] = "watchdog", + [SERVICE_FAILURE_START_LIMIT_HIT] = "start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(service_result, ServiceResult); + +const UnitVTable service_vtable = { + .object_size = sizeof(Service), + .exec_context_offset = offsetof(Service, exec_context), + .cgroup_context_offset = offsetof(Service, cgroup_context), + .kill_context_offset = offsetof(Service, kill_context), + .exec_runtime_offset = offsetof(Service, exec_runtime), + .dynamic_creds_offset = offsetof(Service, dynamic_creds), + + .sections = + "Unit\0" + "Service\0" + "Install\0", + .private_section = "Service", + + .init = service_init, + .done = service_done, + .load = service_load, + .release_resources = service_release_resources, + + .coldplug = service_coldplug, + + .dump = service_dump, + + .start = service_start, + .stop = service_stop, + .reload = service_reload, + + .can_reload = service_can_reload, + + .kill = service_kill, + + .serialize = service_serialize, + .deserialize_item = service_deserialize_item, + + .active_state = service_active_state, + .sub_state_to_string = service_sub_state_to_string, + + .check_gc = service_check_gc, + + .sigchld_event = service_sigchld_event, + + .reset_failed = service_reset_failed, + + .notify_cgroup_empty = service_notify_cgroup_empty_event, + .notify_message = service_notify_message, + + .main_pid = service_main_pid, + .control_pid = service_control_pid, + + .bus_name_owner_change = service_bus_name_owner_change, + + .bus_vtable = bus_service_vtable, + .bus_set_property = bus_service_set_property, + .bus_commit_properties = bus_service_commit_properties, + + .get_timeout = service_get_timeout, + .can_transient = true, + + .status_message_formats = { + .starting_stopping = { + [0] = "Starting %s...", + [1] = "Stopping %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Started %s.", + [JOB_FAILED] = "Failed to start %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Stopped %s.", + [JOB_FAILED] = "Stopped (with error) %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/show-status.c b/src/grp-system/libcore/src/show-status.c new file mode 100644 index 0000000000..1b2a7480d5 --- /dev/null +++ b/src/grp-system/libcore/src/show-status.c @@ -0,0 +1,129 @@ +/*** + This file is part of systemd. + + Copyright 2014 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/show-status.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/terminal-util.h" +#include "systemd-basic/util.h" + +int parse_show_status(const char *v, ShowStatus *ret) { + int r; + + assert(v); + assert(ret); + + if (streq(v, "auto")) { + *ret = SHOW_STATUS_AUTO; + return 0; + } + + r = parse_boolean(v); + if (r < 0) + return r; + + *ret = r ? SHOW_STATUS_YES : SHOW_STATUS_NO; + return 0; +} + +int status_vprintf(const char *status, bool ellipse, bool ephemeral, const char *format, va_list ap) { + static const char status_indent[] = " "; /* "[" STATUS "] " */ + _cleanup_free_ char *s = NULL; + _cleanup_close_ int fd = -1; + struct iovec iovec[6] = {}; + int n = 0; + static bool prev_ephemeral; + + assert(format); + + /* This is independent of logging, as status messages are + * optional and go exclusively to the console. */ + + if (vasprintf(&s, format, ap) < 0) + return log_oom(); + + /* Before you ask: yes, on purpose we open/close the console for each status line we write individually. This + * is a good strategy to avoid PID 1 getting killed by the kernel's SAK concept (it doesn't fix this entirely, + * but minimizes the time window the kernel might end up killing PID 1 due to SAK). It also makes things easier + * for us so that we don't have to recover from hangups and suchlike triggered on the console. */ + + fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (fd < 0) + return fd; + + if (ellipse) { + char *e; + size_t emax, sl; + int c; + + c = fd_columns(fd); + if (c <= 0) + c = 80; + + sl = status ? sizeof(status_indent)-1 : 0; + + emax = c - sl - 1; + if (emax < 3) + emax = 3; + + e = ellipsize(s, emax, 50); + if (e) { + free(s); + s = e; + } + } + + if (prev_ephemeral) + IOVEC_SET_STRING(iovec[n++], "\r" ANSI_ERASE_TO_END_OF_LINE); + prev_ephemeral = ephemeral; + + if (status) { + if (!isempty(status)) { + IOVEC_SET_STRING(iovec[n++], "["); + IOVEC_SET_STRING(iovec[n++], status); + IOVEC_SET_STRING(iovec[n++], "] "); + } else + IOVEC_SET_STRING(iovec[n++], status_indent); + } + + IOVEC_SET_STRING(iovec[n++], s); + if (!ephemeral) + IOVEC_SET_STRING(iovec[n++], "\n"); + + if (writev(fd, iovec, n) < 0) + return -errno; + + return 0; +} + +int status_printf(const char *status, bool ellipse, bool ephemeral, const char *format, ...) { + va_list ap; + int r; + + assert(format); + + va_start(ap, format); + r = status_vprintf(status, ellipse, ephemeral, format, ap); + va_end(ap); + + return r; +} diff --git a/src/grp-system/libcore/src/slice.c b/src/grp-system/libcore/src/slice.c new file mode 100644 index 0000000000..8bb47534fc --- /dev/null +++ b/src/grp-system/libcore/src/slice.c @@ -0,0 +1,361 @@ +/*** + This file is part of systemd. + + Copyright 2013 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "core/slice.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" + +#include "dbus-slice.h" + +static const UnitActiveState state_translation_table[_SLICE_STATE_MAX] = { + [SLICE_DEAD] = UNIT_INACTIVE, + [SLICE_ACTIVE] = UNIT_ACTIVE +}; + +static void slice_init(Unit *u) { + assert(u); + assert(u->load_state == UNIT_STUB); + + u->ignore_on_isolate = true; +} + +static void slice_set_state(Slice *t, SliceState state) { + SliceState old_state; + assert(t); + + old_state = t->state; + t->state = state; + + if (state != old_state) + log_debug("%s changed %s -> %s", + UNIT(t)->id, + slice_state_to_string(old_state), + slice_state_to_string(state)); + + unit_notify(UNIT(t), state_translation_table[old_state], state_translation_table[state], true); +} + +static int slice_add_parent_slice(Slice *s) { + char *a, *dash; + Unit *parent; + int r; + + assert(s); + + if (UNIT_ISSET(UNIT(s)->slice)) + return 0; + + if (unit_has_name(UNIT(s), SPECIAL_ROOT_SLICE)) + return 0; + + a = strdupa(UNIT(s)->id); + dash = strrchr(a, '-'); + if (dash) + strcpy(dash, ".slice"); + else + a = (char*) SPECIAL_ROOT_SLICE; + + r = manager_load_unit(UNIT(s)->manager, a, NULL, NULL, &parent); + if (r < 0) + return r; + + unit_ref_set(&UNIT(s)->slice, parent); + return 0; +} + +static int slice_add_default_dependencies(Slice *s) { + int r; + + assert(s); + + if (!UNIT(s)->default_dependencies) + return 0; + + /* Make sure slices are unloaded on shutdown */ + r = unit_add_two_dependencies_by_name( + UNIT(s), + UNIT_BEFORE, UNIT_CONFLICTS, + SPECIAL_SHUTDOWN_TARGET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +static int slice_verify(Slice *s) { + _cleanup_free_ char *parent = NULL; + int r; + + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (!slice_name_is_valid(UNIT(s)->id)) { + log_unit_error(UNIT(s), "Slice name %s is not valid. Refusing.", UNIT(s)->id); + return -EINVAL; + } + + r = slice_build_parent_slice(UNIT(s)->id, &parent); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Failed to determine parent slice: %m"); + + if (parent ? !unit_has_name(UNIT_DEREF(UNIT(s)->slice), parent) : UNIT_ISSET(UNIT(s)->slice)) { + log_unit_error(UNIT(s), "Located outside of parent slice. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int slice_load_root_slice(Unit *u) { + assert(u); + + if (!unit_has_name(u, SPECIAL_ROOT_SLICE)) + return 0; + + u->perpetual = true; + + /* The root slice is a bit special. For example it is always running and cannot be terminated. Because of its + * special semantics we synthesize it here, instead of relying on the unit file on disk. */ + + u->default_dependencies = false; + u->ignore_on_isolate = true; + + if (!u->description) + u->description = strdup("Root Slice"); + if (!u->documentation) + u->documentation = strv_new("man:systemd.special(7)", NULL); + + return 1; +} + +static int slice_load(Unit *u) { + Slice *s = SLICE(u); + int r; + + assert(s); + assert(u->load_state == UNIT_STUB); + + r = slice_load_root_slice(u); + if (r < 0) + return r; + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED) { + + r = unit_patch_contexts(u); + if (r < 0) + return r; + + r = slice_add_parent_slice(s); + if (r < 0) + return r; + + r = slice_add_default_dependencies(s); + if (r < 0) + return r; + } + + return slice_verify(s); +} + +static int slice_coldplug(Unit *u) { + Slice *t = SLICE(u); + + assert(t); + assert(t->state == SLICE_DEAD); + + if (t->deserialized_state != t->state) + slice_set_state(t, t->deserialized_state); + + return 0; +} + +static void slice_dump(Unit *u, FILE *f, const char *prefix) { + Slice *t = SLICE(u); + + assert(t); + assert(f); + + fprintf(f, + "%sSlice State: %s\n", + prefix, slice_state_to_string(t->state)); + + cgroup_context_dump(&t->cgroup_context, f, prefix); +} + +static int slice_start(Unit *u) { + Slice *t = SLICE(u); + int r; + + assert(t); + assert(t->state == SLICE_DEAD); + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + (void) unit_realize_cgroup(u); + (void) unit_reset_cpu_usage(u); + + slice_set_state(t, SLICE_ACTIVE); + return 1; +} + +static int slice_stop(Unit *u) { + Slice *t = SLICE(u); + + assert(t); + assert(t->state == SLICE_ACTIVE); + + /* We do not need to destroy the cgroup explicitly, + * unit_notify() will do that for us anyway. */ + + slice_set_state(t, SLICE_DEAD); + return 1; +} + +static int slice_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, -1, error); +} + +static int slice_serialize(Unit *u, FILE *f, FDSet *fds) { + Slice *s = SLICE(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", slice_state_to_string(s->state)); + return 0; +} + +static int slice_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Slice *s = SLICE(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + SliceState state; + + state = slice_state_from_string(value); + if (state < 0) + log_debug("Failed to parse state value %s", value); + else + s->deserialized_state = state; + + } else + log_debug("Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState slice_active_state(Unit *u) { + assert(u); + + return state_translation_table[SLICE(u)->state]; +} + +_pure_ static const char *slice_sub_state_to_string(Unit *u) { + assert(u); + + return slice_state_to_string(SLICE(u)->state); +} + +static void slice_enumerate(Manager *m) { + Unit *u; + int r; + + assert(m); + + u = manager_get_unit(m, SPECIAL_ROOT_SLICE); + if (!u) { + r = unit_new_for_name(m, sizeof(Slice), SPECIAL_ROOT_SLICE, &u); + if (r < 0) { + log_error_errno(r, "Failed to allocate the special " SPECIAL_ROOT_SLICE " unit: %m"); + return; + } + } + + u->perpetual = true; + SLICE(u)->deserialized_state = SLICE_ACTIVE; + + unit_add_to_load_queue(u); + unit_add_to_dbus_queue(u); +} + +const UnitVTable slice_vtable = { + .object_size = sizeof(Slice), + .cgroup_context_offset = offsetof(Slice, cgroup_context), + + .sections = + "Unit\0" + "Slice\0" + "Install\0", + .private_section = "Slice", + + .can_transient = true, + + .init = slice_init, + .load = slice_load, + + .coldplug = slice_coldplug, + + .dump = slice_dump, + + .start = slice_start, + .stop = slice_stop, + + .kill = slice_kill, + + .serialize = slice_serialize, + .deserialize_item = slice_deserialize_item, + + .active_state = slice_active_state, + .sub_state_to_string = slice_sub_state_to_string, + + .bus_vtable = bus_slice_vtable, + .bus_set_property = bus_slice_set_property, + .bus_commit_properties = bus_slice_commit_properties, + + .enumerate = slice_enumerate, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Created slice %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Removed slice %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/smack-setup.c b/src/grp-system/libcore/src/smack-setup.c new file mode 100644 index 0000000000..5d94873419 --- /dev/null +++ b/src/grp-system/libcore/src/smack-setup.c @@ -0,0 +1,346 @@ +/*** + This file is part of systemd. + + Copyright (C) 2013 Intel Corporation + Authors: + Nathaniel Chen <nathaniel.chen@intel.com> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published + by the Free Software Foundation; either version 2.1 of the License, + or (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <dirent.h> +#include <errno.h> +#include <fcntl.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "core/smack-setup.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/dirent-util.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/fileio.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/util.h" + +#ifdef HAVE_SMACK + +static int write_access2_rules(const char* srcdir) { + _cleanup_close_ int load2_fd = -1, change_fd = -1; + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *entry; + char buf[NAME_MAX]; + int dfd = -1; + int r = 0; + + load2_fd = open("/sys/fs/smackfs/load2", O_RDWR|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); + if (load2_fd < 0) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open '/sys/fs/smackfs/load2': %m"); + return -errno; /* negative error */ + } + + change_fd = open("/sys/fs/smackfs/change-rule", O_RDWR|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); + if (change_fd < 0) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open '/sys/fs/smackfs/change-rule': %m"); + return -errno; /* negative error */ + } + + /* write rules to load2 or change-rule from every file in the directory */ + dir = opendir(srcdir); + if (!dir) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to opendir '%s': %m", srcdir); + return errno; /* positive on purpose */ + } + + dfd = dirfd(dir); + assert(dfd >= 0); + + FOREACH_DIRENT(entry, dir, return 0) { + int fd; + _cleanup_fclose_ FILE *policy = NULL; + + if (!dirent_is_file(entry)) + continue; + + fd = openat(dfd, entry->d_name, O_RDONLY|O_CLOEXEC); + if (fd < 0) { + if (r == 0) + r = -errno; + log_warning_errno(errno, "Failed to open '%s': %m", entry->d_name); + continue; + } + + policy = fdopen(fd, "re"); + if (!policy) { + if (r == 0) + r = -errno; + safe_close(fd); + log_error_errno(errno, "Failed to open '%s': %m", entry->d_name); + continue; + } + + /* load2 write rules in the kernel require a line buffered stream */ + FOREACH_LINE(buf, policy, + log_error_errno(errno, "Failed to read line from '%s': %m", + entry->d_name)) { + + _cleanup_free_ char *sbj = NULL, *obj = NULL, *acc1 = NULL, *acc2 = NULL; + + if (isempty(truncate_nl(buf))) + continue; + + /* if 3 args -> load rule : subject object access1 */ + /* if 4 args -> change rule : subject object access1 access2 */ + if (sscanf(buf, "%ms %ms %ms %ms", &sbj, &obj, &acc1, &acc2) < 3) { + log_error_errno(errno, "Failed to parse rule '%s' in '%s', ignoring.", buf, entry->d_name); + continue; + } + + if (write(isempty(acc2) ? load2_fd : change_fd, buf, strlen(buf)) < 0) { + if (r == 0) + r = -errno; + log_error_errno(errno, "Failed to write '%s' to '%s' in '%s'", + buf, isempty(acc2) ? "/sys/fs/smackfs/load2" : "/sys/fs/smackfs/change-rule", entry->d_name); + } + } + } + + return r; +} + +static int write_cipso2_rules(const char* srcdir) { + _cleanup_close_ int cipso2_fd = -1; + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *entry; + char buf[NAME_MAX]; + int dfd = -1; + int r = 0; + + cipso2_fd = open("/sys/fs/smackfs/cipso2", O_RDWR|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); + if (cipso2_fd < 0) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open '/sys/fs/smackfs/cipso2': %m"); + return -errno; /* negative error */ + } + + /* write rules to cipso2 from every file in the directory */ + dir = opendir(srcdir); + if (!dir) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to opendir '%s': %m", srcdir); + return errno; /* positive on purpose */ + } + + dfd = dirfd(dir); + assert(dfd >= 0); + + FOREACH_DIRENT(entry, dir, return 0) { + int fd; + _cleanup_fclose_ FILE *policy = NULL; + + if (!dirent_is_file(entry)) + continue; + + fd = openat(dfd, entry->d_name, O_RDONLY|O_CLOEXEC); + if (fd < 0) { + if (r == 0) + r = -errno; + log_error_errno(errno, "Failed to open '%s': %m", entry->d_name); + continue; + } + + policy = fdopen(fd, "re"); + if (!policy) { + if (r == 0) + r = -errno; + safe_close(fd); + log_error_errno(errno, "Failed to open '%s': %m", entry->d_name); + continue; + } + + /* cipso2 write rules in the kernel require a line buffered stream */ + FOREACH_LINE(buf, policy, + log_error_errno(errno, "Failed to read line from '%s': %m", + entry->d_name)) { + + if (isempty(truncate_nl(buf))) + continue; + + if (write(cipso2_fd, buf, strlen(buf)) < 0) { + if (r == 0) + r = -errno; + log_error_errno(errno, "Failed to write '%s' to '/sys/fs/smackfs/cipso2' in '%s'", + buf, entry->d_name); + break; + } + } + } + + return r; +} + +static int write_netlabel_rules(const char* srcdir) { + _cleanup_fclose_ FILE *dst = NULL; + _cleanup_closedir_ DIR *dir = NULL; + struct dirent *entry; + char buf[NAME_MAX]; + int dfd = -1; + int r = 0; + + dst = fopen("/sys/fs/smackfs/netlabel", "we"); + if (!dst) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to open /sys/fs/smackfs/netlabel: %m"); + return -errno; /* negative error */ + } + + /* write rules to dst from every file in the directory */ + dir = opendir(srcdir); + if (!dir) { + if (errno != ENOENT) + log_warning_errno(errno, "Failed to opendir %s: %m", srcdir); + return errno; /* positive on purpose */ + } + + dfd = dirfd(dir); + assert(dfd >= 0); + + FOREACH_DIRENT(entry, dir, return 0) { + int fd; + _cleanup_fclose_ FILE *policy = NULL; + + fd = openat(dfd, entry->d_name, O_RDONLY|O_CLOEXEC); + if (fd < 0) { + if (r == 0) + r = -errno; + log_warning_errno(errno, "Failed to open %s: %m", entry->d_name); + continue; + } + + policy = fdopen(fd, "re"); + if (!policy) { + if (r == 0) + r = -errno; + safe_close(fd); + log_error_errno(errno, "Failed to open %s: %m", entry->d_name); + continue; + } + + /* load2 write rules in the kernel require a line buffered stream */ + FOREACH_LINE(buf, policy, + log_error_errno(errno, "Failed to read line from %s: %m", + entry->d_name)) { + if (!fputs(buf, dst)) { + if (r == 0) + r = -EINVAL; + log_error_errno(errno, "Failed to write line to /sys/fs/smackfs/netlabel"); + break; + } + if (fflush(dst)) { + if (r == 0) + r = -errno; + log_error_errno(errno, "Failed to flush writes to /sys/fs/smackfs/netlabel: %m"); + break; + } + } + } + + return r; +} + +#endif + +int mac_smack_setup(bool *loaded_policy) { + +#ifdef HAVE_SMACK + + int r; + + assert(loaded_policy); + + r = write_access2_rules("/etc/smack/accesses.d/"); + switch(r) { + case -ENOENT: + log_debug("Smack is not enabled in the kernel."); + return 0; + case ENOENT: + log_debug("Smack access rules directory '/etc/smack/accesses.d/' not found"); + return 0; + case 0: + log_info("Successfully loaded Smack policies."); + break; + default: + log_warning_errno(r, "Failed to load Smack access rules, ignoring: %m"); + return 0; + } + +#ifdef SMACK_RUN_LABEL + r = write_string_file("/proc/self/attr/current", SMACK_RUN_LABEL, 0); + if (r < 0) + log_warning_errno(r, "Failed to set SMACK label \"" SMACK_RUN_LABEL "\" on self: %m"); + r = write_string_file("/sys/fs/smackfs/ambient", SMACK_RUN_LABEL, 0); + if (r < 0) + log_warning_errno(r, "Failed to set SMACK ambient label \"" SMACK_RUN_LABEL "\": %m"); + r = write_string_file("/sys/fs/smackfs/netlabel", + "0.0.0.0/0 " SMACK_RUN_LABEL, 0); + if (r < 0) + log_warning_errno(r, "Failed to set SMACK netlabel rule \"0.0.0.0/0 " SMACK_RUN_LABEL "\": %m"); + r = write_string_file("/sys/fs/smackfs/netlabel", "127.0.0.1 -CIPSO", 0); + if (r < 0) + log_warning_errno(r, "Failed to set SMACK netlabel rule \"127.0.0.1 -CIPSO\": %m"); +#endif + + r = write_cipso2_rules("/etc/smack/cipso.d/"); + switch(r) { + case -ENOENT: + log_debug("Smack/CIPSO is not enabled in the kernel."); + return 0; + case ENOENT: + log_debug("Smack/CIPSO access rules directory '/etc/smack/cipso.d/' not found"); + break; + case 0: + log_info("Successfully loaded Smack/CIPSO policies."); + break; + default: + log_warning_errno(r, "Failed to load Smack/CIPSO access rules, ignoring: %m"); + break; + } + + r = write_netlabel_rules("/etc/smack/netlabel.d/"); + switch(r) { + case -ENOENT: + log_debug("Smack/CIPSO is not enabled in the kernel."); + return 0; + case ENOENT: + log_debug("Smack network host rules directory '/etc/smack/netlabel.d/' not found"); + break; + case 0: + log_info("Successfully loaded Smack network host rules."); + break; + default: + log_warning_errno(r, "Failed to load Smack network host rules: %m, ignoring."); + break; + } + + *loaded_policy = true; + +#endif + + return 0; +} diff --git a/src/grp-system/libcore/src/socket.c b/src/grp-system/libcore/src/socket.c new file mode 100644 index 0000000000..9d7a72ef5f --- /dev/null +++ b/src/grp-system/libcore/src/socket.c @@ -0,0 +1,3136 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <arpa/inet.h> +#include <errno.h> +#include <fcntl.h> +#include <mqueue.h> +#include <netinet/tcp.h> +#include <signal.h> +#include <sys/epoll.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <linux/sctp.h> + +#include "core/socket.h" +#include "core/unit.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/copy.h" +#include "systemd-basic/def.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/in-addr-util.h" +#include "systemd-basic/io-util.h" +#include "systemd-basic/label.h" +#include "systemd-basic/log.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/selinux-util.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/smack-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" + +#include "dbus-socket.h" +#include "unit-printf.h" + +struct SocketPeer { + unsigned n_ref; + + Socket *socket; + union sockaddr_union peer; +}; + +static const UnitActiveState state_translation_table[_SOCKET_STATE_MAX] = { + [SOCKET_DEAD] = UNIT_INACTIVE, + [SOCKET_START_PRE] = UNIT_ACTIVATING, + [SOCKET_START_CHOWN] = UNIT_ACTIVATING, + [SOCKET_START_POST] = UNIT_ACTIVATING, + [SOCKET_LISTENING] = UNIT_ACTIVE, + [SOCKET_RUNNING] = UNIT_ACTIVE, + [SOCKET_STOP_PRE] = UNIT_DEACTIVATING, + [SOCKET_STOP_PRE_SIGTERM] = UNIT_DEACTIVATING, + [SOCKET_STOP_PRE_SIGKILL] = UNIT_DEACTIVATING, + [SOCKET_STOP_POST] = UNIT_DEACTIVATING, + [SOCKET_FINAL_SIGTERM] = UNIT_DEACTIVATING, + [SOCKET_FINAL_SIGKILL] = UNIT_DEACTIVATING, + [SOCKET_FAILED] = UNIT_FAILED +}; + +static int socket_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); +static int socket_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); + +static void socket_init(Unit *u) { + Socket *s = SOCKET(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + s->backlog = SOMAXCONN; + s->timeout_usec = u->manager->default_timeout_start_usec; + s->directory_mode = 0755; + s->socket_mode = 0666; + + s->max_connections = 64; + + s->priority = -1; + s->ip_tos = -1; + s->ip_ttl = -1; + s->mark = -1; + + s->exec_context.std_output = u->manager->default_std_output; + s->exec_context.std_error = u->manager->default_std_error; + + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; + + s->trigger_limit.interval = USEC_INFINITY; + s->trigger_limit.burst = (unsigned) -1; +} + +static void socket_unwatch_control_pid(Socket *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +static void socket_cleanup_fd_list(SocketPort *p) { + assert(p); + + close_many(p->auxiliary_fds, p->n_auxiliary_fds); + p->auxiliary_fds = mfree(p->auxiliary_fds); + p->n_auxiliary_fds = 0; +} + +void socket_free_ports(Socket *s) { + SocketPort *p; + + assert(s); + + while ((p = s->ports)) { + LIST_REMOVE(port, s->ports, p); + + sd_event_source_unref(p->event_source); + + socket_cleanup_fd_list(p); + safe_close(p->fd); + free(p->path); + free(p); + } +} + +static void socket_done(Unit *u) { + Socket *s = SOCKET(u); + SocketPeer *p; + + assert(s); + + socket_free_ports(s); + + while ((p = set_steal_first(s->peers_by_address))) + p->socket = NULL; + + s->peers_by_address = set_free(s->peers_by_address); + + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + exec_command_free_array(s->exec_command, _SOCKET_EXEC_COMMAND_MAX); + s->control_command = NULL; + + dynamic_creds_unref(&s->dynamic_creds); + + socket_unwatch_control_pid(s); + + unit_ref_unset(&s->service); + + s->tcp_congestion = mfree(s->tcp_congestion); + s->bind_to_device = mfree(s->bind_to_device); + + s->smack = mfree(s->smack); + s->smack_ip_in = mfree(s->smack_ip_in); + s->smack_ip_out = mfree(s->smack_ip_out); + + strv_free(s->symlinks); + + s->user = mfree(s->user); + s->group = mfree(s->group); + + s->fdname = mfree(s->fdname); + + s->timer_event_source = sd_event_source_unref(s->timer_event_source); +} + +static int socket_arm_timer(Socket *s, usec_t usec) { + int r; + + assert(s); + + if (s->timer_event_source) { + r = sd_event_source_set_time(s->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(s->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + socket_dispatch_timer, s); + if (r < 0) + return r; + + (void) sd_event_source_set_description(s->timer_event_source, "socket-timer"); + + return 0; +} + +int socket_instantiate_service(Socket *s) { + _cleanup_free_ char *prefix = NULL, *name = NULL; + int r; + Unit *u; + + assert(s); + + /* This fills in s->service if it isn't filled in yet. For + * Accept=yes sockets we create the next connection service + * here. For Accept=no this is mostly a NOP since the service + * is figured out at load time anyway. */ + + if (UNIT_DEREF(s->service)) + return 0; + + if (!s->accept) + return 0; + + r = unit_name_to_prefix(UNIT(s)->id, &prefix); + if (r < 0) + return r; + + if (asprintf(&name, "%s@%u.service", prefix, s->n_accepted) < 0) + return -ENOMEM; + + r = manager_load_unit(UNIT(s)->manager, name, NULL, NULL, &u); + if (r < 0) + return r; + + unit_ref_set(&s->service, u); + + return unit_add_two_dependencies(UNIT(s), UNIT_BEFORE, UNIT_TRIGGERS, u, false); +} + +static bool have_non_accept_socket(Socket *s) { + SocketPort *p; + + assert(s); + + if (!s->accept) + return true; + + LIST_FOREACH(port, p, s->ports) { + + if (p->type != SOCKET_SOCKET) + return true; + + if (!socket_address_can_accept(&p->address)) + return true; + } + + return false; +} + +static int socket_add_mount_links(Socket *s) { + SocketPort *p; + int r; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + const char *path = NULL; + + if (p->type == SOCKET_SOCKET) + path = socket_address_get_path(&p->address); + else if (IN_SET(p->type, SOCKET_FIFO, SOCKET_SPECIAL, SOCKET_USB_FUNCTION)) + path = p->path; + + if (!path) + continue; + + r = unit_require_mounts_for(UNIT(s), path); + if (r < 0) + return r; + } + + return 0; +} + +static int socket_add_device_link(Socket *s) { + char *t; + + assert(s); + + if (!s->bind_to_device || streq(s->bind_to_device, "lo")) + return 0; + + t = strjoina("/sys/subsystem/net/devices/", s->bind_to_device); + return unit_add_node_link(UNIT(s), t, false, UNIT_BINDS_TO); +} + +static int socket_add_default_dependencies(Socket *s) { + int r; + assert(s); + + if (!UNIT(s)->default_dependencies) + return 0; + + r = unit_add_dependency_by_name(UNIT(s), UNIT_BEFORE, SPECIAL_SOCKETS_TARGET, NULL, true); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(UNIT(s)->manager)) { + r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + } + + return unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +_pure_ static bool socket_has_exec(Socket *s) { + unsigned i; + assert(s); + + for (i = 0; i < _SOCKET_EXEC_COMMAND_MAX; i++) + if (s->exec_command[i]) + return true; + + return false; +} + +static int socket_add_extras(Socket *s) { + Unit *u = UNIT(s); + int r; + + assert(s); + + /* Pick defaults for the trigger limit, if nothing was explicitly configured. We pick a relatively high limit + * in Accept=yes mode, and a lower limit for Accept=no. Reason: in Accept=yes mode we are invoking accept() + * ourselves before the trigger limit can hit, thus incoming connections are taken off the socket queue quickly + * and reliably. This is different for Accept=no, where the spawned service has to take the incoming traffic + * off the queues, which it might not necessarily do. Moreover, while Accept=no services are supposed to + * process whatever is queued in one go, and thus should normally never have to be started frequently. This is + * different for Accept=yes where each connection is processed by a new service instance, and thus frequent + * service starts are typical. */ + + if (s->trigger_limit.interval == USEC_INFINITY) + s->trigger_limit.interval = 2 * USEC_PER_SEC; + + if (s->trigger_limit.burst == (unsigned) -1) { + if (s->accept) + s->trigger_limit.burst = 200; + else + s->trigger_limit.burst = 20; + } + + if (have_non_accept_socket(s)) { + + if (!UNIT_DEREF(s->service)) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + unit_ref_set(&s->service, x); + } + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, UNIT_DEREF(s->service), true); + if (r < 0) + return r; + } + + r = socket_add_mount_links(s); + if (r < 0) + return r; + + r = socket_add_device_link(s); + if (r < 0) + return r; + + r = unit_patch_contexts(u); + if (r < 0) + return r; + + if (socket_has_exec(s)) { + r = unit_add_exec_dependencies(u, &s->exec_context); + if (r < 0) + return r; + + r = unit_set_default_slice(u); + if (r < 0) + return r; + } + + r = socket_add_default_dependencies(s); + if (r < 0) + return r; + + return 0; +} + +static const char *socket_find_symlink_target(Socket *s) { + const char *found = NULL; + SocketPort *p; + + LIST_FOREACH(port, p, s->ports) { + const char *f = NULL; + + switch (p->type) { + + case SOCKET_FIFO: + f = p->path; + break; + + case SOCKET_SOCKET: + if (p->address.sockaddr.un.sun_path[0] != 0) + f = p->address.sockaddr.un.sun_path; + break; + + default: + break; + } + + if (f) { + if (found) + return NULL; + + found = f; + } + } + + return found; +} + +static int socket_verify(Socket *s) { + assert(s); + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + if (!s->ports) { + log_unit_error(UNIT(s), "Unit lacks Listen setting. Refusing."); + return -EINVAL; + } + + if (s->accept && have_non_accept_socket(s)) { + log_unit_error(UNIT(s), "Unit configured for accepting sockets, but sockets are non-accepting. Refusing."); + return -EINVAL; + } + + if (s->accept && s->max_connections <= 0) { + log_unit_error(UNIT(s), "MaxConnection= setting too small. Refusing."); + return -EINVAL; + } + + if (s->accept && UNIT_DEREF(s->service)) { + log_unit_error(UNIT(s), "Explicit service configuration for accepting socket units not supported. Refusing."); + return -EINVAL; + } + + if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_unit_error(UNIT(s), "Unit has PAM enabled. Kill mode must be set to 'control-group'. Refusing."); + return -EINVAL; + } + + if (!strv_isempty(s->symlinks) && !socket_find_symlink_target(s)) { + log_unit_error(UNIT(s), "Unit has symlinks set but none or more than one node in the file system. Refusing."); + return -EINVAL; + } + + return 0; +} + +static void peer_address_hash_func(const void *p, struct siphash *state) { + const SocketPeer *s = p; + + assert(s); + assert(IN_SET(s->peer.sa.sa_family, AF_INET, AF_INET6)); + + if (s->peer.sa.sa_family == AF_INET) + siphash24_compress(&s->peer.in.sin_addr, sizeof(s->peer.in.sin_addr), state); + else + siphash24_compress(&s->peer.in6.sin6_addr, sizeof(s->peer.in6.sin6_addr), state); +} + +static int peer_address_compare_func(const void *a, const void *b) { + const SocketPeer *x = a, *y = b; + + if (x->peer.sa.sa_family < y->peer.sa.sa_family) + return -1; + if (x->peer.sa.sa_family > y->peer.sa.sa_family) + return 1; + + switch(x->peer.sa.sa_family) { + case AF_INET: + return memcmp(&x->peer.in.sin_addr, &y->peer.in.sin_addr, sizeof(x->peer.in.sin_addr)); + case AF_INET6: + return memcmp(&x->peer.in6.sin6_addr, &y->peer.in6.sin6_addr, sizeof(x->peer.in6.sin6_addr)); + } + assert_not_reached("Black sheep in the family!"); +} + +const struct hash_ops peer_address_hash_ops = { + .hash = peer_address_hash_func, + .compare = peer_address_compare_func +}; + +static int socket_load(Unit *u) { + Socket *s = SOCKET(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = set_ensure_allocated(&s->peers_by_address, &peer_address_hash_ops); + if (r < 0) + return r; + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + /* This is a new unit? Then let's add in some extras */ + r = socket_add_extras(s); + if (r < 0) + return r; + } + + return socket_verify(s); +} + +static SocketPeer *socket_peer_new(void) { + SocketPeer *p; + + p = new0(SocketPeer, 1); + if (!p) + return NULL; + + p->n_ref = 1; + + return p; +} + +SocketPeer *socket_peer_ref(SocketPeer *p) { + if (!p) + return NULL; + + assert(p->n_ref > 0); + p->n_ref++; + + return p; +} + +SocketPeer *socket_peer_unref(SocketPeer *p) { + if (!p) + return NULL; + + assert(p->n_ref > 0); + + p->n_ref--; + + if (p->n_ref > 0) + return NULL; + + if (p->socket) + set_remove(p->socket->peers_by_address, p); + + return mfree(p); +} + +int socket_acquire_peer(Socket *s, int fd, SocketPeer **p) { + _cleanup_(socket_peer_unrefp) SocketPeer *remote = NULL; + SocketPeer sa = {}, *i; + socklen_t salen = sizeof(sa.peer); + int r; + + assert(fd >= 0); + assert(s); + + r = getpeername(fd, &sa.peer.sa, &salen); + if (r < 0) + return log_error_errno(errno, "getpeername failed: %m"); + + if (!IN_SET(sa.peer.sa.sa_family, AF_INET, AF_INET6)) { + *p = NULL; + return 0; + } + + i = set_get(s->peers_by_address, &sa); + if (i) { + *p = socket_peer_ref(i); + return 1; + } + + remote = socket_peer_new(); + if (!remote) + return log_oom(); + + remote->peer = sa.peer; + + r = set_put(s->peers_by_address, remote); + if (r < 0) + return r; + + remote->socket = s; + + *p = remote; + remote = NULL; + + return 1; +} + +_const_ static const char* listen_lookup(int family, int type) { + + if (family == AF_NETLINK) + return "ListenNetlink"; + + if (type == SOCK_STREAM) + return "ListenStream"; + else if (type == SOCK_DGRAM) + return "ListenDatagram"; + else if (type == SOCK_SEQPACKET) + return "ListenSequentialPacket"; + + assert_not_reached("Unknown socket type"); + return NULL; +} + +static void socket_dump(Unit *u, FILE *f, const char *prefix) { + char time_string[FORMAT_TIMESPAN_MAX]; + SocketExecCommand c; + Socket *s = SOCKET(u); + SocketPort *p; + const char *prefix2; + + assert(s); + assert(f); + + prefix = strempty(prefix); + prefix2 = strjoina(prefix, "\t"); + + fprintf(f, + "%sSocket State: %s\n" + "%sResult: %s\n" + "%sBindIPv6Only: %s\n" + "%sBacklog: %u\n" + "%sSocketMode: %04o\n" + "%sDirectoryMode: %04o\n" + "%sKeepAlive: %s\n" + "%sNoDelay: %s\n" + "%sFreeBind: %s\n" + "%sTransparent: %s\n" + "%sBroadcast: %s\n" + "%sPassCredentials: %s\n" + "%sPassSecurity: %s\n" + "%sTCPCongestion: %s\n" + "%sRemoveOnStop: %s\n" + "%sWritable: %s\n" + "%sFDName: %s\n" + "%sSELinuxContextFromNet: %s\n", + prefix, socket_state_to_string(s->state), + prefix, socket_result_to_string(s->result), + prefix, socket_address_bind_ipv6_only_to_string(s->bind_ipv6_only), + prefix, s->backlog, + prefix, s->socket_mode, + prefix, s->directory_mode, + prefix, yes_no(s->keep_alive), + prefix, yes_no(s->no_delay), + prefix, yes_no(s->free_bind), + prefix, yes_no(s->transparent), + prefix, yes_no(s->broadcast), + prefix, yes_no(s->pass_cred), + prefix, yes_no(s->pass_sec), + prefix, strna(s->tcp_congestion), + prefix, yes_no(s->remove_on_stop), + prefix, yes_no(s->writable), + prefix, socket_fdname(s), + prefix, yes_no(s->selinux_context_from_net)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: "PID_FMT"\n", + prefix, s->control_pid); + + if (s->bind_to_device) + fprintf(f, + "%sBindToDevice: %s\n", + prefix, s->bind_to_device); + + if (s->accept) + fprintf(f, + "%sAccepted: %u\n" + "%sNConnections: %u\n" + "%sMaxConnections: %u\n", + prefix, s->n_accepted, + prefix, s->n_connections, + prefix, s->max_connections); + + if (s->priority >= 0) + fprintf(f, + "%sPriority: %i\n", + prefix, s->priority); + + if (s->receive_buffer > 0) + fprintf(f, + "%sReceiveBuffer: %zu\n", + prefix, s->receive_buffer); + + if (s->send_buffer > 0) + fprintf(f, + "%sSendBuffer: %zu\n", + prefix, s->send_buffer); + + if (s->ip_tos >= 0) + fprintf(f, + "%sIPTOS: %i\n", + prefix, s->ip_tos); + + if (s->ip_ttl >= 0) + fprintf(f, + "%sIPTTL: %i\n", + prefix, s->ip_ttl); + + if (s->pipe_size > 0) + fprintf(f, + "%sPipeSize: %zu\n", + prefix, s->pipe_size); + + if (s->mark >= 0) + fprintf(f, + "%sMark: %i\n", + prefix, s->mark); + + if (s->mq_maxmsg > 0) + fprintf(f, + "%sMessageQueueMaxMessages: %li\n", + prefix, s->mq_maxmsg); + + if (s->mq_msgsize > 0) + fprintf(f, + "%sMessageQueueMessageSize: %li\n", + prefix, s->mq_msgsize); + + if (s->reuse_port) + fprintf(f, + "%sReusePort: %s\n", + prefix, yes_no(s->reuse_port)); + + if (s->smack) + fprintf(f, + "%sSmackLabel: %s\n", + prefix, s->smack); + + if (s->smack_ip_in) + fprintf(f, + "%sSmackLabelIPIn: %s\n", + prefix, s->smack_ip_in); + + if (s->smack_ip_out) + fprintf(f, + "%sSmackLabelIPOut: %s\n", + prefix, s->smack_ip_out); + + if (!isempty(s->user) || !isempty(s->group)) + fprintf(f, + "%sSocketUser: %s\n" + "%sSocketGroup: %s\n", + prefix, strna(s->user), + prefix, strna(s->group)); + + if (s->keep_alive_time > 0) + fprintf(f, + "%sKeepAliveTimeSec: %s\n", + prefix, format_timespan(time_string, FORMAT_TIMESPAN_MAX, s->keep_alive_time, USEC_PER_SEC)); + + if (s->keep_alive_interval) + fprintf(f, + "%sKeepAliveIntervalSec: %s\n", + prefix, format_timespan(time_string, FORMAT_TIMESPAN_MAX, s->keep_alive_interval, USEC_PER_SEC)); + + if (s->keep_alive_cnt) + fprintf(f, + "%sKeepAliveProbes: %u\n", + prefix, s->keep_alive_cnt); + + if (s->defer_accept) + fprintf(f, + "%sDeferAcceptSec: %s\n", + prefix, format_timespan(time_string, FORMAT_TIMESPAN_MAX, s->defer_accept, USEC_PER_SEC)); + + LIST_FOREACH(port, p, s->ports) { + + if (p->type == SOCKET_SOCKET) { + const char *t; + int r; + char *k = NULL; + + r = socket_address_print(&p->address, &k); + if (r < 0) + t = strerror(-r); + else + t = k; + + fprintf(f, "%s%s: %s\n", prefix, listen_lookup(socket_address_family(&p->address), p->address.type), t); + free(k); + } else if (p->type == SOCKET_SPECIAL) + fprintf(f, "%sListenSpecial: %s\n", prefix, p->path); + else if (p->type == SOCKET_USB_FUNCTION) + fprintf(f, "%sListenUSBFunction: %s\n", prefix, p->path); + else if (p->type == SOCKET_MQUEUE) + fprintf(f, "%sListenMessageQueue: %s\n", prefix, p->path); + else + fprintf(f, "%sListenFIFO: %s\n", prefix, p->path); + } + + fprintf(f, + "%sTriggerLimitIntervalSec: %s\n" + "%sTriggerLimitBurst: %u\n", + prefix, format_timespan(time_string, FORMAT_TIMESPAN_MAX, s->trigger_limit.interval, USEC_PER_SEC), + prefix, s->trigger_limit.burst); + + exec_context_dump(&s->exec_context, f, prefix); + kill_context_dump(&s->kill_context, f, prefix); + + for (c = 0; c < _SOCKET_EXEC_COMMAND_MAX; c++) { + if (!s->exec_command[c]) + continue; + + fprintf(f, "%s-> %s:\n", + prefix, socket_exec_command_to_string(c)); + + exec_command_dump_list(s->exec_command[c], f, prefix2); + } +} + +static int instance_from_socket(int fd, unsigned nr, char **instance) { + socklen_t l; + char *r; + union sockaddr_union local, remote; + + assert(fd >= 0); + assert(instance); + + l = sizeof(local); + if (getsockname(fd, &local.sa, &l) < 0) + return -errno; + + l = sizeof(remote); + if (getpeername(fd, &remote.sa, &l) < 0) + return -errno; + + switch (local.sa.sa_family) { + + case AF_INET: { + uint32_t + a = be32toh(local.in.sin_addr.s_addr), + b = be32toh(remote.in.sin_addr.s_addr); + + if (asprintf(&r, + "%u-%u.%u.%u.%u:%u-%u.%u.%u.%u:%u", + nr, + a >> 24, (a >> 16) & 0xFF, (a >> 8) & 0xFF, a & 0xFF, + be16toh(local.in.sin_port), + b >> 24, (b >> 16) & 0xFF, (b >> 8) & 0xFF, b & 0xFF, + be16toh(remote.in.sin_port)) < 0) + return -ENOMEM; + + break; + } + + case AF_INET6: { + static const unsigned char ipv4_prefix[] = { + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xFF, 0xFF + }; + + if (memcmp(&local.in6.sin6_addr, ipv4_prefix, sizeof(ipv4_prefix)) == 0 && + memcmp(&remote.in6.sin6_addr, ipv4_prefix, sizeof(ipv4_prefix)) == 0) { + const uint8_t + *a = local.in6.sin6_addr.s6_addr+12, + *b = remote.in6.sin6_addr.s6_addr+12; + + if (asprintf(&r, + "%u-%u.%u.%u.%u:%u-%u.%u.%u.%u:%u", + nr, + a[0], a[1], a[2], a[3], + be16toh(local.in6.sin6_port), + b[0], b[1], b[2], b[3], + be16toh(remote.in6.sin6_port)) < 0) + return -ENOMEM; + } else { + char a[INET6_ADDRSTRLEN], b[INET6_ADDRSTRLEN]; + + if (asprintf(&r, + "%u-%s:%u-%s:%u", + nr, + inet_ntop(AF_INET6, &local.in6.sin6_addr, a, sizeof(a)), + be16toh(local.in6.sin6_port), + inet_ntop(AF_INET6, &remote.in6.sin6_addr, b, sizeof(b)), + be16toh(remote.in6.sin6_port)) < 0) + return -ENOMEM; + } + + break; + } + + case AF_UNIX: { + struct ucred ucred; + int k; + + k = getpeercred(fd, &ucred); + if (k >= 0) { + if (asprintf(&r, + "%u-"PID_FMT"-"UID_FMT, + nr, ucred.pid, ucred.uid) < 0) + return -ENOMEM; + } else if (k == -ENODATA) { + /* This handles the case where somebody is + * connecting from another pid/uid namespace + * (e.g. from outside of our container). */ + if (asprintf(&r, + "%u-unknown", + nr) < 0) + return -ENOMEM; + } else + return k; + + break; + } + + default: + assert_not_reached("Unhandled socket type."); + } + + *instance = r; + return 0; +} + +static void socket_close_fds(Socket *s) { + SocketPort *p; + char **i; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + bool was_open; + + was_open = p->fd >= 0; + + p->event_source = sd_event_source_unref(p->event_source); + p->fd = safe_close(p->fd); + socket_cleanup_fd_list(p); + + /* One little note: we should normally not delete any sockets in the file system here! After all some + * other process we spawned might still have a reference of this fd and wants to continue to use + * it. Therefore we normally delete sockets in the file system before we create a new one, not after we + * stopped using one! That all said, if the user explicitly requested this, we'll delete them here + * anyway, but only then. */ + + if (!was_open || !s->remove_on_stop) + continue; + + switch (p->type) { + + case SOCKET_FIFO: + (void) unlink(p->path); + break; + + case SOCKET_MQUEUE: + (void) mq_unlink(p->path); + break; + + case SOCKET_SOCKET: + (void) socket_address_unlink(&p->address); + break; + + default: + break; + } + } + + if (s->remove_on_stop) + STRV_FOREACH(i, s->symlinks) + (void) unlink(*i); +} + +static void socket_apply_socket_options(Socket *s, int fd) { + int r; + + assert(s); + assert(fd >= 0); + + if (s->keep_alive) { + int b = s->keep_alive; + if (setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &b, sizeof(b)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_KEEPALIVE failed: %m"); + } + + if (s->keep_alive_time) { + int value = s->keep_alive_time / USEC_PER_SEC; + if (setsockopt(fd, SOL_TCP, TCP_KEEPIDLE, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_KEEPIDLE failed: %m"); + } + + if (s->keep_alive_interval) { + int value = s->keep_alive_interval / USEC_PER_SEC; + if (setsockopt(fd, SOL_TCP, TCP_KEEPINTVL, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_KEEPINTVL failed: %m"); + } + + if (s->keep_alive_cnt) { + int value = s->keep_alive_cnt; + if (setsockopt(fd, SOL_TCP, TCP_KEEPCNT, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_KEEPCNT failed: %m"); + } + + if (s->defer_accept) { + int value = s->defer_accept / USEC_PER_SEC; + if (setsockopt(fd, SOL_TCP, TCP_DEFER_ACCEPT, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_DEFER_ACCEPT failed: %m"); + } + + if (s->no_delay) { + int b = s->no_delay; + + if (s->socket_protocol == IPPROTO_SCTP) { + if (setsockopt(fd, SOL_SCTP, SCTP_NODELAY, &b, sizeof(b)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SCTP_NODELAY failed: %m"); + } else { + if (setsockopt(fd, SOL_TCP, TCP_NODELAY, &b, sizeof(b)) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_NODELAY failed: %m"); + } + } + + if (s->broadcast) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST, &one, sizeof(one)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_BROADCAST failed: %m"); + } + + if (s->pass_cred) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_PASSCRED failed: %m"); + } + + if (s->pass_sec) { + int one = 1; + if (setsockopt(fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_PASSSEC failed: %m"); + } + + if (s->priority >= 0) + if (setsockopt(fd, SOL_SOCKET, SO_PRIORITY, &s->priority, sizeof(s->priority)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_PRIORITY failed: %m"); + + if (s->receive_buffer > 0) { + int value = (int) s->receive_buffer; + + /* We first try with SO_RCVBUFFORCE, in case we have the perms for that */ + + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &value, sizeof(value)) < 0) + if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_RCVBUF failed: %m"); + } + + if (s->send_buffer > 0) { + int value = (int) s->send_buffer; + if (setsockopt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &value, sizeof(value)) < 0) + if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &value, sizeof(value)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_SNDBUF failed: %m"); + } + + if (s->mark >= 0) + if (setsockopt(fd, SOL_SOCKET, SO_MARK, &s->mark, sizeof(s->mark)) < 0) + log_unit_warning_errno(UNIT(s), errno, "SO_MARK failed: %m"); + + if (s->ip_tos >= 0) + if (setsockopt(fd, IPPROTO_IP, IP_TOS, &s->ip_tos, sizeof(s->ip_tos)) < 0) + log_unit_warning_errno(UNIT(s), errno, "IP_TOS failed: %m"); + + if (s->ip_ttl >= 0) { + int x; + + r = setsockopt(fd, IPPROTO_IP, IP_TTL, &s->ip_ttl, sizeof(s->ip_ttl)); + + if (socket_ipv6_is_supported()) + x = setsockopt(fd, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &s->ip_ttl, sizeof(s->ip_ttl)); + else { + x = -1; + errno = EAFNOSUPPORT; + } + + if (r < 0 && x < 0) + log_unit_warning_errno(UNIT(s), errno, "IP_TTL/IPV6_UNICAST_HOPS failed: %m"); + } + + if (s->tcp_congestion) + if (setsockopt(fd, SOL_TCP, TCP_CONGESTION, s->tcp_congestion, strlen(s->tcp_congestion)+1) < 0) + log_unit_warning_errno(UNIT(s), errno, "TCP_CONGESTION failed: %m"); + + if (s->smack_ip_in) { + r = mac_smack_apply_fd(fd, SMACK_ATTR_IPIN, s->smack_ip_in); + if (r < 0) + log_unit_error_errno(UNIT(s), r, "mac_smack_apply_ip_in_fd: %m"); + } + + if (s->smack_ip_out) { + r = mac_smack_apply_fd(fd, SMACK_ATTR_IPOUT, s->smack_ip_out); + if (r < 0) + log_unit_error_errno(UNIT(s), r, "mac_smack_apply_ip_out_fd: %m"); + } +} + +static void socket_apply_fifo_options(Socket *s, int fd) { + int r; + + assert(s); + assert(fd >= 0); + + if (s->pipe_size > 0) + if (fcntl(fd, F_SETPIPE_SZ, s->pipe_size) < 0) + log_unit_warning_errno(UNIT(s), errno, "Setting pipe size failed, ignoring: %m"); + + if (s->smack) { + r = mac_smack_apply_fd(fd, SMACK_ATTR_ACCESS, s->smack); + if (r < 0) + log_unit_error_errno(UNIT(s), r, "SMACK relabelling failed, ignoring: %m"); + } +} + +static int fifo_address_create( + const char *path, + mode_t directory_mode, + mode_t socket_mode) { + + _cleanup_close_ int fd = -1; + mode_t old_mask; + struct stat st; + int r; + + assert(path); + + mkdir_parents_label(path, directory_mode); + + r = mac_selinux_create_file_prepare(path, S_IFIFO); + if (r < 0) + return r; + + /* Enforce the right access mode for the fifo */ + old_mask = umask(~ socket_mode); + + /* Include the original umask in our mask */ + (void) umask(~socket_mode | old_mask); + + r = mkfifo(path, socket_mode); + (void) umask(old_mask); + + if (r < 0 && errno != EEXIST) { + r = -errno; + goto fail; + } + + fd = open(path, O_RDWR | O_CLOEXEC | O_NOCTTY | O_NONBLOCK | O_NOFOLLOW); + if (fd < 0) { + r = -errno; + goto fail; + } + + mac_selinux_create_file_clear(); + + if (fstat(fd, &st) < 0) { + r = -errno; + goto fail; + } + + if (!S_ISFIFO(st.st_mode) || + (st.st_mode & 0777) != (socket_mode & ~old_mask) || + st.st_uid != getuid() || + st.st_gid != getgid()) { + r = -EEXIST; + goto fail; + } + + r = fd; + fd = -1; + + return r; + +fail: + mac_selinux_create_file_clear(); + return r; +} + +static int special_address_create(const char *path, bool writable) { + _cleanup_close_ int fd = -1; + struct stat st; + int r; + + assert(path); + + fd = open(path, (writable ? O_RDWR : O_RDONLY)|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW); + if (fd < 0) + return -errno; + + if (fstat(fd, &st) < 0) + return -errno; + + /* Check whether this is a /proc, /sys or /dev file or char device */ + if (!S_ISREG(st.st_mode) && !S_ISCHR(st.st_mode)) + return -EEXIST; + + r = fd; + fd = -1; + + return r; +} + +static int usbffs_address_create(const char *path) { + _cleanup_close_ int fd = -1; + struct stat st; + int r; + + assert(path); + + fd = open(path, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW); + if (fd < 0) + return -errno; + + if (fstat(fd, &st) < 0) + return -errno; + + /* Check whether this is a regular file (ffs endpoint)*/ + if (!S_ISREG(st.st_mode)) + return -EEXIST; + + r = fd; + fd = -1; + + return r; +} + +static int mq_address_create( + const char *path, + mode_t mq_mode, + long maxmsg, + long msgsize) { + + _cleanup_close_ int fd = -1; + struct stat st; + mode_t old_mask; + struct mq_attr _attr, *attr = NULL; + int r; + + assert(path); + + if (maxmsg > 0 && msgsize > 0) { + _attr = (struct mq_attr) { + .mq_flags = O_NONBLOCK, + .mq_maxmsg = maxmsg, + .mq_msgsize = msgsize, + }; + attr = &_attr; + } + + /* Enforce the right access mode for the mq */ + old_mask = umask(~ mq_mode); + + /* Include the original umask in our mask */ + (void) umask(~mq_mode | old_mask); + fd = mq_open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_CREAT, mq_mode, attr); + (void) umask(old_mask); + + if (fd < 0) + return -errno; + + if (fstat(fd, &st) < 0) + return -errno; + + if ((st.st_mode & 0777) != (mq_mode & ~old_mask) || + st.st_uid != getuid() || + st.st_gid != getgid()) + return -EEXIST; + + r = fd; + fd = -1; + + return r; +} + +static int socket_symlink(Socket *s) { + const char *p; + char **i; + + assert(s); + + p = socket_find_symlink_target(s); + if (!p) + return 0; + + STRV_FOREACH(i, s->symlinks) + symlink_label(p, *i); + + return 0; +} + +static int usbffs_write_descs(int fd, Service *s) { + int r; + + if (!s->usb_function_descriptors || !s->usb_function_strings) + return -EINVAL; + + r = copy_file_fd(s->usb_function_descriptors, fd, false); + if (r < 0) + return r; + + return copy_file_fd(s->usb_function_strings, fd, false); +} + +static int usbffs_select_ep(const struct dirent *d) { + return d->d_name[0] != '.' && !streq(d->d_name, "ep0"); +} + +static int usbffs_dispatch_eps(SocketPort *p) { + _cleanup_free_ struct dirent **ent = NULL; + int r, i, n, k; + + r = scandir(p->path, &ent, usbffs_select_ep, alphasort); + if (r < 0) + return -errno; + + n = r; + p->auxiliary_fds = new(int, n); + if (!p->auxiliary_fds) + return -ENOMEM; + + p->n_auxiliary_fds = n; + + k = 0; + for (i = 0; i < n; ++i) { + _cleanup_free_ char *ep = NULL; + + ep = path_make_absolute(ent[i]->d_name, p->path); + if (!ep) + return -ENOMEM; + + path_kill_slashes(ep); + + r = usbffs_address_create(ep); + if (r < 0) + goto fail; + + p->auxiliary_fds[k] = r; + + ++k; + free(ent[i]); + } + + return r; + +fail: + close_many(p->auxiliary_fds, k); + p->auxiliary_fds = mfree(p->auxiliary_fds); + p->n_auxiliary_fds = 0; + + return r; +} + +static int socket_determine_selinux_label(Socket *s, char **ret) { + ExecCommand *c; + int r; + + assert(s); + assert(ret); + + if (s->selinux_context_from_net) { + /* If this is requested, get label from the network label */ + + r = mac_selinux_get_our_label(ret); + if (r == -EOPNOTSUPP) + goto no_label; + + } else { + /* Otherwise, get it from the executable we are about to start */ + r = socket_instantiate_service(s); + if (r < 0) + return r; + + if (!UNIT_ISSET(s->service)) + goto no_label; + + c = SERVICE(UNIT_DEREF(s->service))->exec_command[SERVICE_EXEC_START]; + if (!c) + goto no_label; + + r = mac_selinux_get_create_label_from_exe(c->path, ret); + if (r == -EPERM || r == -EOPNOTSUPP) + goto no_label; + } + + return r; + +no_label: + *ret = NULL; + return 0; +} + +static int socket_open_fds(Socket *s) { + _cleanup_(mac_selinux_freep) char *label = NULL; + bool know_label = false; + SocketPort *p; + int r; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + + if (p->fd >= 0) + continue; + + switch (p->type) { + + case SOCKET_SOCKET: + + if (!know_label) { + /* Figure out label, if we don't it know yet. We do it once, for the first socket where + * we need this and remember it for the rest. */ + + r = socket_determine_selinux_label(s, &label); + if (r < 0) + goto rollback; + + know_label = true; + } + + /* Apply the socket protocol */ + switch (p->address.type) { + + case SOCK_STREAM: + case SOCK_SEQPACKET: + if (s->socket_protocol == IPPROTO_SCTP) + p->address.protocol = s->socket_protocol; + break; + + case SOCK_DGRAM: + if (s->socket_protocol == IPPROTO_UDPLITE) + p->address.protocol = s->socket_protocol; + break; + } + + r = socket_address_listen( + &p->address, + SOCK_CLOEXEC|SOCK_NONBLOCK, + s->backlog, + s->bind_ipv6_only, + s->bind_to_device, + s->reuse_port, + s->free_bind, + s->transparent, + s->directory_mode, + s->socket_mode, + label); + if (r < 0) + goto rollback; + + p->fd = r; + socket_apply_socket_options(s, p->fd); + socket_symlink(s); + break; + + case SOCKET_SPECIAL: + + p->fd = special_address_create(p->path, s->writable); + if (p->fd < 0) { + r = p->fd; + goto rollback; + } + break; + + case SOCKET_FIFO: + + p->fd = fifo_address_create( + p->path, + s->directory_mode, + s->socket_mode); + if (p->fd < 0) { + r = p->fd; + goto rollback; + } + + socket_apply_fifo_options(s, p->fd); + socket_symlink(s); + break; + + case SOCKET_MQUEUE: + + p->fd = mq_address_create( + p->path, + s->socket_mode, + s->mq_maxmsg, + s->mq_msgsize); + if (p->fd < 0) { + r = p->fd; + goto rollback; + } + break; + + case SOCKET_USB_FUNCTION: { + _cleanup_free_ char *ep = NULL; + + ep = path_make_absolute("ep0", p->path); + + p->fd = usbffs_address_create(ep); + if (p->fd < 0) { + r = p->fd; + goto rollback; + } + + r = usbffs_write_descs(p->fd, SERVICE(UNIT_DEREF(s->service))); + if (r < 0) + goto rollback; + + r = usbffs_dispatch_eps(p); + if (r < 0) + goto rollback; + + break; + } + default: + assert_not_reached("Unknown port type"); + } + } + + return 0; + +rollback: + socket_close_fds(s); + return r; +} + +static void socket_unwatch_fds(Socket *s) { + SocketPort *p; + int r; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + if (!p->event_source) + continue; + + r = sd_event_source_set_enabled(p->event_source, SD_EVENT_OFF); + if (r < 0) + log_unit_debug_errno(UNIT(s), r, "Failed to disable event source: %m"); + } +} + +static int socket_watch_fds(Socket *s) { + SocketPort *p; + int r; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + if (p->event_source) { + r = sd_event_source_set_enabled(p->event_source, SD_EVENT_ON); + if (r < 0) + goto fail; + } else { + r = sd_event_add_io(UNIT(s)->manager->event, &p->event_source, p->fd, EPOLLIN, socket_dispatch_io, p); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(p->event_source, "socket-port-io"); + } + } + + return 0; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to watch listening fds: %m"); + socket_unwatch_fds(s); + return r; +} + +enum { + SOCKET_OPEN_NONE, + SOCKET_OPEN_SOME, + SOCKET_OPEN_ALL, +}; + +static int socket_check_open(Socket *s) { + bool have_open = false, have_closed = false; + SocketPort *p; + + assert(s); + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + have_closed = true; + else + have_open = true; + + if (have_open && have_closed) + return SOCKET_OPEN_SOME; + } + + if (have_open) + return SOCKET_OPEN_ALL; + + return SOCKET_OPEN_NONE; +} + +static void socket_set_state(Socket *s, SocketState state) { + SocketState old_state; + assert(s); + + old_state = s->state; + s->state = state; + + if (!IN_SET(state, + SOCKET_START_PRE, + SOCKET_START_CHOWN, + SOCKET_START_POST, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_PRE_SIGKILL, + SOCKET_STOP_POST, + SOCKET_FINAL_SIGTERM, + SOCKET_FINAL_SIGKILL)) { + + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + socket_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; + } + + if (state != SOCKET_LISTENING) + socket_unwatch_fds(s); + + if (!IN_SET(state, + SOCKET_START_CHOWN, + SOCKET_START_POST, + SOCKET_LISTENING, + SOCKET_RUNNING, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_PRE_SIGKILL)) + socket_close_fds(s); + + if (state != old_state) + log_unit_debug(UNIT(s), "Changed %s -> %s", socket_state_to_string(old_state), socket_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); +} + +static int socket_coldplug(Unit *u) { + Socket *s = SOCKET(u); + int r; + + assert(s); + assert(s->state == SOCKET_DEAD); + + if (s->deserialized_state == s->state) + return 0; + + if (s->control_pid > 0 && + pid_is_unwaited(s->control_pid) && + IN_SET(s->deserialized_state, + SOCKET_START_PRE, + SOCKET_START_CHOWN, + SOCKET_START_POST, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_PRE_SIGKILL, + SOCKET_STOP_POST, + SOCKET_FINAL_SIGTERM, + SOCKET_FINAL_SIGKILL)) { + + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + + r = socket_arm_timer(s, usec_add(u->state_change_timestamp.monotonic, s->timeout_usec)); + if (r < 0) + return r; + } + + if (IN_SET(s->deserialized_state, + SOCKET_START_CHOWN, + SOCKET_START_POST, + SOCKET_LISTENING, + SOCKET_RUNNING)) { + + /* Originally, we used to simply reopen all sockets here that we didn't have file descriptors + * for. However, this is problematic, as we won't traverse throught the SOCKET_START_CHOWN state for + * them, and thus the UID/GID wouldn't be right. Hence, instead simply check if we have all fds open, + * and if there's a mismatch, warn loudly. */ + + r = socket_check_open(s); + if (r == SOCKET_OPEN_NONE) + log_unit_warning(UNIT(s), + "Socket unit configuration has changed while unit has been running, " + "no open socket file descriptor left. " + "The socket unit is not functional until restarted."); + else if (r == SOCKET_OPEN_SOME) + log_unit_warning(UNIT(s), + "Socket unit configuration has changed while unit has been running, " + "and some socket file descriptors have not been opened yet. " + "The socket unit is not fully functional until restarted."); + } + + if (s->deserialized_state == SOCKET_LISTENING) { + r = socket_watch_fds(s); + if (r < 0) + return r; + } + + if (!IN_SET(s->deserialized_state, SOCKET_DEAD, SOCKET_FAILED)) + (void) unit_setup_dynamic_creds(u); + + socket_set_state(s, s->deserialized_state); + return 0; +} + +static int socket_spawn(Socket *s, ExecCommand *c, pid_t *_pid) { + _cleanup_free_ char **argv = NULL; + pid_t pid; + int r; + ExecParameters exec_params = { + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, + }; + + assert(s); + assert(c); + assert(_pid); + + (void) unit_realize_cgroup(UNIT(s)); + if (s->reset_cpu_usage) { + (void) unit_reset_cpu_usage(UNIT(s)); + s->reset_cpu_usage = false; + } + + r = unit_setup_exec_runtime(UNIT(s)); + if (r < 0) + return r; + + r = unit_setup_dynamic_creds(UNIT(s)); + if (r < 0) + return r; + + r = socket_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); + if (r < 0) + return r; + + r = unit_full_printf_strv(UNIT(s), c->argv, &argv); + if (r < 0) + return r; + + exec_params.argv = argv; + exec_params.environment = UNIT(s)->manager->environment; + exec_params.flags |= UNIT(s)->manager->confirm_spawn ? EXEC_CONFIRM_SPAWN : 0; + exec_params.cgroup_supported = UNIT(s)->manager->cgroup_supported; + exec_params.cgroup_path = UNIT(s)->cgroup_path; + exec_params.cgroup_delegate = s->cgroup_context.delegate; + exec_params.runtime_prefix = manager_get_runtime_prefix(UNIT(s)->manager); + + r = exec_spawn(UNIT(s), + c, + &s->exec_context, + &exec_params, + s->exec_runtime, + &s->dynamic_creds, + &pid); + if (r < 0) + return r; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + return r; + + *_pid = pid; + return 0; +} + +static int socket_chown(Socket *s, pid_t *_pid) { + pid_t pid; + int r; + + r = socket_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); + if (r < 0) + goto fail; + + /* We have to resolve the user names out-of-process, hence + * let's fork here. It's messy, but well, what can we do? */ + + pid = fork(); + if (pid < 0) + return -errno; + + if (pid == 0) { + SocketPort *p; + uid_t uid = UID_INVALID; + gid_t gid = GID_INVALID; + int ret; + + (void) default_signals(SIGNALS_CRASH_HANDLER, SIGNALS_IGNORE, -1); + (void) ignore_signals(SIGPIPE, -1); + log_forget_fds(); + + if (!isempty(s->user)) { + const char *user = s->user; + + r = get_user_creds(&user, &uid, &gid, NULL, NULL); + if (r < 0) { + ret = EXIT_USER; + goto fail_child; + } + } + + if (!isempty(s->group)) { + const char *group = s->group; + + r = get_group_creds(&group, &gid); + if (r < 0) { + ret = EXIT_GROUP; + goto fail_child; + } + } + + LIST_FOREACH(port, p, s->ports) { + const char *path = NULL; + + if (p->type == SOCKET_SOCKET) + path = socket_address_get_path(&p->address); + else if (p->type == SOCKET_FIFO) + path = p->path; + + if (!path) + continue; + + if (chown(path, uid, gid) < 0) { + r = -errno; + ret = EXIT_CHOWN; + goto fail_child; + } + } + + _exit(0); + + fail_child: + log_open(); + log_error_errno(r, "Failed to chown socket at step %s: %m", exit_status_to_string(ret, EXIT_STATUS_SYSTEMD)); + + _exit(ret); + } + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + goto fail; + + *_pid = pid; + return 0; + +fail: + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + return r; +} + +static void socket_enter_dead(Socket *s, SocketResult f) { + assert(s); + + if (s->result == SOCKET_SUCCESS) + s->result = f; + + socket_set_state(s, s->result != SOCKET_SUCCESS ? SOCKET_FAILED : SOCKET_DEAD); + + exec_runtime_destroy(s->exec_runtime); + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + + exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + + unit_unref_uid_gid(UNIT(s), true); + + dynamic_creds_destroy(&s->dynamic_creds); +} + +static void socket_enter_signal(Socket *s, SocketState state, SocketResult f); + +static void socket_enter_stop_post(Socket *s, SocketResult f) { + int r; + assert(s); + + if (s->result == SOCKET_SUCCESS) + s->result = f; + + socket_unwatch_control_pid(s); + s->control_command_id = SOCKET_EXEC_STOP_POST; + s->control_command = s->exec_command[SOCKET_EXEC_STOP_POST]; + + if (s->control_command) { + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + socket_set_state(s, SOCKET_STOP_POST); + } else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'stop-post' task: %m"); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_signal(Socket *s, SocketState state, SocketResult f) { + int r; + + assert(s); + + if (s->result == SOCKET_SUCCESS) + s->result = f; + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + (state != SOCKET_STOP_PRE_SIGTERM && state != SOCKET_FINAL_SIGTERM) ? + KILL_KILL : KILL_TERMINATE, + -1, + s->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = socket_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); + if (r < 0) + goto fail; + + socket_set_state(s, state); + } else if (state == SOCKET_STOP_PRE_SIGTERM) + socket_enter_signal(s, SOCKET_STOP_PRE_SIGKILL, SOCKET_SUCCESS); + else if (state == SOCKET_STOP_PRE_SIGKILL) + socket_enter_stop_post(s, SOCKET_SUCCESS); + else if (state == SOCKET_FINAL_SIGTERM) + socket_enter_signal(s, SOCKET_FINAL_SIGKILL, SOCKET_SUCCESS); + else + socket_enter_dead(s, SOCKET_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to kill processes: %m"); + + if (state == SOCKET_STOP_PRE_SIGTERM || state == SOCKET_STOP_PRE_SIGKILL) + socket_enter_stop_post(s, SOCKET_FAILURE_RESOURCES); + else + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_stop_pre(Socket *s, SocketResult f) { + int r; + assert(s); + + if (s->result == SOCKET_SUCCESS) + s->result = f; + + socket_unwatch_control_pid(s); + s->control_command_id = SOCKET_EXEC_STOP_PRE; + s->control_command = s->exec_command[SOCKET_EXEC_STOP_PRE]; + + if (s->control_command) { + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + socket_set_state(s, SOCKET_STOP_PRE); + } else + socket_enter_stop_post(s, SOCKET_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'stop-pre' task: %m"); + socket_enter_stop_post(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_listening(Socket *s) { + int r; + assert(s); + + r = socket_watch_fds(s); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to watch sockets: %m"); + goto fail; + } + + socket_set_state(s, SOCKET_LISTENING); + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_start_post(Socket *s) { + int r; + assert(s); + + socket_unwatch_control_pid(s); + s->control_command_id = SOCKET_EXEC_START_POST; + s->control_command = s->exec_command[SOCKET_EXEC_START_POST]; + + if (s->control_command) { + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to run 'start-post' task: %m"); + goto fail; + } + + socket_set_state(s, SOCKET_START_POST); + } else + socket_enter_listening(s); + + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_start_chown(Socket *s) { + int r; + + assert(s); + + r = socket_open_fds(s); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to listen on sockets: %m"); + goto fail; + } + + if (!isempty(s->user) || !isempty(s->group)) { + + socket_unwatch_control_pid(s); + s->control_command_id = SOCKET_EXEC_START_CHOWN; + s->control_command = NULL; + + r = socket_chown(s, &s->control_pid); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to fork 'start-chown' task: %m"); + goto fail; + } + + socket_set_state(s, SOCKET_START_CHOWN); + } else + socket_enter_start_post(s); + + return; + +fail: + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); +} + +static void socket_enter_start_pre(Socket *s) { + int r; + assert(s); + + socket_unwatch_control_pid(s); + s->control_command_id = SOCKET_EXEC_START_PRE; + s->control_command = s->exec_command[SOCKET_EXEC_START_PRE]; + + if (s->control_command) { + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) { + log_unit_warning_errno(UNIT(s), r, "Failed to run 'start-pre' task: %m"); + goto fail; + } + + socket_set_state(s, SOCKET_START_PRE); + } else + socket_enter_start_chown(s); + + return; + +fail: + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); +} + +static void flush_ports(Socket *s) { + SocketPort *p; + + /* Flush all incoming traffic, regardless if actual bytes or new connections, so that this socket isn't busy + * anymore */ + + LIST_FOREACH(port, p, s->ports) { + if (p->fd < 0) + continue; + + (void) flush_accept(p->fd); + (void) flush_fd(p->fd); + } +} + +static void socket_enter_running(Socket *s, int cfd) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + int r; + + /* Note that this call takes possession of the connection fd passed. It either has to assign it somewhere or + * close it. */ + + assert(s); + + /* We don't take connections anymore if we are supposed to shut down anyway */ + if (unit_stop_pending(UNIT(s))) { + + log_unit_debug(UNIT(s), "Suppressing connection request since unit stop is scheduled."); + + if (cfd >= 0) + cfd = safe_close(cfd); + else + flush_ports(s); + + return; + } + + if (!ratelimit_test(&s->trigger_limit)) { + safe_close(cfd); + log_unit_warning(UNIT(s), "Trigger limit hit, refusing further activation."); + socket_enter_stop_pre(s, SOCKET_FAILURE_TRIGGER_LIMIT_HIT); + return; + } + + if (cfd < 0) { + Iterator i; + Unit *other; + bool pending = false; + + /* If there's already a start pending don't bother to + * do anything */ + SET_FOREACH(other, UNIT(s)->dependencies[UNIT_TRIGGERS], i) + if (unit_active_or_pending(other)) { + pending = true; + break; + } + + if (!pending) { + if (!UNIT_ISSET(s->service)) { + log_unit_error(UNIT(s), "Service to activate vanished, refusing activation."); + r = -ENOENT; + goto fail; + } + + r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT_DEREF(s->service), JOB_REPLACE, &error, NULL); + if (r < 0) + goto fail; + } + + socket_set_state(s, SOCKET_RUNNING); + } else { + _cleanup_free_ char *prefix = NULL, *instance = NULL, *name = NULL; + _cleanup_(socket_peer_unrefp) SocketPeer *p = NULL; + Service *service; + + if (s->n_connections >= s->max_connections) { + log_unit_warning(UNIT(s), "Too many incoming connections (%u), dropping connection.", + s->n_connections); + safe_close(cfd); + return; + } + + if (s->max_connections_per_source > 0) { + r = socket_acquire_peer(s, cfd, &p); + if (r < 0) { + safe_close(cfd); + return; + } else if (r > 0 && p->n_ref > s->max_connections_per_source) { + _cleanup_free_ char *t = NULL; + + sockaddr_pretty(&p->peer.sa, FAMILY_ADDRESS_SIZE(p->peer.sa.sa_family), true, false, &t); + + log_unit_warning(UNIT(s), + "Too many incoming connections (%u) from source %s, dropping connection.", + p->n_ref, strnull(t)); + safe_close(cfd); + return; + } + } + + r = socket_instantiate_service(s); + if (r < 0) + goto fail; + + r = instance_from_socket(cfd, s->n_accepted, &instance); + if (r < 0) { + if (r != -ENOTCONN) + goto fail; + + /* ENOTCONN is legitimate if TCP RST was received. + * This connection is over, but the socket unit lives on. */ + log_unit_debug(UNIT(s), "Got ENOTCONN on incoming socket, assuming aborted connection attempt, ignoring."); + safe_close(cfd); + return; + } + + r = unit_name_to_prefix(UNIT(s)->id, &prefix); + if (r < 0) + goto fail; + + r = unit_name_build(prefix, instance, ".service", &name); + if (r < 0) + goto fail; + + r = unit_add_name(UNIT_DEREF(s->service), name); + if (r < 0) + goto fail; + + service = SERVICE(UNIT_DEREF(s->service)); + unit_ref_unset(&s->service); + + s->n_accepted++; + unit_choose_id(UNIT(service), name); + + r = service_set_socket_fd(service, cfd, s, s->selinux_context_from_net); + if (r < 0) + goto fail; + + cfd = -1; /* We passed ownership of the fd to the service now. Forget it here. */ + s->n_connections++; + + service->peer = p; /* Pass ownership of the peer reference */ + p = NULL; + + r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT(service), JOB_REPLACE, &error, NULL); + if (r < 0) { + /* We failed to activate the new service, but it still exists. Let's make sure the service + * closes and forgets the connection fd again, immediately. */ + service_close_socket_fd(service); + goto fail; + } + + /* Notify clients about changed counters */ + unit_add_to_dbus_queue(UNIT(s)); + } + + return; + +fail: + log_unit_warning(UNIT(s), "Failed to queue service startup job (Maybe the service file is missing or not a %s unit?): %s", + cfd >= 0 ? "template" : "non-template", + bus_error_message(&error, r)); + + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); + safe_close(cfd); +} + +static void socket_run_next(Socket *s) { + int r; + + assert(s); + assert(s->control_command); + assert(s->control_command->command_next); + + socket_unwatch_control_pid(s); + + s->control_command = s->control_command->command_next; + + r = socket_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run next task: %m"); + + if (s->state == SOCKET_START_POST) + socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); + else if (s->state == SOCKET_STOP_POST) + socket_enter_dead(s, SOCKET_FAILURE_RESOURCES); + else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_RESOURCES); +} + +static int socket_start(Unit *u) { + Socket *s = SOCKET(u); + int r; + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + if (IN_SET(s->state, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGKILL, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_POST, + SOCKET_FINAL_SIGTERM, + SOCKET_FINAL_SIGKILL)) + return -EAGAIN; + + /* Already on it! */ + if (IN_SET(s->state, + SOCKET_START_PRE, + SOCKET_START_CHOWN, + SOCKET_START_POST)) + return 0; + + /* Cannot run this without the service being around */ + if (UNIT_ISSET(s->service)) { + Service *service; + + service = SERVICE(UNIT_DEREF(s->service)); + + if (UNIT(service)->load_state != UNIT_LOADED) { + log_unit_error(u, "Socket service %s not loaded, refusing.", UNIT(service)->id); + return -ENOENT; + } + + /* If the service is already active we cannot start the + * socket */ + if (service->state != SERVICE_DEAD && + service->state != SERVICE_FAILED && + service->state != SERVICE_AUTO_RESTART) { + log_unit_error(u, "Socket service %s already active, refusing.", UNIT(service)->id); + return -EBUSY; + } + } + + assert(s->state == SOCKET_DEAD || s->state == SOCKET_FAILED); + + r = unit_start_limit_test(u); + if (r < 0) { + socket_enter_dead(s, SOCKET_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + s->result = SOCKET_SUCCESS; + s->reset_cpu_usage = true; + + socket_enter_start_pre(s); + return 1; +} + +static int socket_stop(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + /* Already on it */ + if (IN_SET(s->state, + SOCKET_STOP_PRE, + SOCKET_STOP_PRE_SIGTERM, + SOCKET_STOP_PRE_SIGKILL, + SOCKET_STOP_POST, + SOCKET_FINAL_SIGTERM, + SOCKET_FINAL_SIGKILL)) + return 0; + + /* If there's already something running we go directly into + * kill mode. */ + if (IN_SET(s->state, + SOCKET_START_PRE, + SOCKET_START_CHOWN, + SOCKET_START_POST)) { + socket_enter_signal(s, SOCKET_STOP_PRE_SIGTERM, SOCKET_SUCCESS); + return -EAGAIN; + } + + assert(s->state == SOCKET_LISTENING || s->state == SOCKET_RUNNING); + + socket_enter_stop_pre(s, SOCKET_SUCCESS); + return 1; +} + +static int socket_serialize(Unit *u, FILE *f, FDSet *fds) { + Socket *s = SOCKET(u); + SocketPort *p; + int r; + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", socket_state_to_string(s->state)); + unit_serialize_item(u, f, "result", socket_result_to_string(s->result)); + unit_serialize_item_format(u, f, "n-accepted", "%u", s->n_accepted); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", PID_FMT, s->control_pid); + + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", socket_exec_command_to_string(s->control_command_id)); + + LIST_FOREACH(port, p, s->ports) { + int copy; + + if (p->fd < 0) + continue; + + copy = fdset_put_dup(fds, p->fd); + if (copy < 0) + return copy; + + if (p->type == SOCKET_SOCKET) { + _cleanup_free_ char *t = NULL; + + r = socket_address_print(&p->address, &t); + if (r < 0) + return r; + + if (socket_address_family(&p->address) == AF_NETLINK) + unit_serialize_item_format(u, f, "netlink", "%i %s", copy, t); + else + unit_serialize_item_format(u, f, "socket", "%i %i %s", copy, p->address.type, t); + + } else if (p->type == SOCKET_SPECIAL) + unit_serialize_item_format(u, f, "special", "%i %s", copy, p->path); + else if (p->type == SOCKET_MQUEUE) + unit_serialize_item_format(u, f, "mqueue", "%i %s", copy, p->path); + else if (p->type == SOCKET_USB_FUNCTION) + unit_serialize_item_format(u, f, "ffs", "%i %s", copy, p->path); + else { + assert(p->type == SOCKET_FIFO); + unit_serialize_item_format(u, f, "fifo", "%i %s", copy, p->path); + } + } + + return 0; +} + +static void socket_port_take_fd(SocketPort *p, FDSet *fds, int fd) { + safe_close(p->fd); + p->fd = fdset_remove(fds, fd); +} + +static int socket_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Socket *s = SOCKET(u); + + assert(u); + assert(key); + assert(value); + + if (streq(key, "state")) { + SocketState state; + + state = socket_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + SocketResult f; + + f = socket_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != SOCKET_SUCCESS) + s->result = f; + + } else if (streq(key, "n-accepted")) { + unsigned k; + + if (safe_atou(value, &k) < 0) + log_unit_debug(u, "Failed to parse n-accepted value: %s", value); + else + s->n_accepted += k; + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse control-pid value: %s", value); + else + s->control_pid = pid; + } else if (streq(key, "control-command")) { + SocketExecCommand id; + + id = socket_exec_command_from_string(value); + if (id < 0) + log_unit_debug(u, "Failed to parse exec-command value: %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command[id]; + } + } else if (streq(key, "fifo")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse fifo value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_FIFO && + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else if (streq(key, "special")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse special value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_SPECIAL && + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else if (streq(key, "mqueue")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse mqueue value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_MQUEUE && + streq(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else if (streq(key, "socket")) { + int fd, type, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %i %n", &fd, &type, &skip) < 2 || fd < 0 || type < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse socket value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (socket_address_is(&p->address, value+skip, type)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else if (streq(key, "netlink")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse socket value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (socket_address_is_netlink(&p->address, value+skip)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else if (streq(key, "ffs")) { + int fd, skip = 0; + SocketPort *p; + + if (sscanf(value, "%i %n", &fd, &skip) < 1 || fd < 0 || !fdset_contains(fds, fd)) + log_unit_debug(u, "Failed to parse ffs value: %s", value); + else + LIST_FOREACH(port, p, s->ports) + if (p->type == SOCKET_USB_FUNCTION && + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); + break; + } + + } else + log_unit_debug(UNIT(s), "Unknown serialization key: %s", key); + + return 0; +} + +static void socket_distribute_fds(Unit *u, FDSet *fds) { + Socket *s = SOCKET(u); + SocketPort *p; + + assert(u); + + LIST_FOREACH(port, p, s->ports) { + Iterator i; + int fd; + + if (p->type != SOCKET_SOCKET) + continue; + + if (p->fd >= 0) + continue; + + FDSET_FOREACH(fd, fds, i) { + if (socket_address_matches_fd(&p->address, fd)) { + p->fd = fdset_remove(fds, fd); + s->deserialized_state = SOCKET_LISTENING; + break; + } + } + } +} + +_pure_ static UnitActiveState socket_active_state(Unit *u) { + assert(u); + + return state_translation_table[SOCKET(u)->state]; +} + +_pure_ static const char *socket_sub_state_to_string(Unit *u) { + assert(u); + + return socket_state_to_string(SOCKET(u)->state); +} + +const char* socket_port_type_to_string(SocketPort *p) { + + assert(p); + + switch (p->type) { + + case SOCKET_SOCKET: + + switch (p->address.type) { + + case SOCK_STREAM: + return "Stream"; + + case SOCK_DGRAM: + return "Datagram"; + + case SOCK_SEQPACKET: + return "SequentialPacket"; + + case SOCK_RAW: + if (socket_address_family(&p->address) == AF_NETLINK) + return "Netlink"; + + default: + return NULL; + } + + case SOCKET_SPECIAL: + return "Special"; + + case SOCKET_MQUEUE: + return "MessageQueue"; + + case SOCKET_FIFO: + return "FIFO"; + + case SOCKET_USB_FUNCTION: + return "USBFunction"; + + default: + return NULL; + } +} + +_pure_ static bool socket_check_gc(Unit *u) { + Socket *s = SOCKET(u); + + assert(u); + + return s->n_connections > 0; +} + +static int socket_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + SocketPort *p = userdata; + int cfd = -1; + + assert(p); + assert(fd >= 0); + + if (p->socket->state != SOCKET_LISTENING) + return 0; + + log_unit_debug(UNIT(p->socket), "Incoming traffic"); + + if (revents != EPOLLIN) { + + if (revents & EPOLLHUP) + log_unit_error(UNIT(p->socket), "Got POLLHUP on a listening socket. The service probably invoked shutdown() on it, and should better not do that."); + else + log_unit_error(UNIT(p->socket), "Got unexpected poll event (0x%x) on socket.", revents); + goto fail; + } + + if (p->socket->accept && + p->type == SOCKET_SOCKET && + socket_address_can_accept(&p->address)) { + + for (;;) { + + cfd = accept4(fd, NULL, NULL, SOCK_NONBLOCK); + if (cfd < 0) { + + if (errno == EINTR) + continue; + + log_unit_error_errno(UNIT(p->socket), errno, "Failed to accept socket: %m"); + goto fail; + } + + break; + } + + socket_apply_socket_options(p->socket, cfd); + } + + socket_enter_running(p->socket, cfd); + return 0; + +fail: + socket_enter_stop_pre(p->socket, SOCKET_FAILURE_RESOURCES); + return 0; +} + +static void socket_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Socket *s = SOCKET(u); + SocketResult f; + + assert(s); + assert(pid >= 0); + + if (pid != s->control_pid) + return; + + s->control_pid = 0; + + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) + f = SOCKET_SUCCESS; + else if (code == CLD_EXITED) + f = SOCKET_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SOCKET_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SOCKET_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown sigchld code"); + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, &s->exec_context, pid, code, status); + + if (s->control_command->ignore) + f = SOCKET_SUCCESS; + } + + log_unit_full(u, f == SOCKET_SUCCESS ? LOG_DEBUG : LOG_NOTICE, 0, + "Control process exited, code=%s status=%i", + sigchld_code_to_string(code), status); + + if (s->result == SOCKET_SUCCESS) + s->result = f; + + if (s->control_command && + s->control_command->command_next && + f == SOCKET_SUCCESS) { + + log_unit_debug(u, "Running next command for state %s", socket_state_to_string(s->state)); + socket_run_next(s); + } else { + s->control_command = NULL; + s->control_command_id = _SOCKET_EXEC_COMMAND_INVALID; + + /* No further commands for this step, so let's figure + * out what to do next */ + + log_unit_debug(u, "Got final SIGCHLD for state %s", socket_state_to_string(s->state)); + + switch (s->state) { + + case SOCKET_START_PRE: + if (f == SOCKET_SUCCESS) + socket_enter_start_chown(s); + else + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, f); + break; + + case SOCKET_START_CHOWN: + if (f == SOCKET_SUCCESS) + socket_enter_start_post(s); + else + socket_enter_stop_pre(s, f); + break; + + case SOCKET_START_POST: + if (f == SOCKET_SUCCESS) + socket_enter_listening(s); + else + socket_enter_stop_pre(s, f); + break; + + case SOCKET_STOP_PRE: + case SOCKET_STOP_PRE_SIGTERM: + case SOCKET_STOP_PRE_SIGKILL: + socket_enter_stop_post(s, f); + break; + + case SOCKET_STOP_POST: + case SOCKET_FINAL_SIGTERM: + case SOCKET_FINAL_SIGKILL: + socket_enter_dead(s, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static int socket_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + Socket *s = SOCKET(userdata); + + assert(s); + assert(s->timer_event_source == source); + + switch (s->state) { + + case SOCKET_START_PRE: + log_unit_warning(UNIT(s), "Starting timed out. Terminating."); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_START_CHOWN: + case SOCKET_START_POST: + log_unit_warning(UNIT(s), "Starting timed out. Stopping."); + socket_enter_stop_pre(s, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_PRE: + log_unit_warning(UNIT(s), "Stopping timed out. Terminating."); + socket_enter_signal(s, SOCKET_STOP_PRE_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_PRE_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "Stopping timed out. Killing."); + socket_enter_signal(s, SOCKET_STOP_PRE_SIGKILL, SOCKET_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "Stopping timed out. Skipping SIGKILL. Ignoring."); + socket_enter_stop_post(s, SOCKET_FAILURE_TIMEOUT); + } + break; + + case SOCKET_STOP_PRE_SIGKILL: + log_unit_warning(UNIT(s), "Processes still around after SIGKILL. Ignoring."); + socket_enter_stop_post(s, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_STOP_POST: + log_unit_warning(UNIT(s), "Stopping timed out (2). Terminating."); + socket_enter_signal(s, SOCKET_FINAL_SIGTERM, SOCKET_FAILURE_TIMEOUT); + break; + + case SOCKET_FINAL_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "Stopping timed out (2). Killing."); + socket_enter_signal(s, SOCKET_FINAL_SIGKILL, SOCKET_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "Stopping timed out (2). Skipping SIGKILL. Ignoring."); + socket_enter_dead(s, SOCKET_FAILURE_TIMEOUT); + } + break; + + case SOCKET_FINAL_SIGKILL: + log_unit_warning(UNIT(s), "Still around after SIGKILL (2). Entering failed mode."); + socket_enter_dead(s, SOCKET_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +int socket_collect_fds(Socket *s, int **fds) { + int *rfds, k = 0, n = 0; + SocketPort *p; + + assert(s); + assert(fds); + + /* Called from the service code for requesting our fds */ + + LIST_FOREACH(port, p, s->ports) { + if (p->fd >= 0) + n++; + n += p->n_auxiliary_fds; + } + + if (n <= 0) { + *fds = NULL; + return 0; + } + + rfds = new(int, n); + if (!rfds) + return -ENOMEM; + + LIST_FOREACH(port, p, s->ports) { + int i; + + if (p->fd >= 0) + rfds[k++] = p->fd; + for (i = 0; i < p->n_auxiliary_fds; ++i) + rfds[k++] = p->auxiliary_fds[i]; + } + + assert(k == n); + + *fds = rfds; + return n; +} + +static void socket_reset_failed(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + if (s->state == SOCKET_FAILED) + socket_set_state(s, SOCKET_DEAD); + + s->result = SOCKET_SUCCESS; +} + +void socket_connection_unref(Socket *s) { + assert(s); + + /* The service is dead. Yay! + * + * This is strictly for one-instance-per-connection + * services. */ + + assert(s->n_connections > 0); + s->n_connections--; + + log_unit_debug(UNIT(s), "One connection closed, %u left.", s->n_connections); +} + +static void socket_trigger_notify(Unit *u, Unit *other) { + Socket *s = SOCKET(u); + + assert(u); + assert(other); + + /* Filter out invocations with bogus state */ + if (other->load_state != UNIT_LOADED || other->type != UNIT_SERVICE) + return; + + /* Don't propagate state changes from the service if we are already down */ + if (!IN_SET(s->state, SOCKET_RUNNING, SOCKET_LISTENING)) + return; + + /* We don't care for the service state if we are in Accept=yes mode */ + if (s->accept) + return; + + /* Propagate start limit hit state */ + if (other->start_limit_hit) { + socket_enter_stop_pre(s, SOCKET_FAILURE_SERVICE_START_LIMIT_HIT); + return; + } + + /* Don't propagate anything if there's still a job queued */ + if (other->job) + return; + + if (IN_SET(SERVICE(other)->state, + SERVICE_DEAD, SERVICE_FAILED, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL, + SERVICE_AUTO_RESTART)) + socket_enter_listening(s); + + if (SERVICE(other)->state == SERVICE_RUNNING) + socket_set_state(s, SOCKET_RUNNING); +} + +static int socket_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, SOCKET(u)->control_pid, error); +} + +static int socket_get_timeout(Unit *u, usec_t *timeout) { + Socket *s = SOCKET(u); + usec_t t; + int r; + + if (!s->timer_event_source) + return 0; + + r = sd_event_source_get_time(s->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +char *socket_fdname(Socket *s) { + assert(s); + + /* Returns the name to use for $LISTEN_NAMES. If the user + * didn't specify anything specifically, use the socket unit's + * name as fallback. */ + + if (s->fdname) + return s->fdname; + + return UNIT(s)->id; +} + +static int socket_control_pid(Unit *u) { + Socket *s = SOCKET(u); + + assert(s); + + return s->control_pid; +} + +static const char* const socket_exec_command_table[_SOCKET_EXEC_COMMAND_MAX] = { + [SOCKET_EXEC_START_PRE] = "StartPre", + [SOCKET_EXEC_START_CHOWN] = "StartChown", + [SOCKET_EXEC_START_POST] = "StartPost", + [SOCKET_EXEC_STOP_PRE] = "StopPre", + [SOCKET_EXEC_STOP_POST] = "StopPost" +}; + +DEFINE_STRING_TABLE_LOOKUP(socket_exec_command, SocketExecCommand); + +static const char* const socket_result_table[_SOCKET_RESULT_MAX] = { + [SOCKET_SUCCESS] = "success", + [SOCKET_FAILURE_RESOURCES] = "resources", + [SOCKET_FAILURE_TIMEOUT] = "timeout", + [SOCKET_FAILURE_EXIT_CODE] = "exit-code", + [SOCKET_FAILURE_SIGNAL] = "signal", + [SOCKET_FAILURE_CORE_DUMP] = "core-dump", + [SOCKET_FAILURE_START_LIMIT_HIT] = "start-limit-hit", + [SOCKET_FAILURE_TRIGGER_LIMIT_HIT] = "trigger-limit-hit", + [SOCKET_FAILURE_SERVICE_START_LIMIT_HIT] = "service-start-limit-hit" +}; + +DEFINE_STRING_TABLE_LOOKUP(socket_result, SocketResult); + +const UnitVTable socket_vtable = { + .object_size = sizeof(Socket), + .exec_context_offset = offsetof(Socket, exec_context), + .cgroup_context_offset = offsetof(Socket, cgroup_context), + .kill_context_offset = offsetof(Socket, kill_context), + .exec_runtime_offset = offsetof(Socket, exec_runtime), + .dynamic_creds_offset = offsetof(Socket, dynamic_creds), + + .sections = + "Unit\0" + "Socket\0" + "Install\0", + .private_section = "Socket", + + .init = socket_init, + .done = socket_done, + .load = socket_load, + + .coldplug = socket_coldplug, + + .dump = socket_dump, + + .start = socket_start, + .stop = socket_stop, + + .kill = socket_kill, + + .get_timeout = socket_get_timeout, + + .serialize = socket_serialize, + .deserialize_item = socket_deserialize_item, + .distribute_fds = socket_distribute_fds, + + .active_state = socket_active_state, + .sub_state_to_string = socket_sub_state_to_string, + + .check_gc = socket_check_gc, + + .sigchld_event = socket_sigchld_event, + + .trigger_notify = socket_trigger_notify, + + .reset_failed = socket_reset_failed, + + .control_pid = socket_control_pid, + + .bus_vtable = bus_socket_vtable, + .bus_set_property = bus_socket_set_property, + .bus_commit_properties = bus_socket_commit_properties, + + .status_message_formats = { + /*.starting_stopping = { + [0] = "Starting socket %s...", + [1] = "Stopping socket %s...", + },*/ + .finished_start_job = { + [JOB_DONE] = "Listening on %s.", + [JOB_FAILED] = "Failed to listen on %s.", + [JOB_TIMEOUT] = "Timed out starting %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Closed %s.", + [JOB_FAILED] = "Failed stopping %s.", + [JOB_TIMEOUT] = "Timed out stopping %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/swap.c b/src/grp-system/libcore/src/swap.c new file mode 100644 index 0000000000..85f789e12b --- /dev/null +++ b/src/grp-system/libcore/src/swap.c @@ -0,0 +1,1547 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <sys/epoll.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <libudev.h> + +#include "core/swap.h" +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/exit-status.h" +#include "systemd-basic/fd-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/virt.h" +#include "systemd-shared/fstab-util.h" +#include "systemd-shared/udev-util.h" + +#include "dbus-swap.h" + +static const UnitActiveState state_translation_table[_SWAP_STATE_MAX] = { + [SWAP_DEAD] = UNIT_INACTIVE, + [SWAP_ACTIVATING] = UNIT_ACTIVATING, + [SWAP_ACTIVATING_DONE] = UNIT_ACTIVE, + [SWAP_ACTIVE] = UNIT_ACTIVE, + [SWAP_DEACTIVATING] = UNIT_DEACTIVATING, + [SWAP_ACTIVATING_SIGTERM] = UNIT_DEACTIVATING, + [SWAP_ACTIVATING_SIGKILL] = UNIT_DEACTIVATING, + [SWAP_DEACTIVATING_SIGTERM] = UNIT_DEACTIVATING, + [SWAP_DEACTIVATING_SIGKILL] = UNIT_DEACTIVATING, + [SWAP_FAILED] = UNIT_FAILED +}; + +static int swap_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata); +static int swap_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata); + +static void swap_unset_proc_swaps(Swap *s) { + assert(s); + + if (!s->from_proc_swaps) + return; + + s->parameters_proc_swaps.what = mfree(s->parameters_proc_swaps.what); + + s->from_proc_swaps = false; +} + +static int swap_set_devnode(Swap *s, const char *devnode) { + Hashmap *swaps; + Swap *first; + int r; + + assert(s); + + r = hashmap_ensure_allocated(&UNIT(s)->manager->swaps_by_devnode, &string_hash_ops); + if (r < 0) + return r; + + swaps = UNIT(s)->manager->swaps_by_devnode; + + if (s->devnode) { + first = hashmap_get(swaps, s->devnode); + + LIST_REMOVE(same_devnode, first, s); + if (first) + hashmap_replace(swaps, first->devnode, first); + else + hashmap_remove(swaps, s->devnode); + + s->devnode = mfree(s->devnode); + } + + if (devnode) { + s->devnode = strdup(devnode); + if (!s->devnode) + return -ENOMEM; + + first = hashmap_get(swaps, s->devnode); + LIST_PREPEND(same_devnode, first, s); + + return hashmap_replace(swaps, first->devnode, first); + } + + return 0; +} + +static void swap_init(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + assert(UNIT(s)->load_state == UNIT_STUB); + + s->timeout_usec = u->manager->default_timeout_start_usec; + + s->exec_context.std_output = u->manager->default_std_output; + s->exec_context.std_error = u->manager->default_std_error; + + s->parameters_proc_swaps.priority = s->parameters_fragment.priority = -1; + + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + + u->ignore_on_isolate = true; +} + +static void swap_unwatch_control_pid(Swap *s) { + assert(s); + + if (s->control_pid <= 0) + return; + + unit_unwatch_pid(UNIT(s), s->control_pid); + s->control_pid = 0; +} + +static void swap_done(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + swap_unset_proc_swaps(s); + swap_set_devnode(s, NULL); + + s->what = mfree(s->what); + s->parameters_fragment.what = mfree(s->parameters_fragment.what); + s->parameters_fragment.options = mfree(s->parameters_fragment.options); + + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + exec_command_done_array(s->exec_command, _SWAP_EXEC_COMMAND_MAX); + s->control_command = NULL; + + dynamic_creds_unref(&s->dynamic_creds); + + swap_unwatch_control_pid(s); + + s->timer_event_source = sd_event_source_unref(s->timer_event_source); +} + +static int swap_arm_timer(Swap *s, usec_t usec) { + int r; + + assert(s); + + if (s->timer_event_source) { + r = sd_event_source_set_time(s->timer_event_source, usec); + if (r < 0) + return r; + + return sd_event_source_set_enabled(s->timer_event_source, SD_EVENT_ONESHOT); + } + + if (usec == USEC_INFINITY) + return 0; + + r = sd_event_add_time( + UNIT(s)->manager->event, + &s->timer_event_source, + CLOCK_MONOTONIC, + usec, 0, + swap_dispatch_timer, s); + if (r < 0) + return r; + + (void) sd_event_source_set_description(s->timer_event_source, "swap-timer"); + + return 0; +} + +static int swap_add_device_links(Swap *s) { + assert(s); + + if (!s->what) + return 0; + + if (!s->from_fragment) + return 0; + + if (is_device_path(s->what)) + return unit_add_node_link(UNIT(s), s->what, MANAGER_IS_SYSTEM(UNIT(s)->manager), UNIT_BINDS_TO); + else + /* File based swap devices need to be ordered after + * systemd-remount-fs.service, since they might need a + * writable file system. */ + return unit_add_dependency_by_name(UNIT(s), UNIT_AFTER, SPECIAL_REMOUNT_FS_SERVICE, NULL, true); +} + +static int swap_add_default_dependencies(Swap *s) { + int r; + + assert(s); + + if (!UNIT(s)->default_dependencies) + return 0; + + if (!MANAGER_IS_SYSTEM(UNIT(s)->manager)) + return 0; + + if (detect_container() > 0) + return 0; + + /* swap units generated for the swap dev links are missing the + * ordering dep against the swap target. */ + r = unit_add_dependency_by_name(UNIT(s), UNIT_BEFORE, SPECIAL_SWAP_TARGET, NULL, true); + if (r < 0) + return r; + + return unit_add_two_dependencies_by_name(UNIT(s), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_UMOUNT_TARGET, NULL, true); +} + +static int swap_verify(Swap *s) { + _cleanup_free_ char *e = NULL; + int r; + + if (UNIT(s)->load_state != UNIT_LOADED) + return 0; + + r = unit_name_from_path(s->what, ".swap", &e); + if (r < 0) + return log_unit_error_errno(UNIT(s), r, "Failed to generate unit name from path: %m"); + + if (!unit_has_name(UNIT(s), e)) { + log_unit_error(UNIT(s), "Value of What= and unit name do not match, not loading."); + return -EINVAL; + } + + if (s->exec_context.pam_name && s->kill_context.kill_mode != KILL_CONTROL_GROUP) { + log_unit_error(UNIT(s), "Unit has PAM enabled. Kill mode must be set to 'control-group'. Refusing to load."); + return -EINVAL; + } + + return 0; +} + +static int swap_load_devnode(Swap *s) { + _cleanup_udev_device_unref_ struct udev_device *d = NULL; + struct stat st; + const char *p; + + assert(s); + + if (stat(s->what, &st) < 0 || !S_ISBLK(st.st_mode)) + return 0; + + d = udev_device_new_from_devnum(UNIT(s)->manager->udev, 'b', st.st_rdev); + if (!d) + return 0; + + p = udev_device_get_devnode(d); + if (!p) + return 0; + + return swap_set_devnode(s, p); +} + +static int swap_load(Unit *u) { + int r; + Swap *s = SWAP(u); + + assert(s); + assert(u->load_state == UNIT_STUB); + + /* Load a .swap file */ + r = unit_load_fragment_and_dropin_optional(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + + if (UNIT(s)->fragment_path) + s->from_fragment = true; + + if (!s->what) { + if (s->parameters_fragment.what) + s->what = strdup(s->parameters_fragment.what); + else if (s->parameters_proc_swaps.what) + s->what = strdup(s->parameters_proc_swaps.what); + else { + r = unit_name_to_path(u->id, &s->what); + if (r < 0) + return r; + } + + if (!s->what) + return -ENOMEM; + } + + path_kill_slashes(s->what); + + if (!UNIT(s)->description) { + r = unit_set_description(u, s->what); + if (r < 0) + return r; + } + + r = unit_require_mounts_for(UNIT(s), s->what); + if (r < 0) + return r; + + r = swap_add_device_links(s); + if (r < 0) + return r; + + r = swap_load_devnode(s); + if (r < 0) + return r; + + r = unit_patch_contexts(u); + if (r < 0) + return r; + + r = unit_add_exec_dependencies(u, &s->exec_context); + if (r < 0) + return r; + + r = unit_set_default_slice(u); + if (r < 0) + return r; + + r = swap_add_default_dependencies(s); + if (r < 0) + return r; + } + + return swap_verify(s); +} + +static int swap_setup_unit( + Manager *m, + const char *what, + const char *what_proc_swaps, + int priority, + bool set_flags) { + + _cleanup_free_ char *e = NULL; + bool delete = false; + Unit *u = NULL; + int r; + SwapParameters *p; + + assert(m); + assert(what); + assert(what_proc_swaps); + + r = unit_name_from_path(what, ".swap", &e); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to generate unit name from path: %m"); + + u = manager_get_unit(m, e); + + if (u && + SWAP(u)->from_proc_swaps && + !path_equal(SWAP(u)->parameters_proc_swaps.what, what_proc_swaps)) { + log_error("Swap %s appeared twice with different device paths %s and %s", e, SWAP(u)->parameters_proc_swaps.what, what_proc_swaps); + return -EEXIST; + } + + if (!u) { + delete = true; + + r = unit_new_for_name(m, sizeof(Swap), e, &u); + if (r < 0) + goto fail; + + SWAP(u)->what = strdup(what); + if (!SWAP(u)->what) { + r = -ENOMEM; + goto fail; + } + + unit_add_to_load_queue(u); + } else + delete = false; + + p = &SWAP(u)->parameters_proc_swaps; + + if (!p->what) { + p->what = strdup(what_proc_swaps); + if (!p->what) { + r = -ENOMEM; + goto fail; + } + } + + if (set_flags) { + SWAP(u)->is_active = true; + SWAP(u)->just_activated = !SWAP(u)->from_proc_swaps; + } + + SWAP(u)->from_proc_swaps = true; + + p->priority = priority; + + unit_add_to_dbus_queue(u); + return 0; + +fail: + log_unit_warning_errno(u, r, "Failed to load swap unit: %m"); + + if (delete && u) + unit_free(u); + + return r; +} + +static int swap_process_new(Manager *m, const char *device, int prio, bool set_flags) { + _cleanup_udev_device_unref_ struct udev_device *d = NULL; + struct udev_list_entry *item = NULL, *first = NULL; + const char *dn; + struct stat st; + int r; + + assert(m); + + r = swap_setup_unit(m, device, device, prio, set_flags); + if (r < 0) + return r; + + /* If this is a block device, then let's add duplicates for + * all other names of this block device */ + if (stat(device, &st) < 0 || !S_ISBLK(st.st_mode)) + return 0; + + d = udev_device_new_from_devnum(m->udev, 'b', st.st_rdev); + if (!d) + return 0; + + /* Add the main device node */ + dn = udev_device_get_devnode(d); + if (dn && !streq(dn, device)) + swap_setup_unit(m, dn, device, prio, set_flags); + + /* Add additional units for all symlinks */ + first = udev_device_get_devlinks_list_entry(d); + udev_list_entry_foreach(item, first) { + const char *p; + + /* Don't bother with the /dev/block links */ + p = udev_list_entry_get_name(item); + + if (streq(p, device)) + continue; + + if (path_startswith(p, "/dev/block/")) + continue; + + if (stat(p, &st) >= 0) + if (!S_ISBLK(st.st_mode) || + st.st_rdev != udev_device_get_devnum(d)) + continue; + + swap_setup_unit(m, p, device, prio, set_flags); + } + + return r; +} + +static void swap_set_state(Swap *s, SwapState state) { + SwapState old_state; + Swap *other; + + assert(s); + + old_state = s->state; + s->state = state; + + if (state != SWAP_ACTIVATING && + state != SWAP_ACTIVATING_SIGTERM && + state != SWAP_ACTIVATING_SIGKILL && + state != SWAP_ACTIVATING_DONE && + state != SWAP_DEACTIVATING && + state != SWAP_DEACTIVATING_SIGTERM && + state != SWAP_DEACTIVATING_SIGKILL) { + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + swap_unwatch_control_pid(s); + s->control_command = NULL; + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + } + + if (state != old_state) + log_unit_debug(UNIT(s), "Changed %s -> %s", swap_state_to_string(old_state), swap_state_to_string(state)); + + unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); + + /* If there other units for the same device node have a job + queued it might be worth checking again if it is runnable + now. This is necessary, since swap_start() refuses + operation with EAGAIN if there's already another job for + the same device node queued. */ + LIST_FOREACH_OTHERS(same_devnode, other, s) + if (UNIT(other)->job) + job_add_to_run_queue(UNIT(other)->job); +} + +static int swap_coldplug(Unit *u) { + Swap *s = SWAP(u); + SwapState new_state = SWAP_DEAD; + int r; + + assert(s); + assert(s->state == SWAP_DEAD); + + if (s->deserialized_state != s->state) + new_state = s->deserialized_state; + else if (s->from_proc_swaps) + new_state = SWAP_ACTIVE; + + if (new_state == s->state) + return 0; + + if (s->control_pid > 0 && + pid_is_unwaited(s->control_pid) && + IN_SET(new_state, + SWAP_ACTIVATING, + SWAP_ACTIVATING_SIGTERM, + SWAP_ACTIVATING_SIGKILL, + SWAP_ACTIVATING_DONE, + SWAP_DEACTIVATING, + SWAP_DEACTIVATING_SIGTERM, + SWAP_DEACTIVATING_SIGKILL)) { + + r = unit_watch_pid(UNIT(s), s->control_pid); + if (r < 0) + return r; + + r = swap_arm_timer(s, usec_add(u->state_change_timestamp.monotonic, s->timeout_usec)); + if (r < 0) + return r; + } + + if (!IN_SET(new_state, SWAP_DEAD, SWAP_FAILED)) + (void) unit_setup_dynamic_creds(u); + + swap_set_state(s, new_state); + return 0; +} + +static void swap_dump(Unit *u, FILE *f, const char *prefix) { + Swap *s = SWAP(u); + SwapParameters *p; + + assert(s); + assert(f); + + if (s->from_proc_swaps) + p = &s->parameters_proc_swaps; + else if (s->from_fragment) + p = &s->parameters_fragment; + else + p = NULL; + + fprintf(f, + "%sSwap State: %s\n" + "%sResult: %s\n" + "%sWhat: %s\n" + "%sFrom /proc/swaps: %s\n" + "%sFrom fragment: %s\n", + prefix, swap_state_to_string(s->state), + prefix, swap_result_to_string(s->result), + prefix, s->what, + prefix, yes_no(s->from_proc_swaps), + prefix, yes_no(s->from_fragment)); + + if (s->devnode) + fprintf(f, "%sDevice Node: %s\n", prefix, s->devnode); + + if (p) + fprintf(f, + "%sPriority: %i\n" + "%sOptions: %s\n", + prefix, p->priority, + prefix, strempty(p->options)); + + if (s->control_pid > 0) + fprintf(f, + "%sControl PID: "PID_FMT"\n", + prefix, s->control_pid); + + exec_context_dump(&s->exec_context, f, prefix); + kill_context_dump(&s->kill_context, f, prefix); +} + +static int swap_spawn(Swap *s, ExecCommand *c, pid_t *_pid) { + pid_t pid; + int r; + ExecParameters exec_params = { + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, + }; + + assert(s); + assert(c); + assert(_pid); + + (void) unit_realize_cgroup(UNIT(s)); + if (s->reset_cpu_usage) { + (void) unit_reset_cpu_usage(UNIT(s)); + s->reset_cpu_usage = false; + } + + r = unit_setup_exec_runtime(UNIT(s)); + if (r < 0) + goto fail; + + r = unit_setup_dynamic_creds(UNIT(s)); + if (r < 0) + return r; + + r = swap_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); + if (r < 0) + goto fail; + + exec_params.environment = UNIT(s)->manager->environment; + exec_params.flags |= UNIT(s)->manager->confirm_spawn ? EXEC_CONFIRM_SPAWN : 0; + exec_params.cgroup_supported = UNIT(s)->manager->cgroup_supported; + exec_params.cgroup_path = UNIT(s)->cgroup_path; + exec_params.cgroup_delegate = s->cgroup_context.delegate; + exec_params.runtime_prefix = manager_get_runtime_prefix(UNIT(s)->manager); + + r = exec_spawn(UNIT(s), + c, + &s->exec_context, + &exec_params, + s->exec_runtime, + &s->dynamic_creds, + &pid); + if (r < 0) + goto fail; + + r = unit_watch_pid(UNIT(s), pid); + if (r < 0) + /* FIXME: we need to do something here */ + goto fail; + + *_pid = pid; + + return 0; + +fail: + s->timer_event_source = sd_event_source_unref(s->timer_event_source); + return r; +} + +static void swap_enter_dead(Swap *s, SwapResult f) { + assert(s); + + if (s->result == SWAP_SUCCESS) + s->result = f; + + swap_set_state(s, s->result != SWAP_SUCCESS ? SWAP_FAILED : SWAP_DEAD); + + exec_runtime_destroy(s->exec_runtime); + s->exec_runtime = exec_runtime_unref(s->exec_runtime); + + exec_context_destroy_runtime_directory(&s->exec_context, manager_get_runtime_prefix(UNIT(s)->manager)); + + unit_unref_uid_gid(UNIT(s), true); + + dynamic_creds_destroy(&s->dynamic_creds); +} + +static void swap_enter_active(Swap *s, SwapResult f) { + assert(s); + + if (s->result == SWAP_SUCCESS) + s->result = f; + + swap_set_state(s, SWAP_ACTIVE); +} + +static void swap_enter_signal(Swap *s, SwapState state, SwapResult f) { + int r; + + assert(s); + + if (s->result == SWAP_SUCCESS) + s->result = f; + + r = unit_kill_context( + UNIT(s), + &s->kill_context, + (state != SWAP_ACTIVATING_SIGTERM && state != SWAP_DEACTIVATING_SIGTERM) ? + KILL_KILL : KILL_TERMINATE, + -1, + s->control_pid, + false); + if (r < 0) + goto fail; + + if (r > 0) { + r = swap_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); + if (r < 0) + goto fail; + + swap_set_state(s, state); + } else if (state == SWAP_ACTIVATING_SIGTERM) + swap_enter_signal(s, SWAP_ACTIVATING_SIGKILL, SWAP_SUCCESS); + else if (state == SWAP_DEACTIVATING_SIGTERM) + swap_enter_signal(s, SWAP_DEACTIVATING_SIGKILL, SWAP_SUCCESS); + else + swap_enter_dead(s, SWAP_SUCCESS); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to kill processes: %m"); + swap_enter_dead(s, SWAP_FAILURE_RESOURCES); +} + +static void swap_enter_activating(Swap *s) { + _cleanup_free_ char *opts = NULL; + int r; + + assert(s); + + s->control_command_id = SWAP_EXEC_ACTIVATE; + s->control_command = s->exec_command + SWAP_EXEC_ACTIVATE; + + if (s->from_fragment) { + int priority = -1; + + r = fstab_find_pri(s->parameters_fragment.options, &priority); + if (r < 0) + log_warning_errno(r, "Failed to parse swap priority \"%s\", ignoring: %m", s->parameters_fragment.options); + else if (r == 1 && s->parameters_fragment.priority >= 0) + log_warning("Duplicate swap priority configuration by Priority and Options fields."); + + if (r <= 0 && s->parameters_fragment.priority >= 0) { + if (s->parameters_fragment.options) + r = asprintf(&opts, "%s,pri=%i", s->parameters_fragment.options, s->parameters_fragment.priority); + else + r = asprintf(&opts, "pri=%i", s->parameters_fragment.priority); + if (r < 0) + goto fail; + } + } + + r = exec_command_set(s->control_command, "/sbin/swapon", NULL); + if (r < 0) + goto fail; + + if (s->parameters_fragment.options || opts) { + r = exec_command_append(s->control_command, "-o", + opts ? : s->parameters_fragment.options, NULL); + if (r < 0) + goto fail; + } + + r = exec_command_append(s->control_command, s->what, NULL); + if (r < 0) + goto fail; + + swap_unwatch_control_pid(s); + + r = swap_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + swap_set_state(s, SWAP_ACTIVATING); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'swapon' task: %m"); + swap_enter_dead(s, SWAP_FAILURE_RESOURCES); +} + +static void swap_enter_deactivating(Swap *s) { + int r; + + assert(s); + + s->control_command_id = SWAP_EXEC_DEACTIVATE; + s->control_command = s->exec_command + SWAP_EXEC_DEACTIVATE; + + r = exec_command_set(s->control_command, + "/sbin/swapoff", + s->what, + NULL); + if (r < 0) + goto fail; + + swap_unwatch_control_pid(s); + + r = swap_spawn(s, s->control_command, &s->control_pid); + if (r < 0) + goto fail; + + swap_set_state(s, SWAP_DEACTIVATING); + + return; + +fail: + log_unit_warning_errno(UNIT(s), r, "Failed to run 'swapoff' task: %m"); + swap_enter_active(s, SWAP_FAILURE_RESOURCES); +} + +static int swap_start(Unit *u) { + Swap *s = SWAP(u), *other; + int r; + + assert(s); + + /* We cannot fulfill this request right now, try again later + * please! */ + + if (s->state == SWAP_DEACTIVATING || + s->state == SWAP_DEACTIVATING_SIGTERM || + s->state == SWAP_DEACTIVATING_SIGKILL || + s->state == SWAP_ACTIVATING_SIGTERM || + s->state == SWAP_ACTIVATING_SIGKILL) + return -EAGAIN; + + if (s->state == SWAP_ACTIVATING) + return 0; + + assert(s->state == SWAP_DEAD || s->state == SWAP_FAILED); + + if (detect_container() > 0) + return -EPERM; + + /* If there's a job for another swap unit for the same node + * running, then let's not dispatch this one for now, and wait + * until that other job has finished. */ + LIST_FOREACH_OTHERS(same_devnode, other, s) + if (UNIT(other)->job && UNIT(other)->job->state == JOB_RUNNING) + return -EAGAIN; + + r = unit_start_limit_test(u); + if (r < 0) { + swap_enter_dead(s, SWAP_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + s->result = SWAP_SUCCESS; + s->reset_cpu_usage = true; + + swap_enter_activating(s); + return 1; +} + +static int swap_stop(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + if (s->state == SWAP_DEACTIVATING || + s->state == SWAP_DEACTIVATING_SIGTERM || + s->state == SWAP_DEACTIVATING_SIGKILL || + s->state == SWAP_ACTIVATING_SIGTERM || + s->state == SWAP_ACTIVATING_SIGKILL) + return 0; + + assert(s->state == SWAP_ACTIVATING || + s->state == SWAP_ACTIVATING_DONE || + s->state == SWAP_ACTIVE); + + if (detect_container() > 0) + return -EPERM; + + swap_enter_deactivating(s); + return 1; +} + +static int swap_serialize(Unit *u, FILE *f, FDSet *fds) { + Swap *s = SWAP(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", swap_state_to_string(s->state)); + unit_serialize_item(u, f, "result", swap_result_to_string(s->result)); + + if (s->control_pid > 0) + unit_serialize_item_format(u, f, "control-pid", PID_FMT, s->control_pid); + + if (s->control_command_id >= 0) + unit_serialize_item(u, f, "control-command", swap_exec_command_to_string(s->control_command_id)); + + return 0; +} + +static int swap_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Swap *s = SWAP(u); + + assert(s); + assert(fds); + + if (streq(key, "state")) { + SwapState state; + + state = swap_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + s->deserialized_state = state; + } else if (streq(key, "result")) { + SwapResult f; + + f = swap_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != SWAP_SUCCESS) + s->result = f; + } else if (streq(key, "control-pid")) { + pid_t pid; + + if (parse_pid(value, &pid) < 0) + log_unit_debug(u, "Failed to parse control-pid value: %s", value); + else + s->control_pid = pid; + + } else if (streq(key, "control-command")) { + SwapExecCommand id; + + id = swap_exec_command_from_string(value); + if (id < 0) + log_unit_debug(u, "Failed to parse exec-command value: %s", value); + else { + s->control_command_id = id; + s->control_command = s->exec_command + id; + } + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState swap_active_state(Unit *u) { + assert(u); + + return state_translation_table[SWAP(u)->state]; +} + +_pure_ static const char *swap_sub_state_to_string(Unit *u) { + assert(u); + + return swap_state_to_string(SWAP(u)->state); +} + +_pure_ static bool swap_check_gc(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + return s->from_proc_swaps; +} + +static void swap_sigchld_event(Unit *u, pid_t pid, int code, int status) { + Swap *s = SWAP(u); + SwapResult f; + + assert(s); + assert(pid >= 0); + + if (pid != s->control_pid) + return; + + s->control_pid = 0; + + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) + f = SWAP_SUCCESS; + else if (code == CLD_EXITED) + f = SWAP_FAILURE_EXIT_CODE; + else if (code == CLD_KILLED) + f = SWAP_FAILURE_SIGNAL; + else if (code == CLD_DUMPED) + f = SWAP_FAILURE_CORE_DUMP; + else + assert_not_reached("Unknown code"); + + if (s->result == SWAP_SUCCESS) + s->result = f; + + if (s->control_command) { + exec_status_exit(&s->control_command->exec_status, &s->exec_context, pid, code, status); + + s->control_command = NULL; + s->control_command_id = _SWAP_EXEC_COMMAND_INVALID; + } + + log_unit_full(u, f == SWAP_SUCCESS ? LOG_DEBUG : LOG_NOTICE, 0, + "Swap process exited, code=%s status=%i", sigchld_code_to_string(code), status); + + switch (s->state) { + + case SWAP_ACTIVATING: + case SWAP_ACTIVATING_DONE: + case SWAP_ACTIVATING_SIGTERM: + case SWAP_ACTIVATING_SIGKILL: + + if (f == SWAP_SUCCESS) + swap_enter_active(s, f); + else + swap_enter_dead(s, f); + break; + + case SWAP_DEACTIVATING: + case SWAP_DEACTIVATING_SIGKILL: + case SWAP_DEACTIVATING_SIGTERM: + + swap_enter_dead(s, f); + break; + + default: + assert_not_reached("Uh, control process died at wrong time."); + } + + /* Notify clients about changed exit status */ + unit_add_to_dbus_queue(u); +} + +static int swap_dispatch_timer(sd_event_source *source, usec_t usec, void *userdata) { + Swap *s = SWAP(userdata); + + assert(s); + assert(s->timer_event_source == source); + + switch (s->state) { + + case SWAP_ACTIVATING: + case SWAP_ACTIVATING_DONE: + log_unit_warning(UNIT(s), "Activation timed out. Stopping."); + swap_enter_signal(s, SWAP_ACTIVATING_SIGTERM, SWAP_FAILURE_TIMEOUT); + break; + + case SWAP_DEACTIVATING: + log_unit_warning(UNIT(s), "Deactivation timed out. Stopping."); + swap_enter_signal(s, SWAP_DEACTIVATING_SIGTERM, SWAP_FAILURE_TIMEOUT); + break; + + case SWAP_ACTIVATING_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "Activation timed out. Killing."); + swap_enter_signal(s, SWAP_ACTIVATING_SIGKILL, SWAP_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "Activation timed out. Skipping SIGKILL. Ignoring."); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + } + break; + + case SWAP_DEACTIVATING_SIGTERM: + if (s->kill_context.send_sigkill) { + log_unit_warning(UNIT(s), "Deactivation timed out. Killing."); + swap_enter_signal(s, SWAP_DEACTIVATING_SIGKILL, SWAP_FAILURE_TIMEOUT); + } else { + log_unit_warning(UNIT(s), "Deactivation timed out. Skipping SIGKILL. Ignoring."); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + } + break; + + case SWAP_ACTIVATING_SIGKILL: + case SWAP_DEACTIVATING_SIGKILL: + log_unit_warning(UNIT(s), "Swap process still around after SIGKILL. Ignoring."); + swap_enter_dead(s, SWAP_FAILURE_TIMEOUT); + break; + + default: + assert_not_reached("Timeout at wrong time."); + } + + return 0; +} + +static int swap_load_proc_swaps(Manager *m, bool set_flags) { + unsigned i; + int r = 0; + + assert(m); + + rewind(m->proc_swaps); + + (void) fscanf(m->proc_swaps, "%*s %*s %*s %*s %*s\n"); + + for (i = 1;; i++) { + _cleanup_free_ char *dev = NULL, *d = NULL; + int prio = 0, k; + + k = fscanf(m->proc_swaps, + "%ms " /* device/file */ + "%*s " /* type of swap */ + "%*s " /* swap size */ + "%*s " /* used */ + "%i\n", /* priority */ + &dev, &prio); + if (k != 2) { + if (k == EOF) + break; + + log_warning("Failed to parse /proc/swaps:%u.", i); + continue; + } + + if (cunescape(dev, UNESCAPE_RELAX, &d) < 0) + return log_oom(); + + device_found_node(m, d, true, DEVICE_FOUND_SWAP, set_flags); + + k = swap_process_new(m, d, prio, set_flags); + if (k < 0) + r = k; + } + + return r; +} + +static int swap_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + Manager *m = userdata; + Unit *u; + int r; + + assert(m); + assert(revents & EPOLLPRI); + + r = swap_load_proc_swaps(m, true); + if (r < 0) { + log_error_errno(r, "Failed to reread /proc/swaps: %m"); + + /* Reset flags, just in case, for late calls */ + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_SWAP]) { + Swap *swap = SWAP(u); + + swap->is_active = swap->just_activated = false; + } + + return 0; + } + + manager_dispatch_load_queue(m); + + LIST_FOREACH(units_by_type, u, m->units_by_type[UNIT_SWAP]) { + Swap *swap = SWAP(u); + + if (!swap->is_active) { + /* This has just been deactivated */ + + swap_unset_proc_swaps(swap); + + switch (swap->state) { + + case SWAP_ACTIVE: + swap_enter_dead(swap, SWAP_SUCCESS); + break; + + default: + /* Fire again */ + swap_set_state(swap, swap->state); + break; + } + + if (swap->what) + device_found_node(m, swap->what, false, DEVICE_FOUND_SWAP, true); + + } else if (swap->just_activated) { + + /* New swap entry */ + + switch (swap->state) { + + case SWAP_DEAD: + case SWAP_FAILED: + (void) unit_acquire_invocation_id(UNIT(swap)); + swap_enter_active(swap, SWAP_SUCCESS); + break; + + case SWAP_ACTIVATING: + swap_set_state(swap, SWAP_ACTIVATING_DONE); + break; + + default: + /* Nothing really changed, but let's + * issue an notification call + * nonetheless, in case somebody is + * waiting for this. */ + swap_set_state(swap, swap->state); + break; + } + } + + /* Reset the flags for later calls */ + swap->is_active = swap->just_activated = false; + } + + return 1; +} + +static Unit *swap_following(Unit *u) { + Swap *s = SWAP(u); + Swap *other, *first = NULL; + + assert(s); + + /* If the user configured the swap through /etc/fstab or + * a device unit, follow that. */ + + if (s->from_fragment) + return NULL; + + LIST_FOREACH_OTHERS(same_devnode, other, s) + if (other->from_fragment) + return UNIT(other); + + /* Otherwise, make everybody follow the unit that's named after + * the swap device in the kernel */ + + if (streq_ptr(s->what, s->devnode)) + return NULL; + + LIST_FOREACH_AFTER(same_devnode, other, s) + if (streq_ptr(other->what, other->devnode)) + return UNIT(other); + + LIST_FOREACH_BEFORE(same_devnode, other, s) { + if (streq_ptr(other->what, other->devnode)) + return UNIT(other); + + first = other; + } + + /* Fall back to the first on the list */ + return UNIT(first); +} + +static int swap_following_set(Unit *u, Set **_set) { + Swap *s = SWAP(u), *other; + Set *set; + int r; + + assert(s); + assert(_set); + + if (LIST_JUST_US(same_devnode, s)) { + *_set = NULL; + return 0; + } + + set = set_new(NULL); + if (!set) + return -ENOMEM; + + LIST_FOREACH_OTHERS(same_devnode, other, s) { + r = set_put(set, other); + if (r < 0) + goto fail; + } + + *_set = set; + return 1; + +fail: + set_free(set); + return r; +} + +static void swap_shutdown(Manager *m) { + assert(m); + + m->swap_event_source = sd_event_source_unref(m->swap_event_source); + + m->proc_swaps = safe_fclose(m->proc_swaps); + + m->swaps_by_devnode = hashmap_free(m->swaps_by_devnode); +} + +static void swap_enumerate(Manager *m) { + int r; + + assert(m); + + if (!m->proc_swaps) { + m->proc_swaps = fopen("/proc/swaps", "re"); + if (!m->proc_swaps) { + if (errno == ENOENT) + log_debug("Not swap enabled, skipping enumeration"); + else + log_error_errno(errno, "Failed to open /proc/swaps: %m"); + + return; + } + + r = sd_event_add_io(m->event, &m->swap_event_source, fileno(m->proc_swaps), EPOLLPRI, swap_dispatch_io, m); + if (r < 0) { + log_error_errno(r, "Failed to watch /proc/swaps: %m"); + goto fail; + } + + /* Dispatch this before we dispatch SIGCHLD, so that + * we always get the events from /proc/swaps before + * the SIGCHLD of /sbin/swapon. */ + r = sd_event_source_set_priority(m->swap_event_source, -10); + if (r < 0) { + log_error_errno(r, "Failed to change /proc/swaps priority: %m"); + goto fail; + } + + (void) sd_event_source_set_description(m->swap_event_source, "swap-proc"); + } + + r = swap_load_proc_swaps(m, false); + if (r < 0) + goto fail; + + return; + +fail: + swap_shutdown(m); +} + +int swap_process_device_new(Manager *m, struct udev_device *dev) { + struct udev_list_entry *item = NULL, *first = NULL; + _cleanup_free_ char *e = NULL; + const char *dn; + Swap *s; + int r = 0; + + assert(m); + assert(dev); + + dn = udev_device_get_devnode(dev); + if (!dn) + return 0; + + r = unit_name_from_path(dn, ".swap", &e); + if (r < 0) + return r; + + s = hashmap_get(m->units, e); + if (s) + r = swap_set_devnode(s, dn); + + first = udev_device_get_devlinks_list_entry(dev); + udev_list_entry_foreach(item, first) { + _cleanup_free_ char *n = NULL; + int q; + + q = unit_name_from_path(udev_list_entry_get_name(item), ".swap", &n); + if (q < 0) + return q; + + s = hashmap_get(m->units, n); + if (s) { + q = swap_set_devnode(s, dn); + if (q < 0) + r = q; + } + } + + return r; +} + +int swap_process_device_remove(Manager *m, struct udev_device *dev) { + const char *dn; + int r = 0; + Swap *s; + + dn = udev_device_get_devnode(dev); + if (!dn) + return 0; + + while ((s = hashmap_get(m->swaps_by_devnode, dn))) { + int q; + + q = swap_set_devnode(s, NULL); + if (q < 0) + r = q; + } + + return r; +} + +static void swap_reset_failed(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + if (s->state == SWAP_FAILED) + swap_set_state(s, SWAP_DEAD); + + s->result = SWAP_SUCCESS; +} + +static int swap_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { + return unit_kill_common(u, who, signo, -1, SWAP(u)->control_pid, error); +} + +static int swap_get_timeout(Unit *u, usec_t *timeout) { + Swap *s = SWAP(u); + usec_t t; + int r; + + if (!s->timer_event_source) + return 0; + + r = sd_event_source_get_time(s->timer_event_source, &t); + if (r < 0) + return r; + if (t == USEC_INFINITY) + return 0; + + *timeout = t; + return 1; +} + +static bool swap_supported(void) { + static int supported = -1; + + /* If swap support is not available in the kernel, or we are + * running in a container we don't support swap units, and any + * attempts to starting one should fail immediately. */ + + if (supported < 0) + supported = + access("/proc/swaps", F_OK) >= 0 && + detect_container() <= 0; + + return supported; +} + +static int swap_control_pid(Unit *u) { + Swap *s = SWAP(u); + + assert(s); + + return s->control_pid; +} + +static const char* const swap_exec_command_table[_SWAP_EXEC_COMMAND_MAX] = { + [SWAP_EXEC_ACTIVATE] = "ExecActivate", + [SWAP_EXEC_DEACTIVATE] = "ExecDeactivate", +}; + +DEFINE_STRING_TABLE_LOOKUP(swap_exec_command, SwapExecCommand); + +static const char* const swap_result_table[_SWAP_RESULT_MAX] = { + [SWAP_SUCCESS] = "success", + [SWAP_FAILURE_RESOURCES] = "resources", + [SWAP_FAILURE_TIMEOUT] = "timeout", + [SWAP_FAILURE_EXIT_CODE] = "exit-code", + [SWAP_FAILURE_SIGNAL] = "signal", + [SWAP_FAILURE_CORE_DUMP] = "core-dump", + [SWAP_FAILURE_START_LIMIT_HIT] = "start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(swap_result, SwapResult); + +const UnitVTable swap_vtable = { + .object_size = sizeof(Swap), + .exec_context_offset = offsetof(Swap, exec_context), + .cgroup_context_offset = offsetof(Swap, cgroup_context), + .kill_context_offset = offsetof(Swap, kill_context), + .exec_runtime_offset = offsetof(Swap, exec_runtime), + .dynamic_creds_offset = offsetof(Swap, dynamic_creds), + + .sections = + "Unit\0" + "Swap\0" + "Install\0", + .private_section = "Swap", + + .init = swap_init, + .load = swap_load, + .done = swap_done, + + .coldplug = swap_coldplug, + + .dump = swap_dump, + + .start = swap_start, + .stop = swap_stop, + + .kill = swap_kill, + + .get_timeout = swap_get_timeout, + + .serialize = swap_serialize, + .deserialize_item = swap_deserialize_item, + + .active_state = swap_active_state, + .sub_state_to_string = swap_sub_state_to_string, + + .check_gc = swap_check_gc, + + .sigchld_event = swap_sigchld_event, + + .reset_failed = swap_reset_failed, + + .control_pid = swap_control_pid, + + .bus_vtable = bus_swap_vtable, + .bus_set_property = bus_swap_set_property, + .bus_commit_properties = bus_swap_commit_properties, + + .following = swap_following, + .following_set = swap_following_set, + + .enumerate = swap_enumerate, + .shutdown = swap_shutdown, + .supported = swap_supported, + + .status_message_formats = { + .starting_stopping = { + [0] = "Activating swap %s...", + [1] = "Deactivating swap %s...", + }, + .finished_start_job = { + [JOB_DONE] = "Activated swap %s.", + [JOB_FAILED] = "Failed to activate swap %s.", + [JOB_TIMEOUT] = "Timed out activating swap %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Deactivated swap %s.", + [JOB_FAILED] = "Failed deactivating swap %s.", + [JOB_TIMEOUT] = "Timed out deactivating swap %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/target.c b/src/grp-system/libcore/src/target.c new file mode 100644 index 0000000000..1e5212d75d --- /dev/null +++ b/src/grp-system/libcore/src/target.c @@ -0,0 +1,229 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/target.h" +#include "core/unit.h" +#include "systemd-basic/log.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" + +#include "dbus-target.h" + +static const UnitActiveState state_translation_table[_TARGET_STATE_MAX] = { + [TARGET_DEAD] = UNIT_INACTIVE, + [TARGET_ACTIVE] = UNIT_ACTIVE +}; + +static void target_set_state(Target *t, TargetState state) { + TargetState old_state; + assert(t); + + old_state = t->state; + t->state = state; + + if (state != old_state) + log_debug("%s changed %s -> %s", + UNIT(t)->id, + target_state_to_string(old_state), + target_state_to_string(state)); + + unit_notify(UNIT(t), state_translation_table[old_state], state_translation_table[state], true); +} + +static int target_add_default_dependencies(Target *t) { + + static const UnitDependency deps[] = { + UNIT_REQUIRES, + UNIT_REQUISITE, + UNIT_WANTS, + UNIT_BINDS_TO, + UNIT_PART_OF + }; + + Iterator i; + Unit *other; + int r; + unsigned k; + + assert(t); + + /* Imply ordering for requirement dependencies on target + * units. Note that when the user created a contradicting + * ordering manually we won't add anything in here to make + * sure we don't create a loop. */ + + for (k = 0; k < ELEMENTSOF(deps); k++) + SET_FOREACH(other, UNIT(t)->dependencies[deps[k]], i) { + r = unit_add_default_target_dependency(other, UNIT(t)); + if (r < 0) + return r; + } + + /* Make sure targets are unloaded on shutdown */ + return unit_add_dependency_by_name(UNIT(t), UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int target_load(Unit *u) { + Target *t = TARGET(u); + int r; + + assert(t); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + /* This is a new unit? Then let's add in some extras */ + if (u->load_state == UNIT_LOADED && u->default_dependencies) { + r = target_add_default_dependencies(t); + if (r < 0) + return r; + } + + return 0; +} + +static int target_coldplug(Unit *u) { + Target *t = TARGET(u); + + assert(t); + assert(t->state == TARGET_DEAD); + + if (t->deserialized_state != t->state) + target_set_state(t, t->deserialized_state); + + return 0; +} + +static void target_dump(Unit *u, FILE *f, const char *prefix) { + Target *t = TARGET(u); + + assert(t); + assert(f); + + fprintf(f, + "%sTarget State: %s\n", + prefix, target_state_to_string(t->state)); +} + +static int target_start(Unit *u) { + Target *t = TARGET(u); + int r; + + assert(t); + assert(t->state == TARGET_DEAD); + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + target_set_state(t, TARGET_ACTIVE); + return 1; +} + +static int target_stop(Unit *u) { + Target *t = TARGET(u); + + assert(t); + assert(t->state == TARGET_ACTIVE); + + target_set_state(t, TARGET_DEAD); + return 1; +} + +static int target_serialize(Unit *u, FILE *f, FDSet *fds) { + Target *s = TARGET(u); + + assert(s); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", target_state_to_string(s->state)); + return 0; +} + +static int target_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Target *s = TARGET(u); + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + TargetState state; + + state = target_state_from_string(value); + if (state < 0) + log_debug("Failed to parse state value %s", value); + else + s->deserialized_state = state; + + } else + log_debug("Unknown serialization key '%s'", key); + + return 0; +} + +_pure_ static UnitActiveState target_active_state(Unit *u) { + assert(u); + + return state_translation_table[TARGET(u)->state]; +} + +_pure_ static const char *target_sub_state_to_string(Unit *u) { + assert(u); + + return target_state_to_string(TARGET(u)->state); +} + +const UnitVTable target_vtable = { + .object_size = sizeof(Target), + + .sections = + "Unit\0" + "Target\0" + "Install\0", + + .load = target_load, + .coldplug = target_coldplug, + + .dump = target_dump, + + .start = target_start, + .stop = target_stop, + + .serialize = target_serialize, + .deserialize_item = target_deserialize_item, + + .active_state = target_active_state, + .sub_state_to_string = target_sub_state_to_string, + + .bus_vtable = bus_target_vtable, + + .status_message_formats = { + .finished_start_job = { + [JOB_DONE] = "Reached target %s.", + }, + .finished_stop_job = { + [JOB_DONE] = "Stopped target %s.", + }, + }, +}; diff --git a/src/grp-system/libcore/src/timer.c b/src/grp-system/libcore/src/timer.c new file mode 100644 index 0000000000..8789dce022 --- /dev/null +++ b/src/grp-system/libcore/src/timer.c @@ -0,0 +1,866 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> + +#include "core/timer.h" +#include "core/unit.h" +#include "sd-bus/bus-error.h" +#include "sd-bus/bus-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/fs-util.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/random-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/string-table.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/virt.h" + +#include "dbus-timer.h" + +static const UnitActiveState state_translation_table[_TIMER_STATE_MAX] = { + [TIMER_DEAD] = UNIT_INACTIVE, + [TIMER_WAITING] = UNIT_ACTIVE, + [TIMER_RUNNING] = UNIT_ACTIVE, + [TIMER_ELAPSED] = UNIT_ACTIVE, + [TIMER_FAILED] = UNIT_FAILED +}; + +static int timer_dispatch(sd_event_source *s, uint64_t usec, void *userdata); + +static void timer_init(Unit *u) { + Timer *t = TIMER(u); + + assert(u); + assert(u->load_state == UNIT_STUB); + + t->next_elapse_monotonic_or_boottime = USEC_INFINITY; + t->next_elapse_realtime = USEC_INFINITY; + t->accuracy_usec = u->manager->default_timer_accuracy_usec; + t->remain_after_elapse = true; +} + +void timer_free_values(Timer *t) { + TimerValue *v; + + assert(t); + + while ((v = t->values)) { + LIST_REMOVE(value, t->values, v); + calendar_spec_free(v->calendar_spec); + free(v); + } +} + +static void timer_done(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + + timer_free_values(t); + + t->monotonic_event_source = sd_event_source_unref(t->monotonic_event_source); + t->realtime_event_source = sd_event_source_unref(t->realtime_event_source); + + free(t->stamp_path); +} + +static int timer_verify(Timer *t) { + assert(t); + + if (UNIT(t)->load_state != UNIT_LOADED) + return 0; + + if (!t->values) { + log_unit_error(UNIT(t), "Timer unit lacks value setting. Refusing."); + return -EINVAL; + } + + return 0; +} + +static int timer_add_default_dependencies(Timer *t) { + int r; + TimerValue *v; + + assert(t); + + if (!UNIT(t)->default_dependencies) + return 0; + + r = unit_add_dependency_by_name(UNIT(t), UNIT_BEFORE, SPECIAL_TIMERS_TARGET, NULL, true); + if (r < 0) + return r; + + if (MANAGER_IS_SYSTEM(UNIT(t)->manager)) { + r = unit_add_two_dependencies_by_name(UNIT(t), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_SYSINIT_TARGET, NULL, true); + if (r < 0) + return r; + + LIST_FOREACH(value, v, t->values) { + if (v->base == TIMER_CALENDAR) { + r = unit_add_dependency_by_name(UNIT(t), UNIT_AFTER, SPECIAL_TIME_SYNC_TARGET, NULL, true); + if (r < 0) + return r; + break; + } + } + } + + return unit_add_two_dependencies_by_name(UNIT(t), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); +} + +static int timer_setup_persistent(Timer *t) { + int r; + + assert(t); + + if (!t->persistent) + return 0; + + if (MANAGER_IS_SYSTEM(UNIT(t)->manager)) { + + r = unit_require_mounts_for(UNIT(t), "/var/lib/systemd/timers"); + if (r < 0) + return r; + + t->stamp_path = strappend("/var/lib/systemd/timers/stamp-", UNIT(t)->id); + } else { + const char *e; + + e = getenv("XDG_DATA_HOME"); + if (e) + t->stamp_path = strjoin(e, "/systemd/timers/stamp-", UNIT(t)->id, NULL); + else { + + _cleanup_free_ char *h = NULL; + + r = get_home_dir(&h); + if (r < 0) + return log_unit_error_errno(UNIT(t), r, "Failed to determine home directory: %m"); + + t->stamp_path = strjoin(h, "/.local/share/systemd/timers/stamp-", UNIT(t)->id, NULL); + } + } + + if (!t->stamp_path) + return log_oom(); + + return 0; +} + +static int timer_load(Unit *u) { + Timer *t = TIMER(u); + int r; + + assert(u); + assert(u->load_state == UNIT_STUB); + + r = unit_load_fragment_and_dropin(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_LOADED) { + + if (set_isempty(u->dependencies[UNIT_TRIGGERS])) { + Unit *x; + + r = unit_load_related_unit(u, ".service", &x); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_BEFORE, UNIT_TRIGGERS, x, true); + if (r < 0) + return r; + } + + r = timer_setup_persistent(t); + if (r < 0) + return r; + + r = timer_add_default_dependencies(t); + if (r < 0) + return r; + } + + return timer_verify(t); +} + +static void timer_dump(Unit *u, FILE *f, const char *prefix) { + char buf[FORMAT_TIMESPAN_MAX]; + Timer *t = TIMER(u); + Unit *trigger; + TimerValue *v; + + trigger = UNIT_TRIGGER(u); + + fprintf(f, + "%sTimer State: %s\n" + "%sResult: %s\n" + "%sUnit: %s\n" + "%sPersistent: %s\n" + "%sWakeSystem: %s\n" + "%sAccuracy: %s\n" + "%sRemainAfterElapse: %s\n", + prefix, timer_state_to_string(t->state), + prefix, timer_result_to_string(t->result), + prefix, trigger ? trigger->id : "n/a", + prefix, yes_no(t->persistent), + prefix, yes_no(t->wake_system), + prefix, format_timespan(buf, sizeof(buf), t->accuracy_usec, 1), + prefix, yes_no(t->remain_after_elapse)); + + LIST_FOREACH(value, v, t->values) { + + if (v->base == TIMER_CALENDAR) { + _cleanup_free_ char *p = NULL; + + calendar_spec_to_string(v->calendar_spec, &p); + + fprintf(f, + "%s%s: %s\n", + prefix, + timer_base_to_string(v->base), + strna(p)); + } else { + char timespan1[FORMAT_TIMESPAN_MAX]; + + fprintf(f, + "%s%s: %s\n", + prefix, + timer_base_to_string(v->base), + format_timespan(timespan1, sizeof(timespan1), v->value, 0)); + } + } +} + +static void timer_set_state(Timer *t, TimerState state) { + TimerState old_state; + assert(t); + + old_state = t->state; + t->state = state; + + if (state != TIMER_WAITING) { + t->monotonic_event_source = sd_event_source_unref(t->monotonic_event_source); + t->realtime_event_source = sd_event_source_unref(t->realtime_event_source); + t->next_elapse_monotonic_or_boottime = USEC_INFINITY; + t->next_elapse_realtime = USEC_INFINITY; + } + + if (state != old_state) + log_unit_debug(UNIT(t), "Changed %s -> %s", timer_state_to_string(old_state), timer_state_to_string(state)); + + unit_notify(UNIT(t), state_translation_table[old_state], state_translation_table[state], true); +} + +static void timer_enter_waiting(Timer *t, bool initial); + +static int timer_coldplug(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + assert(t->state == TIMER_DEAD); + + if (t->deserialized_state == t->state) + return 0; + + if (t->deserialized_state == TIMER_WAITING) + timer_enter_waiting(t, false); + else + timer_set_state(t, t->deserialized_state); + + return 0; +} + +static void timer_enter_dead(Timer *t, TimerResult f) { + assert(t); + + if (t->result == TIMER_SUCCESS) + t->result = f; + + timer_set_state(t, t->result != TIMER_SUCCESS ? TIMER_FAILED : TIMER_DEAD); +} + +static void timer_enter_elapsed(Timer *t, bool leave_around) { + assert(t); + + /* If a unit is marked with RemainAfterElapse=yes we leave it + * around even after it elapsed once, so that starting it + * later again does not necessarily mean immediate + * retriggering. We unconditionally leave units with + * TIMER_UNIT_ACTIVE or TIMER_UNIT_INACTIVE triggers around, + * since they might be restarted automatically at any time + * later on. */ + + if (t->remain_after_elapse || leave_around) + timer_set_state(t, TIMER_ELAPSED); + else + timer_enter_dead(t, TIMER_SUCCESS); +} + +static usec_t monotonic_to_boottime(usec_t t) { + usec_t a, b; + + if (t <= 0) + return 0; + + a = now(clock_boottime_or_monotonic()); + b = now(CLOCK_MONOTONIC); + + if (t + a > b) + return t + a - b; + else + return 0; +} + +static void add_random(Timer *t, usec_t *v) { + char s[FORMAT_TIMESPAN_MAX]; + usec_t add; + + assert(t); + assert(v); + + if (t->random_usec == 0) + return; + if (*v == USEC_INFINITY) + return; + + add = random_u64() % t->random_usec; + + if (*v + add < *v) /* overflow */ + *v = (usec_t) -2; /* Highest possible value, that is not USEC_INFINITY */ + else + *v += add; + + log_unit_info(UNIT(t), "Adding %s random time.", format_timespan(s, sizeof(s), add, 0)); +} + +static void timer_enter_waiting(Timer *t, bool initial) { + bool found_monotonic = false, found_realtime = false; + usec_t ts_realtime, ts_monotonic; + usec_t base = 0; + bool leave_around = false; + TimerValue *v; + Unit *trigger; + int r; + + assert(t); + + trigger = UNIT_TRIGGER(UNIT(t)); + if (!trigger) { + log_unit_error(UNIT(t), "Unit to trigger vanished."); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); + return; + } + + /* If we shall wake the system we use the boottime clock + * rather than the monotonic clock. */ + + ts_realtime = now(CLOCK_REALTIME); + ts_monotonic = now(t->wake_system ? clock_boottime_or_monotonic() : CLOCK_MONOTONIC); + t->next_elapse_monotonic_or_boottime = t->next_elapse_realtime = 0; + + LIST_FOREACH(value, v, t->values) { + + if (v->disabled) + continue; + + if (v->base == TIMER_CALENDAR) { + usec_t b; + + /* If we know the last time this was + * triggered, schedule the job based relative + * to that. If we don't just start from + * now. */ + + b = t->last_trigger.realtime > 0 ? t->last_trigger.realtime : ts_realtime; + + r = calendar_spec_next_usec(v->calendar_spec, b, &v->next_elapse); + if (r < 0) + continue; + + if (!found_realtime) + t->next_elapse_realtime = v->next_elapse; + else + t->next_elapse_realtime = MIN(t->next_elapse_realtime, v->next_elapse); + + found_realtime = true; + + } else { + switch (v->base) { + + case TIMER_ACTIVE: + if (state_translation_table[t->state] == UNIT_ACTIVE) + base = UNIT(t)->inactive_exit_timestamp.monotonic; + else + base = ts_monotonic; + break; + + case TIMER_BOOT: + if (detect_container() <= 0) { + /* CLOCK_MONOTONIC equals the uptime on Linux */ + base = 0; + break; + } + /* In a container we don't want to include the time the host + * was already up when the container started, so count from + * our own startup. Fall through. */ + case TIMER_STARTUP: + base = UNIT(t)->manager->userspace_timestamp.monotonic; + break; + + case TIMER_UNIT_ACTIVE: + leave_around = true; + base = trigger->inactive_exit_timestamp.monotonic; + + if (base <= 0) + base = t->last_trigger.monotonic; + + if (base <= 0) + continue; + + break; + + case TIMER_UNIT_INACTIVE: + leave_around = true; + base = trigger->inactive_enter_timestamp.monotonic; + + if (base <= 0) + base = t->last_trigger.monotonic; + + if (base <= 0) + continue; + + break; + + default: + assert_not_reached("Unknown timer base"); + } + + if (t->wake_system) + base = monotonic_to_boottime(base); + + v->next_elapse = base + v->value; + + if (!initial && v->next_elapse < ts_monotonic && IN_SET(v->base, TIMER_ACTIVE, TIMER_BOOT, TIMER_STARTUP)) { + /* This is a one time trigger, disable it now */ + v->disabled = true; + continue; + } + + if (!found_monotonic) + t->next_elapse_monotonic_or_boottime = v->next_elapse; + else + t->next_elapse_monotonic_or_boottime = MIN(t->next_elapse_monotonic_or_boottime, v->next_elapse); + + found_monotonic = true; + } + } + + if (!found_monotonic && !found_realtime) { + log_unit_debug(UNIT(t), "Timer is elapsed."); + timer_enter_elapsed(t, leave_around); + return; + } + + if (found_monotonic) { + char buf[FORMAT_TIMESPAN_MAX]; + usec_t left; + + add_random(t, &t->next_elapse_monotonic_or_boottime); + + left = t->next_elapse_monotonic_or_boottime > ts_monotonic ? t->next_elapse_monotonic_or_boottime - ts_monotonic : 0; + log_unit_debug(UNIT(t), "Monotonic timer elapses in %s.", format_timespan(buf, sizeof(buf), left, 0)); + + if (t->monotonic_event_source) { + r = sd_event_source_set_time(t->monotonic_event_source, t->next_elapse_monotonic_or_boottime); + if (r < 0) + goto fail; + + r = sd_event_source_set_enabled(t->monotonic_event_source, SD_EVENT_ONESHOT); + if (r < 0) + goto fail; + } else { + + r = sd_event_add_time( + UNIT(t)->manager->event, + &t->monotonic_event_source, + t->wake_system ? CLOCK_BOOTTIME_ALARM : CLOCK_MONOTONIC, + t->next_elapse_monotonic_or_boottime, t->accuracy_usec, + timer_dispatch, t); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(t->monotonic_event_source, "timer-monotonic"); + } + + } else if (t->monotonic_event_source) { + + r = sd_event_source_set_enabled(t->monotonic_event_source, SD_EVENT_OFF); + if (r < 0) + goto fail; + } + + if (found_realtime) { + char buf[FORMAT_TIMESTAMP_MAX]; + + add_random(t, &t->next_elapse_realtime); + + log_unit_debug(UNIT(t), "Realtime timer elapses at %s.", format_timestamp(buf, sizeof(buf), t->next_elapse_realtime)); + + if (t->realtime_event_source) { + r = sd_event_source_set_time(t->realtime_event_source, t->next_elapse_realtime); + if (r < 0) + goto fail; + + r = sd_event_source_set_enabled(t->realtime_event_source, SD_EVENT_ONESHOT); + if (r < 0) + goto fail; + } else { + r = sd_event_add_time( + UNIT(t)->manager->event, + &t->realtime_event_source, + t->wake_system ? CLOCK_REALTIME_ALARM : CLOCK_REALTIME, + t->next_elapse_realtime, t->accuracy_usec, + timer_dispatch, t); + if (r < 0) + goto fail; + + (void) sd_event_source_set_description(t->realtime_event_source, "timer-realtime"); + } + + } else if (t->realtime_event_source) { + + r = sd_event_source_set_enabled(t->realtime_event_source, SD_EVENT_OFF); + if (r < 0) + goto fail; + } + + timer_set_state(t, TIMER_WAITING); + return; + +fail: + log_unit_warning_errno(UNIT(t), r, "Failed to enter waiting state: %m"); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); +} + +static void timer_enter_running(Timer *t) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + Unit *trigger; + int r; + + assert(t); + + /* Don't start job if we are supposed to go down */ + if (unit_stop_pending(UNIT(t))) + return; + + trigger = UNIT_TRIGGER(UNIT(t)); + if (!trigger) { + log_unit_error(UNIT(t), "Unit to trigger vanished."); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); + return; + } + + r = manager_add_job(UNIT(t)->manager, JOB_START, trigger, JOB_REPLACE, &error, NULL); + if (r < 0) + goto fail; + + dual_timestamp_get(&t->last_trigger); + + if (t->stamp_path) + touch_file(t->stamp_path, true, t->last_trigger.realtime, UID_INVALID, GID_INVALID, MODE_INVALID); + + timer_set_state(t, TIMER_RUNNING); + return; + +fail: + log_unit_warning(UNIT(t), "Failed to queue unit startup job: %s", bus_error_message(&error, r)); + timer_enter_dead(t, TIMER_FAILURE_RESOURCES); +} + +static int timer_start(Unit *u) { + Timer *t = TIMER(u); + TimerValue *v; + Unit *trigger; + int r; + + assert(t); + assert(t->state == TIMER_DEAD || t->state == TIMER_FAILED); + + trigger = UNIT_TRIGGER(u); + if (!trigger || trigger->load_state != UNIT_LOADED) { + log_unit_error(u, "Refusing to start, unit to trigger not loaded."); + return -ENOENT; + } + + r = unit_start_limit_test(u); + if (r < 0) { + timer_enter_dead(t, TIMER_FAILURE_START_LIMIT_HIT); + return r; + } + + r = unit_acquire_invocation_id(u); + if (r < 0) + return r; + + t->last_trigger = DUAL_TIMESTAMP_NULL; + + /* Reenable all timers that depend on unit activation time */ + LIST_FOREACH(value, v, t->values) + if (v->base == TIMER_ACTIVE) + v->disabled = false; + + if (t->stamp_path) { + struct stat st; + + if (stat(t->stamp_path, &st) >= 0) + t->last_trigger.realtime = timespec_load(&st.st_atim); + else if (errno == ENOENT) + /* The timer has never run before, + * make sure a stamp file exists. + */ + (void) touch_file(t->stamp_path, true, USEC_INFINITY, UID_INVALID, GID_INVALID, MODE_INVALID); + } + + t->result = TIMER_SUCCESS; + timer_enter_waiting(t, true); + return 1; +} + +static int timer_stop(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + assert(t->state == TIMER_WAITING || t->state == TIMER_RUNNING || t->state == TIMER_ELAPSED); + + timer_enter_dead(t, TIMER_SUCCESS); + return 1; +} + +static int timer_serialize(Unit *u, FILE *f, FDSet *fds) { + Timer *t = TIMER(u); + + assert(u); + assert(f); + assert(fds); + + unit_serialize_item(u, f, "state", timer_state_to_string(t->state)); + unit_serialize_item(u, f, "result", timer_result_to_string(t->result)); + + if (t->last_trigger.realtime > 0) + unit_serialize_item_format(u, f, "last-trigger-realtime", "%" PRIu64, t->last_trigger.realtime); + + if (t->last_trigger.monotonic > 0) + unit_serialize_item_format(u, f, "last-trigger-monotonic", "%" PRIu64, t->last_trigger.monotonic); + + return 0; +} + +static int timer_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { + Timer *t = TIMER(u); + int r; + + assert(u); + assert(key); + assert(value); + assert(fds); + + if (streq(key, "state")) { + TimerState state; + + state = timer_state_from_string(value); + if (state < 0) + log_unit_debug(u, "Failed to parse state value: %s", value); + else + t->deserialized_state = state; + } else if (streq(key, "result")) { + TimerResult f; + + f = timer_result_from_string(value); + if (f < 0) + log_unit_debug(u, "Failed to parse result value: %s", value); + else if (f != TIMER_SUCCESS) + t->result = f; + } else if (streq(key, "last-trigger-realtime")) { + + r = safe_atou64(value, &t->last_trigger.realtime); + if (r < 0) + log_unit_debug(u, "Failed to parse last-trigger-realtime value: %s", value); + + } else if (streq(key, "last-trigger-monotonic")) { + + r = safe_atou64(value, &t->last_trigger.monotonic); + if (r < 0) + log_unit_debug(u, "Failed to parse last-trigger-monotonic value: %s", value); + + } else + log_unit_debug(u, "Unknown serialization key: %s", key); + + return 0; +} + +_pure_ static UnitActiveState timer_active_state(Unit *u) { + assert(u); + + return state_translation_table[TIMER(u)->state]; +} + +_pure_ static const char *timer_sub_state_to_string(Unit *u) { + assert(u); + + return timer_state_to_string(TIMER(u)->state); +} + +static int timer_dispatch(sd_event_source *s, uint64_t usec, void *userdata) { + Timer *t = TIMER(userdata); + + assert(t); + + if (t->state != TIMER_WAITING) + return 0; + + log_unit_debug(UNIT(t), "Timer elapsed."); + timer_enter_running(t); + return 0; +} + +static void timer_trigger_notify(Unit *u, Unit *other) { + Timer *t = TIMER(u); + TimerValue *v; + + assert(u); + assert(other); + + if (other->load_state != UNIT_LOADED) + return; + + /* Reenable all timers that depend on unit state */ + LIST_FOREACH(value, v, t->values) + if (v->base == TIMER_UNIT_ACTIVE || + v->base == TIMER_UNIT_INACTIVE) + v->disabled = false; + + switch (t->state) { + + case TIMER_WAITING: + case TIMER_ELAPSED: + + /* Recalculate sleep time */ + timer_enter_waiting(t, false); + break; + + case TIMER_RUNNING: + + if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) { + log_unit_debug(UNIT(t), "Got notified about unit deactivation."); + timer_enter_waiting(t, false); + } + break; + + case TIMER_DEAD: + case TIMER_FAILED: + break; + + default: + assert_not_reached("Unknown timer state"); + } +} + +static void timer_reset_failed(Unit *u) { + Timer *t = TIMER(u); + + assert(t); + + if (t->state == TIMER_FAILED) + timer_set_state(t, TIMER_DEAD); + + t->result = TIMER_SUCCESS; +} + +static void timer_time_change(Unit *u) { + Timer *t = TIMER(u); + + assert(u); + + if (t->state != TIMER_WAITING) + return; + + log_unit_debug(u, "Time change, recalculating next elapse."); + timer_enter_waiting(t, false); +} + +static const char* const timer_base_table[_TIMER_BASE_MAX] = { + [TIMER_ACTIVE] = "OnActiveSec", + [TIMER_BOOT] = "OnBootSec", + [TIMER_STARTUP] = "OnStartupSec", + [TIMER_UNIT_ACTIVE] = "OnUnitActiveSec", + [TIMER_UNIT_INACTIVE] = "OnUnitInactiveSec", + [TIMER_CALENDAR] = "OnCalendar" +}; + +DEFINE_STRING_TABLE_LOOKUP(timer_base, TimerBase); + +static const char* const timer_result_table[_TIMER_RESULT_MAX] = { + [TIMER_SUCCESS] = "success", + [TIMER_FAILURE_RESOURCES] = "resources", + [TIMER_FAILURE_START_LIMIT_HIT] = "start-limit-hit", +}; + +DEFINE_STRING_TABLE_LOOKUP(timer_result, TimerResult); + +const UnitVTable timer_vtable = { + .object_size = sizeof(Timer), + + .sections = + "Unit\0" + "Timer\0" + "Install\0", + .private_section = "Timer", + + .init = timer_init, + .done = timer_done, + .load = timer_load, + + .coldplug = timer_coldplug, + + .dump = timer_dump, + + .start = timer_start, + .stop = timer_stop, + + .serialize = timer_serialize, + .deserialize_item = timer_deserialize_item, + + .active_state = timer_active_state, + .sub_state_to_string = timer_sub_state_to_string, + + .trigger_notify = timer_trigger_notify, + + .reset_failed = timer_reset_failed, + .time_change = timer_time_change, + + .bus_vtable = bus_timer_vtable, + .bus_set_property = bus_timer_set_property, + + .can_transient = true, +}; diff --git a/src/grp-system/libcore/src/transaction.c b/src/grp-system/libcore/src/transaction.c new file mode 100644 index 0000000000..b9b28900e9 --- /dev/null +++ b/src/grp-system/libcore/src/transaction.c @@ -0,0 +1,1101 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <fcntl.h> +#include <unistd.h> + +#include "sd-bus/bus-common-errors.h" +#include "sd-bus/bus-error.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/terminal-util.h" + +#include "dbus-unit.h" +#include "transaction.h" + +static void transaction_unlink_job(Transaction *tr, Job *j, bool delete_dependencies); + +static void transaction_delete_job(Transaction *tr, Job *j, bool delete_dependencies) { + assert(tr); + assert(j); + + /* Deletes one job from the transaction */ + + transaction_unlink_job(tr, j, delete_dependencies); + + job_free(j); +} + +static void transaction_delete_unit(Transaction *tr, Unit *u) { + Job *j; + + /* Deletes all jobs associated with a certain unit from the + * transaction */ + + while ((j = hashmap_get(tr->jobs, u))) + transaction_delete_job(tr, j, true); +} + +void transaction_abort(Transaction *tr) { + Job *j; + + assert(tr); + + while ((j = hashmap_first(tr->jobs))) + transaction_delete_job(tr, j, false); + + assert(hashmap_isempty(tr->jobs)); +} + +static void transaction_find_jobs_that_matter_to_anchor(Job *j, unsigned generation) { + JobDependency *l; + + /* A recursive sweep through the graph that marks all units + * that matter to the anchor job, i.e. are directly or + * indirectly a dependency of the anchor job via paths that + * are fully marked as mattering. */ + + j->matters_to_anchor = true; + j->generation = generation; + + LIST_FOREACH(subject, l, j->subject_list) { + + /* This link does not matter */ + if (!l->matters) + continue; + + /* This unit has already been marked */ + if (l->object->generation == generation) + continue; + + transaction_find_jobs_that_matter_to_anchor(l->object, generation); + } +} + +static void transaction_merge_and_delete_job(Transaction *tr, Job *j, Job *other, JobType t) { + JobDependency *l, *last; + + assert(j); + assert(other); + assert(j->unit == other->unit); + assert(!j->installed); + + /* Merges 'other' into 'j' and then deletes 'other'. */ + + j->type = t; + j->state = JOB_WAITING; + j->irreversible = j->irreversible || other->irreversible; + j->matters_to_anchor = j->matters_to_anchor || other->matters_to_anchor; + + /* Patch us in as new owner of the JobDependency objects */ + last = NULL; + LIST_FOREACH(subject, l, other->subject_list) { + assert(l->subject == other); + l->subject = j; + last = l; + } + + /* Merge both lists */ + if (last) { + last->subject_next = j->subject_list; + if (j->subject_list) + j->subject_list->subject_prev = last; + j->subject_list = other->subject_list; + } + + /* Patch us in as new owner of the JobDependency objects */ + last = NULL; + LIST_FOREACH(object, l, other->object_list) { + assert(l->object == other); + l->object = j; + last = l; + } + + /* Merge both lists */ + if (last) { + last->object_next = j->object_list; + if (j->object_list) + j->object_list->object_prev = last; + j->object_list = other->object_list; + } + + /* Kill the other job */ + other->subject_list = NULL; + other->object_list = NULL; + transaction_delete_job(tr, other, true); +} + +_pure_ static bool job_is_conflicted_by(Job *j) { + JobDependency *l; + + assert(j); + + /* Returns true if this job is pulled in by a least one + * ConflictedBy dependency. */ + + LIST_FOREACH(object, l, j->object_list) + if (l->conflicts) + return true; + + return false; +} + +static int delete_one_unmergeable_job(Transaction *tr, Job *j) { + Job *k; + + assert(j); + + /* Tries to delete one item in the linked list + * j->transaction_next->transaction_next->... that conflicts + * with another one, in an attempt to make an inconsistent + * transaction work. */ + + /* We rely here on the fact that if a merged with b does not + * merge with c, either a or b merge with c neither */ + LIST_FOREACH(transaction, j, j) + LIST_FOREACH(transaction, k, j->transaction_next) { + Job *d; + + /* Is this one mergeable? Then skip it */ + if (job_type_is_mergeable(j->type, k->type)) + continue; + + /* Ok, we found two that conflict, let's see if we can + * drop one of them */ + if (!j->matters_to_anchor && !k->matters_to_anchor) { + + /* Both jobs don't matter, so let's + * find the one that is smarter to + * remove. Let's think positive and + * rather remove stops then starts -- + * except if something is being + * stopped because it is conflicted by + * another unit in which case we + * rather remove the start. */ + + log_unit_debug(j->unit, + "Looking at job %s/%s conflicted_by=%s", + j->unit->id, job_type_to_string(j->type), + yes_no(j->type == JOB_STOP && job_is_conflicted_by(j))); + log_unit_debug(k->unit, + "Looking at job %s/%s conflicted_by=%s", + k->unit->id, job_type_to_string(k->type), + yes_no(k->type == JOB_STOP && job_is_conflicted_by(k))); + + if (j->type == JOB_STOP) { + + if (job_is_conflicted_by(j)) + d = k; + else + d = j; + + } else if (k->type == JOB_STOP) { + + if (job_is_conflicted_by(k)) + d = j; + else + d = k; + } else + d = j; + + } else if (!j->matters_to_anchor) + d = j; + else if (!k->matters_to_anchor) + d = k; + else + return -ENOEXEC; + + /* Ok, we can drop one, so let's do so. */ + log_unit_debug(d->unit, + "Fixing conflicting jobs %s/%s,%s/%s by deleting job %s/%s", + j->unit->id, job_type_to_string(j->type), + k->unit->id, job_type_to_string(k->type), + d->unit->id, job_type_to_string(d->type)); + transaction_delete_job(tr, d, true); + return 0; + } + + return -EINVAL; +} + +static int transaction_merge_jobs(Transaction *tr, sd_bus_error *e) { + Job *j; + Iterator i; + int r; + + assert(tr); + + /* First step, check whether any of the jobs for one specific + * task conflict. If so, try to drop one of them. */ + HASHMAP_FOREACH(j, tr->jobs, i) { + JobType t; + Job *k; + + t = j->type; + LIST_FOREACH(transaction, k, j->transaction_next) { + if (job_type_merge_and_collapse(&t, k->type, j->unit) >= 0) + continue; + + /* OK, we could not merge all jobs for this + * action. Let's see if we can get rid of one + * of them */ + + r = delete_one_unmergeable_job(tr, j); + if (r >= 0) + /* Ok, we managed to drop one, now + * let's ask our callers to call us + * again after garbage collecting */ + return -EAGAIN; + + /* We couldn't merge anything. Failure */ + return sd_bus_error_setf(e, BUS_ERROR_TRANSACTION_JOBS_CONFLICTING, + "Transaction contains conflicting jobs '%s' and '%s' for %s. " + "Probably contradicting requirement dependencies configured.", + job_type_to_string(t), + job_type_to_string(k->type), + k->unit->id); + } + } + + /* Second step, merge the jobs. */ + HASHMAP_FOREACH(j, tr->jobs, i) { + JobType t = j->type; + Job *k; + + /* Merge all transaction jobs for j->unit */ + LIST_FOREACH(transaction, k, j->transaction_next) + assert_se(job_type_merge_and_collapse(&t, k->type, j->unit) == 0); + + while ((k = j->transaction_next)) { + if (tr->anchor_job == k) { + transaction_merge_and_delete_job(tr, k, j, t); + j = k; + } else + transaction_merge_and_delete_job(tr, j, k, t); + } + + assert(!j->transaction_next); + assert(!j->transaction_prev); + } + + return 0; +} + +static void transaction_drop_redundant(Transaction *tr) { + Job *j; + Iterator i; + + /* Goes through the transaction and removes all jobs of the units + * whose jobs are all noops. If not all of a unit's jobs are + * redundant, they are kept. */ + + assert(tr); + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + Job *k; + + LIST_FOREACH(transaction, k, j) { + + if (tr->anchor_job == k || + !job_type_is_redundant(k->type, unit_active_state(k->unit)) || + (k->unit->job && job_type_is_conflicting(k->type, k->unit->job->type))) + goto next_unit; + } + + /* log_debug("Found redundant job %s/%s, dropping.", j->unit->id, job_type_to_string(j->type)); */ + transaction_delete_job(tr, j, false); + goto rescan; + next_unit:; + } +} + +_pure_ static bool unit_matters_to_anchor(Unit *u, Job *j) { + assert(u); + assert(!j->transaction_prev); + + /* Checks whether at least one of the jobs for this unit + * matters to the anchor. */ + + LIST_FOREACH(transaction, j, j) + if (j->matters_to_anchor) + return true; + + return false; +} + +static int transaction_verify_order_one(Transaction *tr, Job *j, Job *from, unsigned generation, sd_bus_error *e) { + Iterator i; + Unit *u; + int r; + + assert(tr); + assert(j); + assert(!j->transaction_prev); + + /* Does a recursive sweep through the ordering graph, looking + * for a cycle. If we find a cycle we try to break it. */ + + /* Have we seen this before? */ + if (j->generation == generation) { + Job *k, *delete; + + /* If the marker is NULL we have been here already and + * decided the job was loop-free from here. Hence + * shortcut things and return right-away. */ + if (!j->marker) + return 0; + + /* So, the marker is not NULL and we already have been + * here. We have a cycle. Let's try to break it. We go + * backwards in our path and try to find a suitable + * job to remove. We use the marker to find our way + * back, since smart how we are we stored our way back + * in there. */ + log_unit_warning(j->unit, + "Found ordering cycle on %s/%s", + j->unit->id, job_type_to_string(j->type)); + + delete = NULL; + for (k = from; k; k = ((k->generation == generation && k->marker != k) ? k->marker : NULL)) { + + /* logging for j not k here to provide consistent narrative */ + log_unit_warning(j->unit, + "Found dependency on %s/%s", + k->unit->id, job_type_to_string(k->type)); + + if (!delete && hashmap_get(tr->jobs, k->unit) && !unit_matters_to_anchor(k->unit, k)) + /* Ok, we can drop this one, so let's + * do so. */ + delete = k; + + /* Check if this in fact was the beginning of + * the cycle */ + if (k == j) + break; + } + + + if (delete) { + const char *status; + /* logging for j not k here to provide consistent narrative */ + log_unit_warning(j->unit, + "Breaking ordering cycle by deleting job %s/%s", + delete->unit->id, job_type_to_string(delete->type)); + log_unit_error(delete->unit, + "Job %s/%s deleted to break ordering cycle starting with %s/%s", + delete->unit->id, job_type_to_string(delete->type), + j->unit->id, job_type_to_string(j->type)); + + if (log_get_show_color()) + status = ANSI_HIGHLIGHT_RED " SKIP " ANSI_NORMAL; + else + status = " SKIP "; + + unit_status_printf(delete->unit, status, + "Ordering cycle found, skipping %s"); + transaction_delete_unit(tr, delete->unit); + return -EAGAIN; + } + + log_error("Unable to break cycle"); + + return sd_bus_error_setf(e, BUS_ERROR_TRANSACTION_ORDER_IS_CYCLIC, + "Transaction order is cyclic. See system logs for details."); + } + + /* Make the marker point to where we come from, so that we can + * find our way backwards if we want to break a cycle. We use + * a special marker for the beginning: we point to + * ourselves. */ + j->marker = from ? from : j; + j->generation = generation; + + /* We assume that the dependencies are bidirectional, and + * hence can ignore UNIT_AFTER */ + SET_FOREACH(u, j->unit->dependencies[UNIT_BEFORE], i) { + Job *o; + + /* Is there a job for this unit? */ + o = hashmap_get(tr->jobs, u); + if (!o) { + /* Ok, there is no job for this in the + * transaction, but maybe there is already one + * running? */ + o = u->job; + if (!o) + continue; + } + + r = transaction_verify_order_one(tr, o, j, generation, e); + if (r < 0) + return r; + } + + /* Ok, let's backtrack, and remember that this entry is not on + * our path anymore. */ + j->marker = NULL; + + return 0; +} + +static int transaction_verify_order(Transaction *tr, unsigned *generation, sd_bus_error *e) { + Job *j; + int r; + Iterator i; + unsigned g; + + assert(tr); + assert(generation); + + /* Check if the ordering graph is cyclic. If it is, try to fix + * that up by dropping one of the jobs. */ + + g = (*generation)++; + + HASHMAP_FOREACH(j, tr->jobs, i) { + r = transaction_verify_order_one(tr, j, NULL, g, e); + if (r < 0) + return r; + } + + return 0; +} + +static void transaction_collect_garbage(Transaction *tr) { + Iterator i; + Job *j; + + assert(tr); + + /* Drop jobs that are not required by any other job */ + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + if (tr->anchor_job == j || j->object_list) { + /* log_debug("Keeping job %s/%s because of %s/%s", */ + /* j->unit->id, job_type_to_string(j->type), */ + /* j->object_list->subject ? j->object_list->subject->unit->id : "root", */ + /* j->object_list->subject ? job_type_to_string(j->object_list->subject->type) : "root"); */ + continue; + } + + /* log_debug("Garbage collecting job %s/%s", j->unit->id, job_type_to_string(j->type)); */ + transaction_delete_job(tr, j, true); + goto rescan; + } +} + +static int transaction_is_destructive(Transaction *tr, JobMode mode, sd_bus_error *e) { + Iterator i; + Job *j; + + assert(tr); + + /* Checks whether applying this transaction means that + * existing jobs would be replaced */ + + HASHMAP_FOREACH(j, tr->jobs, i) { + + /* Assume merged */ + assert(!j->transaction_prev); + assert(!j->transaction_next); + + if (j->unit->job && (mode == JOB_FAIL || j->unit->job->irreversible) && + job_type_is_conflicting(j->unit->job->type, j->type)) + return sd_bus_error_setf(e, BUS_ERROR_TRANSACTION_IS_DESTRUCTIVE, + "Transaction is destructive."); + } + + return 0; +} + +static void transaction_minimize_impact(Transaction *tr) { + Job *j; + Iterator i; + + assert(tr); + + /* Drops all unnecessary jobs that reverse already active jobs + * or that stop a running service. */ + +rescan: + HASHMAP_FOREACH(j, tr->jobs, i) { + LIST_FOREACH(transaction, j, j) { + bool stops_running_service, changes_existing_job; + + /* If it matters, we shouldn't drop it */ + if (j->matters_to_anchor) + continue; + + /* Would this stop a running service? + * Would this change an existing job? + * If so, let's drop this entry */ + + stops_running_service = + j->type == JOB_STOP && UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(j->unit)); + + changes_existing_job = + j->unit->job && + job_type_is_conflicting(j->type, j->unit->job->type); + + if (!stops_running_service && !changes_existing_job) + continue; + + if (stops_running_service) + log_unit_debug(j->unit, + "%s/%s would stop a running service.", + j->unit->id, job_type_to_string(j->type)); + + if (changes_existing_job) + log_unit_debug(j->unit, + "%s/%s would change existing job.", + j->unit->id, job_type_to_string(j->type)); + + /* Ok, let's get rid of this */ + log_unit_debug(j->unit, + "Deleting %s/%s to minimize impact.", + j->unit->id, job_type_to_string(j->type)); + + transaction_delete_job(tr, j, true); + goto rescan; + } + } +} + +static int transaction_apply(Transaction *tr, Manager *m, JobMode mode) { + Iterator i; + Job *j; + int r; + + /* Moves the transaction jobs to the set of active jobs */ + + if (mode == JOB_ISOLATE || mode == JOB_FLUSH) { + + /* When isolating first kill all installed jobs which + * aren't part of the new transaction */ + HASHMAP_FOREACH(j, m->jobs, i) { + assert(j->installed); + + if (j->unit->ignore_on_isolate) + continue; + + if (hashmap_get(tr->jobs, j->unit)) + continue; + + /* Not invalidating recursively. Avoids triggering + * OnFailure= actions of dependent jobs. Also avoids + * invalidating our iterator. */ + job_finish_and_invalidate(j, JOB_CANCELED, false, false); + } + } + + HASHMAP_FOREACH(j, tr->jobs, i) { + /* Assume merged */ + assert(!j->transaction_prev); + assert(!j->transaction_next); + + r = hashmap_put(m->jobs, UINT32_TO_PTR(j->id), j); + if (r < 0) + goto rollback; + } + + while ((j = hashmap_steal_first(tr->jobs))) { + Job *installed_job; + + /* Clean the job dependencies */ + transaction_unlink_job(tr, j, false); + + installed_job = job_install(j); + if (installed_job != j) { + /* j has been merged into a previously installed job */ + if (tr->anchor_job == j) + tr->anchor_job = installed_job; + hashmap_remove(m->jobs, UINT32_TO_PTR(j->id)); + job_free(j); + j = installed_job; + } + + job_add_to_run_queue(j); + job_add_to_dbus_queue(j); + job_start_timer(j); + job_shutdown_magic(j); + } + + return 0; + +rollback: + + HASHMAP_FOREACH(j, tr->jobs, i) + hashmap_remove(m->jobs, UINT32_TO_PTR(j->id)); + + return r; +} + +int transaction_activate(Transaction *tr, Manager *m, JobMode mode, sd_bus_error *e) { + Iterator i; + Job *j; + int r; + unsigned generation = 1; + + assert(tr); + + /* This applies the changes recorded in tr->jobs to + * the actual list of jobs, if possible. */ + + /* Reset the generation counter of all installed jobs. The detection of cycles + * looks at installed jobs. If they had a non-zero generation from some previous + * walk of the graph, the algorithm would break. */ + HASHMAP_FOREACH(j, m->jobs, i) + j->generation = 0; + + /* First step: figure out which jobs matter */ + transaction_find_jobs_that_matter_to_anchor(tr->anchor_job, generation++); + + /* Second step: Try not to stop any running services if + * we don't have to. Don't try to reverse running + * jobs if we don't have to. */ + if (mode == JOB_FAIL) + transaction_minimize_impact(tr); + + /* Third step: Drop redundant jobs */ + transaction_drop_redundant(tr); + + for (;;) { + /* Fourth step: Let's remove unneeded jobs that might + * be lurking. */ + if (mode != JOB_ISOLATE) + transaction_collect_garbage(tr); + + /* Fifth step: verify order makes sense and correct + * cycles if necessary and possible */ + r = transaction_verify_order(tr, &generation, e); + if (r >= 0) + break; + + if (r != -EAGAIN) { + log_warning("Requested transaction contains an unfixable cyclic ordering dependency: %s", bus_error_message(e, r)); + return r; + } + + /* Let's see if the resulting transaction ordering + * graph is still cyclic... */ + } + + for (;;) { + /* Sixth step: let's drop unmergeable entries if + * necessary and possible, merge entries we can + * merge */ + r = transaction_merge_jobs(tr, e); + if (r >= 0) + break; + + if (r != -EAGAIN) { + log_warning("Requested transaction contains unmergeable jobs: %s", bus_error_message(e, r)); + return r; + } + + /* Seventh step: an entry got dropped, let's garbage + * collect its dependencies. */ + if (mode != JOB_ISOLATE) + transaction_collect_garbage(tr); + + /* Let's see if the resulting transaction still has + * unmergeable entries ... */ + } + + /* Eights step: Drop redundant jobs again, if the merging now allows us to drop more. */ + transaction_drop_redundant(tr); + + /* Ninth step: check whether we can actually apply this */ + r = transaction_is_destructive(tr, mode, e); + if (r < 0) { + log_notice("Requested transaction contradicts existing jobs: %s", bus_error_message(e, r)); + return r; + } + + /* Tenth step: apply changes */ + r = transaction_apply(tr, m, mode); + if (r < 0) + return log_warning_errno(r, "Failed to apply transaction: %m"); + + assert(hashmap_isempty(tr->jobs)); + + if (!hashmap_isempty(m->jobs)) { + /* Are there any jobs now? Then make sure we have the + * idle pipe around. We don't really care too much + * whether this works or not, as the idle pipe is a + * feature for cosmetics, not actually useful for + * anything beyond that. */ + + if (m->idle_pipe[0] < 0 && m->idle_pipe[1] < 0 && + m->idle_pipe[2] < 0 && m->idle_pipe[3] < 0) { + (void) pipe2(m->idle_pipe, O_NONBLOCK|O_CLOEXEC); + (void) pipe2(m->idle_pipe + 2, O_NONBLOCK|O_CLOEXEC); + } + } + + return 0; +} + +static Job* transaction_add_one_job(Transaction *tr, JobType type, Unit *unit, bool *is_new) { + Job *j, *f; + + assert(tr); + assert(unit); + + /* Looks for an existing prospective job and returns that. If + * it doesn't exist it is created and added to the prospective + * jobs list. */ + + f = hashmap_get(tr->jobs, unit); + + LIST_FOREACH(transaction, j, f) { + assert(j->unit == unit); + + if (j->type == type) { + if (is_new) + *is_new = false; + return j; + } + } + + j = job_new(unit, type); + if (!j) + return NULL; + + j->generation = 0; + j->marker = NULL; + j->matters_to_anchor = false; + j->irreversible = tr->irreversible; + + LIST_PREPEND(transaction, f, j); + + if (hashmap_replace(tr->jobs, unit, f) < 0) { + LIST_REMOVE(transaction, f, j); + job_free(j); + return NULL; + } + + if (is_new) + *is_new = true; + + /* log_debug("Added job %s/%s to transaction.", unit->id, job_type_to_string(type)); */ + + return j; +} + +static void transaction_unlink_job(Transaction *tr, Job *j, bool delete_dependencies) { + assert(tr); + assert(j); + + if (j->transaction_prev) + j->transaction_prev->transaction_next = j->transaction_next; + else if (j->transaction_next) + hashmap_replace(tr->jobs, j->unit, j->transaction_next); + else + hashmap_remove_value(tr->jobs, j->unit, j); + + if (j->transaction_next) + j->transaction_next->transaction_prev = j->transaction_prev; + + j->transaction_prev = j->transaction_next = NULL; + + while (j->subject_list) + job_dependency_free(j->subject_list); + + while (j->object_list) { + Job *other = j->object_list->matters ? j->object_list->subject : NULL; + + job_dependency_free(j->object_list); + + if (other && delete_dependencies) { + log_unit_debug(other->unit, + "Deleting job %s/%s as dependency of job %s/%s", + other->unit->id, job_type_to_string(other->type), + j->unit->id, job_type_to_string(j->type)); + transaction_delete_job(tr, other, delete_dependencies); + } + } +} + +int transaction_add_job_and_dependencies( + Transaction *tr, + JobType type, + Unit *unit, + Job *by, + bool matters, + bool conflicts, + bool ignore_requirements, + bool ignore_order, + sd_bus_error *e) { + Job *ret; + Iterator i; + Unit *dep; + int r; + bool is_new; + + assert(tr); + assert(type < _JOB_TYPE_MAX); + assert(type < _JOB_TYPE_MAX_IN_TRANSACTION); + assert(unit); + + /* Before adding jobs for this unit, let's ensure that its state has been loaded + * This matters when jobs are spawned as part of coldplugging itself (see e. g. path_coldplug()). + * This way, we "recursively" coldplug units, ensuring that we do not look at state of + * not-yet-coldplugged units. */ + if (MANAGER_IS_RELOADING(unit->manager)) + unit_coldplug(unit); + + /* log_debug("Pulling in %s/%s from %s/%s", */ + /* unit->id, job_type_to_string(type), */ + /* by ? by->unit->id : "NA", */ + /* by ? job_type_to_string(by->type) : "NA"); */ + + if (!IN_SET(unit->load_state, UNIT_LOADED, UNIT_ERROR, UNIT_NOT_FOUND, UNIT_MASKED)) + return sd_bus_error_setf(e, BUS_ERROR_LOAD_FAILED, "Unit %s is not loaded properly.", unit->id); + + if (type != JOB_STOP) { + r = bus_unit_check_load_state(unit, e); + if (r < 0) + return r; + } + + if (!unit_job_is_applicable(unit, type)) + return sd_bus_error_setf(e, BUS_ERROR_JOB_TYPE_NOT_APPLICABLE, + "Job type %s is not applicable for unit %s.", + job_type_to_string(type), unit->id); + + + /* First add the job. */ + ret = transaction_add_one_job(tr, type, unit, &is_new); + if (!ret) + return -ENOMEM; + + ret->ignore_order = ret->ignore_order || ignore_order; + + /* Then, add a link to the job. */ + if (by) { + if (!job_dependency_new(by, ret, matters, conflicts)) + return -ENOMEM; + } else { + /* If the job has no parent job, it is the anchor job. */ + assert(!tr->anchor_job); + tr->anchor_job = ret; + } + + if (is_new && !ignore_requirements && type != JOB_NOP) { + Set *following; + + /* If we are following some other unit, make sure we + * add all dependencies of everybody following. */ + if (unit_following_set(ret->unit, &following) > 0) { + SET_FOREACH(dep, following, i) { + r = transaction_add_job_and_dependencies(tr, type, dep, ret, false, false, false, ignore_order, e); + if (r < 0) { + log_unit_warning(dep, "Cannot add dependency job for, ignoring: %s", bus_error_message(e, r)); + sd_bus_error_free(e); + } + } + + set_free(following); + } + + /* Finally, recursively add in all dependencies. */ + if (type == JOB_START || type == JOB_RESTART) { + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUIRES], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) /* job type not applicable */ + goto fail; + + sd_bus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_BINDS_TO], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) /* job type not applicable */ + goto fail; + + sd_bus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_WANTS], i) { + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, false, false, false, ignore_order, e); + if (r < 0) { + /* unit masked, job type not applicable and unit not found are not considered as errors. */ + log_unit_full(dep, + IN_SET(r, -ERFKILL, -EBADR, -ENOENT) ? LOG_DEBUG : LOG_WARNING, + r, "Cannot add dependency job, ignoring: %s", + bus_error_message(e, r)); + sd_bus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUISITE], i) { + r = transaction_add_job_and_dependencies(tr, JOB_VERIFY_ACTIVE, dep, ret, true, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) /* job type not applicable */ + goto fail; + + sd_bus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONFLICTS], i) { + r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, true, true, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) /* job type not applicable */ + goto fail; + + sd_bus_error_free(e); + } + } + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONFLICTED_BY], i) { + r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, false, false, false, ignore_order, e); + if (r < 0) { + log_unit_warning(dep, + "Cannot add dependency job, ignoring: %s", + bus_error_message(e, r)); + sd_bus_error_free(e); + } + } + + } + + if (type == JOB_STOP || type == JOB_RESTART) { + static const UnitDependency propagate_deps[] = { + UNIT_REQUIRED_BY, + UNIT_REQUISITE_OF, + UNIT_BOUND_BY, + UNIT_CONSISTS_OF, + }; + + JobType ptype; + unsigned j; + + /* We propagate STOP as STOP, but RESTART only + * as TRY_RESTART, in order not to start + * dependencies that are not around. */ + ptype = type == JOB_RESTART ? JOB_TRY_RESTART : type; + + for (j = 0; j < ELEMENTSOF(propagate_deps); j++) + SET_FOREACH(dep, ret->unit->dependencies[propagate_deps[j]], i) { + JobType nt; + + nt = job_type_collapse(ptype, dep); + if (nt == JOB_NOP) + continue; + + r = transaction_add_job_and_dependencies(tr, nt, dep, ret, true, false, false, ignore_order, e); + if (r < 0) { + if (r != -EBADR) /* job type not applicable */ + goto fail; + + sd_bus_error_free(e); + } + } + } + + if (type == JOB_RELOAD) { + + SET_FOREACH(dep, ret->unit->dependencies[UNIT_PROPAGATES_RELOAD_TO], i) { + JobType nt; + + nt = job_type_collapse(JOB_TRY_RELOAD, dep); + if (nt == JOB_NOP) + continue; + + r = transaction_add_job_and_dependencies(tr, nt, dep, ret, false, false, false, ignore_order, e); + if (r < 0) { + log_unit_warning(dep, + "Cannot add dependency reload job, ignoring: %s", + bus_error_message(e, r)); + sd_bus_error_free(e); + } + } + } + + /* JOB_VERIFY_STARTED require no dependency handling */ + } + + return 0; + +fail: + return r; +} + +int transaction_add_isolate_jobs(Transaction *tr, Manager *m) { + Iterator i; + Unit *u; + char *k; + int r; + + assert(tr); + assert(m); + + HASHMAP_FOREACH_KEY(u, k, m->units, i) { + + /* ignore aliases */ + if (u->id != k) + continue; + + if (u->ignore_on_isolate) + continue; + + /* No need to stop inactive jobs */ + if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(u)) && !u->job) + continue; + + /* Is there already something listed for this? */ + if (hashmap_get(tr->jobs, u)) + continue; + + r = transaction_add_job_and_dependencies(tr, JOB_STOP, u, tr->anchor_job, true, false, false, false, NULL); + if (r < 0) + log_unit_warning_errno(u, r, "Cannot add isolate job, ignoring: %m"); + } + + return 0; +} + +Transaction *transaction_new(bool irreversible) { + Transaction *tr; + + tr = new0(Transaction, 1); + if (!tr) + return NULL; + + tr->jobs = hashmap_new(NULL); + if (!tr->jobs) + return mfree(tr); + + tr->irreversible = irreversible; + + return tr; +} + +void transaction_free(Transaction *tr) { + assert(hashmap_isempty(tr->jobs)); + hashmap_free(tr->jobs); + free(tr); +} diff --git a/src/grp-system/libcore/src/transaction.h b/src/grp-system/libcore/src/transaction.h new file mode 100644 index 0000000000..e4066a0963 --- /dev/null +++ b/src/grp-system/libcore/src/transaction.h @@ -0,0 +1,51 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/job.h" +#include "core/manager.h" +#include "core/unit.h" +#include "systemd-basic/hashmap.h" + +typedef struct Transaction Transaction; + +struct Transaction { + /* Jobs to be added */ + Hashmap *jobs; /* Unit object => Job object list 1:1 */ + Job *anchor_job; /* the job the user asked for */ + bool irreversible; +}; + +Transaction *transaction_new(bool irreversible); +void transaction_free(Transaction *tr); + +int transaction_add_job_and_dependencies( + Transaction *tr, + JobType type, + Unit *unit, + Job *by, + bool matters, + bool conflicts, + bool ignore_requirements, + bool ignore_order, + sd_bus_error *e); +int transaction_activate(Transaction *tr, Manager *m, JobMode mode, sd_bus_error *e); +int transaction_add_isolate_jobs(Transaction *tr, Manager *m); +void transaction_abort(Transaction *tr); diff --git a/src/grp-system/libcore/src/unit-printf.c b/src/grp-system/libcore/src/unit-printf.c new file mode 100644 index 0000000000..2f73214295 --- /dev/null +++ b/src/grp-system/libcore/src/unit-printf.c @@ -0,0 +1,305 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/unit.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" +#include "systemd-shared/specifier.h" + +#include "unit-printf.h" + +static int specifier_prefix_and_instance(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + + assert(u); + + return unit_name_to_prefix_and_instance(u->id, ret); +} + +static int specifier_prefix(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + + assert(u); + + return unit_name_to_prefix(u->id, ret); +} + +static int specifier_prefix_unescaped(char specifier, void *data, void *userdata, char **ret) { + _cleanup_free_ char *p = NULL; + Unit *u = userdata; + int r; + + assert(u); + + r = unit_name_to_prefix(u->id, &p); + if (r < 0) + return r; + + return unit_name_unescape(p, ret); +} + +static int specifier_instance_unescaped(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + + assert(u); + + return unit_name_unescape(strempty(u->instance), ret); +} + +static int specifier_filename(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + + assert(u); + + if (u->instance) + return unit_name_path_unescape(u->instance, ret); + else + return unit_name_to_path(u->id, ret); +} + +static int specifier_cgroup(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + char *n; + + assert(u); + + if (u->cgroup_path) + n = strdup(u->cgroup_path); + else + n = unit_default_cgroup_path(u); + if (!n) + return -ENOMEM; + + *ret = n; + return 0; +} + +static int specifier_cgroup_root(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + char *n; + + assert(u); + + n = strdup(u->manager->cgroup_root); + if (!n) + return -ENOMEM; + + *ret = n; + return 0; +} + +static int specifier_cgroup_slice(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + char *n; + + assert(u); + + if (UNIT_ISSET(u->slice)) { + Unit *slice; + + slice = UNIT_DEREF(u->slice); + + if (slice->cgroup_path) + n = strdup(slice->cgroup_path); + else + n = unit_default_cgroup_path(slice); + } else + n = strdup(u->manager->cgroup_root); + if (!n) + return -ENOMEM; + + *ret = n; + return 0; +} + +static int specifier_runtime(char specifier, void *data, void *userdata, char **ret) { + Unit *u = userdata; + const char *e; + char *n = NULL; + + assert(u); + + e = manager_get_runtime_prefix(u->manager); + if (!e) + return -EOPNOTSUPP; + n = strdup(e); + if (!n) + return -ENOMEM; + + *ret = n; + return 0; +} + +static int specifier_user_name(char specifier, void *data, void *userdata, char **ret) { + char *t; + + /* If we are UID 0 (root), this will not result in NSS, + * otherwise it might. This is good, as we want to be able to + * run this in PID 1, where our user ID is 0, but where NSS + * lookups are not allowed. */ + + t = getusername_malloc(); + if (!t) + return -ENOMEM; + + *ret = t; + return 0; +} + +static int specifier_user_id(char specifier, void *data, void *userdata, char **ret) { + + if (asprintf(ret, UID_FMT, getuid()) < 0) + return -ENOMEM; + + return 0; +} + +static int specifier_user_home(char specifier, void *data, void *userdata, char **ret) { + + /* On PID 1 (which runs as root) this will not result in NSS, + * which is good. See above */ + + return get_home_dir(ret); +} + +static int specifier_user_shell(char specifier, void *data, void *userdata, char **ret) { + + /* On PID 1 (which runs as root) this will not result in NSS, + * which is good. See above */ + + return get_shell(ret); +} + +int unit_name_printf(Unit *u, const char* format, char **ret) { + + /* + * This will use the passed string as format string and + * replace the following specifiers: + * + * %n: the full id of the unit (foo@bar.waldo) + * %N: the id of the unit without the suffix (foo@bar) + * %p: the prefix (foo) + * %i: the instance (bar) + */ + + const Specifier table[] = { + { 'n', specifier_string, u->id }, + { 'N', specifier_prefix_and_instance, NULL }, + { 'p', specifier_prefix, NULL }, + { 'i', specifier_string, u->instance }, + { 0, NULL, NULL } + }; + + assert(u); + assert(format); + assert(ret); + + return specifier_printf(format, table, u, ret); +} + +int unit_full_printf(Unit *u, const char *format, char **ret) { + + /* This is similar to unit_name_printf() but also supports + * unescaping. Also, adds a couple of additional codes: + * + * %f the instance if set, otherwise the id + * %c cgroup path of unit + * %r where units in this slice are placed in the cgroup tree + * %R the root of this systemd's instance tree + * %t the runtime directory to place sockets in (e.g. "/run" or $XDG_RUNTIME_DIR) + * %U the UID of the running user + * %u the username of the running user + * %h the homedir of the running user + * %s the shell of the running user + * %m the machine ID of the running system + * %H the host name of the running system + * %b the boot ID of the running system + * %v `uname -r` of the running system + */ + + const Specifier table[] = { + { 'n', specifier_string, u->id }, + { 'N', specifier_prefix_and_instance, NULL }, + { 'p', specifier_prefix, NULL }, + { 'P', specifier_prefix_unescaped, NULL }, + { 'i', specifier_string, u->instance }, + { 'I', specifier_instance_unescaped, NULL }, + + { 'f', specifier_filename, NULL }, + { 'c', specifier_cgroup, NULL }, + { 'r', specifier_cgroup_slice, NULL }, + { 'R', specifier_cgroup_root, NULL }, + { 't', specifier_runtime, NULL }, + + { 'U', specifier_user_id, NULL }, + { 'u', specifier_user_name, NULL }, + { 'h', specifier_user_home, NULL }, + { 's', specifier_user_shell, NULL }, + + { 'm', specifier_machine_id, NULL }, + { 'H', specifier_host_name, NULL }, + { 'b', specifier_boot_id, NULL }, + { 'v', specifier_kernel_release, NULL }, + {} + }; + + assert(u); + assert(format); + assert(ret); + + return specifier_printf(format, table, u, ret); +} + +int unit_full_printf_strv(Unit *u, char **l, char ***ret) { + size_t n; + char **r, **i, **j; + int q; + + /* Applies unit_full_printf to every entry in l */ + + assert(u); + + n = strv_length(l); + r = new(char*, n+1); + if (!r) + return -ENOMEM; + + for (i = l, j = r; *i; i++, j++) { + q = unit_full_printf(u, *i, j); + if (q < 0) + goto fail; + } + + *j = NULL; + *ret = r; + return 0; + +fail: + for (j--; j >= r; j--) + free(*j); + + free(r); + return q; +} diff --git a/src/grp-system/libcore/src/unit-printf.h b/src/grp-system/libcore/src/unit-printf.h new file mode 100644 index 0000000000..7ef76e5bb9 --- /dev/null +++ b/src/grp-system/libcore/src/unit-printf.h @@ -0,0 +1,26 @@ +#pragma once + +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include "core/unit.h" + +int unit_name_printf(Unit *u, const char* text, char **ret); +int unit_full_printf(Unit *u, const char *text, char **ret); +int unit_full_printf_strv(Unit *u, char **l, char ***ret); diff --git a/src/grp-system/libcore/src/unit.c b/src/grp-system/libcore/src/unit.c new file mode 100644 index 0000000000..364208dbad --- /dev/null +++ b/src/grp-system/libcore/src/unit.c @@ -0,0 +1,4284 @@ +/*** + This file is part of systemd. + + Copyright 2010 Lennart Poettering + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <errno.h> +#include <stdlib.h> +#include <string.h> +#include <sys/stat.h> +#include <unistd.h> + +#include <systemd/sd-id128.h> +#include <systemd/sd-messages.h> + +#include "core/execute.h" +#include "core/load-fragment.h" +#include "core/unit.h" +#include "sd-bus/bus-common-errors.h" +#include "sd-bus/bus-util.h" +#include "sd-id128/id128-util.h" +#include "systemd-basic/alloc-util.h" +#include "systemd-basic/cgroup-util.h" +#include "systemd-basic/escape.h" +#include "systemd-basic/fileio-label.h" +#include "systemd-basic/formats-util.h" +#include "systemd-basic/log.h" +#include "systemd-basic/macro.h" +#include "systemd-basic/missing.h" +#include "systemd-basic/mkdir.h" +#include "systemd-basic/parse-util.h" +#include "systemd-basic/path-util.h" +#include "systemd-basic/process-util.h" +#include "systemd-basic/set.h" +#include "systemd-basic/signal-util.h" +#include "systemd-basic/special.h" +#include "systemd-basic/stat-util.h" +#include "systemd-basic/stdio-util.h" +#include "systemd-basic/string-util.h" +#include "systemd-basic/strv.h" +#include "systemd-basic/umask-util.h" +#include "systemd-basic/unit-name.h" +#include "systemd-basic/user-util.h" +#include "systemd-basic/virt.h" +#include "systemd-shared/dropin.h" + +#include "dbus-unit.h" +#include "dbus.h" +#include "load-dropin.h" + +const UnitVTable * const unit_vtable[_UNIT_TYPE_MAX] = { + [UNIT_SERVICE] = &service_vtable, + [UNIT_SOCKET] = &socket_vtable, + [UNIT_BUSNAME] = &busname_vtable, + [UNIT_TARGET] = &target_vtable, + [UNIT_DEVICE] = &device_vtable, + [UNIT_MOUNT] = &mount_vtable, + [UNIT_AUTOMOUNT] = &automount_vtable, + [UNIT_SWAP] = &swap_vtable, + [UNIT_TIMER] = &timer_vtable, + [UNIT_PATH] = &path_vtable, + [UNIT_SLICE] = &slice_vtable, + [UNIT_SCOPE] = &scope_vtable +}; + +static void maybe_warn_about_dependency(Unit *u, const char *other, UnitDependency dependency); + +Unit *unit_new(Manager *m, size_t size) { + Unit *u; + + assert(m); + assert(size >= sizeof(Unit)); + + u = malloc0(size); + if (!u) + return NULL; + + u->names = set_new(&string_hash_ops); + if (!u->names) + return mfree(u); + + u->manager = m; + u->type = _UNIT_TYPE_INVALID; + u->default_dependencies = true; + u->unit_file_state = _UNIT_FILE_STATE_INVALID; + u->unit_file_preset = -1; + u->on_failure_job_mode = JOB_REPLACE; + u->cgroup_inotify_wd = -1; + u->job_timeout = USEC_INFINITY; + u->ref_uid = UID_INVALID; + u->ref_gid = GID_INVALID; + u->cpu_usage_last = NSEC_INFINITY; + + RATELIMIT_INIT(u->start_limit, m->default_start_limit_interval, m->default_start_limit_burst); + RATELIMIT_INIT(u->auto_stop_ratelimit, 10 * USEC_PER_SEC, 16); + + return u; +} + +int unit_new_for_name(Manager *m, size_t size, const char *name, Unit **ret) { + Unit *u; + int r; + + u = unit_new(m, size); + if (!u) + return -ENOMEM; + + r = unit_add_name(u, name); + if (r < 0) { + unit_free(u); + return r; + } + + *ret = u; + return r; +} + +bool unit_has_name(Unit *u, const char *name) { + assert(u); + assert(name); + + return set_contains(u->names, (char*) name); +} + +static void unit_init(Unit *u) { + CGroupContext *cc; + ExecContext *ec; + KillContext *kc; + + assert(u); + assert(u->manager); + assert(u->type >= 0); + + cc = unit_get_cgroup_context(u); + if (cc) { + cgroup_context_init(cc); + + /* Copy in the manager defaults into the cgroup + * context, _before_ the rest of the settings have + * been initialized */ + + cc->cpu_accounting = u->manager->default_cpu_accounting; + cc->io_accounting = u->manager->default_io_accounting; + cc->blockio_accounting = u->manager->default_blockio_accounting; + cc->memory_accounting = u->manager->default_memory_accounting; + cc->tasks_accounting = u->manager->default_tasks_accounting; + + if (u->type != UNIT_SLICE) + cc->tasks_max = u->manager->default_tasks_max; + } + + ec = unit_get_exec_context(u); + if (ec) + exec_context_init(ec); + + kc = unit_get_kill_context(u); + if (kc) + kill_context_init(kc); + + if (UNIT_VTABLE(u)->init) + UNIT_VTABLE(u)->init(u); +} + +int unit_add_name(Unit *u, const char *text) { + _cleanup_free_ char *s = NULL, *i = NULL; + UnitType t; + int r; + + assert(u); + assert(text); + + if (unit_name_is_valid(text, UNIT_NAME_TEMPLATE)) { + + if (!u->instance) + return -EINVAL; + + r = unit_name_replace_instance(text, u->instance, &s); + if (r < 0) + return r; + } else { + s = strdup(text); + if (!s) + return -ENOMEM; + } + + if (set_contains(u->names, s)) + return 0; + if (hashmap_contains(u->manager->units, s)) + return -EEXIST; + + if (!unit_name_is_valid(s, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) + return -EINVAL; + + t = unit_name_to_type(s); + if (t < 0) + return -EINVAL; + + if (u->type != _UNIT_TYPE_INVALID && t != u->type) + return -EINVAL; + + r = unit_name_to_instance(s, &i); + if (r < 0) + return r; + + if (i && !unit_type_may_template(t)) + return -EINVAL; + + /* Ensure that this unit is either instanced or not instanced, + * but not both. Note that we do allow names with different + * instance names however! */ + if (u->type != _UNIT_TYPE_INVALID && !u->instance != !i) + return -EINVAL; + + if (!unit_type_may_alias(t) && !set_isempty(u->names)) + return -EEXIST; + + if (hashmap_size(u->manager->units) >= MANAGER_MAX_NAMES) + return -E2BIG; + + r = set_put(u->names, s); + if (r < 0) + return r; + assert(r > 0); + + r = hashmap_put(u->manager->units, s, u); + if (r < 0) { + (void) set_remove(u->names, s); + return r; + } + + if (u->type == _UNIT_TYPE_INVALID) { + u->type = t; + u->id = s; + u->instance = i; + + LIST_PREPEND(units_by_type, u->manager->units_by_type[t], u); + + unit_init(u); + + i = NULL; + } + + s = NULL; + + unit_add_to_dbus_queue(u); + return 0; +} + +int unit_choose_id(Unit *u, const char *name) { + _cleanup_free_ char *t = NULL; + char *s, *i; + int r; + + assert(u); + assert(name); + + if (unit_name_is_valid(name, UNIT_NAME_TEMPLATE)) { + + if (!u->instance) + return -EINVAL; + + r = unit_name_replace_instance(name, u->instance, &t); + if (r < 0) + return r; + + name = t; + } + + /* Selects one of the names of this unit as the id */ + s = set_get(u->names, (char*) name); + if (!s) + return -ENOENT; + + /* Determine the new instance from the new id */ + r = unit_name_to_instance(s, &i); + if (r < 0) + return r; + + u->id = s; + + free(u->instance); + u->instance = i; + + unit_add_to_dbus_queue(u); + + return 0; +} + +int unit_set_description(Unit *u, const char *description) { + char *s; + + assert(u); + + if (isempty(description)) + s = NULL; + else { + s = strdup(description); + if (!s) + return -ENOMEM; + } + + free(u->description); + u->description = s; + + unit_add_to_dbus_queue(u); + return 0; +} + +bool unit_check_gc(Unit *u) { + UnitActiveState state; + bool inactive; + assert(u); + + if (u->job) + return true; + + if (u->nop_job) + return true; + + state = unit_active_state(u); + inactive = state == UNIT_INACTIVE; + + /* If the unit is inactive and failed and no job is queued for + * it, then release its runtime resources */ + if (UNIT_IS_INACTIVE_OR_FAILED(state) && + UNIT_VTABLE(u)->release_resources) + UNIT_VTABLE(u)->release_resources(u, inactive); + + /* But we keep the unit object around for longer when it is + * referenced or configured to not be gc'ed */ + if (!inactive) + return true; + + if (u->perpetual) + return true; + + if (u->refs) + return true; + + if (sd_bus_track_count(u->bus_track) > 0) + return true; + + if (UNIT_VTABLE(u)->check_gc) + if (UNIT_VTABLE(u)->check_gc(u)) + return true; + + return false; +} + +void unit_add_to_load_queue(Unit *u) { + assert(u); + assert(u->type != _UNIT_TYPE_INVALID); + + if (u->load_state != UNIT_STUB || u->in_load_queue) + return; + + LIST_PREPEND(load_queue, u->manager->load_queue, u); + u->in_load_queue = true; +} + +void unit_add_to_cleanup_queue(Unit *u) { + assert(u); + + if (u->in_cleanup_queue) + return; + + LIST_PREPEND(cleanup_queue, u->manager->cleanup_queue, u); + u->in_cleanup_queue = true; +} + +void unit_add_to_gc_queue(Unit *u) { + assert(u); + + if (u->in_gc_queue || u->in_cleanup_queue) + return; + + if (unit_check_gc(u)) + return; + + LIST_PREPEND(gc_queue, u->manager->gc_queue, u); + u->in_gc_queue = true; + + u->manager->n_in_gc_queue++; +} + +void unit_add_to_dbus_queue(Unit *u) { + assert(u); + assert(u->type != _UNIT_TYPE_INVALID); + + if (u->load_state == UNIT_STUB || u->in_dbus_queue) + return; + + /* Shortcut things if nobody cares */ + if (sd_bus_track_count(u->manager->subscribed) <= 0 && + set_isempty(u->manager->private_buses)) { + u->sent_dbus_new_signal = true; + return; + } + + LIST_PREPEND(dbus_queue, u->manager->dbus_unit_queue, u); + u->in_dbus_queue = true; +} + +static void bidi_set_free(Unit *u, Set *s) { + Iterator i; + Unit *other; + + assert(u); + + /* Frees the set and makes sure we are dropped from the + * inverse pointers */ + + SET_FOREACH(other, s, i) { + UnitDependency d; + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + set_remove(other->dependencies[d], u); + + unit_add_to_gc_queue(other); + } + + set_free(s); +} + +static void unit_remove_transient(Unit *u) { + char **i; + + assert(u); + + if (!u->transient) + return; + + if (u->fragment_path) + (void) unlink(u->fragment_path); + + STRV_FOREACH(i, u->dropin_paths) { + _cleanup_free_ char *p = NULL, *pp = NULL; + + p = dirname_malloc(*i); /* Get the drop-in directory from the drop-in file */ + if (!p) + continue; + + pp = dirname_malloc(p); /* Get the config directory from the drop-in directory */ + if (!pp) + continue; + + /* Only drop transient drop-ins */ + if (!path_equal(u->manager->lookup_paths.transient, pp)) + continue; + + (void) unlink(*i); + (void) rmdir(p); + } +} + +static void unit_free_requires_mounts_for(Unit *u) { + char **j; + + STRV_FOREACH(j, u->requires_mounts_for) { + char s[strlen(*j) + 1]; + + PATH_FOREACH_PREFIX_MORE(s, *j) { + char *y; + Set *x; + + x = hashmap_get2(u->manager->units_requiring_mounts_for, s, (void**) &y); + if (!x) + continue; + + set_remove(x, u); + + if (set_isempty(x)) { + hashmap_remove(u->manager->units_requiring_mounts_for, y); + free(y); + set_free(x); + } + } + } + + u->requires_mounts_for = strv_free(u->requires_mounts_for); +} + +static void unit_done(Unit *u) { + ExecContext *ec; + CGroupContext *cc; + + assert(u); + + if (u->type < 0) + return; + + if (UNIT_VTABLE(u)->done) + UNIT_VTABLE(u)->done(u); + + ec = unit_get_exec_context(u); + if (ec) + exec_context_done(ec); + + cc = unit_get_cgroup_context(u); + if (cc) + cgroup_context_done(cc); +} + +void unit_free(Unit *u) { + UnitDependency d; + Iterator i; + char *t; + + assert(u); + + if (u->transient_file) + fclose(u->transient_file); + + if (!MANAGER_IS_RELOADING(u->manager)) + unit_remove_transient(u); + + bus_unit_send_removed_signal(u); + + unit_done(u); + + sd_bus_slot_unref(u->match_bus_slot); + + sd_bus_track_unref(u->bus_track); + u->deserialized_refs = strv_free(u->deserialized_refs); + + unit_free_requires_mounts_for(u); + + SET_FOREACH(t, u->names, i) + hashmap_remove_value(u->manager->units, t, u); + + if (!sd_id128_is_null(u->invocation_id)) + hashmap_remove_value(u->manager->units_by_invocation_id, &u->invocation_id, u); + + if (u->job) { + Job *j = u->job; + job_uninstall(j); + job_free(j); + } + + if (u->nop_job) { + Job *j = u->nop_job; + job_uninstall(j); + job_free(j); + } + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + bidi_set_free(u, u->dependencies[d]); + + if (u->type != _UNIT_TYPE_INVALID) + LIST_REMOVE(units_by_type, u->manager->units_by_type[u->type], u); + + if (u->in_load_queue) + LIST_REMOVE(load_queue, u->manager->load_queue, u); + + if (u->in_dbus_queue) + LIST_REMOVE(dbus_queue, u->manager->dbus_unit_queue, u); + + if (u->in_cleanup_queue) + LIST_REMOVE(cleanup_queue, u->manager->cleanup_queue, u); + + if (u->in_gc_queue) { + LIST_REMOVE(gc_queue, u->manager->gc_queue, u); + u->manager->n_in_gc_queue--; + } + + if (u->in_cgroup_queue) + LIST_REMOVE(cgroup_queue, u->manager->cgroup_queue, u); + + unit_release_cgroup(u); + + unit_unref_uid_gid(u, false); + + (void) manager_update_failed_units(u->manager, u, false); + set_remove(u->manager->startup_units, u); + + free(u->description); + strv_free(u->documentation); + free(u->fragment_path); + free(u->source_path); + strv_free(u->dropin_paths); + free(u->instance); + + free(u->job_timeout_reboot_arg); + + set_free_free(u->names); + + unit_unwatch_all_pids(u); + + condition_free_list(u->conditions); + condition_free_list(u->asserts); + + free(u->reboot_arg); + + unit_ref_unset(&u->slice); + + while (u->refs) + unit_ref_unset(u->refs); + + free(u); +} + +UnitActiveState unit_active_state(Unit *u) { + assert(u); + + if (u->load_state == UNIT_MERGED) + return unit_active_state(unit_follow_merge(u)); + + /* After a reload it might happen that a unit is not correctly + * loaded but still has a process around. That's why we won't + * shortcut failed loading to UNIT_INACTIVE_FAILED. */ + + return UNIT_VTABLE(u)->active_state(u); +} + +const char* unit_sub_state_to_string(Unit *u) { + assert(u); + + return UNIT_VTABLE(u)->sub_state_to_string(u); +} + +static int complete_move(Set **s, Set **other) { + int r; + + assert(s); + assert(other); + + if (!*other) + return 0; + + if (*s) { + r = set_move(*s, *other); + if (r < 0) + return r; + } else { + *s = *other; + *other = NULL; + } + + return 0; +} + +static int merge_names(Unit *u, Unit *other) { + char *t; + Iterator i; + int r; + + assert(u); + assert(other); + + r = complete_move(&u->names, &other->names); + if (r < 0) + return r; + + set_free_free(other->names); + other->names = NULL; + other->id = NULL; + + SET_FOREACH(t, u->names, i) + assert_se(hashmap_replace(u->manager->units, t, u) == 0); + + return 0; +} + +static int reserve_dependencies(Unit *u, Unit *other, UnitDependency d) { + unsigned n_reserve; + + assert(u); + assert(other); + assert(d < _UNIT_DEPENDENCY_MAX); + + /* + * If u does not have this dependency set allocated, there is no need + * to reserve anything. In that case other's set will be transferred + * as a whole to u by complete_move(). + */ + if (!u->dependencies[d]) + return 0; + + /* merge_dependencies() will skip a u-on-u dependency */ + n_reserve = set_size(other->dependencies[d]) - !!set_get(other->dependencies[d], u); + + return set_reserve(u->dependencies[d], n_reserve); +} + +static void merge_dependencies(Unit *u, Unit *other, const char *other_id, UnitDependency d) { + Iterator i; + Unit *back; + int r; + + assert(u); + assert(other); + assert(d < _UNIT_DEPENDENCY_MAX); + + /* Fix backwards pointers */ + SET_FOREACH(back, other->dependencies[d], i) { + UnitDependency k; + + for (k = 0; k < _UNIT_DEPENDENCY_MAX; k++) { + /* Do not add dependencies between u and itself */ + if (back == u) { + if (set_remove(back->dependencies[k], other)) + maybe_warn_about_dependency(u, other_id, k); + } else { + r = set_remove_and_put(back->dependencies[k], other, u); + if (r == -EEXIST) + set_remove(back->dependencies[k], other); + else + assert(r >= 0 || r == -ENOENT); + } + } + } + + /* Also do not move dependencies on u to itself */ + back = set_remove(other->dependencies[d], u); + if (back) + maybe_warn_about_dependency(u, other_id, d); + + /* The move cannot fail. The caller must have performed a reservation. */ + assert_se(complete_move(&u->dependencies[d], &other->dependencies[d]) == 0); + + other->dependencies[d] = set_free(other->dependencies[d]); +} + +int unit_merge(Unit *u, Unit *other) { + UnitDependency d; + const char *other_id = NULL; + int r; + + assert(u); + assert(other); + assert(u->manager == other->manager); + assert(u->type != _UNIT_TYPE_INVALID); + + other = unit_follow_merge(other); + + if (other == u) + return 0; + + if (u->type != other->type) + return -EINVAL; + + if (!u->instance != !other->instance) + return -EINVAL; + + if (!unit_type_may_alias(u->type)) /* Merging only applies to unit names that support aliases */ + return -EEXIST; + + if (other->load_state != UNIT_STUB && + other->load_state != UNIT_NOT_FOUND) + return -EEXIST; + + if (other->job) + return -EEXIST; + + if (other->nop_job) + return -EEXIST; + + if (!UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) + return -EEXIST; + + if (other->id) + other_id = strdupa(other->id); + + /* Make reservations to ensure merge_dependencies() won't fail */ + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) { + r = reserve_dependencies(u, other, d); + /* + * We don't rollback reservations if we fail. We don't have + * a way to undo reservations. A reservation is not a leak. + */ + if (r < 0) + return r; + } + + /* Merge names */ + r = merge_names(u, other); + if (r < 0) + return r; + + /* Redirect all references */ + while (other->refs) + unit_ref_set(other->refs, u); + + /* Merge dependencies */ + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) + merge_dependencies(u, other, other_id, d); + + other->load_state = UNIT_MERGED; + other->merged_into = u; + + /* If there is still some data attached to the other node, we + * don't need it anymore, and can free it. */ + if (other->load_state != UNIT_STUB) + if (UNIT_VTABLE(other)->done) + UNIT_VTABLE(other)->done(other); + + unit_add_to_dbus_queue(u); + unit_add_to_cleanup_queue(other); + + return 0; +} + +int unit_merge_by_name(Unit *u, const char *name) { + _cleanup_free_ char *s = NULL; + Unit *other; + int r; + + assert(u); + assert(name); + + if (unit_name_is_valid(name, UNIT_NAME_TEMPLATE)) { + if (!u->instance) + return -EINVAL; + + r = unit_name_replace_instance(name, u->instance, &s); + if (r < 0) + return r; + + name = s; + } + + other = manager_get_unit(u->manager, name); + if (other) + return unit_merge(u, other); + + return unit_add_name(u, name); +} + +Unit* unit_follow_merge(Unit *u) { + assert(u); + + while (u->load_state == UNIT_MERGED) + assert_se(u = u->merged_into); + + return u; +} + +int unit_add_exec_dependencies(Unit *u, ExecContext *c) { + int r; + + assert(u); + assert(c); + + if (c->working_directory) { + r = unit_require_mounts_for(u, c->working_directory); + if (r < 0) + return r; + } + + if (c->root_directory) { + r = unit_require_mounts_for(u, c->root_directory); + if (r < 0) + return r; + } + + if (!MANAGER_IS_SYSTEM(u->manager)) + return 0; + + if (c->private_tmp) { + r = unit_require_mounts_for(u, "/tmp"); + if (r < 0) + return r; + + r = unit_require_mounts_for(u, "/var/tmp"); + if (r < 0) + return r; + } + + if (!IN_SET(c->std_output, + EXEC_OUTPUT_JOURNAL, EXEC_OUTPUT_JOURNAL_AND_CONSOLE, + EXEC_OUTPUT_KMSG, EXEC_OUTPUT_KMSG_AND_CONSOLE, + EXEC_OUTPUT_SYSLOG, EXEC_OUTPUT_SYSLOG_AND_CONSOLE) && + !IN_SET(c->std_error, + EXEC_OUTPUT_JOURNAL, EXEC_OUTPUT_JOURNAL_AND_CONSOLE, + EXEC_OUTPUT_KMSG, EXEC_OUTPUT_KMSG_AND_CONSOLE, + EXEC_OUTPUT_SYSLOG, EXEC_OUTPUT_SYSLOG_AND_CONSOLE)) + return 0; + + /* If syslog or kernel logging is requested, make sure our own + * logging daemon is run first. */ + + r = unit_add_dependency_by_name(u, UNIT_AFTER, SPECIAL_JOURNALD_SOCKET, NULL, true); + if (r < 0) + return r; + + return 0; +} + +const char *unit_description(Unit *u) { + assert(u); + + if (u->description) + return u->description; + + return strna(u->id); +} + +void unit_dump(Unit *u, FILE *f, const char *prefix) { + char *t, **j; + UnitDependency d; + Iterator i; + const char *prefix2; + char + timestamp0[FORMAT_TIMESTAMP_MAX], + timestamp1[FORMAT_TIMESTAMP_MAX], + timestamp2[FORMAT_TIMESTAMP_MAX], + timestamp3[FORMAT_TIMESTAMP_MAX], + timestamp4[FORMAT_TIMESTAMP_MAX], + timespan[FORMAT_TIMESPAN_MAX]; + Unit *following; + _cleanup_set_free_ Set *following_set = NULL; + int r; + const char *n; + + assert(u); + assert(u->type >= 0); + + prefix = strempty(prefix); + prefix2 = strjoina(prefix, "\t"); + + fprintf(f, + "%s-> Unit %s:\n" + "%s\tDescription: %s\n" + "%s\tInstance: %s\n" + "%s\tUnit Load State: %s\n" + "%s\tUnit Active State: %s\n" + "%s\tState Change Timestamp: %s\n" + "%s\tInactive Exit Timestamp: %s\n" + "%s\tActive Enter Timestamp: %s\n" + "%s\tActive Exit Timestamp: %s\n" + "%s\tInactive Enter Timestamp: %s\n" + "%s\tGC Check Good: %s\n" + "%s\tNeed Daemon Reload: %s\n" + "%s\tTransient: %s\n" + "%s\tPerpetual: %s\n" + "%s\tSlice: %s\n" + "%s\tCGroup: %s\n" + "%s\tCGroup realized: %s\n" + "%s\tCGroup mask: 0x%x\n" + "%s\tCGroup members mask: 0x%x\n", + prefix, u->id, + prefix, unit_description(u), + prefix, strna(u->instance), + prefix, unit_load_state_to_string(u->load_state), + prefix, unit_active_state_to_string(unit_active_state(u)), + prefix, strna(format_timestamp(timestamp0, sizeof(timestamp0), u->state_change_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp1, sizeof(timestamp1), u->inactive_exit_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp2, sizeof(timestamp2), u->active_enter_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp3, sizeof(timestamp3), u->active_exit_timestamp.realtime)), + prefix, strna(format_timestamp(timestamp4, sizeof(timestamp4), u->inactive_enter_timestamp.realtime)), + prefix, yes_no(unit_check_gc(u)), + prefix, yes_no(unit_need_daemon_reload(u)), + prefix, yes_no(u->transient), + prefix, yes_no(u->perpetual), + prefix, strna(unit_slice_name(u)), + prefix, strna(u->cgroup_path), + prefix, yes_no(u->cgroup_realized), + prefix, u->cgroup_realized_mask, + prefix, u->cgroup_members_mask); + + SET_FOREACH(t, u->names, i) + fprintf(f, "%s\tName: %s\n", prefix, t); + + if (!sd_id128_is_null(u->invocation_id)) + fprintf(f, "%s\tInvocation ID: " SD_ID128_FORMAT_STR "\n", + prefix, SD_ID128_FORMAT_VAL(u->invocation_id)); + + STRV_FOREACH(j, u->documentation) + fprintf(f, "%s\tDocumentation: %s\n", prefix, *j); + + following = unit_following(u); + if (following) + fprintf(f, "%s\tFollowing: %s\n", prefix, following->id); + + r = unit_following_set(u, &following_set); + if (r >= 0) { + Unit *other; + + SET_FOREACH(other, following_set, i) + fprintf(f, "%s\tFollowing Set Member: %s\n", prefix, other->id); + } + + if (u->fragment_path) + fprintf(f, "%s\tFragment Path: %s\n", prefix, u->fragment_path); + + if (u->source_path) + fprintf(f, "%s\tSource Path: %s\n", prefix, u->source_path); + + STRV_FOREACH(j, u->dropin_paths) + fprintf(f, "%s\tDropIn Path: %s\n", prefix, *j); + + if (u->job_timeout != USEC_INFINITY) + fprintf(f, "%s\tJob Timeout: %s\n", prefix, format_timespan(timespan, sizeof(timespan), u->job_timeout, 0)); + + if (u->job_timeout_action != EMERGENCY_ACTION_NONE) + fprintf(f, "%s\tJob Timeout Action: %s\n", prefix, emergency_action_to_string(u->job_timeout_action)); + + if (u->job_timeout_reboot_arg) + fprintf(f, "%s\tJob Timeout Reboot Argument: %s\n", prefix, u->job_timeout_reboot_arg); + + condition_dump_list(u->conditions, f, prefix, condition_type_to_string); + condition_dump_list(u->asserts, f, prefix, assert_type_to_string); + + if (dual_timestamp_is_set(&u->condition_timestamp)) + fprintf(f, + "%s\tCondition Timestamp: %s\n" + "%s\tCondition Result: %s\n", + prefix, strna(format_timestamp(timestamp1, sizeof(timestamp1), u->condition_timestamp.realtime)), + prefix, yes_no(u->condition_result)); + + if (dual_timestamp_is_set(&u->assert_timestamp)) + fprintf(f, + "%s\tAssert Timestamp: %s\n" + "%s\tAssert Result: %s\n", + prefix, strna(format_timestamp(timestamp1, sizeof(timestamp1), u->assert_timestamp.realtime)), + prefix, yes_no(u->assert_result)); + + for (d = 0; d < _UNIT_DEPENDENCY_MAX; d++) { + Unit *other; + + SET_FOREACH(other, u->dependencies[d], i) + fprintf(f, "%s\t%s: %s\n", prefix, unit_dependency_to_string(d), other->id); + } + + if (!strv_isempty(u->requires_mounts_for)) { + fprintf(f, + "%s\tRequiresMountsFor:", prefix); + + STRV_FOREACH(j, u->requires_mounts_for) + fprintf(f, " %s", *j); + + fputs("\n", f); + } + + if (u->load_state == UNIT_LOADED) { + + fprintf(f, + "%s\tStopWhenUnneeded: %s\n" + "%s\tRefuseManualStart: %s\n" + "%s\tRefuseManualStop: %s\n" + "%s\tDefaultDependencies: %s\n" + "%s\tOnFailureJobMode: %s\n" + "%s\tIgnoreOnIsolate: %s\n", + prefix, yes_no(u->stop_when_unneeded), + prefix, yes_no(u->refuse_manual_start), + prefix, yes_no(u->refuse_manual_stop), + prefix, yes_no(u->default_dependencies), + prefix, job_mode_to_string(u->on_failure_job_mode), + prefix, yes_no(u->ignore_on_isolate)); + + if (UNIT_VTABLE(u)->dump) + UNIT_VTABLE(u)->dump(u, f, prefix2); + + } else if (u->load_state == UNIT_MERGED) + fprintf(f, + "%s\tMerged into: %s\n", + prefix, u->merged_into->id); + else if (u->load_state == UNIT_ERROR) + fprintf(f, "%s\tLoad Error Code: %s\n", prefix, strerror(-u->load_error)); + + for (n = sd_bus_track_first(u->bus_track); n; n = sd_bus_track_next(u->bus_track)) + fprintf(f, "%s\tBus Ref: %s\n", prefix, n); + + if (u->job) + job_dump(u->job, f, prefix2); + + if (u->nop_job) + job_dump(u->nop_job, f, prefix2); +} + +/* Common implementation for multiple backends */ +int unit_load_fragment_and_dropin(Unit *u) { + int r; + + assert(u); + + /* Load a .{service,socket,...} file */ + r = unit_load_fragment(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) + return -ENOENT; + + /* Load drop-in directory data */ + r = unit_load_dropin(unit_follow_merge(u)); + if (r < 0) + return r; + + return 0; +} + +/* Common implementation for multiple backends */ +int unit_load_fragment_and_dropin_optional(Unit *u) { + int r; + + assert(u); + + /* Same as unit_load_fragment_and_dropin(), but whether + * something can be loaded or not doesn't matter. */ + + /* Load a .service file */ + r = unit_load_fragment(u); + if (r < 0) + return r; + + if (u->load_state == UNIT_STUB) + u->load_state = UNIT_LOADED; + + /* Load drop-in directory data */ + r = unit_load_dropin(unit_follow_merge(u)); + if (r < 0) + return r; + + return 0; +} + +int unit_add_default_target_dependency(Unit *u, Unit *target) { + assert(u); + assert(target); + + if (target->type != UNIT_TARGET) + return 0; + + /* Only add the dependency if both units are loaded, so that + * that loop check below is reliable */ + if (u->load_state != UNIT_LOADED || + target->load_state != UNIT_LOADED) + return 0; + + /* If either side wants no automatic dependencies, then let's + * skip this */ + if (!u->default_dependencies || + !target->default_dependencies) + return 0; + + /* Don't create loops */ + if (set_get(target->dependencies[UNIT_BEFORE], u)) + return 0; + + return unit_add_dependency(target, UNIT_AFTER, u, true); +} + +static int unit_add_target_dependencies(Unit *u) { + + static const UnitDependency deps[] = { + UNIT_REQUIRED_BY, + UNIT_REQUISITE_OF, + UNIT_WANTED_BY, + UNIT_BOUND_BY + }; + + Unit *target; + Iterator i; + unsigned k; + int r = 0; + + assert(u); + + for (k = 0; k < ELEMENTSOF(deps); k++) + SET_FOREACH(target, u->dependencies[deps[k]], i) { + r = unit_add_default_target_dependency(u, target); + if (r < 0) + return r; + } + + return r; +} + +static int unit_add_slice_dependencies(Unit *u) { + assert(u); + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return 0; + + if (UNIT_ISSET(u->slice)) + return unit_add_two_dependencies(u, UNIT_AFTER, UNIT_REQUIRES, UNIT_DEREF(u->slice), true); + + if (unit_has_name(u, SPECIAL_ROOT_SLICE)) + return 0; + + return unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_REQUIRES, SPECIAL_ROOT_SLICE, NULL, true); +} + +static int unit_add_mount_dependencies(Unit *u) { + char **i; + int r; + + assert(u); + + STRV_FOREACH(i, u->requires_mounts_for) { + char prefix[strlen(*i) + 1]; + + PATH_FOREACH_PREFIX_MORE(prefix, *i) { + _cleanup_free_ char *p = NULL; + Unit *m; + + r = unit_name_from_path(prefix, ".mount", &p); + if (r < 0) + return r; + + m = manager_get_unit(u->manager, p); + if (!m) { + /* Make sure to load the mount unit if + * it exists. If so the dependencies + * on this unit will be added later + * during the loading of the mount + * unit. */ + (void) manager_load_unit_prepare(u->manager, p, NULL, NULL, &m); + continue; + } + if (m == u) + continue; + + if (m->load_state != UNIT_LOADED) + continue; + + r = unit_add_dependency(u, UNIT_AFTER, m, true); + if (r < 0) + return r; + + if (m->fragment_path) { + r = unit_add_dependency(u, UNIT_REQUIRES, m, true); + if (r < 0) + return r; + } + } + } + + return 0; +} + +static int unit_add_startup_units(Unit *u) { + CGroupContext *c; + int r; + + c = unit_get_cgroup_context(u); + if (!c) + return 0; + + if (c->startup_cpu_shares == CGROUP_CPU_SHARES_INVALID && + c->startup_io_weight == CGROUP_WEIGHT_INVALID && + c->startup_blockio_weight == CGROUP_BLKIO_WEIGHT_INVALID) + return 0; + + r = set_ensure_allocated(&u->manager->startup_units, NULL); + if (r < 0) + return r; + + return set_put(u->manager->startup_units, u); +} + +int unit_load(Unit *u) { + int r; + + assert(u); + + if (u->in_load_queue) { + LIST_REMOVE(load_queue, u->manager->load_queue, u); + u->in_load_queue = false; + } + + if (u->type == _UNIT_TYPE_INVALID) + return -EINVAL; + + if (u->load_state != UNIT_STUB) + return 0; + + if (u->transient_file) { + r = fflush_and_check(u->transient_file); + if (r < 0) + goto fail; + + fclose(u->transient_file); + u->transient_file = NULL; + + u->fragment_mtime = now(CLOCK_REALTIME); + } + + if (UNIT_VTABLE(u)->load) { + r = UNIT_VTABLE(u)->load(u); + if (r < 0) + goto fail; + } + + if (u->load_state == UNIT_STUB) { + r = -ENOENT; + goto fail; + } + + if (u->load_state == UNIT_LOADED) { + + r = unit_add_target_dependencies(u); + if (r < 0) + goto fail; + + r = unit_add_slice_dependencies(u); + if (r < 0) + goto fail; + + r = unit_add_mount_dependencies(u); + if (r < 0) + goto fail; + + r = unit_add_startup_units(u); + if (r < 0) + goto fail; + + if (u->on_failure_job_mode == JOB_ISOLATE && set_size(u->dependencies[UNIT_ON_FAILURE]) > 1) { + log_unit_error(u, "More than one OnFailure= dependencies specified but OnFailureJobMode=isolate set. Refusing."); + r = -EINVAL; + goto fail; + } + + unit_update_cgroup_members_masks(u); + } + + assert((u->load_state != UNIT_MERGED) == !u->merged_into); + + unit_add_to_dbus_queue(unit_follow_merge(u)); + unit_add_to_gc_queue(u); + + return 0; + +fail: + u->load_state = u->load_state == UNIT_STUB ? UNIT_NOT_FOUND : UNIT_ERROR; + u->load_error = r; + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); + + log_unit_debug_errno(u, r, "Failed to load configuration: %m"); + + return r; +} + +static bool unit_condition_test_list(Unit *u, Condition *first, const char *(*to_string)(ConditionType t)) { + Condition *c; + int triggered = -1; + + assert(u); + assert(to_string); + + /* If the condition list is empty, then it is true */ + if (!first) + return true; + + /* Otherwise, if all of the non-trigger conditions apply and + * if any of the trigger conditions apply (unless there are + * none) we return true */ + LIST_FOREACH(conditions, c, first) { + int r; + + r = condition_test(c); + if (r < 0) + log_unit_warning(u, + "Couldn't determine result for %s=%s%s%s, assuming failed: %m", + to_string(c->type), + c->trigger ? "|" : "", + c->negate ? "!" : "", + c->parameter); + else + log_unit_debug(u, + "%s=%s%s%s %s.", + to_string(c->type), + c->trigger ? "|" : "", + c->negate ? "!" : "", + c->parameter, + condition_result_to_string(c->result)); + + if (!c->trigger && r <= 0) + return false; + + if (c->trigger && triggered <= 0) + triggered = r > 0; + } + + return triggered != 0; +} + +static bool unit_condition_test(Unit *u) { + assert(u); + + dual_timestamp_get(&u->condition_timestamp); + u->condition_result = unit_condition_test_list(u, u->conditions, condition_type_to_string); + + return u->condition_result; +} + +static bool unit_assert_test(Unit *u) { + assert(u); + + dual_timestamp_get(&u->assert_timestamp); + u->assert_result = unit_condition_test_list(u, u->asserts, assert_type_to_string); + + return u->assert_result; +} + +void unit_status_printf(Unit *u, const char *status, const char *unit_status_msg_format) { + DISABLE_WARNING_FORMAT_NONLITERAL; + manager_status_printf(u->manager, STATUS_TYPE_NORMAL, status, unit_status_msg_format, unit_description(u)); + REENABLE_WARNING; +} + +_pure_ static const char* unit_get_status_message_format(Unit *u, JobType t) { + const char *format; + const UnitStatusMessageFormats *format_table; + + assert(u); + assert(IN_SET(t, JOB_START, JOB_STOP, JOB_RELOAD)); + + if (t != JOB_RELOAD) { + format_table = &UNIT_VTABLE(u)->status_message_formats; + if (format_table) { + format = format_table->starting_stopping[t == JOB_STOP]; + if (format) + return format; + } + } + + /* Return generic strings */ + if (t == JOB_START) + return "Starting %s."; + else if (t == JOB_STOP) + return "Stopping %s."; + else + return "Reloading %s."; +} + +static void unit_status_print_starting_stopping(Unit *u, JobType t) { + const char *format; + + assert(u); + + /* Reload status messages have traditionally not been printed to console. */ + if (!IN_SET(t, JOB_START, JOB_STOP)) + return; + + format = unit_get_status_message_format(u, t); + + DISABLE_WARNING_FORMAT_NONLITERAL; + unit_status_printf(u, "", format); + REENABLE_WARNING; +} + +static void unit_status_log_starting_stopping_reloading(Unit *u, JobType t) { + const char *format; + char buf[LINE_MAX]; + sd_id128_t mid; + + assert(u); + + if (!IN_SET(t, JOB_START, JOB_STOP, JOB_RELOAD)) + return; + + if (log_on_console()) + return; + + /* We log status messages for all units and all operations. */ + + format = unit_get_status_message_format(u, t); + + DISABLE_WARNING_FORMAT_NONLITERAL; + snprintf(buf, sizeof buf, format, unit_description(u)); + REENABLE_WARNING; + + mid = t == JOB_START ? SD_MESSAGE_UNIT_STARTING : + t == JOB_STOP ? SD_MESSAGE_UNIT_STOPPING : + SD_MESSAGE_UNIT_RELOADING; + + /* Note that we deliberately use LOG_MESSAGE() instead of + * LOG_UNIT_MESSAGE() here, since this is supposed to mimic + * closely what is written to screen using the status output, + * which is supposed the highest level, friendliest output + * possible, which means we should avoid the low-level unit + * name. */ + log_struct(LOG_INFO, + LOG_MESSAGE_ID(mid), + LOG_UNIT_ID(u), + LOG_MESSAGE("%s", buf), + NULL); +} + +void unit_status_emit_starting_stopping_reloading(Unit *u, JobType t) { + assert(u); + assert(t >= 0); + assert(t < _JOB_TYPE_MAX); + + unit_status_log_starting_stopping_reloading(u, t); + unit_status_print_starting_stopping(u, t); +} + +int unit_start_limit_test(Unit *u) { + assert(u); + + if (ratelimit_test(&u->start_limit)) { + u->start_limit_hit = false; + return 0; + } + + log_unit_warning(u, "Start request repeated too quickly."); + u->start_limit_hit = true; + + return emergency_action(u->manager, u->start_limit_action, u->reboot_arg, "unit failed"); +} + +/* Errors: + * -EBADR: This unit type does not support starting. + * -EALREADY: Unit is already started. + * -EAGAIN: An operation is already in progress. Retry later. + * -ECANCELED: Too many requests for now. + * -EPROTO: Assert failed + * -EINVAL: Unit not loaded + * -EOPNOTSUPP: Unit type not supported + */ +int unit_start(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + /* If this is already started, then this will succeed. Note + * that this will even succeed if this unit is not startable + * by the user. This is relied on to detect when we need to + * wait for units and when waiting is finished. */ + state = unit_active_state(u); + if (UNIT_IS_ACTIVE_OR_RELOADING(state)) + return -EALREADY; + + /* Units that aren't loaded cannot be started */ + if (u->load_state != UNIT_LOADED) + return -EINVAL; + + /* If the conditions failed, don't do anything at all. If we + * already are activating this call might still be useful to + * speed up activation in case there is some hold-off time, + * but we don't want to recheck the condition in that case. */ + if (state != UNIT_ACTIVATING && + !unit_condition_test(u)) { + log_unit_debug(u, "Starting requested but condition failed. Not starting unit."); + return -EALREADY; + } + + /* If the asserts failed, fail the entire job */ + if (state != UNIT_ACTIVATING && + !unit_assert_test(u)) { + log_unit_notice(u, "Starting requested but asserts failed."); + return -EPROTO; + } + + /* Units of types that aren't supported cannot be + * started. Note that we do this test only after the condition + * checks, so that we rather return condition check errors + * (which are usually not considered a true failure) than "not + * supported" errors (which are considered a failure). + */ + if (!unit_supported(u)) + return -EOPNOTSUPP; + + /* Forward to the main object, if we aren't it. */ + following = unit_following(u); + if (following) { + log_unit_debug(u, "Redirecting start request from %s to %s.", u->id, following->id); + return unit_start(following); + } + + /* If it is stopped, but we cannot start it, then fail */ + if (!UNIT_VTABLE(u)->start) + return -EBADR; + + /* We don't suppress calls to ->start() here when we are + * already starting, to allow this request to be used as a + * "hurry up" call, for example when the unit is in some "auto + * restart" state where it waits for a holdoff timer to elapse + * before it will start again. */ + + unit_add_to_dbus_queue(u); + + return UNIT_VTABLE(u)->start(u); +} + +bool unit_can_start(Unit *u) { + assert(u); + + if (u->load_state != UNIT_LOADED) + return false; + + if (!unit_supported(u)) + return false; + + return !!UNIT_VTABLE(u)->start; +} + +bool unit_can_isolate(Unit *u) { + assert(u); + + return unit_can_start(u) && + u->allow_isolate; +} + +/* Errors: + * -EBADR: This unit type does not support stopping. + * -EALREADY: Unit is already stopped. + * -EAGAIN: An operation is already in progress. Retry later. + */ +int unit_stop(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + state = unit_active_state(u); + if (UNIT_IS_INACTIVE_OR_FAILED(state)) + return -EALREADY; + + following = unit_following(u); + if (following) { + log_unit_debug(u, "Redirecting stop request from %s to %s.", u->id, following->id); + return unit_stop(following); + } + + if (!UNIT_VTABLE(u)->stop) + return -EBADR; + + unit_add_to_dbus_queue(u); + + return UNIT_VTABLE(u)->stop(u); +} + +bool unit_can_stop(Unit *u) { + assert(u); + + if (!unit_supported(u)) + return false; + + if (u->perpetual) + return false; + + return !!UNIT_VTABLE(u)->stop; +} + +/* Errors: + * -EBADR: This unit type does not support reloading. + * -ENOEXEC: Unit is not started. + * -EAGAIN: An operation is already in progress. Retry later. + */ +int unit_reload(Unit *u) { + UnitActiveState state; + Unit *following; + + assert(u); + + if (u->load_state != UNIT_LOADED) + return -EINVAL; + + if (!unit_can_reload(u)) + return -EBADR; + + state = unit_active_state(u); + if (state == UNIT_RELOADING) + return -EALREADY; + + if (state != UNIT_ACTIVE) { + log_unit_warning(u, "Unit cannot be reloaded because it is inactive."); + return -ENOEXEC; + } + + following = unit_following(u); + if (following) { + log_unit_debug(u, "Redirecting reload request from %s to %s.", u->id, following->id); + return unit_reload(following); + } + + unit_add_to_dbus_queue(u); + + return UNIT_VTABLE(u)->reload(u); +} + +bool unit_can_reload(Unit *u) { + assert(u); + + if (!UNIT_VTABLE(u)->reload) + return false; + + if (!UNIT_VTABLE(u)->can_reload) + return true; + + return UNIT_VTABLE(u)->can_reload(u); +} + +static void unit_check_unneeded(Unit *u) { + + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + + static const UnitDependency needed_dependencies[] = { + UNIT_REQUIRED_BY, + UNIT_REQUISITE_OF, + UNIT_WANTED_BY, + UNIT_BOUND_BY, + }; + + Unit *other; + Iterator i; + unsigned j; + int r; + + assert(u); + + /* If this service shall be shut down when unneeded then do + * so. */ + + if (!u->stop_when_unneeded) + return; + + if (!UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) + return; + + for (j = 0; j < ELEMENTSOF(needed_dependencies); j++) + SET_FOREACH(other, u->dependencies[needed_dependencies[j]], i) + if (unit_active_or_pending(other)) + return; + + /* If stopping a unit fails continuously we might enter a stop + * loop here, hence stop acting on the service being + * unnecessary after a while. */ + if (!ratelimit_test(&u->auto_stop_ratelimit)) { + log_unit_warning(u, "Unit not needed anymore, but not stopping since we tried this too often recently."); + return; + } + + log_unit_info(u, "Unit not needed anymore. Stopping."); + + /* Ok, nobody needs us anymore. Sniff. Then let's commit suicide */ + r = manager_add_job(u->manager, JOB_STOP, u, JOB_FAIL, &error, NULL); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to enqueue stop job, ignoring: %s", bus_error_message(&error, r)); +} + +static void unit_check_binds_to(Unit *u) { + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + bool stop = false; + Unit *other; + Iterator i; + int r; + + assert(u); + + if (u->job) + return; + + if (unit_active_state(u) != UNIT_ACTIVE) + return; + + SET_FOREACH(other, u->dependencies[UNIT_BINDS_TO], i) { + if (other->job) + continue; + + if (!UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(other))) + continue; + + stop = true; + break; + } + + if (!stop) + return; + + /* If stopping a unit fails continuously we might enter a stop + * loop here, hence stop acting on the service being + * unnecessary after a while. */ + if (!ratelimit_test(&u->auto_stop_ratelimit)) { + log_unit_warning(u, "Unit is bound to inactive unit %s, but not stopping since we tried this too often recently.", other->id); + return; + } + + assert(other); + log_unit_info(u, "Unit is bound to inactive unit %s. Stopping, too.", other->id); + + /* A unit we need to run is gone. Sniff. Let's stop this. */ + r = manager_add_job(u->manager, JOB_STOP, u, JOB_FAIL, &error, NULL); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to enqueue stop job, ignoring: %s", bus_error_message(&error, r)); +} + +static void retroactively_start_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))); + + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_REPLACE, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_BINDS_TO], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_REPLACE, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) + if (!set_get(u->dependencies[UNIT_AFTER], other) && + !UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_START, other, JOB_FAIL, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_CONFLICTS], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, NULL, NULL); + + SET_FOREACH(other, u->dependencies[UNIT_CONFLICTED_BY], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, NULL, NULL); +} + +static void retroactively_stop_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))); + + /* Pull down units which are bound to us recursively if enabled */ + SET_FOREACH(other, u->dependencies[UNIT_BOUND_BY], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, NULL, NULL); +} + +static void check_unneeded_dependencies(Unit *u) { + Iterator i; + Unit *other; + + assert(u); + assert(UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))); + + /* Garbage collect services that might not be needed anymore, if enabled */ + SET_FOREACH(other, u->dependencies[UNIT_REQUIRES], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_REQUISITE], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); + SET_FOREACH(other, u->dependencies[UNIT_BINDS_TO], i) + if (!UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(other))) + unit_check_unneeded(other); +} + +void unit_start_on_failure(Unit *u) { + Unit *other; + Iterator i; + + assert(u); + + if (set_size(u->dependencies[UNIT_ON_FAILURE]) <= 0) + return; + + log_unit_info(u, "Triggering OnFailure= dependencies."); + + SET_FOREACH(other, u->dependencies[UNIT_ON_FAILURE], i) { + int r; + + r = manager_add_job(u->manager, JOB_START, other, u->on_failure_job_mode, NULL, NULL); + if (r < 0) + log_unit_error_errno(u, r, "Failed to enqueue OnFailure= job: %m"); + } +} + +void unit_trigger_notify(Unit *u) { + Unit *other; + Iterator i; + + assert(u); + + SET_FOREACH(other, u->dependencies[UNIT_TRIGGERED_BY], i) + if (UNIT_VTABLE(other)->trigger_notify) + UNIT_VTABLE(other)->trigger_notify(other, u); +} + +void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_success) { + Manager *m; + bool unexpected; + + assert(u); + assert(os < _UNIT_ACTIVE_STATE_MAX); + assert(ns < _UNIT_ACTIVE_STATE_MAX); + + /* Note that this is called for all low-level state changes, + * even if they might map to the same high-level + * UnitActiveState! That means that ns == os is an expected + * behavior here. For example: if a mount point is remounted + * this function will be called too! */ + + m = u->manager; + + /* Update timestamps for state changes */ + if (!MANAGER_IS_RELOADING(m)) { + dual_timestamp_get(&u->state_change_timestamp); + + if (UNIT_IS_INACTIVE_OR_FAILED(os) && !UNIT_IS_INACTIVE_OR_FAILED(ns)) + u->inactive_exit_timestamp = u->state_change_timestamp; + else if (!UNIT_IS_INACTIVE_OR_FAILED(os) && UNIT_IS_INACTIVE_OR_FAILED(ns)) + u->inactive_enter_timestamp = u->state_change_timestamp; + + if (!UNIT_IS_ACTIVE_OR_RELOADING(os) && UNIT_IS_ACTIVE_OR_RELOADING(ns)) + u->active_enter_timestamp = u->state_change_timestamp; + else if (UNIT_IS_ACTIVE_OR_RELOADING(os) && !UNIT_IS_ACTIVE_OR_RELOADING(ns)) + u->active_exit_timestamp = u->state_change_timestamp; + } + + /* Keep track of failed units */ + (void) manager_update_failed_units(u->manager, u, ns == UNIT_FAILED); + + /* Make sure the cgroup is always removed when we become inactive */ + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + unit_prune_cgroup(u); + + /* Note that this doesn't apply to RemainAfterExit services exiting + * successfully, since there's no change of state in that case. Which is + * why it is handled in service_set_state() */ + if (UNIT_IS_INACTIVE_OR_FAILED(os) != UNIT_IS_INACTIVE_OR_FAILED(ns)) { + ExecContext *ec; + + ec = unit_get_exec_context(u); + if (ec && exec_context_may_touch_console(ec)) { + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) { + m->n_on_console--; + + if (m->n_on_console == 0) + /* unset no_console_output flag, since the console is free */ + m->no_console_output = false; + } else + m->n_on_console++; + } + } + + if (u->job) { + unexpected = false; + + if (u->job->state == JOB_WAITING) + + /* So we reached a different state for this + * job. Let's see if we can run it now if it + * failed previously due to EAGAIN. */ + job_add_to_run_queue(u->job); + + /* Let's check whether this state change constitutes a + * finished job, or maybe contradicts a running job and + * hence needs to invalidate jobs. */ + + switch (u->job->type) { + + case JOB_START: + case JOB_VERIFY_ACTIVE: + + if (UNIT_IS_ACTIVE_OR_RELOADING(ns)) + job_finish_and_invalidate(u->job, JOB_DONE, true, false); + else if (u->job->state == JOB_RUNNING && ns != UNIT_ACTIVATING) { + unexpected = true; + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, ns == UNIT_FAILED ? JOB_FAILED : JOB_DONE, true, false); + } + + break; + + case JOB_RELOAD: + case JOB_RELOAD_OR_START: + case JOB_TRY_RELOAD: + + if (u->job->state == JOB_RUNNING) { + if (ns == UNIT_ACTIVE) + job_finish_and_invalidate(u->job, reload_success ? JOB_DONE : JOB_FAILED, true, false); + else if (ns != UNIT_ACTIVATING && ns != UNIT_RELOADING) { + unexpected = true; + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, ns == UNIT_FAILED ? JOB_FAILED : JOB_DONE, true, false); + } + } + + break; + + case JOB_STOP: + case JOB_RESTART: + case JOB_TRY_RESTART: + + if (UNIT_IS_INACTIVE_OR_FAILED(ns)) + job_finish_and_invalidate(u->job, JOB_DONE, true, false); + else if (u->job->state == JOB_RUNNING && ns != UNIT_DEACTIVATING) { + unexpected = true; + job_finish_and_invalidate(u->job, JOB_FAILED, true, false); + } + + break; + + default: + assert_not_reached("Job type unknown"); + } + + } else + unexpected = true; + + if (!MANAGER_IS_RELOADING(m)) { + + /* If this state change happened without being + * requested by a job, then let's retroactively start + * or stop dependencies. We skip that step when + * deserializing, since we don't want to create any + * additional jobs just because something is already + * activated. */ + + if (unexpected) { + if (UNIT_IS_INACTIVE_OR_FAILED(os) && UNIT_IS_ACTIVE_OR_ACTIVATING(ns)) + retroactively_start_dependencies(u); + else if (UNIT_IS_ACTIVE_OR_ACTIVATING(os) && UNIT_IS_INACTIVE_OR_DEACTIVATING(ns)) + retroactively_stop_dependencies(u); + } + + /* stop unneeded units regardless if going down was expected or not */ + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(ns)) + check_unneeded_dependencies(u); + + if (ns != os && ns == UNIT_FAILED) { + log_unit_notice(u, "Unit entered failed state."); + unit_start_on_failure(u); + } + } + + /* Some names are special */ + if (UNIT_IS_ACTIVE_OR_RELOADING(ns)) { + + if (unit_has_name(u, SPECIAL_DBUS_SERVICE)) + /* The bus might have just become available, + * hence try to connect to it, if we aren't + * yet connected. */ + bus_init(m, true); + + if (u->type == UNIT_SERVICE && + !UNIT_IS_ACTIVE_OR_RELOADING(os) && + !MANAGER_IS_RELOADING(m)) { + /* Write audit record if we have just finished starting up */ + manager_send_unit_audit(m, u, AUDIT_SERVICE_START, true); + u->in_audit = true; + } + + if (!UNIT_IS_ACTIVE_OR_RELOADING(os)) + manager_send_unit_plymouth(m, u); + + } else { + + /* We don't care about D-Bus here, since we'll get an + * asynchronous notification for it anyway. */ + + if (u->type == UNIT_SERVICE && + UNIT_IS_INACTIVE_OR_FAILED(ns) && + !UNIT_IS_INACTIVE_OR_FAILED(os) && + !MANAGER_IS_RELOADING(m)) { + + /* Hmm, if there was no start record written + * write it now, so that we always have a nice + * pair */ + if (!u->in_audit) { + manager_send_unit_audit(m, u, AUDIT_SERVICE_START, ns == UNIT_INACTIVE); + + if (ns == UNIT_INACTIVE) + manager_send_unit_audit(m, u, AUDIT_SERVICE_STOP, true); + } else + /* Write audit record if we have just finished shutting down */ + manager_send_unit_audit(m, u, AUDIT_SERVICE_STOP, ns == UNIT_INACTIVE); + + u->in_audit = false; + } + } + + manager_recheck_journal(m); + unit_trigger_notify(u); + + if (!MANAGER_IS_RELOADING(u->manager)) { + /* Maybe we finished startup and are now ready for + * being stopped because unneeded? */ + unit_check_unneeded(u); + + /* Maybe we finished startup, but something we needed + * has vanished? Let's die then. (This happens when + * something BindsTo= to a Type=oneshot unit, as these + * units go directly from starting to inactive, + * without ever entering started.) */ + unit_check_binds_to(u); + } + + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); +} + +int unit_watch_pid(Unit *u, pid_t pid) { + int q, r; + + assert(u); + assert(pid >= 1); + + /* Watch a specific PID. We only support one or two units + * watching each PID for now, not more. */ + + r = set_ensure_allocated(&u->pids, NULL); + if (r < 0) + return r; + + r = hashmap_ensure_allocated(&u->manager->watch_pids1, NULL); + if (r < 0) + return r; + + r = hashmap_put(u->manager->watch_pids1, PID_TO_PTR(pid), u); + if (r == -EEXIST) { + r = hashmap_ensure_allocated(&u->manager->watch_pids2, NULL); + if (r < 0) + return r; + + r = hashmap_put(u->manager->watch_pids2, PID_TO_PTR(pid), u); + } + + q = set_put(u->pids, PID_TO_PTR(pid)); + if (q < 0) + return q; + + return r; +} + +void unit_unwatch_pid(Unit *u, pid_t pid) { + assert(u); + assert(pid >= 1); + + (void) hashmap_remove_value(u->manager->watch_pids1, PID_TO_PTR(pid), u); + (void) hashmap_remove_value(u->manager->watch_pids2, PID_TO_PTR(pid), u); + (void) set_remove(u->pids, PID_TO_PTR(pid)); +} + +void unit_unwatch_all_pids(Unit *u) { + assert(u); + + while (!set_isempty(u->pids)) + unit_unwatch_pid(u, PTR_TO_PID(set_first(u->pids))); + + u->pids = set_free(u->pids); +} + +void unit_tidy_watch_pids(Unit *u, pid_t except1, pid_t except2) { + Iterator i; + void *e; + + assert(u); + + /* Cleans dead PIDs from our list */ + + SET_FOREACH(e, u->pids, i) { + pid_t pid = PTR_TO_PID(e); + + if (pid == except1 || pid == except2) + continue; + + if (!pid_is_unwaited(pid)) + unit_unwatch_pid(u, pid); + } +} + +bool unit_job_is_applicable(Unit *u, JobType j) { + assert(u); + assert(j >= 0 && j < _JOB_TYPE_MAX); + + switch (j) { + + case JOB_VERIFY_ACTIVE: + case JOB_START: + case JOB_NOP: + /* Note that we don't check unit_can_start() here. That's because .device units and suchlike are not + * startable by us but may appear due to external events, and it thus makes sense to permit enqueing + * jobs for it. */ + return true; + + case JOB_STOP: + /* Similar as above. However, perpetual units can never be stopped (neither explicitly nor due to + * external events), hence it makes no sense to permit enqueing such a request either. */ + return !u->perpetual; + + case JOB_RESTART: + case JOB_TRY_RESTART: + return unit_can_stop(u) && unit_can_start(u); + + case JOB_RELOAD: + case JOB_TRY_RELOAD: + return unit_can_reload(u); + + case JOB_RELOAD_OR_START: + return unit_can_reload(u) && unit_can_start(u); + + default: + assert_not_reached("Invalid job type"); + } +} + +static void maybe_warn_about_dependency(Unit *u, const char *other, UnitDependency dependency) { + assert(u); + + /* Only warn about some unit types */ + if (!IN_SET(dependency, UNIT_CONFLICTS, UNIT_CONFLICTED_BY, UNIT_BEFORE, UNIT_AFTER, UNIT_ON_FAILURE, UNIT_TRIGGERS, UNIT_TRIGGERED_BY)) + return; + + if (streq_ptr(u->id, other)) + log_unit_warning(u, "Dependency %s=%s dropped", unit_dependency_to_string(dependency), u->id); + else + log_unit_warning(u, "Dependency %s=%s dropped, merged into %s", unit_dependency_to_string(dependency), strna(other), u->id); +} + +int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_reference) { + + static const UnitDependency inverse_table[_UNIT_DEPENDENCY_MAX] = { + [UNIT_REQUIRES] = UNIT_REQUIRED_BY, + [UNIT_WANTS] = UNIT_WANTED_BY, + [UNIT_REQUISITE] = UNIT_REQUISITE_OF, + [UNIT_BINDS_TO] = UNIT_BOUND_BY, + [UNIT_PART_OF] = UNIT_CONSISTS_OF, + [UNIT_REQUIRED_BY] = UNIT_REQUIRES, + [UNIT_REQUISITE_OF] = UNIT_REQUISITE, + [UNIT_WANTED_BY] = UNIT_WANTS, + [UNIT_BOUND_BY] = UNIT_BINDS_TO, + [UNIT_CONSISTS_OF] = UNIT_PART_OF, + [UNIT_CONFLICTS] = UNIT_CONFLICTED_BY, + [UNIT_CONFLICTED_BY] = UNIT_CONFLICTS, + [UNIT_BEFORE] = UNIT_AFTER, + [UNIT_AFTER] = UNIT_BEFORE, + [UNIT_ON_FAILURE] = _UNIT_DEPENDENCY_INVALID, + [UNIT_REFERENCES] = UNIT_REFERENCED_BY, + [UNIT_REFERENCED_BY] = UNIT_REFERENCES, + [UNIT_TRIGGERS] = UNIT_TRIGGERED_BY, + [UNIT_TRIGGERED_BY] = UNIT_TRIGGERS, + [UNIT_PROPAGATES_RELOAD_TO] = UNIT_RELOAD_PROPAGATED_FROM, + [UNIT_RELOAD_PROPAGATED_FROM] = UNIT_PROPAGATES_RELOAD_TO, + [UNIT_JOINS_NAMESPACE_OF] = UNIT_JOINS_NAMESPACE_OF, + }; + int r, q = 0, v = 0, w = 0; + Unit *orig_u = u, *orig_other = other; + + assert(u); + assert(d >= 0 && d < _UNIT_DEPENDENCY_MAX); + assert(other); + + u = unit_follow_merge(u); + other = unit_follow_merge(other); + + /* We won't allow dependencies on ourselves. We will not + * consider them an error however. */ + if (u == other) { + maybe_warn_about_dependency(orig_u, orig_other->id, d); + return 0; + } + + if (d == UNIT_BEFORE && other->type == UNIT_DEVICE) { + log_unit_warning(u, "Dependency Before=%s ignored (.device units cannot be delayed)", other->id); + return 0; + } + + r = set_ensure_allocated(&u->dependencies[d], NULL); + if (r < 0) + return r; + + if (inverse_table[d] != _UNIT_DEPENDENCY_INVALID) { + r = set_ensure_allocated(&other->dependencies[inverse_table[d]], NULL); + if (r < 0) + return r; + } + + if (add_reference) { + r = set_ensure_allocated(&u->dependencies[UNIT_REFERENCES], NULL); + if (r < 0) + return r; + + r = set_ensure_allocated(&other->dependencies[UNIT_REFERENCED_BY], NULL); + if (r < 0) + return r; + } + + q = set_put(u->dependencies[d], other); + if (q < 0) + return q; + + if (inverse_table[d] != _UNIT_DEPENDENCY_INVALID && inverse_table[d] != d) { + v = set_put(other->dependencies[inverse_table[d]], u); + if (v < 0) { + r = v; + goto fail; + } + } + + if (add_reference) { + w = set_put(u->dependencies[UNIT_REFERENCES], other); + if (w < 0) { + r = w; + goto fail; + } + + r = set_put(other->dependencies[UNIT_REFERENCED_BY], u); + if (r < 0) + goto fail; + } + + unit_add_to_dbus_queue(u); + return 0; + +fail: + if (q > 0) + set_remove(u->dependencies[d], other); + + if (v > 0) + set_remove(other->dependencies[inverse_table[d]], u); + + if (w > 0) + set_remove(u->dependencies[UNIT_REFERENCES], other); + + return r; +} + +int unit_add_two_dependencies(Unit *u, UnitDependency d, UnitDependency e, Unit *other, bool add_reference) { + int r; + + assert(u); + + r = unit_add_dependency(u, d, other, add_reference); + if (r < 0) + return r; + + return unit_add_dependency(u, e, other, add_reference); +} + +static int resolve_template(Unit *u, const char *name, const char*path, char **buf, const char **ret) { + int r; + + assert(u); + assert(name || path); + assert(buf); + assert(ret); + + if (!name) + name = basename(path); + + if (!unit_name_is_valid(name, UNIT_NAME_TEMPLATE)) { + *buf = NULL; + *ret = name; + return 0; + } + + if (u->instance) + r = unit_name_replace_instance(name, u->instance, buf); + else { + _cleanup_free_ char *i = NULL; + + r = unit_name_to_prefix(u->id, &i); + if (r < 0) + return r; + + r = unit_name_replace_instance(name, i, buf); + } + if (r < 0) + return r; + + *ret = *buf; + return 0; +} + +int unit_add_dependency_by_name(Unit *u, UnitDependency d, const char *name, const char *path, bool add_reference) { + _cleanup_free_ char *buf = NULL; + Unit *other; + int r; + + assert(u); + assert(name || path); + + r = resolve_template(u, name, path, &buf, &name); + if (r < 0) + return r; + + r = manager_load_unit(u->manager, name, path, NULL, &other); + if (r < 0) + return r; + + return unit_add_dependency(u, d, other, add_reference); +} + +int unit_add_two_dependencies_by_name(Unit *u, UnitDependency d, UnitDependency e, const char *name, const char *path, bool add_reference) { + _cleanup_free_ char *buf = NULL; + Unit *other; + int r; + + assert(u); + assert(name || path); + + r = resolve_template(u, name, path, &buf, &name); + if (r < 0) + return r; + + r = manager_load_unit(u->manager, name, path, NULL, &other); + if (r < 0) + return r; + + return unit_add_two_dependencies(u, d, e, other, add_reference); +} + +int set_unit_path(const char *p) { + /* This is mostly for debug purposes */ + if (setenv("SYSTEMD_UNIT_PATH", p, 1) < 0) + return -errno; + + return 0; +} + +char *unit_dbus_path(Unit *u) { + assert(u); + + if (!u->id) + return NULL; + + return unit_dbus_path_from_name(u->id); +} + +char *unit_dbus_path_invocation_id(Unit *u) { + assert(u); + + if (sd_id128_is_null(u->invocation_id)) + return NULL; + + return unit_dbus_path_from_name(u->invocation_id_string); +} + +int unit_set_slice(Unit *u, Unit *slice) { + assert(u); + assert(slice); + + /* Sets the unit slice if it has not been set before. Is extra + * careful, to only allow this for units that actually have a + * cgroup context. Also, we don't allow to set this for slices + * (since the parent slice is derived from the name). Make + * sure the unit we set is actually a slice. */ + + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return -EOPNOTSUPP; + + if (u->type == UNIT_SLICE) + return -EINVAL; + + if (unit_active_state(u) != UNIT_INACTIVE) + return -EBUSY; + + if (slice->type != UNIT_SLICE) + return -EINVAL; + + if (unit_has_name(u, SPECIAL_INIT_SCOPE) && + !unit_has_name(slice, SPECIAL_ROOT_SLICE)) + return -EPERM; + + if (UNIT_DEREF(u->slice) == slice) + return 0; + + /* Disallow slice changes if @u is already bound to cgroups */ + if (UNIT_ISSET(u->slice) && u->cgroup_realized) + return -EBUSY; + + unit_ref_unset(&u->slice); + unit_ref_set(&u->slice, slice); + return 1; +} + +int unit_set_default_slice(Unit *u) { + _cleanup_free_ char *b = NULL; + const char *slice_name; + Unit *slice; + int r; + + assert(u); + + if (UNIT_ISSET(u->slice)) + return 0; + + if (u->instance) { + _cleanup_free_ char *prefix = NULL, *escaped = NULL; + + /* Implicitly place all instantiated units in their + * own per-template slice */ + + r = unit_name_to_prefix(u->id, &prefix); + if (r < 0) + return r; + + /* The prefix is already escaped, but it might include + * "-" which has a special meaning for slice units, + * hence escape it here extra. */ + escaped = unit_name_escape(prefix); + if (!escaped) + return -ENOMEM; + + if (MANAGER_IS_SYSTEM(u->manager)) + b = strjoin("system-", escaped, ".slice", NULL); + else + b = strappend(escaped, ".slice"); + if (!b) + return -ENOMEM; + + slice_name = b; + } else + slice_name = + MANAGER_IS_SYSTEM(u->manager) && !unit_has_name(u, SPECIAL_INIT_SCOPE) + ? SPECIAL_SYSTEM_SLICE + : SPECIAL_ROOT_SLICE; + + r = manager_load_unit(u->manager, slice_name, NULL, NULL, &slice); + if (r < 0) + return r; + + return unit_set_slice(u, slice); +} + +const char *unit_slice_name(Unit *u) { + assert(u); + + if (!UNIT_ISSET(u->slice)) + return NULL; + + return UNIT_DEREF(u->slice)->id; +} + +int unit_load_related_unit(Unit *u, const char *type, Unit **_found) { + _cleanup_free_ char *t = NULL; + int r; + + assert(u); + assert(type); + assert(_found); + + r = unit_name_change_suffix(u->id, type, &t); + if (r < 0) + return r; + if (unit_has_name(u, t)) + return -EINVAL; + + r = manager_load_unit(u->manager, t, NULL, NULL, _found); + assert(r < 0 || *_found != u); + return r; +} + +static int signal_name_owner_changed(sd_bus_message *message, void *userdata, sd_bus_error *error) { + const char *name, *old_owner, *new_owner; + Unit *u = userdata; + int r; + + assert(message); + assert(u); + + r = sd_bus_message_read(message, "sss", &name, &old_owner, &new_owner); + if (r < 0) { + bus_log_parse_error(r); + return 0; + } + + if (UNIT_VTABLE(u)->bus_name_owner_change) + UNIT_VTABLE(u)->bus_name_owner_change(u, name, old_owner, new_owner); + + return 0; +} + +int unit_install_bus_match(Unit *u, sd_bus *bus, const char *name) { + const char *match; + + assert(u); + assert(bus); + assert(name); + + if (u->match_bus_slot) + return -EBUSY; + + match = strjoina("type='signal'," + "sender='org.freedesktop.DBus'," + "path='/org/freedesktop/DBus'," + "interface='org.freedesktop.DBus'," + "member='NameOwnerChanged'," + "arg0='", name, "'"); + + return sd_bus_add_match(bus, &u->match_bus_slot, match, signal_name_owner_changed, u); +} + +int unit_watch_bus_name(Unit *u, const char *name) { + int r; + + assert(u); + assert(name); + + /* Watch a specific name on the bus. We only support one unit + * watching each name for now. */ + + if (u->manager->api_bus) { + /* If the bus is already available, install the match directly. + * Otherwise, just put the name in the list. bus_setup_api() will take care later. */ + r = unit_install_bus_match(u, u->manager->api_bus, name); + if (r < 0) + return log_warning_errno(r, "Failed to subscribe to NameOwnerChanged signal for '%s': %m", name); + } + + r = hashmap_put(u->manager->watch_bus, name, u); + if (r < 0) { + u->match_bus_slot = sd_bus_slot_unref(u->match_bus_slot); + return log_warning_errno(r, "Failed to put bus name to hashmap: %m"); + } + + return 0; +} + +void unit_unwatch_bus_name(Unit *u, const char *name) { + assert(u); + assert(name); + + hashmap_remove_value(u->manager->watch_bus, name, u); + u->match_bus_slot = sd_bus_slot_unref(u->match_bus_slot); +} + +bool unit_can_serialize(Unit *u) { + assert(u); + + return UNIT_VTABLE(u)->serialize && UNIT_VTABLE(u)->deserialize_item; +} + +int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { + int r; + + assert(u); + assert(f); + assert(fds); + + if (unit_can_serialize(u)) { + ExecRuntime *rt; + + r = UNIT_VTABLE(u)->serialize(u, f, fds); + if (r < 0) + return r; + + rt = unit_get_exec_runtime(u); + if (rt) { + r = exec_runtime_serialize(u, rt, f, fds); + if (r < 0) + return r; + } + } + + dual_timestamp_serialize(f, "state-change-timestamp", &u->state_change_timestamp); + + dual_timestamp_serialize(f, "inactive-exit-timestamp", &u->inactive_exit_timestamp); + dual_timestamp_serialize(f, "active-enter-timestamp", &u->active_enter_timestamp); + dual_timestamp_serialize(f, "active-exit-timestamp", &u->active_exit_timestamp); + dual_timestamp_serialize(f, "inactive-enter-timestamp", &u->inactive_enter_timestamp); + + dual_timestamp_serialize(f, "condition-timestamp", &u->condition_timestamp); + dual_timestamp_serialize(f, "assert-timestamp", &u->assert_timestamp); + + if (dual_timestamp_is_set(&u->condition_timestamp)) + unit_serialize_item(u, f, "condition-result", yes_no(u->condition_result)); + + if (dual_timestamp_is_set(&u->assert_timestamp)) + unit_serialize_item(u, f, "assert-result", yes_no(u->assert_result)); + + unit_serialize_item(u, f, "transient", yes_no(u->transient)); + + unit_serialize_item_format(u, f, "cpu-usage-base", "%" PRIu64, u->cpu_usage_base); + if (u->cpu_usage_last != NSEC_INFINITY) + unit_serialize_item_format(u, f, "cpu-usage-last", "%" PRIu64, u->cpu_usage_last); + + if (u->cgroup_path) + unit_serialize_item(u, f, "cgroup", u->cgroup_path); + unit_serialize_item(u, f, "cgroup-realized", yes_no(u->cgroup_realized)); + + if (uid_is_valid(u->ref_uid)) + unit_serialize_item_format(u, f, "ref-uid", UID_FMT, u->ref_uid); + if (gid_is_valid(u->ref_gid)) + unit_serialize_item_format(u, f, "ref-gid", GID_FMT, u->ref_gid); + + if (!sd_id128_is_null(u->invocation_id)) + unit_serialize_item_format(u, f, "invocation-id", SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(u->invocation_id)); + + bus_track_serialize(u->bus_track, f, "ref"); + + if (serialize_jobs) { + if (u->job) { + fprintf(f, "job\n"); + job_serialize(u->job, f); + } + + if (u->nop_job) { + fprintf(f, "job\n"); + job_serialize(u->nop_job, f); + } + } + + /* End marker */ + fputc('\n', f); + return 0; +} + +int unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value) { + assert(u); + assert(f); + assert(key); + + if (!value) + return 0; + + fputs(key, f); + fputc('=', f); + fputs(value, f); + fputc('\n', f); + + return 1; +} + +int unit_serialize_item_escaped(Unit *u, FILE *f, const char *key, const char *value) { + _cleanup_free_ char *c = NULL; + + assert(u); + assert(f); + assert(key); + + if (!value) + return 0; + + c = cescape(value); + if (!c) + return -ENOMEM; + + fputs(key, f); + fputc('=', f); + fputs(c, f); + fputc('\n', f); + + return 1; +} + +int unit_serialize_item_fd(Unit *u, FILE *f, FDSet *fds, const char *key, int fd) { + int copy; + + assert(u); + assert(f); + assert(key); + + if (fd < 0) + return 0; + + copy = fdset_put_dup(fds, fd); + if (copy < 0) + return copy; + + fprintf(f, "%s=%i\n", key, copy); + return 1; +} + +void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *format, ...) { + va_list ap; + + assert(u); + assert(f); + assert(key); + assert(format); + + fputs(key, f); + fputc('=', f); + + va_start(ap, format); + vfprintf(f, format, ap); + va_end(ap); + + fputc('\n', f); +} + +int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { + ExecRuntime **rt = NULL; + size_t offset; + int r; + + assert(u); + assert(f); + assert(fds); + + offset = UNIT_VTABLE(u)->exec_runtime_offset; + if (offset > 0) + rt = (ExecRuntime**) ((uint8_t*) u + offset); + + for (;;) { + char line[LINE_MAX], *l, *v; + size_t k; + + if (!fgets(line, sizeof(line), f)) { + if (feof(f)) + return 0; + return -errno; + } + + char_array_0(line); + l = strstrip(line); + + /* End marker */ + if (isempty(l)) + break; + + k = strcspn(l, "="); + + if (l[k] == '=') { + l[k] = 0; + v = l+k+1; + } else + v = l+k; + + if (streq(l, "job")) { + if (v[0] == '\0') { + /* new-style serialized job */ + Job *j; + + j = job_new_raw(u); + if (!j) + return log_oom(); + + r = job_deserialize(j, f); + if (r < 0) { + job_free(j); + return r; + } + + r = hashmap_put(u->manager->jobs, UINT32_TO_PTR(j->id), j); + if (r < 0) { + job_free(j); + return r; + } + + r = job_install_deserialized(j); + if (r < 0) { + hashmap_remove(u->manager->jobs, UINT32_TO_PTR(j->id)); + job_free(j); + return r; + } + } else /* legacy for pre-44 */ + log_unit_warning(u, "Update from too old systemd versions are unsupported, cannot deserialize job: %s", v); + continue; + } else if (streq(l, "state-change-timestamp")) { + dual_timestamp_deserialize(v, &u->state_change_timestamp); + continue; + } else if (streq(l, "inactive-exit-timestamp")) { + dual_timestamp_deserialize(v, &u->inactive_exit_timestamp); + continue; + } else if (streq(l, "active-enter-timestamp")) { + dual_timestamp_deserialize(v, &u->active_enter_timestamp); + continue; + } else if (streq(l, "active-exit-timestamp")) { + dual_timestamp_deserialize(v, &u->active_exit_timestamp); + continue; + } else if (streq(l, "inactive-enter-timestamp")) { + dual_timestamp_deserialize(v, &u->inactive_enter_timestamp); + continue; + } else if (streq(l, "condition-timestamp")) { + dual_timestamp_deserialize(v, &u->condition_timestamp); + continue; + } else if (streq(l, "assert-timestamp")) { + dual_timestamp_deserialize(v, &u->assert_timestamp); + continue; + } else if (streq(l, "condition-result")) { + + r = parse_boolean(v); + if (r < 0) + log_unit_debug(u, "Failed to parse condition result value %s, ignoring.", v); + else + u->condition_result = r; + + continue; + + } else if (streq(l, "assert-result")) { + + r = parse_boolean(v); + if (r < 0) + log_unit_debug(u, "Failed to parse assert result value %s, ignoring.", v); + else + u->assert_result = r; + + continue; + + } else if (streq(l, "transient")) { + + r = parse_boolean(v); + if (r < 0) + log_unit_debug(u, "Failed to parse transient bool %s, ignoring.", v); + else + u->transient = r; + + continue; + + } else if (STR_IN_SET(l, "cpu-usage-base", "cpuacct-usage-base")) { + + r = safe_atou64(v, &u->cpu_usage_base); + if (r < 0) + log_unit_debug(u, "Failed to parse CPU usage base %s, ignoring.", v); + + continue; + + } else if (streq(l, "cpu-usage-last")) { + + r = safe_atou64(v, &u->cpu_usage_last); + if (r < 0) + log_unit_debug(u, "Failed to read CPU usage last %s, ignoring.", v); + + continue; + + } else if (streq(l, "cgroup")) { + + r = unit_set_cgroup_path(u, v); + if (r < 0) + log_unit_debug_errno(u, r, "Failed to set cgroup path %s, ignoring: %m", v); + + (void) unit_watch_cgroup(u); + + continue; + } else if (streq(l, "cgroup-realized")) { + int b; + + b = parse_boolean(v); + if (b < 0) + log_unit_debug(u, "Failed to parse cgroup-realized bool %s, ignoring.", v); + else + u->cgroup_realized = b; + + continue; + + } else if (streq(l, "ref-uid")) { + uid_t uid; + + r = parse_uid(v, &uid); + if (r < 0) + log_unit_debug(u, "Failed to parse referenced UID %s, ignoring.", v); + else + unit_ref_uid_gid(u, uid, GID_INVALID); + + continue; + + } else if (streq(l, "ref-gid")) { + gid_t gid; + + r = parse_gid(v, &gid); + if (r < 0) + log_unit_debug(u, "Failed to parse referenced GID %s, ignoring.", v); + else + unit_ref_uid_gid(u, UID_INVALID, gid); + + } else if (streq(l, "ref")) { + + r = strv_extend(&u->deserialized_refs, v); + if (r < 0) + log_oom(); + + continue; + } else if (streq(l, "invocation-id")) { + sd_id128_t id; + + r = sd_id128_from_string(v, &id); + if (r < 0) + log_unit_debug(u, "Failed to parse invocation id %s, ignoring.", v); + else { + r = unit_set_invocation_id(u, id); + if (r < 0) + log_unit_warning_errno(u, r, "Failed to set invocation ID for unit: %m"); + } + + continue; + } + + if (unit_can_serialize(u)) { + if (rt) { + r = exec_runtime_deserialize_item(u, rt, l, v, fds); + if (r < 0) { + log_unit_warning(u, "Failed to deserialize runtime parameter '%s', ignoring.", l); + continue; + } + + /* Returns positive if key was handled by the call */ + if (r > 0) + continue; + } + + r = UNIT_VTABLE(u)->deserialize_item(u, l, v, fds); + if (r < 0) + log_unit_warning(u, "Failed to deserialize unit parameter '%s', ignoring.", l); + } + } + + /* Versions before 228 did not carry a state change timestamp. In this case, take the current time. This is + * useful, so that timeouts based on this timestamp don't trigger too early, and is in-line with the logic from + * before 228 where the base for timeouts was not persistent across reboots. */ + + if (!dual_timestamp_is_set(&u->state_change_timestamp)) + dual_timestamp_get(&u->state_change_timestamp); + + return 0; +} + +int unit_add_node_link(Unit *u, const char *what, bool wants, UnitDependency dep) { + Unit *device; + _cleanup_free_ char *e = NULL; + int r; + + assert(u); + + /* Adds in links to the device node that this unit is based on */ + if (isempty(what)) + return 0; + + if (!is_device_path(what)) + return 0; + + /* When device units aren't supported (such as in a + * container), don't create dependencies on them. */ + if (!unit_type_supported(UNIT_DEVICE)) + return 0; + + r = unit_name_from_path(what, ".device", &e); + if (r < 0) + return r; + + r = manager_load_unit(u->manager, e, NULL, NULL, &device); + if (r < 0) + return r; + + r = unit_add_two_dependencies(u, UNIT_AFTER, + MANAGER_IS_SYSTEM(u->manager) ? dep : UNIT_WANTS, + device, true); + if (r < 0) + return r; + + if (wants) { + r = unit_add_dependency(device, UNIT_WANTS, u, false); + if (r < 0) + return r; + } + + return 0; +} + +int unit_coldplug(Unit *u) { + int r = 0, q; + char **i; + + assert(u); + + /* Make sure we don't enter a loop, when coldplugging + * recursively. */ + if (u->coldplugged) + return 0; + + u->coldplugged = true; + + STRV_FOREACH(i, u->deserialized_refs) { + q = bus_unit_track_add_name(u, *i); + if (q < 0 && r >= 0) + r = q; + } + u->deserialized_refs = strv_free(u->deserialized_refs); + + if (UNIT_VTABLE(u)->coldplug) { + q = UNIT_VTABLE(u)->coldplug(u); + if (q < 0 && r >= 0) + r = q; + } + + if (u->job) { + q = job_coldplug(u->job); + if (q < 0 && r >= 0) + r = q; + } + + return r; +} + +static bool fragment_mtime_newer(const char *path, usec_t mtime, bool path_masked) { + struct stat st; + + if (!path) + return false; + + if (stat(path, &st) < 0) + /* What, cannot access this anymore? */ + return true; + + if (path_masked) + /* For masked files check if they are still so */ + return !null_or_empty(&st); + else + /* For non-empty files check the mtime */ + return timespec_load(&st.st_mtim) > mtime; + + return false; +} + +bool unit_need_daemon_reload(Unit *u) { + _cleanup_strv_free_ char **t = NULL; + char **path; + + assert(u); + + /* For unit files, we allow masking… */ + if (fragment_mtime_newer(u->fragment_path, u->fragment_mtime, + u->load_state == UNIT_MASKED)) + return true; + + /* Source paths should not be masked… */ + if (fragment_mtime_newer(u->source_path, u->source_mtime, false)) + return true; + + (void) unit_find_dropin_paths(u, &t); + if (!strv_equal(u->dropin_paths, t)) + return true; + + /* … any drop-ins that are masked are simply omitted from the list. */ + STRV_FOREACH(path, u->dropin_paths) + if (fragment_mtime_newer(*path, u->dropin_mtime, false)) + return true; + + return false; +} + +void unit_reset_failed(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->reset_failed) + UNIT_VTABLE(u)->reset_failed(u); + + RATELIMIT_RESET(u->start_limit); + u->start_limit_hit = false; +} + +Unit *unit_following(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->following) + return UNIT_VTABLE(u)->following(u); + + return NULL; +} + +bool unit_stop_pending(Unit *u) { + assert(u); + + /* This call does check the current state of the unit. It's + * hence useful to be called from state change calls of the + * unit itself, where the state isn't updated yet. This is + * different from unit_inactive_or_pending() which checks both + * the current state and for a queued job. */ + + return u->job && u->job->type == JOB_STOP; +} + +bool unit_inactive_or_pending(Unit *u) { + assert(u); + + /* Returns true if the unit is inactive or going down */ + + if (UNIT_IS_INACTIVE_OR_DEACTIVATING(unit_active_state(u))) + return true; + + if (unit_stop_pending(u)) + return true; + + return false; +} + +bool unit_active_or_pending(Unit *u) { + assert(u); + + /* Returns true if the unit is active or going up */ + + if (UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(u))) + return true; + + if (u->job && + (u->job->type == JOB_START || + u->job->type == JOB_RELOAD_OR_START || + u->job->type == JOB_RESTART)) + return true; + + return false; +} + +int unit_kill(Unit *u, KillWho w, int signo, sd_bus_error *error) { + assert(u); + assert(w >= 0 && w < _KILL_WHO_MAX); + assert(SIGNAL_VALID(signo)); + + if (!UNIT_VTABLE(u)->kill) + return -EOPNOTSUPP; + + return UNIT_VTABLE(u)->kill(u, w, signo, error); +} + +static Set *unit_pid_set(pid_t main_pid, pid_t control_pid) { + Set *pid_set; + int r; + + pid_set = set_new(NULL); + if (!pid_set) + return NULL; + + /* Exclude the main/control pids from being killed via the cgroup */ + if (main_pid > 0) { + r = set_put(pid_set, PID_TO_PTR(main_pid)); + if (r < 0) + goto fail; + } + + if (control_pid > 0) { + r = set_put(pid_set, PID_TO_PTR(control_pid)); + if (r < 0) + goto fail; + } + + return pid_set; + +fail: + set_free(pid_set); + return NULL; +} + +int unit_kill_common( + Unit *u, + KillWho who, + int signo, + pid_t main_pid, + pid_t control_pid, + sd_bus_error *error) { + + int r = 0; + bool killed = false; + + if (IN_SET(who, KILL_MAIN, KILL_MAIN_FAIL)) { + if (main_pid < 0) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no main processes", unit_type_to_string(u->type)); + else if (main_pid == 0) + return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No main process to kill"); + } + + if (IN_SET(who, KILL_CONTROL, KILL_CONTROL_FAIL)) { + if (control_pid < 0) + return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_PROCESS, "%s units have no control processes", unit_type_to_string(u->type)); + else if (control_pid == 0) + return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No control process to kill"); + } + + if (IN_SET(who, KILL_CONTROL, KILL_CONTROL_FAIL, KILL_ALL, KILL_ALL_FAIL)) + if (control_pid > 0) { + if (kill(control_pid, signo) < 0) + r = -errno; + else + killed = true; + } + + if (IN_SET(who, KILL_MAIN, KILL_MAIN_FAIL, KILL_ALL, KILL_ALL_FAIL)) + if (main_pid > 0) { + if (kill(main_pid, signo) < 0) + r = -errno; + else + killed = true; + } + + if (IN_SET(who, KILL_ALL, KILL_ALL_FAIL) && u->cgroup_path) { + _cleanup_set_free_ Set *pid_set = NULL; + int q; + + /* Exclude the main/control pids from being killed via the cgroup */ + pid_set = unit_pid_set(main_pid, control_pid); + if (!pid_set) + return -ENOMEM; + + q = cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, signo, 0, pid_set, NULL, NULL); + if (q < 0 && q != -EAGAIN && q != -ESRCH && q != -ENOENT) + r = q; + else + killed = true; + } + + if (r == 0 && !killed && IN_SET(who, KILL_ALL_FAIL, KILL_CONTROL_FAIL)) + return -ESRCH; + + return r; +} + +int unit_following_set(Unit *u, Set **s) { + assert(u); + assert(s); + + if (UNIT_VTABLE(u)->following_set) + return UNIT_VTABLE(u)->following_set(u, s); + + *s = NULL; + return 0; +} + +UnitFileState unit_get_unit_file_state(Unit *u) { + int r; + + assert(u); + + if (u->unit_file_state < 0 && u->fragment_path) { + r = unit_file_get_state( + u->manager->unit_file_scope, + NULL, + basename(u->fragment_path), + &u->unit_file_state); + if (r < 0) + u->unit_file_state = UNIT_FILE_BAD; + } + + return u->unit_file_state; +} + +int unit_get_unit_file_preset(Unit *u) { + assert(u); + + if (u->unit_file_preset < 0 && u->fragment_path) + u->unit_file_preset = unit_file_query_preset( + u->manager->unit_file_scope, + NULL, + basename(u->fragment_path)); + + return u->unit_file_preset; +} + +Unit* unit_ref_set(UnitRef *ref, Unit *u) { + assert(ref); + assert(u); + + if (ref->unit) + unit_ref_unset(ref); + + ref->unit = u; + LIST_PREPEND(refs, u->refs, ref); + return u; +} + +void unit_ref_unset(UnitRef *ref) { + assert(ref); + + if (!ref->unit) + return; + + /* We are about to drop a reference to the unit, make sure the garbage collection has a look at it as it might + * be unreferenced now. */ + unit_add_to_gc_queue(ref->unit); + + LIST_REMOVE(refs, ref->unit->refs, ref); + ref->unit = NULL; +} + +static int user_from_unit_name(Unit *u, char **ret) { + + static const uint8_t hash_key[] = { + 0x58, 0x1a, 0xaf, 0xe6, 0x28, 0x58, 0x4e, 0x96, + 0xb4, 0x4e, 0xf5, 0x3b, 0x8c, 0x92, 0x07, 0xec + }; + + _cleanup_free_ char *n = NULL; + int r; + + r = unit_name_to_prefix(u->id, &n); + if (r < 0) + return r; + + if (valid_user_group_name(n)) { + *ret = n; + n = NULL; + return 0; + } + + /* If we can't use the unit name as a user name, then let's hash it and use that */ + if (asprintf(ret, "_du%016" PRIx64, siphash24(n, strlen(n), hash_key)) < 0) + return -ENOMEM; + + return 0; +} + +int unit_patch_contexts(Unit *u) { + CGroupContext *cc; + ExecContext *ec; + unsigned i; + int r; + + assert(u); + + /* Patch in the manager defaults into the exec and cgroup + * contexts, _after_ the rest of the settings have been + * initialized */ + + ec = unit_get_exec_context(u); + if (ec) { + /* This only copies in the ones that need memory */ + for (i = 0; i < _RLIMIT_MAX; i++) + if (u->manager->rlimit[i] && !ec->rlimit[i]) { + ec->rlimit[i] = newdup(struct rlimit, u->manager->rlimit[i], 1); + if (!ec->rlimit[i]) + return -ENOMEM; + } + + if (MANAGER_IS_USER(u->manager) && + !ec->working_directory) { + + r = get_home_dir(&ec->working_directory); + if (r < 0) + return r; + + /* Allow user services to run, even if the + * home directory is missing */ + ec->working_directory_missing_ok = true; + } + + if (MANAGER_IS_USER(u->manager) && + (ec->syscall_whitelist || + !set_isempty(ec->syscall_filter) || + !set_isempty(ec->syscall_archs) || + ec->address_families_whitelist || + !set_isempty(ec->address_families))) + ec->no_new_privileges = true; + + if (ec->private_devices) + ec->capability_bounding_set &= ~((UINT64_C(1) << CAP_MKNOD) | (UINT64_C(1) << CAP_SYS_RAWIO)); + + if (ec->protect_kernel_modules) + ec->capability_bounding_set &= ~(UINT64_C(1) << CAP_SYS_MODULE); + + if (ec->dynamic_user) { + if (!ec->user) { + r = user_from_unit_name(u, &ec->user); + if (r < 0) + return r; + } + + if (!ec->group) { + ec->group = strdup(ec->user); + if (!ec->group) + return -ENOMEM; + } + + /* If the dynamic user option is on, let's make sure that the unit can't leave its UID/GID + * around in the file system or on IPC objects. Hence enforce a strict sandbox. */ + + ec->private_tmp = true; + ec->remove_ipc = true; + ec->protect_system = PROTECT_SYSTEM_STRICT; + if (ec->protect_home == PROTECT_HOME_NO) + ec->protect_home = PROTECT_HOME_READ_ONLY; + } + } + + cc = unit_get_cgroup_context(u); + if (cc) { + + if (ec && + ec->private_devices && + cc->device_policy == CGROUP_AUTO) + cc->device_policy = CGROUP_CLOSED; + } + + return 0; +} + +ExecContext *unit_get_exec_context(Unit *u) { + size_t offset; + assert(u); + + if (u->type < 0) + return NULL; + + offset = UNIT_VTABLE(u)->exec_context_offset; + if (offset <= 0) + return NULL; + + return (ExecContext*) ((uint8_t*) u + offset); +} + +KillContext *unit_get_kill_context(Unit *u) { + size_t offset; + assert(u); + + if (u->type < 0) + return NULL; + + offset = UNIT_VTABLE(u)->kill_context_offset; + if (offset <= 0) + return NULL; + + return (KillContext*) ((uint8_t*) u + offset); +} + +CGroupContext *unit_get_cgroup_context(Unit *u) { + size_t offset; + + if (u->type < 0) + return NULL; + + offset = UNIT_VTABLE(u)->cgroup_context_offset; + if (offset <= 0) + return NULL; + + return (CGroupContext*) ((uint8_t*) u + offset); +} + +ExecRuntime *unit_get_exec_runtime(Unit *u) { + size_t offset; + + if (u->type < 0) + return NULL; + + offset = UNIT_VTABLE(u)->exec_runtime_offset; + if (offset <= 0) + return NULL; + + return *(ExecRuntime**) ((uint8_t*) u + offset); +} + +static const char* unit_drop_in_dir(Unit *u, UnitSetPropertiesMode mode) { + assert(u); + + if (!IN_SET(mode, UNIT_RUNTIME, UNIT_PERSISTENT)) + return NULL; + + if (u->transient) /* Redirect drop-ins for transient units always into the transient directory. */ + return u->manager->lookup_paths.transient; + + if (mode == UNIT_RUNTIME) + return u->manager->lookup_paths.runtime_control; + + if (mode == UNIT_PERSISTENT) + return u->manager->lookup_paths.persistent_control; + + return NULL; +} + +int unit_write_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data) { + _cleanup_free_ char *p = NULL, *q = NULL; + const char *dir, *wrapped; + int r; + + assert(u); + + if (u->transient_file) { + /* When this is a transient unit file in creation, then let's not create a new drop-in but instead + * write to the transient unit file. */ + fputs(data, u->transient_file); + fputc('\n', u->transient_file); + return 0; + } + + if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) + return 0; + + dir = unit_drop_in_dir(u, mode); + if (!dir) + return -EINVAL; + + wrapped = strjoina("# This is a drop-in unit file extension, created via \"systemctl set-property\"\n" + "# or an equivalent operation. Do not edit.\n", + data, + "\n"); + + r = drop_in_file(dir, u->id, 50, name, &p, &q); + if (r < 0) + return r; + + (void) mkdir_p(p, 0755); + r = write_string_file_atomic_label(q, wrapped); + if (r < 0) + return r; + + r = strv_push(&u->dropin_paths, q); + if (r < 0) + return r; + q = NULL; + + strv_uniq(u->dropin_paths); + + u->dropin_mtime = now(CLOCK_REALTIME); + + return 0; +} + +int unit_write_drop_in_format(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *format, ...) { + _cleanup_free_ char *p = NULL; + va_list ap; + int r; + + assert(u); + assert(name); + assert(format); + + if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) + return 0; + + va_start(ap, format); + r = vasprintf(&p, format, ap); + va_end(ap); + + if (r < 0) + return -ENOMEM; + + return unit_write_drop_in(u, mode, name, p); +} + +int unit_write_drop_in_private(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data) { + const char *ndata; + + assert(u); + assert(name); + assert(data); + + if (!UNIT_VTABLE(u)->private_section) + return -EINVAL; + + if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) + return 0; + + ndata = strjoina("[", UNIT_VTABLE(u)->private_section, "]\n", data); + + return unit_write_drop_in(u, mode, name, ndata); +} + +int unit_write_drop_in_private_format(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *format, ...) { + _cleanup_free_ char *p = NULL; + va_list ap; + int r; + + assert(u); + assert(name); + assert(format); + + if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) + return 0; + + va_start(ap, format); + r = vasprintf(&p, format, ap); + va_end(ap); + + if (r < 0) + return -ENOMEM; + + return unit_write_drop_in_private(u, mode, name, p); +} + +int unit_make_transient(Unit *u) { + FILE *f; + char *path; + + assert(u); + + if (!UNIT_VTABLE(u)->can_transient) + return -EOPNOTSUPP; + + path = strjoin(u->manager->lookup_paths.transient, "/", u->id, NULL); + if (!path) + return -ENOMEM; + + /* Let's open the file we'll write the transient settings into. This file is kept open as long as we are + * creating the transient, and is closed in unit_load(), as soon as we start loading the file. */ + + RUN_WITH_UMASK(0022) { + f = fopen(path, "we"); + if (!f) { + free(path); + return -errno; + } + } + + if (u->transient_file) + fclose(u->transient_file); + u->transient_file = f; + + free(u->fragment_path); + u->fragment_path = path; + + u->source_path = mfree(u->source_path); + u->dropin_paths = strv_free(u->dropin_paths); + u->fragment_mtime = u->source_mtime = u->dropin_mtime = 0; + + u->load_state = UNIT_STUB; + u->load_error = 0; + u->transient = true; + + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); + + fputs("# This is a transient unit file, created programmatically via the systemd API. Do not edit.\n", + u->transient_file); + + return 0; +} + +static void log_kill(pid_t pid, int sig, void *userdata) { + _cleanup_free_ char *comm = NULL; + + (void) get_process_comm(pid, &comm); + + /* Don't log about processes marked with brackets, under the assumption that these are temporary processes + only, like for example systemd's own PAM stub process. */ + if (comm && comm[0] == '(') + return; + + log_unit_notice(userdata, + "Killing process " PID_FMT " (%s) with signal SIG%s.", + pid, + strna(comm), + signal_to_string(sig)); +} + +static int operation_to_signal(KillContext *c, KillOperation k) { + assert(c); + + switch (k) { + + case KILL_TERMINATE: + case KILL_TERMINATE_AND_LOG: + return c->kill_signal; + + case KILL_KILL: + return SIGKILL; + + case KILL_ABORT: + return SIGABRT; + + default: + assert_not_reached("KillOperation unknown"); + } +} + +int unit_kill_context( + Unit *u, + KillContext *c, + KillOperation k, + pid_t main_pid, + pid_t control_pid, + bool main_pid_alien) { + + bool wait_for_exit = false, send_sighup; + cg_kill_log_func_t log_func; + int sig, r; + + assert(u); + assert(c); + + /* Kill the processes belonging to this unit, in preparation for shutting the unit down. Returns > 0 if we + * killed something worth waiting for, 0 otherwise. */ + + if (c->kill_mode == KILL_NONE) + return 0; + + sig = operation_to_signal(c, k); + + send_sighup = + c->send_sighup && + IN_SET(k, KILL_TERMINATE, KILL_TERMINATE_AND_LOG) && + sig != SIGHUP; + + log_func = + k != KILL_TERMINATE || + IN_SET(sig, SIGKILL, SIGABRT) ? log_kill : NULL; + + if (main_pid > 0) { + if (log_func) + log_func(main_pid, sig, u); + + r = kill_and_sigcont(main_pid, sig); + if (r < 0 && r != -ESRCH) { + _cleanup_free_ char *comm = NULL; + (void) get_process_comm(main_pid, &comm); + + log_unit_warning_errno(u, r, "Failed to kill main process " PID_FMT " (%s), ignoring: %m", main_pid, strna(comm)); + } else { + if (!main_pid_alien) + wait_for_exit = true; + + if (r != -ESRCH && send_sighup) + (void) kill(main_pid, SIGHUP); + } + } + + if (control_pid > 0) { + if (log_func) + log_func(control_pid, sig, u); + + r = kill_and_sigcont(control_pid, sig); + if (r < 0 && r != -ESRCH) { + _cleanup_free_ char *comm = NULL; + (void) get_process_comm(control_pid, &comm); + + log_unit_warning_errno(u, r, "Failed to kill control process " PID_FMT " (%s), ignoring: %m", control_pid, strna(comm)); + } else { + wait_for_exit = true; + + if (r != -ESRCH && send_sighup) + (void) kill(control_pid, SIGHUP); + } + } + + if (u->cgroup_path && + (c->kill_mode == KILL_CONTROL_GROUP || (c->kill_mode == KILL_MIXED && k == KILL_KILL))) { + _cleanup_set_free_ Set *pid_set = NULL; + + /* Exclude the main/control pids from being killed via the cgroup */ + pid_set = unit_pid_set(main_pid, control_pid); + if (!pid_set) + return -ENOMEM; + + r = cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, + sig, + CGROUP_SIGCONT|CGROUP_IGNORE_SELF, + pid_set, + log_func, u); + if (r < 0) { + if (r != -EAGAIN && r != -ESRCH && r != -ENOENT) + log_unit_warning_errno(u, r, "Failed to kill control group %s, ignoring: %m", u->cgroup_path); + + } else if (r > 0) { + + /* FIXME: For now, on the legacy hierarchy, we + * will not wait for the cgroup members to die + * if we are running in a container or if this + * is a delegation unit, simply because cgroup + * notification is unreliable in these + * cases. It doesn't work at all in + * containers, and outside of containers it + * can be confused easily by left-over + * directories in the cgroup — which however + * should not exist in non-delegated units. On + * the unified hierarchy that's different, + * there we get proper events. Hence rely on + * them.*/ + + if (cg_unified(SYSTEMD_CGROUP_CONTROLLER) > 0 || + (detect_container() == 0 && !unit_cgroup_delegate(u))) + wait_for_exit = true; + + if (send_sighup) { + set_free(pid_set); + + pid_set = unit_pid_set(main_pid, control_pid); + if (!pid_set) + return -ENOMEM; + + cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, + SIGHUP, + CGROUP_IGNORE_SELF, + pid_set, + NULL, NULL); + } + } + } + + return wait_for_exit; +} + +int unit_require_mounts_for(Unit *u, const char *path) { + char prefix[strlen(path) + 1], *p; + int r; + + assert(u); + assert(path); + + /* Registers a unit for requiring a certain path and all its + * prefixes. We keep a simple array of these paths in the + * unit, since its usually short. However, we build a prefix + * table for all possible prefixes so that new appearing mount + * units can easily determine which units to make themselves a + * dependency of. */ + + if (!path_is_absolute(path)) + return -EINVAL; + + p = strdup(path); + if (!p) + return -ENOMEM; + + path_kill_slashes(p); + + if (!path_is_safe(p)) { + free(p); + return -EPERM; + } + + if (strv_contains(u->requires_mounts_for, p)) { + free(p); + return 0; + } + + r = strv_consume(&u->requires_mounts_for, p); + if (r < 0) + return r; + + PATH_FOREACH_PREFIX_MORE(prefix, p) { + Set *x; + + x = hashmap_get(u->manager->units_requiring_mounts_for, prefix); + if (!x) { + char *q; + + r = hashmap_ensure_allocated(&u->manager->units_requiring_mounts_for, &string_hash_ops); + if (r < 0) + return r; + + q = strdup(prefix); + if (!q) + return -ENOMEM; + + x = set_new(NULL); + if (!x) { + free(q); + return -ENOMEM; + } + + r = hashmap_put(u->manager->units_requiring_mounts_for, q, x); + if (r < 0) { + free(q); + set_free(x); + return r; + } + } + + r = set_put(x, u); + if (r < 0) + return r; + } + + return 0; +} + +int unit_setup_exec_runtime(Unit *u) { + ExecRuntime **rt; + size_t offset; + Iterator i; + Unit *other; + + offset = UNIT_VTABLE(u)->exec_runtime_offset; + assert(offset > 0); + + /* Check if there already is an ExecRuntime for this unit? */ + rt = (ExecRuntime**) ((uint8_t*) u + offset); + if (*rt) + return 0; + + /* Try to get it from somebody else */ + SET_FOREACH(other, u->dependencies[UNIT_JOINS_NAMESPACE_OF], i) { + + *rt = unit_get_exec_runtime(other); + if (*rt) { + exec_runtime_ref(*rt); + return 0; + } + } + + return exec_runtime_make(rt, unit_get_exec_context(u), u->id); +} + +int unit_setup_dynamic_creds(Unit *u) { + ExecContext *ec; + DynamicCreds *dcreds; + size_t offset; + + assert(u); + + offset = UNIT_VTABLE(u)->dynamic_creds_offset; + assert(offset > 0); + dcreds = (DynamicCreds*) ((uint8_t*) u + offset); + + ec = unit_get_exec_context(u); + assert(ec); + + if (!ec->dynamic_user) + return 0; + + return dynamic_creds_acquire(dcreds, u->manager, ec->user, ec->group); +} + +bool unit_type_supported(UnitType t) { + if (_unlikely_(t < 0)) + return false; + if (_unlikely_(t >= _UNIT_TYPE_MAX)) + return false; + + if (!unit_vtable[t]->supported) + return true; + + return unit_vtable[t]->supported(); +} + +void unit_warn_if_dir_nonempty(Unit *u, const char* where) { + int r; + + assert(u); + assert(where); + + r = dir_is_empty(where); + if (r > 0) + return; + if (r < 0) { + log_unit_warning_errno(u, r, "Failed to check directory %s: %m", where); + return; + } + + log_struct(LOG_NOTICE, + LOG_MESSAGE_ID(SD_MESSAGE_OVERMOUNTING), + LOG_UNIT_ID(u), + LOG_UNIT_MESSAGE(u, "Directory %s to mount over is not empty, mounting anyway.", where), + "WHERE=%s", where, + NULL); +} + +int unit_fail_if_symlink(Unit *u, const char* where) { + int r; + + assert(u); + assert(where); + + r = is_symlink(where); + if (r < 0) { + log_unit_debug_errno(u, r, "Failed to check symlink %s, ignoring: %m", where); + return 0; + } + if (r == 0) + return 0; + + log_struct(LOG_ERR, + LOG_MESSAGE_ID(SD_MESSAGE_OVERMOUNTING), + LOG_UNIT_ID(u), + LOG_UNIT_MESSAGE(u, "Mount on symlink %s not allowed.", where), + "WHERE=%s", where, + NULL); + + return -ELOOP; +} + +bool unit_is_pristine(Unit *u) { + assert(u); + + /* Check if the unit already exists or is already around, + * in a number of different ways. Note that to cater for unit + * types such as slice, we are generally fine with units that + * are marked UNIT_LOADED even though nothing was + * actually loaded, as those unit types don't require a file + * on disk to validly load. */ + + return !(!IN_SET(u->load_state, UNIT_NOT_FOUND, UNIT_LOADED) || + u->fragment_path || + u->source_path || + !strv_isempty(u->dropin_paths) || + u->job || + u->merged_into); +} + +pid_t unit_control_pid(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->control_pid) + return UNIT_VTABLE(u)->control_pid(u); + + return 0; +} + +pid_t unit_main_pid(Unit *u) { + assert(u); + + if (UNIT_VTABLE(u)->main_pid) + return UNIT_VTABLE(u)->main_pid(u); + + return 0; +} + +static void unit_unref_uid_internal( + Unit *u, + uid_t *ref_uid, + bool destroy_now, + void (*_manager_unref_uid)(Manager *m, uid_t uid, bool destroy_now)) { + + assert(u); + assert(ref_uid); + assert(_manager_unref_uid); + + /* Generic implementation of both unit_unref_uid() and unit_unref_gid(), under the assumption that uid_t and + * gid_t are actually the same time, with the same validity rules. + * + * Drops a reference to UID/GID from a unit. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (!uid_is_valid(*ref_uid)) + return; + + _manager_unref_uid(u->manager, *ref_uid, destroy_now); + *ref_uid = UID_INVALID; +} + +void unit_unref_uid(Unit *u, bool destroy_now) { + unit_unref_uid_internal(u, &u->ref_uid, destroy_now, manager_unref_uid); +} + +void unit_unref_gid(Unit *u, bool destroy_now) { + unit_unref_uid_internal(u, (uid_t*) &u->ref_gid, destroy_now, manager_unref_gid); +} + +static int unit_ref_uid_internal( + Unit *u, + uid_t *ref_uid, + uid_t uid, + bool clean_ipc, + int (*_manager_ref_uid)(Manager *m, uid_t uid, bool clean_ipc)) { + + int r; + + assert(u); + assert(ref_uid); + assert(uid_is_valid(uid)); + assert(_manager_ref_uid); + + /* Generic implementation of both unit_ref_uid() and unit_ref_guid(), under the assumption that uid_t and gid_t + * are actually the same type, and have the same validity rules. + * + * Adds a reference on a specific UID/GID to this unit. Each unit referencing the same UID/GID maintains a + * reference so that we can destroy the UID/GID's IPC resources as soon as this is requested and the counter + * drops to zero. */ + + assert_cc(sizeof(uid_t) == sizeof(gid_t)); + assert_cc(UID_INVALID == (uid_t) GID_INVALID); + + if (*ref_uid == uid) + return 0; + + if (uid_is_valid(*ref_uid)) /* Already set? */ + return -EBUSY; + + r = _manager_ref_uid(u->manager, uid, clean_ipc); + if (r < 0) + return r; + + *ref_uid = uid; + return 1; +} + +int unit_ref_uid(Unit *u, uid_t uid, bool clean_ipc) { + return unit_ref_uid_internal(u, &u->ref_uid, uid, clean_ipc, manager_ref_uid); +} + +int unit_ref_gid(Unit *u, gid_t gid, bool clean_ipc) { + return unit_ref_uid_internal(u, (uid_t*) &u->ref_gid, (uid_t) gid, clean_ipc, manager_ref_gid); +} + +static int unit_ref_uid_gid_internal(Unit *u, uid_t uid, gid_t gid, bool clean_ipc) { + int r = 0, q = 0; + + assert(u); + + /* Reference both a UID and a GID in one go. Either references both, or neither. */ + + if (uid_is_valid(uid)) { + r = unit_ref_uid(u, uid, clean_ipc); + if (r < 0) + return r; + } + + if (gid_is_valid(gid)) { + q = unit_ref_gid(u, gid, clean_ipc); + if (q < 0) { + if (r > 0) + unit_unref_uid(u, false); + + return q; + } + } + + return r > 0 || q > 0; +} + +int unit_ref_uid_gid(Unit *u, uid_t uid, gid_t gid) { + ExecContext *c; + int r; + + assert(u); + + c = unit_get_exec_context(u); + + r = unit_ref_uid_gid_internal(u, uid, gid, c ? c->remove_ipc : false); + if (r < 0) + return log_unit_warning_errno(u, r, "Couldn't add UID/GID reference to unit, proceeding without: %m"); + + return r; +} + +void unit_unref_uid_gid(Unit *u, bool destroy_now) { + assert(u); + + unit_unref_uid(u, destroy_now); + unit_unref_gid(u, destroy_now); +} + +void unit_notify_user_lookup(Unit *u, uid_t uid, gid_t gid) { + int r; + + assert(u); + + /* This is invoked whenever one of the forked off processes let's us know the UID/GID its user name/group names + * resolved to. We keep track of which UID/GID is currently assigned in order to be able to destroy its IPC + * objects when no service references the UID/GID anymore. */ + + r = unit_ref_uid_gid(u, uid, gid); + if (r > 0) + bus_unit_send_change_signal(u); +} + +int unit_set_invocation_id(Unit *u, sd_id128_t id) { + int r; + + assert(u); + + /* Set the invocation ID for this unit. If we cannot, this will not roll back, but reset the whole thing. */ + + if (sd_id128_equal(u->invocation_id, id)) + return 0; + + if (!sd_id128_is_null(u->invocation_id)) + (void) hashmap_remove_value(u->manager->units_by_invocation_id, &u->invocation_id, u); + + if (sd_id128_is_null(id)) { + r = 0; + goto reset; + } + + r = hashmap_ensure_allocated(&u->manager->units_by_invocation_id, &id128_hash_ops); + if (r < 0) + goto reset; + + u->invocation_id = id; + sd_id128_to_string(id, u->invocation_id_string); + + r = hashmap_put(u->manager->units_by_invocation_id, &u->invocation_id, u); + if (r < 0) + goto reset; + + return 0; + +reset: + u->invocation_id = SD_ID128_NULL; + u->invocation_id_string[0] = 0; + return r; +} + +int unit_acquire_invocation_id(Unit *u) { + sd_id128_t id; + int r; + + assert(u); + + r = sd_id128_randomize(&id); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to generate invocation ID for unit: %m"); + + r = unit_set_invocation_id(u, id); + if (r < 0) + return log_unit_error_errno(u, r, "Failed to set invocation ID for unit: %m"); + + return 0; +} |