diff options
Diffstat (limited to 'src/core')
133 files changed, 15121 insertions, 7988 deletions
diff --git a/src/core/.gitignore b/src/core/.gitignore index f293bbdc93..465b4fcc20 100644 --- a/src/core/.gitignore +++ b/src/core/.gitignore @@ -1,2 +1,3 @@ /macros.systemd +/triggers.systemd /systemd.pc diff --git a/src/core/audit-fd.c b/src/core/audit-fd.c index 5a18e263a8..76afe3fe15 100644 --- a/src/core/audit-fd.c +++ b/src/core/audit-fd.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,13 +19,15 @@ #include <errno.h> + #include "audit-fd.h" #ifdef HAVE_AUDIT -#include <stdbool.h> #include <libaudit.h> +#include <stdbool.h> +#include "fd-util.h" #include "log.h" #include "util.h" diff --git a/src/core/audit-fd.h b/src/core/audit-fd.h index 8b58289dc5..0eccb59210 100644 --- a/src/core/audit-fd.h +++ b/src/core/audit-fd.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/automount.c b/src/core/automount.c index d847dc1629..7d7a0a6e46 100644 --- a/src/core/automount.c +++ b/src/core/automount.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,29 +18,37 @@ ***/ #include <errno.h> -#include <limits.h> -#include <sys/mount.h> -#include <unistd.h> #include <fcntl.h> +#include <limits.h> +#include <linux/auto_dev-ioctl.h> +#include <linux/auto_fs4.h> #include <sys/epoll.h> +#include <sys/mount.h> #include <sys/stat.h> -#include <linux/auto_fs4.h> -#include <linux/auto_dev-ioctl.h> +#include <unistd.h> -#include "unit.h" +#include "alloc-util.h" +#include "async.h" #include "automount.h" -#include "mount.h" -#include "unit-name.h" -#include "special.h" +#include "bus-error.h" +#include "bus-util.h" +#include "dbus-automount.h" +#include "fd-util.h" +#include "formats-util.h" +#include "io-util.h" #include "label.h" #include "mkdir.h" +#include "mount-util.h" +#include "mount.h" +#include "parse-util.h" #include "path-util.h" -#include "dbus-automount.h" -#include "bus-util.h" -#include "bus-error.h" -#include "formats-util.h" #include "process-util.h" -#include "async.h" +#include "special.h" +#include "stdio-util.h" +#include "string-table.h" +#include "string-util.h" +#include "unit-name.h" +#include "unit.h" static const UnitActiveState state_translation_table[_AUTOMOUNT_STATE_MAX] = { [AUTOMOUNT_DEAD] = UNIT_INACTIVE, @@ -69,6 +75,9 @@ 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); @@ -81,34 +90,14 @@ static void automount_init(Unit *u) { UNIT(a)->ignore_on_isolate = true; } -static void repeat_unmount(const char *path) { - assert(path); - - for (;;) { - /* If there are multiple mounts on a mount point, this - * removes them all */ - - if (umount2(path, MNT_DETACH) >= 0) - continue; - - if (errno != EINVAL) - log_error_errno(errno, "Failed to unmount: %m"); - - break; - } -} - -static int automount_send_ready(Automount *a, Set *tokens, int status); - static void unmount_autofs(Automount *a) { + int r; + assert(a); if (a->pipe_fd < 0) return; - automount_send_ready(a, a->tokens, -EHOSTDOWN); - automount_send_ready(a, a->expire_tokens, -EHOSTDOWN); - a->pipe_event_source = sd_event_source_unref(a->pipe_event_source); a->pipe_fd = safe_close(a->pipe_fd); @@ -116,8 +105,14 @@ static void unmount_autofs(Automount *a) { * around */ if (a->where && (UNIT(a)->manager->exit_code != MANAGER_RELOAD && - UNIT(a)->manager->exit_code != MANAGER_REEXECUTE)) - repeat_unmount(a->where); + 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) { @@ -127,26 +122,22 @@ static void automount_done(Unit *u) { unmount_autofs(a); - free(a->where); - a->where = NULL; + a->where = mfree(a->where); - set_free(a->tokens); - a->tokens = NULL; - set_free(a->expire_tokens); - a->expire_tokens = NULL; + 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; - int r; assert(a); - r = path_get_parent(a->where, &parent); - if (r < 0) - return r; + parent = dirname_malloc(a->where); + if (!parent) + return -ENOMEM; return unit_require_mounts_for(UNIT(a), parent); } @@ -156,7 +147,10 @@ static int automount_add_default_dependencies(Automount *a) { assert(a); - if (UNIT(a)->manager->running_as != MANAGER_SYSTEM) + 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); @@ -227,11 +221,9 @@ static int automount_load(Unit *u) { if (r < 0) return r; - if (UNIT(a)->default_dependencies) { - r = automount_add_default_dependencies(a); - if (r < 0) - return r; - } + r = automount_add_default_dependencies(a); + if (r < 0) + return r; } return automount_verify(a); @@ -244,6 +236,9 @@ static void automount_set_state(Automount *a, AutomountState state) { 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); @@ -276,6 +271,11 @@ static int automount_coldplug(Unit *u) { 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); @@ -306,7 +306,7 @@ static void automount_dump(Unit *u, FILE *f, const char *prefix) { static void automount_enter_dead(Automount *a, AutomountResult f) { assert(a); - if (f != AUTOMOUNT_SUCCESS) + if (a->result == AUTOMOUNT_SUCCESS) a->result = f; automount_set_state(a, a->result != AUTOMOUNT_SUCCESS ? AUTOMOUNT_FAILED : AUTOMOUNT_DEAD); @@ -417,7 +417,7 @@ static int autofs_send_ready(int dev_autofs_fd, int ioctl_fd, uint32_t token, in init_autofs_dev_ioctl(¶m); param.ioctlfd = ioctl_fd; - if (status) { + if (status != 0) { param.fail.token = token; param.fail.status = status; } else @@ -444,7 +444,7 @@ static int automount_send_ready(Automount *a, Set *tokens, int status) { if (ioctl_fd < 0) return ioctl_fd; - if (status) + if (status != 0) log_unit_debug_errno(UNIT(a), status, "Sending failure: %m"); else log_unit_debug(UNIT(a), "Sending success."); @@ -471,51 +471,68 @@ static int automount_send_ready(Automount *a, Set *tokens, int status) { return r; } -int automount_update_mount(Automount *a, MountState old_state, MountState state) { +static void automount_trigger_notify(Unit *u, Unit *other) { + Automount *a = AUTOMOUNT(u); + int r; + assert(a); + assert(other); - switch (state) { - case MOUNT_MOUNTED: - case MOUNT_REMOUNTING: - automount_send_ready(a, a->tokens, 0); - break; - case MOUNT_DEAD: - case MOUNT_UNMOUNTING: - case MOUNT_MOUNTING_SIGTERM: - case MOUNT_MOUNTING_SIGKILL: - case MOUNT_REMOUNTING_SIGTERM: - case MOUNT_REMOUNTING_SIGKILL: - case MOUNT_UNMOUNTING_SIGTERM: - case MOUNT_UNMOUNTING_SIGKILL: - case MOUNT_FAILED: - if (old_state != state) - automount_send_ready(a, a->tokens, -ENODEV); - break; - default: - break; + /* 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; } - switch (state) { - case MOUNT_DEAD: - automount_send_ready(a, a->expire_tokens, 0); - break; - case MOUNT_MOUNTING: - case MOUNT_MOUNTING_DONE: - case MOUNT_MOUNTING_SIGTERM: - case MOUNT_MOUNTING_SIGKILL: - case MOUNT_REMOUNTING_SIGTERM: - case MOUNT_REMOUNTING_SIGKILL: - case MOUNT_UNMOUNTING_SIGTERM: - case MOUNT_UNMOUNTING_SIGKILL: - case MOUNT_FAILED: - if (old_state != state) - automount_send_ready(a, a->expire_tokens, -ENODEV); - break; - default: - break; + /* 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); } - return 0; + 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) { @@ -603,12 +620,16 @@ static void automount_enter_waiting(Automount *a) { return; fail: + log_unit_error_errno(UNIT(a), r, "Failed to initialize automounter: %m"); + safe_close_pair(p); - if (mounted) - repeat_unmount(a->where); + if (mounted) { + r = repeat_unmount(a->where, MNT_DETACH); + if (r < 0) + log_error_errno(r, "Failed to unmount, ignoring: %m"); + } - log_unit_error_errno(UNIT(a), r, "Failed to initialize automounter: %m"); automount_enter_dead(a, AUTOMOUNT_FAILURE_RESOURCES); } @@ -633,8 +654,6 @@ static void *expire_thread(void *p) { return NULL; } -static int automount_start_expire(Automount *a); - 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; @@ -672,7 +691,10 @@ static int automount_start_expire(Automount *a) { assert(a); - timeout = now(CLOCK_MONOTONIC) + MAX(a->timeout_idle_usec/10, USEC_PER_SEC); + 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); @@ -695,8 +717,17 @@ static int automount_start_expire(Automount *a) { 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_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; struct stat st; int r; @@ -722,18 +753,21 @@ static void automount_enter_runnning(Automount *a) { if (!S_ISDIR(st.st_mode) || st.st_dev != a->dev_id) log_unit_info(UNIT(a), "Automount point already active?"); else { - r = manager_add_job(UNIT(a)->manager, JOB_START, UNIT_TRIGGER(UNIT(a)), - JOB_REPLACE, true, &error, NULL); + 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; } } - 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); return; @@ -743,6 +777,8 @@ fail: 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); @@ -752,8 +788,21 @@ static int automount_start(Unit *u) { return -EEXIST; } - if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + 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); @@ -772,8 +821,9 @@ static int automount_stop(Unit *u) { static int automount_serialize(Unit *u, FILE *f, FDSet *fds) { Automount *a = AUTOMOUNT(u); - void *p; Iterator i; + void *p; + int r; assert(a); assert(f); @@ -788,15 +838,9 @@ static int automount_serialize(Unit *u, FILE *f, FDSet *fds) { SET_FOREACH(p, a->expire_tokens, i) unit_serialize_item_format(u, f, "expire-token", "%u", PTR_TO_UINT(p)); - if (a->pipe_fd >= 0) { - int copy; - - copy = fdset_put_dup(fds, a->pipe_fd); - if (copy < 0) - return copy; - - unit_serialize_item_format(u, f, "pipe-fd", "%i", copy); - } + r = unit_serialize_item_fd(u, f, fds, "pipe-fd", a->pipe_fd); + if (r < 0) + return r; return 0; } @@ -901,9 +945,11 @@ static bool automount_check_gc(Unit *u) { } static int automount_dispatch_io(sd_event_source *s, int fd, uint32_t events, void *userdata) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _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); @@ -950,7 +996,7 @@ static int automount_dispatch_io(sd_event_source *s, int fd, uint32_t events, vo case autofs_ptype_expire_direct: log_unit_debug(UNIT(a), "Got direct umount request on %s", a->where); - (void) sd_event_source_set_enabled(a->expire_event_source, SD_EVENT_OFF); + automount_stop_expire(a); r = set_ensure_allocated(&a->expire_tokens, NULL); if (r < 0) { @@ -963,7 +1009,26 @@ static int automount_dispatch_io(sd_event_source *s, int fd, uint32_t events, vo log_unit_error_errno(UNIT(a), r, "Failed to remember token: %m"); goto fail; } - r = manager_add_job(UNIT(a)->manager, JOB_STOP, UNIT_TRIGGER(UNIT(a)), JOB_REPLACE, true, &error, NULL); + + /* 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; @@ -1008,18 +1073,11 @@ static bool automount_supported(void) { return supported; } -static const char* const automount_state_table[_AUTOMOUNT_STATE_MAX] = { - [AUTOMOUNT_DEAD] = "dead", - [AUTOMOUNT_WAITING] = "waiting", - [AUTOMOUNT_RUNNING] = "running", - [AUTOMOUNT_FAILED] = "failed" -}; - -DEFINE_STRING_TABLE_LOOKUP(automount_state, AutomountState); - static const char* const automount_result_table[_AUTOMOUNT_RESULT_MAX] = { [AUTOMOUNT_SUCCESS] = "success", - [AUTOMOUNT_FAILURE_RESOURCES] = "resources" + [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); @@ -1032,9 +1090,6 @@ const UnitVTable automount_vtable = { "Automount\0" "Install\0", - .no_alias = true, - .no_instances = true, - .init = automount_init, .load = automount_load, .done = automount_done, @@ -1054,10 +1109,14 @@ const UnitVTable automount_vtable = { .check_gc = automount_check_gc, + .trigger_notify = automount_trigger_notify, + .reset_failed = automount_reset_failed, - .bus_interface = "org.freedesktop.systemd1.Automount", .bus_vtable = bus_automount_vtable, + .bus_set_property = bus_automount_set_property, + + .can_transient = true, .shutdown = automount_shutdown, .supported = automount_supported, @@ -1066,7 +1125,6 @@ const UnitVTable automount_vtable = { .finished_start_job = { [JOB_DONE] = "Set up automount %s.", [JOB_FAILED] = "Failed to set up automount %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", }, .finished_stop_job = { [JOB_DONE] = "Unset automount %s.", diff --git a/src/core/automount.h b/src/core/automount.h index 2a50fef68d..76a201178e 100644 --- a/src/core/automount.h +++ b/src/core/automount.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -25,18 +23,11 @@ typedef struct Automount Automount; #include "unit.h" -typedef enum AutomountState { - AUTOMOUNT_DEAD, - AUTOMOUNT_WAITING, - AUTOMOUNT_RUNNING, - AUTOMOUNT_FAILED, - _AUTOMOUNT_STATE_MAX, - _AUTOMOUNT_STATE_INVALID = -1 -} AutomountState; - 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; @@ -64,10 +55,5 @@ struct Automount { extern const UnitVTable automount_vtable; -int automount_update_mount(Automount *a, MountState old_state, MountState state); - -const char* automount_state_to_string(AutomountState i) _const_; -AutomountState automount_state_from_string(const char *s) _pure_; - const char* automount_result_to_string(AutomountResult i) _const_; AutomountResult automount_result_from_string(const char *s) _pure_; diff --git a/src/core/bus-endpoint.c b/src/core/bus-endpoint.c deleted file mode 100644 index 0c4b3e7c8b..0000000000 --- a/src/core/bus-endpoint.c +++ /dev/null @@ -1,134 +0,0 @@ -/*** - 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 "kdbus.h" -#include "bus-kernel.h" -#include "bus-policy.h" -#include "bus-endpoint.h" - -int bus_kernel_set_endpoint_policy(int fd, uid_t uid, BusEndpoint *ep) { - - struct kdbus_cmd *update; - struct kdbus_item *n; - BusEndpointPolicy *po; - Iterator i; - size_t size; - int r; - - size = ALIGN8(offsetof(struct kdbus_cmd, items)); - - HASHMAP_FOREACH(po, ep->policy_hash, i) { - size += ALIGN8(offsetof(struct kdbus_item, str) + strlen(po->name) + 1); - size += ALIGN8(offsetof(struct kdbus_item, policy_access) + sizeof(struct kdbus_policy_access)); - } - - update = alloca0_align(size, 8); - update->size = size; - - n = update->items; - - HASHMAP_FOREACH(po, ep->policy_hash, i) { - n->type = KDBUS_ITEM_NAME; - n->size = offsetof(struct kdbus_item, str) + strlen(po->name) + 1; - strcpy(n->str, po->name); - n = KDBUS_ITEM_NEXT(n); - - 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_USER; - n->policy_access.access = bus_kernel_translate_access(po->access); - n->policy_access.id = uid; - - n = KDBUS_ITEM_NEXT(n); - } - - r = ioctl(fd, KDBUS_CMD_ENDPOINT_UPDATE, update); - if (r < 0) - return -errno; - - return 0; -} - -int bus_endpoint_new(BusEndpoint **ep) { - assert(ep); - - *ep = new0(BusEndpoint, 1); - if (!*ep) - return -ENOMEM; - - return 0; -} - -int bus_endpoint_add_policy(BusEndpoint *ep, const char *name, BusPolicyAccess access) { - _cleanup_free_ BusEndpointPolicy *po = NULL; - _cleanup_free_ char *key = NULL; - int r; - - assert(ep); - assert(name); - assert(access > _BUS_POLICY_ACCESS_INVALID && access < _BUS_POLICY_ACCESS_MAX); - - /* check if we already have this name in the policy list. If we do, see if the new access level - * is higher than the exising one, and upgrade the entry in that case. Otherwise, do nothing. - */ - - if (ep->policy_hash) { - po = hashmap_get(ep->policy_hash, name); - if (po) { - if (po->access < access) - po->access = access; - - return 0; - } - } else { - ep->policy_hash = hashmap_new(&string_hash_ops); - if (!ep->policy_hash) - return -ENOMEM; - } - - po = new0(BusEndpointPolicy, 1); - if (!po) - return -ENOMEM; - - key = strdup(name); - if (!key) - return -ENOMEM; - - po->name = key; - po->access = access; - - r = hashmap_put(ep->policy_hash, key, po); - if (r < 0) - return r; - - po = NULL; - key = NULL; - return 0; -} - -void bus_endpoint_free(BusEndpoint *endpoint) { - if (!endpoint) - return; - - hashmap_free_free_free(endpoint->policy_hash); - free(endpoint); -} diff --git a/src/core/bus-endpoint.h b/src/core/bus-endpoint.h deleted file mode 100644 index 4a31f4c4be..0000000000 --- a/src/core/bus-endpoint.h +++ /dev/null @@ -1,44 +0,0 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - -#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/>. -***/ - -typedef struct BusEndpoint BusEndpoint; -typedef struct BusEndpointPolicy BusEndpointPolicy; - -#include "hashmap.h" -#include "bus-policy.h" - -struct BusEndpointPolicy { - char *name; - BusPolicyAccess access; -}; - -struct BusEndpoint { - Hashmap *policy_hash; -}; - -int bus_endpoint_new(BusEndpoint **ep); -void bus_endpoint_free(BusEndpoint *endpoint); - -int bus_endpoint_add_policy(BusEndpoint *ep, const char *name, BusPolicyAccess access); - -int bus_kernel_set_endpoint_policy(int fd, uid_t uid, BusEndpoint *ep); diff --git a/src/core/bus-policy.c b/src/core/bus-policy.c index a6a8fcd4d3..4907c268e8 100644 --- a/src/core/bus-policy.c +++ b/src/core/bus-policy.c @@ -19,10 +19,13 @@ #include <stdlib.h> -#include "kdbus.h" -#include "util.h" +#include "alloc-util.h" #include "bus-kernel.h" #include "bus-policy.h" +#include "kdbus.h" +#include "string-table.h" +#include "user-util.h" +#include "util.h" int bus_kernel_translate_access(BusPolicyAccess access) { assert(access >= 0); diff --git a/src/core/bus-policy.h b/src/core/bus-policy.h index 3b04f5457a..5b2c4d5953 100644 --- a/src/core/bus-policy.h +++ b/src/core/bus-policy.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -21,9 +19,9 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include "kdbus.h" #include "list.h" #include "macro.h" -#include "kdbus.h" typedef struct BusNamePolicy BusNamePolicy; diff --git a/src/core/busname.c b/src/core/busname.c index 2085721546..b96ec09e67 100644 --- a/src/core/busname.c +++ b/src/core/busname.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,17 +19,23 @@ #include <sys/mman.h> -#include "special.h" -#include "formats-util.h" -#include "signal-util.h" -#include "bus-kernel.h" +#include "alloc-util.h" #include "bus-internal.h" +#include "bus-kernel.h" +#include "bus-policy.h" #include "bus-util.h" +#include "busname.h" +#include "dbus-busname.h" +#include "fd-util.h" +#include "formats-util.h" #include "kdbus.h" -#include "bus-policy.h" +#include "parse-util.h" +#include "process-util.h" #include "service.h" -#include "dbus-busname.h" -#include "busname.h" +#include "signal-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" static const UnitActiveState state_translation_table[_BUSNAME_STATE_MAX] = { [BUSNAME_DEAD] = UNIT_INACTIVE, @@ -95,8 +99,7 @@ static void busname_done(Unit *u) { assert(n); - free(n->name); - n->name = NULL; + n->name = mfree(n->name); busname_free_policy(n); busname_unwatch_control_pid(n); @@ -107,29 +110,27 @@ static void busname_done(Unit *u) { n->timer_event_source = sd_event_source_unref(n->timer_event_source); } -static int busname_arm_timer(BusName *n) { +static int busname_arm_timer(BusName *n, usec_t usec) { int r; assert(n); - if (n->timeout_usec <= 0) { - n->timer_event_source = sd_event_source_unref(n->timer_event_source); - return 0; - } - if (n->timer_event_source) { - r = sd_event_source_set_time(n->timer_event_source, now(CLOCK_MONOTONIC) + n->timeout_usec); + 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, - now(CLOCK_MONOTONIC) + n->timeout_usec, 0, + usec, 0, busname_dispatch_timer, n); if (r < 0) return r; @@ -148,7 +149,7 @@ static int busname_add_default_default_dependencies(BusName *n) { if (r < 0) return r; - if (UNIT(n)->manager->running_as == MANAGER_SYSTEM) { + 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; @@ -317,7 +318,7 @@ static int busname_open_fd(BusName *n) { if (n->starter_fd >= 0) return 0; - mode = UNIT(n)->manager->running_as == MANAGER_SYSTEM ? "system" : "user"; + 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"); @@ -359,16 +360,15 @@ static int busname_coldplug(Unit *u) { if (n->deserialized_state == n->state) return 0; - if (IN_SET(n->deserialized_state, BUSNAME_MAKING, BUSNAME_SIGTERM, BUSNAME_SIGKILL)) { - - if (n->control_pid <= 0) - return -EBADMSG; + 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); + r = busname_arm_timer(n, usec_add(u->state_change_timestamp.monotonic, n->timeout_usec)); if (r < 0) return r; } @@ -393,7 +393,7 @@ static int busname_make_starter(BusName *n, pid_t *_pid) { pid_t pid; int r; - r = busname_arm_timer(n); + r = busname_arm_timer(n, usec_add(now(CLOCK_MONOTONIC), n->timeout_usec)); if (r < 0) goto fail; @@ -442,7 +442,7 @@ fail: static void busname_enter_dead(BusName *n, BusNameResult f) { assert(n); - if (f != BUSNAME_SUCCESS) + if (n->result == BUSNAME_SUCCESS) n->result = f; busname_set_state(n, n->result != BUSNAME_SUCCESS ? BUSNAME_FAILED : BUSNAME_DEAD); @@ -454,7 +454,7 @@ static void busname_enter_signal(BusName *n, BusNameState state, BusNameResult f assert(n); - if (f != BUSNAME_SUCCESS) + if (n->result == BUSNAME_SUCCESS) n->result = f; kill_context_init(&kill_context); @@ -471,7 +471,7 @@ static void busname_enter_signal(BusName *n, BusNameState state, BusNameResult f } if (r > 0) { - r = busname_arm_timer(n); + 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; @@ -553,7 +553,7 @@ fail: } static void busname_enter_running(BusName *n) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; bool pending = false; Unit *other; Iterator i; @@ -586,7 +586,13 @@ static void busname_enter_running(BusName *n) { } if (!pending) { - r = manager_add_job(UNIT(n)->manager, JOB_START, UNIT_DEREF(n->service), JOB_REPLACE, true, &error, NULL); + 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; } @@ -601,6 +607,7 @@ fail: static int busname_start(Unit *u) { BusName *n = BUSNAME(u); + int r; assert(n); @@ -626,6 +633,16 @@ static int busname_start(Unit *u) { 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); @@ -657,6 +674,7 @@ static int busname_stop(Unit *u) { static int busname_serialize(Unit *u, FILE *f, FDSet *fds) { BusName *n = BUSNAME(u); + int r; assert(n); assert(f); @@ -668,15 +686,9 @@ static int busname_serialize(Unit *u, FILE *f, FDSet *fds) { if (n->control_pid > 0) unit_serialize_item_format(u, f, "control-pid", PID_FMT, n->control_pid); - if (n->starter_fd >= 0) { - int copy; - - copy = fdset_put_dup(fds, n->starter_fd); - if (copy < 0) - return copy; - - unit_serialize_item_format(u, f, "starter-fd", "%i", copy); - } + r = unit_serialize_item_fd(u, f, fds, "starter-fd", n->starter_fd); + if (r < 0) + return r; return 0; } @@ -860,7 +872,7 @@ static void busname_sigchld_event(Unit *u, pid_t pid, int code, int status) { n->control_pid = 0; - if (is_clean_exit(code, status, NULL)) + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) f = BUSNAME_SUCCESS; else if (code == CLD_EXITED) f = BUSNAME_FAILURE_EXIT_CODE; @@ -874,7 +886,7 @@ static void busname_sigchld_event(Unit *u, pid_t pid, int code, int status) { 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 (f != BUSNAME_SUCCESS) + if (n->result == BUSNAME_SUCCESS) n->result = f; switch (n->state) { @@ -942,7 +954,6 @@ static void busname_reset_failed(Unit *u) { static void busname_trigger_notify(Unit *u, Unit *other) { BusName *n = BUSNAME(u); - Service *s; assert(n); assert(other); @@ -950,60 +961,57 @@ static void busname_trigger_notify(Unit *u, Unit *other) { if (!IN_SET(n->state, BUSNAME_RUNNING, BUSNAME_LISTENING)) return; - if (other->load_state != UNIT_LOADED || other->type != UNIT_SERVICE) + if (other->start_limit_hit) { + busname_enter_dead(n, BUSNAME_FAILURE_SERVICE_START_LIMIT_HIT); return; + } - s = SERVICE(other); + if (other->load_state != UNIT_LOADED || other->type != UNIT_SERVICE) + return; - if (s->state == SERVICE_FAILED && s->result == SERVICE_FAILURE_START_LIMIT) - busname_enter_dead(n, BUSNAME_FAILURE_SERVICE_FAILED_PERMANENT); - else if (IN_SET(s->state, - SERVICE_DEAD, SERVICE_FAILED, - SERVICE_STOP, SERVICE_STOP_SIGTERM, SERVICE_STOP_SIGKILL, - SERVICE_STOP_POST, SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL, - SERVICE_AUTO_RESTART)) + 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, uint64_t *timeout) { +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, timeout); + 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) { - static int supported = -1; - - if (supported < 0) - supported = is_kdbus_available(); - - return supported; + return false; } -static const char* const busname_state_table[_BUSNAME_STATE_MAX] = { - [BUSNAME_DEAD] = "dead", - [BUSNAME_MAKING] = "making", - [BUSNAME_REGISTERED] = "registered", - [BUSNAME_LISTENING] = "listening", - [BUSNAME_RUNNING] = "running", - [BUSNAME_SIGTERM] = "sigterm", - [BUSNAME_SIGKILL] = "sigkill", - [BUSNAME_FAILED] = "failed", -}; +static int busname_control_pid(Unit *u) { + BusName *n = BUSNAME(u); + + assert(n); -DEFINE_STRING_TABLE_LOOKUP(busname_state, BusNameState); + return n->control_pid; +} static const char* const busname_result_table[_BUSNAME_RESULT_MAX] = { [BUSNAME_SUCCESS] = "success", @@ -1012,7 +1020,8 @@ static const char* const busname_result_table[_BUSNAME_RESULT_MAX] = { [BUSNAME_FAILURE_EXIT_CODE] = "exit-code", [BUSNAME_FAILURE_SIGNAL] = "signal", [BUSNAME_FAILURE_CORE_DUMP] = "core-dump", - [BUSNAME_FAILURE_SERVICE_FAILED_PERMANENT] = "service-failed-permanent", + [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); @@ -1026,9 +1035,6 @@ const UnitVTable busname_vtable = { "Install\0", .private_section = "BusName", - .no_alias = true, - .no_instances = true, - .init = busname_init, .done = busname_done, .load = busname_load, @@ -1058,20 +1064,18 @@ const UnitVTable busname_vtable = { .supported = busname_supported, - .bus_interface = "org.freedesktop.systemd1.BusName", + .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.", - [JOB_DEPENDENCY] = "Dependency failed for %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/core/busname.h b/src/core/busname.h index 69528a2aef..a8562db458 100644 --- a/src/core/busname.h +++ b/src/core/busname.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -24,19 +22,8 @@ typedef struct BusName BusName; typedef struct BusNamePolicy BusNamePolicy; - -typedef enum BusNameState { - BUSNAME_DEAD, - BUSNAME_MAKING, - BUSNAME_REGISTERED, - BUSNAME_LISTENING, - BUSNAME_RUNNING, - BUSNAME_SIGTERM, - BUSNAME_SIGKILL, - BUSNAME_FAILED, - _BUSNAME_STATE_MAX, - _BUSNAME_STATE_INVALID = -1 -} BusNameState; +#include "unit.h" +#include "bus-policy.h" typedef enum BusNameResult { BUSNAME_SUCCESS, @@ -45,7 +32,8 @@ typedef enum BusNameResult { BUSNAME_FAILURE_EXIT_CODE, BUSNAME_FAILURE_SIGNAL, BUSNAME_FAILURE_CORE_DUMP, - BUSNAME_FAILURE_SERVICE_FAILED_PERMANENT, + BUSNAME_FAILURE_START_LIMIT_HIT, + BUSNAME_FAILURE_SERVICE_START_LIMIT_HIT, _BUSNAME_RESULT_MAX, _BUSNAME_RESULT_INVALID = -1 } BusNameResult; @@ -77,8 +65,5 @@ struct BusName { extern const UnitVTable busname_vtable; -const char* busname_state_to_string(BusNameState i) _const_; -BusNameState busname_state_from_string(const char *s) _pure_; - const char* busname_result_to_string(BusNameResult i) _const_; BusNameResult busname_result_from_string(const char *s) _pure_; diff --git a/src/core/cgroup.c b/src/core/cgroup.c index 6474e08bd2..20bdbc39d0 100644 --- a/src/core/cgroup.c +++ b/src/core/cgroup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -22,27 +20,63 @@ #include <fcntl.h> #include <fnmatch.h> -#include "process-util.h" -#include "path-util.h" -#include "special.h" +#include "alloc-util.h" #include "cgroup-util.h" #include "cgroup.h" +#include "fd-util.h" +#include "fileio.h" +#include "fs-util.h" +#include "parse-util.h" +#include "path-util.h" +#include "process-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "stdio-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_shares = (unsigned long) -1; - c->startup_cpu_shares = (unsigned long) -1; - c->memory_limit = (uint64_t) -1; - c->blockio_weight = (unsigned long) -1; - c->startup_blockio_weight = (unsigned long) -1; - + 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) { @@ -54,6 +88,24 @@ void cgroup_context_free_device_allow(CGroupContext *c, CGroupDeviceAllow *a) { 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); @@ -75,6 +127,12 @@ void cgroup_context_free_blockio_device_bandwidth(CGroupContext *c, CGroupBlockI 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); @@ -86,6 +144,8 @@ void cgroup_context_done(CGroupContext *c) { } void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { + CGroupIODeviceLimit *il; + CGroupIODeviceWeight *iw; CGroupBlockIODeviceBandwidth *b; CGroupBlockIODeviceWeight *w; CGroupDeviceAllow *a; @@ -98,25 +158,47 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { fprintf(f, "%sCPUAccounting=%s\n" + "%sIOAccounting=%s\n" "%sBlockIOAccounting=%s\n" "%sMemoryAccounting=%s\n" - "%sCPUShares=%lu\n" - "%sStartupCPUShares=%lu\n" + "%sTasksAccounting=%s\n" + "%sCPUWeight=%" PRIu64 "\n" + "%sStartupCPUWeight=%" PRIu64 "\n" + "%sCPUShares=%" PRIu64 "\n" + "%sStartupCPUShares=%" PRIu64 "\n" "%sCPUQuotaPerSecSec=%s\n" - "%sBlockIOWeight=%lu\n" - "%sStartupBlockIOWeight=%lu\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)); @@ -127,9 +209,30 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *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 %lu", + "%sBlockIODeviceWeight=%s %" PRIu64, prefix, w->path, w->weight); @@ -137,16 +240,22 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) { LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { char buf[FORMAT_BYTES_MAX]; - fprintf(f, - "%s%s=%s %s\n", - prefix, - b->read ? "BlockIOReadBandwidth" : "BlockIOWriteBandwidth", - b->path, - format_bytes(buf, sizeof(buf), b->bandwidth)); + 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_blkio_device(const char *p, dev_t *dev) { +static int lookup_block_device(const char *p, dev_t *dev) { struct stat st; int r; @@ -201,7 +310,7 @@ static int whitelist_device(const char *path, const char *node, const char *acc) r = cg_set_attribute("devices", path, "devices.allow", buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, + 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; @@ -272,7 +381,7 @@ static int whitelist_major(const char *path, const char *name, char type, const r = cg_set_attribute("devices", path, "devices.allow", buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, + log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to set devices.allow on %s: %m", path); } @@ -283,10 +392,261 @@ fail: return -errno; } -void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const char *path, ManagerState state) { +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); @@ -304,98 +664,233 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha * cgroup trees (assuming we are running in a container then), * and missing cgroups, i.e. EROFS and ENOENT. */ - if ((mask & CGROUP_CPU) && !is_root) { - char buf[MAX(DECIMAL_STR_MAX(unsigned long), DECIMAL_STR_MAX(usec_t)) + 1]; + 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); - sprintf(buf, "%lu\n", - IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_cpu_shares != (unsigned long) -1 ? c->startup_cpu_shares : - c->cpu_shares != (unsigned long) -1 ? c->cpu_shares : 1024); - r = cg_set_attribute("cpu", path, "cpu.shares", buf); - if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.shares on %s: %m", path); + if (cg_all_unified() > 0) { + uint64_t weight; - sprintf(buf, USEC_FMT "\n", CGROUP_CPU_QUOTA_PERIOD_USEC); - r = cg_set_attribute("cpu", path, "cpu.cfs_period_us", buf); - if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.cfs_period_us on %s: %m", path); + if (has_weight) + weight = cgroup_context_cpu_weight(c, state); + else if (has_shares) { + uint64_t shares = cgroup_context_cpu_shares(c, state); - if (c->cpu_quota_per_sec_usec != USEC_INFINITY) { - sprintf(buf, USEC_FMT "\n", c->cpu_quota_per_sec_usec * CGROUP_CPU_QUOTA_PERIOD_USEC / USEC_PER_SEC); - r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", buf); - } else - r = cg_set_attribute("cpu", path, "cpu.cfs_quota_us", "-1"); - if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set cpu.cfs_quota_us on %s: %m", path); + 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_shares) + shares = cgroup_context_cpu_shares(c, state); + else 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 + shares = CGROUP_CPU_SHARES_DEFAULT; + + cgroup_apply_legacy_cpu_config(u, shares, c->cpu_quota_per_sec_usec); + } } - if (mask & CGROUP_BLKIO) { - char buf[MAX3(DECIMAL_STR_MAX(unsigned long)+1, - DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(unsigned long)*1, - DECIMAL_STR_MAX(dev_t)*2+2+DECIMAL_STR_MAX(uint64_t)+1)]; - CGroupBlockIODeviceWeight *w; - CGroupBlockIODeviceBandwidth *b; + 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) { - sprintf(buf, "%lu\n", IN_SET(state, MANAGER_STARTING, MANAGER_INITIALIZING) && c->startup_blockio_weight != (unsigned long) -1 ? c->startup_blockio_weight : - c->blockio_weight != (unsigned long) -1 ? c->blockio_weight : 1000); - r = cg_set_attribute("blkio", path, "blkio.weight", buf); + 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_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set blkio.weight on %s: %m", path); - - /* FIXME: no way to reset this list */ - LIST_FOREACH(device_weights, w, c->blockio_device_weights) { - dev_t dev; - - r = lookup_blkio_device(w->path, &dev); - if (r < 0) - continue; - - sprintf(buf, "%u:%u %lu", major(dev), minor(dev), w->weight); - r = cg_set_attribute("blkio", path, "blkio.weight_device", buf); - if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set blkio.weight_device on %s: %m", path); + 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); + } } } - /* FIXME: no way to reset this list */ - LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { - const char *a; - dev_t dev; + /* Apply limits and free ones without config. */ + if (has_io) { + CGroupIODeviceLimit *l, *next; - r = lookup_blkio_device(b->path, &dev); - if (r < 0) - continue; + 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]; - a = b->read ? "blkio.throttle.read_bps_device" : "blkio.throttle.write_bps_device"; + limits[CGROUP_IO_RBPS_MAX] = b->rbps; + limits[CGROUP_IO_WBPS_MAX] = b->wbps; - sprintf(buf, "%u:%u %" PRIu64 "\n", major(dev), minor(dev), b->bandwidth); - r = cg_set_attribute("blkio", path, a, buf); + 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_blockio) + weight = cgroup_context_blkio_weight(c, state); + else 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 + weight = CGROUP_BLKIO_WEIGHT_DEFAULT; + + xsprintf(buf, "%" PRIu64 "\n", weight); + r = cg_set_attribute("blkio", path, "blkio.weight", buf); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set %s on %s: %m", a, path); + log_unit_full(u, IN_SET(r, -ENOENT, -EROFS, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to set blkio.weight: %m"); + + 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); + } else 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); + } + } + } + + /* Apply limits and free ones without config. */ + 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); + } + } else 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); + } } } - if ((mask & CGROUP_MEMORY) && !is_root) { - if (c->memory_limit != (uint64_t) -1) { + if ((mask & CGROUP_MASK_MEMORY) && !is_root) { + if (cg_all_unified() > 0) { + uint64_t max = c->memory_max; + uint64_t swap_max = c->memory_swap_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 = c->memory_limit; - sprintf(buf, "%" PRIu64 "\n", c->memory_limit); - r = cg_set_attribute("memory", path, "memory.limit_in_bytes", buf); - } else - r = cg_set_attribute("memory", path, "memory.limit_in_bytes", "-1"); + if (val == CGROUP_LIMIT_MAX) { + val = c->memory_max; - if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to set memory.limit_in_bytes on %s: %m", path); + if (val != CGROUP_LIMIT_MAX) + log_cgroup_compat(u, "Applying MemoryMax %" PRIi64 " as MemoryLimit", c->memory_max); + } + + 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_DEVICE) && !is_root) { + if ((mask & CGROUP_MASK_DEVICES) && !is_root) { CGroupDeviceAllow *a; /* Changing the devices list of a populated cgroup @@ -407,8 +902,8 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha else r = cg_set_attribute("devices", path, "devices.allow", "a"); if (r < 0) - log_full_errno(IN_SET(r, -ENOENT, -EROFS, -EINVAL) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to reset devices.list on %s: %m", path); + 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)) { @@ -419,7 +914,10 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha "/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 */ + "/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; @@ -454,66 +952,92 @@ void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const cha else if (startswith(a->path, "char-")) whitelist_major(path, a->path + 5, 'c', acc); else - log_debug("Ignoring device %s while writing cgroup attribute.", a->path); + 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"); + } } -CGroupControllerMask cgroup_context_get_mask(CGroupContext *c) { - CGroupControllerMask mask = 0; +CGroupMask cgroup_context_get_mask(CGroupContext *c) { + CGroupMask mask = 0; /* Figure out which controllers we need */ if (c->cpu_accounting || - c->cpu_shares != (unsigned long) -1 || - c->startup_cpu_shares != (unsigned long) -1 || + cgroup_context_has_cpu_weight(c) || + cgroup_context_has_cpu_shares(c) || c->cpu_quota_per_sec_usec != USEC_INFINITY) - mask |= CGROUP_CPUACCT | CGROUP_CPU; + mask |= CGROUP_MASK_CPUACCT | CGROUP_MASK_CPU; - if (c->blockio_accounting || - c->blockio_weight != (unsigned long) -1 || - c->startup_blockio_weight != (unsigned long) -1 || - c->blockio_device_weights || - c->blockio_device_bandwidths) - mask |= CGROUP_BLKIO; + 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 != (uint64_t) -1) - mask |= CGROUP_MEMORY; + 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_DEVICE; + mask |= CGROUP_MASK_DEVICES; + + if (c->tasks_accounting || + c->tasks_max != (uint64_t) -1) + mask |= CGROUP_MASK_PIDS; return mask; } -CGroupControllerMask unit_get_cgroup_mask(Unit *u) { +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 the process we fork into it is known to drop - * privileges anyway, and shouldn't get access to the - * controllers anyway. */ + * 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)) - return _CGROUP_CONTROLLER_MASK_ALL; + if (!e || + exec_context_maintains_privileges(e) || + cg_all_unified() > 0) + return _CGROUP_MASK_ALL; } return cgroup_context_get_mask(c); } -CGroupControllerMask unit_get_members_mask(Unit *u) { +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; @@ -532,7 +1056,7 @@ CGroupControllerMask unit_get_members_mask(Unit *u) { continue; u->cgroup_members_mask |= - unit_get_cgroup_mask(member) | + unit_get_own_mask(member) | unit_get_members_mask(member); } } @@ -541,19 +1065,52 @@ CGroupControllerMask unit_get_members_mask(Unit *u) { return u->cgroup_members_mask; } -CGroupControllerMask unit_get_siblings_mask(Unit *u) { +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_cgroup_mask(u) | unit_get_members_mask(u); + 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; } -CGroupControllerMask unit_get_target_mask(Unit *u) { - CGroupControllerMask mask; +CGroupMask unit_get_enable_mask(Unit *u) { + CGroupMask mask; - mask = unit_get_cgroup_mask(u) | unit_get_members_mask(u) | unit_get_siblings_mask(u); + /* 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; @@ -562,13 +1119,13 @@ CGroupControllerMask unit_get_target_mask(Unit *u) { /* 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) { - CGroupControllerMask m; + CGroupMask m; bool more; assert(u); /* Calculate subtree mask */ - m = unit_get_cgroup_mask(u) | unit_get_members_mask(u); + m = unit_get_subtree_mask(u); /* See if anything changed from the previous invocation. If * not, we're done. */ @@ -608,7 +1165,7 @@ void unit_update_cgroup_members_masks(Unit *u) { } } -static const char *migrate_callback(CGroupControllerMask mask, void *userdata) { +static const char *migrate_callback(CGroupMask mask, void *userdata) { Unit *u = userdata; assert(mask != 0); @@ -626,7 +1183,115 @@ static const char *migrate_callback(CGroupControllerMask mask, void *userdata) { return NULL; } -static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) { +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; @@ -643,25 +1308,30 @@ static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) { if (!path) return log_oom(); - r = hashmap_put(u->manager->cgroup_unit, path, u); - if (r < 0) { - log_error(r == -EEXIST ? "cgroup %s exists already: %s" : "hashmap_put failed for %s: %s", path, strerror(-r)); - return r; - } - if (r > 0) { - u->cgroup_path = path; - path = NULL; - } + 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, mask, u->cgroup_path); + r = cg_create_everywhere(u->manager->cgroup_supported, target_mask, u->cgroup_path); if (r < 0) - return log_error_errno(r, "Failed to create cgroup %s: %m", u->cgroup_path); + 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 = mask; + u->cgroup_realized_mask = target_mask; + u->cgroup_enabled_mask = enable_mask; if (u->type != UNIT_SLICE && !c->delegate) { @@ -670,7 +1340,7 @@ static int unit_create_cgroups(Unit *u, CGroupControllerMask mask) { * 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_warning_errno(r, "Failed to migrate cgroup from to %s: %m", u->cgroup_path); + log_unit_warning_errno(u, r, "Failed to migrate cgroup from to %s, ignoring: %m", u->cgroup_path); } return 0; @@ -691,10 +1361,30 @@ int unit_attach_pids_to_cgroup(Unit *u) { return 0; } -static bool unit_has_mask_realized(Unit *u, CGroupControllerMask mask) { +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 == mask; + 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. @@ -704,7 +1394,7 @@ static bool unit_has_mask_realized(Unit *u, CGroupControllerMask mask) { * * Returns 0 on success and < 0 on failure. */ static int unit_realize_cgroup_now(Unit *u, ManagerState state) { - CGroupControllerMask mask; + CGroupMask target_mask, enable_mask; int r; assert(u); @@ -714,9 +1404,10 @@ static int unit_realize_cgroup_now(Unit *u, ManagerState state) { u->in_cgroup_queue = false; } - mask = unit_get_target_mask(u); + target_mask = unit_get_target_mask(u); + enable_mask = unit_get_enable_mask(u); - if (unit_has_mask_realized(u, mask)) + if (unit_has_mask_realized(u, target_mask, enable_mask)) return 0; /* First, realize parents */ @@ -727,12 +1418,13 @@ static int unit_realize_cgroup_now(Unit *u, ManagerState state) { } /* And then do the real work */ - r = unit_create_cgroups(u, mask); + r = unit_create_cgroup(u, target_mask, enable_mask); if (r < 0) return r; /* Finally, apply the necessary attributes. */ - cgroup_context_apply(unit_get_cgroup_context(u), mask, u->cgroup_path, state); + cgroup_context_apply(u, target_mask, state); + cgroup_xattr_apply(u); return 0; } @@ -759,7 +1451,7 @@ unsigned manager_dispatch_cgroup_queue(Manager *m) { r = unit_realize_cgroup_now(i, state); if (r < 0) - log_warning_errno(r, "Failed to realize cgroups for queued unit %s: %m", i->id); + log_warning_errno(r, "Failed to realize cgroups for queued unit %s, ignoring: %m", i->id); n++; } @@ -795,7 +1487,7 @@ static void unit_queue_siblings(Unit *u) { /* 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))) + if (unit_has_mask_realized(m, unit_get_target_mask(m), unit_get_enable_mask(m))) continue; unit_add_to_cgroup_queue(m); @@ -806,12 +1498,9 @@ static void unit_queue_siblings(Unit *u) { } int unit_realize_cgroup(Unit *u) { - CGroupContext *c; - assert(u); - c = unit_get_cgroup_context(u); - if (!c) + if (!UNIT_HAS_CGROUP_CONTEXT(u)) return 0; /* So, here's the deal: when realizing the cgroups for this @@ -832,39 +1521,70 @@ int unit_realize_cgroup(Unit *u) { return unit_realize_cgroup_now(u, manager_state(u->manager)); } -void unit_destroy_cgroup_if_empty(Unit *u) { +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; - r = cg_trim_everywhere(u->manager->cgroup_supported, u->cgroup_path, !unit_has_name(u, SPECIAL_ROOT_SLICE)); + (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_debug_errno(r, "Failed to destroy cgroup %s: %m", u->cgroup_path); + log_unit_debug_errno(u, r, "Failed to destroy cgroup %s, ignoring: %m", u->cgroup_path); return; } - hashmap_remove(u->manager->cgroup_unit, u->cgroup_path); + if (is_root_slice) + return; + + unit_release_cgroup(u); - free(u->cgroup_path); - u->cgroup_path = NULL; u->cgroup_realized = false; u->cgroup_realized_mask = 0; + u->cgroup_enabled_mask = 0; } -pid_t unit_search_main_pid(Unit *u) { +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 0; + return -ENXIO; - if (cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, &f) < 0) - return 0; + 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) { @@ -874,93 +1594,285 @@ pid_t unit_search_main_pid(Unit *u) { continue; /* Ignore processes that aren't our kids */ - if (get_parent_of_pid(npid, &ppid) >= 0 && ppid != mypid) + if (get_process_ppid(npid, &ppid) >= 0 && ppid != mypid) continue; - if (pid != 0) { + 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. */ - pid = 0; - break; - } + + return -ENODATA; pid = npid; } - return pid; + *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; - int r; + CGroupController c; + int r, all_unified, systemd_unified; + char *e; assert(m); /* 1. Determine hierarchy */ - free(m->cgroup_root); - m->cgroup_root = NULL; - + 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"); - /* LEGACY: Already in /system.slice? If so, let's cut this - * off. This is to support live upgrades from older systemd - * versions where PID 1 was moved there. */ - if (m->running_as == MANAGER_SYSTEM) { - char *e; + /* 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"); - if (e) - *e = 0; + 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. */ - if (streq(m->cgroup_root, "/")) - m->cgroup_root[0] = 0; + 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"); - log_debug("Using cgroup controller " SYSTEMD_CGROUP_CONTROLLER ". File system hierarchy is at %s.", path); + 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 (m->running_as == MANAGER_SYSTEM) { + 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 + else if (r == 0) log_debug("Release agent already installed."); } - /* 4. Make sure we are in the root cgroup */ - r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, 0); + /* 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) - return log_error_errno(r, "Failed to create root cgroup hierarchy: %m"); + 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... */ - cg_set_attribute("memory", "/", "memory.use_hierarchy", "1"); + if (!all_unified) + (void) cg_set_attribute("memory", "/", "memory.use_hierarchy", "1"); } /* 7. Figure out which controllers are supported */ - m->cgroup_supported = cg_mask_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; } @@ -971,12 +1883,16 @@ void manager_shutdown_cgroup(Manager *m, bool delete) { /* We can't really delete the group, since we are in it. But * let's trim it. */ if (delete && m->cgroup_root) - cg_trim(SYSTEMD_CGROUP_CONTROLLER, m->cgroup_root, false); + (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); - free(m->cgroup_root); - m->cgroup_root = NULL; + m->cgroup_root = mfree(m->cgroup_root); } Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) { @@ -995,8 +1911,8 @@ Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) { char *e; e = strrchr(p, '/'); - if (e == p || !e) - return NULL; + if (!e || e == p) + return hashmap_get(m->cgroup_unit, SPECIAL_ROOT_SLICE); *e = 0; @@ -1006,13 +1922,13 @@ Unit* manager_get_unit_by_cgroup(Manager *m, const char *cgroup) { } } -Unit *manager_get_unit_by_pid(Manager *m, pid_t pid) { +Unit *manager_get_unit_by_pid_cgroup(Manager *m, pid_t pid) { _cleanup_free_ char *cgroup = NULL; int r; assert(m); - if (pid <= 1) + if (pid <= 0) return NULL; r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &cgroup); @@ -1022,26 +1938,41 @@ Unit *manager_get_unit_by_pid(Manager *m, pid_t pid) { 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; - int r; 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; - r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, true); - if (r <= 0) - return r; - - if (UNIT_VTABLE(u)->notify_cgroup_empty) - UNIT_VTABLE(u)->notify_cgroup_empty(u); - - unit_add_to_gc_queue(u); - return 0; + return unit_notify_cgroup_empty(u); } int unit_get_memory_current(Unit *u, uint64_t *ret) { @@ -1054,10 +1985,13 @@ int unit_get_memory_current(Unit *u, uint64_t *ret) { if (!u->cgroup_path) return -ENODATA; - if ((u->cgroup_realized_mask & CGROUP_MEMORY) == 0) + if ((u->cgroup_realized_mask & CGROUP_MASK_MEMORY) == 0) return -ENODATA; - r = cg_get_attribute("memory", u->cgroup_path, "memory.usage_in_bytes", &v); + 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) @@ -1066,9 +2000,8 @@ int unit_get_memory_current(Unit *u, uint64_t *ret) { return safe_atou64(v, ret); } -static int unit_get_cpu_usage_raw(Unit *u, nsec_t *ret) { +int unit_get_tasks_current(Unit *u, uint64_t *ret) { _cleanup_free_ char *v = NULL; - uint64_t ns; int r; assert(u); @@ -1077,18 +2010,60 @@ static int unit_get_cpu_usage_raw(Unit *u, nsec_t *ret) { if (!u->cgroup_path) return -ENODATA; - if ((u->cgroup_realized_mask & CGROUP_CPUACCT) == 0) + if ((u->cgroup_realized_mask & CGROUP_MASK_PIDS) == 0) return -ENODATA; - r = cg_get_attribute("cpuacct", u->cgroup_path, "cpuacct.usage", &v); + r = cg_get_attribute("pids", u->cgroup_path, "pids.current", &v); if (r == -ENOENT) return -ENODATA; if (r < 0) return r; - r = safe_atou64(v, &ns); - 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; @@ -1098,16 +2073,33 @@ 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->cpuacct_usage_base) - ns -= u->cpuacct_usage_base; + if (ns > u->cpu_usage_base) + ns -= u->cpu_usage_base; else ns = 0; - *ret = ns; + u->cpu_usage_last = ns; + if (ret) + *ret = ns; + return 0; } @@ -1117,16 +2109,60 @@ int unit_reset_cpu_usage(Unit *u) { assert(u); + u->cpu_usage_last = NSEC_INFINITY; + r = unit_get_cpu_usage_raw(u, &ns); if (r < 0) { - u->cpuacct_usage_base = 0; + u->cpu_usage_base = 0; return r; } - u->cpuacct_usage_base = ns; + 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", diff --git a/src/core/cgroup.h b/src/core/cgroup.h index 869ddae8c4..4cd168f63e 100644 --- a/src/core/cgroup.h +++ b/src/core/cgroup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -25,9 +23,12 @@ #include "list.h" #include "time-util.h" +#include "cgroup-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; @@ -55,30 +56,59 @@ struct CGroupDeviceAllow { 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; - unsigned long weight; + uint64_t weight; }; struct CGroupBlockIODeviceBandwidth { LIST_FIELDS(CGroupBlockIODeviceBandwidth, device_bandwidths); char *path; - uint64_t bandwidth; - bool read; + uint64_t rbps; + uint64_t wbps; }; struct CGroupContext { bool cpu_accounting; + bool io_accounting; bool blockio_accounting; bool memory_accounting; + bool tasks_accounting; - unsigned long cpu_shares; - unsigned long startup_cpu_shares; + /* For unified hierarchy */ + uint64_t cpu_weight; + uint64_t startup_cpu_weight; usec_t cpu_quota_per_sec_usec; - unsigned long blockio_weight; - unsigned long startup_blockio_weight; + 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); @@ -87,31 +117,44 @@ struct CGroupContext { CGroupDevicePolicy device_policy; LIST_HEAD(CGroupDeviceAllow, device_allow); + /* Common */ + uint64_t tasks_max; + bool delegate; }; #include "unit.h" -#include "cgroup-util.h" void cgroup_context_init(CGroupContext *c); void cgroup_context_done(CGroupContext *c); void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix); -void cgroup_context_apply(CGroupContext *c, CGroupControllerMask mask, const char *path, ManagerState state); -CGroupControllerMask cgroup_context_get_mask(CGroupContext *c); +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); -CGroupControllerMask unit_get_cgroup_mask(Unit *u); -CGroupControllerMask unit_get_siblings_mask(Unit *u); -CGroupControllerMask unit_get_members_mask(Unit *u); -CGroupControllerMask unit_get_target_mask(Unit *u); +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_destroy_cgroup_if_empty(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); @@ -120,15 +163,25 @@ 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); -pid_t unit_search_main_pid(Unit *u); - -int manager_notify_cgroup_empty(Manager *m, const char *group); +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/core/dbus-automount.c b/src/core/dbus-automount.c index 5162ce34cb..26212b3a95 100644 --- a/src/core/dbus-automount.c +++ b/src/core/dbus-automount.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,8 +18,9 @@ ***/ #include "automount.h" -#include "dbus-automount.h" #include "bus-util.h" +#include "dbus-automount.h" +#include "string-util.h" static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, automount_result, AutomountResult); @@ -33,3 +32,57 @@ const sd_bus_vtable bus_automount_vtable[] = { 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/core/dbus-automount.h b/src/core/dbus-automount.h index a2b124d756..f41adda2a6 100644 --- a/src/core/dbus-automount.h +++ b/src/core/dbus-automount.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,3 +21,5 @@ 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/core/dbus-busname.c b/src/core/dbus-busname.c index b1ceb05b1a..cf816ba15b 100644 --- a/src/core/dbus-busname.c +++ b/src/core/dbus-busname.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,10 +17,11 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" +#include "bus-util.h" #include "busname.h" #include "dbus-busname.h" -#include "bus-util.h" +#include "string-util.h" +#include "unit.h" static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, busname_result, BusNameResult); diff --git a/src/core/dbus-busname.h b/src/core/dbus-busname.h index ea55b6c8c9..8643d1a404 100644 --- a/src/core/dbus-busname.h +++ b/src/core/dbus-busname.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/dbus-cgroup.c b/src/core/dbus-cgroup.c index 4a9df06016..c4067a95bf 100644 --- a/src/core/dbus-cgroup.c +++ b/src/core/dbus-cgroup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,14 +17,83 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include "alloc-util.h" #include "bus-util.h" -#include "path-util.h" #include "cgroup-util.h" #include "cgroup.h" #include "dbus-cgroup.h" +#include "fd-util.h" +#include "fileio.h" +#include "path-util.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, @@ -79,11 +146,17 @@ static int property_get_blockio_device_bandwidths( 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 (streq(property, "BlockIOReadBandwidth") != b->read) + if (v == CGROUP_LIMIT_MAX) continue; - r = sd_bus_message_append(reply, "(st)", b->path, b->bandwidth); + r = sd_bus_message_append(reply, "(st)", b->path, v); if (r < 0) return r; } @@ -133,41 +206,39 @@ static int property_get_device_allow( return sd_bus_message_close_container(reply); } -static int property_get_ulong_as_u64( - sd_bus *bus, - const char *path, - const char *interface, - const char *property, - sd_bus_message *reply, - void *userdata, - sd_bus_error *error) { - - unsigned long *ul = userdata; - - assert(bus); - assert(reply); - assert(ul); - - return sd_bus_message_append(reply, "t", *ul == (unsigned long) -1 ? (uint64_t) -1 : (uint64_t) *ul); -} - 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("CPUShares", "t", property_get_ulong_as_u64, offsetof(CGroupContext, cpu_shares), 0), - SD_BUS_PROPERTY("StartupCPUShares", "t", property_get_ulong_as_u64, offsetof(CGroupContext, startup_cpu_shares), 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", property_get_ulong_as_u64, offsetof(CGroupContext, blockio_weight), 0), - SD_BUS_PROPERTY("StartupBlockIOWeight", "t", property_get_ulong_as_u64, offsetof(CGroupContext, startup_blockio_weight), 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 }; @@ -212,6 +283,7 @@ int bus_cgroup_set_property( UnitSetPropertiesMode mode, sd_bus_error *error) { + CGroupIOLimitType iol_type; int r; assert(u); @@ -228,56 +300,96 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->cpu_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_CPUACCT; + 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, "CPUShares")) { - uint64_t u64; - unsigned long ul; + } else if (streq(name, "CPUWeight")) { + uint64_t weight; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &weight); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul <= 0 || (uint64_t) ul != u64) - return sd_bus_error_set_errnof(error, EINVAL, "CPUShares value out of range"); + 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->cpu_shares = ul; - u->cgroup_realized_mask &= ~CGROUP_CPU; - unit_write_drop_in_private_format(u, mode, name, "CPUShares=%lu", ul); + 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, "StartupCPUShares")) { - uint64_t u64; - unsigned long ul; + } else if (streq(name, "CPUShares")) { + uint64_t shares; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &shares); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul <= 0 || (uint64_t) ul != u64) - return sd_bus_error_set_errnof(error, EINVAL, "StartupCPUShares value out of range"); + 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 = ul; - u->cgroup_realized_mask &= ~CGROUP_CPU; - unit_write_drop_in_private_format(u, mode, name, "StartupCPUShares=%lu", ul); + 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; @@ -294,12 +406,229 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->cpu_quota_per_sec_usec = u64; - u->cgroup_realized_mask &= ~CGROUP_CPU; + 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; @@ -309,61 +638,57 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->blockio_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_BLKIO; + 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 u64; - unsigned long ul; + uint64_t weight; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &weight); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul < 10 || ul > 1000) - return sd_bus_error_set_errnof(error, EINVAL, "BlockIOWeight value out of range"); - } + 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 = ul; - u->cgroup_realized_mask &= ~CGROUP_BLKIO; - unit_write_drop_in_private_format(u, mode, name, "BlockIOWeight=%lu", ul); + 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 u64; - unsigned long ul; + uint64_t weight; - r = sd_bus_message_read(message, "t", &u64); + r = sd_bus_message_read(message, "t", &weight); if (r < 0) return r; - if (u64 == (uint64_t) -1) - ul = (unsigned long) -1; - else { - ul = (unsigned long) u64; - if (ul < 10 || ul > 1000) - return sd_bus_error_set_errnof(error, EINVAL, "StartupBlockIOWeight value out of range"); - } + 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 = ul; - u->cgroup_realized_mask &= ~CGROUP_BLKIO; - unit_write_drop_in_private_format(u, mode, name, "StartupBlockIOWeight=%lu", ul); + 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 (streq(name, "BlockIOReadBandwidth") || streq(name, "BlockIOWriteBandwidth")) { + } else if (STR_IN_SET(name, "BlockIOReadBandwidth", "BlockIOWriteBandwidth")) { const char *path; bool read = true; unsigned n = 0; @@ -382,7 +707,7 @@ int bus_cgroup_set_property( CGroupBlockIODeviceBandwidth *a = NULL, *b; LIST_FOREACH(device_bandwidths, b, c->blockio_device_bandwidths) { - if (path_equal(path, b->path) && read == b->read) { + if (path_equal(path, b->path)) { a = b; break; } @@ -393,7 +718,8 @@ int bus_cgroup_set_property( if (!a) return -ENOMEM; - a->read = read; + a->rbps = CGROUP_LIMIT_MAX; + a->wbps = CGROUP_LIMIT_MAX; a->path = strdup(path); if (!a->path) { free(a); @@ -403,7 +729,10 @@ int bus_cgroup_set_property( LIST_PREPEND(device_bandwidths, c->blockio_device_bandwidths, a); } - a->bandwidth = u64; + if (read) + a->rbps = u64; + else + a->wbps = u64; } n++; @@ -416,36 +745,41 @@ int bus_cgroup_set_property( return r; if (mode != UNIT_CHECK) { - CGroupBlockIODeviceBandwidth *a, *next; + CGroupBlockIODeviceBandwidth *a; _cleanup_free_ char *buf = NULL; _cleanup_fclose_ FILE *f = NULL; size_t size = 0; if (n == 0) { - LIST_FOREACH_SAFE(device_bandwidths, a, next, c->blockio_device_bandwidths) - if (a->read == read) - cgroup_context_free_blockio_device_bandwidth(c, a); + LIST_FOREACH(device_bandwidths, a, c->blockio_device_bandwidths) { + if (read) + a->rbps = CGROUP_LIMIT_MAX; + else + a->wbps = CGROUP_LIMIT_MAX; + } } - u->cgroup_realized_mask &= ~CGROUP_BLKIO; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); f = open_memstream(&buf, &size); if (!f) return -ENOMEM; - if (read) { + if (read) { fputs("BlockIOReadBandwidth=\n", f); - LIST_FOREACH(device_bandwidths, a, c->blockio_device_bandwidths) - if (a->read) - fprintf(f, "BlockIOReadBandwidth=%s %" PRIu64 "\n", a->path, a->bandwidth); + 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->read) - fprintf(f, "BlockIOWriteBandwidth=%s %" PRIu64 "\n", a->path, a->bandwidth); + if (a->wbps != CGROUP_LIMIT_MAX) + fprintf(f, "BlockIOWriteBandwidth=%s %" PRIu64 "\n", a->path, a->wbps); } - fflush(f); + r = fflush_and_check(f); + if (r < 0) + return r; unit_write_drop_in_private(u, mode, name, buf); } @@ -453,17 +787,16 @@ int bus_cgroup_set_property( } else if (streq(name, "BlockIODeviceWeight")) { const char *path; - uint64_t u64; + 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, &u64)) > 0) { - unsigned long ul = u64; + while ((r = sd_bus_message_read(message, "(st)", &path, &weight)) > 0) { - if (ul < 10 || ul > 1000) + 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) { @@ -489,7 +822,7 @@ int bus_cgroup_set_property( LIST_PREPEND(device_weights,c->blockio_device_weights, a); } - a->weight = ul; + a->weight = weight; } n++; @@ -510,7 +843,7 @@ int bus_cgroup_set_property( cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); } - u->cgroup_realized_mask &= ~CGROUP_BLKIO; + unit_invalidate_cgroup(u, CGROUP_MASK_BLKIO); f = open_memstream(&buf, &size); if (!f) @@ -518,9 +851,11 @@ int bus_cgroup_set_property( fputs("BlockIODeviceWeight=\n", f); LIST_FOREACH(device_weights, a, c->blockio_device_weights) - fprintf(f, "BlockIODeviceWeight=%s %lu\n", a->path, a->weight); + fprintf(f, "BlockIODeviceWeight=%s %" PRIu64 "\n", a->path, a->weight); - fflush(f); + r = fflush_and_check(f); + if (r < 0) + return r; unit_write_drop_in_private(u, mode, name, buf); } @@ -535,23 +870,112 @@ int bus_cgroup_set_property( if (mode != UNIT_CHECK) { c->memory_accounting = b; - u->cgroup_realized_mask &= ~CGROUP_MEMORY; + 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; - u->cgroup_realized_mask &= ~CGROUP_MEMORY; - unit_write_drop_in_private_format(u, mode, name, "%s=%" PRIu64, name, 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; @@ -569,13 +993,9 @@ int bus_cgroup_set_property( return -EINVAL; if (mode != UNIT_CHECK) { - char *buf; - c->device_policy = p; - u->cgroup_realized_mask &= ~CGROUP_DEVICE; - - buf = strjoina("DevicePolicy=", policy); - unit_write_drop_in_private(u, mode, name, buf); + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); + unit_write_drop_in_private_format(u, mode, name, "DevicePolicy=%s", policy); } return 1; @@ -591,6 +1011,7 @@ int bus_cgroup_set_property( 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)) @@ -651,7 +1072,7 @@ int bus_cgroup_set_property( cgroup_context_free_device_allow(c, c->device_allow); } - u->cgroup_realized_mask &= ~CGROUP_DEVICE; + unit_invalidate_cgroup(u, CGROUP_MASK_DEVICES); f = open_memstream(&buf, &size); if (!f) @@ -661,12 +1082,69 @@ int bus_cgroup_set_property( 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" : ""); - fflush(f); + 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) { diff --git a/src/core/dbus-cgroup.h b/src/core/dbus-cgroup.h index c2a3910f3d..b2212fe44e 100644 --- a/src/core/dbus-cgroup.h +++ b/src/core/dbus-cgroup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "cgroup.h" extern const sd_bus_vtable bus_cgroup_vtable[]; diff --git a/src/core/dbus-device.c b/src/core/dbus-device.c index cb156fd37c..e1a12224d3 100644 --- a/src/core/dbus-device.c +++ b/src/core/dbus-device.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,9 +17,9 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "device.h" #include "dbus-device.h" +#include "device.h" +#include "unit.h" const sd_bus_vtable bus_device_vtable[] = { SD_BUS_VTABLE_START(0), diff --git a/src/core/dbus-device.h b/src/core/dbus-device.h index 10e945e402..eb1d8c3278 100644 --- a/src/core/dbus-device.h +++ b/src/core/dbus-device.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/dbus-execute.c b/src/core/dbus-execute.c index a9f7971cde..eec4500c8c 100644 --- a/src/core/dbus-execute.c +++ b/src/core/dbus-execute.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -25,27 +23,36 @@ #include <seccomp.h> #endif +#include "af-list.h" +#include "alloc-util.h" #include "bus-util.h" -#include "missing.h" -#include "ioprio.h" -#include "strv.h" -#include "fileio.h" -#include "execute.h" -#include "capability.h" +#include "capability-util.h" +#include "dbus-execute.h" #include "env-util.h" -#include "af-list.h" +#include "execute.h" +#include "fd-util.h" +#include "fileio.h" +#include "ioprio.h" +#include "missing.h" #include "namespace.h" +#include "parse-util.h" #include "path-util.h" -#include "dbus-execute.h" - +#include "process-util.h" +#include "rlimit-util.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" #endif +#include "strv.h" +#include "syslog-util.h" +#include "user-util.h" +#include "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); @@ -81,45 +88,6 @@ static int property_get_environment_files( return sd_bus_message_close_container(reply); } -static int property_get_rlimit( - sd_bus *bus, - const char *path, - const char *interface, - const char *property, - sd_bus_message *reply, - void *userdata, - sd_bus_error *error) { - - struct rlimit *rl; - uint64_t u; - rlim_t x; - - assert(bus); - assert(reply); - assert(userdata); - - rl = *(struct rlimit**) userdata; - if (rl) - x = rl->rlim_max; - else { - struct rlimit buf = {}; - int z; - - z = rlimit_from_string(property); - assert(z >= 0); - - getrlimit(z, &buf); - x = buf.rlim_max; - } - - /* rlim_t might have different sizes, let's map - * RLIMIT_INFINITY to (uint64_t) -1, so that it is the same on - * all archs */ - u = x == RLIM_INFINITY ? (uint64_t) -1 : (uint64_t) x; - - return sd_bus_message_append(reply, "t", u); -} - static int property_get_oom_score_adjust( sd_bus *bus, const char *path, @@ -144,7 +112,7 @@ static int property_get_oom_score_adjust( n = 0; if (read_one_line_file("/proc/self/oom_score_adj", &t) >= 0) - safe_atoi(t, &n); + safe_atoi32(t, &n); } return sd_bus_message_append(reply, "i", n); @@ -172,7 +140,7 @@ static int property_get_nice( else { errno = 0; n = getpriority(PRIO_PROCESS, 0); - if (errno != 0) + if (errno > 0) n = 0; } @@ -324,12 +292,10 @@ static int property_get_capability_bounding_set( assert(reply); assert(c); - /* We store this negated internally, to match the kernel, but - * we expose it normalized. */ - return sd_bus_message_append(reply, "t", ~c->capability_bounding_set_drop); + return sd_bus_message_append(reply, "t", c->capability_bounding_set); } -static int property_get_capabilities( +static int property_get_ambient_capabilities( sd_bus *bus, const char *path, const char *interface, @@ -339,22 +305,27 @@ static int property_get_capabilities( sd_bus_error *error) { ExecContext *c = userdata; - _cleanup_cap_free_charp_ char *t = NULL; - const char *s; assert(bus); assert(reply); assert(c); - if (c->capabilities) - s = t = cap_to_text(c->capabilities, NULL); - else - s = ""; + 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) { - if (!s) - return -ENOMEM; + assert(bus); + assert(reply); - return sd_bus_message_append(reply, "s", s); + return sd_bus_message_append(reply, "s", ""); } static int property_get_syscall_filter( @@ -593,28 +564,108 @@ static int property_get_address_families( 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)); +} + 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", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CPU]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitFSIZE", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_FSIZE]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitDATA", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_DATA]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitSTACK", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_STACK]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitCORE", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_CORE]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitRSS", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RSS]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitNOFILE", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NOFILE]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitAS", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_AS]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitNPROC", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NPROC]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitMEMLOCK", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MEMLOCK]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitLOCKS", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_LOCKS]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitSIGPENDING", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_SIGPENDING]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitMSGQUEUE", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_MSGQUEUE]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitNICE", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_NICE]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitRTPRIO", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTPRIO]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("LimitRTTIME", "t", property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("WorkingDirectory", "s", NULL, offsetof(ExecContext, working_directory), 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), @@ -635,24 +686,36 @@ const sd_bus_vtable bus_exec_vtable[] = { 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("Capabilities", "s", property_get_capabilities, 0, 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_dirs), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("ReadOnlyDirectories", "as", NULL, offsetof(ExecContext, read_only_dirs), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("InaccessibleDirectories", "as", NULL, offsetof(ExecContext, inaccessible_dirs), 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("PrivateNetwork", "b", bus_property_get_bool, offsetof(ExecContext, private_network), 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("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), @@ -665,6 +728,8 @@ const sd_bus_vtable bus_exec_vtable[] = { 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 }; @@ -766,7 +831,8 @@ int bus_exec_context_set_transient_property( UnitSetPropertiesMode mode, sd_bus_error *error) { - int r; + const char *soft = NULL; + int r, ri; assert(u); assert(c); @@ -780,23 +846,17 @@ int bus_exec_context_set_transient_property( if (r < 0) return r; - if (mode != UNIT_CHECK) { - - if (isempty(uu)) { - free(c->user); - c->user = NULL; - } else { - char *t; + 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); - t = strdup(uu); - if (!t) - return -ENOMEM; + if (mode != UNIT_CHECK) { - free(c->user); - c->user = t; - } + 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\n", uu); + unit_write_drop_in_private_format(u, mode, name, "User=%s", uu); } return 1; @@ -808,27 +868,70 @@ int bus_exec_context_set_transient_property( 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)) { - free(c->group); - c->group = NULL; - } else { - char *t; + if (isempty(gg)) + c->group = mfree(c->group); + else if (free_and_strdup(&c->group, gg) < 0) + return -ENOMEM; - t = strdup(gg); - if (!t) - return -ENOMEM; + unit_write_drop_in_private_format(u, mode, name, "Group=%s", gg); + } - free(c->group); - c->group = t; - } + 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"); - unit_write_drop_in_private_format(u, mode, name, "Group=%s\n", gg); + 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; @@ -836,37 +939,72 @@ int bus_exec_context_set_transient_property( if (r < 0) return r; - if (n < PRIO_MIN || n >= PRIO_MAX) + 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", n); + unit_write_drop_in_private_format(u, mode, name, "Nice=%i", n); } return 1; - } else if (streq(name, "TTYPath")) { - const char *tty; + } else if (STR_IN_SET(name, "TTYPath", "RootDirectory")) { + const char *s; - r = sd_bus_message_read(message, "s", &tty); + r = sd_bus_message_read(message, "s", &s); if (r < 0) return r; - if (!path_is_absolute(tty)) - return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "TTY device not absolute path"); + 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) { - char *t; + 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; - t = strdup(tty); - if (!t) - return -ENOMEM; + 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; - free(c->tty_path); - c->tty_path = t; + if (s[0] == '-') { + missing_ok = true; + s++; + } else + missing_ok = false; - unit_write_drop_in_private_format(u, mode, name, "TTYPath=%s\n", tty); + 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; @@ -886,7 +1024,7 @@ int bus_exec_context_set_transient_property( if (mode != UNIT_CHECK) { c->std_input = p; - unit_write_drop_in_private_format(u, mode, name, "StandardInput=%s\n", exec_input_to_string(p)); + unit_write_drop_in_private_format(u, mode, name, "StandardInput=%s", exec_input_to_string(p)); } return 1; @@ -907,7 +1045,7 @@ int bus_exec_context_set_transient_property( if (mode != UNIT_CHECK) { c->std_output = p; - unit_write_drop_in_private_format(u, mode, name, "StandardOutput=%s\n", exec_output_to_string(p)); + unit_write_drop_in_private_format(u, mode, name, "StandardOutput=%s", exec_output_to_string(p)); } return 1; @@ -927,7 +1065,112 @@ int bus_exec_context_set_transient_property( if (mode != UNIT_CHECK) { c->std_error = p; - unit_write_drop_in_private_format(u, mode, name, "StandardError=%s\n", exec_output_to_string(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, + "IgnoreSIGPIPE", "TTYVHangup", "TTYReset", + "PrivateTmp", "PrivateDevices", "PrivateNetwork", "PrivateUsers", + "NoNewPrivileges", "SyslogLevelPrefix", "MemoryDenyWriteExecute", + "RestrictRealtime", "DynamicUser", "RemoveIPC", "ProtectKernelTunables", + "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, "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; @@ -947,23 +1190,338 @@ int bus_exec_context_set_transient_property( _cleanup_free_ char *joined = NULL; char **e; - e = strv_env_merge(2, c->environment, l); - if (!e) - return -ENOMEM; + 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; - strv_free(c->environment); - c->environment = e; + 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; - joined = strv_join_quoted(c->environment); - if (!joined) + 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, "Environment=%s\n", joined); + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, strempty(s)); } return 1; - } else if (rlimit_from_string(name) >= 0) { + } + + 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; @@ -981,22 +1539,36 @@ int bus_exec_context_set_transient_property( } if (mode != UNIT_CHECK) { - int z; - - z = rlimit_from_string(name); - - if (!c->rlimit[z]) { - c->rlimit[z] = new(struct rlimit, 1); - if (!c->rlimit[z]) + _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; } - c->rlimit[z]->rlim_cur = c->rlimit[z]->rlim_max = x; - - if (x == RLIM_INFINITY) - unit_write_drop_in_private_format(u, mode, name, "%s=infinity\n", name); - else - unit_write_drop_in_private_format(u, mode, name, "%s=%" PRIu64 "\n", name, rl); + unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, f); } return 1; diff --git a/src/core/dbus-execute.h b/src/core/dbus-execute.h index e4c2d5ddf6..d0aa8e1dd5 100644 --- a/src/core/dbus-execute.h +++ b/src/core/dbus-execute.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "execute.h" #define BUS_EXEC_STATUS_VTABLE(prefix, offset, flags) \ diff --git a/src/core/dbus-job.c b/src/core/dbus-job.c index cd6b909426..ccf7453d47 100644 --- a/src/core/dbus-job.c +++ b/src/core/dbus-job.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,12 +17,15 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "log.h" #include "sd-bus.h" -#include "selinux-access.h" -#include "job.h" + +#include "alloc-util.h" #include "dbus-job.h" #include "dbus.h" +#include "job.h" +#include "log.h" +#include "selinux-access.h" +#include "string-util.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); @@ -74,7 +75,7 @@ int bus_job_method_cancel(sd_bus_message *message, void *userdata, sd_bus_error 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); + job_finish_and_invalidate(j, JOB_CANCELED, true, false); return sd_bus_reply_method_return(message, NULL); } @@ -90,7 +91,7 @@ const sd_bus_vtable bus_job_vtable[] = { }; static int send_new_signal(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; _cleanup_free_ char *p = NULL; Job *j = userdata; int r; @@ -150,7 +151,7 @@ void bus_job_send_change_signal(Job *j) { } static int send_removed_signal(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; _cleanup_free_ char *p = NULL; Job *j = userdata; int r; diff --git a/src/core/dbus-job.h b/src/core/dbus-job.h index fb5f1b513e..024d06719e 100644 --- a/src/core/dbus-job.h +++ b/src/core/dbus-job.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "job.h" extern const sd_bus_vtable bus_job_vtable[]; diff --git a/src/core/dbus-kill.c b/src/core/dbus-kill.c index 3b8116281c..8c65be65fa 100644 --- a/src/core/dbus-kill.c +++ b/src/core/dbus-kill.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,11 +17,10 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "signal-util.h" #include "bus-util.h" - -#include "kill.h" #include "dbus-kill.h" +#include "kill.h" +#include "signal-util.h" static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_kill_mode, kill_mode, KillMode); @@ -61,12 +58,12 @@ int bus_kill_context_set_transient_property( k = kill_mode_from_string(m); if (k < 0) - return -EINVAL; + 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\n", kill_mode_to_string(k)); + unit_write_drop_in_private_format(u, mode, name, "KillMode=%s", kill_mode_to_string(k)); } return 1; @@ -78,13 +75,13 @@ int bus_kill_context_set_transient_property( if (r < 0) return r; - if (sig <= 0 || sig >= _NSIG) + 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\n", signal_to_string(sig)); + unit_write_drop_in_private_format(u, mode, name, "KillSignal=%s", signal_to_string(sig)); } return 1; @@ -99,7 +96,7 @@ int bus_kill_context_set_transient_property( if (mode != UNIT_CHECK) { c->send_sighup = b; - unit_write_drop_in_private_format(u, mode, name, "SendSIGHUP=%s\n", yes_no(b)); + unit_write_drop_in_private_format(u, mode, name, "SendSIGHUP=%s", yes_no(b)); } return 1; @@ -114,7 +111,7 @@ int bus_kill_context_set_transient_property( if (mode != UNIT_CHECK) { c->send_sigkill = b; - unit_write_drop_in_private_format(u, mode, name, "SendSIGKILL=%s\n", yes_no(b)); + unit_write_drop_in_private_format(u, mode, name, "SendSIGKILL=%s", yes_no(b)); } return 1; diff --git a/src/core/dbus-kill.h b/src/core/dbus-kill.h index 7c15f3a90b..b9b18811e3 100644 --- a/src/core/dbus-kill.h +++ b/src/core/dbus-kill.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,8 +20,9 @@ ***/ #include "sd-bus.h" -#include "unit.h" + #include "kill.h" +#include "unit.h" extern const sd_bus_vtable bus_kill_vtable[]; diff --git a/src/core/dbus-manager.c b/src/core/dbus-manager.c index d8b39bdf5f..12eb55cb7f 100644 --- a/src/core/dbus-manager.c +++ b/src/core/dbus-manager.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,27 +18,34 @@ ***/ #include <errno.h> +#include <sys/prctl.h> #include <unistd.h> -#include "log.h" -#include "strv.h" +#include "alloc-util.h" +#include "architecture.h" #include "build.h" -#include "install.h" -#include "selinux-access.h" -#include "watchdog.h" +#include "bus-common-errors.h" #include "clock-util.h" -#include "path-util.h" -#include "virt.h" -#include "architecture.h" -#include "env-util.h" -#include "dbus.h" +#include "dbus-execute.h" #include "dbus-job.h" #include "dbus-manager.h" #include "dbus-unit.h" -#include "dbus-snapshot.h" -#include "dbus-execute.h" -#include "bus-common-errors.h" +#include "dbus.h" +#include "env-util.h" +#include "fd-util.h" +#include "fileio.h" #include "formats-util.h" +#include "install.h" +#include "log.h" +#include "path-util.h" +#include "selinux-access.h" +#include "stat-util.h" +#include "string-util.h" +#include "strv.h" +#include "syslog-util.h" +#include "user-util.h" +#include "virt.h" +#include "watchdog.h" static int property_get_version( sd_bus *bus, @@ -81,14 +86,21 @@ static int property_get_virtualization( void *userdata, sd_bus_error *error) { - const char *id = NULL; + int v; assert(bus); assert(reply); - detect_virtualization(&id); + v = detect_virtualization(); - return sd_bus_message_append(reply, "s", id); + /* 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( @@ -115,8 +127,7 @@ static int property_get_tainted( void *userdata, sd_bus_error *error) { - char buf[sizeof("split-usr:mtab-not-symlink:cgroups-missing:local-hwclock:")] = "", *e = buf; - _cleanup_free_ char *p = NULL; + char buf[sizeof("split-usr:cgroups-missing:local-hwclock:")] = "", *e = buf; Manager *m = userdata; assert(bus); @@ -126,13 +137,10 @@ static int property_get_tainted( if (m->taint_usr) e = stpcpy(e, "split-usr:"); - if (readlink_malloc("/etc/mtab", &p) < 0) - e = stpcpy(e, "mtab-not-symlink:"); - if (access("/proc/cgroups", F_OK) < 0) e = stpcpy(e, "cgroups-missing:"); - if (clock_is_localtime() > 0) + if (clock_is_localtime(NULL) > 0) e = stpcpy(e, "local-hwclock:"); /* remove the last ':' */ @@ -220,7 +228,10 @@ static int property_set_log_level( if (r < 0) return r; - return log_set_max_level_from_string(t); + 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( @@ -343,6 +354,21 @@ static int property_set_runtime_watchdog( 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; @@ -360,7 +386,7 @@ static int method_get_unit(sd_bus_message *message, void *userdata, sd_bus_error return r; if (isempty(name)) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _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); @@ -412,7 +438,7 @@ static int method_get_unit_by_pid(sd_bus_message *message, void *userdata, sd_bu return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid PID " PID_FMT, pid); if (pid == 0) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _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) @@ -438,6 +464,64 @@ static int method_get_unit_by_pid(sd_bus_message *message, void *userdata, sd_bu 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; @@ -455,7 +539,7 @@ static int method_load_unit(sd_bus_message *message, void *userdata, sd_bus_erro return r; if (isempty(name)) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _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); @@ -606,13 +690,163 @@ static int method_set_unit_properties(sd_bus_message *message, void *userdata, s 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); + 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, @@ -620,6 +854,7 @@ static int transient_unit_from_message( Unit **unit, sd_bus_error *error) { + UnitType t; Unit *u; int r; @@ -627,12 +862,18 @@ static int transient_unit_from_message( 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 (u->load_state != UNIT_NOT_FOUND || - set_size(u->dependencies[UNIT_REFERENCED_BY]) > 0) + 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 @@ -646,6 +887,17 @@ static int transient_unit_from_message( 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; @@ -656,8 +908,6 @@ static int transient_aux_units_from_message( sd_bus_message *message, sd_bus_error *error) { - Unit *u; - char *name = NULL; int r; assert(m); @@ -668,20 +918,17 @@ static int transient_aux_units_from_message( 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 && r != -EEXIST) + if (r < 0) return r; - if (r != -EEXIST) { - r = unit_load(u); - if (r < 0) - return r; - } - r = sd_bus_message_exit_container(message); if (r < 0) return r; @@ -700,7 +947,6 @@ static int method_start_transient_unit(sd_bus_message *message, void *userdata, const char *name, *smode; Manager *m = userdata; JobMode mode; - UnitType t; Unit *u; int r; @@ -715,13 +961,6 @@ static int method_start_transient_unit(sd_bus_message *message, void *userdata, if (r < 0) return r; - t = unit_name_to_type(name); - if (t < 0) - return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid unit 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)); - 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); @@ -740,13 +979,6 @@ static int method_start_transient_unit(sd_bus_message *message, void *userdata, if (r < 0) return r; - /* And load this stub fully */ - r = unit_load(u); - if (r < 0) - return r; - - manager_dispatch_load_queue(m); - /* Finally, start it */ return bus_unit_queue_job(message, u, JOB_START, mode, false, error); } @@ -846,8 +1078,8 @@ static int method_reset_failed(sd_bus_message *message, void *userdata, sd_bus_e 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) { - _cleanup_bus_message_unref_ sd_bus_message *reply = 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; @@ -872,42 +1104,20 @@ static int list_units_filtered(sd_bus_message *message, void *userdata, sd_bus_e return r; HASHMAP_FOREACH_KEY(u, k, m->units, i) { - _cleanup_free_ char *unit_path = NULL, *job_path = NULL; - Unit *following; - if (k != u->id) continue; - following = unit_following(u); - 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; - 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; - } + if (!strv_isempty(patterns) && + !strv_fnmatch_or_empty(patterns, u->id, FNM_NOESCAPE)) + continue; - r = 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 : "/"); + r = reply_unit_info(reply, u); if (r < 0) return r; } @@ -920,7 +1130,7 @@ static int list_units_filtered(sd_bus_message *message, void *userdata, sd_bus_e } static int method_list_units(sd_bus_message *message, void *userdata, sd_bus_error *error) { - return list_units_filtered(message, userdata, error, NULL); + 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) { @@ -931,11 +1141,27 @@ static int method_list_units_filtered(sd_bus_message *message, void *userdata, s if (r < 0) return r; - return list_units_filtered(message, userdata, error, states); + 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_bus_message_unref_ sd_bus_message *reply = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; Manager *m = userdata; Iterator i; Job *j; @@ -1069,74 +1295,15 @@ static int method_dump(sd_bus_message *message, void *userdata, sd_bus_error *er manager_dump_units(m, f, NULL); manager_dump_jobs(m, f, NULL); - fflush(f); - - if (ferror(f)) - return -ENOMEM; - - return sd_bus_reply_method_return(message, "s", dump); -} - -static int method_create_snapshot(sd_bus_message *message, void *userdata, sd_bus_error *error) { - _cleanup_free_ char *path = NULL; - Manager *m = userdata; - const char *name; - int cleanup; - Snapshot *s = NULL; - 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, "sb", &name, &cleanup); + r = fflush_and_check(f); if (r < 0) return r; - if (isempty(name)) - name = NULL; - - 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 = snapshot_create(m, name, cleanup, error, &s); - if (r < 0) - return r; - - path = unit_dbus_path(UNIT(s)); - if (!path) - return -ENOMEM; - - return sd_bus_reply_method_return(message, "o", path); + return sd_bus_reply_method_return(message, "s", dump); } -static int method_remove_snapshot(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 does not exist.", name); - - if (u->type != UNIT_SNAPSHOT) - return sd_bus_error_setf(error, BUS_ERROR_NO_SUCH_UNIT, "Unit %s is not a snapshot", name); - - return bus_snapshot_method_remove(message, u, error); +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) { @@ -1206,8 +1373,10 @@ static int method_exit(sd_bus_message *message, void *userdata, sd_bus_error *er if (r < 0) return r; - if (m->running_as == MANAGER_SYSTEM) - return sd_bus_error_setf(error, SD_BUS_ERROR_NOT_SUPPORTED, "Exit is only supported for user service managers."); + /* 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; @@ -1225,7 +1394,7 @@ static int method_reboot(sd_bus_message *message, void *userdata, sd_bus_error * if (r < 0) return r; - if (m->running_as != MANAGER_SYSTEM) + 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; @@ -1244,7 +1413,7 @@ static int method_poweroff(sd_bus_message *message, void *userdata, sd_bus_error if (r < 0) return r; - if (m->running_as != MANAGER_SYSTEM) + 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; @@ -1263,7 +1432,7 @@ static int method_halt(sd_bus_message *message, void *userdata, sd_bus_error *er if (r < 0) return r; - if (m->running_as != MANAGER_SYSTEM) + 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; @@ -1282,7 +1451,7 @@ static int method_kexec(sd_bus_message *message, void *userdata, sd_bus_error *e if (r < 0) return r; - if (m->running_as != MANAGER_SYSTEM) + 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; @@ -1303,7 +1472,7 @@ static int method_switch_root(sd_bus_message *message, void *userdata, sd_bus_er if (r < 0) return r; - if (m->running_as != MANAGER_SYSTEM) + 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); @@ -1455,8 +1624,87 @@ static int method_unset_and_set_environment(sd_bus_message *message, void *userd return sd_bus_reply_method_return(message, NULL); } -static int method_list_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { - _cleanup_bus_message_unref_ sd_bus_message *reply = 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; @@ -1480,7 +1728,7 @@ static int method_list_unit_files(sd_bus_message *message, void *userdata, sd_bu if (!h) return -ENOMEM; - r = unit_file_get_list(m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, NULL, h); + r = unit_file_get_list(m->unit_file_scope, NULL, h, states, patterns); if (r < 0) goto fail; @@ -1508,11 +1756,30 @@ fail: 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; - UnitFileScope scope; int r; assert(message); @@ -1528,11 +1795,9 @@ static int method_get_unit_file_state(sd_bus_message *message, void *userdata, s if (r < 0) return r; - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - state = unit_file_get_state(scope, NULL, name); - if (state < 0) - return state; + 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)); } @@ -1540,7 +1805,6 @@ static int method_get_unit_file_state(sd_bus_message *message, void *userdata, s 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; - UnitFileScope scope; int r; assert(message); @@ -1552,9 +1816,7 @@ static int method_get_default_target(sd_bus_message *message, void *userdata, sd if (r < 0) return r; - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = unit_file_get_default(scope, NULL, &default_target); + r = unit_file_get_default(m->unit_file_scope, NULL, &default_target); if (r < 0) return r; @@ -1562,7 +1824,7 @@ static int method_get_default_target(sd_bus_message *message, void *userdata, sd } static int send_unit_files_changed(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *message = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; int r; assert(bus); @@ -1581,11 +1843,11 @@ static int reply_unit_file_changes_and_free( UnitFileChange *changes, unsigned n_changes) { - _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *reply = NULL; unsigned i; int r; - if (n_changes > 0) { + 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"); @@ -1605,20 +1867,25 @@ static int reply_unit_file_changes_and_free( if (r < 0) goto fail; - for (i = 0; i < n_changes; i++) { - r = sd_bus_message_append( - reply, "(sss)", - unit_file_change_type_to_string(changes[i].type), - changes[i].path, - changes[i].source); - 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: @@ -1626,10 +1893,61 @@ fail: 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, - const char *verb, int (*call)(UnitFileScope scope, bool runtime, const char *root_dir, char *files[], bool force, UnitFileChange **changes, unsigned *n_changes), bool carries_install_info, sd_bus_error *error) { @@ -1637,7 +1955,6 @@ static int method_enable_unit_files_generic( _cleanup_strv_free_ char **l = NULL; UnitFileChange *changes = NULL; unsigned n_changes = 0; - UnitFileScope scope; int runtime, force, r; assert(message); @@ -1651,35 +1968,29 @@ static int method_enable_unit_files_generic( if (r < 0) return r; - r = mac_selinux_unit_access_check_strv(l, message, m, verb, error); - 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 */ - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = call(scope, runtime, NULL, l, force, &changes, &n_changes); + r = call(m->unit_file_scope, runtime, NULL, l, force, &changes, &n_changes); if (r < 0) - return r; + 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, "enable", unit_file_enable, true, 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, "enable", unit_file_reenable, true, 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, "enable", unit_file_link, false, error); + return method_enable_unit_files_generic(message, userdata, unit_file_link, false, error); } static int unit_file_preset_without_mode(UnitFileScope scope, bool runtime, const char *root_dir, char **files, bool force, UnitFileChange **changes, unsigned *n_changes) { @@ -1687,11 +1998,11 @@ static int unit_file_preset_without_mode(UnitFileScope scope, bool runtime, cons } static int method_preset_unit_files(sd_bus_message *message, void *userdata, sd_bus_error *error) { - return method_enable_unit_files_generic(message, userdata, "enable", unit_file_preset_without_mode, true, 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, "disable", unit_file_mask, false, 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) { @@ -1701,7 +2012,6 @@ static int method_preset_unit_files_with_mode(sd_bus_message *message, void *use unsigned n_changes = 0; Manager *m = userdata; UnitFilePresetMode mm; - UnitFileScope scope; int runtime, force, r; const char *mode; @@ -1724,36 +2034,28 @@ static int method_preset_unit_files_with_mode(sd_bus_message *message, void *use return -EINVAL; } - r = mac_selinux_unit_access_check_strv(l, message, m, "enable", error); - 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 */ - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = unit_file_preset(scope, runtime, NULL, l, mm, force, &changes, &n_changes); + r = unit_file_preset(m->unit_file_scope, runtime, NULL, l, mm, force, &changes, &n_changes); if (r < 0) - return r; + 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, const - char *verb, + Manager *m, int (*call)(UnitFileScope scope, bool runtime, 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; - UnitFileScope scope; int r, runtime; assert(message); @@ -1767,38 +2069,58 @@ static int method_disable_unit_files_generic( if (r < 0) return r; - r = mac_selinux_unit_access_check_strv(l, message, m, verb, error); - if (r < 0) - return r; - - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - 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(scope, runtime, NULL, l, &changes, &n_changes); + r = call(m->unit_file_scope, runtime, NULL, l, &changes, &n_changes); if (r < 0) - return r; + 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, "disable", unit_file_disable, 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, "enable", unit_file_unmask, 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; - UnitFileScope scope; const char *name; int force, r; @@ -1819,11 +2141,9 @@ static int method_set_default_target(sd_bus_message *message, void *userdata, sd if (r == 0) return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = unit_file_set_default(scope, NULL, name, force, &changes, &n_changes); + r = unit_file_set_default(m->unit_file_scope, NULL, name, force, &changes, &n_changes); if (r < 0) - return r; + return install_error(error, r, changes, n_changes); return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); } @@ -1833,7 +2153,6 @@ static int method_preset_all_unit_files(sd_bus_message *message, void *userdata, unsigned n_changes = 0; Manager *m = userdata; UnitFilePresetMode mm; - UnitFileScope scope; const char *mode; int force, runtime, r; @@ -1862,11 +2181,9 @@ static int method_preset_all_unit_files(sd_bus_message *message, void *userdata, if (r == 0) return 1; /* No authorization for now, but the async polkit stuff will call us again when it has it */ - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = unit_file_preset_all(scope, runtime, NULL, mm, force, &changes, &n_changes); + r = unit_file_preset_all(m->unit_file_scope, runtime, NULL, mm, force, &changes, &n_changes); if (r < 0) - return r; + return install_error(error, r, changes, n_changes); return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); } @@ -1876,10 +2193,8 @@ static int method_add_dependency_unit_files(sd_bus_message *message, void *userd Manager *m = userdata; UnitFileChange *changes = NULL; unsigned n_changes = 0; - UnitFileScope scope; int runtime, force, r; - char *target; - char *type; + char *target, *type; UnitDependency dep; assert(message); @@ -1903,15 +2218,9 @@ static int method_add_dependency_unit_files(sd_bus_message *message, void *userd if (dep < 0) return -EINVAL; - r = mac_selinux_unit_access_check_strv(l, message, m, "enable", error); - if (r < 0) - return r; - - scope = m->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER; - - r = unit_file_add_dependency(scope, runtime, NULL, l, target, dep, force, &changes, &n_changes); + r = unit_file_add_dependency(m->unit_file_scope, runtime, NULL, l, target, dep, force, &changes, &n_changes); if (r < 0) - return r; + return install_error(error, r, changes, n_changes); return reply_unit_file_changes_and_free(m, message, -1, changes, n_changes); } @@ -1947,16 +2256,63 @@ const sd_bus_vtable bus_manager_vtable[] = { 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.unit_path), 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), @@ -1969,19 +2325,24 @@ const sd_bus_vtable bus_manager_vtable[] = { 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_create_snapshot, SD_BUS_VTABLE_UNPRIVILEGED), - SD_BUS_METHOD("RemoveSnapshot", "s", NULL, method_remove_snapshot, 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), @@ -1994,6 +2355,7 @@ const sd_bus_vtable bus_manager_vtable[] = { 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), @@ -2003,10 +2365,14 @@ const sd_bus_vtable bus_manager_vtable[] = { 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("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), @@ -2020,7 +2386,7 @@ const sd_bus_vtable bus_manager_vtable[] = { }; static int send_finished(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *message = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; usec_t *times = userdata; int r; @@ -2068,7 +2434,7 @@ void bus_manager_send_finished( } static int send_reloading(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *message = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *message = NULL; int r; assert(bus); diff --git a/src/core/dbus-manager.h b/src/core/dbus-manager.h index 5bdf6e17ab..36a2e9481b 100644 --- a/src/core/dbus-manager.h +++ b/src/core/dbus-manager.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/dbus-mount.c b/src/core/dbus-mount.c index 24813c6d20..76a7a7ce97 100644 --- a/src/core/dbus-mount.c +++ b/src/core/dbus-mount.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,13 +17,14 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "mount.h" +#include "bus-util.h" +#include "dbus-cgroup.h" #include "dbus-execute.h" #include "dbus-kill.h" -#include "dbus-cgroup.h" #include "dbus-mount.h" -#include "bus-util.h" +#include "mount.h" +#include "string-util.h" +#include "unit.h" static int property_get_what( sd_bus *bus, @@ -117,7 +116,11 @@ const sd_bus_vtable bus_mount_vtable[] = { 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), @@ -158,6 +161,9 @@ static int bus_mount_set_transient_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; } diff --git a/src/core/dbus-mount.h b/src/core/dbus-mount.h index f7004d252f..ec16166d36 100644 --- a/src/core/dbus-mount.h +++ b/src/core/dbus-mount.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_mount_vtable[]; diff --git a/src/core/dbus-path.c b/src/core/dbus-path.c index 683561999b..1e153e503f 100644 --- a/src/core/dbus-path.c +++ b/src/core/dbus-path.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,10 +17,11 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "path.h" -#include "dbus-path.h" #include "bus-util.h" +#include "dbus-path.h" +#include "path.h" +#include "string-util.h" +#include "unit.h" static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, path_result, PathResult); diff --git a/src/core/dbus-path.h b/src/core/dbus-path.h index 389b0d7f9b..d3c19e0c2b 100644 --- a/src/core/dbus-path.h +++ b/src/core/dbus-path.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/dbus-scope.c b/src/core/dbus-scope.c index f8fb373bf0..1abaf9f658 100644 --- a/src/core/dbus-scope.c +++ b/src/core/dbus-scope.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,17 +17,18 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "selinux-access.h" -#include "unit.h" -#include "scope.h" -#include "dbus.h" -#include "bus-util.h" -#include "bus-internal.h" +#include "alloc-util.h" #include "bus-common-errors.h" -#include "dbus-unit.h" +#include "bus-internal.h" +#include "bus-util.h" #include "dbus-cgroup.h" #include "dbus-kill.h" #include "dbus-scope.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "scope.h" +#include "selinux-access.h" +#include "unit.h" static int bus_scope_abandon(sd_bus_message *message, void *userdata, sd_bus_error *error) { Scope *s = userdata; @@ -148,7 +147,7 @@ static int bus_scope_set_transient_property( if (r < 0) return r; - unit_write_drop_in_format(UNIT(s), mode, name, "[Scope]\nTimeoutStopSec="USEC_FMT"us\n", s->timeout_stop_usec); + 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) @@ -204,7 +203,7 @@ int bus_scope_commit_properties(Unit *u) { } int bus_scope_send_request_stop(Scope *s) { - _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; _cleanup_free_ char *p = NULL; int r; @@ -226,5 +225,5 @@ int bus_scope_send_request_stop(Scope *s) { if (r < 0) return r; - return sd_bus_send_to(UNIT(s)->manager->api_bus, m, /* s->controller */ NULL, NULL); + return sd_bus_send_to(UNIT(s)->manager->api_bus, m, s->controller, NULL); } diff --git a/src/core/dbus-scope.h b/src/core/dbus-scope.h index 33beda47b7..270306f508 100644 --- a/src/core/dbus-scope.h +++ b/src/core/dbus-scope.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_scope_vtable[]; diff --git a/src/core/dbus-service.c b/src/core/dbus-service.c index e1f3d56495..3c55e0f7fe 100644 --- a/src/core/dbus-service.c +++ b/src/core/dbus-service.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,15 +17,20 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "strv.h" -#include "path-util.h" -#include "unit.h" -#include "service.h" +#include "alloc-util.h" +#include "async.h" +#include "bus-util.h" +#include "dbus-cgroup.h" #include "dbus-execute.h" #include "dbus-kill.h" -#include "dbus-cgroup.h" #include "dbus-service.h" -#include "bus-util.h" +#include "fd-util.h" +#include "fileio.h" +#include "path-util.h" +#include "service.h" +#include "string-util.h" +#include "strv.h" +#include "unit.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); @@ -44,12 +47,9 @@ const sd_bus_vtable bus_service_vtable[] = { 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("StartLimitInterval", "t", bus_property_get_usec, offsetof(Service, start_limit.interval), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("StartLimitBurst", "u", bus_property_get_unsigned, offsetof(Service, start_limit.burst), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("StartLimitAction", "s", property_get_failure_action, offsetof(Service, start_limit_action), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_PROPERTY("RebootArgument", "s", NULL, offsetof(Service, reboot_arg), SD_BUS_VTABLE_PROPERTY_CONST), SD_BUS_PROPERTY("FailureAction", "s", property_get_failure_action, offsetof(Service, failure_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), @@ -58,10 +58,16 @@ const sd_bus_vtable bus_service_vtable[] = { 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", NULL, offsetof(Service, n_fd_store_max), 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), @@ -69,6 +75,12 @@ const sd_bus_vtable bus_service_vtable[] = { 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_failure_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 }; @@ -94,7 +106,7 @@ static int bus_service_set_transient_property( if (mode != UNIT_CHECK) { s->remain_after_exit = b; - unit_write_drop_in_private_format(UNIT(s), mode, name, "RemainAfterExit=%s\n", yes_no(b)); + unit_write_drop_in_private_format(UNIT(s), mode, name, "RemainAfterExit=%s", yes_no(b)); } return 1; @@ -113,7 +125,53 @@ static int bus_service_set_transient_property( if (mode != UNIT_CHECK) { s->type = k; - unit_write_drop_in_private_format(UNIT(s), mode, name, "Type=%s\n", service_type_to_string(s->type)); + 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; @@ -209,7 +267,9 @@ static int bus_service_set_transient_property( a); } - fflush(f); + r = fflush_and_check(f); + if (r < 0) + return r; unit_write_drop_in_private(UNIT(s), mode, name, buf); } diff --git a/src/core/dbus-service.h b/src/core/dbus-service.h index aab9f7aa26..769a53769e 100644 --- a/src/core/dbus-service.h +++ b/src/core/dbus-service.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_service_vtable[]; diff --git a/src/core/dbus-slice.c b/src/core/dbus-slice.c index 09e78d1f33..e37f50b283 100644 --- a/src/core/dbus-slice.c +++ b/src/core/dbus-slice.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,10 +17,10 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "slice.h" #include "dbus-cgroup.h" #include "dbus-slice.h" +#include "slice.h" +#include "unit.h" const sd_bus_vtable bus_slice_vtable[] = { SD_BUS_VTABLE_START(0), diff --git a/src/core/dbus-slice.h b/src/core/dbus-slice.h index eadc3b1a9c..52ceebb135 100644 --- a/src/core/dbus-slice.h +++ b/src/core/dbus-slice.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_slice_vtable[]; diff --git a/src/core/dbus-snapshot.c b/src/core/dbus-snapshot.c deleted file mode 100644 index cfe44c9c15..0000000000 --- a/src/core/dbus-snapshot.c +++ /dev/null @@ -1,55 +0,0 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - -/*** - 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 "selinux-access.h" -#include "unit.h" -#include "dbus.h" -#include "snapshot.h" -#include "dbus-snapshot.h" - -int bus_snapshot_method_remove(sd_bus_message *message, void *userdata, sd_bus_error *error) { - Snapshot *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 */ - - snapshot_remove(s); - - return sd_bus_reply_method_return(message, NULL); -} - -const sd_bus_vtable bus_snapshot_vtable[] = { - SD_BUS_VTABLE_START(0), - SD_BUS_PROPERTY("Cleanup", "b", bus_property_get_bool, offsetof(Snapshot, cleanup), SD_BUS_VTABLE_PROPERTY_CONST), - SD_BUS_METHOD("Remove", NULL, NULL, bus_snapshot_method_remove, SD_BUS_VTABLE_UNPRIVILEGED), - SD_BUS_VTABLE_END -}; diff --git a/src/core/dbus-snapshot.h b/src/core/dbus-snapshot.h deleted file mode 100644 index 9288f44e15..0000000000 --- a/src/core/dbus-snapshot.h +++ /dev/null @@ -1,28 +0,0 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - -#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 "sd-bus.h" - -extern const sd_bus_vtable bus_snapshot_vtable[]; - -int bus_snapshot_method_remove(sd_bus_message *message, void *userdata, sd_bus_error *error); diff --git a/src/core/dbus-socket.c b/src/core/dbus-socket.c index 02599a9e55..21adb64e15 100644 --- a/src/core/dbus-socket.c +++ b/src/core/dbus-socket.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,12 +17,14 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "socket.h" -#include "dbus-execute.h" +#include "alloc-util.h" +#include "bus-util.h" #include "dbus-cgroup.h" +#include "dbus-execute.h" #include "dbus-socket.h" -#include "bus-util.h" +#include "socket.h" +#include "string-util.h" +#include "unit.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); @@ -68,6 +68,7 @@ static int property_get_listen( case SOCKET_SPECIAL: case SOCKET_MQUEUE: case SOCKET_FIFO: + case SOCKET_USB_FUNCTION: a = p->path; break; @@ -83,6 +84,25 @@ static int property_get_listen( 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), @@ -94,6 +114,7 @@ const sd_bus_vtable bus_socket_vtable[] = { 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), @@ -116,6 +137,7 @@ const sd_bus_vtable bus_socket_vtable[] = { 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), @@ -126,6 +148,12 @@ const sd_bus_vtable bus_socket_vtable[] = { 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), diff --git a/src/core/dbus-socket.h b/src/core/dbus-socket.h index 17164d9871..7a792c7a89 100644 --- a/src/core/dbus-socket.h +++ b/src/core/dbus-socket.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_socket_vtable[]; diff --git a/src/core/dbus-swap.c b/src/core/dbus-swap.c index 0093371306..85a2c26b98 100644 --- a/src/core/dbus-swap.c +++ b/src/core/dbus-swap.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,12 +18,13 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "swap.h" -#include "dbus-execute.h" +#include "bus-util.h" #include "dbus-cgroup.h" +#include "dbus-execute.h" #include "dbus-swap.h" -#include "bus-util.h" +#include "string-util.h" +#include "swap.h" +#include "unit.h" static int property_get_priority( sd_bus *bus, @@ -85,6 +84,8 @@ const sd_bus_vtable bus_swap_vtable[] = { 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 diff --git a/src/core/dbus-swap.h b/src/core/dbus-swap.h index 9469f68ab8..5238471f98 100644 --- a/src/core/dbus-swap.h +++ b/src/core/dbus-swap.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,6 +21,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_swap_vtable[]; diff --git a/src/core/dbus-target.c b/src/core/dbus-target.c index 350f5c3ed2..6858b1ce72 100644 --- a/src/core/dbus-target.c +++ b/src/core/dbus-target.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,8 +17,8 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" #include "dbus-target.h" +#include "unit.h" const sd_bus_vtable bus_target_vtable[] = { SD_BUS_VTABLE_START(0), diff --git a/src/core/dbus-target.h b/src/core/dbus-target.h index 4c4297bc9e..9be5ce06b7 100644 --- a/src/core/dbus-target.h +++ b/src/core/dbus-target.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -21,5 +19,6 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include "sd-bus.h" extern const sd_bus_vtable bus_target_vtable[]; diff --git a/src/core/dbus-timer.c b/src/core/dbus-timer.c index 74a9914358..efbb0e8915 100644 --- a/src/core/dbus-timer.c +++ b/src/core/dbus-timer.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,11 +17,12 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" -#include "timer.h" -#include "dbus-timer.h" +#include "alloc-util.h" #include "bus-util.h" +#include "dbus-timer.h" #include "strv.h" +#include "timer.h" +#include "unit.h" static BUS_DEFINE_PROPERTY_GET_ENUM(property_get_result, timer_result, TimerResult); @@ -157,7 +156,7 @@ static int property_get_next_elapse_monotonic( usec_t a, b; a = now(CLOCK_MONOTONIC); - b = now(CLOCK_BOOTTIME); + b = now(clock_boottime_or_monotonic()); if (t->next_elapse_monotonic_or_boottime + a > b) x = t->next_elapse_monotonic_or_boottime + a - b; @@ -179,8 +178,10 @@ const sd_bus_vtable bus_timer_vtable[] = { 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 }; @@ -219,7 +220,7 @@ static int bus_timer_set_transient_property( if (mode != UNIT_CHECK) { char time[FORMAT_TIMESPAN_MAX]; - unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s\n", name, format_timespan(time, sizeof(time), u, USEC_PER_MSEC)); + 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) @@ -248,12 +249,11 @@ static int bus_timer_set_transient_property( if (r < 0) return r; - unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s\n", name, str); + unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s", name, str); v = new0(TimerValue, 1); if (!v) { - if (c) - calendar_spec_free(c); + calendar_spec_free(c); return -ENOMEM; } @@ -265,25 +265,38 @@ static int bus_timer_set_transient_property( return 1; - } else if (streq(name, "AccuracySec")) { - + } 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) { - char time[FORMAT_TIMESPAN_MAX]; - t->accuracy_usec = u; - unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s\n", name, format_timespan(time, sizeof(time), u, USEC_PER_MSEC)); + unit_write_drop_in_private_format(UNIT(t), mode, name, "AccuracySec=" USEC_FMT "us", u); } return 1; - } else if (streq(name, "WakeSystem")) { + } 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); @@ -292,11 +305,24 @@ static int bus_timer_set_transient_property( if (mode != UNIT_CHECK) { t->wake_system = b; - unit_write_drop_in_private_format(UNIT(t), mode, name, "%s=%s\n", name, yes_no(t->wake_system)); + 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; diff --git a/src/core/dbus-timer.h b/src/core/dbus-timer.h index 103172f055..39053dc4a2 100644 --- a/src/core/dbus-timer.h +++ b/src/core/dbus-timer.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_timer_vtable[]; diff --git a/src/core/dbus-unit.c b/src/core/dbus-unit.c index 1892725f91..245912fc0f 100644 --- a/src/core/dbus-unit.c +++ b/src/core/dbus-unit.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,13 +18,22 @@ ***/ #include "sd-bus.h" + +#include "alloc-util.h" +#include "bus-common-errors.h" +#include "cgroup-util.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "fd-util.h" +#include "locale-util.h" #include "log.h" +#include "process-util.h" #include "selinux-access.h" -#include "cgroup-util.h" +#include "signal-util.h" +#include "special.h" +#include "string-util.h" #include "strv.h" -#include "bus-common-errors.h" -#include "dbus.h" -#include "dbus-unit.h" +#include "user-util.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); @@ -112,6 +119,22 @@ static int property_get_dependencies( 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, @@ -377,7 +400,7 @@ static int property_get_load_error( void *userdata, sd_bus_error *error) { - _cleanup_bus_error_free_ sd_bus_error e = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error e = SD_BUS_ERROR_NULL; Unit *u = userdata; assert(bus); @@ -390,6 +413,38 @@ static int property_get_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, @@ -399,13 +454,24 @@ int bus_unit_method_start_generic( 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 '$(unit)'."), + }; int r; assert(message); assert(u); assert(job_type >= 0 && job_type < _JOB_TYPE_MAX); - r = mac_selinux_unit_access_check(u, message, job_type == JOB_STOP ? "stop" : "start", error); + r = mac_selinux_unit_access_check( + u, message, + job_type_to_access_method(job_type), + error); if (r < 0) return r; @@ -417,7 +483,21 @@ int bus_unit_method_start_generic( if (mode < 0) return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Job mode %s invalid", smode); - r = bus_verify_manage_units_async(u->manager, message, error); + if (reload_if_possible) + verb = strjoin("reload-or-", job_type_to_string(job_type), NULL); + else + verb = strdup(job_type_to_string(job_type)); + if (!verb) + return -ENOMEM; + + r = bus_verify_manage_units_async_full( + u, + verb, + CAP_SYS_ADMIN, + job_type < _JOB_TYPE_MAX ? polkit_message_for_job[job_type] : NULL, + true, + message, + error); if (r < 0) return r; if (r == 0) @@ -480,10 +560,17 @@ int bus_unit_method_kill(sd_bus_message *message, void *userdata, sd_bus_error * return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid who argument %s", swho); } - if (signo <= 0 || signo >= _NSIG) + if (!SIGNAL_VALID(signo)) return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Signal number out of range."); - r = bus_verify_manage_units_async_for_kill(u->manager, message, error); + r = bus_verify_manage_units_async_full( + u, + "kill", + CAP_KILL, + N_("Authentication is required to kill '$(unit)'."), + true, + message, + error); if (r < 0) return r; if (r == 0) @@ -507,7 +594,14 @@ int bus_unit_method_reset_failed(sd_bus_message *message, void *userdata, sd_bus if (r < 0) return r; - r = bus_verify_manage_units_async(u->manager, message, error); + r = bus_verify_manage_units_async_full( + u, + "reset-failed", + CAP_SYS_ADMIN, + N_("Authentication is required to reset the \"failed\" state of '$(unit)'."), + true, + message, + error); if (r < 0) return r; if (r == 0) @@ -533,7 +627,14 @@ int bus_unit_method_set_properties(sd_bus_message *message, void *userdata, sd_b if (r < 0) return r; - r = bus_verify_manage_units_async(u->manager, message, error); + r = bus_verify_manage_units_async_full( + u, + "set-property", + CAP_SYS_ADMIN, + N_("Authentication is required to set properties on '$(unit)'."), + true, + message, + error); if (r < 0) return r; if (r == 0) @@ -546,6 +647,53 @@ int bus_unit_method_set_properties(sd_bus_message *message, void *userdata, sd_b 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), @@ -553,16 +701,12 @@ const sd_bus_vtable bus_unit_vtable[] = { 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("RequiresOverridable", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUIRES_OVERRIDABLE]), 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("RequisiteOverridable", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUISITE_OVERRIDABLE]), 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("RequiredByOverridable", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUIRED_BY_OVERRIDABLE]), 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("RequisiteOfOverridable", "as", property_get_dependencies, offsetof(Unit, dependencies[UNIT_REQUISITE_OF_OVERRIDABLE]), 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), @@ -587,6 +731,7 @@ const sd_bus_vtable bus_unit_vtable[] = { 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), @@ -603,7 +748,6 @@ const sd_bus_vtable bus_unit_vtable[] = { 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("IgnoreOnSnapshot", "b", bus_property_get_bool, offsetof(Unit, ignore_on_snapshot), 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_failure_action, offsetof(Unit, job_timeout_action), SD_BUS_VTABLE_PROPERTY_CONST), @@ -616,6 +760,11 @@ const sd_bus_vtable bus_unit_vtable[] = { 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("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_failure_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), @@ -627,7 +776,15 @@ const sd_bus_vtable bus_unit_vtable[] = { 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 }; @@ -673,6 +830,30 @@ static int property_get_current_memory( 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, @@ -697,17 +878,188 @@ static int property_get_cpu_usage( 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", NULL, offsetof(Unit, cgroup_path), 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_bus_message_unref_ sd_bus_message *m = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; _cleanup_free_ char *p = NULL; Unit *u = userdata; int r; @@ -753,7 +1105,7 @@ static int send_changed_signal(sd_bus *bus, void *userdata) { r = sd_bus_emit_properties_changed_strv( bus, p, - UNIT_VTABLE(u)->bus_interface, + unit_dbus_interface_from_type(u->type), NULL); if (r < 0) return r; @@ -784,7 +1136,7 @@ void bus_unit_send_change_signal(Unit *u) { } static int send_removed_signal(sd_bus *bus, void *userdata) { - _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + _cleanup_(sd_bus_message_unrefp) sd_bus_message *m = NULL; _cleanup_free_ char *p = NULL; Unit *u = userdata; int r; @@ -816,7 +1168,7 @@ void bus_unit_send_removed_signal(Unit *u) { int r; assert(u); - if (!u->sent_dbus_new_signal) + if (!u->sent_dbus_new_signal || u->in_dbus_queue) bus_unit_send_change_signal(u); if (!u->id) @@ -844,20 +1196,20 @@ int bus_unit_queue_job( 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_RELOAD; + type = JOB_TRY_RELOAD; } - r = mac_selinux_unit_access_check( - u, message, - (type == JOB_START || type == JOB_RESTART || type == JOB_TRY_RESTART) ? "start" : - type == JOB_STOP ? "stop" : "reload", error); - if (r < 0) - return r; - if (type == JOB_STOP && (u->load_state == UNIT_NOT_FOUND || u->load_state == UNIT_ERROR) && unit_active_state(u) == UNIT_INACTIVE) @@ -865,10 +1217,11 @@ int bus_unit_queue_job( 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_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, true, error, &j); + r = manager_add_job(u->manager, type, u, mode, error, &j); if (r < 0) return r; @@ -916,7 +1269,7 @@ static int bus_unit_set_transient_property( if (r < 0) return r; - unit_write_drop_in_format(u, mode, name, "[Unit]\nDescription=%s\n", d); + unit_write_drop_in_format(u, mode, name, "[Unit]\nDescription=%s", d); } return 1; @@ -930,43 +1283,49 @@ static int bus_unit_set_transient_property( if (mode != UNIT_CHECK) { u->default_dependencies = b; - unit_write_drop_in_format(u, mode, name, "[Unit]\nDefaultDependencies=%s\n", yes_no(b)); + unit_write_drop_in_format(u, mode, name, "[Unit]\nDefaultDependencies=%s", yes_no(b)); } return 1; - } else if (streq(name, "Slice") && unit_get_cgroup_context(u)) { + } 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) || !endswith(s, ".slice")) - return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid slice name %s", s); + 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); - if (isempty(s)) { - if (mode != UNIT_CHECK) { - unit_ref_unset(&u->slice); - unit_remove_drop_in(u, mode, name); - } - } else { - Unit *slice; + /* 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; - r = manager_load_unit(u->manager, s, NULL, error, &slice); + 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; - if (slice->type != UNIT_SLICE) - return -EINVAL; - - if (mode != UNIT_CHECK) { - unit_ref_set(&u->slice, slice); - unit_write_drop_in_private_format(u, mode, name, "Slice=%s\n", s); - } + unit_write_drop_in_private_format(u, mode, name, "Slice=%s", s); } return 1; + } else if (STR_IN_SET(name, "Requires", "RequiresOverridable", "Requisite", "RequisiteOverridable", @@ -981,9 +1340,15 @@ static int bus_unit_set_transient_property( UnitDependency d; const char *other; - d = unit_dependency_from_string(name); - if (d < 0) - return -EINVAL; + 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) @@ -1004,7 +1369,7 @@ static int bus_unit_set_transient_property( if (!label) return -ENOMEM; - unit_write_drop_in_format(u, mode, label, "[Unit]\n%s=%s\n", name, other); + unit_write_drop_in_format(u, mode, label, "[Unit]\n%s=%s", name, other); } } @@ -1016,6 +1381,29 @@ static int bus_unit_set_transient_property( 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; @@ -1102,3 +1490,89 @@ int bus_unit_set_properties( 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/core/dbus-unit.h b/src/core/dbus-unit.h index b622e0ae8d..b280de7a1d 100644 --- a/src/core/dbus-unit.h +++ b/src/core/dbus-unit.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,6 +20,7 @@ ***/ #include "sd-bus.h" + #include "unit.h" extern const sd_bus_vtable bus_unit_vtable[]; @@ -34,6 +33,15 @@ int bus_unit_method_start_generic(sd_bus_message *message, Unit *u, JobType job_ 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_queue_job(sd_bus_message *message, Unit *u, JobType type, JobMode mode, bool reload_if_possible, 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/core/dbus.c b/src/core/dbus.c index 057653a8b5..070974fe66 100644 --- a/src/core/dbus.c +++ b/src/core/dbus.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,29 +17,34 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/epoll.h> #include <errno.h> +#include <sys/epoll.h> #include <unistd.h> #include "sd-bus.h" -#include "log.h" -#include "strv.h" -#include "mkdir.h" -#include "missing.h" -#include "dbus-unit.h" -#include "dbus-job.h" -#include "dbus-manager.h" + +#include "alloc-util.h" +#include "bus-common-errors.h" +#include "bus-error.h" +#include "bus-internal.h" +#include "bus-util.h" +#include "dbus-cgroup.h" #include "dbus-execute.h" +#include "dbus-job.h" #include "dbus-kill.h" -#include "dbus-cgroup.h" -#include "special.h" +#include "dbus-manager.h" +#include "dbus-unit.h" #include "dbus.h" -#include "bus-util.h" -#include "bus-error.h" -#include "bus-common-errors.h" -#include "strxcpyx.h" -#include "bus-internal.h" +#include "fd-util.h" +#include "log.h" +#include "missing.h" +#include "mkdir.h" #include "selinux-access.h" +#include "special.h" +#include "string-util.h" +#include "strv.h" +#include "strxcpyx.h" +#include "user-util.h" #define CONNECTIONS_MAX 4096 @@ -68,28 +71,42 @@ int bus_send_queued_message(Manager *m) { 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_bus_creds_unref_ sd_bus_creds *creds = NULL; - const char *cgroup, *me; + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; Manager *m = userdata; + const char *cgroup; uid_t sender_uid; - sd_bus *bus; int r; assert(message); assert(m); - /* ignore recursive events sent by us on the system/user bus */ - bus = sd_bus_message_get_bus(message); - if (!sd_bus_is_server(bus)) { - r = sd_bus_get_unique_name(bus, &me); - if (r < 0) - return r; - - if (streq_ptr(sd_bus_message_get_sender(message), me)) - return 0; - } - /* only accept org.freedesktop.systemd1.Agent from UID=0 */ r = sd_bus_query_sender_creds(message, SD_BUS_CREDS_EUID, &creds); if (r < 0) @@ -107,16 +124,6 @@ static int signal_agent_released(sd_bus_message *message, void *userdata, sd_bus } manager_notify_cgroup_empty(m, cgroup); - - /* if running as system-instance, forward under our name */ - if (m->running_as == MANAGER_SYSTEM && m->system_bus) { - r = sd_bus_message_rewind(message, 1); - if (r >= 0) - r = sd_bus_send(m->system_bus, message, NULL); - if (r < 0) - log_warning_errno(r, "Failed to forward Released message: %m"); - } - return 0; } @@ -140,31 +147,9 @@ static int signal_disconnected(sd_bus_message *message, void *userdata, sd_bus_e return 0; } -static int signal_name_owner_changed(sd_bus_message *message, void *userdata, sd_bus_error *error) { - const char *name, *old_owner, *new_owner; - Manager *m = userdata; - int r; - - assert(message); - assert(m); - - r = sd_bus_message_read(message, "sss", &name, &old_owner, &new_owner); - if (r < 0) { - bus_log_parse_error(r); - return 0; - } - - manager_dispatch_bus_name_owner_changed( - m, name, - isempty(old_owner) ? NULL : old_owner, - isempty(new_owner) ? NULL : new_owner); - - return 0; -} - static int signal_activation_request(sd_bus_message *message, void *userdata, sd_bus_error *ret_error) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; + _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; @@ -194,7 +179,7 @@ static int signal_activation_request(sd_bus_message *message, void *userdata, sd goto failed; } - r = manager_add_job(m, JOB_START, u, JOB_REPLACE, true, &error, NULL); + r = manager_add_job(m, JOB_START, u, JOB_REPLACE, &error, NULL); if (r < 0) goto failed; @@ -262,7 +247,7 @@ static int mac_selinux_filter(sd_bus_message *message, void *userdata, sd_bus_er } if (streq_ptr(path, "/org/freedesktop/systemd1/unit/self")) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _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); @@ -321,7 +306,7 @@ static int find_unit(Manager *m, sd_bus *bus, const char *path, Unit **unit, sd_ assert(path); if (streq_ptr(path, "/org/freedesktop/systemd1/unit/self")) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; sd_bus_message *message; pid_t pid; @@ -378,7 +363,7 @@ static int bus_unit_interface_find(sd_bus *bus, const char *path, const char *in if (r <= 0) return r; - if (!streq_ptr(interface, UNIT_VTABLE(u)->bus_interface)) + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) return 0; *found = u; @@ -400,10 +385,10 @@ static int bus_unit_cgroup_find(sd_bus *bus, const char *path, const char *inter if (r <= 0) return r; - if (!streq_ptr(interface, UNIT_VTABLE(u)->bus_interface)) + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) return 0; - if (!unit_get_cgroup_context(u)) + if (!UNIT_HAS_CGROUP_CONTEXT(u)) return 0; *found = u; @@ -426,7 +411,7 @@ static int bus_cgroup_context_find(sd_bus *bus, const char *path, const char *in if (r <= 0) return r; - if (!streq_ptr(interface, UNIT_VTABLE(u)->bus_interface)) + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) return 0; c = unit_get_cgroup_context(u); @@ -453,7 +438,7 @@ static int bus_exec_context_find(sd_bus *bus, const char *path, const char *inte if (r <= 0) return r; - if (!streq_ptr(interface, UNIT_VTABLE(u)->bus_interface)) + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) return 0; c = unit_get_exec_context(u); @@ -480,7 +465,7 @@ static int bus_kill_context_find(sd_bus *bus, const char *path, const char *inte if (r <= 0) return r; - if (!streq_ptr(interface, UNIT_VTABLE(u)->bus_interface)) + if (!streq_ptr(interface, unit_dbus_interface_from_type(u->type))) return 0; c = unit_get_kill_context(u); @@ -577,30 +562,34 @@ static int bus_setup_api_vtables(Manager *m, sd_bus *bus) { return log_error_errno(r, "Failed to add job enumerator: %m"); for (t = 0; t < _UNIT_TYPE_MAX; t++) { - r = sd_bus_add_fallback_vtable(bus, NULL, "/org/freedesktop/systemd1/unit", unit_vtable[t]->bus_interface, unit_vtable[t]->bus_vtable, bus_unit_interface_find, m); + 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", unit_vtable[t]->bus_interface); + 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", unit_vtable[t]->bus_interface, bus_unit_cgroup_vtable, bus_unit_cgroup_find, m); + 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", unit_vtable[t]->bus_interface); + 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", unit_vtable[t]->bus_interface, bus_cgroup_vtable, bus_cgroup_context_find, m); + 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", unit_vtable[t]->bus_interface); + 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", unit_vtable[t]->bus_interface, bus_exec_vtable, bus_exec_context_find, m); + 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", unit_vtable[t]->bus_interface); + 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", unit_vtable[t]->bus_interface, bus_kill_vtable, bus_kill_context_find, m); + 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", unit_vtable[t]->bus_interface); + return log_error_errno(r, "Failed to register kill vtable for %s: %m", interface); } } @@ -630,7 +619,7 @@ static int bus_setup_disconnected_match(Manager *m, sd_bus *bus) { } static int bus_on_connection(sd_event_source *s, int fd, uint32_t revents, void *userdata) { - _cleanup_bus_unref_ sd_bus *bus = NULL; + _cleanup_(sd_bus_unrefp) sd_bus *bus = NULL; _cleanup_close_ int nfd = -1; Manager *m = userdata; sd_id128_t id; @@ -705,25 +694,6 @@ static int bus_on_connection(sd_event_source *s, int fd, uint32_t revents, void return 0; } - if (m->running_as == MANAGER_SYSTEM) { - /* When we run as system instance we get the Released - * signal via a direct connection */ - - 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 new connection bus: %m"); - return 0; - } - } - r = bus_setup_disconnected_match(m, bus); if (r < 0) return 0; @@ -747,9 +717,11 @@ static int bus_on_connection(sd_event_source *s, int fd, uint32_t revents, void return 0; } -static int bus_list_names(Manager *m, sd_bus *bus) { +int manager_sync_bus_names(Manager *m, sd_bus *bus) { _cleanup_strv_free_ char **names = NULL; - char **i; + const char *name; + Iterator i; + Unit *u; int r; assert(m); @@ -759,16 +731,64 @@ static int bus_list_names(Manager *m, sd_bus *bus) { if (r < 0) return log_error_errno(r, "Failed to get initial list of names: %m"); - /* This is a bit hacky, we say the owner of the name is the - * name itself, because we don't want the extra traffic to - * figure out the real owner. */ - STRV_FOREACH(i, names) - manager_dispatch_bus_name_owner_changed(m, *i, NULL, *i); + /* 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); @@ -786,17 +806,11 @@ static int bus_setup_api(Manager *m, sd_bus *bus) { if (r < 0) return r; - r = sd_bus_add_match( - bus, - NULL, - "type='signal'," - "sender='org.freedesktop.DBus'," - "path='/org/freedesktop/DBus'," - "interface='org.freedesktop.DBus'," - "member='NameOwnerChanged'", - signal_name_owner_changed, m); - if (r < 0) - log_warning_errno(r, "Failed to subscribe to NameOwnerChanged signal: %m"); + 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, @@ -819,24 +833,26 @@ static int bus_setup_api(Manager *m, sd_bus *bus) { if (r < 0) return log_error_errno(r, "Failed to register name: %m"); - bus_list_names(m, bus); + 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_bus_unref_ sd_bus *bus = NULL; + _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 (m->running_as == MANAGER_SYSTEM && m->system_bus) + if (MANAGER_IS_SYSTEM(m) && m->system_bus) bus = sd_bus_ref(m->system_bus); else { - if (m->running_as == MANAGER_SYSTEM) + if (MANAGER_IS_SYSTEM(m)) r = sd_bus_open_system(&bus); else r = sd_bus_open_user(&bus); @@ -875,8 +891,8 @@ static int bus_setup_system(Manager *m, sd_bus *bus) { assert(m); assert(bus); - /* On kdbus or if we are a user instance we get the Released message via the system bus */ - if (m->running_as == MANAGER_USER || m->kdbus_fd >= 0) { + /* 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, @@ -894,14 +910,14 @@ static int bus_setup_system(Manager *m, sd_bus *bus) { } static int bus_init_system(Manager *m) { - _cleanup_bus_unref_ sd_bus *bus = NULL; + _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 (m->running_as == MANAGER_SYSTEM && m->api_bus) { + if (MANAGER_IS_SYSTEM(m) && m->api_bus) { m->system_bus = sd_bus_ref(m->api_bus); return 0; } @@ -948,18 +964,14 @@ static int bus_init_private(Manager *m) { if (m->private_listen_fd >= 0) return 0; - /* We don't need the private socket if we have kdbus */ - if (m->kdbus_fd >= 0) - return 0; - - if (m->running_as == MANAGER_SYSTEM) { + 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 = offsetof(union sockaddr_union, un.sun_path) + strlen("/run/systemd/private"); + salen = SOCKADDR_UN_LEN(sa.un); } else { size_t left = sizeof(sa.un.sun_path); char *p = sa.un.sun_path; @@ -1051,7 +1063,7 @@ static void destroy_bus(Manager *m, sd_bus **bus) { /* Possibly flush unwritten data, but only if we are * unprivileged, since we don't want to sync here */ - if (m->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(m)) sd_bus_flush(*bus); /* And destroy the object */ @@ -1071,12 +1083,10 @@ void bus_done(Manager *m) { while ((b = set_steal_first(m->private_buses))) destroy_bus(m, &b); - set_free(m->private_buses); - m->private_buses = NULL; + m->private_buses = set_free(m->private_buses); m->subscribed = sd_bus_track_unref(m->subscribed); - strv_free(m->deserialized_subscribed); - m->deserialized_subscribed = NULL; + 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); @@ -1154,82 +1164,73 @@ int bus_foreach_bus( return ret; } -void bus_track_serialize(sd_bus_track *t, FILE *f) { +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)) - fprintf(f, "subscribed=%s\n", n); -} - -int bus_track_deserialize_item(char ***l, const char *line) { - const char *e; - int r; + for (n = sd_bus_track_first(t); n; n = sd_bus_track_next(t)) { + int c, j; - assert(l); - assert(line); + c = sd_bus_track_count_name(t, n); - e = startswith(line, "subscribed="); - if (!e) - return 0; - - r = strv_extend(l, e); - if (r < 0) - return r; - - return 1; + 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, char ***l) { +int bus_track_coldplug(Manager *m, sd_bus_track **t, bool recursive, char **l) { + char **i; int r = 0; assert(m); assert(t); - assert(l); - - if (!strv_isempty(*l) && m->api_bus) { - char **i; - if (!*t) { - r = sd_bus_track_new(m->api_bus, t, NULL, NULL); - if (r < 0) - return r; - } + if (strv_isempty(l)) + return 0; - r = 0; - STRV_FOREACH(i, *l) { - int k; + if (!m->api_bus) + return 0; - k = sd_bus_track_add_name(*t, *i); - if (k < 0) - r = k; - } + if (!*t) { + r = sd_bus_track_new(m->api_bus, t, NULL, NULL); + if (r < 0) + return r; } - strv_free(*l); - *l = NULL; + 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", false, UID_INVALID, &m->polkit_registry, error); -} - -/* Same as bus_verify_manage_unit_async(), but checks for CAP_KILL instead of CAP_SYS_ADMIN */ -int bus_verify_manage_units_async_for_kill(Manager *m, sd_bus_message *call, sd_bus_error *error) { - return bus_verify_polkit_async(call, CAP_KILL, "org.freedesktop.systemd1.manage-units", false, UID_INVALID, &m->polkit_registry, 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", false, UID_INVALID, &m->polkit_registry, 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", false, UID_INVALID, &m->polkit_registry, 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", false, UID_INVALID, &m->polkit_registry, 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/core/dbus.h b/src/core/dbus.h index 4832722069..a092ed9d76 100644 --- a/src/core/dbus.h +++ b/src/core/dbus.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -30,14 +28,16 @@ void bus_done(Manager *m); int bus_fdset_add_all(Manager *m, FDSet *fds); -void bus_track_serialize(sd_bus_track *t, FILE *f); -int bus_track_deserialize_item(char ***l, const char *line); -int bus_track_coldplug(Manager *m, sd_bus_track **t, char ***l); +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_units_async_for_kill(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/core/device.c b/src/core/device.c index e7efcf0f0a..8a3e888e5e 100644 --- a/src/core/device.c +++ b/src/core/device.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,16 +19,21 @@ #include <errno.h> #include <sys/epoll.h> -#include <libudev.h> -#include "log.h" -#include "unit-name.h" +#include "libudev.h" + +#include "alloc-util.h" #include "dbus-device.h" +#include "device.h" +#include "log.h" +#include "parse-util.h" #include "path-util.h" +#include "stat-util.h" +#include "string-util.h" +#include "swap.h" #include "udev-util.h" +#include "unit-name.h" #include "unit.h" -#include "swap.h" -#include "device.h" static const UnitActiveState state_translation_table[_DEVICE_STATE_MAX] = { [DEVICE_DEAD] = UNIT_INACTIVE, @@ -60,8 +63,7 @@ static void device_unset_sysfs(Device *d) { else hashmap_remove(devices, d->sysfs); - free(d->sysfs); - d->sysfs = NULL; + d->sysfs = mfree(d->sysfs); } static int device_set_sysfs(Device *d, const char *sysfs) { @@ -113,7 +115,6 @@ static void device_init(Unit *u) { u->job_timeout = u->manager->default_timeout_start_usec; u->ignore_on_isolate = true; - u->ignore_on_snapshot = true; } static void device_done(Unit *u) { @@ -264,7 +265,7 @@ static int device_add_udev_wants(Unit *u, struct udev_device *dev) { assert(u); assert(dev); - property = u->manager->running_as == MANAGER_USER ? "MANAGER_USER_WANTS" : "SYSTEMD_WANTS"; + property = MANAGER_IS_USER(u->manager) ? "SYSTEMD_USER_WANTS" : "SYSTEMD_WANTS"; wants = udev_device_get_property_value(dev, property); if (!wants) return 0; @@ -312,12 +313,19 @@ static int device_setup_unit(Manager *m, struct udev_device *dev, const char *pa u = manager_get_unit(m, e); - if (u && - sysfs && - DEVICE(u)->sysfs && - !path_equal(DEVICE(u)->sysfs, sysfs)) { - log_unit_debug(u, "Device %s appeared twice with different sysfs paths %s and %s", e, DEVICE(u)->sysfs, sysfs); - return -EEXIST; + /* 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) { @@ -456,6 +464,10 @@ static void device_update_found_one(Device *d, bool add, DeviceFound found, bool 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. */ @@ -595,11 +607,10 @@ static void device_shutdown(Manager *m) { m->udev_monitor = NULL; } - hashmap_free(m->devices_by_sysfs); - m->devices_by_sysfs = NULL; + m->devices_by_sysfs = hashmap_free(m->devices_by_sysfs); } -static int device_enumerate(Manager *m) { +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; @@ -609,7 +620,7 @@ static int device_enumerate(Manager *m) { if (!m->udev_monitor) { m->udev_monitor = udev_monitor_new_from_netlink(m->udev, "udev"); if (!m->udev_monitor) { - r = -ENOMEM; + log_oom(); goto fail; } @@ -619,37 +630,49 @@ static int device_enumerate(Manager *m) { (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) + 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) + 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) + 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) { - r = -ENOMEM; + log_oom(); goto fail; } r = udev_enumerate_add_match_tag(e, "systemd"); - if (r < 0) + 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) + 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) + 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) { @@ -672,13 +695,10 @@ static int device_enumerate(Manager *m) { device_update_found_by_sysfs(m, sysfs, true, DEVICE_FOUND_UDEV, false); } - return 0; + return; fail: - log_error_errno(r, "Failed to enumerate devices: %m"); - device_shutdown(m); - return r; } static int device_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { @@ -818,14 +838,6 @@ int device_found_node(Manager *m, const char *node, bool add, DeviceFound found, return device_update_found_by_name(m, node, add, found, now); } -static const char* const device_state_table[_DEVICE_STATE_MAX] = { - [DEVICE_DEAD] = "dead", - [DEVICE_TENTATIVE] = "tentative", - [DEVICE_PLUGGED] = "plugged", -}; - -DEFINE_STRING_TABLE_LOOKUP(device_state, DeviceState); - const UnitVTable device_vtable = { .object_size = sizeof(Device), .sections = @@ -833,8 +845,6 @@ const UnitVTable device_vtable = { "Device\0" "Install\0", - .no_instances = true, - .init = device_init, .done = device_done, .load = unit_load_fragment_and_dropin_optional, @@ -849,7 +859,6 @@ const UnitVTable device_vtable = { .active_state = device_active_state, .sub_state_to_string = device_sub_state_to_string, - .bus_interface = "org.freedesktop.systemd1.Device", .bus_vtable = bus_device_vtable, .following = device_following, diff --git a/src/core/device.h b/src/core/device.h index 10ab113176..184a1a349b 100644 --- a/src/core/device.h +++ b/src/core/device.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,16 +21,6 @@ typedef struct Device Device; -/* We simply watch devices, we cannot plug/unplug them. That - * simplifies the state engine greatly */ -typedef enum DeviceState { - DEVICE_DEAD, - DEVICE_TENTATIVE, /* mounted or swapped, but not (yet) announced by udev */ - DEVICE_PLUGGED, /* announced by udev */ - _DEVICE_STATE_MAX, - _DEVICE_STATE_INVALID = -1 -} DeviceState; - typedef enum DeviceFound { DEVICE_NOT_FOUND = 0, DEVICE_FOUND_UDEV = 1, @@ -56,7 +44,4 @@ struct Device { extern const UnitVTable device_vtable; -const char* device_state_to_string(DeviceState i) _const_; -DeviceState device_state_from_string(const char *s) _pure_; - int device_found_node(Manager *m, const char *node, bool add, DeviceFound found, bool now); diff --git a/src/core/dynamic-user.c b/src/core/dynamic-user.c new file mode 100644 index 0000000000..1043da3eb7 --- /dev/null +++ b/src/core/dynamic-user.c @@ -0,0 +1,796 @@ +/*** + 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 "dynamic-user.h" +#include "fd-util.h" +#include "fs-util.h" +#include "parse-util.h" +#include "random-util.h" +#include "stdio-util.h" +#include "string-util.h" +#include "user-util.h" +#include "fileio.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); + free(d); + + return NULL; +} + +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/core/dynamic-user.h b/src/core/dynamic-user.h new file mode 100644 index 0000000000..0b8bce1a72 --- /dev/null +++ b/src/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/core/execute.c b/src/core/execute.c index 21721dc240..0c983f4953 100644 --- a/src/core/execute.c +++ b/src/core/execute.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,17 +19,21 @@ #include <errno.h> #include <fcntl.h> -#include <unistd.h> -#include <string.h> -#include <signal.h> -#include <sys/socket.h> -#include <sys/un.h> -#include <sys/prctl.h> -#include <sys/stat.h> +#include <glob.h> #include <grp.h> #include <poll.h> -#include <glob.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/socket.h> +#include <sys/stat.h> +#include <sys/un.h> +#include <unistd.h> +#include <utmpx.h> #ifdef HAVE_PAM #include <security/pam_appl.h> @@ -50,45 +52,55 @@ #endif #include "sd-messages.h" -#include "rm-rf.h" -#include "strv.h" -#include "macro.h" -#include "capability.h" -#include "util.h" -#include "log.h" -#include "ioprio.h" -#include "securebits.h" -#include "namespace.h" -#include "exit-status.h" -#include "missing.h" -#include "utmp-wtmp.h" + +#include "af-list.h" +#include "alloc-util.h" +#ifdef HAVE_APPARMOR +#include "apparmor-util.h" +#endif +#include "async.h" +#include "barrier.h" +#include "cap-list.h" +#include "capability-util.h" #include "def.h" -#include "path-util.h" #include "env-util.h" -#include "fileio.h" -#include "unit.h" -#include "async.h" -#include "selinux-util.h" #include "errno-list.h" -#include "af-list.h" -#include "mkdir.h" -#include "smack-util.h" -#include "bus-endpoint.h" -#include "cap-list.h" +#include "execute.h" +#include "exit-status.h" +#include "fd-util.h" +#include "fileio.h" #include "formats-util.h" +#include "fs-util.h" +#include "glob-util.h" +#include "io-util.h" +#include "ioprio.h" +#include "log.h" +#include "macro.h" +#include "missing.h" +#include "mkdir.h" +#include "namespace.h" +#include "parse-util.h" +#include "path-util.h" #include "process-util.h" -#include "terminal-util.h" -#include "signal-util.h" - -#ifdef HAVE_APPARMOR -#include "apparmor-util.h" -#endif - +#include "rlimit-util.h" +#include "rm-rf.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" #endif - -#include "execute.h" +#include "securebits.h" +#include "selinux-util.h" +#include "signal-util.h" +#include "smack-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "syslog-util.h" +#include "terminal-util.h" +#include "unit.h" +#include "user-util.h" +#include "util.h" +#include "utmp-wtmp.h" #define IDLE_TIMEOUT_USEC (5*USEC_PER_SEC) #define IDLE_TIMEOUT2_USEC (1*USEC_PER_SEC) @@ -121,7 +133,8 @@ static int shift_fds(int fds[], unsigned n_fds) { if (fds[i] == i+3) continue; - if ((nfd = fcntl(fds[i], F_DUPFD, i+3)) < 0) + nfd = fcntl(fds[i], F_DUPFD, i + 3); + if (nfd < 0) return -errno; safe_close(fds[i]); @@ -155,48 +168,89 @@ static int flags_fds(const int fds[], unsigned n_fds, bool nonblock) { for (i = 0; i < n_fds; i++) { - if ((r = fd_nonblock(fds[i], nonblock)) < 0) + 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 */ - if ((r = fd_cloexec(fds[i], false)) < 0) + r = fd_cloexec(fds[i], false); + if (r < 0) return r; } return 0; } -_pure_ static const char *tty_path(const ExecContext *context) { +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) { +static void exec_context_tty_reset(const ExecContext *context, const ExecParameters *p) { + const char *path; + assert(context); - if (context->tty_vhangup) - terminal_vhangup(tty_path(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) - reset_terminal(tty_path(context)); + 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 && context->tty_path) - vt_disallocate(context->tty_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 - o == EXEC_OUTPUT_TTY || - o == EXEC_OUTPUT_SYSLOG_AND_CONSOLE || - o == EXEC_OUTPUT_KMSG_AND_CONSOLE || - o == EXEC_OUTPUT_JOURNAL_AND_CONSOLE; + 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) { @@ -244,7 +298,7 @@ static int connect_journal_socket(int fd, uid_t uid, gid_t gid) { } } - r = connect(fd, &sa.sa, offsetof(struct sockaddr_un, sun_path) + strlen(sa.un.sun_path)); + r = connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); if (r < 0) r = -errno; @@ -261,7 +315,15 @@ static int connect_journal_socket(int fd, uid_t uid, gid_t gid) { return r; } -static int connect_logger_as(const ExecContext *context, ExecOutput output, const char *ident, const char *unit_id, int nfd, uid_t uid, gid_t gid) { +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); @@ -282,7 +344,7 @@ static int connect_logger_as(const ExecContext *context, ExecOutput output, cons return -errno; } - fd_inc_sndbuf(fd, SNDBUF_SIZE); + (void) fd_inc_sndbuf(fd, SNDBUF_SIZE); dprintf(fd, "%s\n" @@ -293,18 +355,18 @@ static int connect_logger_as(const ExecContext *context, ExecOutput output, cons "%i\n" "%i\n", context->syslog_identifier ? context->syslog_identifier : ident, - unit_id, + 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) { - r = dup2(fd, nfd) < 0 ? -errno : nfd; - safe_close(fd); - } else - r = nfd; + if (fd == nfd) + return nfd; + + r = dup2(fd, nfd) < 0 ? -errno : nfd; + safe_close(fd); return r; } @@ -314,7 +376,8 @@ static int open_terminal_as(const char *path, mode_t mode, int nfd) { assert(path); assert(nfd >= 0); - if ((fd = open_terminal(path, mode | O_NOCTTY)) < 0) + fd = open_terminal(path, mode | O_NOCTTY); + if (fd < 0) return fd; if (fd != nfd) { @@ -326,13 +389,6 @@ static int open_terminal_as(const char *path, mode_t mode, int nfd) { return r; } -static bool is_terminal_input(ExecInput i) { - return - i == EXEC_INPUT_TTY || - i == EXEC_INPUT_TTY_FORCE || - i == EXEC_INPUT_TTY_FAIL; -} - static int fixup_input(ExecInput std_input, int socket_fd, bool apply_tty_stdin) { if (is_terminal_input(std_input) && !apply_tty_stdin) @@ -352,12 +408,28 @@ static int fixup_output(ExecOutput std_output, int socket_fd) { return std_output; } -static int setup_input(const ExecContext *context, int socket_fd, bool apply_tty_stdin) { +static int setup_input( + const ExecContext *context, + const ExecParameters *params, + int socket_fd) { + 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, apply_tty_stdin); + i = fixup_input(context->std_input, socket_fd, params->flags & EXEC_APPLY_TTY_STDIN); switch (i) { @@ -369,7 +441,7 @@ static int setup_input(const ExecContext *context, int socket_fd, bool apply_tty case EXEC_INPUT_TTY_FAIL: { int fd, r; - fd = acquire_terminal(tty_path(context), + fd = acquire_terminal(exec_context_tty_path(context), i == EXEC_INPUT_TTY_FAIL, i == EXEC_INPUT_TTY_FORCE, false, @@ -394,16 +466,45 @@ static int setup_input(const ExecContext *context, int socket_fd, bool apply_tty } } -static int setup_output(Unit *unit, const ExecContext *context, int fileno, int socket_fd, const char *ident, bool apply_tty_stdin, uid_t uid, gid_t gid) { +static int setup_output( + Unit *unit, + const ExecContext *context, + const ExecParameters *params, + int fileno, + int socket_fd, + 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, apply_tty_stdin); + 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) { @@ -430,7 +531,7 @@ static int setup_output(Unit *unit, const ExecContext *context, int fileno, int } 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(tty_path(context), O_WRONLY, fileno); + 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) @@ -454,7 +555,7 @@ static int setup_output(Unit *unit, const ExecContext *context, int fileno, int return dup2(STDIN_FILENO, fileno) < 0 ? -errno : fileno; /* We don't reset the terminal if this is just about output */ - return open_terminal_as(tty_path(context), O_WRONLY, fileno); + return open_terminal_as(exec_context_tty_path(context), O_WRONLY, fileno); case EXEC_OUTPUT_SYSLOG: case EXEC_OUTPUT_SYSLOG_AND_CONSOLE: @@ -462,10 +563,21 @@ static int setup_output(Unit *unit, const ExecContext *context, int fileno, int case EXEC_OUTPUT_KMSG_AND_CONSOLE: case EXEC_OUTPUT_JOURNAL: case EXEC_OUTPUT_JOURNAL_AND_CONSOLE: - r = connect_logger_as(context, o, ident, unit->id, fileno, uid, gid); + 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; @@ -483,6 +595,10 @@ static int chown_terminal(int fd, uid_t uid) { 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); @@ -496,9 +612,9 @@ static int chown_terminal(int fd, uid_t uid) { return 0; } -static int setup_confirm_stdio(int *_saved_stdin, - int *_saved_stdout) { - int fd = -1, saved_stdin, saved_stdout = -1, r; +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); @@ -508,10 +624,8 @@ static int setup_confirm_stdio(int *_saved_stdin, return -errno; saved_stdout = fcntl(STDOUT_FILENO, F_DUPFD, 3); - if (saved_stdout < 0) { - r = errno; - goto fail; - } + if (saved_stdout < 0) + return -errno; fd = acquire_terminal( "/dev/console", @@ -519,39 +633,33 @@ static int setup_confirm_stdio(int *_saved_stdin, false, false, DEFAULT_CONFIRM_USEC); - if (fd < 0) { - r = fd; - goto fail; - } + if (fd < 0) + return fd; r = chown_terminal(fd, getuid()); if (r < 0) - goto fail; + return r; - if (dup2(fd, STDIN_FILENO) < 0) { - r = -errno; - goto fail; - } + r = reset_terminal_fd(fd, true); + if (r < 0) + return r; - if (dup2(fd, STDOUT_FILENO) < 0) { - r = -errno; - goto fail; - } + 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; - return 0; - -fail: - safe_close(saved_stdout); - safe_close(saved_stdin); - safe_close(fd); + saved_stdin = saved_stdout = -1; - return r; + return 0; } _printf_(1, 2) static int write_confirm_message(const char *format, ...) { @@ -571,9 +679,7 @@ _printf_(1, 2) static int write_confirm_message(const char *format, ...) { return 0; } -static int restore_confirm_stdio(int *saved_stdin, - int *saved_stdout) { - +static int restore_confirm_stdio(int *saved_stdin, int *saved_stdout) { int r = 0; assert(saved_stdin); @@ -589,8 +695,8 @@ static int restore_confirm_stdio(int *saved_stdin, if (dup2(*saved_stdout, STDOUT_FILENO) < 0) r = -errno; - safe_close(*saved_stdin); - safe_close(*saved_stdout); + *saved_stdin = safe_close(*saved_stdin); + *saved_stdout = safe_close(*saved_stdout); return r; } @@ -624,14 +730,6 @@ static int enforce_groups(const ExecContext *context, const char *username, gid_ * we avoid NSS lookups for gid=0. */ if (context->group || username) { - - if (context->group) { - const char *g = context->group; - - if ((r = get_group_creds(&g, &gid)) < 0) - return r; - } - /* First step, initialize groups from /etc/groups */ if (username && gid != 0) { if (initgroups(username, gid) < 0) @@ -657,7 +755,8 @@ static int enforce_groups(const ExecContext *context, const char *username, gid_ return -ENOMEM; if (keep_groups) { - if ((k = getgroups(ngroups_max, gids)) < 0) { + k = getgroups(ngroups_max, gids); + if (k < 0) { free(gids); return -errno; } @@ -682,9 +781,10 @@ static int enforce_groups(const ExecContext *context, const char *username, gid_ k++; } - if (setgroups(k, gids) < 0) { + r = maybe_setgroups(k, gids); + if (r < 0) { free(gids); - return -errno; + return r; } free(gids); @@ -696,15 +796,10 @@ static int enforce_groups(const ExecContext *context, const char *username, gid_ static int enforce_user(const ExecContext *context, uid_t uid) { assert(context); - /* Sets (but doesn't lookup) the uid and make sure we keep the + /* Sets (but doesn't look up) the uid and make sure we keep the * capabilities while doing so. */ - if (context->capabilities) { - _cleanup_cap_free_ cap_t d = NULL; - static const cap_value_t bits[] = { - CAP_SETUID, /* Necessary so that we can run setresuid() below */ - CAP_SETPCAP /* Necessary so that we can set PR_SET_SECUREBITS later on */ - }; + 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 @@ -716,23 +811,9 @@ static int enforce_user(const ExecContext *context, uid_t uid) { if (prctl(PR_SET_SECUREBITS, sb) < 0) return -errno; } - - /* Second step: set the capabilities. This will reduce - * the capabilities to the minimum we need. */ - - d = cap_dup(context->capabilities); - if (!d) - return -errno; - - if (cap_set_flag(d, CAP_EFFECTIVE, ELEMENTSOF(bits), bits, CAP_SET) < 0 || - cap_set_flag(d, CAP_PERMITTED, ELEMENTSOF(bits), bits, CAP_SET) < 0) - return -errno; - - if (cap_set_proc(d) < 0) - return -errno; } - /* Third step: actually set the uids */ + /* Second step: actually set the uids */ if (setresuid(uid, uid, uid) < 0) return -errno; @@ -757,31 +838,36 @@ static int null_conv( 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 ***pam_env, + 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; - int err; - char **e = NULL; + 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(pam_env); + 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 @@ -790,6 +876,10 @@ static int setup_pam( * 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; @@ -805,6 +895,12 @@ static int setup_pam( 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; @@ -829,15 +925,17 @@ static int setup_pam( parent_pid = getpid(); pam_pid = fork(); - if (pam_pid < 0) + if (pam_pid < 0) { + r = -errno; goto fail; + } if (pam_pid == 0) { - int sig; - int r = EXIT_PAM; + 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 */ @@ -852,8 +950,14 @@ static int setup_pam( * 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_error_errno(r, "Error: Failed to setresuid() in sd-pam: %m"); + log_warning_errno(errno, "Failed to setresuid() in sd-pam: %m"); (void) ignore_signals(SIGPIPE, -1); @@ -865,6 +969,11 @@ static int setup_pam( 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) { @@ -893,13 +1002,15 @@ static int setup_pam( goto child_finish; } - r = 0; + ret = 0; child_finish: pam_end(handle, pam_code | flags); - _exit(r); + _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; @@ -911,19 +1022,22 @@ static int setup_pam( * might have opened it, but we don't want this fd around. */ closelog(); - *pam_env = e; - e = NULL; + /* 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)); - err = -EPERM; /* PAM errors do not map to errno */ - } else { - log_error_errno(errno, "PAM failed: %m"); - err = -errno; - } + r = -EPERM; /* PAM errors do not map to errno */ + } else + log_error_errno(r, "PAM failed: %m"); if (handle) { if (close_session) @@ -933,17 +1047,13 @@ fail: } strv_free(e); - closelog(); - if (pam_pid > 1) { - kill(pam_pid, SIGTERM); - kill(pam_pid, SIGCONT); - } - - return err; -} + return r; +#else + return 0; #endif +} static void rename_process_from_path(const char *path) { char process_name[11]; @@ -978,7 +1088,17 @@ static void rename_process_from_path(const char *path) { #ifdef HAVE_SECCOMP -static int apply_seccomp(const ExecContext *c) { +static bool skip_seccomp_unavailable(const Unit* u, const char* msg) { + if (!is_seccomp_available()) { + log_open(); + log_unit_debug(u, "SECCOMP features not detected in the kernel, skipping %s", msg); + log_close(); + return true; + } + return false; +} + +static int apply_seccomp(const Unit* u, const ExecContext *c) { uint32_t negative_action, action; scmp_filter_ctx *seccomp; Iterator i; @@ -987,6 +1107,9 @@ static int apply_seccomp(const ExecContext *c) { 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); @@ -1027,13 +1150,16 @@ finish: return r; } -static int apply_address_families(const ExecContext *c) { +static int apply_address_families(const Unit* u, const ExecContext *c) { scmp_filter_ctx *seccomp; Iterator i; int r; assert(c); + if (skip_seccomp_unavailable(u, "RestrictAddressFamilies=")) + return 0; + seccomp = seccomp_init(SCMP_ACT_ALLOW); if (!seccomp) return -ENOMEM; @@ -1148,14 +1274,246 @@ finish: 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; + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return -ENOMEM; + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) + goto finish; + + 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_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_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; + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return -ENOMEM; + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) + goto finish; + + /* 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_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_protect_sysctl(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; + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return -ENOMEM; + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) + goto finish; + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + SCMP_SYS(_sysctl), + 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_private_devices(Unit *u, const ExecContext *c) { + const SystemCallFilterSet *set; + scmp_filter_ctx *seccomp; + const char *sys; + bool syscalls_found = false; + int r; + + assert(c); + + /* If PrivateDevices= is set, also turn off iopl and all @raw-io syscalls. */ + + if (skip_seccomp_unavailable(u, "PrivateDevices=")) + return 0; + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return -ENOMEM; + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) + goto finish; + + for (set = syscall_filter_sets; set->set_name; set++) + if (streq(set->set_name, "@raw-io")) { + syscalls_found = true; + break; + } + + /* We should never fail here */ + if (!syscalls_found) { + r = -EOPNOTSUPP; + goto finish; + } + + NULSTR_FOREACH(sys, set->value) { + int id; + bool add = true; + +#ifndef __NR_s390_pci_mmio_read + if (streq(sys, "s390_pci_mmio_read")) + add = false; +#endif +#ifndef __NR_s390_pci_mmio_write + if (streq(sys, "s390_pci_mmio_write")) + add = false; +#endif + + if (!add) + continue; + + id = seccomp_syscall_resolve_name(sys); + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EPERM), + id, 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; +} + #endif static void do_idle_pipe_dance(int idle_pipe[4]) { assert(idle_pipe); - - safe_close(idle_pipe[1]); - safe_close(idle_pipe[2]); + idle_pipe[1] = safe_close(idle_pipe[1]); + idle_pipe[2] = safe_close(idle_pipe[2]); if (idle_pipe[0] >= 0) { int r; @@ -1163,41 +1521,49 @@ static void do_idle_pipe_dance(int idle_pipe[4]) { 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. */ - r = write(idle_pipe[3], "x", 1); - if (r > 0) + 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); } - safe_close(idle_pipe[0]); + idle_pipe[0] = safe_close(idle_pipe[0]); } - safe_close(idle_pipe[3]); + idle_pipe[3] = safe_close(idle_pipe[3]); } static int build_environment( + Unit *u, const ExecContext *c, + const ExecParameters *p, unsigned n_fds, - usec_t watchdog_usec, 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*, 10); + 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; @@ -1205,14 +1571,33 @@ static int build_environment( 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 (watchdog_usec > 0) { + 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, watchdog_usec) < 0) + 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; } @@ -1243,19 +1628,42 @@ static int build_environment( our_env[n_env++] = x; } - if (is_terminal_input(c->std_input) || - c->std_output == EXEC_OUTPUT_TTY || - c->std_error == EXEC_OUTPUT_TTY || - c->tty_path) { + 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); - x = strdup(default_term_for_tty(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 <= 10); + assert(n_env <= 12); *ret = our_env; our_env = NULL; @@ -1263,6 +1671,34 @@ static int build_environment( 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, @@ -1271,9 +1707,9 @@ static bool exec_needs_mount_namespace( assert(context); assert(params); - if (!strv_isempty(context->read_write_dirs) || - !strv_isempty(context->read_only_dirs) || - !strv_isempty(context->inaccessible_dirs)) + 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) @@ -1282,38 +1718,412 @@ static bool exec_needs_mount_namespace( if (context->private_tmp && runtime && (runtime->tmp_dir || runtime->var_tmp_dir)) return true; - if (params->bus_endpoint_path) - return true; - if (context->private_devices || context->protect_system != PROTECT_SYSTEM_NO || - context->protect_home != PROTECT_HOME_NO) + context->protect_home != PROTECT_HOME_NO || + context->protect_kernel_tunables || + 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); /* The case where the above is the same */ + else + uid_map = strdup("0 0 1\n"); + 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 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 || + 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 *fds, unsigned n_fds, char **files_env, + int user_lookup_fd, int *exit_status) { - _cleanup_strv_free_ char **our_env = NULL, **pam_env = NULL, **final_env = NULL, **final_argv = NULL; + _cleanup_strv_free_ char **our_env = NULL, **pass_env = NULL, **accum_env = NULL, **final_argv = NULL; _cleanup_free_ char *mac_selinux_context_net = NULL; - const char *username = NULL, *home = NULL, *shell = NULL; - unsigned n_dont_close = 0; - int dont_close[n_fds + 4]; + const char *username = NULL, *home = NULL, *shell = NULL, *wd; + 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; - bool needs_mount_namespace; assert(unit); assert(command); @@ -1349,22 +2159,7 @@ static int exec_child( log_forget_fds(); - 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 (params->bus_endpoint_fd >= 0) - dont_close[n_dont_close++] = params->bus_endpoint_fd; - if (runtime) { - if (runtime->netns_storage_socket[0] >= 0) - dont_close[n_dont_close++] = runtime->netns_storage_socket[0]; - if (runtime->netns_storage_socket[1] >= 0) - dont_close[n_dont_close++] = runtime->netns_storage_socket[1]; - } - - r = close_all_fds(dont_close, n_dont_close); + r = close_remaining_fds(params, runtime, dcreds, user_lookup_fd, socket_fd, fds, n_fds); if (r < 0) { *exit_status = EXIT_FDS; return r; @@ -1376,9 +2171,9 @@ static int exec_child( return -errno; } - exec_context_tty_reset(context); + exec_context_tty_reset(context, params); - if (params->confirm_spawn) { + if (params->flags & EXEC_CONFIRM_SPAWN) { char response; r = ask_for_confirmation(&response, argv); @@ -1397,33 +2192,78 @@ static int exec_child( } } - if (context->user) { - username = context->user; - r = get_user_creds(&username, &uid, &gid, &home, &shell); + 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 { + if (context->user) { + username = context->user; + r = get_user_creds_clean(&username, &uid, &gid, &home, &shell); + if (r < 0) { + *exit_status = EXIT_USER; + return r; + } + + /* Note that we don't set $HOME or $SHELL if they are not particularly enlightening anyway + * (i.e. are "/" or "/bin/nologin"). */ + } + + if (context->group) { + const char *g = context->group; + + r = get_group_creds(&g, &gid); + 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) - fd_nonblock(socket_fd, false); + (void) fd_nonblock(socket_fd, false); - r = setup_input(context, socket_fd, params->apply_tty_stdin); + r = setup_input(context, params, socket_fd); if (r < 0) { *exit_status = EXIT_STDIN; return r; } - r = setup_output(unit, context, STDOUT_FILENO, socket_fd, basename(command->path), params->apply_tty_stdin, uid, gid); + r = setup_output(unit, context, params, STDOUT_FILENO, socket_fd, 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, STDERR_FILENO, socket_fd, basename(command->path), params->apply_tty_stdin, uid, gid); + r = setup_output(unit, context, params, STDERR_FILENO, socket_fd, basename(command->path), uid, gid, &journal_stream_dev, &journal_stream_ino); if (r < 0) { *exit_status = EXIT_STDERR; return r; @@ -1504,9 +2344,13 @@ static int exec_child( } if (context->utmp_id) - utmp_put_init_process(context->utmp_id, getpid(), getsid(0), context->tty_path); + 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 && is_terminal_input(context->std_input)) { + if (context->user) { r = chown_terminal(STDIN_FILENO, uid); if (r < 0) { *exit_status = EXIT_STDIN; @@ -1514,16 +2358,6 @@ static int exec_child( } } - if (params->bus_endpoint_fd >= 0 && context->bus_endpoint) { - uid_t ep_uid = (uid == UID_INVALID) ? 0 : uid; - - r = bus_kernel_set_endpoint_policy(params->bus_endpoint_fd, ep_uid, context->bus_endpoint); - if (r < 0) { - *exit_status = EXIT_BUS_ENDPOINT; - 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. */ @@ -1543,44 +2377,65 @@ static int exec_child( } if (!strv_isempty(context->runtime_directory) && params->runtime_prefix) { - char **rt; - - STRV_FOREACH(rt, context->runtime_directory) { - _cleanup_free_ char *p; + r = setup_runtime_directory(context, params, uid, gid); + if (r < 0) { + *exit_status = EXIT_RUNTIME_DIRECTORY; + return r; + } + } - p = strjoin(params->runtime_prefix, "/", *rt, NULL); - if (!p) { - *exit_status = EXIT_RUNTIME_DIRECTORY; - return -ENOMEM; - } + 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 = mkdir_safe_label(p, context->runtime_directory_mode, uid, gid); - if (r < 0) { - *exit_status = EXIT_RUNTIME_DIRECTORY; - return r; - } - } + r = build_pass_environment(context, &pass_env); + if (r < 0) { + *exit_status = EXIT_MEMORY; + return r; } - if (params->apply_permissions) { - r = enforce_groups(context, username, gid); - if (r < 0) { - *exit_status = EXIT_GROUP; - 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); - umask(context->umask); + (void) umask(context->umask); -#ifdef HAVE_PAM - if (params->apply_permissions && context->pam_name && username) { - r = setup_pam(context->pam_name, username, uid, context->tty_path, &pam_env, fds, n_fds); + if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { + r = setup_smack(context, command); if (r < 0) { - *exit_status = EXIT_PAM; + *exit_status = EXIT_SMACK_PROCESS_LABEL; return r; } + + 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; + } + } } -#endif if (context->private_network && runtime && runtime->netns_storage_socket[0] >= 0) { r = setup_netns(runtime->netns_storage_socket); @@ -1591,8 +2446,8 @@ static int exec_child( } needs_mount_namespace = exec_needs_mount_namespace(context, params, runtime); - if (needs_mount_namespace) { + _cleanup_free_ char **rw = NULL; char *tmp = NULL, *var = NULL; /* The runtime struct only contains the parent @@ -1608,15 +2463,22 @@ static int exec_child( var = strjoina(runtime->var_tmp_dir, "/tmp"); } + r = compile_read_write_paths(context, params, &rw); + if (r < 0) { + *exit_status = EXIT_NAMESPACE; + return r; + } + r = setup_namespace( - params->apply_chroot ? context->root_directory : NULL, - context->read_write_dirs, - context->read_only_dirs, - context->inaccessible_dirs, + (params->flags & EXEC_APPLY_CHROOT) ? context->root_directory : NULL, + rw, + context->read_only_paths, + context->inaccessible_paths, tmp, var, - params->bus_endpoint_path, context->private_devices, + context->protect_kernel_tunables, + context->protect_control_groups, context->protect_home, context->protect_system, context->mount_flags); @@ -1634,28 +2496,37 @@ static int exec_child( } } - if (params->apply_chroot) { + if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) { + r = enforce_groups(context, username, gid); + if (r < 0) { + *exit_status = EXIT_GROUP; + return r; + } + } + + 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_namespace && context->root_directory) if (chroot(context->root_directory) < 0) { *exit_status = EXIT_CHROOT; return -errno; } - if (chdir(context->working_directory ?: "/") < 0 && + if (chdir(wd) < 0 && !context->working_directory_missing_ok) { *exit_status = EXIT_CHDIR; return -errno; } } else { - _cleanup_free_ char *d = NULL; - - if (asprintf(&d, "%s/%s", - context->root_directory ?: "", - context->working_directory ?: "") < 0) { - *exit_status = EXIT_MEMORY; - return -ENOMEM; - } + const char *d; + d = strjoina(strempty(context->root_directory), "/", strempty(wd)); if (chdir(d) < 0 && !context->working_directory_missing_ok) { *exit_status = EXIT_CHDIR; @@ -1664,7 +2535,12 @@ static int exec_child( } #ifdef HAVE_SELINUX - if (params->apply_permissions && mac_selinux_use() && params->selinux_context_net && socket_fd >= 0) { + 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; @@ -1673,6 +2549,14 @@ static int exec_child( } #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 @@ -1689,44 +2573,47 @@ static int exec_child( return r; } - if (params->apply_permissions) { + 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; - if (setrlimit_closest(i, context->rlimit[i]) < 0) { + r = setrlimit_closest(i, context->rlimit[i]); + if (r < 0) { *exit_status = EXIT_LIMITS; - return -errno; + return r; } } - if (context->capability_bounding_set_drop) { - r = capability_bounding_set_drop(context->capability_bounding_set_drop, false); - if (r < 0) { - *exit_status = EXIT_CAPABILITIES; - 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; } } -#ifdef HAVE_SMACK - if (context->smack_process_label) { - r = mac_smack_apply_pid(0, context->smack_process_label); + if (!cap_test_all(context->capability_bounding_set)) { + r = capability_bounding_set_drop(context->capability_bounding_set, false); if (r < 0) { - *exit_status = EXIT_SMACK_PROCESS_LABEL; + *exit_status = EXIT_CAPABILITIES; return r; } } -#ifdef SMACK_DEFAULT_PROCESS_LABEL - else { - r = mac_smack_apply_pid(0, SMACK_DEFAULT_PROCESS_LABEL); + + /* 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_SMACK_PROCESS_LABEL; + *exit_status = EXIT_CAPABILITIES; return r; } } -#endif -#endif if (context->user) { r = enforce_user(context, uid); @@ -1734,44 +2621,85 @@ static int exec_child( *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; + } } /* 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) != context->secure_bits) - if (prctl(PR_SET_SECUREBITS, context->secure_bits) < 0) { + if (prctl(PR_GET_SECUREBITS) != secure_bits) + if (prctl(PR_SET_SECUREBITS, secure_bits) < 0) { *exit_status = EXIT_SECUREBITS; return -errno; } - if (context->capabilities) - if (cap_set_proc(context->capabilities) < 0) { - *exit_status = EXIT_CAPABILITIES; - return -errno; - } - - if (context->no_new_privileges) + 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->address_families_whitelist || - !set_isempty(context->address_families)) { - r = apply_address_families(context); + if (context_has_address_families(context)) { + r = apply_address_families(unit, context); if (r < 0) { *exit_status = EXIT_ADDRESS_FAMILIES; return r; } } - if (context->syscall_whitelist || - !set_isempty(context->syscall_filter) || - !set_isempty(context->syscall_archs)) { - r = apply_seccomp(context); + 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->private_devices) { + r = apply_private_devices(unit, context); + if (r < 0) { + *exit_status = EXIT_SECCOMP; + return r; + } + } + + if (context_has_syscall_filters(context)) { + r = apply_seccomp(unit, context); if (r < 0) { *exit_status = EXIT_SECCOMP; return r; @@ -1804,32 +2732,12 @@ static int exec_child( #endif } - r = build_environment(context, n_fds, params->watchdog_usec, home, username, shell, &our_env); - if (r < 0) { - *exit_status = EXIT_MEMORY; - return r; - } - - final_env = strv_env_merge(5, - params->environment, - our_env, - context->environment, - files_env, - pam_env, - NULL); - if (!final_env) { - *exit_status = EXIT_MEMORY; - return -ENOMEM; - } - - final_argv = replace_env_argv(argv, final_env); + final_argv = replace_env_argv(argv, accum_env); if (!final_argv) { *exit_status = EXIT_MEMORY; return -ENOMEM; } - final_env = strv_env_clean(final_env); - if (_unlikely_(log_get_max_level() >= LOG_DEBUG)) { _cleanup_free_ char *line; @@ -1845,7 +2753,7 @@ static int exec_child( } } - execve(command->path, final_argv, final_env); + execve(command->path, final_argv, accum_env); *exit_status = EXIT_EXEC; return -errno; } @@ -1855,6 +2763,7 @@ int exec_spawn(Unit *unit, const ExecContext *context, const ExecParameters *params, ExecRuntime *runtime, + DynamicCreds *dcreds, pid_t *ret) { _cleanup_strv_free_ char **files_env = NULL; @@ -1903,7 +2812,7 @@ int exec_spawn(Unit *unit, NULL); pid = fork(); if (pid < 0) - return log_unit_error_errno(unit, r, "Failed to fork: %m"); + return log_unit_error_errno(unit, errno, "Failed to fork: %m"); if (pid == 0) { int exit_status; @@ -1913,10 +2822,12 @@ int exec_spawn(Unit *unit, context, params, runtime, + dcreds, argv, socket_fd, fds, n_fds, files_env, + unit->manager->user_lookup_fds[1], &exit_status); if (r < 0) { log_open(); @@ -1961,6 +2872,7 @@ void exec_context_init(ExecContext *c) { 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) { @@ -1968,80 +2880,40 @@ void exec_context_done(ExecContext *c) { assert(c); - strv_free(c->environment); - c->environment = NULL; - - strv_free(c->environment_files); - c->environment_files = NULL; - - for (l = 0; l < ELEMENTSOF(c->rlimit); l++) { - free(c->rlimit[l]); - c->rlimit[l] = NULL; - } - - free(c->working_directory); - c->working_directory = NULL; - free(c->root_directory); - c->root_directory = NULL; - - free(c->tty_path); - c->tty_path = NULL; + c->environment = strv_free(c->environment); + c->environment_files = strv_free(c->environment_files); + c->pass_environment = strv_free(c->pass_environment); - free(c->syslog_identifier); - c->syslog_identifier = NULL; + for (l = 0; l < ELEMENTSOF(c->rlimit); l++) + c->rlimit[l] = mfree(c->rlimit[l]); - free(c->user); - c->user = NULL; + 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); - free(c->group); - c->group = NULL; + c->supplementary_groups = strv_free(c->supplementary_groups); - strv_free(c->supplementary_groups); - c->supplementary_groups = NULL; + c->pam_name = mfree(c->pam_name); - free(c->pam_name); - c->pam_name = NULL; - - if (c->capabilities) { - cap_free(c->capabilities); - c->capabilities = NULL; - } - - strv_free(c->read_only_dirs); - c->read_only_dirs = NULL; - - strv_free(c->read_write_dirs); - c->read_write_dirs = NULL; - - strv_free(c->inaccessible_dirs); - c->inaccessible_dirs = NULL; + 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); - free(c->utmp_id); - c->utmp_id = NULL; - - free(c->selinux_context); - c->selinux_context = NULL; - - free(c->apparmor_profile); - c->apparmor_profile = NULL; + c->utmp_id = mfree(c->utmp_id); + c->selinux_context = mfree(c->selinux_context); + c->apparmor_profile = mfree(c->apparmor_profile); - set_free(c->syscall_filter); - c->syscall_filter = NULL; + c->syscall_filter = set_free(c->syscall_filter); + c->syscall_archs = set_free(c->syscall_archs); + c->address_families = set_free(c->address_families); - set_free(c->syscall_archs); - c->syscall_archs = NULL; - - set_free(c->address_families); - c->address_families = NULL; - - strv_free(c->runtime_directory); - c->runtime_directory = NULL; - - bus_endpoint_free(c->bus_endpoint); - c->bus_endpoint = NULL; + c->runtime_directory = strv_free(c->runtime_directory); } int exec_context_destroy_runtime_directory(ExecContext *c, const char *runtime_prefix) { @@ -2071,11 +2943,9 @@ int exec_context_destroy_runtime_directory(ExecContext *c, const char *runtime_p void exec_command_done(ExecCommand *c) { assert(c); - free(c->path); - c->path = NULL; + c->path = mfree(c->path); - strv_free(c->argv); - c->argv = NULL; + c->argv = strv_free(c->argv); } void exec_command_done_array(ExecCommand *c, unsigned n) { @@ -2133,7 +3003,7 @@ int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l) { if (fn[0] == '-') { ignore = true; - fn ++; + fn++; } if (!path_is_absolute(fn)) { @@ -2151,7 +3021,7 @@ int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l) { continue; strv_free(r); - return errno ? -errno : -EINVAL; + return errno > 0 ? -errno : -EINVAL; } count = pglob.gl_pathc; if (count == 0) { @@ -2203,7 +3073,10 @@ int exec_context_load_environment(Unit *unit, const ExecContext *c, char ***l) { static bool tty_may_match_dev_console(const char *tty) { _cleanup_free_ char *active = NULL; - char *console; + char *console; + + if (!tty) + return true; if (startswith(tty, "/dev/")) tty += 5; @@ -2222,11 +3095,14 @@ static bool tty_may_match_dev_console(const char *tty) { } bool exec_context_may_touch_console(ExecContext *ec) { - return (ec->tty_reset || ec->tty_vhangup || ec->tty_vt_disallocate || + + 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(tty_path(ec)); + tty_may_match_dev_console(exec_context_tty_path(ec)); } static void strv_fprintf(FILE *f, char **l) { @@ -2239,7 +3115,7 @@ static void strv_fprintf(FILE *f, char **l) { } void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { - char **e; + char **e, **d; unsigned i; assert(c); @@ -2253,21 +3129,31 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { "%sRootDirectory: %s\n" "%sNonBlocking: %s\n" "%sPrivateTmp: %s\n" - "%sPrivateNetwork: %s\n" "%sPrivateDevices: %s\n" + "%sProtectKernelTunables: %s\n" + "%sProtectControlGroups: %s\n" + "%sPrivateNetwork: %s\n" + "%sPrivateUsers: %s\n" "%sProtectHome: %s\n" "%sProtectSystem: %s\n" - "%sIgnoreSIGPIPE: %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_network), prefix, yes_no(c->private_devices), + prefix, yes_no(c->protect_kernel_tunables), + 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->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); @@ -2275,6 +3161,14 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { 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", @@ -2286,9 +3180,12 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { prefix, c->oom_score_adjust); for (i = 0; i < RLIM_NLIMITS; i++) - if (c->rlimit[i]) - fprintf(f, "%s%s: "RLIM_FMT"\n", + 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; @@ -2369,14 +3266,6 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { prefix, strna(lvl_str)); } - if (c->capabilities) { - _cleanup_cap_free_charp_ char *t; - - t = cap_to_text(c->capabilities, NULL); - if (t) - fprintf(f, "%sCapabilities: %s\n", prefix, t); - } - if (c->secure_bits) fprintf(f, "%sSecure Bits:%s%s%s%s%s%s\n", prefix, @@ -2387,12 +3276,23 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { (c->secure_bits & 1<<SECURE_NOROOT) ? " noroot" : "", (c->secure_bits & 1<<SECURE_NOROOT_LOCKED) ? "noroot-locked" : ""); - if (c->capability_bounding_set_drop) { + 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_drop & ((uint64_t) 1ULL << (uint64_t) 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); @@ -2403,6 +3303,8 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { 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); @@ -2412,21 +3314,21 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { if (c->pam_name) fprintf(f, "%sPAMName: %s\n", prefix, c->pam_name); - if (strv_length(c->read_write_dirs) > 0) { - fprintf(f, "%sReadWriteDirs:", prefix); - strv_fprintf(f, c->read_write_dirs); + 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_dirs) > 0) { - fprintf(f, "%sReadOnlyDirs:", prefix); - strv_fprintf(f, c->read_only_dirs); + 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_dirs) > 0) { - fprintf(f, "%sInaccessibleDirs:", prefix); - strv_fprintf(f, c->inaccessible_dirs); + if (strv_length(c->inaccessible_paths) > 0) { + fprintf(f, "%sInaccessiblePaths:", prefix); + strv_fprintf(f, c->inaccessible_paths); fputs("\n", f); } @@ -2493,7 +3395,7 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { fputc('\n', f); } - if (c->syscall_errno != 0) + if (c->syscall_errno > 0) fprintf(f, "%sSystemCallErrorNumber: %s\n", prefix, strna(errno_to_name(c->syscall_errno))); @@ -2507,7 +3409,7 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) { bool exec_context_maintains_privileges(ExecContext *c) { assert(c); - /* Returns true if the process forked off would run run under + /* Returns true if the process forked off would run under * an unchanged UID or as root. */ if (!c->user) @@ -2543,7 +3445,7 @@ void exec_status_exit(ExecStatus *s, ExecContext *context, pid_t pid, int code, if (context->utmp_id) utmp_put_dead_process(context->utmp_id, pid, code, status); - exec_context_tty_reset(context); + exec_context_tty_reset(context, NULL); } } @@ -2562,12 +3464,12 @@ void exec_status_dump(ExecStatus *s, FILE *f, const char *prefix) { "%sPID: "PID_FMT"\n", prefix, s->pid); - if (s->start_timestamp.realtime > 0) + 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 (s->exit_timestamp.realtime > 0) + if (dual_timestamp_is_set(&s->exit_timestamp)) fprintf(f, "%sExit Timestamp: %s\n" "%sExit Code: %s\n" @@ -2701,7 +3603,7 @@ int exec_command_append(ExecCommand *c, const char *path, ...) { if (!l) return -ENOMEM; - r = strv_extend_strv(&c->argv, l); + r = strv_extend_strv(&c->argv, l, false); if (r < 0) return r; @@ -2742,7 +3644,7 @@ int exec_runtime_make(ExecRuntime **rt, ExecContext *c, const char *id) { return r; if (c->private_network && (*rt)->netns_storage_socket[0] < 0) { - if (socketpair(AF_UNIX, SOCK_DGRAM, 0, (*rt)->netns_storage_socket) < 0) + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, (*rt)->netns_storage_socket) < 0) return -errno; } @@ -2954,3 +3856,11 @@ static const char* const exec_output_table[_EXEC_OUTPUT_MAX] = { }; 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/core/execute.h b/src/core/execute.h index f5d5c1dee7..449180c903 100644 --- a/src/core/execute.h +++ b/src/core/execute.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -27,16 +25,24 @@ typedef struct ExecContext ExecContext; typedef struct ExecRuntime ExecRuntime; typedef struct ExecParameters ExecParameters; -#include <sys/capability.h> +#include <sched.h> #include <stdbool.h> #include <stdio.h> -#include <sched.h> +#include <sys/capability.h> -#include "list.h" +#include "cgroup-util.h" #include "fdset.h" +#include "list.h" #include "missing.h" #include "namespace.h" -#include "bus-endpoint.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, @@ -76,7 +82,8 @@ struct ExecCommand { char **argv; ExecStatus exec_status; LIST_FIELDS(ExecCommand, command); /* useful for chaining commands */ - bool ignore; + bool ignore:1; + bool privileged:1; }; struct ExecRuntime { @@ -85,16 +92,20 @@ struct ExecRuntime { 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; @@ -112,6 +123,8 @@ struct ExecContext { nsec_t timer_slack_nsec; + bool stdio_as_fds; + char *tty_path; bool tty_reset; @@ -120,7 +133,7 @@ struct ExecContext { bool ignore_sigpipe; - /* Since resolving these names might might involve socket + /* 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. */ @@ -131,6 +144,7 @@ struct ExecContext { char *pam_name; char *utmp_id; + ExecUtmpMode utmp_mode; bool selinux_context_ignore; char *selinux_context; @@ -141,12 +155,11 @@ struct ExecContext { bool smack_process_label_ignore; char *smack_process_label; - char **read_write_dirs, **read_only_dirs, **inaccessible_dirs; + char **read_write_paths, **read_only_paths, **inaccessible_paths; unsigned long mount_flags; - uint64_t capability_bounding_set_drop; - - cap_t capabilities; + uint64_t capability_bounding_set; + uint64_t capability_ambient_set; int secure_bits; int syslog_priority; @@ -158,11 +171,17 @@ struct ExecContext { bool private_tmp; bool private_network; bool private_devices; + bool private_users; ProtectSystem protect_system; ProtectHome protect_home; + bool protect_kernel_tunables; + 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 @@ -183,43 +202,64 @@ struct ExecContext { 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; - - /* custom dbus enpoint */ - BusEndpoint *bus_endpoint; }; -#include "cgroup.h" -#include "cgroup-util.h" +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; - int *fds; unsigned n_fds; char **environment; - bool apply_permissions; - bool apply_chroot; - bool apply_tty_stdin; - bool confirm_spawn; - bool selinux_context_net; - CGroupControllerMask cgroup_supported; + + 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; - bool cgroup_delegate; + const char *runtime_prefix; + usec_t watchdog_usec; + int *idle_pipe; - char *bus_endpoint_path; - int bus_endpoint_fd; + + int stdin_fd; + int stdout_fd; + int stderr_fd; }; +#include "unit.h" +#include "dynamic-user.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); @@ -265,3 +305,6 @@ 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/core/failure-action.c b/src/core/failure-action.c index b06a7d2ae5..ddae46190f 100644 --- a/src/core/failure-action.c +++ b/src/core/failure-action.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -23,16 +21,17 @@ #include <sys/reboot.h> #include <linux/reboot.h> -#include "bus-util.h" #include "bus-error.h" -#include "special.h" +#include "bus-util.h" #include "failure-action.h" +#include "special.h" +#include "string-table.h" #include "terminal-util.h" static void log_and_status(Manager *m, const char *message) { log_warning("%s", message); manager_status_printf(m, STATUS_TYPE_EMERGENCY, - ANSI_HIGHLIGHT_RED_ON " !! " ANSI_HIGHLIGHT_OFF, + ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, "%s", message); } @@ -41,8 +40,6 @@ int failure_action( FailureAction action, const char *reboot_arg) { - int r; - assert(m); assert(action >= 0); assert(action < _FAILURE_ACTION_MAX); @@ -50,7 +47,7 @@ int failure_action( if (action == FAILURE_ACTION_NONE) return -ECANCELED; - if (m->running_as == MANAGER_USER) { + if (!MANAGER_IS_SYSTEM(m)) { /* Downgrade all options to simply exiting if we run * in user mode */ @@ -61,24 +58,20 @@ int failure_action( switch (action) { - case FAILURE_ACTION_REBOOT: { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - + case FAILURE_ACTION_REBOOT: log_and_status(m, "Rebooting as result of failure."); - update_reboot_param_file(reboot_arg); - r = manager_add_job_by_name(m, JOB_START, SPECIAL_REBOOT_TARGET, JOB_REPLACE, true, &error, NULL); - if (r < 0) - log_error("Failed to reboot: %s.", bus_error_message(&error, r)); + (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 FAILURE_ACTION_REBOOT_FORCE: log_and_status(m, "Forcibly rebooting as result of failure."); - update_reboot_param_file(reboot_arg); + (void) update_reboot_parameter_and_warn(reboot_arg); m->exit_code = MANAGER_REBOOT; + break; case FAILURE_ACTION_REBOOT_IMMEDIATE: @@ -86,26 +79,20 @@ int failure_action( sync(); - if (reboot_arg) { + 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 FAILURE_ACTION_POWEROFF: { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - + case FAILURE_ACTION_POWEROFF: log_and_status(m, "Powering off as result of failure."); - - r = manager_add_job_by_name(m, JOB_START, SPECIAL_POWEROFF_TARGET, JOB_REPLACE, true, &error, NULL); - if (r < 0) - log_error("Failed to poweroff: %s.", bus_error_message(&error, r)); - + (void) manager_add_job_by_name_and_warn(m, JOB_START, SPECIAL_POWEROFF_TARGET, JOB_REPLACE_IRREVERSIBLY, NULL); break; - } case FAILURE_ACTION_POWEROFF_FORCE: log_and_status(m, "Forcibly powering off as result of failure."); diff --git a/src/core/failure-action.h b/src/core/failure-action.h index 1af4dd987b..1adac4ad5c 100644 --- a/src/core/failure-action.h +++ b/src/core/failure-action.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/hostname-setup.c b/src/core/hostname-setup.c index 932ddbf95a..68be52856b 100644 --- a/src/core/hostname-setup.c +++ b/src/core/hostname-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,16 +17,18 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <stdio.h> #include <errno.h> +#include <stdio.h> #include <stdlib.h> -#include "macro.h" -#include "util.h" -#include "log.h" +#include "alloc-util.h" #include "fileio.h" -#include "hostname-util.h" #include "hostname-setup.h" +#include "hostname-util.h" +#include "log.h" +#include "macro.h" +#include "string-util.h" +#include "util.h" int hostname_setup(void) { int r; @@ -59,8 +59,9 @@ int hostname_setup(void) { hn = "localhost"; } - if (sethostname_idempotent(hn) < 0) - return log_warning_errno(errno, "Failed to set hostname to <%s>: %m", hn); + 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/core/hostname-setup.h b/src/core/hostname-setup.h index 8dc3a9e1d8..73e8c75c71 100644 --- a/src/core/hostname-setup.h +++ b/src/core/hostname-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/ima-setup.c b/src/core/ima-setup.c index 42a3e97459..d1b0ce76ef 100644 --- a/src/core/ima-setup.c +++ b/src/core/ima-setup.c @@ -1,11 +1,9 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** 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 + 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 @@ -21,12 +19,14 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <unistd.h> #include <errno.h> +#include <unistd.h> +#include "fd-util.h" +#include "fileio.h" #include "ima-setup.h" -#include "util.h" #include "log.h" +#include "util.h" #define IMA_SECFS_DIR "/sys/kernel/security/ima" #define IMA_SECFS_POLICY IMA_SECFS_DIR "/policy" diff --git a/src/core/ima-setup.h b/src/core/ima-setup.h index 14b56d1fc2..472b58cb00 100644 --- a/src/core/ima-setup.h +++ b/src/core/ima-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -7,7 +5,7 @@ Copyright 2010 Lennart Poettering Copyright (C) 2012 Roberto Sassu - Politecnico di Torino, Italy - TORSEC group -- http://security.polito.it + 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 diff --git a/src/core/job.c b/src/core/job.c index 1448e5b69a..7faf2ef686 100644 --- a/src/core/job.c +++ b/src/core/job.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -23,17 +21,25 @@ #include "sd-id128.h" #include "sd-messages.h" -#include "set.h" -#include "unit.h" -#include "macro.h" -#include "strv.h" -#include "log.h" -#include "dbus-job.h" -#include "special.h" + +#include "alloc-util.h" #include "async.h" -#include "virt.h" +#include "dbus-job.h" #include "dbus.h" +#include "escape.h" +#include "job.h" +#include "log.h" +#include "macro.h" +#include "parse-util.h" +#include "set.h" +#include "special.h" +#include "stdio-util.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" #include "terminal-util.h" +#include "unit.h" +#include "virt.h" Job* job_new_raw(Unit *unit) { Job *j; @@ -131,7 +137,7 @@ void job_uninstall(Job *j) { /* Detach from next 'bigger' objects */ /* daemon-reload should be transparent to job observers */ - if (j->manager->n_reloading <= 0) + if (!MANAGER_IS_RELOADING(j->manager)) bus_job_send_removed_signal(j); *pj = NULL; @@ -168,7 +174,6 @@ static void job_merge_into_installed(Job *j, Job *other) { else assert(other->type == JOB_NOP); - j->override = j->override || other->override; j->irreversible = j->irreversible || other->irreversible; j->ignore_order = j->ignore_order || other->ignore_order; } @@ -186,7 +191,7 @@ Job* job_install(Job *j) { if (uj) { if (job_type_is_conflicting(uj->type, j->type)) - job_finish_and_invalidate(uj, JOB_CANCELED, false); + job_finish_and_invalidate(uj, JOB_CANCELED, false, false); else { /* not conflicting, i.e. mergeable */ @@ -217,7 +222,7 @@ Job* job_install(Job *j) { *pj = j; j->installed = true; - j->manager->n_installed_jobs ++; + 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); @@ -300,12 +305,10 @@ void job_dump(Job *j, FILE*f, const char *prefix) { "%s-> Job %u:\n" "%s\tAction: %s -> %s\n" "%s\tState: %s\n" - "%s\tForced: %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->override), prefix, yes_no(j->irreversible)); } @@ -400,6 +403,13 @@ JobType job_type_collapse(JobType t, Unit *u) { 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)) @@ -495,10 +505,58 @@ static void job_change_type(Job *j, JobType 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; - uint32_t id; - Manager *m = j->manager; assert(j); assert(j->installed); @@ -517,23 +575,9 @@ int job_run_and_invalidate(Job *j) { job_set_state(j, JOB_RUNNING); job_add_to_dbus_queue(j); - /* While we execute this operation the job might go away (for - * example: because it 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. */ - id = j->id; switch (j->type) { - case JOB_START: - r = unit_start(j->unit); - - /* If this unit cannot be started, then simply wait */ - if (r == -EBADR) - r = 0; - break; - case JOB_VERIFY_ACTIVE: { UnitActiveState t = unit_active_state(j->unit); if (UNIT_IS_ACTIVE_OR_RELOADING(t)) @@ -545,17 +589,19 @@ int job_run_and_invalidate(Job *j) { break; } + case JOB_START: case JOB_STOP: case JOB_RESTART: - r = unit_stop(j->unit); + r = job_perform_on_unit(&j); - /* If this unit cannot stopped, then simply wait. */ + /* If the unit type does not support starting/stopping, + * then simply wait. */ if (r == -EBADR) r = 0; break; case JOB_RELOAD: - r = unit_reload(j->unit); + r = job_perform_on_unit(&j); break; case JOB_NOP: @@ -566,183 +612,146 @@ int job_run_and_invalidate(Job *j) { assert_not_reached("Unknown job type"); } - j = manager_get_job(m, id); if (j) { if (r == -EALREADY) - r = job_finish_and_invalidate(j, JOB_DONE, true); + r = job_finish_and_invalidate(j, JOB_DONE, true, true); else if (r == -EBADR) - r = job_finish_and_invalidate(j, JOB_SKIPPED, true); + r = job_finish_and_invalidate(j, JOB_SKIPPED, true, false); else if (r == -ENOEXEC) - r = job_finish_and_invalidate(j, JOB_INVALID, true); + r = job_finish_and_invalidate(j, JOB_INVALID, true, false); else if (r == -EPROTO) - r = job_finish_and_invalidate(j, JOB_ASSERT, true); + r = job_finish_and_invalidate(j, JOB_ASSERT, true, false); else if (r == -EOPNOTSUPP) - r = job_finish_and_invalidate(j, JOB_UNSUPPORTED, true); + 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); + 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) { - const UnitStatusMessageFormats *format_table; - - assert(u); - assert(t >= 0); - assert(t < _JOB_TYPE_MAX); - - format_table = &UNIT_VTABLE(u)->status_message_formats; - if (!format_table) - return NULL; - if (t == JOB_START) - return format_table->finished_start_job[result]; - else if (t == JOB_STOP || t == JOB_RESTART) - return format_table->finished_stop_job[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.", + }; - return NULL; -} - -_pure_ static const char *job_get_status_message_format_try_harder(Unit *u, JobType t, JobResult result) { + const UnitStatusMessageFormats *format_table; const char *format; assert(u); assert(t >= 0); assert(t < _JOB_TYPE_MAX); - format = job_get_status_message_format(u, t, result); - if (format) - return format; + 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) { - if (result == JOB_DONE) - return "Started %s."; - else if (result == JOB_TIMEOUT) - return "Timed out starting %s."; - else if (result == JOB_FAILED) - return "Failed to start %s."; - else if (result == JOB_DEPENDENCY) - return "Dependency failed for %s."; - else if (result == JOB_ASSERT) - return "Assertion failed for %s."; - else if (result == JOB_UNSUPPORTED) - return "Starting of %s not supported."; - } else if (t == JOB_STOP || t == JOB_RESTART) { - if (result == JOB_DONE) - return "Stopped %s."; - else if (result == JOB_FAILED) - return "Stopped (with error) %s."; - else if (result == JOB_TIMEOUT) - return "Timed out stoppping %s."; - } else if (t == JOB_RELOAD) { - if (result == JOB_DONE) - return "Reloaded %s."; - else if (result == JOB_FAILED) - return "Reload failed for %s."; - else if (result == JOB_TIMEOUT) - return "Timed out reloading %s."; - } + 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 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); - DISABLE_WARNING_FORMAT_NONLITERAL; - - if (t == JOB_START) { - format = job_get_status_message_format(u, t, result); - if (!format) - return; - - switch (result) { - - case JOB_DONE: - if (u->condition_result) - unit_status_printf(u, ANSI_GREEN_ON " OK " ANSI_HIGHLIGHT_OFF, format); - break; - - case JOB_TIMEOUT: - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON " TIME " ANSI_HIGHLIGHT_OFF, format); - break; - - case JOB_FAILED: { - _cleanup_free_ char *quoted = NULL; - - quoted = shell_maybe_quote(u->id); - - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON "FAILED" ANSI_HIGHLIGHT_OFF, format); - manager_status_printf(u->manager, STATUS_TYPE_NORMAL, NULL, "See 'systemctl status %s' for details.", strna(quoted)); - break; - } - - case JOB_DEPENDENCY: - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_YELLOW_ON "DEPEND" ANSI_HIGHLIGHT_OFF, format); - break; - - case JOB_ASSERT: - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_YELLOW_ON "ASSERT" ANSI_HIGHLIGHT_OFF, format); - break; - - case JOB_UNSUPPORTED: - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_YELLOW_ON "UNSUPP" ANSI_HIGHLIGHT_OFF, format); - break; - - default: - ; - } - - } else if (t == JOB_STOP || t == JOB_RESTART) { - - format = job_get_status_message_format(u, t, result); - if (!format) - return; + /* Reload status messages have traditionally not been printed to console. */ + if (t == JOB_RELOAD) + return; - switch (result) { + format = job_get_status_message_format(u, t, result); + if (!format) + return; - case JOB_TIMEOUT: - manager_flip_auto_status(u->manager, true); - unit_status_printf(u, ANSI_HIGHLIGHT_RED_ON " TIME " ANSI_HIGHLIGHT_OFF, format); - break; + if (log_get_show_color()) + status = strjoina(statuses[result].color, statuses[result].word, ANSI_NORMAL); + else + status = statuses[result].word; - case JOB_DONE: - case JOB_FAILED: - unit_status_printf(u, ANSI_GREEN_ON " OK " ANSI_HIGHLIGHT_OFF, format); - break; + if (result != JOB_DONE) + manager_flip_auto_status(u->manager, true); - default: - ; - } + DISABLE_WARNING_FORMAT_NONLITERAL; + unit_status_printf(u, status, format); + REENABLE_WARNING; - } else if (t == JOB_VERIFY_ACTIVE) { + if (t == JOB_START && result == JOB_FAILED) { + _cleanup_free_ char *quoted; - /* 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. */ - if (result == JOB_SKIPPED) - unit_status_printf(u, ANSI_HIGHLIGHT_ON " INFO " ANSI_HIGHLIGHT_OFF, "%s is not active."); + quoted = shell_maybe_quote(u->id); + manager_status_printf(u->manager, STATUS_TYPE_NORMAL, NULL, "See 'systemctl status %s' for details.", strna(quoted)); } - - REENABLE_WARNING; } 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); @@ -754,40 +763,54 @@ static void job_log_status_message(Unit *u, JobType t, JobResult result) { if (log_on_console()) return; - format = job_get_status_message_format_try_harder(u, t, result); + format = job_get_status_message_format(u, t, result); if (!format) return; DISABLE_WARNING_FORMAT_NONLITERAL; - snprintf(buf, sizeof(buf), format, unit_description(u)); + xsprintf(buf, format, unit_description(u)); REENABLE_WARNING; - if (t == JOB_START) { - sd_id128_t mid; + switch (t) { + case JOB_START: mid = result == JOB_DONE ? SD_MESSAGE_UNIT_STARTED : SD_MESSAGE_UNIT_FAILED; - log_struct(result == JOB_DONE ? LOG_INFO : LOG_ERR, - LOG_MESSAGE_ID(mid), - LOG_UNIT_ID(u), - LOG_MESSAGE("%s", buf), - "RESULT=%s", job_result_to_string(result), - NULL); + break; - } else if (t == JOB_STOP) - log_struct(result == JOB_DONE ? LOG_INFO : LOG_ERR, - LOG_MESSAGE_ID(SD_MESSAGE_UNIT_STOPPED), - LOG_UNIT_ID(u), - LOG_MESSAGE("%s", buf), - "RESULT=%s", job_result_to_string(result), - NULL); + case JOB_RELOAD: + mid = SD_MESSAGE_UNIT_RELOADED; + break; - else if (t == JOB_RELOAD) - log_struct(result == JOB_DONE ? LOG_INFO : LOG_ERR, - LOG_MESSAGE_ID(SD_MESSAGE_UNIT_RELOADED), + 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) { @@ -804,11 +827,11 @@ static void job_fail_dependencies(Unit *u, UnitDependency d) { if (!IN_SET(j->type, JOB_START, JOB_VERIFY_ACTIVE)) continue; - job_finish_and_invalidate(j, JOB_DEPENDENCY, true); + job_finish_and_invalidate(j, JOB_DEPENDENCY, true, false); } } -int job_finish_and_invalidate(Job *j, JobResult result, bool recursive) { +int job_finish_and_invalidate(Job *j, JobResult result, bool recursive, bool already) { Unit *u; Unit *other; JobType t; @@ -825,8 +848,9 @@ int job_finish_and_invalidate(Job *j, JobResult result, bool recursive) { log_unit_debug(u, "Job %s/%s finished, result=%s", u->id, job_type_to_string(t), job_result_to_string(result)); - job_print_status_message(u, t, result); - job_log_status_message(u, t, 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); @@ -842,7 +866,7 @@ int job_finish_and_invalidate(Job *j, JobResult result, bool recursive) { } if (result == JOB_FAILED || result == JOB_INVALID) - j->manager->n_failed_jobs ++; + j->manager->n_failed_jobs++; job_uninstall(j); job_free(j); @@ -853,8 +877,6 @@ int job_finish_and_invalidate(Job *j, JobResult result, bool recursive) { job_fail_dependencies(u, UNIT_REQUIRED_BY); job_fail_dependencies(u, UNIT_REQUISITE_OF); job_fail_dependencies(u, UNIT_BOUND_BY); - job_fail_dependencies(u, UNIT_REQUIRED_BY_OVERRIDABLE); - job_fail_dependencies(u, UNIT_REQUISITE_OF_OVERRIDABLE); } else if (t == JOB_STOP) job_fail_dependencies(u, UNIT_CONFLICTED_BY); } @@ -903,7 +925,7 @@ static int job_dispatch_timer(sd_event_source *s, uint64_t monotonic, void *user 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); + job_finish_and_invalidate(j, JOB_TIMEOUT, true, false); failure_action(u->manager, u->job_timeout_action, u->job_timeout_reboot_arg); @@ -918,14 +940,14 @@ int job_start_timer(Job *j) { j->begin_usec = now(CLOCK_MONOTONIC); - if (j->unit->job_timeout <= 0) + if (j->unit->job_timeout == USEC_INFINITY) return 0; r = sd_event_add_time( j->manager->event, &j->timer_event_source, CLOCK_MONOTONIC, - j->begin_usec + j->unit->job_timeout, 0, + usec_add(j->begin_usec, j->unit->job_timeout), 0, job_dispatch_timer, j); if (r < 0) return r; @@ -975,11 +997,13 @@ char *job_dbus_path(Job *j) { return p; } -int job_serialize(Job *j, FILE *f, FDSet *fds) { +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-override=%s\n", yes_no(j->override)); 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)); @@ -987,15 +1011,16 @@ int job_serialize(Job *j, FILE *f, FDSet *fds) { if (j->begin_usec > 0) fprintf(f, "job-begin="USEC_FMT"\n", j->begin_usec); - bus_track_serialize(j->clients, f); + bus_track_serialize(j->clients, f, "subscribed"); /* End marker */ fputc('\n', f); return 0; } -int job_deserialize(Job *j, FILE *f, FDSet *fds) { +int job_deserialize(Job *j, FILE *f) { assert(j); + assert(f); for (;;) { char line[LINE_MAX], *l, *v; @@ -1047,15 +1072,6 @@ int job_deserialize(Job *j, FILE *f, FDSet *fds) { else job_set_state(j, s); - } else if (streq(l, "job-override")) { - int b; - - b = parse_boolean(v); - if (b < 0) - log_debug("Failed to parse job override flag %s", v); - else - j->override = j->override || b; - } else if (streq(l, "job-irreversible")) { int b; @@ -1094,7 +1110,7 @@ int job_deserialize(Job *j, FILE *f, FDSet *fds) { } else if (streq(l, "subscribed")) { if (strv_extend(&j->deserialized_clients, v) < 0) - return log_oom(); + log_oom(); } } } @@ -1106,24 +1122,22 @@ int job_coldplug(Job *j) { /* After deserialization is complete and the bus connection * set up again, let's start watching our subscribers again */ - r = bus_track_coldplug(j->manager, &j->clients, &j->deserialized_clients); - if (r < 0) - return r; + (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 == 0) + if (j->begin_usec == 0 || j->unit->job_timeout == USEC_INFINITY) return 0; - if (j->timer_event_source) - j->timer_event_source = sd_event_source_unref(j->timer_event_source); + 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, - j->begin_usec + j->unit->job_timeout, 0, + 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"); @@ -1147,7 +1161,7 @@ void job_shutdown_magic(Job *j) { if (j->type != JOB_START) return; - if (j->unit->manager->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(j->unit->manager)) return; if (!unit_has_name(j->unit, SPECIAL_SHUTDOWN_TARGET)) @@ -1156,16 +1170,16 @@ void job_shutdown_magic(Job *j) { /* In case messages on console has been disabled on boot */ j->unit->manager->no_console_output = false; - if (detect_container(NULL) > 0) + if (detect_container() > 0) return; asynchronous_sync(); } -int job_get_timeout(Job *j, uint64_t *timeout) { +int job_get_timeout(Job *j, usec_t *timeout) { + usec_t x = USEC_INFINITY, y = USEC_INFINITY; Unit *u = j->unit; - uint64_t x = -1, y = -1; - int r = 0, q = 0; + int r; assert(u); @@ -1173,20 +1187,18 @@ int job_get_timeout(Job *j, uint64_t *timeout) { r = sd_event_source_get_time(j->timer_event_source, &x); if (r < 0) return r; - r = 1; } if (UNIT_VTABLE(u)->get_timeout) { - q = UNIT_VTABLE(u)->get_timeout(u, &y); - if (q < 0) - return q; + r = UNIT_VTABLE(u)->get_timeout(u, &y); + if (r < 0) + return r; } - if (r == 0 && q == 0) + if (x == USEC_INFINITY && y == USEC_INFINITY) return 0; *timeout = MIN(x, y); - return 1; } @@ -1205,6 +1217,7 @@ static const char* const job_type_table[_JOB_TYPE_MAX] = { [JOB_RELOAD_OR_START] = "reload-or-start", [JOB_RESTART] = "restart", [JOB_TRY_RESTART] = "try-restart", + [JOB_TRY_RELOAD] = "try-reload", [JOB_NOP] = "nop", }; @@ -1235,3 +1248,15 @@ static const char* const job_result_table[_JOB_RESULT_MAX] = { }; 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/core/job.h b/src/core/job.h index 1d1b10f1d3..85368f0d30 100644 --- a/src/core/job.h +++ b/src/core/job.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,6 +21,11 @@ #include <stdbool.h> +#include "sd-event.h" + +#include "list.h" +#include "unit-name.h" + typedef struct Job Job; typedef struct JobDependency JobDependency; typedef enum JobType JobType; @@ -61,6 +64,9 @@ enum JobType { * 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 @@ -105,9 +111,7 @@ enum JobResult { _JOB_RESULT_INVALID = -1 }; -#include "sd-event.h" #include "unit.h" -#include "list.h" struct JobDependency { /* Encodes that the 'subject' job needs the 'object' job in @@ -160,7 +164,6 @@ struct Job { bool installed:1; bool in_run_queue:1; bool matters_to_anchor:1; - bool override:1; bool in_dbus_queue:1; bool sent_dbus_new_signal:1; bool ignore_order:1; @@ -174,8 +177,8 @@ 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, FDSet *fds); -int job_deserialize(Job *j, FILE *f, FDSet *fds); +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); @@ -216,12 +219,14 @@ 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); +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_; @@ -234,4 +239,4 @@ 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_; -int job_get_timeout(Job *j, uint64_t *timeout) _pure_; +const char* job_type_to_access_method(JobType t); diff --git a/src/core/kill.c b/src/core/kill.c index 2de71c6bf9..6854587d54 100644 --- a/src/core/kill.c +++ b/src/core/kill.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,9 +17,10 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "util.h" -#include "signal-util.h" #include "kill.h" +#include "signal-util.h" +#include "string-table.h" +#include "util.h" void kill_context_init(KillContext *c) { assert(c); @@ -60,7 +59,10 @@ 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_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/core/kill.h b/src/core/kill.h index d5f125fa41..b3d2056cb0 100644 --- a/src/core/kill.h +++ b/src/core/kill.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -50,6 +48,9 @@ typedef enum KillWho { KILL_MAIN, KILL_CONTROL, KILL_ALL, + KILL_MAIN_FAIL, + KILL_CONTROL_FAIL, + KILL_ALL_FAIL, _KILL_WHO_MAX, _KILL_WHO_INVALID = -1 } KillWho; diff --git a/src/core/killall.c b/src/core/killall.c index 2a9d72c901..a8b814e868 100644 --- a/src/core/killall.c +++ b/src/core/killall.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,21 +17,24 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/wait.h> -#include <signal.h> #include <errno.h> +#include <signal.h> +#include <sys/wait.h> #include <unistd.h> -#include "util.h" -#include "killall.h" -#include "set.h" +#include "alloc-util.h" +#include "def.h" +#include "fd-util.h" #include "formats-util.h" +#include "killall.h" +#include "parse-util.h" #include "process-util.h" +#include "set.h" +#include "string-util.h" #include "terminal-util.h" +#include "util.h" -#define TIMEOUT_USEC (10 * USEC_PER_SEC) - -static bool ignore_proc(pid_t pid) { +static bool ignore_proc(pid_t pid, bool warn_rootfs) { _cleanup_fclose_ FILE *f = NULL; char c; const char *p; @@ -68,7 +69,22 @@ static bool ignore_proc(pid_t pid) { * spree. * * http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */ - if (count == 1 && c == '@') + 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; @@ -82,7 +98,7 @@ static void wait_for_children(Set *pids, sigset_t *mask) { if (set_isempty(pids)) return; - until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC; + until = now(CLOCK_MONOTONIC) + DEFAULT_TIMEOUT_USEC; for (;;) { struct timespec ts; int k; @@ -108,7 +124,7 @@ static void wait_for_children(Set *pids, sigset_t *mask) { return; } - set_remove(pids, ULONG_TO_PTR(pid)); + (void) set_remove(pids, PID_TO_PTR(pid)); } /* Now explicitly check who might be remaining, who @@ -117,7 +133,7 @@ static void wait_for_children(Set *pids, sigset_t *mask) { /* We misuse getpgid as a check whether a * process still exists. */ - if (getpgid((pid_t) PTR_TO_ULONG(p)) >= 0) + if (getpgid(PTR_TO_PID(p)) >= 0) continue; if (errno != ESRCH) @@ -167,7 +183,7 @@ static int killall(int sig, Set *pids, bool send_sighup) { if (parse_pid(d->d_name, &pid) < 0) continue; - if (ignore_proc(pid)) + if (ignore_proc(pid, sig == SIGKILL && !in_initrd())) continue; if (sig == SIGKILL) { @@ -179,7 +195,7 @@ static int killall(int sig, Set *pids, bool send_sighup) { if (kill(pid, sig) >= 0) { if (pids) { - r = set_put(pids, ULONG_TO_PTR(pid)); + r = set_put(pids, PID_TO_PTR(pid)); if (r < 0) log_oom(); } diff --git a/src/core/killall.h b/src/core/killall.h index 986cf432cb..acc2439f00 100644 --- a/src/core/killall.h +++ b/src/core/killall.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/kmod-setup.c b/src/core/kmod-setup.c index fc6d2f4acb..fd1021f706 100644 --- a/src/core/kmod-setup.c +++ b/src/core/kmod-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,17 +17,17 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <unistd.h> #include <string.h> +#include <unistd.h> #ifdef HAVE_KMOD #include <libkmod.h> #endif -#include "macro.h" -#include "capability.h" #include "bus-util.h" +#include "capability-util.h" #include "kmod-setup.h" +#include "macro.h" #ifdef HAVE_KMOD static void systemd_kmod_log( @@ -66,9 +64,6 @@ int kmod_setup(void) { /* this should never be a module */ { "unix", "/proc/net/unix", true, true, NULL }, - /* IPC is needed before we bring up any other services */ - { "kdbus", "/sys/fs/kdbus", false, false, is_kdbus_wanted }, - #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 }, @@ -112,7 +107,7 @@ int kmod_setup(void) { r = kmod_module_probe_insert_module(mod, KMOD_PROBE_APPLY_BLACKLIST, NULL, NULL, NULL, NULL); if (r == 0) - log_info("Inserted module '%s'", kmod_module_get_name(mod)); + 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 { diff --git a/src/core/kmod-setup.h b/src/core/kmod-setup.h index 24dcdddfa4..685f4df301 100644 --- a/src/core/kmod-setup.h +++ b/src/core/kmod-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/load-dropin.c b/src/core/load-dropin.c index 11566af51b..f83fa09301 100644 --- a/src/core/load-dropin.c +++ b/src/core/load-dropin.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,13 +18,13 @@ ***/ -#include "unit.h" +#include "conf-parser.h" #include "load-dropin.h" +#include "load-fragment.h" #include "log.h" #include "strv.h" #include "unit-name.h" -#include "conf-parser.h" -#include "load-fragment.h" +#include "unit.h" static int add_dependency_consumer( UnitDependency dependency, @@ -46,6 +44,7 @@ static int add_dependency_consumer( } int unit_load_dropin(Unit *u) { + _cleanup_strv_free_ char **l = NULL; Iterator i; char *t, **f; int r; @@ -57,7 +56,7 @@ int unit_load_dropin(Unit *u) { SET_FOREACH(t, u->names, i) { char **p; - STRV_FOREACH(p, u->manager->lookup_paths.unit_path) { + 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, @@ -65,10 +64,19 @@ int unit_load_dropin(Unit *u) { } } - r = unit_find_dropin_paths(u, &u->dropin_paths); + 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, diff --git a/src/core/load-dropin.h b/src/core/load-dropin.h index 1e018c4525..942d26724e 100644 --- a/src/core/load-dropin.h +++ b/src/core/load-dropin.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -21,13 +19,13 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" #include "dropin.h" +#include "unit.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.unit_path, + return unit_file_find_dropin_paths(u->manager->lookup_paths.search_path, u->manager->unit_path_cache, u->names, paths); diff --git a/src/core/load-fragment-gperf.gperf.m4 b/src/core/load-fragment-gperf.gperf.m4 index aae81c80cb..c49c1d6732 100644 --- a/src/core/load-fragment-gperf.gperf.m4 +++ b/src/core/load-fragment-gperf.gperf.m4 @@ -17,11 +17,11 @@ struct ConfigPerfItem; %% m4_dnl Define the context options only once m4_define(`EXEC_CONTEXT_CONFIG_ITEMS', -`$1.WorkingDirectory, config_parse_unit_path_printf, 0, offsetof($1, exec_context.working_directory) +`$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_unit_string_printf, 0, offsetof($1, exec_context.user) -$1.Group, config_parse_unit_string_printf, 0, offsetof($1, exec_context.group) -$1.SupplementaryGroups, config_parse_strv, 0, offsetof($1, exec_context.supplementary_groups) +$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) @@ -33,6 +33,8 @@ $1.CPUAffinity, config_parse_exec_cpu_affinity, 0, $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_input, 0, offsetof($1, exec_context.std_input) $1.StandardOutput, config_parse_output, 0, offsetof($1, exec_context.std_output) $1.StandardError, config_parse_output, 0, offsetof($1, exec_context.std_error) @@ -44,19 +46,24 @@ $1.SyslogIdentifier, config_parse_unit_string_printf, 0, $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_exec_capabilities, 0, offsetof($1, exec_context) +$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_bounding_set, 0, offsetof($1, exec_context.capability_bounding_set_drop) +$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) @@ -74,12 +81,18 @@ $1.LimitMSGQUEUE, config_parse_limit, RLIMIT_MSGQ $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_dirs) -$1.ReadOnlyDirectories, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.read_only_dirs) -$1.InaccessibleDirectories, config_parse_namespace_path_strv, 0, offsetof($1, exec_context.inaccessible_dirs) +$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.PrivateNetwork, config_parse_bool, 0, offsetof($1, exec_context.private_network) $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.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) @@ -91,6 +104,7 @@ m4_ifdef(`HAVE_PAM', `$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') @@ -105,33 +119,48 @@ 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_kill_signal, 0, offsetof($1, kill_context.kill_signal)' +$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.Delegate, config_parse_bool, 0, offsetof($1, cgroup_context.delegate)' +$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.RequiresOverridable, config_parse_unit_deps, UNIT_REQUIRES_OVERRIDABLE, 0 Unit.Requisite, config_parse_unit_deps, UNIT_REQUISITE, 0 -Unit.RequisiteOverridable, config_parse_unit_deps, UNIT_REQUISITE_OVERRIDABLE, 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 @@ -145,6 +174,8 @@ Unit.ReloadPropagatedFrom, config_parse_unit_deps, UNIT_RELOAD 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) @@ -154,10 +185,16 @@ Unit.DefaultDependencies, config_parse_bool, 0, 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_bool, 0, offsetof(Unit, ignore_on_snapshot) -Unit.JobTimeoutSec, config_parse_sec, 0, offsetof(Unit, job_timeout) +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_failure_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_failure_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) @@ -205,14 +242,16 @@ Service.ExecReload, config_parse_exec, SERVICE_EXE 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, offsetof(Service, timeout_start_usec) -Service.TimeoutStartSec, config_parse_service_timeout, 0, offsetof(Service, timeout_start_usec) -Service.TimeoutStopSec, config_parse_service_timeout, 0, offsetof(Service, timeout_stop_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) -Service.StartLimitInterval, config_parse_sec, 0, offsetof(Service, start_limit.interval) -Service.StartLimitBurst, config_parse_unsigned, 0, offsetof(Service, start_limit.burst) -Service.StartLimitAction, config_parse_failure_action, 0, offsetof(Service, start_limit_action) -Service.RebootArgument, config_parse_string, 0, offsetof(Service, reboot_arg) +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_failure_action, 0, offsetof(Unit, start_limit_action) +Service.RebootArgument, config_parse_string, 0, offsetof(Unit, reboot_arg) Service.FailureAction, config_parse_failure_action, 0, offsetof(Service, failure_action) Service.Type, config_parse_service_type, 0, offsetof(Service, type) Service.Restart, config_parse_service_restart, 0, offsetof(Service, restart) @@ -229,7 +268,9 @@ Service.BusName, config_parse_bus_name, 0, 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_bus_endpoint_policy, 0, offsetof(Service, exec_context) +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 @@ -241,6 +282,8 @@ Socket.ListenFIFO, config_parse_socket_listen, SOCKET_FIFO 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 @@ -249,12 +292,14 @@ Socket.ExecStartPost, config_parse_exec, SOCKET_EXEC 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_unit_string_printf, 0, offsetof(Socket, user) -Socket.SocketGroup, config_parse_unit_string_printf, 0, offsetof(Socket, group) +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) @@ -279,7 +324,10 @@ Socket.MessageQueueMaxMessages, config_parse_long, 0, 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) @@ -310,6 +358,8 @@ Mount.Type, config_parse_string, 0, 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 @@ -334,7 +384,9 @@ Timer.OnUnitActiveSec, config_parse_timer, 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 diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c index a48cb4029a..06c156a623 100644 --- a/src/core/load-fragment.c +++ b/src/core/load-fragment.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,44 +18,55 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <linux/oom.h> #include <errno.h> -#include <string.h> #include <fcntl.h> -#include <sched.h> #include <linux/fs.h> -#include <sys/stat.h> -#include <sys/resource.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 "unit.h" -#include "strv.h" +#include "af-list.h" +#include "alloc-util.h" +#include "bus-error.h" +#include "bus-internal.h" +#include "bus-util.h" +#include "cap-list.h" +#include "capability-util.h" +#include "cgroup.h" #include "conf-parser.h" +#include "cpu-set-util.h" +#include "env-util.h" +#include "errno-list.h" +#include "escape.h" +#include "fd-util.h" +#include "fs-util.h" +#include "ioprio.h" #include "load-fragment.h" #include "log.h" -#include "ioprio.h" -#include "securebits.h" #include "missing.h" -#include "unit-name.h" -#include "unit-printf.h" -#include "utf8.h" +#include "parse-util.h" #include "path-util.h" -#include "env-util.h" -#include "cgroup.h" -#include "bus-util.h" -#include "bus-error.h" -#include "errno-list.h" -#include "af-list.h" -#include "cap-list.h" -#include "signal-util.h" -#include "bus-internal.h" - +#include "process-util.h" +#include "rlimit-util.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" #endif +#include "securebits.h" +#include "signal-util.h" +#include "stat-util.h" +#include "string-util.h" +#include "strv.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "unit.h" +#include "user-util.h" +#include "utf8.h" +#include "web-util.h" int config_parse_warn_compat( const char *unit, @@ -74,15 +83,15 @@ int config_parse_warn_compat( switch(reason) { case DISABLED_CONFIGURATION: - log_syntax(unit, LOG_DEBUG, filename, line, EINVAL, + 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, EINVAL, + 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, EINVAL, + log_syntax(unit, LOG_INFO, filename, line, 0, "Support for option %s= has not yet been enabled and it is ignored", lvalue); break; }; @@ -90,52 +99,73 @@ int config_parse_warn_compat( 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) { +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 *word, *state; - size_t l; + const char *p; assert(filename); assert(lvalue); assert(rvalue); - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *t = NULL, *k = NULL; + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL, *k = NULL; int r; - t = strndup(word, l); - if (!t) + 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, t, &k); + r = unit_name_printf(u, word, &k); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve specifiers, ignoring: %s", strerror(-r)); + 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: %s", k, strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add dependency on %s, ignoring: %m", k); } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Invalid syntax, ignoring."); 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, @@ -166,16 +196,17 @@ int config_parse_unit_string_printf( 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) { +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; @@ -187,24 +218,25 @@ int config_parse_unit_strv_printf(const char *unit, 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: %s", rvalue, strerror(-r)); + 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 ? k : rvalue, data, userdata); + 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) { +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; @@ -217,7 +249,7 @@ int config_parse_unit_path_printf(const char *unit, 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: %s", rvalue, strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve unit specifiers on %s, ignoring: %m", rvalue); return 0; } @@ -256,17 +288,17 @@ int config_parse_unit_path_strv_printf( 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: %s", t, strerror(-r)); + 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_invalid_utf8(unit, LOG_ERR, filename, line, EINVAL, rvalue); + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, rvalue); return 0; } if (!path_is_absolute(k)) { - log_syntax(unit, LOG_ERR, filename, line, -r, "Symlink path %s is not absolute, ignoring: %s", k, strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, 0, "Symlink path %s is not absolute, ignoring: %m", k); return 0; } @@ -279,7 +311,7 @@ int config_parse_unit_path_strv_printf( k = NULL; } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Invalid syntax, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid syntax, ignoring."); return 0; } @@ -322,12 +354,8 @@ int config_parse_socket_listen(const char *unit, p->type = ltype; r = unit_full_printf(UNIT(s), rvalue, &p->path); if (r < 0) { - p->path = strdup(rvalue); - if (!p->path) - return log_oom(); - else - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve unit specifiers on %s, ignoring: %s", rvalue, strerror(-r)); + 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); @@ -337,14 +365,14 @@ int config_parse_socket_listen(const char *unit, 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: %s", rvalue, strerror(-r)); + 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 ?: rvalue); + 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse address value, ignoring: %s", rvalue); return 0; } @@ -353,14 +381,14 @@ int config_parse_socket_listen(const char *unit, 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: %s", rvalue, strerror(-r)); + 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 ? k : rvalue); + 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse address value, ignoring: %s", rvalue); return 0; } @@ -374,13 +402,14 @@ int config_parse_socket_listen(const char *unit, } if (socket_address_family(&p->address) != AF_LOCAL && p->address.type == SOCK_SEQPACKET) { - log_syntax(unit, LOG_ERR, filename, line, EOPNOTSUPP, - "Address family not supported, ignoring: %s", rvalue); + 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) { @@ -393,6 +422,37 @@ int config_parse_socket_listen(const char *unit, 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, @@ -420,8 +480,7 @@ int config_parse_socket_bind(const char *unit, r = parse_boolean(rvalue); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse bind IPv6 only value, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse bind IPv6 only value, ignoring: %s", rvalue); return 0; } @@ -432,16 +491,17 @@ int config_parse_socket_bind(const char *unit, 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) { +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; @@ -451,16 +511,13 @@ int config_parse_exec_nice(const char *unit, assert(rvalue); assert(data); - r = safe_atoi(rvalue, &priority); + r = parse_nice(rvalue, &priority); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to parse nice priority, ignoring: %s. ", rvalue); - return 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); - if (priority < PRIO_MIN || priority >= PRIO_MAX) { - log_syntax(unit, LOG_ERR, filename, line, ERANGE, - "Nice priority out of range, ignoring: %s", rvalue); return 0; } @@ -491,14 +548,12 @@ int config_parse_exec_oom_score_adjust(const char* unit, 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); + 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, ERANGE, - "OOM score adjust value out of range, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "OOM score adjust value out of range, ignoring: %s", rvalue); return 0; } @@ -531,9 +586,7 @@ int config_parse_exec( assert(e); e += ltype; - rvalue += strspn(rvalue, WHITESPACE); - p = rvalue; if (isempty(rvalue)) { /* An empty assignment resets the list */ @@ -541,55 +594,55 @@ int config_parse_exec( return 0; } + p = rvalue; do { - int i; + _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; - _cleanup_free_ ExecCommand *nce = NULL; - _cleanup_free_ char *path = NULL, *firstword = NULL; char *f; - bool separate_argv0 = false, ignore = false; + int i; semicolon = false; - r = unquote_first_word_and_warn(&p, &firstword, UNQUOTE_CUNESCAPE, unit, filename, line, rvalue); + 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 < 2; i++) { - /* We accept an absolute path as first argument, or - * alternatively an absolute prefixed with @ to allow - * overriding of argv[0]. */ + 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 ++; + f++; } if (isempty(f)) { /* First word is either "-" or "@" with no command. */ - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Empty path in command line, ignoring: \"%s\"", rvalue); + 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, EINVAL, - "Executable path contains special characters, ignoring: %s", rvalue); + 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, EINVAL, - "Executable path is not absolute, ignoring: %s", rvalue); + 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, EINVAL, - "Executable path specifies a directory, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Executable path specifies a directory, ignoring: %s", rvalue); return 0; } @@ -614,20 +667,20 @@ int config_parse_exec( path_kill_slashes(path); - for (;;) { + 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++; p += strspn(p, WHITESPACE); semicolon = true; break; } /* Check for \; explicitly, to not confuse it with \\; - * or "\;" or "\\;" etc. unquote_first_word would + * 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; @@ -642,7 +695,7 @@ int config_parse_exec( continue; } - r = unquote_first_word_and_warn(&p, &word, UNQUOTE_CUNESCAPE, unit, filename, line, rvalue); + 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) @@ -656,8 +709,7 @@ int config_parse_exec( } if (!n || !n[0]) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Empty executable name or zeroeth argument, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Empty executable name or zeroeth argument, ignoring: %s", rvalue); return 0; } @@ -668,6 +720,7 @@ int config_parse_exec( nce->argv = n; nce->path = path; nce->ignore = ignore; + nce->privileged = privileged; exec_command_append_list(e, nce); @@ -685,16 +738,17 @@ int config_parse_exec( 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) { +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; @@ -705,6 +759,11 @@ int config_parse_socket_bindtodevice(const char* unit, 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(); @@ -741,8 +800,7 @@ int config_parse_exec_io_class(const char *unit, x = ioprio_class_from_string(rvalue); if (x < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse IO scheduling class, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse IO scheduling class, ignoring: %s", rvalue); return 0; } @@ -773,8 +831,7 @@ int config_parse_exec_io_priority(const char *unit, 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse IO priority, ignoring: %s", rvalue); return 0; } @@ -806,8 +863,7 @@ int config_parse_exec_cpu_sched_policy(const char *unit, x = sched_policy_from_string(rvalue); if (x < 0) { - log_syntax(unit, LOG_ERR, filename, line, -x, - "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); return 0; } @@ -840,8 +896,7 @@ int config_parse_exec_cpu_sched_prio(const char *unit, 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CPU scheduling policy, ignoring: %s", rvalue); return 0; } @@ -850,8 +905,7 @@ int config_parse_exec_cpu_sched_prio(const char *unit, max = sched_get_priority_max(c->cpu_sched_policy); if (i < min || i > max) { - log_syntax(unit, LOG_ERR, filename, line, ERANGE, - "CPU scheduling priority is out of range, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "CPU scheduling priority is out of range, ignoring: %s", rvalue); return 0; } @@ -873,83 +927,29 @@ int config_parse_exec_cpu_affinity(const char *unit, void *userdata) { ExecContext *c = data; - const char *word, *state; - size_t l; + _cleanup_cpu_free_ cpu_set_t *cpuset = NULL; + int ncpus; assert(filename); assert(lvalue); assert(rvalue); assert(data); - if (isempty(rvalue)) { - /* An empty assignment resets the CPU list */ - if (c->cpuset) - CPU_FREE(c->cpuset); - c->cpuset = NULL; - return 0; - } - - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *t = NULL; - int r; - unsigned cpu; - - t = strndup(word, l); - if (!t) - return log_oom(); - - r = safe_atou(t, &cpu); - - if (!c->cpuset) { - c->cpuset = cpu_set_malloc(&c->cpuset_ncpus); - if (!c->cpuset) - return log_oom(); - } - - if (r < 0 || cpu >= c->cpuset_ncpus) { - log_syntax(unit, LOG_ERR, filename, line, ERANGE, - "Failed to parse CPU affinity '%s', ignoring: %s", t, rvalue); - return 0; - } - - CPU_SET_S(cpu, CPU_ALLOC_SIZE(c->cpuset_ncpus), c->cpuset); - } - if (!isempty(state)) - log_syntax(unit, LOG_WARNING, filename, line, EINVAL, - "Trailing garbage, ignoring."); - - return 0; -} + ncpus = parse_cpu_set_and_warn(rvalue, &cpuset, unit, filename, line, lvalue); + if (ncpus < 0) + return ncpus; -int config_parse_exec_capabilities(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) { + if (c->cpuset) + CPU_FREE(c->cpuset); - ExecContext *c = data; - cap_t cap; - - assert(filename); - assert(lvalue); - assert(rvalue); - assert(data); - - cap = cap_from_text(rvalue); - if (!cap) { - log_syntax(unit, LOG_ERR, filename, line, errno, - "Failed to parse capabilities, ignoring: %s", rvalue); - return 0; + if (ncpus == 0) + /* An empty assignment resets the CPU list */ + c->cpuset = NULL; + else { + c->cpuset = cpuset; + cpuset = NULL; } - - if (c->capabilities) - cap_free(c->capabilities); - c->capabilities = cap; + c->cpuset_ncpus = ncpus; return 0; } @@ -994,34 +994,32 @@ int config_parse_exec_secure_bits(const char *unit, else if (first_word(word, "noroot-locked")) c->secure_bits |= 1<<SECURE_NOROOT_LOCKED; else { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse secure bits, ignoring: %s", rvalue); + 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, EINVAL, - "Invalid syntax, garbage at the end, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid syntax, garbage at the end, ignoring."); return 0; } -int config_parse_bounding_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_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_bounding_set_drop = data; - const char *word, *state; - size_t l; + uint64_t *capability_set = data; + uint64_t sum = 0, initial = 0; bool invert = false; - uint64_t sum = 0; + const char *p; assert(filename); assert(lvalue); @@ -1033,80 +1031,84 @@ int config_parse_bounding_set(const char *unit, rvalue++; } - /* Note that we store this inverted internally, since the - * kernel wants it like this. But we actually expose it - * non-inverted everywhere to have a fully normalized - * interface. */ + if (strcmp(lvalue, "CapabilityBoundingSet") == 0) + initial = CAP_ALL; /* initialized to all bits on */ + /* else "AmbientCapabilities" initialized to all bits off */ - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *t = NULL; - int cap; + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL; + int cap, r; - t = strndup(word, l); - if (!t) + 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(t); + cap = capability_from_name(word); if (cap < 0) { - log_syntax(unit, LOG_ERR, filename, line, errno, "Failed to parse capability in bounding set, ignoring: %s", t); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse capability in bounding/ambient set, ignoring: %s", word); continue; } - sum |= ((uint64_t) 1ULL) << (uint64_t) cap; + sum |= ((uint64_t) UINT64_C(1)) << (uint64_t) cap; } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); - if (invert) - *capability_bounding_set_drop |= sum; + sum = invert ? ~sum : sum; + + if (sum == 0 || *capability_set == initial) + /* "" or uninitialized data -> replace */ + *capability_set = sum; else - *capability_bounding_set_drop |= ~sum; + /* 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) { +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; - unsigned long long u; + struct rlimit **rl = data, d = {}; + int r; assert(filename); assert(lvalue); assert(rvalue); assert(data); - rl += ltype; - - if (streq(rvalue, "infinity")) - u = (unsigned long long) RLIM_INFINITY; - else { - int r; - - r = safe_atollu(rvalue, &u); - if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to parse resource value, ignoring: %s", rvalue); - return 0; - } + 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) { - *rl = new(struct rlimit, 1); - if (!*rl) + if (rl[ltype]) + *rl[ltype] = d; + else { + rl[ltype] = newdup(struct rlimit, &d, 1); + if (!rl[ltype]) return log_oom(); } - (*rl)->rlim_cur = (*rl)->rlim_max = (rlim_t) u; return 0; } @@ -1132,8 +1134,7 @@ int config_parse_sysv_priority(const char *unit, 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse SysV start priority, ignoring: %s", rvalue); return 0; } @@ -1142,38 +1143,9 @@ int config_parse_sysv_priority(const char *unit, } #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_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 *sig = data; - int r; - - assert(filename); - assert(lvalue); - assert(rvalue); - assert(sig); - - r = signal_from_string_try_harder(rvalue); - if (r <= 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to parse kill signal, ignoring: %s", rvalue); - return 0; - } - - *sig = r; - return 0; -} - int config_parse_exec_mount_flags(const char *unit, const char *filename, unsigned line, @@ -1185,38 +1157,28 @@ int config_parse_exec_mount_flags(const char *unit, void *data, void *userdata) { - ExecContext *c = data; - const char *word, *state; - size_t l; + unsigned long flags = 0; + ExecContext *c = data; assert(filename); assert(lvalue); assert(rvalue); assert(data); - FOREACH_WORD_SEPARATOR(word, l, rvalue, ", ", state) { - _cleanup_free_ char *t; - - t = strndup(word, l); - if (!t) - return log_oom(); - - if (streq(t, "shared")) - flags = MS_SHARED; - else if (streq(t, "slave")) - flags = MS_SLAVE; - else if (streq(t, "private")) - flags = MS_PRIVATE; - else { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Failed to parse mount flag %s, ignoring: %s", t, rvalue); - return 0; - } + 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; } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Trailing garbage, ignoring."); c->mount_flags = flags; + return 0; } @@ -1244,8 +1206,7 @@ int config_parse_exec_selinux_context( assert(data); if (isempty(rvalue)) { - free(c->selinux_context); - c->selinux_context = NULL; + c->selinux_context = mfree(c->selinux_context); c->selinux_context_ignore = false; return 0; } @@ -1258,8 +1219,7 @@ int config_parse_exec_selinux_context( r = unit_name_printf(u, rvalue, &k); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve specifiers, ignoring: %s", strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); return 0; } @@ -1294,8 +1254,7 @@ int config_parse_exec_apparmor_profile( assert(data); if (isempty(rvalue)) { - free(c->apparmor_profile); - c->apparmor_profile = NULL; + c->apparmor_profile = mfree(c->apparmor_profile); c->apparmor_profile_ignore = false; return 0; } @@ -1308,8 +1267,7 @@ int config_parse_exec_apparmor_profile( r = unit_name_printf(u, rvalue, &k); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve specifiers, ignoring: %s", strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); return 0; } @@ -1344,8 +1302,7 @@ int config_parse_exec_smack_process_label( assert(data); if (isempty(rvalue)) { - free(c->smack_process_label); - c->smack_process_label = NULL; + c->smack_process_label = mfree(c->smack_process_label); c->smack_process_label_ignore = false; return 0; } @@ -1358,8 +1315,7 @@ int config_parse_exec_smack_process_label( r = unit_name_printf(u, rvalue, &k); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve specifiers, ignoring: %s", strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %m"); return 0; } @@ -1382,10 +1338,13 @@ int config_parse_timer(const char *unit, void *userdata) { Timer *t = data; - usec_t u = 0; + 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); @@ -1400,23 +1359,24 @@ int config_parse_timer(const char *unit, b = timer_base_from_string(lvalue); if (b < 0) { - log_syntax(unit, LOG_ERR, filename, line, -b, - "Failed to parse timer base, ignoring: %s", lvalue); + 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(rvalue, &c) < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse calendar specification, ignoring: %s", - rvalue); + 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(rvalue, &u) < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse timer value, ignoring: %s", - rvalue); + if (parse_sec(k, &usec) < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse timer value, ignoring: %s", k); return 0; } } @@ -1428,7 +1388,7 @@ int config_parse_timer(const char *unit, } v->base = b; - v->value = u; + v->value = usec; v->calendar_spec = c; LIST_PREPEND(value, t->values, v); @@ -1459,33 +1419,30 @@ int config_parse_trigger_unit( assert(data); if (!set_isempty(u->dependencies[UNIT_TRIGGERS])) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Multiple units to trigger specified, ignoring: %s", rvalue); + 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: %s", strerror(-r)); + 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 ?: rvalue); + type = unit_name_to_type(p); if (type < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Unit type not valid, ignoring: %s", rvalue); + 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, EINVAL, - "Trigger cannot be of same type, ignoring: %s", rvalue); + 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 ?: rvalue, NULL, true); + 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: %s", p ?: rvalue, strerror(-r)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add trigger on %s, ignoring: %m", p); return 0; } @@ -1522,25 +1479,18 @@ int config_parse_path_spec(const char *unit, b = path_type_from_string(lvalue); if (b < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse path type, ignoring: %s", lvalue); + 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) { - k = strdup(rvalue); - if (!k) - return log_oom(); - else - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to resolve unit specifiers on %s. Ignoring.", - rvalue); + 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, EINVAL, - "Path is not absolute, ignoring: %s", k); + log_syntax(unit, LOG_ERR, filename, line, 0, "Path is not absolute, ignoring: %s", k); return 0; } @@ -1571,11 +1521,11 @@ int config_parse_socket_service( void *data, void *userdata) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_free_ char *p = NULL; Socket *s = data; - int r; Unit *x; - _cleanup_free_ char *p = NULL; + int r; assert(filename); assert(lvalue); @@ -1589,13 +1539,13 @@ int config_parse_socket_service( } if (!endswith(p, ".service")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Unit must be of type service, ignoring: %s", rvalue); + 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)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load unit %s, ignoring: %s", rvalue, bus_error_message(&error, r)); return 0; } @@ -1604,6 +1554,50 @@ int config_parse_socket_service( 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; + } + + free(s->fdname); + s->fdname = p; + p = NULL; + + return 0; +} + int config_parse_service_sockets( const char *unit, const char *filename, @@ -1617,8 +1611,7 @@ int config_parse_service_sockets( void *userdata) { Service *s = data; - const char *word, *state; - size_t l; + const char *p; int r; assert(filename); @@ -1626,21 +1619,28 @@ int config_parse_service_sockets( assert(rvalue); assert(data); - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *t = NULL, *k = NULL; + p = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL, *k = NULL; - t = strndup(word, l); - if (!t) + 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), t, &k); + 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, EINVAL, "Unit must be of type socket, ignoring: %s", k); + log_syntax(unit, LOG_ERR, filename, line, 0, "Unit must be of type socket, ignoring: %s", k); continue; } @@ -1652,8 +1652,6 @@ int config_parse_service_sockets( if (r < 0) log_syntax(unit, LOG_ERR, filename, line, r, "Failed to add dependency on %s, ignoring: %m", k); } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Trailing garbage, ignoring."); return 0; } @@ -1686,25 +1684,27 @@ int config_parse_bus_name( } if (!service_name_is_valid(k)) { - log_syntax(unit, LOG_ERR, filename, line, r, "Invalid bus name %s, ignoring.", 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) { +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); @@ -1712,16 +1712,180 @@ int config_parse_service_timeout(const char *unit, assert(rvalue); assert(s); - r = config_parse_sec(unit, filename, line, section, section_line, lvalue, ltype, - rvalue, data, userdata); - if (r < 0) - return r; + /* This is called for three cases: TimeoutSec=, TimeoutStopSec= and TimeoutStartSec=. */ - if (streq(lvalue, "TimeoutSec")) { - s->start_timeout_defined = true; - s->timeout_stop_usec = s->timeout_start_usec; - } else if (streq(lvalue, "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; } @@ -1738,7 +1902,7 @@ int config_parse_busname_service( void *data, void *userdata) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; BusName *n = data; int r; Unit *x; @@ -1751,21 +1915,18 @@ int config_parse_busname_service( 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); + 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, EINVAL, - "Unit must be of type service, ignoring: %s", rvalue); + 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)); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load unit %s, ignoring: %s", rvalue, bus_error_message(&error, r)); return 0; } @@ -1815,8 +1976,7 @@ int config_parse_bus_policy( access_str = strpbrk(id_str, WHITESPACE); if (!access_str) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid busname policy value '%s'", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid busname policy value '%s'", rvalue); return 0; } @@ -1826,8 +1986,7 @@ int config_parse_bus_policy( p->access = bus_policy_access_from_string(access_str); if (p->access < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid busname policy access type '%s'", access_str); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid busname policy access type '%s'", access_str); return 0; } @@ -1840,7 +1999,7 @@ int config_parse_bus_policy( return 0; } -int config_parse_bus_endpoint_policy( +int config_parse_working_directory( const char *unit, const char *filename, unsigned line, @@ -1852,48 +2011,56 @@ int config_parse_bus_endpoint_policy( void *data, void *userdata) { - _cleanup_free_ char *name = NULL; - BusPolicyAccess access; ExecContext *c = data; - char *access_str; + Unit *u = userdata; + bool missing_ok; int r; assert(filename); assert(lvalue); assert(rvalue); - assert(data); + assert(c); + assert(u); - name = strdup(rvalue); - if (!name) - return log_oom(); + if (rvalue[0] == '-') { + missing_ok = true; + rvalue++; + } else + missing_ok = false; - access_str = strpbrk(name, WHITESPACE); - if (!access_str) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid endpoint policy value '%s'", rvalue); - return 0; - } + if (streq(rvalue, "~")) { + c->working_directory_home = true; + c->working_directory = mfree(c->working_directory); + } else { + _cleanup_free_ char *k = NULL; - *access_str = '\0'; - access_str++; - access_str += strspn(access_str, WHITESPACE); + 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; + } - access = bus_policy_access_from_string(access_str); - if (access <= _BUS_POLICY_ACCESS_INVALID || - access >= _BUS_POLICY_ACCESS_MAX) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid endpoint policy access type '%s'", access_str); - return 0; - } + path_kill_slashes(k); - if (!c->bus_endpoint) { - r = bus_endpoint_new(&c->bus_endpoint); + if (!utf8_is_valid(k)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, rvalue); + return 0; + } - if (r < 0) - return r; + 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(c->working_directory); + c->working_directory = k; + k = NULL; + + c->working_directory_home = false; } - return bus_endpoint_add_policy(c->bus_endpoint, name, access); + c->working_directory_missing_ok = missing_ok; + return 0; } int config_parse_unit_env_file(const char *unit, @@ -1910,7 +2077,6 @@ int config_parse_unit_env_file(const char *unit, char ***env = data; Unit *u = userdata; _cleanup_free_ char *n = NULL; - const char *s; int r; assert(filename); @@ -1920,24 +2086,22 @@ int config_parse_unit_env_file(const char *unit, if (isempty(rvalue)) { /* Empty assignment frees the list */ - strv_free(*env); - *env = NULL; + *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); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } - s = n ?: rvalue; - if (!path_is_absolute(s[0] == '-' ? s + 1 : s)) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Path '%s' is not absolute, ignoring.", s); + 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, s); + r = strv_extend(env, n); if (r < 0) return log_oom(); @@ -1969,24 +2133,26 @@ int config_parse_environ(const char *unit, if (isempty(rvalue)) { /* Empty assignment resets the list */ - strv_free(*env); - *env = NULL; + *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); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); + return 0; + } } - if (!k) + if (!k) { k = strdup(rvalue); - if (!k) - return log_oom(); + if (!k) + return log_oom(); + } FOREACH_WORD_QUOTED(word, l, k, state) { - _cleanup_free_ char *n; + _cleanup_free_ char *n = NULL; char **x; r = cunescape_length(word, l, 0, &n); @@ -1996,7 +2162,7 @@ int config_parse_environ(const char *unit, } if (!env_assignment_is_valid(n)) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, "Invalid environment assignment, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid environment assignment, ignoring: %s", rvalue); continue; } @@ -2008,8 +2174,71 @@ int config_parse_environ(const char *unit, *env = x; } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + 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; } @@ -2034,8 +2263,7 @@ int config_parse_ip_tos(const char *unit, x = ip_tos_from_string(rvalue); if (x < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse IP TOS value, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse IP TOS value, ignoring: %s", rvalue); return 0; } @@ -2083,12 +2311,12 @@ int config_parse_unit_condition_path( 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); + 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, EINVAL, "Path in condition not absolute, ignoring: %s", p); + log_syntax(unit, LOG_ERR, filename, line, 0, "Path in condition not absolute, ignoring: %s", p); return 0; } @@ -2140,7 +2368,7 @@ int config_parse_unit_condition_string( 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers, ignoring: %s", rvalue); return 0; } @@ -2189,7 +2417,7 @@ int config_parse_unit_condition_null( 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); + log_syntax(unit, LOG_ERR, filename, line, b, "Failed to parse boolean value in condition, ignoring: %s", rvalue); return 0; } @@ -2237,20 +2465,18 @@ int config_parse_unit_requires_mounts_for( return log_oom(); if (!utf8_is_valid(n)) { - log_invalid_utf8(unit, LOG_ERR, filename, line, EINVAL, rvalue); + 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); + 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, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } @@ -2277,8 +2503,7 @@ int config_parse_documentation(const char *unit, if (isempty(rvalue)) { /* Empty assignment resets the list */ - strv_free(u->documentation); - u->documentation = NULL; + u->documentation = strv_free(u->documentation); return 0; } @@ -2292,8 +2517,7 @@ int config_parse_documentation(const char *unit, if (documentation_url_is_valid(*a)) *(b++) = *a; else { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid URL, ignoring: %s", *a); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid URL, ignoring: %s", *a); free(*a); } } @@ -2304,6 +2528,55 @@ int config_parse_documentation(const char *unit, } #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 == '@') { + const SystemCallFilterSet *set; + + for (set = syscall_filter_sets; set->set_name; set++) + if (streq(set->set_name, t)) { + const char *sys; + + NULSTR_FOREACH(sys, set->value) { + r = syscall_filter_parse_one(unit, filename, line, c, invert, sys, false); + if (r < 0) + return r; + } + break; + } + } else { + int id; + + id = seccomp_syscall_resolve_name(t); + if (id == __NR_SCMP_ERROR) { + if (warn) + log_syntax(unit, LOG_ERR, 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 + set_remove(c->syscall_filter, INT_TO_PTR(id + 1)); + } + return 0; +} + int config_parse_syscall_filter( const char *unit, const char *filename, @@ -2316,13 +2589,6 @@ int config_parse_syscall_filter( void *data, void *userdata) { - static const char default_syscalls[] = - "execve\0" - "exit\0" - "exit_group\0" - "rt_sigreturn\0" - "sigreturn\0"; - ExecContext *c = data; Unit *u = userdata; bool invert = false; @@ -2337,8 +2603,7 @@ int config_parse_syscall_filter( if (isempty(rvalue)) { /* Empty assignment resets the list */ - set_free(c->syscall_filter); - c->syscall_filter = NULL; + c->syscall_filter = set_free(c->syscall_filter); c->syscall_whitelist = false; return 0; } @@ -2357,62 +2622,33 @@ int config_parse_syscall_filter( /* Allow everything but the ones listed */ c->syscall_whitelist = false; else { - const char *i; - /* Allow nothing but the ones listed */ c->syscall_whitelist = true; /* Accept default syscalls if we are on a whitelist */ - NULSTR_FOREACH(i, default_syscalls) { - int id; - - id = seccomp_syscall_resolve_name(i); - if (id < 0) - continue; - - r = set_put(c->syscall_filter, INT_TO_PTR(id + 1)); - if (r == 0) - continue; - if (r < 0) - return log_oom(); - } + r = syscall_filter_parse_one(unit, filename, line, c, false, "@default", false); + if (r < 0) + return r; } } FOREACH_WORD_QUOTED(word, l, rvalue, state) { _cleanup_free_ char *t = NULL; - int id; t = strndup(word, l); if (!t) return log_oom(); - id = seccomp_syscall_resolve_name(t); - if (id < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse system call, ignoring: %s", t); - continue; - } - - /* 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) - continue; - if (r < 0) - return log_oom(); - } else - set_remove(c->syscall_filter, INT_TO_PTR(id + 1)); + r = syscall_filter_parse_one(unit, filename, line, c, invert, t, true); + if (r < 0) + return r; } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); /* Turn on NNP, but only if it wasn't configured explicitly * before, and only if we are in user mode. */ - if (!c->no_new_privileges_set && u->manager->running_as == MANAGER_USER) + if (!c->no_new_privileges_set && MANAGER_IS_USER(u->manager)) c->no_new_privileges = true; return 0; @@ -2436,8 +2672,7 @@ int config_parse_syscall_archs( int r; if (isempty(rvalue)) { - set_free(*archs); - *archs = NULL; + *archs = set_free(*archs); return 0; } @@ -2455,8 +2690,7 @@ int config_parse_syscall_archs( r = seccomp_arch_from_string(t, &a); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse system call architecture, ignoring: %s", t); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse system call architecture, ignoring: %s", t); continue; } @@ -2467,8 +2701,7 @@ int config_parse_syscall_archs( return log_oom(); } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } @@ -2500,8 +2733,7 @@ int config_parse_syscall_errno( e = errno_from_name(rvalue); if (e < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse error number, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse error number, ignoring: %s", rvalue); return 0; } @@ -2533,8 +2765,7 @@ int config_parse_address_families( if (isempty(rvalue)) { /* Empty assignment resets the list */ - set_free(c->address_families); - c->address_families = NULL; + c->address_families = set_free(c->address_families); c->address_families_whitelist = false; return 0; } @@ -2562,8 +2793,7 @@ int config_parse_address_families( af = af_from_name(t); if (af <= 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse address family, ignoring: %s", t); + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse address family, ignoring: %s", t); continue; } @@ -2580,8 +2810,7 @@ int config_parse_address_families( set_remove(c->address_families, INT_TO_PTR(af)); } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } @@ -2600,7 +2829,7 @@ int config_parse_unit_slice( void *userdata) { _cleanup_free_ char *k = NULL; - Unit *u = userdata, *slice; + Unit *u = userdata, *slice = NULL; int r; assert(filename); @@ -2609,35 +2838,29 @@ int config_parse_unit_slice( 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); - if (!k) { - k = strdup(rvalue); - if (!k) - return log_oom(); + 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); + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to load slice unit %s. Ignoring.", k); return 0; } - if (slice->type != UNIT_SLICE) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Slice unit %s is not a slice. Ignoring.", k); + 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; } - unit_ref_set(&u->slice, slice); return 0; } DEFINE_CONFIG_PARSE_ENUM(config_parse_device_policy, cgroup_device_policy, CGroupDevicePolicy, "Failed to parse device policy"); -int config_parse_cpu_shares( +int config_parse_cpu_weight( const char *unit, const char *filename, unsigned line, @@ -2649,26 +2872,47 @@ int config_parse_cpu_shares( void *data, void *userdata) { - unsigned long *shares = data, lu; + uint64_t *weight = data; int r; assert(filename); assert(lvalue); assert(rvalue); - if (isempty(rvalue)) { - *shares = (unsigned long) -1; + 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; } - r = safe_atolu(rvalue, &lu); - if (r < 0 || lu <= 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "CPU shares '%s' invalid. Ignoring.", rvalue); + 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; } - *shares = lu; return 0; } @@ -2685,7 +2929,7 @@ int config_parse_cpu_quota( void *userdata) { CGroupContext *c = data; - double percent; + int r; assert(filename); assert(lvalue); @@ -2696,25 +2940,67 @@ int config_parse_cpu_quota( return 0; } - if (!endswith(rvalue, "%")) { - - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "CPU quota '%s' not ending in '%%'. Ignoring.", rvalue); + r = parse_percent_unbounded(rvalue); + if (r <= 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "CPU quota '%s' invalid. Ignoring.", rvalue); return 0; } - if (sscanf(rvalue, "%lf%%", &percent) != 1 || percent <= 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "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; + } } - c->cpu_quota_per_sec_usec = (usec_t) (percent * USEC_PER_SEC / 100); + 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_memory_limit( +int config_parse_tasks_max( const char *unit, const char *filename, unsigned line, @@ -2726,25 +3012,36 @@ int config_parse_memory_limit( void *data, void *userdata) { - CGroupContext *c = data; - off_t bytes; + uint64_t *tasks_max = data, v; + Unit *u = userdata; int r; if (isempty(rvalue)) { - c->memory_limit = (uint64_t) -1; + *tasks_max = u->manager->default_tasks_max; return 0; } - assert_cc(sizeof(uint64_t) == sizeof(off_t)); + if (streq(rvalue, "infinity")) { + *tasks_max = CGROUP_LIMIT_MAX; + return 0; + } - r = parse_size(rvalue, 1024, &bytes); + r = parse_percent(rvalue); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Memory limit '%s' invalid. Ignoring.", rvalue); + 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; } - c->memory_limit = (uint64_t) bytes; + *tasks_max = v; return 0; } @@ -2760,11 +3057,12 @@ int config_parse_device_allow( void *data, void *userdata) { - _cleanup_free_ char *path = NULL; + _cleanup_free_ char *path = NULL, *t = NULL; CGroupContext *c = data; CGroupDeviceAllow *a; - const char *m; + const char *m = NULL; size_t n; + int r; if (isempty(rvalue)) { while (c->device_allow) @@ -2773,26 +3071,30 @@ int config_parse_device_allow( return 0; } - n = strcspn(rvalue, WHITESPACE); - path = strndup(rvalue, n); + 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 (!startswith(path, "/dev/") && - !startswith(path, "block-") && - !startswith(path, "char-")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid device node path '%s'. Ignoring.", path); + if (!is_deviceallow_pattern(path)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); return 0; } - m = rvalue + n + strspn(rvalue + n, WHITESPACE); + m = t + n + strspn(t + n, WHITESPACE); if (isempty(m)) m = "rwm"; if (!in_charset(m, "rwm")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid device rights '%s'. Ignoring.", m); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device rights '%s'. Ignoring.", m); return 0; } @@ -2810,7 +3112,7 @@ int config_parse_device_allow( return 0; } -int config_parse_blockio_weight( +int config_parse_io_weight( const char *unit, const char *filename, unsigned line, @@ -2822,30 +3124,23 @@ int config_parse_blockio_weight( void *data, void *userdata) { - unsigned long *weight = data, lu; + uint64_t *weight = data; int r; assert(filename); assert(lvalue); assert(rvalue); - if (isempty(rvalue)) { - *weight = (unsigned long) -1; - return 0; - } - - r = safe_atolu(rvalue, &lu); - if (r < 0 || lu < 10 || lu > 1000) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Block IO weight '%s' invalid. Ignoring.", 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; } - *weight = lu; return 0; } -int config_parse_blockio_device_weight( +int config_parse_io_device_weight( const char *unit, const char *filename, unsigned line, @@ -2858,10 +3153,10 @@ int config_parse_blockio_device_weight( void *userdata) { _cleanup_free_ char *path = NULL; - CGroupBlockIODeviceWeight *w; + CGroupIODeviceWeight *w; CGroupContext *c = data; - unsigned long lu; const char *weight; + uint64_t u; size_t n; int r; @@ -2870,17 +3165,18 @@ int config_parse_blockio_device_weight( assert(rvalue); if (isempty(rvalue)) { - while (c->blockio_device_weights) - cgroup_context_free_blockio_device_weight(c, c->blockio_device_weights); + 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; - if (!*weight) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Expected block device and device weight. Ignoring."); + weight += strspn(weight, WHITESPACE); + + if (isempty(weight)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected block device and device weight. Ignoring."); return 0; } @@ -2889,33 +3185,32 @@ int config_parse_blockio_device_weight( return log_oom(); if (!path_startswith(path, "/dev")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid device node path '%s'. Ignoring.", path); + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid device node path '%s'. Ignoring.", path); return 0; } - weight += strspn(weight, WHITESPACE); - r = safe_atolu(weight, &lu); - if (r < 0 || lu < 10 || lu > 1000) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Block IO weight '%s' invalid. Ignoring.", rvalue); + 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; } - w = new0(CGroupBlockIODeviceWeight, 1); + assert(u != CGROUP_WEIGHT_INVALID); + + w = new0(CGroupIODeviceWeight, 1); if (!w) return log_oom(); w->path = path; path = NULL; - w->weight = lu; + w->weight = u; - LIST_PREPEND(device_weights, c->blockio_device_weights, w); + LIST_PREPEND(device_weights, c->io_device_weights, w); return 0; } -int config_parse_blockio_bandwidth( +int config_parse_io_limit( const char *unit, const char *filename, unsigned line, @@ -2928,11 +3223,11 @@ int config_parse_blockio_bandwidth( void *userdata) { _cleanup_free_ char *path = NULL; - CGroupBlockIODeviceBandwidth *b; + CGroupIODeviceLimit *l = NULL, *t; CGroupContext *c = data; - const char *bandwidth; - off_t bytes; - bool read; + CGroupIOLimitType type; + const char *limit; + uint64_t num; size_t n; int r; @@ -2940,25 +3235,135 @@ int config_parse_blockio_bandwidth( assert(lvalue); assert(rvalue); - read = streq("BlockIOReadBandwidth", lvalue); + type = cgroup_io_limit_type_from_string(lvalue); + assert(type >= 0); if (isempty(rvalue)) { - CGroupBlockIODeviceBandwidth *next; + 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); - LIST_FOREACH_SAFE (device_bandwidths, b, next, c->blockio_device_bandwidths) - if (b->read == read) - cgroup_context_free_blockio_device_bandwidth(c, b); + 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); - bandwidth = rvalue + n; - bandwidth += strspn(bandwidth, WHITESPACE); + weight = rvalue + n; + weight += strspn(weight, WHITESPACE); - if (!*bandwidth) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Expected space separated pair of device node and bandwidth. Ignoring."); + if (isempty(weight)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Expected block device and device weight. Ignoring."); return 0; } @@ -2967,35 +3372,32 @@ int config_parse_blockio_bandwidth( return log_oom(); if (!path_startswith(path, "/dev")) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Invalid device node path '%s'. Ignoring.", path); + 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, EINVAL, - "Block IO Bandwidth '%s' invalid. Ignoring.", rvalue); + 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; } - b = new0(CGroupBlockIODeviceBandwidth, 1); - if (!b) + assert(u != CGROUP_BLKIO_WEIGHT_INVALID); + + w = new0(CGroupBlockIODeviceWeight, 1); + if (!w) return log_oom(); - b->path = path; + w->path = path; path = NULL; - b->bandwidth = (uint64_t) bytes; - b->read = read; - LIST_PREPEND(device_bandwidths, c->blockio_device_bandwidths, b); + w->weight = u; + LIST_PREPEND(device_weights, c->blockio_device_weights, w); return 0; } -DEFINE_CONFIG_PARSE_ENUM(config_parse_job_mode, job_mode, JobMode, "Failed to parse job mode"); - -int config_parse_job_mode_isolate( +int config_parse_blockio_bandwidth( const char *unit, const char *filename, unsigned line, @@ -3007,25 +3409,84 @@ int config_parse_job_mode_isolate( void *data, void *userdata) { - JobMode *m = data; + _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); - r = parse_boolean(rvalue); - if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse boolean, ignoring: %s", 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; } - *m = r ? JOB_ISOLATE : JOB_REPLACE; + 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; } -int config_parse_personality( +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, @@ -3037,21 +3498,20 @@ int config_parse_personality( void *data, void *userdata) { - unsigned long *personality = data, p; + JobMode *m = data; + int r; assert(filename); assert(lvalue); assert(rvalue); - assert(personality); - p = personality_from_string(rvalue); - if (p == PERSONALITY_INVALID) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Failed to parse personality, ignoring: %s", 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; } - *personality = p; + *m = r ? JOB_ISOLATE : JOB_REPLACE; return 0; } @@ -3068,6 +3528,7 @@ int config_parse_runtime_directory( void *userdata) { char***rt = data; + Unit *u = userdata; const char *word, *state; size_t l; int r; @@ -3079,21 +3540,25 @@ int config_parse_runtime_directory( if (isempty(rvalue)) { /* Empty assignment resets the list */ - strv_free(*rt); - *rt = NULL; + *rt = strv_free(*rt); return 0; } FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *n; + _cleanup_free_ char *t = NULL, *n = NULL; - n = strndup(word, l); - if (!n) + 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, EINVAL, - "Runtime directory is not valid, ignoring assignment: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, 0, "Runtime directory is not valid, ignoring assignment: %s", rvalue); continue; } @@ -3104,8 +3569,7 @@ int config_parse_runtime_directory( n = NULL; } if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } @@ -3152,15 +3616,13 @@ int config_parse_set_status( val = signal_from_string_try_harder(temp); if (val <= 0) { - log_syntax(unit, LOG_ERR, filename, line, -val, - "Failed to parse value, ignoring: %s", word); + 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, ERANGE, - "Value %d is outside range 0-255, ignoring", val); + log_syntax(unit, LOG_ERR, filename, line, 0, "Value %d is outside range 0-255, ignoring", val); continue; } set = &status_set->status; @@ -3172,14 +3634,12 @@ int config_parse_set_status( r = set_put(*set, INT_TO_PTR(val)); if (r < 0) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Unable to store: %s", word); + 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, EINVAL, - "Trailing garbage, ignoring."); + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } @@ -3197,8 +3657,8 @@ int config_parse_namespace_path_strv( void *userdata) { char*** sv = data; - const char *word, *state; - size_t l; + const char *prev; + const char *cur; int r; assert(filename); @@ -3208,42 +3668,47 @@ int config_parse_namespace_path_strv( if (isempty(rvalue)) { /* Empty assignment resets the list */ - strv_free(*sv); - *sv = NULL; + *sv = strv_free(*sv); return 0; } - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - _cleanup_free_ char *n; + prev = cur = rvalue; + for (;;) { + _cleanup_free_ char *word = NULL; int offset; - n = strndup(word, l); - if (!n) + 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(n)) { - log_invalid_utf8(unit, LOG_ERR, filename, line, EINVAL, rvalue); + if (!utf8_is_valid(word)) { + log_syntax_invalid_utf8(unit, LOG_ERR, filename, line, word); + prev = cur; continue; } - offset = n[0] == '-'; - if (!path_is_absolute(n + offset)) { - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Not an absolute path, ignoring: %s", rvalue); + 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(n); + path_kill_slashes(word + offset); - r = strv_push(sv, n); + r = strv_push(sv, word); if (r < 0) return log_oom(); - n = NULL; + prev = cur; + word = NULL; } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); return 0; } @@ -3270,8 +3735,7 @@ int config_parse_no_new_privileges( k = parse_boolean(rvalue); if (k < 0) { - log_syntax(unit, LOG_ERR, filename, line, -k, - "Failed to parse boolean value, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse boolean value, ignoring: %s", rvalue); return 0; } @@ -3302,7 +3766,7 @@ int config_parse_protect_home( assert(data); /* Our enum shall be a superset of booleans, hence first try - * to parse as as boolean, and then as enum */ + * to parse as boolean, and then as enum */ k = parse_boolean(rvalue); if (k > 0) @@ -3313,9 +3777,8 @@ int config_parse_protect_home( ProtectHome h; h = protect_home_from_string(rvalue); - if (h < 0){ - log_syntax(unit, LOG_ERR, filename, line, -h, - "Failed to parse protect home value, ignoring: %s", rvalue); + if (h < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse protect home value, ignoring: %s", rvalue); return 0; } @@ -3346,7 +3809,7 @@ int config_parse_protect_system( assert(data); /* Our enum shall be a superset of booleans, hence first try - * to parse as as boolean, and then as enum */ + * to parse as boolean, and then as enum */ k = parse_boolean(rvalue); if (k > 0) @@ -3357,9 +3820,8 @@ int config_parse_protect_system( ProtectSystem s; s = protect_system_from_string(rvalue); - if (s < 0){ - log_syntax(unit, LOG_ERR, filename, line, -s, - "Failed to parse protect system value, ignoring: %s", rvalue); + if (s < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse protect system value, ignoring: %s", rvalue); return 0; } @@ -3372,10 +3834,10 @@ int config_parse_protect_system( #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; - char *id = NULL; assert(filename); assert(*filename); @@ -3397,7 +3859,6 @@ static int open_follow(char **filename, FILE **_f, Set *names, char **_final) { * 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); @@ -3437,6 +3898,7 @@ static int open_follow(char **filename, FILE **_f, Set *names, char **_final) { *_f = f; *_final = id; + return 0; } @@ -3461,7 +3923,19 @@ static int merge_by_names(Unit **u, Set *names, const char *id) { * ours? Then let's try it the other way * round */ - other = manager_get_unit((*u)->manager, k); + /* 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) { @@ -3485,13 +3959,13 @@ static int merge_by_names(Unit **u, Set *names, const char *id) { } static int load_from_path(Unit *u, const char *path) { - int r; _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); @@ -3508,9 +3982,7 @@ static int load_from_path(Unit *u, const char *path) { r = open_follow(&filename, &f, symlink_names, &id); if (r < 0) { - free(filename); - filename = NULL; - + filename = mfree(filename); if (r != -ENOENT) return r; } @@ -3518,7 +3990,7 @@ static int load_from_path(Unit *u, const char *path) { } else { char **p; - STRV_FOREACH(p, u->manager->lookup_paths.unit_path) { + 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 @@ -3532,20 +4004,22 @@ static int load_from_path(Unit *u, const char *path) { 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; - if (r < 0) { - free(filename); - filename = NULL; - - if (r != -ENOENT) - return r; - - /* Empty the symlink names for the next run */ - set_clear_free(symlink_names); - continue; - } - - break; + /* Empty the symlink names for the next run */ + set_clear_free(symlink_names); } } @@ -3553,6 +4027,11 @@ static int load_from_path(Unit *u, const char *path) { /* 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) @@ -3566,10 +4045,12 @@ static int load_from_path(Unit *u, const char *path) { if (fstat(fileno(f), &st) < 0) return -errno; - if (null_or_empty(&st)) + if (null_or_empty(&st)) { u->load_state = UNIT_MASKED; - else { + 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, @@ -3584,8 +4065,6 @@ static int load_from_path(Unit *u, const char *path) { u->fragment_path = filename; filename = NULL; - u->fragment_mtime = timespec_load(&st.st_mtim); - if (u->source_path) { if (stat(u->source_path, &st) >= 0) u->source_mtime = timespec_load(&st.st_mtim); @@ -3605,6 +4084,11 @@ int unit_load_fragment(Unit *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 */ @@ -3635,13 +4119,11 @@ int unit_load_fragment(Unit *u) { if (r < 0) return r; - if (u->load_state == UNIT_STUB) { + 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. */ - free(u->fragment_path); - u->fragment_path = NULL; - } + u->fragment_path = mfree(u->fragment_path); } /* Look for a template */ @@ -3691,7 +4173,7 @@ void unit_dump_config_items(FILE *f) { { config_parse_int, "INTEGER" }, { config_parse_unsigned, "UNSIGNED" }, { config_parse_iec_size, "SIZE" }, - { config_parse_iec_off, "SIZE" }, + { config_parse_iec_uint64, "SIZE" }, { config_parse_si_size, "SIZE" }, { config_parse_bool, "BOOLEAN" }, { config_parse_string, "STRING" }, @@ -3711,9 +4193,8 @@ void unit_dump_config_items(FILE *f) { { config_parse_input, "INPUT" }, { config_parse_log_facility, "FACILITY" }, { config_parse_log_level, "LEVEL" }, - { config_parse_exec_capabilities, "CAPABILITIES" }, { config_parse_exec_secure_bits, "SECUREBITS" }, - { config_parse_bounding_set, "BOUNDINGSET" }, + { config_parse_capability_set, "BOUNDINGSET" }, { config_parse_limit, "LIMIT" }, { config_parse_unit_deps, "UNIT [...]" }, { config_parse_exec, "PATH [ARGUMENT [...]]" }, @@ -3723,7 +4204,7 @@ void unit_dump_config_items(FILE *f) { { config_parse_sysv_priority, "SYSVPRIORITY" }, #endif { config_parse_kill_mode, "KILLMODE" }, - { config_parse_kill_signal, "SIGNAL" }, + { config_parse_signal, "SIGNAL" }, { config_parse_socket_listen, "SOCKET [...]" }, { config_parse_socket_bind, "SOCKETBIND" }, { config_parse_socket_bindtodevice, "NETWORKINTERFACE" }, @@ -3755,9 +4236,13 @@ void unit_dump_config_items(FILE *f) { { 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" }, diff --git a/src/core/load-fragment.h b/src/core/load-fragment.h index ce10d03c3f..8b688740cf 100644 --- a/src/core/load-fragment.h +++ b/src/core/load-fragment.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -31,12 +29,14 @@ 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); @@ -52,9 +52,8 @@ int config_parse_exec_io_priority(const char *unit, const char *filename, unsign 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_capabilities(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_bounding_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_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); @@ -67,7 +66,6 @@ int config_parse_service_sockets(const char *unit, const char *filename, unsigne 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_bus_endpoint_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_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); @@ -81,18 +79,24 @@ int config_parse_syscall_filter(const char *unit, const char *filename, unsigned 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_personality(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); @@ -104,6 +108,12 @@ int config_parse_cpu_quota(const char *unit, const char *filename, unsigned line 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, unsigned length); diff --git a/src/core/locale-setup.c b/src/core/locale-setup.c index 108072ca9f..ccf61d29fb 100644 --- a/src/core/locale-setup.c +++ b/src/core/locale-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,23 +17,24 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <stdlib.h> #include <errno.h> +#include <stdlib.h> +#include "env-util.h" +#include "fileio.h" #include "locale-setup.h" +#include "locale-util.h" +#include "string-util.h" +#include "strv.h" #include "util.h" #include "virt.h" -#include "fileio.h" -#include "strv.h" -#include "env-util.h" -#include "locale-util.h" int locale_setup(char ***environment) { char **add; char *variables[_VARIABLE_LC_MAX] = {}; int r = 0, i; - if (detect_container(NULL) <= 0) { + if (detect_container() <= 0) { r = parse_env_file("/proc/cmdline", WHITESPACE, "locale.LANG", &variables[VARIABLE_LANG], "locale.LANGUAGE", &variables[VARIABLE_LANGUAGE], diff --git a/src/core/locale-setup.h b/src/core/locale-setup.h index 62c654c37c..3b97497afe 100644 --- a/src/core/locale-setup.h +++ b/src/core/locale-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/loopback-setup.c b/src/core/loopback-setup.c index 4503fc9dcc..04062a7910 100644 --- a/src/core/loopback-setup.c +++ b/src/core/loopback-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -23,12 +21,13 @@ #include <stdlib.h> #include "sd-netlink.h" -#include "netlink-util.h" -#include "missing.h" + #include "loopback-setup.h" +#include "missing.h" +#include "netlink-util.h" static int start_loopback(sd_netlink *rtnl) { - _cleanup_netlink_message_unref_ sd_netlink_message *req = NULL; + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; int r; r = sd_rtnl_message_new_link(rtnl, &req, RTM_SETLINK, LOOPBACK_IFINDEX); @@ -47,7 +46,7 @@ static int start_loopback(sd_netlink *rtnl) { } static bool check_loopback(sd_netlink *rtnl) { - _cleanup_netlink_message_unref_ sd_netlink_message *req = NULL, *reply = NULL; + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL, *reply = NULL; unsigned flags; int r; @@ -67,7 +66,7 @@ static bool check_loopback(sd_netlink *rtnl) { } int loopback_setup(void) { - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; + _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL; int r; r = sd_netlink_open(&rtnl); diff --git a/src/core/loopback-setup.h b/src/core/loopback-setup.h index dd83cf13a7..e7547b8a26 100644 --- a/src/core/loopback-setup.h +++ b/src/core/loopback-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/machine-id-setup.c b/src/core/machine-id-setup.c index 8e26362546..76dfcfa6d7 100644 --- a/src/core/machine-id-setup.c +++ b/src/core/machine-id-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,196 +17,92 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <unistd.h> -#include <stdio.h> -#include <errno.h> -#include <string.h> #include <fcntl.h> +#include <sched.h> #include <sys/mount.h> +#include <unistd.h> -#include "systemd/sd-id128.h" +#include "sd-id128.h" +#include "alloc-util.h" +#include "fd-util.h" +#include "fs-util.h" +#include "id128-util.h" +#include "log.h" #include "machine-id-setup.h" #include "macro.h" -#include "util.h" #include "mkdir.h" -#include "log.h" -#include "virt.h" -#include "fileio.h" +#include "mount-util.h" #include "path-util.h" #include "process-util.h" +#include "stat-util.h" +#include "string-util.h" +#include "umask-util.h" +#include "util.h" +#include "virt.h" -static int shorten_uuid(char destination[34], const char source[36]) { - unsigned i, j; - - assert(destination); - assert(source); - - /* Converts a UUID into a machine ID, by lowercasing it and - * removing dashes. Validates everything. */ - - for (i = 0, j = 0; i < 36 && j < 32; i++) { - int t; - - t = unhexchar(source[i]); - if (t < 0) - continue; - - destination[j++] = hexchar(t); - } - - if (i != 36 || j != 32) - return -EINVAL; - - destination[32] = '\n'; - destination[33] = 0; - return 0; -} - -static int read_machine_id(int fd, char id[34]) { - char id_to_validate[34]; +static int generate_machine_id(const char *root, sd_id128_t *ret) { + const char *dbus_machine_id; + _cleanup_close_ int fd = -1; int r; - assert(fd >= 0); - assert(id); - - /* Reads a machine ID from a file, validates it, and returns - * it. The returned ID ends in a newline. */ - - r = loop_read_exact(fd, id_to_validate, 33, false); - if (r < 0) - return r; - - if (id_to_validate[32] != '\n') - return -EINVAL; - - id_to_validate[32] = 0; - - if (!id128_is_valid(id_to_validate)) - return -EINVAL; - - memcpy(id, id_to_validate, 32); - id[32] = '\n'; - id[33] = 0; - return 0; -} - -static int write_machine_id(int fd, char id[34]) { - assert(fd >= 0); - assert(id); - - if (lseek(fd, 0, SEEK_SET) < 0) - return -errno; - - return loop_write(fd, id, 33, false); -} - -static int generate_machine_id(char id[34], const char *root) { - int fd, r; - unsigned char *p; - sd_id128_t buf; - char *q; - const char *vm_id, *dbus_machine_id; - - assert(id); - - if (isempty(root)) - dbus_machine_id = "/var/lib/dbus/machine-id"; - else - dbus_machine_id = strjoina(root, "/var/lib/dbus/machine-id"); + assert(ret); /* First, try reading the D-Bus machine id, unless it is a symlink */ + dbus_machine_id = prefix_roota(root, "/var/lib/dbus/machine-id"); fd = open(dbus_machine_id, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); if (fd >= 0) { - r = read_machine_id(fd, id); - safe_close(fd); - - if (r >= 0) { + if (id128_read_fd(fd, ID128_PLAIN, ret) >= 0) { log_info("Initializing machine ID from D-Bus machine ID."); return 0; } + + fd = safe_close(fd); } if (isempty(root)) { /* If that didn't work, see if we are running in a container, * and a machine ID was passed in via $container_uuid the way * libvirt/LXC does it */ - r = detect_container(NULL); - if (r > 0) { + + if (detect_container() > 0) { _cleanup_free_ char *e = NULL; - r = getenv_for_pid(1, "container_uuid", &e); - if (r > 0) { - r = shorten_uuid(id, e); - if (r >= 0) { - log_info("Initializing machine ID from container UUID."); - return 0; - } + if (getenv_for_pid(1, "container_uuid", &e) > 0 && + sd_id128_from_string(e, ret) >= 0) { + log_info("Initializing machine ID from container UUID."); + return 0; } - } else { + } else if (detect_vm() == VIRTUALIZATION_KVM) { + /* If we are not running in a container, see if we are * running in qemu/kvm and a machine ID was passed in * via -uuid on the qemu/kvm command line */ - r = detect_vm(&vm_id); - if (r > 0 && streq(vm_id, "kvm")) { - char uuid[36]; - - fd = open("/sys/class/dmi/id/product_uuid", O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); - if (fd >= 0) { - r = loop_read_exact(fd, uuid, 36, false); - safe_close(fd); - - if (r >= 0) { - r = shorten_uuid(id, uuid); - if (r >= 0) { - log_info("Initializing machine ID from KVM UUID."); - return 0; - } - } - } + if (id128_read("/sys/class/dmi/id/product_uuid", ID128_UUID, ret) >= 0) { + log_info("Initializing machine ID from KVM UUID."); + return 0; } } } /* If that didn't work, generate a random machine id */ - r = sd_id128_randomize(&buf); + r = sd_id128_randomize(ret); if (r < 0) - return log_error_errno(r, "Failed to open /dev/urandom: %m"); - - for (p = buf.bytes, q = id; p < buf.bytes + sizeof(buf); p++, q += 2) { - q[0] = hexchar(*p >> 4); - q[1] = hexchar(*p & 15); - } - - id[32] = '\n'; - id[33] = 0; + return log_error_errno(r, "Failed to generate randomized : %m"); log_info("Initializing machine ID from random generator."); - return 0; } -int machine_id_setup(const char *root) { +int machine_id_setup(const char *root, sd_id128_t machine_id, sd_id128_t *ret) { const char *etc_machine_id, *run_machine_id; _cleanup_close_ int fd = -1; - bool writable = true; - char id[34]; /* 32 + \n + \0 */ + bool writable; int r; - if (isempty(root)) { - etc_machine_id = "/etc/machine-id"; - run_machine_id = "/run/machine-id"; - } else { - char *x; - - x = strjoina(root, "/etc/machine-id"); - etc_machine_id = path_kill_slashes(x); - - x = strjoina(root, "/run/machine-id"); - run_machine_id = path_kill_slashes(x); - } + etc_machine_id = prefix_roota(root, "/etc/machine-id"); RUN_WITH_UMASK(0000) { /* We create this 0444, to indicate that this isn't really @@ -216,7 +110,7 @@ int machine_id_setup(const char *root) { * will be owned by root it doesn't matter much, but maybe * people look. */ - mkdir_parents(etc_machine_id, 0755); + (void) mkdir_parents(etc_machine_id, 0755); fd = open(etc_machine_id, O_RDWR|O_CREAT|O_CLOEXEC|O_NOCTTY, 0444); if (fd < 0) { int old_errno = errno; @@ -237,31 +131,38 @@ int machine_id_setup(const char *root) { } writable = false; - } + } else + writable = true; } - if (read_machine_id(fd, id) >= 0) - return 0; + /* A we got a valid machine ID argument, that's what counts */ + if (sd_id128_is_null(machine_id)) { - /* Hmm, so, the id currently stored is not useful, then let's - * generate one */ + /* Try to read any existing machine ID */ + if (id128_read_fd(fd, ID128_PLAIN, ret) >= 0) + return 0; - r = generate_machine_id(id, root); - if (r < 0) - return r; + /* Hmm, so, the id currently stored is not useful, then let's generate one */ + r = generate_machine_id(root, &machine_id); + if (r < 0) + return r; + + if (lseek(fd, 0, SEEK_SET) == (off_t) -1) + return log_error_errno(errno, "Failed to seek: %m"); + } if (writable) - if (write_machine_id(fd, id) >= 0) - return 0; + if (id128_write_fd(fd, ID128_PLAIN, machine_id, true) >= 0) + goto finish; fd = safe_close(fd); - /* Hmm, we couldn't write it? So let's write it to - * /run/machine-id as a replacement */ + /* Hmm, we couldn't write it? So let's write it to /run/machine-id as a replacement */ - RUN_WITH_UMASK(0022) { - r = write_string_file(run_machine_id, id, WRITE_STRING_FILE_CREATE); - } + run_machine_id = prefix_roota(root, "/run/machine-id"); + + RUN_WITH_UMASK(0022) + r = id128_write(run_machine_id, ID128_PLAIN, machine_id, false); if (r < 0) { (void) unlink(run_machine_id); return log_error_errno(r, "Cannot write %s: %m", run_machine_id); @@ -277,7 +178,11 @@ int machine_id_setup(const char *root) { /* Mark the mount read-only */ if (mount(NULL, etc_machine_id, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, NULL) < 0) - log_warning_errno(errno, "Failed to make transient %s read-only: %m", etc_machine_id); + log_warning_errno(errno, "Failed to make transient %s read-only, ignoring: %m", etc_machine_id); + +finish: + if (ret) + *ret = machine_id; return 0; } @@ -285,23 +190,20 @@ int machine_id_setup(const char *root) { int machine_id_commit(const char *root) { _cleanup_close_ int fd = -1, initial_mntns_fd = -1; const char *etc_machine_id; - char id[34]; /* 32 + \n + \0 */ + sd_id128_t id; int r; - if (isempty(root)) - etc_machine_id = "/etc/machine-id"; - else { - char *x; + /* Replaces a tmpfs bind mount of /etc/machine-id by a proper file, atomically. For this, the umount is removed + * in a mount namespace, a new file is created at the right place. Afterwards the mount is also removed in the + * original mount namespace, thus revealing the file that was just created. */ - x = strjoina(root, "/etc/machine-id"); - etc_machine_id = path_kill_slashes(x); - } + etc_machine_id = prefix_roota(root, "/etc/machine-id"); r = path_is_mount_point(etc_machine_id, 0); if (r < 0) return log_error_errno(r, "Failed to determine whether %s is a mount point: %m", etc_machine_id); if (r == 0) { - log_debug("%s is is not a mount point. Nothing to do.", etc_machine_id); + log_debug("%s is not a mount point. Nothing to do.", etc_machine_id); return 0; } @@ -310,10 +212,6 @@ int machine_id_commit(const char *root) { if (fd < 0) return log_error_errno(errno, "Cannot open %s: %m", etc_machine_id); - r = read_machine_id(fd, id); - if (r < 0) - return log_error_errno(r, "We didn't find a valid machine ID in %s.", etc_machine_id); - r = fd_is_temporary_fs(fd); if (r < 0) return log_error_errno(r, "Failed to determine whether %s is on a temporary file system: %m", etc_machine_id); @@ -322,10 +220,14 @@ int machine_id_commit(const char *root) { return -EROFS; } + r = id128_read_fd(fd, ID128_PLAIN, &id); + if (r < 0) + return log_error_errno(r, "We didn't find a valid machine ID in %s.", etc_machine_id); + fd = safe_close(fd); /* Store current mount namespace */ - r = namespace_open(0, NULL, &initial_mntns_fd, NULL, NULL); + r = namespace_open(0, NULL, &initial_mntns_fd, NULL, NULL, NULL); if (r < 0) return log_error_errno(r, "Can't fetch current mount namespace: %m"); @@ -340,18 +242,12 @@ int machine_id_commit(const char *root) { return log_error_errno(errno, "Failed to unmount transient %s file in our private namespace: %m", etc_machine_id); /* Update a persistent version of etc_machine_id */ - fd = open(etc_machine_id, O_RDWR|O_CREAT|O_CLOEXEC|O_NOCTTY, 0444); - if (fd < 0) - return log_error_errno(errno, "Cannot open for writing %s. This is mandatory to get a persistent machine-id: %m", etc_machine_id); - - r = write_machine_id(fd, id); + r = id128_write(etc_machine_id, ID128_PLAIN, id, true); if (r < 0) - return log_error_errno(r, "Cannot write %s: %m", etc_machine_id); - - fd = safe_close(fd); + return log_error_errno(r, "Cannot write %s. This is mandatory to get a persistent machine ID: %m", etc_machine_id); /* Return to initial namespace and proceed a lazy tmpfs unmount */ - r = namespace_enter(-1, initial_mntns_fd, -1, -1); + r = namespace_enter(-1, initial_mntns_fd, -1, -1, -1); if (r < 0) return log_warning_errno(r, "Failed to switch back to initial mount namespace: %m.\nWe'll keep transient %s file until next reboot.", etc_machine_id); diff --git a/src/core/machine-id-setup.h b/src/core/machine-id-setup.h index f7707c3bf9..29f4620646 100644 --- a/src/core/machine-id-setup.h +++ b/src/core/machine-id-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,4 +20,4 @@ ***/ int machine_id_commit(const char *root); -int machine_id_setup(const char *root); +int machine_id_setup(const char *root, sd_id128_t requested, sd_id128_t *ret); diff --git a/src/core/macros.systemd.in b/src/core/macros.systemd.in index bea6ef1da3..6e8a3b3e3d 100644 --- a/src/core/macros.systemd.in +++ b/src/core/macros.systemd.in @@ -1,4 +1,4 @@ -# -*- Mode: makefile; indent-tabs-mode: t -*- */ +# -*- Mode: rpm-spec; indent-tabs-mode: nil -*- */ # # This file is part of systemd. # @@ -29,6 +29,8 @@ %_sysusersdir @sysusersdir@ %_sysctldir @sysctldir@ %_binfmtdir @binfmtdir@ +%_systemdgeneratordir @systemgeneratordir@ +%_systemdusergeneratordir @usergeneratordir@ %systemd_requires \ Requires(post): systemd \ @@ -36,20 +38,25 @@ Requires(preun): systemd \ Requires(postun): systemd \ %{nil} +%systemd_ordering \ +OrderWithRequires(post): systemd \ +OrderWithRequires(preun): systemd \ +OrderWithRequires(postun): systemd \ +%{nil} + %systemd_post() \ if [ $1 -eq 1 ] ; then \ # Initial installation \ - systemctl preset %{?*} >/dev/null 2>&1 || : \ + systemctl --no-reload preset %{?*} >/dev/null 2>&1 || : \ fi \ %{nil} -%systemd_user_post() %systemd_post --user --global %{?*} +%systemd_user_post() %{expand:%systemd_post \\--user \\--global %%{?*}} %systemd_preun() \ if [ $1 -eq 0 ] ; then \ # Package removal, not upgrade \ - systemctl --no-reload disable %{?*} > /dev/null 2>&1 || : \ - systemctl stop %{?*} > /dev/null 2>&1 || : \ + systemctl --no-reload disable --now %{?*} > /dev/null 2>&1 || : \ fi \ %{nil} @@ -60,14 +67,11 @@ if [ $1 -eq 0 ] ; then \ fi \ %{nil} -%systemd_postun() \ -systemctl daemon-reload >/dev/null 2>&1 || : \ -%{nil} +%systemd_postun() %{nil} %systemd_user_postun() %{nil} %systemd_postun_with_restart() \ -systemctl daemon-reload >/dev/null 2>&1 || : \ if [ $1 -ge 1 ] ; then \ # Package upgrade, not uninstall \ systemctl try-restart %{?*} >/dev/null 2>&1 || : \ diff --git a/src/core/main.c b/src/core/main.c index 6ae8b51544..61f3828a36 100644 --- a/src/core/main.c +++ b/src/core/main.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,81 +17,96 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <stdio.h> #include <errno.h> -#include <string.h> -#include <unistd.h> -#include <sys/stat.h> +#include <fcntl.h> #include <getopt.h> #include <signal.h> -#include <fcntl.h> -#include <sys/prctl.h> +#include <stdio.h> +#include <string.h> #include <sys/mount.h> - -#ifdef HAVE_VALGRIND_VALGRIND_H -#include <valgrind/valgrind.h> -#endif +#include <sys/prctl.h> +#include <sys/reboot.h> +#include <sys/stat.h> +#include <unistd.h> #ifdef HAVE_SECCOMP #include <seccomp.h> #endif +#ifdef HAVE_VALGRIND_VALGRIND_H +#include <valgrind/valgrind.h> +#endif -#include "sd-daemon.h" #include "sd-bus.h" -#include "log.h" -#include "fdset.h" -#include "special.h" -#include "conf-parser.h" -#include "missing.h" -#include "pager.h" +#include "sd-daemon.h" + +#include "alloc-util.h" +#include "architecture.h" #include "build.h" -#include "strv.h" +#include "bus-error.h" +#include "bus-util.h" +#include "capability-util.h" +#include "clock-util.h" +#include "conf-parser.h" +#include "cpu-set-util.h" +#include "dbus-manager.h" #include "def.h" -#include "virt.h" -#include "architecture.h" -#include "watchdog.h" -#include "switch-root.h" -#include "capability.h" -#include "killall.h" #include "env-util.h" -#include "clock-util.h" +#include "fd-util.h" +#include "fdset.h" #include "fileio.h" -#include "bus-error.h" -#include "bus-util.h" -#include "selinux-util.h" #include "formats-util.h" -#include "process-util.h" -#include "terminal-util.h" -#include "signal-util.h" -#include "manager.h" -#include "dbus-manager.h" +#include "fs-util.h" +#include "hostname-setup.h" +#include "ima-setup.h" +#include "killall.h" +#include "kmod-setup.h" #include "load-fragment.h" - -#include "mount-setup.h" +#include "log.h" #include "loopback-setup.h" -#include "hostname-setup.h" #include "machine-id-setup.h" +#include "manager.h" +#include "missing.h" +#include "mount-setup.h" +#include "pager.h" +#include "parse-util.h" +#include "proc-cmdline.h" +#include "process-util.h" +#include "raw-clone.h" +#include "rlimit-util.h" +#ifdef HAVE_SECCOMP +#include "seccomp-util.h" +#endif #include "selinux-setup.h" -#include "ima-setup.h" +#include "selinux-util.h" +#include "signal-util.h" #include "smack-setup.h" -#include "kmod-setup.h" +#include "special.h" +#include "stat-util.h" +#include "stdio-util.h" +#include "strv.h" +#include "switch-root.h" +#include "terminal-util.h" +#include "umask-util.h" +#include "user-util.h" +#include "virt.h" +#include "watchdog.h" static enum { ACTION_RUN, ACTION_HELP, ACTION_VERSION, ACTION_TEST, - ACTION_DUMP_CONFIGURATION_ITEMS, - ACTION_DONE + ACTION_DUMP_CONFIGURATION_ITEMS } arg_action = ACTION_RUN; static char *arg_default_unit = NULL; -static ManagerRunningAs arg_running_as = _MANAGER_RUNNING_AS_INVALID; +static bool arg_system = false; static bool arg_dump_core = true; -static bool arg_crash_shell = false; static int arg_crash_chvt = -1; +static bool arg_crash_shell = false; +static bool arg_crash_reboot = false; static bool arg_confirm_spawn = false; static ShowStatus arg_show_status = _SHOW_STATUS_UNSET; static bool arg_switched_root = false; -static int arg_no_pager = -1; +static bool arg_no_pager = false; static char ***arg_join_controllers = NULL; static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL; static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT; @@ -106,68 +119,76 @@ static usec_t arg_runtime_watchdog = 0; static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE; static char **arg_default_environment = NULL; static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {}; -static uint64_t arg_capability_bounding_set_drop = 0; +static uint64_t arg_capability_bounding_set = CAP_ALL; static nsec_t arg_timer_slack_nsec = NSEC_INFINITY; static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE; static Set* arg_syscall_archs = NULL; static FILE* arg_serialization = NULL; static bool arg_default_cpu_accounting = false; +static bool arg_default_io_accounting = false; static bool arg_default_blockio_accounting = false; static bool arg_default_memory_accounting = false; +static bool arg_default_tasks_accounting = true; +static uint64_t arg_default_tasks_max = UINT64_MAX; +static sd_id128_t arg_machine_id = {}; +static CADBurstAction arg_cad_burst_action = CAD_BURST_ACTION_REBOOT; -static void nop_handler(int sig) {} +noreturn static void freeze_or_reboot(void) { -static void pager_open_if_enabled(void) { + if (arg_crash_reboot) { + log_notice("Rebooting in 10s..."); + (void) sleep(10); - if (arg_no_pager <= 0) - return; + log_notice("Rebooting now..."); + (void) reboot(RB_AUTOBOOT); + log_emergency_errno(errno, "Failed to reboot: %m"); + } - pager_open(false); + log_emergency("Freezing execution."); + freeze(); } noreturn static void crash(int sig) { + struct sigaction sa; + pid_t pid; if (getpid() != 1) /* Pass this on immediately, if this is not PID 1 */ - raise(sig); + (void) raise(sig); else if (!arg_dump_core) log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig)); else { - struct sigaction sa = { - .sa_handler = nop_handler, + sa = (struct sigaction) { + .sa_handler = nop_signal_handler, .sa_flags = SA_NOCLDSTOP|SA_RESTART, }; - pid_t pid; /* We want to wait for the core process, hence let's enable SIGCHLD */ - sigaction(SIGCHLD, &sa, NULL); + (void) sigaction(SIGCHLD, &sa, NULL); - pid = raw_clone(SIGCHLD, NULL); + pid = raw_clone(SIGCHLD); if (pid < 0) log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig)); - else if (pid == 0) { - struct rlimit rl = {}; - /* Enable default signal handler for core dump */ - zero(sa); - sa.sa_handler = SIG_DFL; - sigaction(sig, &sa, NULL); - /* Don't limit the core dump size */ - rl.rlim_cur = RLIM_INFINITY; - rl.rlim_max = RLIM_INFINITY; - setrlimit(RLIMIT_CORE, &rl); + sa = (struct sigaction) { + .sa_handler = SIG_DFL, + }; + (void) sigaction(sig, &sa, NULL); + + /* Don't limit the coredump size */ + (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)); /* Just to be sure... */ (void) chdir("/"); /* Raise the signal again */ pid = raw_getpid(); - kill(pid, sig); /* raise() would kill the parent */ + (void) kill(pid, sig); /* raise() would kill the parent */ assert_not_reached("We shouldn't be here..."); - _exit(1); + _exit(EXIT_FAILURE); } else { siginfo_t status; int r; @@ -182,44 +203,45 @@ noreturn static void crash(int sig) { pid, sigchld_code_to_string(status.si_code), status.si_status, strna(status.si_code == CLD_EXITED - ? exit_status_to_string(status.si_status, EXIT_STATUS_FULL) + ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL) : signal_to_string(status.si_status))); else log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid); } } - if (arg_crash_chvt) - chvt(arg_crash_chvt); + if (arg_crash_chvt >= 0) + (void) chvt(arg_crash_chvt); - if (arg_crash_shell) { - struct sigaction sa = { - .sa_handler = SIG_IGN, - .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART, - }; - pid_t pid; + sa = (struct sigaction) { + .sa_handler = SIG_IGN, + .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART, + }; - log_info("Executing crash shell in 10s..."); - sleep(10); + /* Let the kernel reap children for us */ + (void) sigaction(SIGCHLD, &sa, NULL); - /* Let the kernel reap children for us */ - assert_se(sigaction(SIGCHLD, &sa, NULL) == 0); + if (arg_crash_shell) { + log_notice("Executing crash shell in 10s..."); + (void) sleep(10); - pid = raw_clone(SIGCHLD, NULL); + pid = raw_clone(SIGCHLD); if (pid < 0) log_emergency_errno(errno, "Failed to fork off crash shell: %m"); else if (pid == 0) { - make_console_stdio(); - execle("/bin/sh", "/bin/sh", NULL, environ); + (void) setsid(); + (void) make_console_stdio(); + (void) execle("/bin/sh", "/bin/sh", NULL, environ); log_emergency_errno(errno, "execle() failed: %m"); - _exit(1); - } else - log_info("Successfully spawned crash shell as PID "PID_FMT".", pid); + _exit(EXIT_FAILURE); + } else { + log_info("Spawned crash shell as PID "PID_FMT".", pid); + (void) wait_for_terminate(pid, NULL); + } } - log_emergency("Freezing execution."); - freeze(); + freeze_or_reboot(); } static void install_crash_handler(void) { @@ -253,37 +275,40 @@ static int console_setup(void) { return 0; } -static int set_default_unit(const char *u) { - char *c; +static int parse_crash_chvt(const char *value) { + int b; - assert(u); + if (safe_atoi(value, &arg_crash_chvt) >= 0) + return 0; - c = strdup(u); - if (!c) - return -ENOMEM; + b = parse_boolean(value); + if (b < 0) + return b; + + if (b > 0) + arg_crash_chvt = 0; /* switch to where kmsg goes */ + else + arg_crash_chvt = -1; /* turn off switching */ + + return 0; +} + +static int set_machine_id(const char *m) { + sd_id128_t t; + assert(m); + + if (sd_id128_from_string(m, &t) < 0) + return -EINVAL; - free(arg_default_unit); - arg_default_unit = c; + if (sd_id128_is_null(t)) + return -EINVAL; + arg_machine_id = t; return 0; } static int parse_proc_cmdline_item(const char *key, const char *value) { - static const char * const rlmap[] = { - "emergency", SPECIAL_EMERGENCY_TARGET, - "-b", SPECIAL_EMERGENCY_TARGET, - "rescue", SPECIAL_RESCUE_TARGET, - "single", SPECIAL_RESCUE_TARGET, - "-s", SPECIAL_RESCUE_TARGET, - "s", SPECIAL_RESCUE_TARGET, - "S", SPECIAL_RESCUE_TARGET, - "1", SPECIAL_RESCUE_TARGET, - "2", SPECIAL_MULTI_USER_TARGET, - "3", SPECIAL_MULTI_USER_TARGET, - "4", SPECIAL_MULTI_USER_TARGET, - "5", SPECIAL_GRAPHICAL_TARGET, - }; int r; assert(key); @@ -291,12 +316,12 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { if (streq(key, "systemd.unit") && value) { if (!in_initrd()) - return set_default_unit(value); + return free_and_strdup(&arg_default_unit, value); } else if (streq(key, "rd.systemd.unit") && value) { if (in_initrd()) - return set_default_unit(value); + return free_and_strdup(&arg_default_unit, value); } else if (streq(key, "systemd.dump_core") && value) { @@ -306,6 +331,11 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { else arg_dump_core = r; + } else if (streq(key, "systemd.crash_chvt") && value) { + + if (parse_crash_chvt(value) < 0) + log_warning("Failed to parse crash chvt switch %s. Ignoring.", value); + } else if (streq(key, "systemd.crash_shell") && value) { r = parse_boolean(value); @@ -314,12 +344,13 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { else arg_crash_shell = r; - } else if (streq(key, "systemd.crash_chvt") && value) { + } else if (streq(key, "systemd.crash_reboot") && value) { - if (safe_atoi(value, &r) < 0) - log_warning("Failed to parse crash chvt switch %s. Ignoring.", value); + r = parse_boolean(value); + if (r < 0) + log_warning("Failed to parse crash reboot switch %s. Ignoring.", value); else - arg_crash_chvt = r; + arg_crash_reboot = r; } else if (streq(key, "systemd.confirm_spawn") && value) { @@ -364,6 +395,12 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { } else log_warning("Environment variable name '%s' is not valid. Ignoring.", value); + } else if (streq(key, "systemd.machine_id") && value) { + + r = set_machine_id(value); + if (r < 0) + log_warning("MachineID '%s' is not valid. Ignoring.", value); + } else if (streq(key, "quiet") && !value) { if (arg_show_status == _SHOW_STATUS_UNSET) @@ -374,16 +411,25 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { /* Note that log_parse_environment() handles 'debug' * too, and sets the log level to LOG_DEBUG. */ - if (detect_container(NULL) > 0) + if (detect_container() > 0) log_set_target(LOG_TARGET_CONSOLE); - } else if (!in_initrd() && !value) { - unsigned i; + } else if (!value) { + const char *target; /* SysV compatibility */ - for (i = 0; i < ELEMENTSOF(rlmap); i += 2) - if (streq(key, rlmap[i])) - return set_default_unit(rlmap[i+1]); + target = runlevel_to_target(key); + if (target) + return free_and_strdup(&arg_default_unit, target); + + } else if (streq(key, "systemd.default_timeout_start_sec") && value) { + + r = parse_sec(value, &arg_default_timeout_start_usec); + if (r < 0) + log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value); + + if (arg_default_timeout_start_usec <= 0) + arg_default_timeout_start_usec = USEC_INFINITY; } return 0; @@ -409,9 +455,9 @@ static int parse_proc_cmdline_item(const char *key, const char *value) { \ r = func(rvalue); \ if (r < 0) \ - log_syntax(unit, LOG_ERR, filename, line, -r, \ - "Invalid " descr "'%s': %s", \ - rvalue, strerror(-r)); \ + log_syntax(unit, LOG_ERR, filename, line, r, \ + "Invalid " descr "'%s': %m", \ + rvalue); \ \ return 0; \ } @@ -433,49 +479,15 @@ static int config_parse_cpu_affinity2( void *data, void *userdata) { - const char *word, *state; - size_t l; - cpu_set_t *c = NULL; - unsigned ncpus = 0; + _cleanup_cpu_free_ cpu_set_t *c = NULL; + int ncpus; - assert(filename); - assert(lvalue); - assert(rvalue); + ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue); + if (ncpus < 0) + return ncpus; - FOREACH_WORD_QUOTED(word, l, rvalue, state) { - char *t; - int r; - unsigned cpu; - - if (!(t = strndup(word, l))) - return log_oom(); - - r = safe_atou(t, &cpu); - free(t); - - if (!c) - if (!(c = cpu_set_malloc(&ncpus))) - return log_oom(); - - if (r < 0 || cpu >= ncpus) { - log_syntax(unit, LOG_ERR, filename, line, -r, - "Failed to parse CPU affinity '%s'", rvalue); - CPU_FREE(c); - return -EBADMSG; - } - - CPU_SET_S(cpu, CPU_ALLOC_SIZE(ncpus), c); - } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); - - if (c) { - if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0) - log_warning("Failed to set CPU affinity: %m"); - - CPU_FREE(c); - } + if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0) + log_warning("Failed to set CPU affinity: %m"); return 0; } @@ -502,29 +514,38 @@ static int config_parse_show_status( k = parse_show_status(rvalue, b); if (k < 0) { - log_syntax(unit, LOG_ERR, filename, line, -k, - "Failed to parse show status setting, ignoring: %s", rvalue); + log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue); return 0; } return 0; } -static void strv_free_free(char ***l) { - char ***i; +static int config_parse_crash_chvt( + 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) { - if (!l) - return; + int r; - for (i = l; *i; i++) - strv_free(*i); + assert(filename); + assert(lvalue); + assert(rvalue); - free(l); -} + r = parse_crash_chvt(rvalue); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue); + return 0; + } -static void free_join_controllers(void) { - strv_free_free(arg_join_controllers); - arg_join_controllers = NULL; + return 0; } static int config_parse_join_controllers(const char *unit, @@ -538,26 +559,31 @@ static int config_parse_join_controllers(const char *unit, void *data, void *userdata) { + const char *whole_rvalue = rvalue; unsigned n = 0; - const char *word, *state; - size_t length; assert(filename); assert(lvalue); assert(rvalue); - free_join_controllers(); + arg_join_controllers = strv_free_free(arg_join_controllers); - FOREACH_WORD_QUOTED(word, length, rvalue, state) { - char *s, **l; - - s = strndup(word, length); - if (!s) - return log_oom(); + for (;;) { + _cleanup_free_ char *word = NULL; + char **l; + int r; - l = strv_split(s, ","); - free(s); + r = extract_first_word(&rvalue, &word, WHITESPACE, EXTRACT_QUOTES); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Invalid value for %s: %s", lvalue, whole_rvalue); + return r; + } + if (r == 0) + break; + l = strv_split(word, ","); + if (!l) + return log_oom(); strv_uniq(l); if (strv_length(l) <= 1) { @@ -591,7 +617,7 @@ static int config_parse_join_controllers(const char *unit, for (a = arg_join_controllers; *a; a++) { if (strv_overlap(*a, l)) { - if (strv_extend_strv(&l, *a) < 0) { + if (strv_extend_strv(&l, *a, false) < 0) { strv_free(l); strv_free_free(t); return log_oom(); @@ -617,13 +643,14 @@ static int config_parse_join_controllers(const char *unit, arg_join_controllers = t; } } - if (!isempty(state)) - log_syntax(unit, LOG_ERR, filename, line, EINVAL, - "Trailing garbage, ignoring."); + if (!isempty(rvalue)) + log_syntax(unit, LOG_ERR, filename, line, 0, "Trailing garbage, ignoring."); return 0; } +static DEFINE_CONFIG_PARSE_ENUM(config_parse_cad_burst_action, cad_burst_action, CADBurstAction, "Failed to parse service restart specifier"); + static int parse_config_file(void) { const ConfigTableItem items[] = { @@ -632,14 +659,16 @@ static int parse_config_file(void) { { "Manager", "LogColor", config_parse_color, 0, NULL }, { "Manager", "LogLocation", config_parse_location, 0, NULL }, { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, + { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL }, + { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL }, { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, + { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot }, { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, - { "Manager", "CrashChVT", config_parse_int, 0, &arg_crash_chvt }, { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL }, { "Manager", "JoinControllers", config_parse_join_controllers, 0, &arg_join_controllers }, { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog }, { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog }, - { "Manager", "CapabilityBoundingSet", config_parse_bounding_set, 0, &arg_capability_bounding_set_drop }, + { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set }, #ifdef HAVE_SECCOMP { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, #endif @@ -650,37 +679,54 @@ static int parse_config_file(void) { { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec }, { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec }, { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec }, - { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, + { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */ + { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval }, { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst }, { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment }, - { "Manager", "DefaultLimitCPU", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CPU] }, - { "Manager", "DefaultLimitFSIZE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_FSIZE] }, - { "Manager", "DefaultLimitDATA", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_DATA] }, - { "Manager", "DefaultLimitSTACK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_STACK] }, - { "Manager", "DefaultLimitCORE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_CORE] }, - { "Manager", "DefaultLimitRSS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RSS] }, - { "Manager", "DefaultLimitNOFILE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NOFILE] }, - { "Manager", "DefaultLimitAS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_AS] }, - { "Manager", "DefaultLimitNPROC", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NPROC] }, - { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MEMLOCK] }, - { "Manager", "DefaultLimitLOCKS", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_LOCKS] }, - { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_SIGPENDING] }, - { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_MSGQUEUE] }, - { "Manager", "DefaultLimitNICE", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_NICE] }, - { "Manager", "DefaultLimitRTPRIO", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTPRIO] }, - { "Manager", "DefaultLimitRTTIME", config_parse_limit, 0, &arg_default_rlimit[RLIMIT_RTTIME] }, + { "Manager", "DefaultLimitCPU", config_parse_limit, RLIMIT_CPU, arg_default_rlimit }, + { "Manager", "DefaultLimitFSIZE", config_parse_limit, RLIMIT_FSIZE, arg_default_rlimit }, + { "Manager", "DefaultLimitDATA", config_parse_limit, RLIMIT_DATA, arg_default_rlimit }, + { "Manager", "DefaultLimitSTACK", config_parse_limit, RLIMIT_STACK, arg_default_rlimit }, + { "Manager", "DefaultLimitCORE", config_parse_limit, RLIMIT_CORE, arg_default_rlimit }, + { "Manager", "DefaultLimitRSS", config_parse_limit, RLIMIT_RSS, arg_default_rlimit }, + { "Manager", "DefaultLimitNOFILE", config_parse_limit, RLIMIT_NOFILE, arg_default_rlimit }, + { "Manager", "DefaultLimitAS", config_parse_limit, RLIMIT_AS, arg_default_rlimit }, + { "Manager", "DefaultLimitNPROC", config_parse_limit, RLIMIT_NPROC, arg_default_rlimit }, + { "Manager", "DefaultLimitMEMLOCK", config_parse_limit, RLIMIT_MEMLOCK, arg_default_rlimit }, + { "Manager", "DefaultLimitLOCKS", config_parse_limit, RLIMIT_LOCKS, arg_default_rlimit }, + { "Manager", "DefaultLimitSIGPENDING", config_parse_limit, RLIMIT_SIGPENDING, arg_default_rlimit }, + { "Manager", "DefaultLimitMSGQUEUE", config_parse_limit, RLIMIT_MSGQUEUE, arg_default_rlimit }, + { "Manager", "DefaultLimitNICE", config_parse_limit, RLIMIT_NICE, arg_default_rlimit }, + { "Manager", "DefaultLimitRTPRIO", config_parse_limit, RLIMIT_RTPRIO, arg_default_rlimit }, + { "Manager", "DefaultLimitRTTIME", config_parse_limit, RLIMIT_RTTIME, arg_default_rlimit }, { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_default_cpu_accounting }, + { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting }, { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting }, { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting }, + { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting }, + { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max }, + { "Manager", "CtrlAltDelBurstAction", config_parse_cad_burst_action, 0, &arg_cad_burst_action}, {} }; const char *fn, *conf_dirs_nulstr; - fn = arg_running_as == MANAGER_SYSTEM ? PKGSYSCONFDIR "/system.conf" : PKGSYSCONFDIR "/user.conf"; - conf_dirs_nulstr = arg_running_as == MANAGER_SYSTEM ? CONF_DIRS_NULSTR("systemd/system.conf") : CONF_DIRS_NULSTR("systemd/user.conf"); - config_parse_many(fn, conf_dirs_nulstr, "Manager\0", - config_item_table_lookup, items, false, NULL); + fn = arg_system ? + PKGSYSCONFDIR "/system.conf" : + PKGSYSCONFDIR "/user.conf"; + + conf_dirs_nulstr = arg_system ? + CONF_PATHS_NULSTR("systemd/system.conf.d") : + CONF_PATHS_NULSTR("systemd/user.conf.d"); + + config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, false, NULL); + + /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY + * like everywhere else. */ + if (arg_default_timeout_start_usec <= 0) + arg_default_timeout_start_usec = USEC_INFINITY; + if (arg_default_timeout_stop_usec <= 0) + arg_default_timeout_stop_usec = USEC_INFINITY; return 0; } @@ -698,8 +744,11 @@ static void manager_set_defaults(Manager *m) { m->default_start_limit_interval = arg_default_start_limit_interval; m->default_start_limit_burst = arg_default_start_limit_burst; m->default_cpu_accounting = arg_default_cpu_accounting; + m->default_io_accounting = arg_default_io_accounting; m->default_blockio_accounting = arg_default_blockio_accounting; m->default_memory_accounting = arg_default_memory_accounting; + m->default_tasks_accounting = arg_default_tasks_accounting; + m->default_tasks_max = arg_default_tasks_max; manager_set_default_rlimits(m, arg_default_rlimit); manager_environment_add(m, NULL, arg_default_environment); @@ -720,13 +769,16 @@ static int parse_argv(int argc, char *argv[]) { ARG_VERSION, ARG_DUMP_CONFIGURATION_ITEMS, ARG_DUMP_CORE, + ARG_CRASH_CHVT, ARG_CRASH_SHELL, + ARG_CRASH_REBOOT, ARG_CONFIRM_SPAWN, ARG_SHOW_STATUS, ARG_DESERIALIZE, ARG_SWITCHED_ROOT, ARG_DEFAULT_STD_OUTPUT, - ARG_DEFAULT_STD_ERROR + ARG_DEFAULT_STD_ERROR, + ARG_MACHINE_ID }; static const struct option options[] = { @@ -743,13 +795,16 @@ static int parse_argv(int argc, char *argv[]) { { "version", no_argument, NULL, ARG_VERSION }, { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS }, { "dump-core", optional_argument, NULL, ARG_DUMP_CORE }, + { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT }, { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL }, + { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT }, { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN }, { "show-status", optional_argument, NULL, ARG_SHOW_STATUS }, { "deserialize", required_argument, NULL, ARG_DESERIALIZE }, { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT }, { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, }, { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, }, + { "machine-id", required_argument, NULL, ARG_MACHINE_ID }, {} }; @@ -828,24 +883,22 @@ static int parse_argv(int argc, char *argv[]) { case ARG_UNIT: - r = set_default_unit(optarg); + r = free_and_strdup(&arg_default_unit, optarg); if (r < 0) return log_error_errno(r, "Failed to set default unit %s: %m", optarg); break; case ARG_SYSTEM: - arg_running_as = MANAGER_SYSTEM; + arg_system = true; break; case ARG_USER: - arg_running_as = MANAGER_USER; + arg_system = false; break; case ARG_TEST: arg_action = ACTION_TEST; - if (arg_no_pager < 0) - arg_no_pager = true; break; case ARG_NO_PAGER: @@ -861,21 +914,42 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_DUMP_CORE: - r = optarg ? parse_boolean(optarg) : 1; - if (r < 0) { - log_error("Failed to parse dump core boolean %s.", optarg); - return r; + if (!optarg) + arg_dump_core = true; + else { + r = parse_boolean(optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse dump core boolean: %s", optarg); + arg_dump_core = r; } - arg_dump_core = r; + break; + + case ARG_CRASH_CHVT: + r = parse_crash_chvt(optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse crash virtual terminal index: %s", optarg); break; case ARG_CRASH_SHELL: - r = optarg ? parse_boolean(optarg) : 1; - if (r < 0) { - log_error("Failed to parse crash shell boolean %s.", optarg); - return r; + if (!optarg) + arg_crash_shell = true; + else { + r = parse_boolean(optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg); + arg_crash_shell = r; + } + break; + + case ARG_CRASH_REBOOT: + if (!optarg) + arg_crash_reboot = true; + else { + r = parse_boolean(optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse crash shell boolean: %s", optarg); + arg_crash_reboot = r; } - arg_crash_shell = r; break; case ARG_CONFIRM_SPAWN: @@ -905,18 +979,16 @@ static int parse_argv(int argc, char *argv[]) { r = safe_atoi(optarg, &fd); if (r < 0 || fd < 0) { log_error("Failed to parse deserialize option %s.", optarg); - return r < 0 ? r : -EINVAL; + return -EINVAL; } - fd_cloexec(fd, true); + (void) fd_cloexec(fd, true); f = fdopen(fd, "r"); if (!f) return log_error_errno(errno, "Failed to open serialization fd: %m"); - if (arg_serialization) - fclose(arg_serialization); - + safe_fclose(arg_serialization); arg_serialization = f; break; @@ -926,10 +998,14 @@ static int parse_argv(int argc, char *argv[]) { arg_switched_root = true; break; + case ARG_MACHINE_ID: + r = set_machine_id(optarg); + if (r < 0) + return log_error_errno(r, "MachineID '%s' is not valid.", optarg); + break; + case 'h': arg_action = ACTION_HELP; - if (arg_no_pager < 0) - arg_no_pager = true; break; case 'D': @@ -976,14 +1052,16 @@ static int help(void) { " --unit=UNIT Set default unit\n" " --system Run a system instance, even if PID != 1\n" " --user Run a user instance\n" - " --dump-core[=0|1] Dump core on crash\n" - " --crash-shell[=0|1] Run shell on crash\n" - " --confirm-spawn[=0|1] Ask for confirmation when spawning processes\n" - " --show-status[=0|1] Show status updates on the console during bootup\n" + " --dump-core[=BOOL] Dump core on crash\n" + " --crash-vt=NR Change to specified VT on crash\n" + " --crash-reboot[=BOOL] Reboot on crash\n" + " --crash-shell[=BOOL] Run shell on crash\n" + " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n" + " --show-status[=BOOL] Show status updates on the console during bootup\n" " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n" " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n" - " --log-color[=0|1] Highlight important log messages\n" - " --log-location[=0|1] Include code location in log messages\n" + " --log-color[=BOOL] Highlight important log messages\n" + " --log-location[=BOOL] Include code location in log messages\n" " --default-standard-output= Set default standard output for services\n" " --default-standard-error= Set default standard error output for services\n", program_invocation_short_name); @@ -991,16 +1069,9 @@ static int help(void) { return 0; } -static int version(void) { - puts(PACKAGE_STRING); - puts(SYSTEMD_FEATURES); - - return 0; -} - static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching_root) { - FILE *f = NULL; - FDSet *fds = NULL; + _cleanup_fdset_free_ FDSet *fds = NULL; + _cleanup_fclose_ FILE *f = NULL; int r; assert(m); @@ -1008,57 +1079,39 @@ static int prepare_reexecute(Manager *m, FILE **_f, FDSet **_fds, bool switching assert(_fds); r = manager_open_serialization(m, &f); - if (r < 0) { - log_error_errno(r, "Failed to create serialization file: %m"); - goto fail; - } + if (r < 0) + return log_error_errno(r, "Failed to create serialization file: %m"); /* Make sure nothing is really destructed when we shut down */ - m->n_reloading ++; + m->n_reloading++; bus_manager_send_reloading(m, true); fds = fdset_new(); - if (!fds) { - r = -ENOMEM; - log_error_errno(r, "Failed to allocate fd set: %m"); - goto fail; - } + if (!fds) + return log_oom(); r = manager_serialize(m, f, fds, switching_root); - if (r < 0) { - log_error_errno(r, "Failed to serialize state: %m"); - goto fail; - } + if (r < 0) + return log_error_errno(r, "Failed to serialize state: %m"); - if (fseeko(f, 0, SEEK_SET) < 0) { - log_error_errno(errno, "Failed to rewind serialization fd: %m"); - goto fail; - } + if (fseeko(f, 0, SEEK_SET) == (off_t) -1) + return log_error_errno(errno, "Failed to rewind serialization fd: %m"); r = fd_cloexec(fileno(f), false); - if (r < 0) { - log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m"); - goto fail; - } + if (r < 0) + return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m"); r = fdset_cloexec(fds, false); - if (r < 0) { - log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m"); - goto fail; - } + if (r < 0) + return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m"); *_f = f; *_fds = fds; - return 0; - -fail: - fdset_free(fds); - - if (f) - fclose(f); + f = NULL; + fds = NULL; - return r; + return 0; } static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { @@ -1093,32 +1146,6 @@ static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { return 0; } -static void test_mtab(void) { - - static const char ok[] = - "/proc/self/mounts\0" - "/proc/mounts\0" - "../proc/self/mounts\0" - "../proc/mounts\0"; - - _cleanup_free_ char *p = NULL; - int r; - - /* Check that /etc/mtab is a symlink to the right place or - * non-existing. But certainly not a file, or a symlink to - * some weird place... */ - - r = readlink_malloc("/etc/mtab", &p); - if (r == -ENOENT) - return; - if (r >= 0 && nulstr_contains(ok, p)) - return; - - log_warning("/etc/mtab is not a symlink or not pointing to /proc/self/mounts. " - "This is not supported anymore. " - "Please make sure to replace this file by a symlink to avoid incorrect or misleading mount(8) output."); -} - static void test_usr(void) { /* Check that /usr is not a separate fs */ @@ -1142,15 +1169,19 @@ static int initialize_join_controllers(void) { return -ENOMEM; arg_join_controllers[0] = strv_new("cpu", "cpuacct", NULL); - arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL); - arg_join_controllers[2] = NULL; + if (!arg_join_controllers[0]) + goto oom; - if (!arg_join_controllers[0] || !arg_join_controllers[1]) { - free_join_controllers(); - return -ENOMEM; - } + arg_join_controllers[1] = strv_new("net_cls", "net_prio", NULL); + if (!arg_join_controllers[1]) + goto oom; + arg_join_controllers[2] = NULL; return 0; + +oom: + arg_join_controllers = strv_free_free(arg_join_controllers); + return -ENOMEM; } static int enforce_syscall_archs(Set *archs) { @@ -1160,6 +1191,9 @@ static int enforce_syscall_archs(Set *archs) { void *id; int r; + if (!is_seccomp_available()) + return 0; + seccomp = seccomp_init(SCMP_ACT_ALLOW); if (!seccomp) return log_oom(); @@ -1200,30 +1234,107 @@ static int status_welcome(void) { "PRETTY_NAME", &pretty_name, "ANSI_COLOR", &ansi_color, NULL); - if (r == -ENOENT) { + if (r == -ENOENT) r = parse_env_file("/usr/lib/os-release", NEWLINE, "PRETTY_NAME", &pretty_name, "ANSI_COLOR", &ansi_color, NULL); - } if (r < 0 && r != -ENOENT) log_warning_errno(r, "Failed to read os-release file: %m"); - return status_printf(NULL, false, false, - "\nWelcome to \x1B[%sm%s\x1B[0m!\n", - isempty(ansi_color) ? "1" : ansi_color, - isempty(pretty_name) ? "Linux" : pretty_name); + if (log_get_show_color()) + return status_printf(NULL, false, false, + "\nWelcome to \x1B[%sm%s\x1B[0m!\n", + isempty(ansi_color) ? "1" : ansi_color, + isempty(pretty_name) ? "Linux" : pretty_name); + else + return status_printf(NULL, false, false, + "\nWelcome to %s!\n", + isempty(pretty_name) ? "Linux" : pretty_name); } static int write_container_id(void) { const char *c; + int r; c = getenv("container"); if (isempty(c)) return 0; - return write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE); + RUN_WITH_UMASK(0022) + r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE); + if (r < 0) + return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m"); + + return 1; +} + +static int bump_unix_max_dgram_qlen(void) { + _cleanup_free_ char *qlen = NULL; + unsigned long v; + int r; + + /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel + * default of 16 is simply too low. We set the value really + * really early during boot, so that it is actually applied to + * all our sockets, including the $NOTIFY_SOCKET one. */ + + r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen); + if (r < 0) + return log_warning_errno(r, "Failed to read AF_UNIX datagram queue length, ignoring: %m"); + + r = safe_atolu(qlen, &v); + if (r < 0) + return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length, ignoring: %m"); + + if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN) + return 0; + + qlen = mfree(qlen); + if (asprintf(&qlen, "%lu\n", DEFAULT_UNIX_MAX_DGRAM_QLEN) < 0) + return log_oom(); + + r = write_string_file("/proc/sys/net/unix/max_dgram_qlen", qlen, 0); + if (r < 0) + return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, + "Failed to bump AF_UNIX datagram queue length, ignoring: %m"); + + return 1; +} + +static int fixup_environment(void) { + _cleanup_free_ char *term = NULL; + int r; + + /* We expect the environment to be set correctly + * if run inside a container. */ + if (detect_container() > 0) + return 0; + + /* When started as PID1, the kernel uses /dev/console + * for our stdios and uses TERM=linux whatever the + * backend device used by the console. We try to make + * a better guess here since some consoles might not + * have support for color mode for example. + * + * However if TERM was configured through the kernel + * command line then leave it alone. */ + + r = get_proc_cmdline_key("TERM=", &term); + if (r < 0) + return r; + + if (r == 0) { + term = strdup(default_term_for_tty("/dev/console")); + if (!term) + return -ENOMEM; + } + + if (setenv("TERM", term, 1) < 0) + return -errno; + + return 0; } int main(int argc, char *argv[]) { @@ -1255,7 +1366,6 @@ int main(int argc, char *argv[]) { /* This is compatibility support for SysV, where * calling init as a user is identical to telinit. */ - errno = -ENOENT; execv(SYSTEMCTL_BINARY_PATH, argv); log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m"); return 1; @@ -1286,18 +1396,16 @@ int main(int argc, char *argv[]) { saved_argv = argv; saved_argc = argc; - log_show_color(isatty(STDERR_FILENO) > 0); log_set_upgrade_syslog_to_journal(true); /* Disable the umask logic */ if (getpid() == 1) umask(0); - if (getpid() == 1 && detect_container(NULL) <= 0) { + if (getpid() == 1 && detect_container() <= 0) { /* Running outside of a container as PID 1 */ - arg_running_as = MANAGER_SYSTEM; - make_null_stdio(); + arg_system = true; log_set_target(LOG_TARGET_KMSG); log_open(); @@ -1305,28 +1413,32 @@ int main(int argc, char *argv[]) { initrd_timestamp = userspace_timestamp; if (!skip_setup) { - mount_setup_early(); + r = mount_setup_early(); + if (r < 0) { + error_message = "Failed to early mount API filesystems"; + goto finish; + } dual_timestamp_get(&security_start_timestamp); if (mac_selinux_setup(&loaded_policy) < 0) { error_message = "Failed to load SELinux policy"; goto finish; - } else if (ima_setup() < 0) { - error_message = "Failed to load IMA policy"; - goto finish; } else if (mac_smack_setup(&loaded_policy) < 0) { error_message = "Failed to load SMACK policy"; goto finish; + } else if (ima_setup() < 0) { + error_message = "Failed to load IMA policy"; + goto finish; } dual_timestamp_get(&security_finish_timestamp); } - if (mac_selinux_init(NULL) < 0) { + if (mac_selinux_init() < 0) { error_message = "Failed to initialize SELinux policy"; goto finish; } if (!skip_setup) { - if (clock_is_localtime() > 0) { + if (clock_is_localtime(NULL) > 0) { int min; /* @@ -1346,7 +1458,7 @@ int main(int argc, char *argv[]) { /* * Do a dummy very first call to seal the kernel's time warp magic. * - * Do not call this this from inside the initrd. The initrd might not + * Do not call this from inside the initrd. The initrd might not * carry /etc/adjtime with LOCAL, but the real system could be set up * that way. In such case, we need to delay the time-warp or the sealing * until we reach the real system. @@ -1356,8 +1468,14 @@ int main(int argc, char *argv[]) { * saving time change. All kernel local time concepts will be treated * as UTC that way. */ - clock_reset_timewarp(); + (void) clock_reset_timewarp(); } + + r = clock_apply_epoch(); + if (r < 0) + log_error_errno(r, "Current system time is before build time, but cannot correct: %m"); + else if (r > 0) + log_info("System time before build time, advancing clock."); } /* Set the default for later on, but don't actually @@ -1370,7 +1488,7 @@ int main(int argc, char *argv[]) { } else if (getpid() == 1) { /* Running inside a container, as PID 1 */ - arg_running_as = MANAGER_SYSTEM; + arg_system = true; log_set_target(LOG_TARGET_CONSOLE); log_close_console(); /* force reopen of /dev/console */ log_open(); @@ -1380,27 +1498,43 @@ int main(int argc, char *argv[]) { /* clear the kernel timestamp, * because we are in a container */ - kernel_timestamp.monotonic = 0ULL; - kernel_timestamp.realtime = 0ULL; - + kernel_timestamp = DUAL_TIMESTAMP_NULL; } else { /* Running as user instance */ - arg_running_as = MANAGER_USER; + arg_system = false; log_set_target(LOG_TARGET_AUTO); log_open(); /* clear the kernel timestamp, * because we are not PID 1 */ - kernel_timestamp.monotonic = 0ULL; - kernel_timestamp.realtime = 0ULL; + kernel_timestamp = DUAL_TIMESTAMP_NULL; } - /* Initialize default unit */ - r = set_default_unit(SPECIAL_DEFAULT_TARGET); - if (r < 0) { - log_emergency_errno(r, "Failed to set default unit %s: %m", SPECIAL_DEFAULT_TARGET); - error_message = "Failed to set default unit"; - goto finish; + if (getpid() == 1) { + /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit) + * will process core dumps for system services by default. */ + if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0) + log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m"); + + /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored + * until the systemd-coredump tool is enabled via sysctl. */ + if (!skip_setup) + (void) write_string_file("/proc/sys/kernel/core_pattern", "|/bin/false", 0); + } + + if (arg_system) { + if (fixup_environment() < 0) { + error_message = "Failed to fix up PID1 environment"; + goto finish; + } + + /* Try to figure out if we can use colors with the console. No + * need to do that for user instances since they never log + * into the console. */ + log_show_color(colors_enabled()); + r = make_null_stdio(); + if (r < 0) + log_warning_errno(r, "Failed to redirect standard streams to /dev/null: %m"); } r = initialize_join_controllers(); @@ -1428,12 +1562,14 @@ int main(int argc, char *argv[]) { (void) reset_all_signal_handlers(); (void) ignore_signals(SIGNALS_IGNORE, -1); + arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U); + if (parse_config_file() < 0) { error_message = "Failed to parse config file"; goto finish; } - if (arg_running_as == MANAGER_SYSTEM) { + if (arg_system) { r = parse_proc_cmdline(parse_proc_cmdline_item); if (r < 0) log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); @@ -1448,30 +1584,40 @@ int main(int argc, char *argv[]) { goto finish; } + /* Initialize default unit */ + if (!arg_default_unit) { + arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET); + if (!arg_default_unit) { + r = log_oom(); + error_message = "Failed to set default unit"; + goto finish; + } + } + if (arg_action == ACTION_TEST && geteuid() == 0) { log_error("Don't run test mode as root."); goto finish; } - if (arg_running_as == MANAGER_USER && + if (!arg_system && arg_action == ACTION_RUN && sd_booted() <= 0) { log_error("Trying to run as user instance, but the system has not been booted with systemd."); goto finish; } - if (arg_running_as == MANAGER_SYSTEM && + if (arg_system && arg_action == ACTION_RUN && running_in_chroot() > 0) { log_error("Cannot be run in a chroot() environment."); goto finish; } - if (arg_action == ACTION_TEST) + if (arg_action == ACTION_TEST || arg_action == ACTION_HELP) { + pager_open(arg_no_pager, false); skip_setup = true; - - pager_open_if_enabled(); + } if (arg_action == ACTION_HELP) { retval = help(); @@ -1480,15 +1626,13 @@ int main(int argc, char *argv[]) { retval = version(); goto finish; } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { + pager_open(arg_no_pager, false); unit_dump_config_items(stdout); retval = EXIT_SUCCESS; goto finish; - } else if (arg_action == ACTION_DONE) { - retval = EXIT_SUCCESS; - goto finish; } - if (arg_running_as == MANAGER_USER && + if (!arg_system && !getenv("XDG_RUNTIME_DIR")) { log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set."); goto finish; @@ -1511,7 +1655,7 @@ int main(int argc, char *argv[]) { if (arg_serialization) assert_se(fdset_remove(fds, fileno(arg_serialization)) >= 0); - if (arg_running_as == MANAGER_SYSTEM) + if (arg_system) /* Become a session leader if we aren't one yet. */ setsid(); @@ -1520,7 +1664,7 @@ int main(int argc, char *argv[]) { /* Reset the console, but only if this is really init and we * are freshly booted */ - if (arg_running_as == MANAGER_SYSTEM && arg_action == ACTION_RUN) { + if (arg_system && arg_action == ACTION_RUN) { /* If we are init, we connect stdin/stdout/stderr to * /dev/null and make sure we don't have a controlling @@ -1547,15 +1691,15 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_running_as == MANAGER_SYSTEM) { - const char *virtualization = NULL; + if (arg_system) { + int v; log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")", arg_action == ACTION_TEST ? "test " : "" ); - detect_virtualization(&virtualization); - if (virtualization) - log_info("Detected virtualization %s.", virtualization); + v = detect_virtualization(); + if (v > 0) + log_info("Detected virtualization %s.", virtualization_to_string(v)); write_container_id(); @@ -1583,33 +1727,33 @@ int main(int argc, char *argv[]) { arg_action == ACTION_TEST ? " test" : "", getuid(), t); } - if (arg_running_as == MANAGER_SYSTEM && !skip_setup) { + if (arg_system && !skip_setup) { if (arg_show_status > 0) status_welcome(); hostname_setup(); - machine_id_setup(NULL); + machine_id_setup(NULL, arg_machine_id, NULL); loopback_setup(); + bump_unix_max_dgram_qlen(); - test_mtab(); test_usr(); } - if (arg_running_as == MANAGER_SYSTEM && arg_runtime_watchdog > 0) + if (arg_system && arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY) watchdog_set_timeout(&arg_runtime_watchdog); if (arg_timer_slack_nsec != NSEC_INFINITY) if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0) log_error_errno(errno, "Failed to adjust timer slack: %m"); - if (arg_capability_bounding_set_drop) { - r = capability_bounding_set_drop_usermode(arg_capability_bounding_set_drop); + if (!cap_test_all(arg_capability_bounding_set)) { + r = capability_bounding_set_drop_usermode(arg_capability_bounding_set); if (r < 0) { log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m"); error_message = "Failed to drop capability bounding set of usermode helpers"; goto finish; } - r = capability_bounding_set_drop(arg_capability_bounding_set_drop, true); + r = capability_bounding_set_drop(arg_capability_bounding_set, true); if (r < 0) { log_emergency_errno(r, "Failed to drop capability bounding set: %m"); error_message = "Failed to drop capability bounding set"; @@ -1625,24 +1769,24 @@ int main(int argc, char *argv[]) { } } - if (arg_running_as == MANAGER_USER) + if (!arg_system) /* Become reaper of our children */ if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) log_warning_errno(errno, "Failed to make us a subreaper: %m"); - if (arg_running_as == MANAGER_SYSTEM) { + if (arg_system) { bump_rlimit_nofile(&saved_rlimit_nofile); if (empty_etc) { r = unit_file_preset_all(UNIT_FILE_SYSTEM, false, NULL, UNIT_FILE_PRESET_ENABLE_ONLY, false, NULL, 0); if (r < 0) - log_warning_errno(r, "Failed to populate /etc with preset unit settings, ignoring: %m"); + log_full_errno(r == -EEXIST ? LOG_NOTICE : LOG_WARNING, r, "Failed to populate /etc with preset unit settings, ignoring: %m"); else log_info("Populated /etc with preset unit settings."); } } - r = manager_new(arg_running_as, arg_action == ACTION_TEST, &m); + r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, arg_action == ACTION_TEST, &m); if (r < 0) { log_emergency_errno(r, "Failed to allocate manager object: %m"); error_message = "Failed to allocate manager object"; @@ -1657,6 +1801,7 @@ int main(int argc, char *argv[]) { m->initrd_timestamp = initrd_timestamp; m->security_start_timestamp = security_start_timestamp; m->security_finish_timestamp = security_finish_timestamp; + m->cad_burst_action = arg_cad_burst_action; manager_set_defaults(m); manager_set_show_status(m, arg_show_status); @@ -1673,16 +1818,12 @@ int main(int argc, char *argv[]) { /* This will close all file descriptors that were opened, but * not claimed by any unit. */ - fdset_free(fds); - fds = NULL; + fds = fdset_free(fds); - if (arg_serialization) { - fclose(arg_serialization); - arg_serialization = NULL; - } + arg_serialization = safe_fclose(arg_serialization); if (queue_default_job) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; Unit *target = NULL; Job *default_unit_job; @@ -1722,11 +1863,13 @@ int main(int argc, char *argv[]) { manager_dump_units(m, stdout, "\t"); } - r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, false, &error, &default_unit_job); + r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job); if (r == -EPERM) { log_debug("Default target could not be isolated, starting instead: %s", bus_error_message(&error, r)); - r = manager_add_job(m, JOB_START, target, JOB_REPLACE, false, &error, &default_unit_job); + sd_bus_error_free(&error); + + r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job); if (r < 0) { log_emergency("Failed to start default target: %s", bus_error_message(&error, r)); error_message = "Failed to start default target"; @@ -1763,11 +1906,6 @@ int main(int argc, char *argv[]) { switch (m->exit_code) { - case MANAGER_EXIT: - retval = EXIT_SUCCESS; - log_debug("Exit."); - goto finish; - case MANAGER_RELOAD: log_info("Reloading."); @@ -1785,7 +1923,7 @@ int main(int argc, char *argv[]) { case MANAGER_REEXECUTE: if (prepare_reexecute(m, &arg_serialization, &fds, false) < 0) { - error_message = "Failed to prepare for reexection"; + error_message = "Failed to prepare for reexecution"; goto finish; } @@ -1801,7 +1939,7 @@ int main(int argc, char *argv[]) { if (!switch_root_init) if (prepare_reexecute(m, &arg_serialization, &fds, true) < 0) { - error_message = "Failed to prepare for reexection"; + error_message = "Failed to prepare for reexecution"; goto finish; } @@ -1809,11 +1947,21 @@ int main(int argc, char *argv[]) { log_notice("Switching root."); goto finish; + case MANAGER_EXIT: + retval = m->return_value; + + if (MANAGER_IS_USER(m)) { + log_debug("Exit."); + goto finish; + } + + /* fallthrough */ case MANAGER_REBOOT: case MANAGER_POWEROFF: case MANAGER_HALT: case MANAGER_KEXEC: { static const char * const table[_MANAGER_EXIT_CODE_MAX] = { + [MANAGER_EXIT] = "exit", [MANAGER_REBOOT] = "reboot", [MANAGER_POWEROFF] = "poweroff", [MANAGER_HALT] = "halt", @@ -1837,23 +1985,16 @@ finish: if (m) arg_shutdown_watchdog = m->shutdown_watchdog; - m = manager_free(m); - - for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) { - free(arg_default_rlimit[j]); - arg_default_rlimit[j] = NULL; - } - free(arg_default_unit); - arg_default_unit = NULL; - - free_join_controllers(); + m = manager_free(m); - strv_free(arg_default_environment); - arg_default_environment = NULL; + for (j = 0; j < ELEMENTSOF(arg_default_rlimit); j++) + arg_default_rlimit[j] = mfree(arg_default_rlimit[j]); - set_free(arg_syscall_archs); - arg_syscall_archs = NULL; + arg_default_unit = mfree(arg_default_unit); + arg_join_controllers = strv_free_free(arg_join_controllers); + arg_default_environment = strv_free(arg_default_environment); + arg_syscall_archs = set_free(arg_syscall_archs); mac_selinux_finish(); @@ -1870,7 +2011,7 @@ finish: * that the new systemd can pass the kernel default to * its child processes */ if (saved_rlimit_nofile.rlim_cur > 0) - setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile); + (void) setrlimit(RLIMIT_NOFILE, &saved_rlimit_nofile); if (switch_root_dir) { /* Kill all remaining processes from the @@ -1905,17 +2046,22 @@ finish: args[i++] = SYSTEMD_BINARY_PATH; if (switch_root_dir) args[i++] = "--switched-root"; - args[i++] = arg_running_as == MANAGER_SYSTEM ? "--system" : "--user"; + args[i++] = arg_system ? "--system" : "--user"; args[i++] = "--deserialize"; args[i++] = sfd; args[i++] = NULL; - /* do not pass along the environment we inherit from the kernel or initrd */ - if (switch_root_dir) - clearenv(); - assert(i <= args_size); - execv(args[0], (char* const*) args); + + /* + * We want valgrind to print its memory usage summary before reexecution. + * Valgrind won't do this is on its own on exec(), but it will do it on exit(). + * Hence, to ensure we get a summary here, fork() off a child, let it exit() cleanly, + * so that it prints the summary, and wait() for it in the parent, before proceeding into the exec(). + */ + valgrind_summary_hack(); + + (void) execv(args[0], (char* const*) args); } /* Try the fallback, if there is any, without any @@ -1924,18 +2070,11 @@ finish: * getopt() in argv[], and some cleanups in envp[], * but let's hope that doesn't matter.) */ - if (arg_serialization) { - fclose(arg_serialization); - arg_serialization = NULL; - } - - if (fds) { - fdset_free(fds); - fds = NULL; - } + arg_serialization = safe_fclose(arg_serialization); + fds = fdset_free(fds); /* Reopen the console */ - make_console_stdio(); + (void) make_console_stdio(); for (j = 1, i = 1; j < (unsigned) argc; j++) args[i++] = argv[j]; @@ -1949,33 +2088,26 @@ finish: if (switch_root_init) { args[0] = switch_root_init; - execv(args[0], (char* const*) args); + (void) execv(args[0], (char* const*) args); log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m"); } args[0] = "/sbin/init"; - execv(args[0], (char* const*) args); + (void) execv(args[0], (char* const*) args); if (errno == ENOENT) { log_warning("No /sbin/init, trying fallback"); args[0] = "/bin/sh"; args[1] = NULL; - execv(args[0], (char* const*) args); + (void) execv(args[0], (char* const*) args); log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m"); } else log_warning_errno(errno, "Failed to execute /sbin/init, giving up: %m"); } - if (arg_serialization) { - fclose(arg_serialization); - arg_serialization = NULL; - } - - if (fds) { - fdset_free(fds); - fds = NULL; - } + arg_serialization = safe_fclose(arg_serialization); + fds = fdset_free(fds); #ifdef HAVE_VALGRIND_VALGRIND_H /* If we are PID 1 and running under valgrind, then let's exit @@ -1988,7 +2120,8 @@ finish: if (shutdown_verb) { char log_level[DECIMAL_STR_MAX(int) + 1]; - const char* command_line[9] = { + char exit_code[DECIMAL_STR_MAX(uint8_t) + 1]; + const char* command_line[11] = { SYSTEMD_SHUTDOWN_BINARY_PATH, shutdown_verb, "--log-level", log_level, @@ -2003,6 +2136,7 @@ finish: xsprintf(log_level, "%d", log_get_max_level()); switch (log_get_target()) { + case LOG_TARGET_KMSG: case LOG_TARGET_JOURNAL_OR_KMSG: case LOG_TARGET_SYSLOG_OR_KMSG: @@ -2025,9 +2159,15 @@ finish: if (log_get_show_location()) command_line[pos++] = "--log-location"; + if (streq(shutdown_verb, "exit")) { + command_line[pos++] = "--exit-code"; + command_line[pos++] = exit_code; + xsprintf(exit_code, "%d", retval); + } + assert(pos < ELEMENTSOF(command_line)); - if (arm_reboot_watchdog && arg_shutdown_watchdog > 0) { + if (arm_reboot_watchdog && arg_shutdown_watchdog > 0 && arg_shutdown_watchdog != USEC_INFINITY) { char *e; /* If we reboot let's set the shutdown @@ -2038,15 +2178,15 @@ finish: /* Tell the binary how often to ping, ignore failure */ if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0) - strv_push(&env_block, e); + (void) strv_push(&env_block, e); } else watchdog_close(true); /* Avoid the creation of new processes forked by the * kernel; at this point, we will not listen to the * signals anyway */ - if (detect_container(NULL) <= 0) - cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER); + if (detect_container() <= 0) + (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER); execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block); log_error_errno(errno, "Failed to execute shutdown binary, %s: %m", @@ -2056,9 +2196,9 @@ finish: if (getpid() == 1) { if (error_message) manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, - ANSI_HIGHLIGHT_RED_ON "!!!!!!" ANSI_HIGHLIGHT_OFF, + ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL, "%s, freezing.", error_message); - freeze(); + freeze_or_reboot(); } return retval; diff --git a/src/core/manager.c b/src/core/manager.c index a1f37bbbb3..3569249788 100644 --- a/src/core/manager.c +++ b/src/core/manager.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,19 +17,19 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include <dirent.h> #include <errno.h> -#include <string.h> +#include <fcntl.h> +#include <linux/kd.h> #include <signal.h> -#include <sys/wait.h> -#include <unistd.h> -#include <sys/inotify.h> +#include <string.h> #include <sys/epoll.h> -#include <sys/reboot.h> +#include <sys/inotify.h> #include <sys/ioctl.h> -#include <linux/kd.h> -#include <fcntl.h> -#include <dirent.h> +#include <sys/reboot.h> #include <sys/timerfd.h> +#include <sys/wait.h> +#include <unistd.h> #ifdef HAVE_AUDIT #include <libaudit.h> @@ -40,40 +38,57 @@ #include "sd-daemon.h" #include "sd-messages.h" -#include "hashmap.h" -#include "macro.h" -#include "strv.h" -#include "log.h" -#include "util.h" -#include "mkdir.h" -#include "ratelimit.h" -#include "locale-setup.h" -#include "unit-name.h" -#include "missing.h" -#include "rm-rf.h" -#include "path-lookup.h" -#include "special.h" -#include "exit-status.h" -#include "virt.h" -#include "watchdog.h" -#include "path-util.h" +#include "alloc-util.h" #include "audit-fd.h" #include "boot-timestamps.h" -#include "env-util.h" #include "bus-common-errors.h" #include "bus-error.h" -#include "bus-util.h" #include "bus-kernel.h" -#include "time-util.h" -#include "process-util.h" -#include "terminal-util.h" -#include "signal-util.h" -#include "dbus.h" -#include "dbus-unit.h" +#include "bus-util.h" +#include "clean-ipc.h" #include "dbus-job.h" #include "dbus-manager.h" +#include "dbus-unit.h" +#include "dbus.h" +#include "dirent-util.h" +#include "env-util.h" +#include "escape.h" +#include "exit-status.h" +#include "fd-util.h" +#include "fileio.h" +#include "fs-util.h" +#include "hashmap.h" +#include "io-util.h" +#include "locale-setup.h" +#include "log.h" +#include "macro.h" #include "manager.h" +#include "missing.h" +#include "mkdir.h" +#include "parse-util.h" +#include "path-lookup.h" +#include "path-util.h" +#include "process-util.h" +#include "ratelimit.h" +#include "rm-rf.h" +#include "signal-util.h" +#include "special.h" +#include "stat-util.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "terminal-util.h" +#include "time-util.h" #include "transaction.h" +#include "umask-util.h" +#include "unit-name.h" +#include "user-util.h" +#include "util.h" +#include "virt.h" +#include "watchdog.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) @@ -81,13 +96,14 @@ #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_undo_generators(Manager *m); static void manager_watch_jobs_in_progress(Manager *m) { usec_t next; @@ -111,7 +127,7 @@ static void manager_watch_jobs_in_progress(Manager *m) { (void) sd_event_source_set_description(m->jobs_in_progress_event_source, "manager-jobs-in-progress"); } -#define CYLON_BUFFER_EXTRA (2*(sizeof(ANSI_RED_ON)-1) + sizeof(ANSI_HIGHLIGHT_RED_ON)-1 + 2*(sizeof(ANSI_HIGHLIGHT_OFF)-1)) +#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; @@ -122,23 +138,28 @@ static void draw_cylon(char buffer[], size_t buflen, unsigned width, unsigned po if (pos > 1) { if (pos > 2) p = mempset(p, ' ', pos-2); - p = stpcpy(p, ANSI_RED_ON); + if (log_get_show_color()) + p = stpcpy(p, ANSI_RED); *p++ = '*'; } if (pos > 0 && pos <= width) { - p = stpcpy(p, ANSI_HIGHLIGHT_RED_ON); + if (log_get_show_color()) + p = stpcpy(p, ANSI_HIGHLIGHT_RED); *p++ = '*'; } - p = stpcpy(p, ANSI_HIGHLIGHT_OFF); + if (log_get_show_color()) + p = stpcpy(p, ANSI_NORMAL); if (pos < width) { - p = stpcpy(p, ANSI_RED_ON); + if (log_get_show_color()) + p = stpcpy(p, ANSI_RED); *p++ = '*'; if (pos < width-1) p = mempset(p, ' ', width-1-pos); - strcpy(p, ANSI_HIGHLIGHT_OFF); + if (log_get_show_color()) + strcpy(p, ANSI_NORMAL); } } @@ -220,7 +241,7 @@ static int have_ask_password(void) { errno = 0; de = readdir(dir); - if (!de && errno != 0) + if (!de && errno > 0) return -errno; if (!de) return false; @@ -250,8 +271,8 @@ static int manager_dispatch_ask_password_fd(sd_event_source *source, static void manager_close_ask_password(Manager *m) { assert(m); - m->ask_password_inotify_fd = safe_close(m->ask_password_inotify_fd); 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; } @@ -317,6 +338,8 @@ static int manager_watch_idle_pipe(Manager *m) { 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); } @@ -365,6 +388,9 @@ static int enable_special_signals(Manager *m) { 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. */ @@ -467,15 +493,15 @@ static int manager_setup_signals(Manager *m) { (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. */ - r = sd_event_source_set_priority(m->signal_event_source, -5); + /* 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 (m->running_as == MANAGER_SYSTEM) + if (MANAGER_IS_SYSTEM(m)) return enable_special_signals(m); return 0; @@ -493,15 +519,17 @@ static void manager_clean_environment(Manager *m) { "MANAGERPID", "LISTEN_PID", "LISTEN_FDS", + "LISTEN_FDNAMES", "WATCHDOG_PID", "WATCHDOG_USEC", + "INVOCATION_ID", NULL); } static int manager_default_environment(Manager *m) { assert(m); - if (m->running_as == MANAGER_SYSTEM) { + 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. @@ -529,50 +557,55 @@ static int manager_default_environment(Manager *m) { return 0; } - -int manager_new(ManagerRunningAs running_as, bool test_run, Manager **_m) { - - static const char * const unit_log_fields[_MANAGER_RUNNING_AS_MAX] = { - [MANAGER_SYSTEM] = "UNIT=", - [MANAGER_USER] = "USER_UNIT=", - }; - - static const char * const unit_log_format_strings[_MANAGER_RUNNING_AS_MAX] = { - [MANAGER_SYSTEM] = "UNIT=%s", - [MANAGER_USER] = "USER_UNIT=%s", - }; - +int manager_new(UnitFileScope scope, bool test_run, Manager **_m) { Manager *m; int r; assert(_m); - assert(running_as >= 0); - assert(running_as < _MANAGER_RUNNING_AS_MAX); + 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 (running_as == MANAGER_SYSTEM && detect_container(NULL) <= 0) + if (MANAGER_IS_SYSTEM(m) && detect_container() <= 0) boot_timestamps(&m->userspace_timestamp, &m->firmware_timestamp, &m->loader_timestamp); #endif - m->running_as = running_as; - m->exit_code = _MANAGER_EXIT_CODE_INVALID; - m->default_timer_accuracy_usec = USEC_PER_MINUTE; - /* Prepare log fields we can use for structured logging */ - m->unit_log_field = unit_log_fields[running_as]; - m->unit_log_format_string = unit_log_format_strings[running_as]; + 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->signal_fd = m->time_change_fd = m->dev_autofs_fd = m->private_listen_fd = m->kdbus_fd = m->utab_inotify_fd = -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->ask_password_inotify_fd = -1; m->have_ask_password = -EINVAL; /* we don't know */ + m->first_boot = -1; m->test_run = test_run; @@ -599,14 +632,6 @@ int manager_new(ManagerRunningAs running_as, bool test_run, Manager **_m) { if (r < 0) goto fail; - r = set_ensure_allocated(&m->startup_units, NULL); - if (r < 0) - goto fail; - - r = set_ensure_allocated(&m->failed_units, NULL); - if (r < 0) - goto fail; - r = sd_event_default(&m->event); if (r < 0) goto fail; @@ -643,9 +668,8 @@ int manager_new(ManagerRunningAs running_as, bool test_run, Manager **_m) { goto fail; } - /* Note that we set up neither kdbus, nor the notify fd - * here. We do that after deserialization, since they might - * have gotten serialized across the reexec. */ + /* 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; @@ -669,29 +693,25 @@ static int manager_setup_notify(Manager *m) { .sa.sa_family = AF_UNIX, }; static const int one = 1; + const char *e; /* First free all secondary fields */ - free(m->notify_socket); - m->notify_socket = NULL; + 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"); - if (m->running_as == MANAGER_SYSTEM) - m->notify_socket = strdup("/run/systemd/notify"); - else { - const char *e; + fd_inc_rcvbuf(fd, NOTIFY_RCVBUF_SIZE); - e = getenv("XDG_RUNTIME_DIR"); - if (!e) { - log_error_errno(errno, "XDG_RUNTIME_DIR is not set: %m"); - return -EINVAL; - } - - m->notify_socket = strappend(e, "/systemd/notify"); + 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(); @@ -699,7 +719,7 @@ static int manager_setup_notify(Manager *m) { (void) unlink(m->notify_socket); strncpy(sa.un.sun_path, m->notify_socket, sizeof(sa.un.sun_path)-1); - r = bind(fd, &sa.sa, offsetof(struct sockaddr_un, sun_path) + strlen(sa.un.sun_path)); + 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); @@ -718,9 +738,9 @@ static int manager_setup_notify(Manager *m) { if (r < 0) return log_error_errno(r, "Failed to allocate notify event source: %m"); - /* Process signals 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, -7); + /* 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"); @@ -730,24 +750,128 @@ static int manager_setup_notify(Manager *m) { return 0; } -static int manager_setup_kdbus(Manager *m) { - _cleanup_free_ char *p = NULL; +static int manager_setup_cgroups_agent(Manager *m) { - assert(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 (m->test_run || m->kdbus_fd >= 0) + if (!MANAGER_IS_SYSTEM(m)) return 0; - if (!is_kdbus_available()) - return -ESOCKTNOSUPPORT; - m->kdbus_fd = bus_kernel_create_bus( - m->running_as == MANAGER_SYSTEM ? "system" : "user", - m->running_as == MANAGER_SYSTEM, &p); + 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); - if (m->kdbus_fd < 0) - return log_debug_errno(m->kdbus_fd, "Failed to set up kdbus: %m"); + 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); - log_debug("Successfully set up kdbus on %s", p); + (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; } @@ -761,9 +885,8 @@ static int manager_connect_bus(Manager *m, bool reexecuting) { return 0; try_bus_connect = - m->kdbus_fd >= 0 || reexecuting || - (m->running_as == MANAGER_USER && getenv("DBUS_SESSION_BUS_ADDRESS")); + (MANAGER_IS_USER(m) && getenv("DBUS_SESSION_BUS_ADDRESS")); /* Try to connect to the buses, if possible. */ return bus_init(m, try_bus_connect); @@ -793,6 +916,18 @@ enum { _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; @@ -802,6 +937,7 @@ static void unit_gc_sweep(Unit *u, unsigned gc_marker) { 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; @@ -842,7 +978,7 @@ bad: return; good: - u->gc_marker = gc_marker + GC_OFFSET_GOOD; + unit_gc_mark_good(u, gc_marker); } static unsigned manager_dispatch_gc_queue(Manager *m) { @@ -925,11 +1061,15 @@ Manager* manager_free(Manager *m) { * around */ manager_shutdown_cgroup(m, m->exit_code != MANAGER_REEXECUTE); - manager_undo_generators(m); + 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); @@ -940,15 +1080,17 @@ Manager* manager_free(Manager *m) { 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->idle_pipe_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(m->kdbus_fd); + safe_close_pair(m->user_lookup_fds); manager_close_ask_password(m); @@ -969,17 +1111,19 @@ Manager* manager_free(Manager *m) { free(m->switch_root_init); for (i = 0; i < _RLIMIT_MAX; i++) - free(m->rlimit[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); + free(m); return NULL; } -int manager_enumerate(Manager *m) { - int r = 0; +void manager_enumerate(Manager *m) { UnitType c; assert(m); @@ -987,8 +1131,6 @@ int manager_enumerate(Manager *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++) { - int q; - if (!unit_type_supported(c)) { log_debug("Unit type .%s is not supported on this system.", unit_type_to_string(c)); continue; @@ -997,13 +1139,10 @@ int manager_enumerate(Manager *m) { if (!unit_vtable[c]->enumerate) continue; - q = unit_vtable[c]->enumerate(m); - if (q < 0) - r = q; + unit_vtable[c]->enumerate(m); } manager_dispatch_load_queue(m); - return r; } static void manager_coldplug(Manager *m) { @@ -1029,7 +1168,6 @@ static void manager_coldplug(Manager *m) { static void manager_build_unit_path_cache(Manager *m) { char **i; - _cleanup_closedir_ DIR *d = NULL; int r; assert(m); @@ -1038,29 +1176,27 @@ static void manager_build_unit_path_cache(Manager *m) { m->unit_path_cache = set_new(&string_hash_ops); if (!m->unit_path_cache) { - log_error("Failed to allocate unit path cache."); - return; + 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.unit_path) { + 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_error_errno(errno, "Failed to open directory %s: %m", *i); + log_warning_errno(errno, "Failed to open directory %s, ignoring: %m", *i); continue; } - while ((de = readdir(d))) { + FOREACH_DIRENT(de, d, r = -errno; goto fail) { char *p; - if (hidden_file(de->d_name)) - continue; - p = strjoin(streq(*i, "/") ? "" : *i, "/", de->d_name, NULL); if (!p) { r = -ENOMEM; @@ -1071,25 +1207,18 @@ static void manager_build_unit_path_cache(Manager *m) { if (r < 0) goto fail; } - - closedir(d); - d = NULL; } return; fail: - log_error_errno(r, "Failed to build unit path cache: %m"); - - set_free_free(m->unit_path_cache); - m->unit_path_cache = NULL; + 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 int manager_distribute_fds(Manager *m, FDSet *fds) { - Unit *u; +static void manager_distribute_fds(Manager *m, FDSet *fds) { Iterator i; - int r; + Unit *u; assert(m); @@ -1098,14 +1227,11 @@ static int manager_distribute_fds(Manager *m, FDSet *fds) { if (fdset_size(fds) <= 0) break; - if (UNIT_VTABLE(u)->distribute_fds) { - r = UNIT_VTABLE(u)->distribute_fds(u, fds); - if (r < 0) - return r; - } - } + if (!UNIT_VTABLE(u)->distribute_fds) + continue; - return 0; + UNIT_VTABLE(u)->distribute_fds(u, fds); + } } int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { @@ -1113,32 +1239,35 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { assert(m); - dual_timestamp_get(&m->generators_start_timestamp); - r = manager_run_generators(m); - dual_timestamp_get(&m->generators_finish_timestamp); + r = lookup_paths_init(&m->lookup_paths, m->unit_file_scope, 0, NULL); if (r < 0) return r; - r = lookup_paths_init( - &m->lookup_paths, m->running_as, true, - NULL, - m->generator_unit_path, - m->generator_unit_path_early, - m->generator_unit_path_late); + /* 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 ++; + m->n_reloading++; /* First, enumerate what we can from all config files */ dual_timestamp_get(&m->units_load_start_timestamp); - r = manager_enumerate(m); + manager_enumerate(m); dual_timestamp_get(&m->units_load_finish_timestamp); /* Second, deserialize if there is something to deserialize */ @@ -1149,11 +1278,7 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { * useful to allow container managers to pass some file * descriptors to us pre-initialized. This enables * socket-based activation of entire containers. */ - if (fdset_size(fds) > 0) { - q = manager_distribute_fds(m, fds); - if (q < 0 && r == 0) - r = q; - } + manager_distribute_fds(m, fds); /* We might have deserialized the notify fd, but if we didn't * then let's create the bus now */ @@ -1161,18 +1286,33 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { if (q < 0 && r == 0) r = q; - /* We might have deserialized the kdbus control fd, but if we - * didn't, then let's create the bus now. */ - manager_setup_kdbus(m); - manager_connect_bus(m, !!serialization); - bus_track_coldplug(m, &m->subscribed, &m->deserialized_subscribed); + 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 --; + m->n_reloading--; /* Let's wait for the UnitNew/JobNew messages being * sent, before we notify that the reload is @@ -1183,7 +1323,7 @@ int manager_startup(Manager *m, FILE *serialization, FDSet *fds) { return r; } -int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, bool override, sd_bus_error *e, Job **_ret) { +int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, sd_bus_error *e, Job **_ret) { int r; Transaction *tr; @@ -1206,7 +1346,7 @@ int manager_add_job(Manager *m, JobType type, Unit *unit, JobMode mode, bool ove if (!tr) return -ENOMEM; - r = transaction_add_job_and_dependencies(tr, type, unit, NULL, true, override, false, + 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) @@ -1238,7 +1378,7 @@ tr_abort: return r; } -int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, bool override, 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) { Unit *unit; int r; @@ -1251,7 +1391,23 @@ int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode if (r < 0) return r; - return manager_add_job(m, type, unit, mode, override, e, _ret); + 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) { @@ -1318,8 +1474,12 @@ int manager_load_unit_prepare( t = unit_name_to_type(name); - if (t == _UNIT_TYPE_INVALID || !unit_name_is_valid(name, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) + 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) { @@ -1412,7 +1572,7 @@ void manager_clear_jobs(Manager *m) { while ((j = hashmap_first(m->jobs))) /* No need to recurse. We're cancelling all jobs. */ - job_finish_and_invalidate(j, JOB_CANCELED, false); + job_finish_and_invalidate(j, JOB_CANCELED, false, false); } static int manager_dispatch_run_queue(sd_event_source *source, void *userdata) { @@ -1478,13 +1638,41 @@ static unsigned manager_dispatch_dbus_queue(Manager *m) { return n; } -static void manager_invoke_notify_message(Manager *m, Unit *u, pid_t pid, char *buf, size_t n, FDSet *fds) { +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); - assert(n > 0); tags = strv_split(buf, "\n\r"); if (!tags) { @@ -1494,14 +1682,43 @@ static void manager_invoke_notify_message(Manager *m, Unit *u, pid_t pid, char * if (UNIT_VTABLE(u)->notify_message) UNIT_VTABLE(u)->notify_message(u, pid, tags, fds); - else - log_unit_debug(u, "Got notification message for unit. Ignoring."); + 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; - int r; assert(m); assert(m->notify_fd == fd); @@ -1511,114 +1728,108 @@ static int manager_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t return 0; } - for (;;) { - _cleanup_fdset_free_ FDSet *fds = NULL; - 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; - bool found = false; - Unit *u1, *u2, *u3; - int *fd_array = NULL; - unsigned n_fds = 0; - - n = recvmsg(m->notify_fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); - if (n < 0) { - if (errno == EAGAIN || errno == EINTR) - break; + 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 */ - return -errno; - } + /* 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) { + 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); + 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))) { + } 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); - } + ucred = (struct ucred*) CMSG_DATA(cmsg); } + } - if (n_fds > 0) { - assert(fd_array); + 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); - return log_oom(); - } + 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."); - continue; - } + if (!ucred || ucred->pid <= 0) { + log_warning("Received notify message without valid credentials. Ignoring."); + return 0; + } - if ((size_t) n >= sizeof(buf)) { - log_warning("Received notify message exceeded maximum size. Ignoring."); - continue; - } + 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; + } - buf[n] = 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(m, ucred->pid); - if (u1) { - manager_invoke_notify_message(m, u1, ucred->pid, buf, n, fds); - found = true; - } + /* 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, LONG_TO_PTR(ucred->pid)); - if (u2 && u2 != u1) { - manager_invoke_notify_message(m, u2, ucred->pid, buf, n, fds); - found = true; - } + 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, LONG_TO_PTR(ucred->pid)); - if (u3 && u3 != u2 && u3 != u1) { - manager_invoke_notify_message(m, u3, ucred->pid, buf, n, fds); - found = true; - } + 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 (!found) - log_warning("Cannot find unit for notify message of PID "PID_FMT".", ucred->pid); + 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 auxiliary fds with notification message, closing all."); - } + 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, siginfo_t *si) { +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); - UNIT_VTABLE(u)->sigchld_event(u, si->si_pid, si->si_code, si->si_status); + + 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) { @@ -1660,13 +1871,13 @@ static int manager_dispatch_sigchld(Manager *m) { /* And now figure out the unit this belongs * to, it might be multiple... */ - u1 = manager_get_unit_by_pid(m, si.si_pid); + 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, LONG_TO_PTR(si.si_pid)); + 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, LONG_TO_PTR(si.si_pid)); + u3 = hashmap_get(m->watch_pids2, PID_TO_PTR(si.si_pid)); if (u3 && u3 != u2 && u3 != u1) invoke_sigchld_event(m, u3, &si); } @@ -1684,23 +1895,53 @@ static int manager_dispatch_sigchld(Manager *m) { } static int manager_start_target(Manager *m, const char *name, JobMode mode) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _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, true, &error, NULL); + 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 == CAD_BURST_ACTION_IGNORE) + manager_start_target(m, SPECIAL_CTRL_ALT_DEL_TARGET, JOB_REPLACE_IRREVERSIBLY); + else { + switch (m->cad_burst_action) { + + case CAD_BURST_ACTION_REBOOT: + m->exit_code = MANAGER_REBOOT; + break; + + case CAD_BURST_ACTION_POWEROFF: + m->exit_code = MANAGER_POWEROFF; + break; + + default: + assert_not_reached("Unknown action."); + } + + log_notice("Ctrl-Alt-Del was pressed more than 7 times within 2s, performing immediate %s.", + cad_burst_action_to_string(m->cad_burst_action)); + status_printf(NULL, true, false, "Ctrl-Alt-Del was pressed more than 7 times within 2s, performing immediate %s.", + cad_burst_action_to_string(m->cad_burst_action)); + } +} + 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); @@ -1713,18 +1954,21 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t 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 (n >= 0) - return -EIO; - - if (errno == EINTR || errno == EAGAIN) + if (IN_SET(errno, EINTR, EAGAIN)) break; - return -errno; + /* 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 && m->running_as == MANAGER_USER) + (sfsi.ssi_signo == SIGTERM && MANAGER_IS_USER(m)) ? LOG_DEBUG : LOG_INFO, &sfsi); @@ -1735,7 +1979,7 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t break; case SIGTERM: - if (m->running_as == MANAGER_SYSTEM) { + if (MANAGER_IS_SYSTEM(m)) { /* This is for compatibility with the * original sysvinit */ m->exit_code = MANAGER_REEXECUTE; @@ -1745,20 +1989,8 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t /* Fall through */ case SIGINT: - if (m->running_as == MANAGER_SYSTEM) { - - /* If the user presses C-A-D more than - * 7 times within 2s, we reboot - * immediately. */ - - if (ratelimit_test(&m->ctrl_alt_del_ratelimit)) - manager_start_target(m, SPECIAL_CTRL_ALT_DEL_TARGET, JOB_REPLACE_IRREVERSIBLY); - else { - log_notice("Ctrl-Alt-Del was pressed more than 7 times within 2s, rebooting immediately."); - status_printf(NULL, true, false, "Ctrl-Alt-Del was pressed more than 7 times within 2s, rebooting immediately."); - m->exit_code = MANAGER_REBOOT; - } - + if (MANAGER_IS_SYSTEM(m)) { + manager_handle_ctrl_alt_del(m); break; } @@ -1771,14 +2003,14 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t break; case SIGWINCH: - if (m->running_as == MANAGER_SYSTEM) + 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 (m->running_as == MANAGER_SYSTEM) + if (MANAGER_IS_SYSTEM(m)) manager_start_target(m, SPECIAL_SIGPWR_TARGET, JOB_REPLACE); /* This is a nop on non-init */ @@ -1809,20 +2041,16 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t f = open_memstream(&dump, &size); if (!f) { - log_warning("Failed to allocate memory stream."); + log_warning_errno(errno, "Failed to allocate memory stream: %m"); break; } manager_dump_units(m, f, "\t"); manager_dump_jobs(m, f, "\t"); - if (ferror(f)) { - log_warning("Failed to write status stream"); - break; - } - - if (fflush(f)) { - log_warning("Failed to flush status stream"); + r = fflush_and_check(f); + if (r < 0) { + log_warning_errno(r, "Failed to write status stream: %m"); break; } @@ -1872,27 +2100,25 @@ static int manager_dispatch_signal_fd(sd_event_source *source, int fd, uint32_t switch (sfsi.ssi_signo - SIGRTMIN) { case 20: - log_debug("Enabling showing of status."); manager_set_show_status(m, SHOW_STATUS_YES); break; case 21: - log_debug("Disabling showing of status."); manager_set_show_status(m, SHOW_STATUS_NO); break; case 22: log_set_max_level(LOG_DEBUG); - log_notice("Setting log level to debug."); + log_info("Setting log level to debug."); break; case 23: log_set_max_level(LOG_INFO); - log_notice("Setting log level to info."); + log_info("Setting log level to info."); break; case 24: - if (m->running_as == MANAGER_USER) { + if (MANAGER_IS_USER(m)) { m->exit_code = MANAGER_EXIT; return 0; } @@ -1963,7 +2189,6 @@ static int manager_dispatch_idle_pipe_fd(sd_event_source *source, int fd, uint32 m->no_console_output = m->n_on_console > 0; - m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); manager_close_idle_pipe(m); return 0; @@ -1996,8 +2221,7 @@ int manager_loop(Manager *m) { m->exit_code = MANAGER_OK; /* Release the path cache */ - set_free_free(m->unit_path_cache); - m->unit_path_cache = NULL; + m->unit_path_cache = set_free_free(m->unit_path_cache); manager_check_finished(m); @@ -2010,14 +2234,13 @@ int manager_loop(Manager *m) { while (m->exit_code == MANAGER_OK) { usec_t wait_usec; - if (m->runtime_watchdog > 0 && m->running_as == MANAGER_SYSTEM) + 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); - continue; } if (manager_dispatch_load_queue(m) > 0) @@ -2036,7 +2259,7 @@ int manager_loop(Manager *m) { continue; /* Sleep for half the watchdog time */ - if (m->runtime_watchdog > 0 && m->running_as == MANAGER_SYSTEM) { + 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; @@ -2053,6 +2276,7 @@ int manager_loop(Manager *m) { 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; @@ -2064,12 +2288,25 @@ int manager_load_unit_from_dbus_path(Manager *m, const char *s, sd_bus_error *e, 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; } @@ -2107,16 +2344,16 @@ void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success) { 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 (m->n_reloading > 0) - return; - - if (m->running_as != MANAGER_SYSTEM) + if (MANAGER_IS_RELOADING(m)) return; if (u->type != UNIT_SERVICE) @@ -2142,21 +2379,20 @@ void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success) { } void manager_send_unit_plymouth(Manager *m, Unit *u) { - union sockaddr_union sa = PLYMOUTH_SOCKET; - - int n = 0; + 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 (m->n_reloading > 0) + if (MANAGER_IS_RELOADING(m)) return; - if (m->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(m)) return; - if (detect_container(NULL) > 0) + if (detect_container() > 0) return; if (u->type != UNIT_SERVICE && @@ -2172,7 +2408,7 @@ void manager_send_unit_plymouth(Manager *m, Unit *u) { return; } - if (connect(fd, &sa.sa, offsetof(struct sockaddr_un, sun_path) + 1 + strlen(sa.un.sun_path+1)) < 0) { + 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"); @@ -2190,24 +2426,6 @@ void manager_send_unit_plymouth(Manager *m, Unit *u) { log_error_errno(errno, "Failed to write Plymouth message: %m"); } -void manager_dispatch_bus_name_owner_changed( - Manager *m, - const char *name, - const char* old_owner, - const char *new_owner) { - - Unit *u; - - assert(m); - assert(name); - - u = hashmap_get(m->watch_bus, name); - if (!u) - return; - - UNIT_VTABLE(u)->bus_name_owner_change(u, name, old_owner, new_owner); -} - int manager_open_serialization(Manager *m, FILE **_f) { const char *path; int fd = -1; @@ -2215,8 +2433,8 @@ int manager_open_serialization(Manager *m, FILE **_f) { assert(_f); - path = m->running_as == MANAGER_SYSTEM ? "/run/systemd" : "/tmp"; - fd = open_tmpfile(path, O_RDWR|O_CLOEXEC); + path = MANAGER_IS_SYSTEM(m) ? "/run/systemd" : "/tmp"; + fd = open_tmpfile_unlinkable(path, O_RDWR|O_CLOEXEC); if (fd < 0) return -errno; @@ -2244,7 +2462,7 @@ int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { assert(f); assert(fds); - m->n_reloading ++; + 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)); @@ -2290,17 +2508,38 @@ int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { fprintf(f, "notify-socket=%s\n", m->notify_socket); } - if (m->kdbus_fd >= 0) { + if (m->cgroups_agent_fd >= 0) { int copy; - copy = fdset_put_dup(fds, m->kdbus_fd); + copy = fdset_put_dup(fds, m->cgroups_agent_fd); if (copy < 0) return copy; - fprintf(f, "kdbus-fd=%i\n", 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); + 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); @@ -2314,13 +2553,13 @@ int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root) { r = unit_serialize(u, f, fds, !switching_root); if (r < 0) { - m->n_reloading --; + m->n_reloading--; return r; } } assert(m->n_reloading > 0); - m->n_reloading --; + m->n_reloading--; if (ferror(f)) return -EIO; @@ -2340,7 +2579,7 @@ int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { log_debug("Deserializing state..."); - m->n_reloading ++; + m->n_reloading++; for (;;) { char line[LINE_MAX], *l; @@ -2457,25 +2696,42 @@ int manager_deserialize(Manager *m, FILE *f, FDSet *fds) { free(m->notify_socket); m->notify_socket = n; - } else if (startswith(l, "kdbus-fd=")) { + } else if (startswith(l, "cgroups-agent-fd=")) { int fd; - if (safe_atoi(l + 9, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) - log_debug("Failed to parse kdbus fd: %s", l + 9); + 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 { - safe_close(m->kdbus_fd); - m->kdbus_fd = fdset_remove(fds, fd); + 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 { - int k; + } else if (startswith(l, "user-lookup=")) { + int fd0, fd1; - k = bus_track_deserialize_item(&m->deserialized_subscribed, l); - if (k < 0) - log_debug_errno(k, "Failed to deserialize bus tracker object: %m"); - else if (k == 0) - log_debug("Unknown serialization item '%s'", l); - } + 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 (;;) { @@ -2508,7 +2764,7 @@ finish: r = -EIO; assert(m->n_reloading > 0); - m->n_reloading --; + m->n_reloading--; return r; } @@ -2524,51 +2780,48 @@ int manager_reload(Manager *m) { if (r < 0) return r; - m->n_reloading ++; + m->n_reloading++; bus_manager_send_reloading(m, true); fds = fdset_new(); if (!fds) { - m->n_reloading --; + m->n_reloading--; return -ENOMEM; } r = manager_serialize(m, f, fds, false); if (r < 0) { - m->n_reloading --; + m->n_reloading--; return r; } if (fseeko(f, 0, SEEK_SET) < 0) { - m->n_reloading --; + m->n_reloading--; return -errno; } /* From here on there is no way back. */ manager_clear_jobs_and_units(m); - manager_undo_generators(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); - /* Find new unit paths */ - q = manager_run_generators(m); + q = lookup_paths_init(&m->lookup_paths, m->unit_file_scope, 0, NULL); if (q < 0 && r >= 0) r = q; - q = lookup_paths_init( - &m->lookup_paths, m->running_as, true, - NULL, - m->generator_unit_path, - m->generator_unit_path_early, - m->generator_unit_path_late); + /* 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 */ - q = manager_enumerate(m); - if (q < 0 && r >= 0) - r = q; + manager_enumerate(m); /* Second, deserialize our stored data */ q = manager_deserialize(m, f, fds); @@ -2583,9 +2836,28 @@ int manager_reload(Manager *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--; @@ -2594,12 +2866,6 @@ int manager_reload(Manager *m) { return r; } -bool manager_is_reloading_or_reexecuting(Manager *m) { - assert(m); - - return m->n_reloading != 0; -} - void manager_reset_failed(Manager *m) { Unit *u; Iterator i; @@ -2631,7 +2897,7 @@ static void manager_notify_finished(Manager *m) { if (m->test_run) return; - if (m->running_as == MANAGER_SYSTEM && detect_container(NULL) <= 0) { + if (MANAGER_IS_SYSTEM(m) && detect_container() <= 0) { /* Note that m->kernel_usec.monotonic is always at 0, * and m->firmware_usec.monotonic and @@ -2694,12 +2960,9 @@ static void manager_notify_finished(Manager *m) { } void manager_check_finished(Manager *m) { - Unit *u = NULL; - Iterator i; - assert(m); - if (m->n_reloading > 0) + if (MANAGER_IS_RELOADING(m)) return; /* Verify that we are actually running currently. Initially @@ -2709,11 +2972,9 @@ void manager_check_finished(Manager *m) { 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); + (void) sd_event_source_set_time(m->jobs_in_progress_event_source, now(CLOCK_MONOTONIC) + JOBS_IN_PROGRESS_WAIT_USEC); return; } @@ -2721,7 +2982,6 @@ void manager_check_finished(Manager *m) { manager_flip_auto_status(m, false); /* Notify Type=idle units that we are done now */ - m->idle_pipe_event_source = sd_event_source_unref(m->idle_pipe_event_source); manager_close_idle_pipe(m); /* Turn off confirm spawn now */ @@ -2740,83 +3000,7 @@ void manager_check_finished(Manager *m) { manager_notify_finished(m); - SET_FOREACH(u, m->startup_units, i) - if (u->cgroup_path) - cgroup_context_apply(unit_get_cgroup_context(u), unit_get_cgroup_mask(u), u->cgroup_path, manager_state(m)); -} - -static int create_generator_dir(Manager *m, char **generator, const char *name) { - char *p; - int r; - - assert(m); - assert(generator); - assert(name); - - if (*generator) - return 0; - - if (m->running_as == MANAGER_SYSTEM && getpid() == 1) { - /* systemd --system, not running --test */ - - p = strappend("/run/systemd/", name); - if (!p) - return log_oom(); - - r = mkdir_p_label(p, 0755); - if (r < 0) { - log_error_errno(r, "Failed to create generator directory %s: %m", p); - free(p); - return r; - } - } else if (m->running_as == MANAGER_USER) { - const char *s = NULL; - - s = getenv("XDG_RUNTIME_DIR"); - if (!s) - return -EINVAL; - p = strjoin(s, "/systemd/", name, NULL); - if (!p) - return log_oom(); - - r = mkdir_p_label(p, 0755); - if (r < 0) { - log_error_errno(r, "Failed to create generator directory %s: %m", p); - free(p); - return r; - } - } else { - /* systemd --system --test */ - - p = strjoin("/tmp/systemd-", name, ".XXXXXX", NULL); - if (!p) - return log_oom(); - - if (!mkdtemp(p)) { - log_error_errno(errno, "Failed to create generator directory %s: %m", - p); - free(p); - return -errno; - } - } - - *generator = p; - return 0; -} - -static void trim_generator_dir(Manager *m, char **generator) { - assert(m); - assert(generator); - - if (!*generator) - return; - - if (rmdir(*generator) >= 0) { - free(*generator); - *generator = NULL; - } - - return; + manager_invalidate_startup_units(m); } static int manager_run_generators(Manager *m) { @@ -2830,72 +3014,40 @@ static int manager_run_generators(Manager *m) { if (m->test_run) return 0; - paths = generator_paths(m->running_as); + 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) { - r = access(*path, F_OK); - if (r == 0) + 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 = create_generator_dir(m, &m->generator_unit_path, "generator"); - if (r < 0) - goto finish; - - r = create_generator_dir(m, &m->generator_unit_path_early, "generator.early"); - if (r < 0) - goto finish; - - r = create_generator_dir(m, &m->generator_unit_path_late, "generator.late"); + 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->generator_unit_path; - argv[2] = m->generator_unit_path_early; - argv[3] = m->generator_unit_path_late; + 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: - trim_generator_dir(m, &m->generator_unit_path); - trim_generator_dir(m, &m->generator_unit_path_early); - trim_generator_dir(m, &m->generator_unit_path_late); + lookup_paths_trim_generator(&m->lookup_paths); return r; } -static void remove_generator_dir(Manager *m, char **generator) { - assert(m); - assert(generator); - - if (!*generator) - return; - - strv_remove(m->lookup_paths.unit_path, *generator); - (void) rm_rf(*generator, REMOVE_ROOT); - - free(*generator); - *generator = NULL; -} - -static void manager_undo_generators(Manager *m) { - assert(m); - - remove_generator_dir(m, &m->generator_unit_path); - remove_generator_dir(m, &m->generator_unit_path_early); - remove_generator_dir(m, &m->generator_unit_path_late); -} - int manager_environment_add(Manager *m, char **minus, char **plus) { char **a = NULL, **b = NULL, **l; assert(m); @@ -2940,6 +3092,8 @@ int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit) { assert(m); for (i = 0; i < _RLIMIT_MAX; i++) { + m->rlimit[i] = mfree(m->rlimit[i]); + if (!default_rlimit[i]) continue; @@ -2956,7 +3110,7 @@ void manager_recheck_journal(Manager *m) { assert(m); - if (m->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(m)) return; u = manager_get_unit(m, SPECIAL_JOURNALD_SOCKET); @@ -2980,21 +3134,24 @@ 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 (m->running_as != MANAGER_SYSTEM) + 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) - touch("/run/systemd/show-status"); + (void) touch("/run/systemd/show-status"); else - unlink("/run/systemd/show-status"); + (void) unlink("/run/systemd/show-status"); } static bool manager_get_show_status(Manager *m, StatusType type) { assert(m); - if (m->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(m)) return false; if (m->no_console_output) @@ -3016,15 +3173,17 @@ static bool manager_get_show_status(Manager *m, StatusType type) { void manager_set_first_boot(Manager *m, bool b) { assert(m); - if (m->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(m)) return; - m->first_boot = b; + if (m->first_boot != (int) b) { + if (b) + (void) touch("/run/systemd/first-boot"); + else + (void) unlink("/run/systemd/first-boot"); + } - if (m->first_boot) - touch("/run/systemd/first-boot"); - else - unlink("/run/systemd/first-boot"); + m->first_boot = b; } void manager_status_printf(Manager *m, StatusType type, const char *status, const char *format, ...) { @@ -3045,30 +3204,6 @@ void manager_status_printf(Manager *m, StatusType type, const char *status, cons va_end(ap); } -int manager_get_unit_by_path(Manager *m, const char *path, const char *suffix, Unit **_found) { - _cleanup_free_ char *p = NULL; - Unit *found; - int r; - - assert(m); - assert(path); - assert(suffix); - assert(_found); - - r = unit_name_from_path(path, suffix, &p); - if (r < 0) - return r; - - found = manager_get_unit(m, p); - if (!found) { - *_found = NULL; - return 0; - } - - *_found = found; - return 1; -} - Set *manager_get_units_requiring_mounts_for(Manager *m, const char *path) { char p[strlen(path)+1]; @@ -3084,13 +3219,14 @@ Set *manager_get_units_requiring_mounts_for(Manager *m, const char *path) { const char *manager_get_runtime_prefix(Manager *m) { assert(m); - return m->running_as == MANAGER_SYSTEM ? + return MANAGER_IS_SYSTEM(m) ? "/run" : getenv("XDG_RUNTIME_DIR"); } -void manager_update_failed_units(Manager *m, Unit *u, bool failed) { +int manager_update_failed_units(Manager *m, Unit *u, bool failed) { unsigned size; + int r; assert(m); assert(u->manager == m); @@ -3098,13 +3234,19 @@ void manager_update_failed_units(Manager *m, Unit *u, bool failed) { 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) - log_oom(); + return log_oom(); } else - set_remove(m->failed_units, u); + (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) { @@ -3124,18 +3266,18 @@ ManagerState manager_state(Manager *m) { /* 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_TRY_RESTART, JOB_RELOAD_OR_START)) + 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_TRY_RESTART, JOB_RELOAD_OR_START)))) + (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_TRY_RESTART, JOB_RELOAD_OR_START)))) + (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 */ @@ -3145,6 +3287,300 @@ ManagerState manager_state(Manager *m) { 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", @@ -3155,3 +3591,11 @@ static const char *const manager_state_table[_MANAGER_STATE_MAX] = { }; DEFINE_STRING_TABLE_LOOKUP(manager_state, ManagerState); + +static const char *const cad_burst_action_table[_CAD_BURST_ACTION_MAX] = { + [CAD_BURST_ACTION_IGNORE] = "ignore", + [CAD_BURST_ACTION_REBOOT] = "reboot-force", + [CAD_BURST_ACTION_POWEROFF] = "poweroff-force", +}; + +DEFINE_STRING_TABLE_LOOKUP(cad_burst_action, CADBurstAction); diff --git a/src/core/manager.h b/src/core/manager.h index 4ef869d14a..29fe14e10b 100644 --- a/src/core/manager.h +++ b/src/core/manager.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -21,13 +19,15 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include <libmount.h> #include <stdbool.h> #include <stdio.h> #include "sd-bus.h" #include "sd-event.h" -#include "fdset.h" + #include "cgroup-util.h" +#include "fdset.h" #include "hashmap.h" #include "list.h" #include "ratelimit.h" @@ -62,17 +62,25 @@ typedef enum ManagerExitCode { _MANAGER_EXIT_CODE_INVALID = -1 } ManagerExitCode; +typedef enum CADBurstAction { + CAD_BURST_ACTION_IGNORE, + CAD_BURST_ACTION_REBOOT, + CAD_BURST_ACTION_POWEROFF, + _CAD_BURST_ACTION_MAX, + _CAD_BURST_ACTION_INVALID = -1 +} CADBurstAction; + typedef enum StatusType { STATUS_TYPE_EPHEMERAL, STATUS_TYPE_NORMAL, STATUS_TYPE_EMERGENCY, } StatusType; +#include "execute.h" #include "job.h" #include "path-lookup.h" -#include "execute.h" -#include "unit-name.h" #include "show-status.h" +#include "unit-name.h" struct Manager { /* Note that the set of units we know of is allowed to be @@ -81,6 +89,7 @@ struct Manager { /* 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 @@ -132,6 +141,9 @@ struct Manager { 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; @@ -140,8 +152,10 @@ struct Manager { sd_event_source *jobs_in_progress_event_source; - unsigned n_snapshots; + int user_lookup_fds[2]; + sd_event_source *user_lookup_event_source; + UnitFileScope unit_file_scope; LookupPaths lookup_paths; Set *unit_path_cache; @@ -164,10 +178,6 @@ struct Manager { dual_timestamp units_load_start_timestamp; dual_timestamp units_load_finish_timestamp; - char *generator_unit_path; - char *generator_unit_path_early; - char *generator_unit_path_late; - struct udev* udev; /* Data specific to the device subsystem */ @@ -176,10 +186,8 @@ struct Manager { Hashmap *devices_by_sysfs; /* Data specific to the mount subsystem */ - FILE *proc_self_mountinfo; + struct libmnt_monitor *mount_monitor; sd_event_source *mount_event_source; - int utab_inotify_fd; - sd_event_source *mount_utab_event_source; /* Data specific to the swap filesystem */ FILE *proc_swaps; @@ -215,28 +223,36 @@ struct Manager { /* Data specific to the cgroup subsystem */ Hashmap *cgroup_unit; - CGroupControllerMask cgroup_supported; + CGroupMask cgroup_supported; char *cgroup_root; - int gc_marker; - unsigned n_in_gc_queue; + /* 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 */ - ManagerRunningAs running_as; ManagerExitCode exit_code:5; bool dispatching_load_queue:1; bool dispatching_dbus_queue:1; bool taint_usr:1; - bool first_boot: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; @@ -250,8 +266,11 @@ struct Manager { 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]; @@ -284,38 +303,52 @@ struct Manager { * value where Unit objects are contained. */ Hashmap *units_requiring_mounts_for; - /* Reference to the kdbus bus control fd */ - int kdbus_fd; - /* Used for processing polkit authorization responses */ Hashmap *polkit_registry; - /* When the user hits C-A-D more than 7 times per 2s, reboot immediately... */ + /* 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; + CADBurstAction 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 */ }; -int manager_new(ManagerRunningAs running_as, bool test_run, Manager **m); +#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); -int manager_enumerate(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_unit_by_path(Manager *m, const char *path, const char *suffix, Unit **_found); - 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, bool force, sd_bus_error *e, Job **_ret); -int manager_add_job_by_name(Manager *m, JobType type, const char *name, JobMode mode, bool force, sd_bus_error *e, Job **_ret); +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); @@ -329,8 +362,6 @@ int manager_set_default_rlimits(Manager *m, struct rlimit **default_rlimit); int manager_loop(Manager *m); -void manager_dispatch_bus_name_owner_changed(Manager *m, const char *name, const char* old_owner, const char *new_owner); - int manager_open_serialization(Manager *m, FILE **_f); int manager_serialize(Manager *m, FILE *f, FDSet *fds, bool switching_root); @@ -338,8 +369,6 @@ int manager_deserialize(Manager *m, FILE *f, FDSet *fds); int manager_reload(Manager *m); -bool manager_is_reloading_or_reexecuting(Manager *m) _pure_; - void manager_reset_failed(Manager *m); void manager_send_unit_audit(Manager *m, Unit *u, int type, bool success); @@ -363,7 +392,25 @@ const char *manager_get_runtime_prefix(Manager *m); ManagerState manager_state(Manager *m); -void manager_update_failed_units(Manager *m, Unit *u, bool failed); +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_; + +const char *cad_burst_action_to_string(CADBurstAction a) _const_; +CADBurstAction cad_burst_action_from_string(const char *s) _pure_; diff --git a/src/core/mount-setup.c b/src/core/mount-setup.c index 1782d40720..ca63a93e8b 100644 --- a/src/core/mount-setup.c +++ b/src/core/mount-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,28 +17,32 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/mount.h> #include <errno.h> +#include <ftw.h> #include <stdlib.h> +#include <sys/mount.h> #include <unistd.h> -#include <ftw.h> -#include "mount-setup.h" -#include "dev-setup.h" +#include "alloc-util.h" #include "bus-util.h" +#include "cgroup-util.h" +#include "dev-setup.h" +#include "efivars.h" +#include "fs-util.h" +#include "label.h" #include "log.h" #include "macro.h" -#include "util.h" -#include "label.h" -#include "set.h" -#include "strv.h" +#include "missing.h" #include "mkdir.h" +#include "mount-setup.h" +#include "mount-util.h" #include "path-util.h" -#include "missing.h" -#include "virt.h" -#include "efivars.h" +#include "set.h" #include "smack-util.h" -#include "cgroup-util.h" +#include "strv.h" +#include "user-util.h" +#include "util.h" +#include "virt.h" typedef enum MountMode { MNT_NONE = 0, @@ -93,20 +95,22 @@ static const MountPoint mount_table[] = { #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, - NULL, MNT_FATAL|MNT_IN_CONTAINER }, + 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, - NULL, MNT_IN_CONTAINER }, + cg_is_legacy_systemd_controller_wanted, MNT_IN_CONTAINER }, { "cgroup", "/sys/fs/cgroup/systemd", "cgroup", "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV, - NULL, MNT_FATAL|MNT_IN_CONTAINER }, + 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 - { "kdbusfs", "/sys/fs/kdbus", "kdbusfs", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, - is_kdbus_wanted, MNT_IN_CONTAINER }, }; /* These are API file systems that might be mounted by other software, @@ -153,24 +157,26 @@ static int mount_one(const MountPoint *p, bool relabel) { /* Relabel first, just in case */ if (relabel) - label_fix(p->where, true, true); + (void) label_fix(p->where, true, true); r = path_is_mount_point(p->where, AT_SYMLINK_FOLLOW); - if (r < 0 && r != -ENOENT) - return r; + 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(NULL) > 0) + 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) - mkdir_p_label(p->where, 0755); + (void) mkdir_p_label(p->where, 0755); else - mkdir_p(p->where, 0755); + (void) mkdir_p(p->where, 0755); log_debug("Mounting %s to %s of type %s with options %s.", p->what, @@ -183,40 +189,47 @@ static int mount_one(const MountPoint *p, bool relabel) { p->type, p->flags, p->options) < 0) { - log_full((p->mode & MNT_FATAL) ? LOG_ERR : LOG_DEBUG, "Failed to mount %s at %s: %m", p->type, p->where); + 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) - label_fix(p->where, false, false); + (void) label_fix(p->where, false, false); return 1; } -int mount_setup_early(void) { +static int mount_points_setup(unsigned n, bool loaded_policy) { unsigned i; int r = 0; - 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 */ - for (i = 0; i < N_EARLY_MOUNT; i ++) { + for (i = 0; i < n; i ++) { int j; - j = mount_one(mount_table + i, false); - if (r == 0) + 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); @@ -296,7 +309,17 @@ int mount_cgroup_controllers(char ***join_controllers) { return log_oom(); r = symlink(options, t); - if (r < 0 && errno != EEXIST) + 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); } } @@ -334,16 +357,9 @@ static int nftw_cb( #endif int mount_setup(bool loaded_policy) { - unsigned i; int r = 0; - for (i = 0; i < ELEMENTSOF(mount_table); i ++) { - int j; - - j = mount_one(mount_table + i, loaded_policy); - if (r == 0) - r = j; - } + r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy); if (r < 0) return r; @@ -360,6 +376,7 @@ int mount_setup(bool loaded_policy) { 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); @@ -380,7 +397,7 @@ int mount_setup(bool loaded_policy) { * 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(NULL) <= 0) + 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"); @@ -389,9 +406,16 @@ int mount_setup(bool loaded_policy) { * really needs to stay for good, otherwise software that * copied sd-daemon.c into their sources will misdetect * systemd. */ - mkdir_label("/run/systemd", 0755); - mkdir_label("/run/systemd/system", 0755); - mkdir_label("/run/systemd/inaccessible", 0000); + (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/core/mount-setup.h b/src/core/mount-setup.h index b32fbc5a52..647bd770ae 100644 --- a/src/core/mount-setup.h +++ b/src/core/mount-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/mount.c b/src/core/mount.c index 851b41351e..15619dffe3 100644 --- a/src/core/mount.c +++ b/src/core/mount.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,27 +18,33 @@ ***/ #include <errno.h> +#include <signal.h> #include <stdio.h> #include <sys/epoll.h> -#include <signal.h> -#include <libmount.h> -#include <sys/inotify.h> -#include "manager.h" -#include "unit.h" -#include "mount.h" -#include "log.h" #include "sd-messages.h" -#include "strv.h" -#include "mkdir.h" -#include "path-util.h" -#include "mount-setup.h" -#include "unit-name.h" + +#include "alloc-util.h" #include "dbus-mount.h" -#include "special.h" +#include "escape.h" #include "exit-status.h" -#include "fstab-util.h" #include "formats-util.h" +#include "fstab-util.h" +#include "log.h" +#include "manager.h" +#include "mkdir.h" +#include "mount-setup.h" +#include "mount-util.h" +#include "mount.h" +#include "parse-util.h" +#include "path-util.h" +#include "process-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "unit-name.h" +#include "unit.h" #define RETRY_UMOUNT_MAX 32 @@ -82,6 +86,15 @@ static bool mount_is_network(const MountParameters *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); @@ -100,6 +113,28 @@ static bool mount_is_auto(const MountParameters *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); @@ -146,29 +181,27 @@ static void mount_init(Unit *u) { u->ignore_on_isolate = true; } -static int mount_arm_timer(Mount *m) { +static int mount_arm_timer(Mount *m, usec_t usec) { int r; assert(m); - if (m->timeout_usec <= 0) { - m->timer_event_source = sd_event_source_unref(m->timer_event_source); - return 0; - } - if (m->timer_event_source) { - r = sd_event_source_set_time(m->timer_event_source, now(CLOCK_MONOTONIC) + m->timeout_usec); + 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, - now(CLOCK_MONOTONIC) + m->timeout_usec, 0, + usec, 0, mount_dispatch_timer, m); if (r < 0) return r; @@ -203,8 +236,7 @@ static void mount_done(Unit *u) { assert(m); - free(m->where); - m->where = NULL; + m->where = mfree(m->where); mount_parameters_done(&m->parameters_proc_self_mountinfo); mount_parameters_done(&m->parameters_fragment); @@ -213,6 +245,8 @@ static void mount_done(Unit *u) { 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); @@ -249,9 +283,10 @@ static int mount_add_mount_links(Mount *m) { if (!path_equal(m->where, "/")) { /* Adds in links to other mount points that might lie further * up in the hierarchy */ - r = path_get_parent(m->where, &parent); - if (r < 0) - return r; + + parent = dirname_malloc(m->where); + if (!parent) + return -ENOMEM; r = unit_require_mounts_for(UNIT(m), parent); if (r < 0) @@ -259,12 +294,12 @@ static int mount_add_mount_links(Mount *m) { } /* Adds in links to other mount points that might be needed - * for the source path (if this is a bind mount) to be + * 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_network(pm)) { + (mount_is_bind(pm) || mount_is_loop(pm) || !mount_is_network(pm))) { r = unit_require_mounts_for(UNIT(m), pm->what); if (r < 0) @@ -326,10 +361,10 @@ static int mount_add_device_links(Mount *m) { if (path_equal(m->where, "/")) return 0; - if (mount_is_auto(p) && UNIT(m)->manager->running_as == MANAGER_SYSTEM) + 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); + 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; @@ -342,7 +377,7 @@ static int mount_add_quota_links(Mount *m) { assert(m); - if (UNIT(m)->manager->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(UNIT(m)->manager)) return 0; p = get_mount_parameters_fragment(m); @@ -366,8 +401,8 @@ static int mount_add_quota_links(Mount *m) { static bool should_umount(Mount *m) { MountParameters *p; - if (path_equal(m->where, "/") || - path_equal(m->where, "/usr")) + if (PATH_IN_SET(m->where, "/", "/usr") || + path_startswith(m->where, "/run/initramfs")) return false; p = get_mount_parameters(m); @@ -379,22 +414,26 @@ static bool should_umount(Mount *m) { } static int mount_add_default_dependencies(Mount *m) { - const char *after, *after2, *online; MountParameters *p; + const char *after; int r; assert(m); - if (UNIT(m)->manager->running_as != MANAGER_SYSTEM) + if (!UNIT(m)->default_dependencies) + return 0; + + if (!MANAGER_IS_SYSTEM(UNIT(m)->manager)) return 0; - /* We do not add any default dependencies to / and /usr, 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_equal(m->where, "/") || - path_equal(m->where, "/usr") || + /* 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")) @@ -405,30 +444,34 @@ static int mount_add_default_dependencies(Mount *m) { return 0; if (mount_is_network(p)) { - after = SPECIAL_REMOTE_FS_PRE_TARGET; - after2 = SPECIAL_NETWORK_TARGET; - online = SPECIAL_NETWORK_ONLINE_TARGET; - } else { - after = SPECIAL_LOCAL_FS_PRE_TARGET; - after2 = NULL; - online = NULL; - } - - r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after, NULL, true); - if (r < 0) - return r; + /* 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. */ - if (after2) { - r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, after2, NULL, true); + r = unit_add_dependency_by_name(UNIT(m), UNIT_AFTER, SPECIAL_NETWORK_TARGET, NULL, true); if (r < 0) return r; - } - if (online) { - r = unit_add_two_dependencies_by_name(UNIT(m), UNIT_WANTS, UNIT_AFTER, online, NULL, true); + /* 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); @@ -441,6 +484,7 @@ static int mount_add_default_dependencies(Mount *m) { static int mount_verify(Mount *m) { _cleanup_free_ char *e = NULL; + MountParameters *p; int r; assert(m); @@ -465,7 +509,8 @@ static int mount_verify(Mount *m) { return -EINVAL; } - if (UNIT(m)->fragment_path && !m->parameters_fragment.what) { + p = get_mount_parameters_fragment(m); + if (p && !p->what) { log_unit_error(UNIT(m), "What= setting is missing. Refusing."); return -EBADMSG; } @@ -521,15 +566,13 @@ static int mount_add_extras(Mount *m) { if (r < 0) return r; - r = unit_add_default_slice(u, &m->cgroup_context); + r = unit_set_default_slice(u); if (r < 0) return r; - if (u->default_dependencies) { - r = mount_add_default_dependencies(m); - if (r < 0) - return r; - } + r = mount_add_default_dependencies(m); + if (r < 0) + return r; return 0; } @@ -559,23 +602,6 @@ static int mount_load(Unit *u) { return mount_verify(m); } -static int mount_notify_automount(Mount *m, MountState old_state, MountState state) { - Unit *p; - int r; - Iterator i; - - assert(m); - - SET_FOREACH(p, UNIT(m)->dependencies[UNIT_TRIGGERED_BY], i) - if (p->type == UNIT_AUTOMOUNT) { - r = automount_update_mount(AUTOMOUNT(p), old_state, state); - if (r < 0) - return r; - } - - return 0; -} - static void mount_set_state(Mount *m, MountState state) { MountState old_state; assert(m); @@ -583,24 +609,13 @@ static void mount_set_state(Mount *m, MountState state) { old_state = m->state; m->state = state; - if (state != MOUNT_MOUNTING && - state != MOUNT_MOUNTING_DONE && - state != MOUNT_REMOUNTING && - state != MOUNT_UNMOUNTING && - state != MOUNT_MOUNTING_SIGTERM && - state != MOUNT_MOUNTING_SIGKILL && - state != MOUNT_UNMOUNTING_SIGTERM && - state != MOUNT_UNMOUNTING_SIGKILL && - state != MOUNT_REMOUNTING_SIGTERM && - state != MOUNT_REMOUNTING_SIGKILL) { + 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; } - mount_notify_automount(m, old_state, state); - if (state != old_state) log_unit_debug(UNIT(m), "Changed %s -> %s", mount_state_to_string(old_state), mount_state_to_string(state)); @@ -624,29 +639,22 @@ static int mount_coldplug(Unit *u) { if (new_state == m->state) return 0; - if (new_state == MOUNT_MOUNTING || - new_state == MOUNT_MOUNTING_DONE || - new_state == MOUNT_REMOUNTING || - new_state == MOUNT_UNMOUNTING || - new_state == MOUNT_MOUNTING_SIGTERM || - new_state == MOUNT_MOUNTING_SIGKILL || - new_state == MOUNT_UNMOUNTING_SIGTERM || - new_state == MOUNT_UNMOUNTING_SIGKILL || - new_state == MOUNT_REMOUNTING_SIGTERM || - new_state == MOUNT_REMOUNTING_SIGKILL) { - - if (m->control_pid <= 0) - return -EBADMSG; + 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); + 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; } @@ -669,7 +677,10 @@ static void mount_dump(Unit *u, FILE *f, const char *prefix) { "%sOptions: %s\n" "%sFrom /proc/self/mountinfo: %s\n" "%sFrom fragment: %s\n" - "%sDirectoryMode: %04o\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, @@ -678,7 +689,10 @@ static void mount_dump(Unit *u, FILE *f, const char *prefix) { 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, 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, @@ -693,10 +707,10 @@ static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) { pid_t pid; int r; ExecParameters exec_params = { - .apply_permissions = true, - .apply_chroot = true, - .apply_tty_stdin = true, - .bus_endpoint_fd = -1, + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, }; assert(m); @@ -711,14 +725,18 @@ static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) { r = unit_setup_exec_runtime(UNIT(m)); if (r < 0) - goto fail; + return r; - r = mount_arm_timer(m); + r = unit_setup_dynamic_creds(UNIT(m)); if (r < 0) - goto fail; + 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.confirm_spawn = UNIT(m)->manager->confirm_spawn; + 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; @@ -729,43 +747,43 @@ static int mount_spawn(Mount *m, ExecCommand *c, pid_t *_pid) { &m->exec_context, &exec_params, m->exec_runtime, + &m->dynamic_creds, &pid); if (r < 0) - goto fail; + return r; r = unit_watch_pid(UNIT(m), pid); if (r < 0) /* FIXME: we need to do something here */ - goto fail; + return r; *_pid = pid; return 0; - -fail: - m->timer_event_source = sd_event_source_unref(m->timer_event_source); - - return r; } static void mount_enter_dead(Mount *m, MountResult f) { assert(m); - if (f != MOUNT_SUCCESS) + 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)); - mount_set_state(m, m->result != MOUNT_SUCCESS ? MOUNT_FAILED : MOUNT_DEAD); + 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 (f != MOUNT_SUCCESS) + if (m->result == MOUNT_SUCCESS) m->result = f; mount_set_state(m, MOUNT_MOUNTED); @@ -776,7 +794,7 @@ static void mount_enter_signal(Mount *m, MountState state, MountResult f) { assert(m); - if (f != MOUNT_SUCCESS) + if (m->result == MOUNT_SUCCESS) m->result = f; r = unit_kill_context( @@ -791,7 +809,7 @@ static void mount_enter_signal(Mount *m, MountState state, MountResult f) { goto fail; if (r > 0) { - r = mount_arm_timer(m); + r = mount_arm_timer(m, usec_add(now(CLOCK_MONOTONIC), m->timeout_usec)); if (r < 0) goto fail; @@ -812,7 +830,7 @@ static void mount_enter_signal(Mount *m, MountState state, MountResult f) { fail: log_unit_warning_errno(UNIT(m), r, "Failed to kill processes: %m"); - if (state == MOUNT_REMOUNTING_SIGTERM || state == MOUNT_REMOUNTING_SIGKILL) + 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); @@ -834,8 +852,10 @@ static void mount_enter_unmounting(Mount *m) { m->control_command = m->exec_command + MOUNT_EXEC_UNMOUNT; r = exec_command_set(m->control_command, UMOUNT_PATH, m->where, NULL); - if (r >= 0 && UNIT(m)->manager->running_as == MANAGER_SYSTEM) - r = exec_command_append(m->control_command, "-n", 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; @@ -876,22 +896,18 @@ static void mount_enter_mounting(Mount *m) { if (p && mount_is_bind(p)) (void) mkdir_p_label(p->what, m->directory_mode); - if (m->from_fragment) { + if (p) { _cleanup_free_ char *opts = NULL; - r = fstab_filter_options(m->parameters_fragment.options, - "nofail\0" "noauto\0" "auto\0", NULL, NULL, &opts); + 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, - m->parameters_fragment.what, m->where, NULL); - if (r >= 0 && UNIT(m)->manager->running_as == MANAGER_SYSTEM) - r = exec_command_append(m->control_command, "-n", NULL); + 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 && m->parameters_fragment.fstype) - r = exec_command_append(m->control_command, "-t", m->parameters_fragment.fstype, 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 @@ -917,29 +933,29 @@ fail: 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; - if (m->from_fragment) { + p = get_mount_parameters_fragment(m); + if (p) { const char *o; - if (m->parameters_fragment.options) - o = strjoina("remount,", m->parameters_fragment.options); + if (p->options) + o = strjoina("remount,", p->options); else o = "remount"; r = exec_command_set(m->control_command, MOUNT_PATH, - m->parameters_fragment.what, m->where, + p->what, m->where, "-o", o, NULL); - if (r >= 0 && UNIT(m)->manager->running_as == MANAGER_SYSTEM) - r = exec_command_append(m->control_command, "-n", NULL); if (r >= 0 && m->sloppy_options) r = exec_command_append(m->control_command, "-s", NULL); - if (r >= 0 && m->parameters_fragment.fstype) - r = exec_command_append(m->control_command, "-t", m->parameters_fragment.fstype, NULL); + if (r >= 0 && p->fstype) + r = exec_command_append(m->control_command, "-t", p->fstype, NULL); } else r = -ENOENT; @@ -964,23 +980,35 @@ fail: 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 (m->state == MOUNT_UNMOUNTING || - m->state == MOUNT_UNMOUNTING_SIGTERM || - m->state == MOUNT_UNMOUNTING_SIGKILL || - m->state == MOUNT_MOUNTING_SIGTERM || - m->state == MOUNT_MOUNTING_SIGKILL) + 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(m->state == MOUNT_DEAD || m->state == MOUNT_FAILED); + 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; @@ -996,19 +1024,21 @@ static int mount_stop(Unit *u) { assert(m); /* Already on it */ - if (m->state == MOUNT_UNMOUNTING || - m->state == MOUNT_UNMOUNTING_SIGKILL || - m->state == MOUNT_UNMOUNTING_SIGTERM || - m->state == MOUNT_MOUNTING_SIGTERM || - m->state == MOUNT_MOUNTING_SIGKILL) + if (IN_SET(m->state, + MOUNT_UNMOUNTING, + MOUNT_UNMOUNTING_SIGKILL, + MOUNT_UNMOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGTERM, + MOUNT_MOUNTING_SIGKILL)) return 0; - assert(m->state == MOUNT_MOUNTING || - m->state == MOUNT_MOUNTING_DONE || - m->state == MOUNT_MOUNTED || - m->state == MOUNT_REMOUNTING || - m->state == MOUNT_REMOUNTING_SIGTERM || - m->state == MOUNT_REMOUNTING_SIGKILL); + 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; @@ -1025,7 +1055,7 @@ static int mount_reload(Unit *u) { assert(m->state == MOUNT_MOUNTED); mount_enter_remounting(m); - return 0; + return 1; } static int mount_serialize(Unit *u, FILE *f, FDSet *fds) { @@ -1136,7 +1166,7 @@ static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) { m->control_pid = 0; - if (is_clean_exit(code, status, NULL)) + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) f = MOUNT_SUCCESS; else if (code == CLD_EXITED) f = MOUNT_FAILURE_EXIT_CODE; @@ -1147,7 +1177,7 @@ static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) { else assert_not_reached("Unknown code"); - if (f != MOUNT_SUCCESS) + if (m->result == MOUNT_SUCCESS) m->result = f; if (m->control_command) { @@ -1174,9 +1204,10 @@ static void mount_sigchld_event(Unit *u, pid_t pid, int code, int status) { case MOUNT_MOUNTING_SIGKILL: case MOUNT_MOUNTING_SIGTERM: - if (f == MOUNT_SUCCESS) - mount_enter_mounted(m, f); - else if (m->from_proc_self_mountinfo) + 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); @@ -1382,7 +1413,7 @@ static int mount_setup_unit( goto fail; } - if (m->running_as == MANAGER_SYSTEM) { + if (MANAGER_IS_SYSTEM(m)) { const char* target; target = mount_needs_network(options, fstype) ? SPECIAL_REMOTE_FS_TARGET : SPECIAL_LOCAL_FS_TARGET; @@ -1410,7 +1441,7 @@ static int mount_setup_unit( } } - if (m->running_as == MANAGER_SYSTEM && + 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 @@ -1542,92 +1573,91 @@ static int mount_load_proc_self_mountinfo(Manager *m, bool set_flags) { } static void mount_shutdown(Manager *m) { + assert(m); m->mount_event_source = sd_event_source_unref(m->mount_event_source); - m->mount_utab_event_source = sd_event_source_unref(m->mount_utab_event_source); - if (m->proc_self_mountinfo) { - fclose(m->proc_self_mountinfo); - m->proc_self_mountinfo = NULL; - } - m->utab_inotify_fd = safe_close(m->utab_inotify_fd); + mnt_unref_monitor(m->mount_monitor); + m->mount_monitor = NULL; } -static int mount_get_timeout(Unit *u, uint64_t *timeout) { +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, timeout); + 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 mount_enumerate(Manager *m) { +static void mount_enumerate(Manager *m) { int r; + assert(m); mnt_init_debug(0); - if (!m->proc_self_mountinfo) { - m->proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); - if (!m->proc_self_mountinfo) - return -errno; + if (!m->mount_monitor) { + int fd; - r = sd_event_add_io(m->event, &m->mount_event_source, fileno(m->proc_self_mountinfo), EPOLLPRI, mount_dispatch_io, m); - if (r < 0) + m->mount_monitor = mnt_new_monitor(); + if (!m->mount_monitor) { + log_oom(); goto fail; + } - /* Dispatch this before we dispatch SIGCHLD, so that - * we always get the events from /proc/self/mountinfo - * before the SIGCHLD of /usr/bin/mount. */ - r = sd_event_source_set_priority(m->mount_event_source, -10); - if (r < 0) + 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; + } - (void) sd_event_source_set_description(m->mount_event_source, "mount-mountinfo-dispatch"); - } - - if (m->utab_inotify_fd < 0) { - m->utab_inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); - if (m->utab_inotify_fd < 0) { - r = -errno; + 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; } - (void) mkdir_p_label("/run/mount", 0755); - - r = inotify_add_watch(m->utab_inotify_fd, "/run/mount", IN_MOVED_TO); + /* mnt_unref_monitor() will close the fd */ + fd = r = mnt_monitor_get_fd(m->mount_monitor); if (r < 0) { - r = -errno; + log_error_errno(r, "Failed to acquire watch file descriptor: %m"); goto fail; } - r = sd_event_add_io(m->event, &m->mount_utab_event_source, m->utab_inotify_fd, EPOLLIN, mount_dispatch_io, m); - if (r < 0) + 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_utab_event_source, -10); - if (r < 0) + 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_utab_event_source, "mount-utab-dispatch"); + (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 0; + return; fail: mount_shutdown(m); - return r; } static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, void *userdata) { @@ -1639,45 +1669,27 @@ static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, int r; assert(m); - assert(revents & (EPOLLPRI | EPOLLIN)); - - /* The manager calls this for every fd event happening on the - * /proc/self/mountinfo file, which informs us about mounting - * table changes, and for /run/mount events which we watch - * for mount options. */ + assert(revents & EPOLLIN); - if (fd == m->utab_inotify_fd) { + if (fd == mnt_monitor_get_fd(m->mount_monitor)) { bool rescan = false; - /* FIXME: We *really* need to replace this with - * libmount's own API for this, we should not hardcode - * internal behaviour of libmount here. */ - - for (;;) { - union inotify_event_buffer buffer; - struct inotify_event *e; - ssize_t l; - - l = read(fd, &buffer, sizeof(buffer)); - if (l < 0) { - if (errno == EAGAIN || errno == EINTR) - break; - - log_error_errno(errno, "Failed to read utab inotify: %m"); - break; - } - - FOREACH_INOTIFY_EVENT(e, buffer, l) { - /* Only care about changes to utab, - * but we have to monitor the - * directory to reliably get - * notifications about when utab is - * replaced using rename(2) */ - if ((e->mask & IN_Q_OVERFLOW) || streq(e->name, "utab")) - rescan = true; - } - } - + /* 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; } @@ -1722,6 +1734,7 @@ static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, /* 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; @@ -1737,9 +1750,10 @@ static int mount_dispatch_io(sd_event_source *source, int fd, uint32_t revents, case MOUNT_DEAD: case MOUNT_FAILED: - /* This has just been mounted by - * somebody else, follow the state - * change. */ + + /* 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; @@ -1798,23 +1812,13 @@ 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 const char* const mount_state_table[_MOUNT_STATE_MAX] = { - [MOUNT_DEAD] = "dead", - [MOUNT_MOUNTING] = "mounting", - [MOUNT_MOUNTING_DONE] = "mounting-done", - [MOUNT_MOUNTED] = "mounted", - [MOUNT_REMOUNTING] = "remounting", - [MOUNT_UNMOUNTING] = "unmounting", - [MOUNT_MOUNTING_SIGTERM] = "mounting-sigterm", - [MOUNT_MOUNTING_SIGKILL] = "mounting-sigkill", - [MOUNT_REMOUNTING_SIGTERM] = "remounting-sigterm", - [MOUNT_REMOUNTING_SIGKILL] = "remounting-sigkill", - [MOUNT_UNMOUNTING_SIGTERM] = "unmounting-sigterm", - [MOUNT_UNMOUNTING_SIGKILL] = "unmounting-sigkill", - [MOUNT_FAILED] = "failed" -}; +static int mount_control_pid(Unit *u) { + Mount *m = MOUNT(u); -DEFINE_STRING_TABLE_LOOKUP(mount_state, MountState); + assert(m); + + return m->control_pid; +} static const char* const mount_exec_command_table[_MOUNT_EXEC_COMMAND_MAX] = { [MOUNT_EXEC_MOUNT] = "ExecMount", @@ -1830,7 +1834,8 @@ static const char* const mount_result_table[_MOUNT_RESULT_MAX] = { [MOUNT_FAILURE_TIMEOUT] = "timeout", [MOUNT_FAILURE_EXIT_CODE] = "exit-code", [MOUNT_FAILURE_SIGNAL] = "signal", - [MOUNT_FAILURE_CORE_DUMP] = "core-dump" + [MOUNT_FAILURE_CORE_DUMP] = "core-dump", + [MOUNT_FAILURE_START_LIMIT_HIT] = "start-limit-hit", }; DEFINE_STRING_TABLE_LOOKUP(mount_result, MountResult); @@ -1841,6 +1846,7 @@ const UnitVTable mount_vtable = { .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" @@ -1848,9 +1854,6 @@ const UnitVTable mount_vtable = { "Install\0", .private_section = "Mount", - .no_alias = true, - .no_instances = true, - .init = mount_init, .load = mount_load, .done = mount_done, @@ -1877,7 +1880,8 @@ const UnitVTable mount_vtable = { .reset_failed = mount_reset_failed, - .bus_interface = "org.freedesktop.systemd1.Mount", + .control_pid = mount_control_pid, + .bus_vtable = bus_mount_vtable, .bus_set_property = bus_mount_set_property, .bus_commit_properties = bus_mount_commit_properties, @@ -1897,7 +1901,6 @@ const UnitVTable mount_vtable = { .finished_start_job = { [JOB_DONE] = "Mounted %s.", [JOB_FAILED] = "Failed to mount %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", [JOB_TIMEOUT] = "Timed out mounting %s.", }, .finished_stop_job = { diff --git a/src/core/mount.h b/src/core/mount.h index 280ea0d638..9f7326ba6a 100644 --- a/src/core/mount.h +++ b/src/core/mount.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -24,25 +22,7 @@ typedef struct Mount Mount; #include "kill.h" -#include "execute.h" - -typedef enum MountState { - MOUNT_DEAD, - MOUNT_MOUNTING, /* /usr/bin/mount is running, but the mount is not done yet. */ - MOUNT_MOUNTING_DONE, /* /usr/bin/mount is running, and the mount is done. */ - MOUNT_MOUNTED, - MOUNT_REMOUNTING, - MOUNT_UNMOUNTING, - MOUNT_MOUNTING_SIGTERM, - MOUNT_MOUNTING_SIGKILL, - MOUNT_REMOUNTING_SIGTERM, - MOUNT_REMOUNTING_SIGKILL, - MOUNT_UNMOUNTING_SIGTERM, - MOUNT_UNMOUNTING_SIGKILL, - MOUNT_FAILED, - _MOUNT_STATE_MAX, - _MOUNT_STATE_INVALID = -1 -} MountState; +#include "dynamic-user.h" typedef enum MountExecCommand { MOUNT_EXEC_MOUNT, @@ -59,6 +39,7 @@ typedef enum MountResult { MOUNT_FAILURE_EXIT_CODE, MOUNT_FAILURE_SIGNAL, MOUNT_FAILURE_CORE_DUMP, + MOUNT_FAILURE_START_LIMIT_HIT, _MOUNT_RESULT_MAX, _MOUNT_RESULT_INVALID = -1 } MountResult; @@ -90,6 +71,9 @@ struct Mount { bool sloppy_options; + bool lazy_unmount; + bool force_unmount; + MountResult result; MountResult reload_result; @@ -104,6 +88,7 @@ struct Mount { CGroupContext cgroup_context; ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; MountState state, deserialized_state; @@ -120,9 +105,6 @@ extern const UnitVTable mount_vtable; void mount_fd_event(Manager *m, int events); -const char* mount_state_to_string(MountState i) _const_; -MountState mount_state_from_string(const char *s) _pure_; - const char* mount_exec_command_to_string(MountExecCommand i) _const_; MountExecCommand mount_exec_command_from_string(const char *s) _pure_; diff --git a/src/core/namespace.c b/src/core/namespace.c index 045321e1d4..43a2f4ba6e 100644 --- a/src/core/namespace.c +++ b/src/core/namespace.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,23 +18,34 @@ ***/ #include <errno.h> -#include <sys/mount.h> -#include <string.h> +#include <sched.h> #include <stdio.h> -#include <unistd.h> +#include <string.h> +#include <sys/mount.h> #include <sys/stat.h> -#include <sched.h> +#include <unistd.h> #include <linux/fs.h> -#include "strv.h" -#include "util.h" -#include "path-util.h" -#include "missing.h" -#include "loopback-setup.h" +#include "alloc-util.h" #include "dev-setup.h" -#include "selinux-util.h" -#include "namespace.h" +#include "fd-util.h" +#include "fs-util.h" +#include "loopback-setup.h" +#include "missing.h" #include "mkdir.h" +#include "mount-util.h" +#include "namespace.h" +#include "path-util.h" +#include "selinux-util.h" +#include "socket-util.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "umask-util.h" +#include "user-util.h" +#include "util.h" + +#define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) typedef enum MountMode { /* This is ordered by priority! */ @@ -45,62 +54,230 @@ typedef enum MountMode { PRIVATE_TMP, PRIVATE_VAR_TMP, PRIVATE_DEV, - PRIVATE_BUS_ENDPOINT, - READWRITE + READWRITE, } MountMode; typedef struct BindMount { - const char *path; + const char *path; /* stack memory, doesn't need to be freed explicitly */ + char *chased; /* malloc()ed memory, needs to be freed */ MountMode mode; - bool done; - bool ignore; + 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 */ +}; + +/* + * 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; - (*p)->ignore = false; - (*p)->done = false; - - if ((mode == INACCESSIBLE || mode == READONLY || mode == READWRITE) && (*i)[0] == '-') { - (*p)->ignore = true; + if (IN_SET(mode, INACCESSIBLE, READONLY, READWRITE) && startswith(*i, "-")) { (*i)++; + ignore = true; } if (!path_is_absolute(*i)) return -EINVAL; - (*p)->path = *i; - (*p)->mode = mode; + set_bind_mount(p, *i, mode, ignore); (*p)++; } return 0; } -static int mount_path_compare(const void *a, const void *b) { - const BindMount *p = a, *q = b; - int d; +static int append_target_mounts(BindMount **p, const char *root_directory, const TargetMount *mounts, const size_t size) { + unsigned i; - d = path_compare(p->path, q->path); + 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_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; +} - if (d == 0) { - /* If the paths are equal, check the mode */ - if (p->mode < q->mode) - return -1; +static int append_protect_system(BindMount **p, const char *root_directory, ProtectSystem protect_system) { + int r = 0; - if (p->mode > q->mode) - return 1; + 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 */ - return d; + 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) { @@ -109,16 +286,110 @@ static void drop_duplicates(BindMount *m, unsigned *n) { 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 */ - if (previous && path_equal(f->path, previous->path)) + /* 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++; } @@ -148,7 +419,7 @@ static int mount_dev(BindMount *m) { dev = strjoina(temporary_mount, "/dev"); (void) mkdir(dev, 0755); - if (mount("tmpfs", dev, "tmpfs", MS_NOSUID|MS_STRICTATIME, "mode=755") < 0) { + if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { r = -errno; goto fail; } @@ -232,6 +503,8 @@ static int mount_dev(BindMount *m) { */ (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; @@ -262,79 +535,6 @@ fail: return r; } -static int mount_kdbus(BindMount *m) { - - char temporary_mount[] = "/tmp/kdbus-dev-XXXXXX"; - _cleanup_free_ char *basepath = NULL; - _cleanup_umask_ mode_t u; - char *busnode = NULL, *root; - struct stat st; - int r; - - assert(m); - - u = umask(0000); - - if (!mkdtemp(temporary_mount)) - return log_error_errno(errno, "Failed create temp dir: %m"); - - root = strjoina(temporary_mount, "/kdbus"); - (void) mkdir(root, 0755); - if (mount("tmpfs", root, "tmpfs", MS_NOSUID|MS_STRICTATIME, "mode=777") < 0) { - r = -errno; - goto fail; - } - - /* create a new /dev/null dev node copy so we have some fodder to - * bind-mount the custom endpoint over. */ - if (stat("/dev/null", &st) < 0) { - log_error_errno(errno, "Failed to stat /dev/null: %m"); - r = -errno; - goto fail; - } - - busnode = strjoina(root, "/bus"); - if (mknod(busnode, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { - log_error_errno(errno, "mknod() for %s failed: %m", busnode); - r = -errno; - goto fail; - } - - r = mount(m->path, busnode, NULL, MS_BIND, NULL); - if (r < 0) { - log_error_errno(errno, "bind mount of %s failed: %m", m->path); - r = -errno; - goto fail; - } - - basepath = dirname_malloc(m->path); - if (!basepath) { - r = -ENOMEM; - goto fail; - } - - if (mount(root, basepath, NULL, MS_MOVE, NULL) < 0) { - log_error_errno(errno, "bind mount of %s failed: %m", basepath); - r = -errno; - goto fail; - } - - rmdir(temporary_mount); - return 0; - -fail: - if (busnode) { - umount(busnode); - unlink(busnode); - } - - umount(root); - rmdir(root); - rmdir(temporary_mount); - - return r; -} - static int apply_mount( BindMount *m, const char *tmp_dir, @@ -345,23 +545,41 @@ static int apply_mount( assert(m); + log_debug("Applying namespace mount on %s", m->path); + switch (m->mode) { - case INACCESSIBLE: + case INACCESSIBLE: { + struct stat target; /* First, get rid of everything that is below if there * is anything... Then, overmount it with an - * inaccessible directory. */ - umount_recursive(m->path, 0); + * inaccessible path. */ + (void) umount_recursive(m->path, 0); - what = "/run/systemd/inaccessible"; + 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: - /* Nothing to mount here, we just later toggle the - * MS_RDONLY bit for the mount point */ - return 0; + + 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; @@ -374,85 +592,155 @@ static int apply_mount( case PRIVATE_DEV: return mount_dev(m); - case PRIVATE_BUS_ENDPOINT: - return mount_kdbus(m); - default: assert_not_reached("Unknown mode"); } assert(what); - r = mount(what, m->path, NULL, MS_BIND|MS_REC, NULL); - if (r >= 0) - log_debug("Successfully mounted %s to %s", what, m->path); - else if (m->ignore && errno == ENOENT) + 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 make_read_only(BindMount *m) { +static int chase_all_symlinks(const char *root_directory, BindMount *m, unsigned *n) { + BindMount *f, *t; int r; assert(m); + assert(n); - if (IN_SET(m->mode, INACCESSIBLE, READONLY)) - r = bind_remount_recursive(m->path, true); - else if (IN_SET(m->mode, READWRITE, PRIVATE_TMP, PRIVATE_VAR_TMP, PRIVATE_DEV)) - r = bind_remount_recursive(m->path, false); - else - r = 0; + /* 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. */ - if (m->ignore && r == -ENOENT) - return 0; + for (f = m, t = m; f < m+*n; f++) { - return r; + 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( + char** read_write_paths, + char** read_only_paths, + char** inaccessible_paths, + const char* tmp_dir, + const char* var_tmp_dir, + bool private_dev, + bool protect_sysctl, + bool protect_cgroups, + 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) + + private_dev + + (protect_sysctl ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + + (protect_cgroups ? 1 : 0) + + protect_home_cnt + protect_system_cnt; } int setup_namespace( const char* root_directory, - char** read_write_dirs, - char** read_only_dirs, - char** inaccessible_dirs, + char** read_write_paths, + char** read_only_paths, + char** inaccessible_paths, const char* tmp_dir, const char* var_tmp_dir, - const char* bus_endpoint_path, bool private_dev, + bool protect_sysctl, + bool protect_cgroups, 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; - if (unshare(CLONE_NEWNS) < 0) - return -errno; + n = namespace_calculate_mounts(read_write_paths, + read_only_paths, + inaccessible_paths, + tmp_dir, var_tmp_dir, + private_dev, protect_sysctl, + protect_cgroups, protect_home, + protect_system); - n = !!tmp_dir + !!var_tmp_dir + !!bus_endpoint_path + - strv_length(read_write_dirs) + - strv_length(read_only_dirs) + - strv_length(inaccessible_dirs) + - private_dev + - (protect_home != PROTECT_HOME_NO ? 3 : 0) + - (protect_system != PROTECT_SYSTEM_NO ? 2 : 0) + - (protect_system == PROTECT_SYSTEM_FULL ? 1 : 0); + /* 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_dirs, READWRITE); + r = append_mounts(&m, read_write_paths, READWRITE); if (r < 0) return r; - r = append_mounts(&m, read_only_dirs, READONLY); + r = append_mounts(&m, read_only_paths, READONLY); if (r < 0) return r; - r = append_mounts(&m, inaccessible_dirs, INACCESSIBLE); + r = append_mounts(&m, inaccessible_paths, INACCESSIBLE); if (r < 0) return r; @@ -474,101 +762,112 @@ int setup_namespace( m++; } - if (bus_endpoint_path) { - m->path = prefix_roota(root_directory, bus_endpoint_path); - m->mode = PRIVATE_BUS_ENDPOINT; - m++; - } - - if (protect_home != PROTECT_HOME_NO) { - const char *home_dir, *run_user_dir, *root_dir; + if (protect_sysctl) + append_protect_kernel_tunables(&m, root_directory); - home_dir = prefix_roota(root_directory, "/home"); - home_dir = strjoina("-", home_dir); - run_user_dir = prefix_roota(root_directory, "/run/user"); - run_user_dir = strjoina("-", run_user_dir); - root_dir = prefix_roota(root_directory, "/root"); - root_dir = strjoina("-", root_dir); - - r = append_mounts(&m, STRV_MAKE(home_dir, run_user_dir, root_dir), - protect_home == PROTECT_HOME_READ_ONLY ? READONLY : INACCESSIBLE); - if (r < 0) - return r; + if (protect_cgroups) { + m->path = prefix_roota(root_directory, "/sys/fs/cgroup"); + m->mode = READONLY; + m++; } - if (protect_system != PROTECT_SYSTEM_NO) { - const char *usr_dir, *boot_dir, *etc_dir; - - usr_dir = prefix_roota(root_directory, "/usr"); - boot_dir = prefix_roota(root_directory, "/boot"); - boot_dir = strjoina("-", boot_dir); - etc_dir = prefix_roota(root_directory, "/etc"); + r = append_protect_home(&m, root_directory, protect_home); + if (r < 0) + return r; - r = append_mounts(&m, protect_system == PROTECT_SYSTEM_FULL - ? STRV_MAKE(usr_dir, boot_dir, etc_dir) - : STRV_MAKE(usr_dir, boot_dir), READONLY); - 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 (n > 0 || root_directory) { + 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) - return -errno; + if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { + r = -errno; + goto finish; + } } if (root_directory) { - /* Turn directory into bind mount */ - if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0) - return -errno; + /* 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 fail; + 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); + r = make_read_only(m, blacklist); if (r < 0) - goto fail; + 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); - - /* at this point, we cannot rollback */ if (r < 0) - return r; + 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) { - /* at this point, we cannot rollback */ - return -errno; + r = -errno; + goto finish; } - return 0; + r = 0; -fail: - if (n > 0) { - for (m = mounts; m < mounts + n; ++m) - if (m->done) - (void) umount2(m->path, MNT_DETACH); - } +finish: + for (m = mounts; m < mounts + n; m++) + free(m->chased); return r; } @@ -645,16 +944,7 @@ int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { int setup_netns(int netns_storage_socket[2]) { _cleanup_close_ int netns = -1; - union { - struct cmsghdr cmsghdr; - uint8_t buf[CMSG_SPACE(sizeof(int))]; - } control = {}; - struct msghdr mh = { - .msg_control = &control, - .msg_controllen = sizeof(control), - }; - struct cmsghdr *cmsg; - int r; + int r, q; assert(netns_storage_socket); assert(netns_storage_socket[0] >= 0); @@ -671,12 +961,8 @@ int setup_netns(int netns_storage_socket[2]) { if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) return -errno; - if (recvmsg(netns_storage_socket[0], &mh, MSG_DONTWAIT|MSG_CMSG_CLOEXEC) < 0) { - if (errno != EAGAIN) { - r = -errno; - goto fail; - } - + 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) { @@ -693,15 +979,13 @@ int setup_netns(int netns_storage_socket[2]) { } r = 1; - } else { - /* Yay, found something, so let's join the namespace */ - CMSG_FOREACH(cmsg, &mh) - if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SCM_RIGHTS) { - assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int))); - netns = *(int*) CMSG_DATA(cmsg); - } + } 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; @@ -710,21 +994,14 @@ int setup_netns(int netns_storage_socket[2]) { r = 0; } - 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), &netns, sizeof(int)); - mh.msg_controllen = cmsg->cmsg_len; - - if (sendmsg(netns_storage_socket[1], &mh, MSG_DONTWAIT|MSG_NOSIGNAL) < 0) { - r = -errno; + q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); + if (q < 0) { + r = q; goto fail; } fail: - lockf(netns_storage_socket[0], F_ULOCK, 0); - + (void) lockf(netns_storage_socket[0], F_ULOCK, 0); return r; } @@ -740,6 +1017,7 @@ 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/core/namespace.h b/src/core/namespace.h index 00ab22bf2e..6505bcc499 100644 --- a/src/core/namespace.h +++ b/src/core/namespace.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -37,18 +35,20 @@ typedef enum ProtectSystem { PROTECT_SYSTEM_NO, PROTECT_SYSTEM_YES, PROTECT_SYSTEM_FULL, + PROTECT_SYSTEM_STRICT, _PROTECT_SYSTEM_MAX, _PROTECT_SYSTEM_INVALID = -1 } ProtectSystem; int setup_namespace(const char *chroot, - char **read_write_dirs, - char **read_only_dirs, - char **inaccessible_dirs, + char **read_write_paths, + char **read_only_paths, + char **inaccessible_paths, const char *tmp_dir, const char *var_tmp_dir, - const char *endpoint_path, bool private_dev, + bool protect_sysctl, + bool protect_cgroups, ProtectHome protect_home, ProtectSystem protect_system, unsigned long mount_flags); diff --git a/src/core/org.freedesktop.systemd1.conf b/src/core/org.freedesktop.systemd1.conf index 6a7a37ee92..6caa15b0b8 100644 --- a/src/core/org.freedesktop.systemd1.conf +++ b/src/core/org.freedesktop.systemd1.conf @@ -54,6 +54,10 @@ <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="GetUnitByInvocationID"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="LoadUnit"/> <allow send_destination="org.freedesktop.systemd1" @@ -70,14 +74,26 @@ <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="ListUnitsByPatterns"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="ListUnitFiles"/> <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="ListUnitFilesByPatterns"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="GetUnitFileState"/> <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="GetUnitProcesses"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="ListJobs"/> <allow send_destination="org.freedesktop.systemd1" @@ -96,6 +112,14 @@ send_interface="org.freedesktop.systemd1.Manager" send_member="GetDefaultTarget"/> + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="LookupDynamicUserByName"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="LookupDynamicUserByUID"/> + <!-- Managed via polkit or other criteria --> <allow send_destination="org.freedesktop.systemd1" @@ -144,6 +168,10 @@ <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="ListUnitsByNames"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="StartTransientUnit"/> <allow send_destination="org.freedesktop.systemd1" @@ -160,6 +188,14 @@ <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="RefUnit"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" + send_member="UnrefUnit"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="EnableUnitFiles"/> <allow send_destination="org.freedesktop.systemd1" @@ -176,6 +212,10 @@ <allow send_destination="org.freedesktop.systemd1" send_interface="org.freedesktop.systemd1.Manager" + send_member="RevertUnitFiles"/> + + <allow send_destination="org.freedesktop.systemd1" + send_interface="org.freedesktop.systemd1.Manager" send_member="PresetUnitFiles"/> <allow send_destination="org.freedesktop.systemd1" diff --git a/src/core/path.c b/src/core/path.c index 20995d920c..83f794be89 100644 --- a/src/core/path.c +++ b/src/core/path.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,20 +17,26 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/inotify.h> -#include <sys/epoll.h> #include <errno.h> +#include <sys/epoll.h> +#include <sys/inotify.h> #include <unistd.h> -#include "unit.h" -#include "unit-name.h" -#include "path.h" -#include "mkdir.h" +#include "bus-error.h" +#include "bus-util.h" #include "dbus-path.h" -#include "special.h" +#include "fd-util.h" +#include "fs-util.h" +#include "glob-util.h" #include "macro.h" -#include "bus-util.h" -#include "bus-error.h" +#include "mkdir.h" +#include "path.h" +#include "special.h" +#include "stat-util.h" +#include "string-table.h" +#include "string-util.h" +#include "unit-name.h" +#include "unit.h" static const UnitActiveState state_translation_table[_PATH_STATE_MAX] = { [PATH_DEAD] = UNIT_INACTIVE, @@ -106,16 +110,14 @@ int path_spec_watch(PathSpec *s, sd_event_io_handler_t handler) { } else { exists = true; - /* Path exists, we don't need to watch parent - too closely. */ + /* Path exists, we don't need to watch parent too closely. */ if (oldslash) { char *cut2 = oldslash + (oldslash == s->path); char tmp2 = *cut2; *cut2 = '\0'; - inotify_add_watch(s->inotify_fd, s->path, IN_MOVE_SELF); - /* Error is ignored, the worst can happen is - we get spurious events. */ + (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; } @@ -309,20 +311,20 @@ static int path_add_default_dependencies(Path *p) { assert(p); - r = unit_add_dependency_by_name(UNIT(p), UNIT_BEFORE, - SPECIAL_PATHS_TARGET, NULL, true); + 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 (UNIT(p)->manager->running_as == MANAGER_SYSTEM) { - r = unit_add_two_dependencies_by_name(UNIT(p), UNIT_AFTER, UNIT_REQUIRES, - SPECIAL_SYSINIT_TARGET, NULL, true); + 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); + return unit_add_two_dependencies_by_name(UNIT(p), UNIT_BEFORE, UNIT_CONFLICTS, SPECIAL_SHUTDOWN_TARGET, NULL, true); } static int path_load(Unit *u) { @@ -354,11 +356,9 @@ static int path_load(Unit *u) { if (r < 0) return r; - if (UNIT(p)->default_dependencies) { - r = path_add_default_dependencies(p); - if (r < 0) - return r; - } + r = path_add_default_dependencies(p); + if (r < 0) + return r; } return path_verify(p); @@ -454,14 +454,15 @@ static int path_coldplug(Unit *u) { static void path_enter_dead(Path *p, PathResult f) { assert(p); - if (f != PATH_SUCCESS) + 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_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + Unit *trigger; int r; assert(p); @@ -470,8 +471,14 @@ static void path_enter_running(Path *p) { if (unit_stop_pending(UNIT(p))) return; - r = manager_add_job(UNIT(p)->manager, JOB_START, UNIT_TRIGGER(UNIT(p)), - JOB_REPLACE, true, &error, NULL); + 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; @@ -552,12 +559,27 @@ static void path_mkdir(Path *p) { 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); - if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + 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); @@ -715,15 +737,6 @@ static void path_reset_failed(Unit *u) { p->result = PATH_SUCCESS; } -static const char* const path_state_table[_PATH_STATE_MAX] = { - [PATH_DEAD] = "dead", - [PATH_WAITING] = "waiting", - [PATH_RUNNING] = "running", - [PATH_FAILED] = "failed" -}; - -DEFINE_STRING_TABLE_LOOKUP(path_state, PathState); - static const char* const path_type_table[_PATH_TYPE_MAX] = { [PATH_EXISTS] = "PathExists", [PATH_EXISTS_GLOB] = "PathExistsGlob", @@ -737,6 +750,7 @@ 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); @@ -770,6 +784,5 @@ const UnitVTable path_vtable = { .reset_failed = path_reset_failed, - .bus_interface = "org.freedesktop.systemd1.Path", .bus_vtable = bus_path_vtable }; diff --git a/src/core/path.h b/src/core/path.h index dec39333e4..4230c8fb99 100644 --- a/src/core/path.h +++ b/src/core/path.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -26,15 +24,6 @@ typedef struct PathSpec PathSpec; #include "unit.h" -typedef enum PathState { - PATH_DEAD, - PATH_WAITING, - PATH_RUNNING, - PATH_FAILED, - _PATH_STATE_MAX, - _PATH_STATE_INVALID = -1 -} PathState; - typedef enum PathType { PATH_EXISTS, PATH_EXISTS_GLOB, @@ -73,6 +62,7 @@ static inline bool path_spec_owns_inotify_fd(PathSpec *s, int fd) { typedef enum PathResult { PATH_SUCCESS, PATH_FAILURE_RESOURCES, + PATH_FAILURE_START_LIMIT_HIT, _PATH_RESULT_MAX, _PATH_RESULT_INVALID = -1 } PathResult; @@ -96,9 +86,6 @@ void path_free_specs(Path *p); extern const UnitVTable path_vtable; -const char* path_state_to_string(PathState i) _const_; -PathState path_state_from_string(const char *s) _pure_; - const char* path_type_to_string(PathType i) _const_; PathType path_type_from_string(const char *s) _pure_; diff --git a/src/core/scope.c b/src/core/scope.c index ab1769b46b..e7583f6d89 100644 --- a/src/core/scope.c +++ b/src/core/scope.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -22,13 +20,17 @@ #include <errno.h> #include <unistd.h> -#include "unit.h" -#include "scope.h" -#include "log.h" +#include "alloc-util.h" #include "dbus-scope.h" +#include "load-dropin.h" +#include "log.h" +#include "scope.h" #include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" #include "unit-name.h" -#include "load-dropin.h" +#include "unit.h" static const UnitActiveState state_translation_table[_SCOPE_STATE_MAX] = { [SCOPE_DEAD] = UNIT_INACTIVE, @@ -48,9 +50,7 @@ static void scope_init(Unit *u) { assert(u->load_state == UNIT_STUB); s->timeout_stop_usec = u->manager->default_timeout_stop_usec; - - UNIT(s)->ignore_on_isolate = true; - UNIT(s)->ignore_on_snapshot = true; + u->ignore_on_isolate = true; } static void scope_done(Unit *u) { @@ -63,29 +63,27 @@ static void scope_done(Unit *u) { s->timer_event_source = sd_event_source_unref(s->timer_event_source); } -static int scope_arm_timer(Scope *s) { +static int scope_arm_timer(Scope *s, usec_t usec) { int r; assert(s); - if (s->timeout_stop_usec <= 0) { - s->timer_event_source = sd_event_source_unref(s->timer_event_source); - return 0; - } - if (s->timer_event_source) { - r = sd_event_source_set_time(s->timer_event_source, now(CLOCK_MONOTONIC) + s->timeout_stop_usec); + 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, - now(CLOCK_MONOTONIC) + s->timeout_stop_usec, 0, + usec, 0, scope_dispatch_timer, s); if (r < 0) return r; @@ -119,6 +117,9 @@ static int scope_add_default_dependencies(Scope *s) { 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), @@ -136,7 +137,9 @@ static int scope_verify(Scope *s) { if (UNIT(s)->load_state != UNIT_LOADED) return 0; - if (set_isempty(UNIT(s)->pids) && UNIT(s)->manager->n_reloading <= 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; } @@ -151,24 +154,23 @@ static int scope_load(Unit *u) { assert(s); assert(u->load_state == UNIT_STUB); - if (!u->transient && UNIT(s)->manager->n_reloading <= 0) + if (!u->transient && !MANAGER_IS_RELOADING(u->manager)) + /* Refuse to load non-transient scope units, but allow them while reloading. */ return -ENOENT; - u->load_state = UNIT_LOADED; - - r = unit_load_dropin(u); + r = unit_load_fragment_and_dropin_optional(u); if (r < 0) return r; - r = unit_patch_contexts(u); - if (r < 0) - return r; + if (u->load_state == UNIT_LOADED) { + r = unit_patch_contexts(u); + if (r < 0) + return r; - r = unit_add_default_slice(u, &s->cgroup_context); - if (r < 0) - return r; + r = unit_set_default_slice(u); + if (r < 0) + return r; - if (u->default_dependencies) { r = scope_add_default_dependencies(s); if (r < 0) return r; @@ -184,20 +186,19 @@ static int scope_coldplug(Unit *u) { assert(s); assert(s->state == SCOPE_DEAD); - if (s->deserialized_state != s->state) { - - if (IN_SET(s->deserialized_state, SCOPE_STOP_SIGKILL, SCOPE_STOP_SIGTERM)) { - r = scope_arm_timer(s); - if (r < 0) - return r; - } - - if (!IN_SET(s->deserialized_state, SCOPE_DEAD, SCOPE_FAILED)) - unit_watch_all_pids(UNIT(s)); + if (s->deserialized_state == s->state) + return 0; - scope_set_state(s, s->deserialized_state); + 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; } @@ -220,7 +221,7 @@ static void scope_dump(Unit *u, FILE *f, const char *prefix) { static void scope_enter_dead(Scope *s, ScopeResult f) { assert(s); - if (f != SCOPE_SUCCESS) + if (s->result == SCOPE_SUCCESS) s->result = f; scope_set_state(s, s->result != SCOPE_SUCCESS ? SCOPE_FAILED : SCOPE_DEAD); @@ -232,14 +233,14 @@ static void scope_enter_signal(Scope *s, ScopeState state, ScopeResult f) { assert(s); - if (f != SCOPE_SUCCESS) + 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 beserk mode */ + * SIGTERM berserk mode */ if (state == SCOPE_STOP_SIGTERM) skip_signal = bus_scope_send_request_stop(s) > 0; @@ -247,7 +248,9 @@ static void scope_enter_signal(Scope *s, ScopeState state, ScopeResult f) { r = unit_kill_context( UNIT(s), &s->kill_context, - state != SCOPE_STOP_SIGTERM ? KILL_KILL : KILL_TERMINATE, + state != SCOPE_STOP_SIGTERM ? KILL_KILL : + s->was_abandoned ? KILL_TERMINATE_AND_LOG : + KILL_TERMINATE, -1, -1, false); if (r < 0) goto fail; @@ -255,7 +258,7 @@ static void scope_enter_signal(Scope *s, ScopeState state, ScopeResult f) { r = 1; if (r > 0) { - r = scope_arm_timer(s); + r = scope_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_stop_usec)); if (r < 0) goto fail; @@ -279,6 +282,9 @@ static int scope_start(Unit *u) { assert(s); + if (unit_has_name(u, SPECIAL_INIT_SCOPE)) + return -EPERM; + if (s->state == SCOPE_FAILED) return -EPERM; @@ -289,9 +295,13 @@ static int scope_start(Unit *u) { assert(s->state == SCOPE_DEAD); - if (!u->transient && UNIT(s)->manager->n_reloading <= 0) + 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); @@ -339,17 +349,21 @@ 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, uint64_t *timeout) { +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, timeout); + 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; } @@ -361,6 +375,7 @@ static int scope_serialize(Unit *u, FILE *f, FDSet *fds) { 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; } @@ -381,6 +396,14 @@ static int scope_deserialize_item(Unit *u, const char *key, const char *value, F 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); @@ -393,15 +416,10 @@ static bool scope_check_gc(Unit *u) { /* Never clean up scopes that still have a process around, * even if the scope is formally dead. */ - if (u->cgroup_path) { - int r; - - r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, true); - if (r <= 0) - return true; - } + if (!u->cgroup_path) + return false; - return false; + return cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path) <= 0; } static void scope_notify_cgroup_empty_event(Unit *u) { @@ -425,8 +443,9 @@ static void scope_sigchld_event(Unit *u, pid_t pid, int code, int status) { 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 */ - if (set_isempty(u->pids)) + /* 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); } @@ -464,11 +483,14 @@ static int scope_dispatch_timer(sd_event_source *source, usec_t usec, void *user 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; - free(s->controller); - s->controller = NULL; + 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 @@ -499,16 +521,51 @@ _pure_ static const char *scope_sub_state_to_string(Unit *u) { return scope_state_to_string(SCOPE(u)->state); } -static const char* const scope_state_table[_SCOPE_STATE_MAX] = { - [SCOPE_DEAD] = "dead", - [SCOPE_RUNNING] = "running", - [SCOPE_ABANDONED] = "abandoned", - [SCOPE_STOP_SIGTERM] = "stop-sigterm", - [SCOPE_STOP_SIGKILL] = "stop-sigkill", - [SCOPE_FAILED] = "failed", -}; +static void scope_enumerate(Manager *m) { + Unit *u; + int r; + + assert(m); -DEFINE_STRING_TABLE_LOOKUP(scope_state, ScopeState); + /* 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) { + u = unit_new(m, sizeof(Scope)); + if (!u) { + log_oom(); + return; + } + + r = unit_add_name(u, SPECIAL_INIT_SCOPE); + if (r < 0) { + unit_free(u); + log_error_errno(r, "Failed to add init.scope name"); + return; + } + } + + u->transient = true; + u->default_dependencies = false; + u->no_gc = true; + u->ignore_on_isolate = true; + u->refuse_manual_start = true; + u->refuse_manual_stop = true; + SCOPE(u)->deserialized_state = SCOPE_RUNNING; + 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)"); + + 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", @@ -529,8 +586,7 @@ const UnitVTable scope_vtable = { "Install\0", .private_section = "Scope", - .no_alias = true, - .no_instances = true, + .can_transient = true, .init = scope_init, .load = scope_load, @@ -561,10 +617,9 @@ const UnitVTable scope_vtable = { .notify_cgroup_empty = scope_notify_cgroup_empty_event, - .bus_interface = "org.freedesktop.systemd1.Scope", .bus_vtable = bus_scope_vtable, .bus_set_property = bus_scope_set_property, .bus_commit_properties = bus_scope_commit_properties, - .can_transient = true + .enumerate = scope_enumerate, }; diff --git a/src/core/scope.h b/src/core/scope.h index 4452fe2c94..eaf8e8b447 100644 --- a/src/core/scope.h +++ b/src/core/scope.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,18 +21,9 @@ typedef struct Scope Scope; +#include "cgroup.h" #include "kill.h" - -typedef enum ScopeState { - SCOPE_DEAD, - SCOPE_RUNNING, - SCOPE_ABANDONED, - SCOPE_STOP_SIGTERM, - SCOPE_STOP_SIGKILL, - SCOPE_FAILED, - _SCOPE_STATE_MAX, - _SCOPE_STATE_INVALID = -1 -} ScopeState; +#include "unit.h" typedef enum ScopeResult { SCOPE_SUCCESS, @@ -56,6 +45,7 @@ struct Scope { usec_t timeout_stop_usec; char *controller; + bool was_abandoned; sd_event_source *timer_event_source; }; @@ -64,8 +54,5 @@ extern const UnitVTable scope_vtable; int scope_abandon(Scope *s); -const char* scope_state_to_string(ScopeState i) _const_; -ScopeState scope_state_from_string(const char *s) _pure_; - const char* scope_result_to_string(ScopeResult i) _const_; ScopeResult scope_result_from_string(const char *s) _pure_; diff --git a/src/core/selinux-access.c b/src/core/selinux-access.c index e9a9a020de..2b96a9551b 100644 --- a/src/core/selinux-access.c +++ b/src/core/selinux-access.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -23,21 +21,25 @@ #ifdef HAVE_SELINUX -#include <stdio.h> #include <errno.h> -#include <selinux/selinux.h> #include <selinux/avc.h> +#include <selinux/selinux.h> +#include <stdio.h> #ifdef HAVE_AUDIT #include <libaudit.h> #endif #include "sd-bus.h" + +#include "alloc-util.h" +#include "audit-fd.h" #include "bus-util.h" -#include "util.h" #include "log.h" +#include "path-util.h" #include "selinux-util.h" -#include "audit-fd.h" +#include "stdio-util.h" #include "strv.h" +#include "util.h" static bool initialized = false; @@ -108,6 +110,7 @@ static int callback_type_to_priority(int type) { */ _printf_(2, 3) static int log_callback(int type, const char *fmt, ...) { va_list ap; + const char *fmt2; #ifdef HAVE_AUDIT int fd; @@ -129,63 +132,47 @@ _printf_(2, 3) static int log_callback(int type, const char *fmt, ...) { } #endif + fmt2 = strjoina("selinux: ", fmt); + va_start(ap, fmt); - log_internalv(LOG_AUTH | callback_type_to_priority(type), - 0, __FILE__, __LINE__, __FUNCTION__, fmt, ap); + log_internalv(LOG_AUTH | callback_type_to_priority(type), 0, __FILE__, __LINE__, __FUNCTION__, fmt2, ap); va_end(ap); return 0; } -/* - Function must be called once to initialize the SELinux AVC environment. - Sets up callbacks. - If you want to cleanup memory you should need to call selinux_access_finish. -*/ -static int access_init(void) { - int r = 0; - - if (avc_open(NULL, 0)) - return log_error_errno(errno, "avc_open() failed: %m"); +static int access_init(sd_bus_error *error) { - selinux_set_callback(SELINUX_CB_AUDIT, (union selinux_callback) audit_callback); - selinux_set_callback(SELINUX_CB_LOG, (union selinux_callback) log_callback); - - if (security_getenforce() < 0){ - r = -errno; - avc_destroy(); - } - - return r; -} - -static int mac_selinux_access_init(sd_bus_error *error) { - int r; + if (!mac_selinux_use()) + return 0; if (initialized) - return 0; + return 1; - if (!mac_selinux_use()) - return 0; + if (avc_open(NULL, 0) != 0) { + int enforce, saved_errno = errno; - r = access_init(); - if (r < 0) - return sd_bus_error_set(error, SD_BUS_ERROR_ACCESS_DENIED, "Failed to initialize SELinux."); + enforce = security_getenforce(); + log_full_errno(enforce != 0 ? LOG_ERR : LOG_WARNING, saved_errno, "Failed to open the SELinux AVC: %m"); - initialized = true; - return 0; -} -#endif + /* 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; -void mac_selinux_access_free(void) { + /* 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)); + } -#ifdef HAVE_SELINUX - if (!initialized) - return; + selinux_set_callback(SELINUX_CB_AUDIT, (union selinux_callback) audit_callback); + selinux_set_callback(SELINUX_CB_LOG, (union selinux_callback) log_callback); - avc_destroy(); - initialized = false; -#endif + initialized = true; + return 1; } /* @@ -200,12 +187,11 @@ int mac_selinux_generic_access_check( const char *permission, sd_bus_error *error) { -#ifdef HAVE_SELINUX - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _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; - security_context_t fcon = NULL; + char *fcon = NULL; char **cmdline = NULL; int r = 0; @@ -213,11 +199,8 @@ int mac_selinux_generic_access_check( assert(permission); assert(error); - if (!mac_selinux_use()) - return 0; - - r = mac_selinux_access_init(error); - if (r < 0) + r = access_init(error); + if (r <= 0) return r; r = sd_bus_query_sender_creds( @@ -245,7 +228,7 @@ int mac_selinux_generic_access_check( if (path) { /* Get the file context of the unit file */ - r = getfilecon(path, &fcon); + 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; @@ -253,7 +236,7 @@ int mac_selinux_generic_access_check( tclass = "service"; } else { - r = getcon(&fcon); + 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; @@ -284,31 +267,17 @@ finish: } return r; -#else - return 0; -#endif } -int mac_selinux_unit_access_check_strv( - char **units, +#else + +int mac_selinux_generic_access_check( sd_bus_message *message, - Manager *m, + const char *path, const char *permission, sd_bus_error *error) { -#ifdef HAVE_SELINUX - char **i; - Unit *u; - int r; - - STRV_FOREACH(i, units) { - u = manager_get_unit(m, *i); - if (u) { - r = mac_selinux_unit_access_check(u, message, permission, error); - if (r < 0) - return r; - } - } -#endif return 0; } + +#endif diff --git a/src/core/selinux-access.h b/src/core/selinux-access.h index b5758e2e42..f46370d020 100644 --- a/src/core/selinux-access.h +++ b/src/core/selinux-access.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,15 +20,12 @@ ***/ #include "sd-bus.h" + #include "bus-util.h" #include "manager.h" -void mac_selinux_access_free(void); - int mac_selinux_generic_access_check(sd_bus_message *message, const char *path, const char *permission, sd_bus_error *error); -int mac_selinux_unit_access_check_strv(char **units, sd_bus_message *message, Manager *m, const char *permission, sd_bus_error *error); - #ifdef HAVE_SELINUX #define mac_selinux_access_check(message, permission, error) \ @@ -38,7 +33,7 @@ int mac_selinux_unit_access_check_strv(char **units, sd_bus_message *message, Ma #define mac_selinux_unit_access_check(unit, message, permission, error) \ ({ \ - Unit *_unit = (unit); \ + const Unit *_unit = (unit); \ mac_selinux_generic_access_check((message), _unit->source_path ?: _unit->fragment_path, (permission), (error)); \ }) diff --git a/src/core/selinux-setup.c b/src/core/selinux-setup.c index a4678500e6..527aa8add0 100644 --- a/src/core/selinux-setup.c +++ b/src/core/selinux-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,21 +17,23 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <unistd.h> -#include <stdio.h> #include <errno.h> +#include <stdio.h> +#include <unistd.h> #ifdef HAVE_SELINUX #include <selinux/selinux.h> #endif +#include "log.h" +#include "macro.h" #include "selinux-setup.h" #include "selinux-util.h" -#include "macro.h" +#include "string-util.h" #include "util.h" -#include "log.h" #ifdef HAVE_SELINUX +_printf_(2,3) static int null_log(int type, const char *fmt, ...) { return 0; } @@ -44,7 +44,7 @@ int mac_selinux_setup(bool *loaded_policy) { #ifdef HAVE_SELINUX int enforce = 0; usec_t before_load, after_load; - security_context_t con; + char *con; int r; union selinux_callback cb; bool initialized = false; @@ -77,24 +77,22 @@ int mac_selinux_setup(bool *loaded_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]; - char *label; mac_selinux_retest(); /* Transition to the new context */ r = mac_selinux_get_create_label_from_exe(SYSTEMD_BINARY_PATH, &label); - if (r < 0 || label == NULL) { + if (r < 0 || !label) { log_open(); log_error("Failed to compute init label, ignoring."); } else { - r = setcon(label); + r = setcon_raw(label); log_open(); if (r < 0) log_error("Failed to transition into init label '%s', ignoring.", label); - - mac_selinux_free(label); } after_load = now(CLOCK_MONOTONIC); diff --git a/src/core/selinux-setup.h b/src/core/selinux-setup.h index 9ac2276576..7b613249b0 100644 --- a/src/core/selinux-setup.h +++ b/src/core/selinux-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/service.c b/src/core/service.c index d72ff54daa..63045ede55 100644 --- a/src/core/service.c +++ b/src/core/service.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -23,31 +21,38 @@ #include <signal.h> #include <unistd.h> +#include "alloc-util.h" #include "async.h" -#include "manager.h" -#include "unit.h" -#include "service.h" -#include "load-fragment.h" -#include "load-dropin.h" -#include "log.h" -#include "strv.h" -#include "unit-name.h" -#include "unit-printf.h" +#include "bus-error.h" +#include "bus-kernel.h" +#include "bus-util.h" #include "dbus-service.h" -#include "special.h" -#include "exit-status.h" #include "def.h" -#include "path-util.h" -#include "util.h" -#include "utf8.h" #include "env-util.h" +#include "escape.h" +#include "exit-status.h" +#include "fd-util.h" #include "fileio.h" -#include "bus-error.h" -#include "bus-util.h" -#include "bus-kernel.h" #include "formats-util.h" +#include "fs-util.h" +#include "load-dropin.h" +#include "load-fragment.h" +#include "log.h" +#include "manager.h" +#include "parse-util.h" +#include "path-util.h" #include "process-util.h" +#include "service.h" #include "signal-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "unit.h" +#include "utf8.h" +#include "util.h" static const UnitActiveState state_translation_table[_SERVICE_STATE_MAX] = { [SERVICE_DEAD] = UNIT_INACTIVE, @@ -105,13 +110,12 @@ static void service_init(Unit *u) { 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->bus_endpoint_fd = -1; + s->stdin_fd = s->stdout_fd = s->stderr_fd = -1; s->guess_main_pid = true; - RATELIMIT_INIT(s->start_limit, u->manager->default_start_limit_interval, u->manager->default_start_limit_burst); - s->control_command_id = _SERVICE_EXEC_COMMAND_INVALID; } @@ -142,8 +146,7 @@ static void service_unwatch_pid_file(Service *s) { 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); - free(s->pid_file_pathspec); - s->pid_file_pathspec = NULL; + s->pid_file_pathspec = mfree(s->pid_file_pathspec); } static int service_set_main_pid(Service *s, pid_t pid) { @@ -168,7 +171,7 @@ static int service_set_main_pid(Service *s, pid_t pid) { s->main_pid = pid; s->main_pid_known = true; - if (get_parent_of_pid(pid, &ppid) >= 0 && ppid != getpid()) { + 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 @@ -177,20 +180,17 @@ static int service_set_main_pid(Service *s, pid_t pid) { return 0; } -static void service_close_socket_fd(Service *s) { +void service_close_socket_fd(Service *s) { assert(s); - s->socket_fd = asynchronous_close(s->socket_fd); -} + /* Undo the effect of service_set_socket_fd(). */ -static void service_connection_unref(Service *s) { - assert(s); - - if (!UNIT_ISSET(s->accept_socket)) - return; + s->socket_fd = asynchronous_close(s->socket_fd); - socket_connection_unref(SOCKET(UNIT_DEREF(s->accept_socket))); - unit_ref_unset(&s->accept_socket); + 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) { @@ -200,16 +200,27 @@ static void service_stop_watchdog(Service *s) { 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); - if (s->watchdog_usec <= 0) + 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, s->watchdog_timestamp.monotonic + s->watchdog_usec); + 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; @@ -221,7 +232,7 @@ static void service_start_watchdog(Service *s) { UNIT(s)->manager->event, &s->watchdog_event_source, CLOCK_MONOTONIC, - s->watchdog_timestamp.monotonic + s->watchdog_usec, 0, + 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"); @@ -246,6 +257,17 @@ static void service_reset_watchdog(Service *s) { 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) @@ -262,6 +284,7 @@ static void service_fd_store_unlink(ServiceFDStore *fs) { sd_event_source_unref(fs->event_source); } + free(fs->fdname); safe_close(fs->fd); free(fs); } @@ -271,11 +294,15 @@ static void service_release_resources(Unit *u) { assert(s); - if (!s->fd_store) + if (!s->fd_store && s->stdin_fd < 0 && s->stdout_fd < 0 && s->stderr_fd < 0) return; log_unit_debug(u, "Releasing all 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); + while (s->fd_store) service_fd_store_unlink(s->fd_store); @@ -287,20 +314,16 @@ static void service_done(Unit *u) { assert(s); - free(s->pid_file); - s->pid_file = NULL; - - free(s->status_text); - s->status_text = NULL; - - free(s->reboot_arg); - s->reboot_arg = NULL; + 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); @@ -313,13 +336,13 @@ static void service_done(Unit *u) { if (s->bus_name) { unit_unwatch_bus_name(u, s->bus_name); - free(s->bus_name); - s->bus_name = NULL; + s->bus_name = mfree(s->bus_name); } - s->bus_endpoint_fd = safe_close(s->bus_endpoint_fd); + s->bus_name_owner = mfree(s->bus_name_owner); + service_close_socket_fd(s); - service_connection_unref(s); + s->peer = socket_peer_unref(s->peer); unit_ref_unset(&s->accept_socket); @@ -341,7 +364,7 @@ static int on_fd_store_io(sd_event_source *e, int fd, uint32_t revents, void *us return 0; } -static int service_add_fd_store(Service *s, int fd) { +static int service_add_fd_store(Service *s, int fd, const char *name) { ServiceFDStore *fs; int r; @@ -368,9 +391,15 @@ static int service_add_fd_store(Service *s, int fd) { 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; } @@ -383,7 +412,7 @@ static int service_add_fd_store(Service *s, int fd) { return 1; } -static int service_add_fd_store_set(Service *s, FDSet *fds) { +static int service_add_fd_store_set(Service *s, FDSet *fds, const char *name) { int r; assert(s); @@ -398,10 +427,9 @@ static int service_add_fd_store_set(Service *s, FDSet *fds) { if (fd < 0) break; - r = service_add_fd_store(s, fd); + r = service_add_fd_store(s, fd, name); if (r < 0) return log_unit_error_errno(UNIT(s), r, "Couldn't add fd to fd store: %m"); - if (r > 0) { log_unit_debug(UNIT(s), "Added fd to fd store."); fd = -1; @@ -409,7 +437,7 @@ static int service_add_fd_store_set(Service *s, FDSet *fds) { } if (fdset_size(fds) > 0) - log_unit_warning(UNIT(s), "Tried to store more fds than FDStoreMax=%u allows, closing remaining.", s->n_fd_store_max); + log_unit_warning(UNIT(s), "Tried to store more fds than FileDescriptorStoreMax=%u allows, closing remaining.", s->n_fd_store_max); return 0; } @@ -420,18 +448,21 @@ static int service_arm_timer(Service *s, usec_t usec) { assert(s); if (s->timer_event_source) { - r = sd_event_source_set_time(s->timer_event_source, now(CLOCK_MONOTONIC) + usec); + 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, - now(CLOCK_MONOTONIC) + usec, 0, + usec, 0, service_dispatch_timer, s); if (r < 0) return r; @@ -490,6 +521,15 @@ static int service_verify(Service *s) { 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; } @@ -498,15 +538,38 @@ static int service_add_default_dependencies(Service *s) { assert(s); + if (!UNIT(s)->default_dependencies) + return 0; + /* Add a number of automatic dependencies useful for the * majority of services. */ - /* First, pull in base system */ - r = unit_add_two_dependencies_by_name(UNIT(s), UNIT_AFTER, UNIT_REQUIRES, SPECIAL_BASIC_TARGET, NULL, true); + 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; - /* Second, activate normal shutdown */ + /* 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); } @@ -528,6 +591,32 @@ static void service_fix_output(Service *s) { 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; @@ -545,7 +634,7 @@ static int service_add_extras(Service *s) { /* Oneshot services have disabled start timeout by default */ if (s->type == SERVICE_ONESHOT && !s->start_timeout_defined) - s->timeout_start_usec = 0; + s->timeout_start_usec = USEC_INFINITY; service_fix_output(s); @@ -557,7 +646,7 @@ static int service_add_extras(Service *s) { if (r < 0) return r; - r = unit_add_default_slice(UNIT(s), &s->cgroup_context); + r = unit_set_default_slice(UNIT(s)); if (r < 0) return r; @@ -567,24 +656,13 @@ static int service_add_extras(Service *s) { if (s->watchdog_usec > 0 && s->notify_access == NOTIFY_NONE) s->notify_access = NOTIFY_MAIN; - if (s->bus_name) { - const char *n; - - n = strjoina(s->bus_name, ".busname"); - r = unit_add_dependency_by_name(UNIT(s), UNIT_AFTER, n, NULL, true); - if (r < 0) - return r; - - r = unit_watch_bus_name(UNIT(s), s->bus_name); - if (r < 0) - return r; - } + r = service_add_default_dependencies(s); + if (r < 0) + return r; - if (UNIT(s)->default_dependencies) { - r = service_add_default_dependencies(s); - if (r < 0) - return r; - } + r = service_setup_bus_name(s); + if (r < 0) + return r; return 0; } @@ -683,6 +761,11 @@ static void service_dump(Unit *u, FILE *f, const char *prefix) { 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); @@ -701,13 +784,12 @@ static void service_dump(Unit *u, FILE *f, const char *prefix) { fprintf(f, "%sStatus Text: %s\n", prefix, s->status_text); - if (s->n_fd_store_max > 0) { + 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) { @@ -765,8 +847,8 @@ static int service_load_pid_file(Service *s, bool may_warn) { return 0; } -static int service_search_main_pid(Service *s) { - pid_t pid; +static void service_search_main_pid(Service *s) { + pid_t pid = 0; int r; assert(s); @@ -774,30 +856,24 @@ static int service_search_main_pid(Service *s) { /* If we know it anyway, don't ever fallback to unreliable * heuristics */ if (s->main_pid_known) - return 0; + return; if (!s->guess_main_pid) - return 0; + return; assert(s->main_pid <= 0); - pid = unit_search_main_pid(UNIT(s)); - if (pid <= 0) - return -ENOENT; + if (unit_search_main_pid(UNIT(s), &pid) < 0) + return; log_unit_debug(UNIT(s), "Main PID guessed: "PID_FMT, pid); - r = service_set_main_pid(s, pid); - if (r < 0) - return r; + if (service_set_main_pid(s, pid) < 0) + return; r = unit_watch_pid(UNIT(s), pid); - if (r < 0) { + 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); - return r; - } - - return 0; } static void service_set_state(Service *s, ServiceState state) { @@ -815,6 +891,7 @@ static void service_set_state(Service *s, ServiceState state) { 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, @@ -848,18 +925,16 @@ static void service_set_state(Service *s, ServiceState state) { 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)) { + !(state == SERVICE_DEAD && UNIT(s)->job)) service_close_socket_fd(s); - service_connection_unref(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 && UNIT(s)->manager->n_reloading <= 0) - unit_destroy_cgroup_if_empty(UNIT(s)); + 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 @@ -874,7 +949,7 @@ static void service_set_state(Service *s, ServiceState state) { if (ec && exec_context_may_touch_console(ec)) { Manager *m = UNIT(s)->manager; - m->n_on_console --; + 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; @@ -885,7 +960,37 @@ static void service_set_state(Service *s, ServiceState 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); - 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) { @@ -895,125 +1000,138 @@ static int service_coldplug(Unit *u) { assert(s); assert(s->state == SERVICE_DEAD); - if (s->deserialized_state != s->state) { - - if (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)) { - - usec_t k; + if (s->deserialized_state == s->state) + return 0; - k = IN_SET(s->deserialized_state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RELOAD) ? s->timeout_start_usec : s->timeout_stop_usec; + r = service_arm_timer(s, service_coldplug_timeout(s)); + if (r < 0) + return r; - /* For the start/stop timeouts 0 means off */ - if (k > 0) { - r = service_arm_timer(s, k); - 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->deserialized_state == SERVICE_AUTO_RESTART) { + 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; + } - /* The restart timeouts 0 means immediately */ - r = service_arm_timer(s, s->restart_usec); - 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 (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 (IN_SET(s->deserialized_state, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) + service_start_watchdog(s); - if (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)) + (void) unit_setup_dynamic_creds(u); - if (!IN_SET(s->deserialized_state, SERVICE_DEAD, SERVICE_FAILED, SERVICE_AUTO_RESTART)) - unit_watch_all_pids(UNIT(s)); + if (UNIT_ISSET(s->accept_socket)) { + Socket* socket = SOCKET(UNIT_DEREF(s->accept_socket)); - if (IN_SET(s->deserialized_state, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD)) - service_start_watchdog(s); + if (socket->max_connections_per_source > 0) { + SocketPeer *peer; - service_set_state(s, s->deserialized_state); + /* 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, unsigned *n_fds) { +static int service_collect_fds(Service *s, int **fds, char ***fd_names) { + _cleanup_strv_free_ char **rfd_names = NULL; _cleanup_free_ int *rfds = NULL; - unsigned rn_fds = 0; - Iterator i; - int r; - Unit *u; + int rn_fds = 0, r; assert(s); assert(fds); - assert(n_fds); + assert(fd_names); - if (s->socket_fd >= 0) - return 0; + if (s->socket_fd >= 0) { - SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERED_BY], i) { - int *cfds; - unsigned cn_fds; - Socket *sock; + /* Pass the per-connection socket */ - if (u->type != UNIT_SOCKET) - continue; + rfds = new(int, 1); + if (!rfds) + return -ENOMEM; + rfds[0] = s->socket_fd; - sock = SOCKET(u); + rfd_names = strv_new("connection", NULL); + if (!rfd_names) + return -ENOMEM; - r = socket_collect_fds(sock, &cfds, &cn_fds); - if (r < 0) - return r; + rn_fds = 1; + } else { + Iterator i; + Unit *u; - if (cn_fds <= 0) { - free(cfds); - continue; - } + /* Pass all our configured sockets for singleton services */ - if (!rfds) { - rfds = cfds; - rn_fds = cn_fds; - } else { - int *t; + SET_FOREACH(u, UNIT(s)->dependencies[UNIT_TRIGGERED_BY], i) { + _cleanup_free_ int *cfds = NULL; + Socket *sock; + int cn_fds; - t = realloc(rfds, (rn_fds + cn_fds) * sizeof(int)); - if (!t) { - free(cfds); - return -ENOMEM; - } + 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; - memcpy(t + rn_fds, cfds, cn_fds * sizeof(int)); - rfds = t; - rn_fds += cn_fds; + cfds = NULL; + } else { + int *t; - free(cfds); + 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)); @@ -1021,49 +1139,68 @@ static int service_collect_fds(Service *s, int **fds, unsigned *n_fds) { return -ENOMEM; rfds = t; - LIST_FOREACH(fd_store, fs, s->fd_store) - rfds[rn_fds++] = fs->fd; + + 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; - *n_fds = rn_fds; + *fd_names = rfd_names; rfds = NULL; - return 0; + rfd_names = NULL; + + return rn_fds; } static int service_spawn( Service *s, ExecCommand *c, usec_t timeout, - bool pass_fds, - bool apply_permissions, - bool apply_chroot, - bool apply_tty_stdin, - bool is_control, + ExecFlags flags, pid_t *_pid) { - pid_t pid; - int r; - int *fds = NULL; - _cleanup_free_ int *fdsbuf = NULL; + _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; - _cleanup_free_ char *bus_endpoint_path = NULL; - _cleanup_strv_free_ char - **argv = NULL, **final_env = NULL, **our_env = NULL; const char *path; + pid_t pid; + ExecParameters exec_params = { - .apply_permissions = apply_permissions, - .apply_chroot = apply_chroot, - .apply_tty_stdin = apply_tty_stdin, - .bus_endpoint_fd = -1, - .selinux_context_net = s->socket_fd_selinux_context_net + .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)); @@ -1072,161 +1209,150 @@ static int service_spawn( r = unit_setup_exec_runtime(UNIT(s)); if (r < 0) - goto fail; + return r; + + r = unit_setup_dynamic_creds(UNIT(s)); + if (r < 0) + return r; - if (pass_fds || + 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) { - if (s->socket_fd >= 0) { - fds = &s->socket_fd; - n_fds = 1; - } else { - r = service_collect_fds(s, &fdsbuf, &n_fds); - if (r < 0) - goto fail; + r = service_collect_fds(s, &fds, &fd_names); + if (r < 0) + return r; - fds = fdsbuf; - } + n_fds = r; } - if (timeout > 0) { - r = service_arm_timer(s, timeout); - if (r < 0) - goto fail; - } else - s->timer_event_source = sd_event_source_unref(s->timer_event_source); + 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) - goto fail; + return r; - our_env = new0(char*, 6); - if (!our_env) { - r = -ENOMEM; - goto fail; - } + our_env = new0(char*, 9); + if (!our_env) + return -ENOMEM; - if (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) { - r = -ENOMEM; - goto fail; - } + 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) { - r = -ENOMEM; - goto fail; - } + if (asprintf(our_env + n_env++, "MAINPID="PID_FMT, s->main_pid) < 0) + return -ENOMEM; - if (UNIT(s)->manager->running_as != MANAGER_SYSTEM) - if (asprintf(our_env + n_env++, "MANAGERPID="PID_FMT, getpid()) < 0) { - r = -ENOMEM; - goto fail; - } + if (MANAGER_IS_USER(UNIT(s)->manager)) + if (asprintf(our_env + n_env++, "MANAGERPID="PID_FMT, getpid()) < 0) + return -ENOMEM; - if (UNIT_DEREF(s->accept_socket)) { + 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; - goto fail; + + /* 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 (IN_SET(sa.sa.sa_family, AF_INET, AF_INET6)) { + 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) - goto fail; + return r; t = strappend("REMOTE_ADDR=", addr); - if (!t) { - r = -ENOMEM; - goto fail; - } + if (!t) + return -ENOMEM; our_env[n_env++] = t; port = sockaddr_port(&sa.sa); - if (port < 0) { - r = port; - goto fail; - } + if (port < 0) + return port; - if (asprintf(&t, "REMOTE_PORT=%u", port) < 0) { - r = -ENOMEM; - goto fail; - } + if (asprintf(&t, "REMOTE_PORT=%u", port) < 0) + return -ENOMEM; our_env[n_env++] = t; } } - final_env = strv_env_merge(2, UNIT(s)->manager->environment, our_env, NULL); - if (!final_env) { - r = -ENOMEM; - goto fail; + 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; + } } - if (is_control && UNIT(s)->cgroup_path) { + 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"); - cg_create(SYSTEMD_CGROUP_CONTROLLER, path); + (void) cg_create(SYSTEMD_CGROUP_CONTROLLER, path); } else path = UNIT(s)->cgroup_path; - if (s->exec_context.bus_endpoint) { - r = bus_kernel_create_endpoint(UNIT(s)->manager->running_as == MANAGER_SYSTEM ? "system" : "user", - UNIT(s)->id, &bus_endpoint_path); - if (r < 0) - goto fail; - - /* Pass the fd to the exec_params so that the child process can upload the policy. - * Keep a reference to the fd in the service, so the endpoint is kept alive as long - * as the service is running. */ - exec_params.bus_endpoint_fd = s->bus_endpoint_fd = r; - } - 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.environment = final_env; - exec_params.confirm_spawn = UNIT(s)->manager->confirm_spawn; + 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.bus_endpoint_path = bus_endpoint_path; + 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) - goto fail; + return r; r = unit_watch_pid(UNIT(s), pid); if (r < 0) /* FIXME: we need to do something here */ - goto fail; + return r; *_pid = pid; return 0; - -fail: - if (timeout) - s->timer_event_source = sd_event_source_unref(s->timer_event_source); - - return r; } static int main_pid_good(Service *s) { @@ -1268,7 +1394,7 @@ static int cgroup_good(Service *s) { if (!UNIT(s)->cgroup_path) return 0; - r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, UNIT(s)->cgroup_path, true); + r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, UNIT(s)->cgroup_path); if (r < 0) return r; @@ -1322,19 +1448,19 @@ static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) int r; assert(s); - if (f != SERVICE_SUCCESS) + 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)); - failure_action(UNIT(s)->manager, s->failure_action, s->reboot_arg); + failure_action(UNIT(s)->manager, s->failure_action, UNIT(s)->reboot_arg); } if (allow_restart && service_shall_restart(s)) { - r = service_arm_timer(s, s->restart_usec); + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->restart_usec)); if (r < 0) goto fail; @@ -1348,14 +1474,20 @@ static void service_enter_dead(Service *s, ServiceResult f, bool allow_restart) exec_runtime_destroy(s->exec_runtime); s->exec_runtime = exec_runtime_unref(s->exec_runtime); - /* Also, remove the runtime directory in */ + /* 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) - unlink_noerrno(s->pid_file); + (void) unlink(s->pid_file); return; @@ -1368,7 +1500,7 @@ static void service_enter_stop_post(Service *s, ServiceResult f) { int r; assert(s); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; service_unwatch_control_pid(s); @@ -1381,11 +1513,7 @@ static void service_enter_stop_post(Service *s, ServiceResult f) { r = service_spawn(s, s->control_command, s->timeout_stop_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - true, - true, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_IS_CONTROL|EXEC_SETENV_RESULT, &s->control_pid); if (r < 0) goto fail; @@ -1425,7 +1553,7 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f assert(s); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; unit_watch_all_pids(UNIT(s)); @@ -1442,11 +1570,9 @@ static void service_enter_signal(Service *s, ServiceState state, ServiceResult f goto fail; if (r > 0) { - if (s->timeout_stop_usec > 0) { - r = service_arm_timer(s, s->timeout_stop_usec); - if (r < 0) - goto fail; - } + 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) @@ -1474,8 +1600,7 @@ static void service_enter_stop_by_notify(Service *s) { unit_watch_all_pids(UNIT(s)); - if (s->timeout_stop_usec > 0) - service_arm_timer(s, s->timeout_stop_usec); + 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); @@ -1486,7 +1611,7 @@ static void service_enter_stop(Service *s, ServiceResult f) { assert(s); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; service_unwatch_control_pid(s); @@ -1499,11 +1624,7 @@ static void service_enter_stop(Service *s, ServiceResult f) { r = service_spawn(s, s->control_command, s->timeout_stop_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - false, - true, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL|EXEC_SETENV_RESULT, &s->control_pid); if (r < 0) goto fail; @@ -1519,18 +1640,35 @@ fail: 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) { - int main_pid_ok, cgroup_ok; assert(s); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; - main_pid_ok = main_pid_good(s); - cgroup_ok = cgroup_good(s); + service_unwatch_control_pid(s); - if ((main_pid_ok > 0 || (main_pid_ok < 0 && cgroup_ok != 0)) && - (s->bus_name_good || s->type != SERVICE_DBUS)) { + if (service_good(s)) { /* If there are any queued up sd_notify() * notifications, process them now */ @@ -1538,8 +1676,10 @@ static void service_enter_running(Service *s, ServiceResult f) { service_enter_reload_by_notify(s); else if (s->notify_state == NOTIFY_STOPPING) service_enter_stop_by_notify(s); - else + 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); @@ -1561,11 +1701,7 @@ static void service_enter_start_post(Service *s) { r = service_spawn(s, s->control_command, s->timeout_start_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - false, - true, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL, &s->control_pid); if (r < 0) goto fail; @@ -1588,11 +1724,12 @@ static void service_kill_control_processes(Service *s) { return; p = strjoina(UNIT(s)->cgroup_path, "/control"); - cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, p, SIGKILL, true, true, true, NULL); + 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; @@ -1624,19 +1761,22 @@ static void service_enter_start(Service *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, - IN_SET(s->type, SERVICE_FORKING, SERVICE_DBUS, SERVICE_NOTIFY, SERVICE_ONESHOT) ? s->timeout_start_usec : 0, - true, - true, - true, - true, - false, + timeout, + EXEC_PASS_FDS|EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_SET_WATCHDOG, &pid); if (r < 0) goto fail; - if (s->type == SERVICE_SIMPLE || s->type == SERVICE_IDLE) { + if (IN_SET(s->type, SERVICE_SIMPLE, SERVICE_IDLE)) { /* For simple services we immediately start * the START_POST binaries. */ @@ -1651,9 +1791,7 @@ static void service_enter_start(Service *s) { s->control_pid = pid; service_set_state(s, SERVICE_START); - } else if (s->type == SERVICE_ONESHOT || - s->type == SERVICE_DBUS || - s->type == SERVICE_NOTIFY) { + } 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. */ @@ -1692,11 +1830,7 @@ static void service_enter_start_pre(Service *s) { r = service_spawn(s, s->control_command, s->timeout_start_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - true, - true, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL|EXEC_APPLY_TTY_STDIN, &s->control_pid); if (r < 0) goto fail; @@ -1713,7 +1847,7 @@ fail: } static void service_enter_restart(Service *s) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; int r; assert(s); @@ -1722,7 +1856,7 @@ static void service_enter_restart(Service *s) { /* 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, s->restart_usec); + r = service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->restart_usec)); if (r < 0) goto fail; @@ -1733,7 +1867,7 @@ static void service_enter_restart(Service *s) { * 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, false, &error, NULL); + r = manager_add_job(UNIT(s)->manager, JOB_RESTART, UNIT(s), JOB_FAIL, &error, NULL); if (r < 0) goto fail; @@ -1752,9 +1886,7 @@ fail: static void service_enter_reload_by_notify(Service *s) { assert(s); - if (s->timeout_start_usec > 0) - service_arm_timer(s, s->timeout_start_usec); - + service_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_start_usec)); service_set_state(s, SERVICE_RELOAD); } @@ -1764,6 +1896,7 @@ static void service_enter_reload(Service *s) { 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) { @@ -1772,11 +1905,7 @@ static void service_enter_reload(Service *s) { r = service_spawn(s, s->control_command, s->timeout_start_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - false, - true, + EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_IS_CONTROL, &s->control_pid); if (r < 0) goto fail; @@ -1794,6 +1923,7 @@ fail: } static void service_run_next_control(Service *s) { + usec_t timeout; int r; assert(s); @@ -1805,15 +1935,17 @@ static void service_run_next_control(Service *s) { 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, - IN_SET(s->state, SERVICE_START_PRE, SERVICE_START, SERVICE_START_POST, SERVICE_RUNNING, SERVICE_RELOAD) ? s->timeout_start_usec : s->timeout_stop_usec, - false, - !s->permissions_start_only, - !s->root_directory_start_only, - s->control_command_id == SERVICE_EXEC_START_PRE || - s->control_command_id == SERVICE_EXEC_STOP_POST, - true, + 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; @@ -1851,11 +1983,7 @@ static void service_run_next_main(Service *s) { r = service_spawn(s, s->main_command, s->timeout_start_usec, - true, - true, - true, - true, - false, + EXEC_PASS_FDS|EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN|EXEC_SET_WATCHDOG, &pid); if (r < 0) goto fail; @@ -1869,17 +1997,6 @@ fail: service_enter_stop(s, SERVICE_FAILURE_RESOURCES); } -static int service_start_limit_test(Service *s) { - assert(s); - - if (ratelimit_test(&s->start_limit)) - return 0; - - log_unit_warning(UNIT(s), "Start request repeated too quickly."); - - return failure_action(UNIT(s)->manager, s->start_limit_action, s->reboot_arg); -} - static int service_start(Unit *u) { Service *s = SERVICE(u); int r; @@ -1910,12 +2027,16 @@ static int service_start(Unit *u) { assert(IN_SET(s->state, SERVICE_DEAD, SERVICE_FAILED)); /* Make sure we don't enter a busy loop of some kind. */ - r = service_start_limit_test(s); + r = unit_start_limit_test(u); if (r < 0) { - service_enter_dead(s, SERVICE_FAILURE_START_LIMIT, false); + 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; @@ -1923,12 +2044,14 @@ static int service_start(Unit *u) { s->forbid_restart = false; s->reset_cpu_usage = true; - free(s->status_text); - s->status_text = NULL; + 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; } @@ -1974,7 +2097,7 @@ static int service_reload(Unit *u) { assert(s->state == SERVICE_RUNNING || s->state == SERVICE_EXITED); service_enter_reload(s); - return 0; + return 1; } _pure_ static bool service_can_reload(Unit *u) { @@ -1988,6 +2111,7 @@ _pure_ static bool service_can_reload(Unit *u) { static int service_serialize(Unit *u, FILE *f, FDSet *fds) { Service *s = SERVICE(u); ServiceFDStore *fs; + int r; assert(u); assert(f); @@ -2005,13 +2129,11 @@ static int service_serialize(Unit *u, FILE *f, FDSet *fds) { 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); - if (s->status_text) { - _cleanup_free_ char *c = NULL; - - c = cescape(s->status_text); - unit_serialize_item(u, f, "status-text", strempty(c)); - } + 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 @@ -2019,34 +2141,37 @@ static int service_serialize(Unit *u, FILE *f, FDSet *fds) { if (s->control_command_id >= 0) unit_serialize_item(u, f, "control-command", service_exec_command_to_string(s->control_command_id)); - if (s->socket_fd >= 0) { - int copy; - - copy = fdset_put_dup(fds, s->socket_fd); - if (copy < 0) - return copy; + 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; - unit_serialize_item_format(u, f, "socket-fd", "%i", copy); + 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; } - if (s->bus_endpoint_fd >= 0) { - int copy; - - copy = fdset_put_dup(fds, s->bus_endpoint_fd); - if (copy < 0) - return copy; - - unit_serialize_item_format(u, f, "endpoint-fd", "%i", copy); - } + 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; - unit_serialize_item_format(u, f, "fd-store-fd", "%i", 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) { @@ -2060,11 +2185,12 @@ static int service_serialize(Unit *u, FILE *f, FDSet *fds) { } } - if (dual_timestamp_is_set(&s->watchdog_timestamp)) - dual_timestamp_serialize(f, "watchdog-timestamp", &s->watchdog_timestamp); + dual_timestamp_serialize(f, "watchdog-timestamp", &s->watchdog_timestamp); - if (s->forbid_restart) - unit_serialize_item(u, f, "forbid-restart", yes_no(s->forbid_restart)); + 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; } @@ -2136,6 +2262,10 @@ static int service_deserialize_item(Unit *u, const char *key, const char *value, 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; @@ -2157,6 +2287,17 @@ static int service_deserialize_item(Unit *u, const char *key, const char *value, 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; @@ -2166,22 +2307,25 @@ static int service_deserialize_item(Unit *u, const char *key, const char *value, asynchronous_close(s->socket_fd); s->socket_fd = fdset_remove(fds, fd); } - } else if (streq(key, "endpoint-fd")) { - int fd; - - if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, fd)) - log_unit_debug(u, "Failed to parse endpoint-fd value: %s", value); - else { - safe_close(s->bus_endpoint_fd); - s->bus_endpoint_fd = fdset_remove(fds, fd); - } } else if (streq(key, "fd-store-fd")) { + const char *fdv; + size_t pf; int fd; - if (safe_atoi(value, &fd) < 0 || fd < 0 || !fdset_contains(fds, 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 { - r = service_add_fd_store(s, fd); + _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 if (r > 0) @@ -2223,6 +2367,44 @@ static int service_deserialize_item(Unit *u, const char *key, const char *value, 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); @@ -2260,14 +2442,6 @@ static bool service_check_gc(Unit *u) { return false; } -_pure_ static bool service_check_snapshot(Unit *u) { - Service *s = SERVICE(u); - - assert(s); - - return s->socket_fd < 0; -} - static int service_retry_pid_file(Service *s) { int r; @@ -2430,8 +2604,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { assert(s); assert(pid >= 0); - if (UNIT(s)->fragment_path ? is_clean_exit(code, status, &s->success_status) : - is_clean_exit_lsb(code, status, &s->success_status)) + 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; @@ -2484,7 +2657,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { "EXIT_STATUS=%i", status, NULL); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; if (s->main_command && @@ -2565,7 +2738,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { "Control process exited, code=%s status=%i", sigchld_code_to_string(code), status); - if (f != SERVICE_SUCCESS) + if (s->result == SERVICE_SUCCESS) s->result = f; /* Immediately get rid of the cgroup, so that the @@ -2635,7 +2808,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { case SERVICE_START_POST: if (f != SERVICE_SUCCESS) { - service_enter_stop(s, f); + service_enter_signal(s, SERVICE_STOP_SIGTERM, f); break; } @@ -2656,10 +2829,9 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { break; case SERVICE_RELOAD: - if (f == SERVICE_SUCCESS) { - service_load_pid_file(s, true); - service_search_main_pid(s); - } + 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); @@ -2704,8 +2876,9 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) { 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 */ - if (set_isempty(u->pids)) + /* 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); } @@ -2725,11 +2898,17 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us 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. Stopping."); + 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; @@ -2803,12 +2982,15 @@ static int service_dispatch_timer(sd_event_source *source, usec_t usec, void *us 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), s->watchdog_usec, 1)); + format_timespan(t, sizeof(t), watchdog_usec, 1)); service_enter_signal(s, SERVICE_STOP_SIGABRT, SERVICE_FAILURE_WATCHDOG); @@ -2930,13 +3112,28 @@ static void service_notify_message(Unit *u, pid_t pid, char **tags, FDSet *fds) } /* Interpret WATCHDOG= */ - if (strv_find(tags, "WATCHDOG=1")) { + if (strv_find(tags, "WATCHDOG=1")) service_reset_watchdog(s); - } - /* Add the passed fds to the fd store */ if (strv_find(tags, "FDSTORE=1")) { - service_add_fd_store_set(s, fds); + 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 */ @@ -2944,17 +3141,21 @@ static void service_notify_message(Unit *u, pid_t pid, char **tags, FDSet *fds) unit_add_to_dbus_queue(u); } -static int service_get_timeout(Unit *u, uint64_t *timeout) { +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, timeout); + 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; } @@ -2982,6 +3183,13 @@ static void service_bus_name_owner_change( 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 @@ -2998,7 +3206,7 @@ static void service_bus_name_owner_change( s->state == SERVICE_RUNNING || s->state == SERVICE_RELOAD)) { - _cleanup_bus_creds_unref_ sd_bus_creds *creds = NULL; + _cleanup_(sd_bus_creds_unrefp) sd_bus_creds *creds = NULL; pid_t pid; /* Try to acquire PID from bus service */ @@ -3022,9 +3230,8 @@ int service_set_socket_fd(Service *s, int fd, Socket *sock, bool selinux_context 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. */ + /* 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; @@ -3035,7 +3242,7 @@ int service_set_socket_fd(Service *s, int fd, Socket *sock, bool selinux_context if (s->state != SERVICE_DEAD) return -EAGAIN; - if (getpeername_pretty(fd, &peer) >= 0) { + if (getpeername_pretty(fd, true, &peer) >= 0) { if (UNIT(s)->description) { _cleanup_free_ char *a; @@ -3052,12 +3259,15 @@ int service_set_socket_fd(Service *s, int fd, Socket *sock, bool selinux_context 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 unit_add_two_dependencies(UNIT(sock), UNIT_BEFORE, UNIT_TRIGGERS, UNIT(s), false); + return 0; } static void service_reset_failed(Unit *u) { @@ -3070,8 +3280,6 @@ static void service_reset_failed(Unit *u) { s->result = SERVICE_SUCCESS; s->reload_result = SERVICE_SUCCESS; - - RATELIMIT_RESET(s->start_limit); } static int service_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { @@ -3080,26 +3288,21 @@ static int service_kill(Unit *u, KillWho who, int signo, sd_bus_error *error) { return unit_kill_common(u, who, signo, s->main_pid, s->control_pid, error); } -static const char* const service_state_table[_SERVICE_STATE_MAX] = { - [SERVICE_DEAD] = "dead", - [SERVICE_START_PRE] = "start-pre", - [SERVICE_START] = "start", - [SERVICE_START_POST] = "start-post", - [SERVICE_RUNNING] = "running", - [SERVICE_EXITED] = "exited", - [SERVICE_RELOAD] = "reload", - [SERVICE_STOP] = "stop", - [SERVICE_STOP_SIGABRT] = "stop-sigabrt", - [SERVICE_STOP_SIGTERM] = "stop-sigterm", - [SERVICE_STOP_SIGKILL] = "stop-sigkill", - [SERVICE_STOP_POST] = "stop-post", - [SERVICE_FINAL_SIGTERM] = "final-sigterm", - [SERVICE_FINAL_SIGKILL] = "final-sigkill", - [SERVICE_FAILED] = "failed", - [SERVICE_AUTO_RESTART] = "auto-restart", -}; +static int service_main_pid(Unit *u) { + Service *s = SERVICE(u); -DEFINE_STRING_TABLE_LOOKUP(service_state, ServiceState); + 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", @@ -3160,7 +3363,7 @@ static const char* const service_result_table[_SERVICE_RESULT_MAX] = { [SERVICE_FAILURE_SIGNAL] = "signal", [SERVICE_FAILURE_CORE_DUMP] = "core-dump", [SERVICE_FAILURE_WATCHDOG] = "watchdog", - [SERVICE_FAILURE_START_LIMIT] = "start-limit" + [SERVICE_FAILURE_START_LIMIT_HIT] = "start-limit-hit", }; DEFINE_STRING_TABLE_LOOKUP(service_result, ServiceResult); @@ -3171,6 +3374,7 @@ const UnitVTable service_vtable = { .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" @@ -3202,7 +3406,6 @@ const UnitVTable service_vtable = { .sub_state_to_string = service_sub_state_to_string, .check_gc = service_check_gc, - .check_snapshot = service_check_snapshot, .sigchld_event = service_sigchld_event, @@ -3211,9 +3414,11 @@ const UnitVTable service_vtable = { .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_interface = "org.freedesktop.systemd1.Service", .bus_vtable = bus_service_vtable, .bus_set_property = bus_service_set_property, .bus_commit_properties = bus_service_commit_properties, @@ -3229,13 +3434,10 @@ const UnitVTable service_vtable = { .finished_start_job = { [JOB_DONE] = "Started %s.", [JOB_FAILED] = "Failed to start %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", - [JOB_TIMEOUT] = "Timed out starting %s.", }, .finished_stop_job = { [JOB_DONE] = "Stopped %s.", [JOB_FAILED] = "Stopped (with error) %s.", - [JOB_TIMEOUT] = "Timed out stopping %s.", }, }, }; diff --git a/src/core/service.h b/src/core/service.h index 7da0a93961..888007cc0b 100644 --- a/src/core/service.h +++ b/src/core/service.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -24,31 +22,10 @@ typedef struct Service Service; typedef struct ServiceFDStore ServiceFDStore; +#include "exit-status.h" +#include "kill.h" #include "path.h" #include "ratelimit.h" -#include "kill.h" -#include "exit-status.h" - -typedef enum ServiceState { - SERVICE_DEAD, - SERVICE_START_PRE, - SERVICE_START, - SERVICE_START_POST, - SERVICE_RUNNING, - SERVICE_EXITED, /* Nothing is running anymore, but RemainAfterExit is true hence this is OK */ - SERVICE_RELOAD, - SERVICE_STOP, /* No STOP_PRE state, instead just register multiple STOP executables */ - SERVICE_STOP_SIGABRT, /* Watchdog timeout */ - SERVICE_STOP_SIGTERM, - SERVICE_STOP_SIGKILL, - SERVICE_STOP_POST, - SERVICE_FINAL_SIGTERM, /* In case the STOP_POST executable hangs, we shoot that down, too */ - SERVICE_FINAL_SIGKILL, - SERVICE_FAILED, - SERVICE_AUTO_RESTART, - _SERVICE_STATE_MAX, - _SERVICE_STATE_INVALID = -1 -} ServiceState; typedef enum ServiceRestart { SERVICE_RESTART_NO, @@ -103,13 +80,13 @@ typedef enum NotifyState { typedef enum ServiceResult { SERVICE_SUCCESS, - SERVICE_FAILURE_RESOURCES, + 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, + SERVICE_FAILURE_START_LIMIT_HIT, _SERVICE_RESULT_MAX, _SERVICE_RESULT_INVALID = -1 } ServiceResult; @@ -118,6 +95,7 @@ struct ServiceFDStore { Service *service; int fd; + char *fdname; sd_event_source *event_source; LIST_FIELDS(ServiceFDStore, fd_store); @@ -138,9 +116,12 @@ struct Service { 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]; @@ -167,13 +148,13 @@ struct Service { /* 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; - int bus_endpoint_fd; - bool permissions_start_only; bool root_directory_start_only; bool remain_after_exit; @@ -192,14 +173,12 @@ struct Service { bool reset_cpu_usage:1; char *bus_name; + char *bus_name_owner; /* unique name of the current owner */ char *status_text; int status_errno; - RateLimit start_limit; - FailureAction start_limit_action; FailureAction failure_action; - char *reboot_arg; UnitRef accept_socket; @@ -212,14 +191,19 @@ struct Service { 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); - -const char* service_state_to_string(ServiceState i) _const_; -ServiceState service_state_from_string(const char *s) _pure_; +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_; diff --git a/src/core/show-status.c b/src/core/show-status.c index 02b1be73e3..59ebdc7219 100644 --- a/src/core/show-status.c +++ b/src/core/show-status.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,7 +17,13 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include "alloc-util.h" +#include "fd-util.h" +#include "io-util.h" +#include "parse-util.h" #include "show-status.h" +#include "string-util.h" +#include "terminal-util.h" #include "util.h" int parse_show_status(const char *v, ShowStatus *ret) { @@ -40,3 +44,81 @@ int parse_show_status(const char *v, ShowStatus *ret) { *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(); + + 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/core/show-status.h b/src/core/show-status.h index a2b2153746..9a29e72645 100644 --- a/src/core/show-status.h +++ b/src/core/show-status.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -21,6 +19,10 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ +#include <stdbool.h> + +#include "macro.h" + /* Manager status */ typedef enum ShowStatus { @@ -32,3 +34,6 @@ typedef enum ShowStatus { } 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/core/shutdown.c b/src/core/shutdown.c index aba16b4689..a795d875bb 100644 --- a/src/core/shutdown.c +++ b/src/core/shutdown.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,35 +17,39 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/mman.h> -#include <sys/reboot.h> -#include <linux/reboot.h> -#include <sys/stat.h> -#include <sys/mount.h> #include <errno.h> -#include <unistd.h> +#include <getopt.h> +#include <linux/reboot.h> #include <signal.h> #include <stdbool.h> #include <stdlib.h> -#include <getopt.h> +#include <sys/mman.h> +#include <sys/mount.h> +#include <sys/reboot.h> +#include <sys/stat.h> +#include <unistd.h> -#include "missing.h" -#include "log.h" +#include "alloc-util.h" +#include "cgroup-util.h" +#include "def.h" #include "fileio.h" +#include "killall.h" +#include "log.h" +#include "missing.h" +#include "parse-util.h" +#include "process-util.h" +#include "string-util.h" +#include "switch-root.h" +#include "terminal-util.h" #include "umount.h" #include "util.h" #include "virt.h" #include "watchdog.h" -#include "killall.h" -#include "cgroup-util.h" -#include "def.h" -#include "switch-root.h" -#include "process-util.h" -#include "terminal-util.h" #define FINALIZE_ATTEMPTS 50 static char* arg_verb; +static uint8_t arg_exit_code; static int parse_argv(int argc, char *argv[]) { enum { @@ -55,6 +57,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_LOG_TARGET, ARG_LOG_COLOR, ARG_LOG_LOCATION, + ARG_EXIT_CODE, }; static const struct option options[] = { @@ -62,6 +65,7 @@ static int parse_argv(int argc, char *argv[]) { { "log-target", required_argument, NULL, ARG_LOG_TARGET }, { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, + { "exit-code", required_argument, NULL, ARG_EXIT_CODE }, {} }; @@ -110,6 +114,13 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_EXIT_CODE: + r = safe_atou8(optarg, &arg_exit_code); + if (r < 0) + log_error("Failed to parse exit code %s, ignoring", optarg); + + break; + case '\001': if (!arg_verb) arg_verb = optarg; @@ -146,7 +157,6 @@ static int switch_root_initramfs(void) { return switch_root("/run/initramfs", "/oldroot", false, MS_BIND); } - int main(int argc, char *argv[]) { bool need_umount, need_swapoff, need_loop_detach, need_dm_detach; bool in_container, use_watchdog = false; @@ -183,27 +193,34 @@ int main(int argc, char *argv[]) { cmd = RB_HALT_SYSTEM; else if (streq(arg_verb, "kexec")) cmd = LINUX_REBOOT_CMD_KEXEC; + else if (streq(arg_verb, "exit")) + cmd = 0; /* ignored, just checking that arg_verb is valid */ else { r = -EINVAL; log_error("Unknown action '%s'.", arg_verb); goto error; } - cg_get_root_path(&cgroup); + (void) cg_get_root_path(&cgroup); + in_container = detect_container() > 0; use_watchdog = !!getenv("WATCHDOG_USEC"); - /* lock us into memory */ + /* Lock us into memory */ mlockall(MCL_CURRENT|MCL_FUTURE); + /* Synchronize everything that is not written to disk yet at this point already. This is a good idea so that + * slow IO is processed here already and the final process killing spree is not impacted by processes + * desperately trying to sync IO to disk within their timeout. */ + if (!in_container) + sync(); + log_info("Sending SIGTERM to remaining processes..."); broadcast_signal(SIGTERM, true, true); log_info("Sending SIGKILL to remaining processes..."); broadcast_signal(SIGKILL, true, false); - in_container = detect_container(NULL) > 0; - need_umount = !in_container; need_swapoff = !in_container; need_loop_detach = !in_container; @@ -332,13 +349,23 @@ int main(int argc, char *argv[]) { need_loop_detach ? " loop devices," : "", need_dm_detach ? " DM devices," : ""); - /* The kernel will automaticall flush ATA disks and suchlike - * on reboot(), but the file systems need to be synce'd - * explicitly in advance. So let's do this here, but not - * needlessly slow down containers. */ + /* The kernel will automatically flush ATA disks and suchlike on reboot(), but the file systems need to be + * sync'ed explicitly in advance. So let's do this here, but not needlessly slow down containers. Note that we + * sync'ed things already once above, but we did some more work since then which might have caused IO, hence + * let's doit once more. */ if (!in_container) sync(); + if (streq(arg_verb, "exit")) { + if (in_container) + exit(arg_exit_code); + else { + /* We cannot exit() on the host, fallback on another + * method. */ + cmd = RB_POWER_OFF; + } + } + switch (cmd) { case LINUX_REBOOT_CMD_KEXEC: @@ -374,9 +401,14 @@ int main(int argc, char *argv[]) { if (!in_container) { _cleanup_free_ char *param = NULL; - if (read_one_line_file(REBOOT_PARAM_FILE, ¶m) >= 0) { + r = read_one_line_file("/run/systemd/reboot-param", ¶m); + if (r < 0) + log_warning_errno(r, "Failed to read reboot parameter file: %m"); + + if (!isempty(param)) { log_info("Rebooting with argument '%s'.", param); syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_RESTART2, param); + log_warning_errno(errno, "Failed to reboot with parameter, retrying without: %m"); } } @@ -404,11 +436,9 @@ int main(int argc, char *argv[]) { exit(0); } - log_error_errno(errno, "Failed to invoke reboot(): %m"); - r = -errno; + r = log_error_errno(errno, "Failed to invoke reboot(): %m"); error: log_emergency_errno(r, "Critical error while doing system shutdown: %m"); - freeze(); } diff --git a/src/core/slice.c b/src/core/slice.c index e52bf71515..03fe797f27 100644 --- a/src/core/slice.c +++ b/src/core/slice.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,18 +19,28 @@ #include <errno.h> -#include "unit.h" -#include "slice.h" -#include "log.h" +#include "alloc-util.h" #include "dbus-slice.h" +#include "log.h" +#include "slice.h" #include "special.h" +#include "string-util.h" +#include "strv.h" #include "unit-name.h" +#include "unit.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); @@ -82,6 +90,9 @@ static int slice_add_default_dependencies(Slice *s) { 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), @@ -93,7 +104,6 @@ static int slice_add_default_dependencies(Slice *s) { return 0; } - static int slice_verify(Slice *s) { _cleanup_free_ char *parent = NULL; int r; @@ -125,6 +135,7 @@ static int slice_load(Unit *u) { int r; assert(s); + assert(u->load_state == UNIT_STUB); r = unit_load_fragment_and_dropin_optional(u); if (r < 0) @@ -141,11 +152,9 @@ static int slice_load(Unit *u) { if (r < 0) return r; - if (u->default_dependencies) { - r = slice_add_default_dependencies(s); - if (r < 0) - return r; - } + r = slice_add_default_dependencies(s); + if (r < 0) + return r; } return slice_verify(s); @@ -178,10 +187,15 @@ static void slice_dump(Unit *u, FILE *f, const char *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); @@ -252,12 +266,43 @@ _pure_ static const char *slice_sub_state_to_string(Unit *u) { return slice_state_to_string(SLICE(u)->state); } -static const char* const slice_state_table[_SLICE_STATE_MAX] = { - [SLICE_DEAD] = "dead", - [SLICE_ACTIVE] = "active" -}; +static void slice_enumerate(Manager *m) { + Unit *u; + int r; + + assert(m); + + u = manager_get_unit(m, SPECIAL_ROOT_SLICE); + if (!u) { + u = unit_new(m, sizeof(Slice)); + if (!u) { + log_oom(); + return; + } + + r = unit_add_name(u, SPECIAL_ROOT_SLICE); + if (r < 0) { + unit_free(u); + log_error_errno(r, "Failed to add -.slice name"); + return; + } + } -DEFINE_STRING_TABLE_LOOKUP(slice_state, SliceState); + u->default_dependencies = false; + u->no_gc = true; + u->ignore_on_isolate = true; + u->refuse_manual_start = true; + u->refuse_manual_stop = true; + SLICE(u)->deserialized_state = SLICE_ACTIVE; + + if (!u->description) + u->description = strdup("Root Slice"); + if (!u->documentation) + (void) strv_extend(&u->documentation, "man:systemd.special(7)"); + + unit_add_to_load_queue(u); + unit_add_to_dbus_queue(u); +} const UnitVTable slice_vtable = { .object_size = sizeof(Slice), @@ -269,9 +314,9 @@ const UnitVTable slice_vtable = { "Install\0", .private_section = "Slice", - .no_alias = true, - .no_instances = true, + .can_transient = true, + .init = slice_init, .load = slice_load, .coldplug = slice_coldplug, @@ -289,15 +334,15 @@ const UnitVTable slice_vtable = { .active_state = slice_active_state, .sub_state_to_string = slice_sub_state_to_string, - .bus_interface = "org.freedesktop.systemd1.Slice", .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.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", }, .finished_stop_job = { [JOB_DONE] = "Removed slice %s.", diff --git a/src/core/slice.h b/src/core/slice.h index ac648e56f8..c9f3f61067 100644 --- a/src/core/slice.h +++ b/src/core/slice.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,14 +21,6 @@ typedef struct Slice Slice; - -typedef enum SliceState { - SLICE_DEAD, - SLICE_ACTIVE, - _SLICE_STATE_MAX, - _SLICE_STATE_INVALID = -1 -} SliceState; - struct Slice { Unit meta; @@ -40,6 +30,3 @@ struct Slice { }; extern const UnitVTable slice_vtable; - -const char* slice_state_to_string(SliceState i) _const_; -SliceState slice_state_from_string(const char *s) _pure_; diff --git a/src/core/smack-setup.c b/src/core/smack-setup.c index cbe7d0b4a9..5a6d11cfa1 100644 --- a/src/core/smack-setup.c +++ b/src/core/smack-setup.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,18 +19,22 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <stdio.h> +#include <dirent.h> #include <errno.h> -#include <string.h> -#include <stdlib.h> #include <fcntl.h> -#include <dirent.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "alloc-util.h" +#include "dirent-util.h" +#include "fd-util.h" +#include "fileio.h" +#include "log.h" #include "macro.h" #include "smack-setup.h" +#include "string-util.h" #include "util.h" -#include "fileio.h" -#include "log.h" #ifdef HAVE_SMACK @@ -193,6 +195,75 @@ static int write_cipso2_rules(const char* srcdir) { 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) { @@ -215,16 +286,24 @@ int mac_smack_setup(bool *loaded_policy) { log_info("Successfully loaded Smack policies."); break; default: - log_warning("Failed to load Smack access rules: %s, ignoring.", - strerror(abs(r))); + 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) - log_warning("Failed to set SMACK label \"%s\" on self: %s", - SMACK_RUN_LABEL, strerror(-r)); + 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/"); @@ -234,14 +313,29 @@ int mac_smack_setup(bool *loaded_policy) { return 0; case ENOENT: log_debug("Smack/CIPSO access rules directory '/etc/smack/cipso.d/' not found"); - return 0; + break; case 0: log_info("Successfully loaded Smack/CIPSO policies."); break; default: - log_warning("Failed to load Smack/CIPSO access rules: %s, ignoring.", - strerror(abs(r))); + 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; diff --git a/src/core/smack-setup.h b/src/core/smack-setup.h index 1cab7718ff..78164c85e6 100644 --- a/src/core/smack-setup.h +++ b/src/core/smack-setup.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/snapshot.c b/src/core/snapshot.c deleted file mode 100644 index 1e634b9bc1..0000000000 --- a/src/core/snapshot.c +++ /dev/null @@ -1,308 +0,0 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - -/*** - 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 "unit.h" -#include "snapshot.h" -#include "unit-name.h" -#include "dbus-snapshot.h" -#include "bus-common-errors.h" - -static const UnitActiveState state_translation_table[_SNAPSHOT_STATE_MAX] = { - [SNAPSHOT_DEAD] = UNIT_INACTIVE, - [SNAPSHOT_ACTIVE] = UNIT_ACTIVE -}; - -static void snapshot_init(Unit *u) { - Snapshot *s = SNAPSHOT(u); - - assert(s); - assert(UNIT(s)->load_state == UNIT_STUB); - - UNIT(s)->ignore_on_isolate = true; - UNIT(s)->ignore_on_snapshot = true; - UNIT(s)->allow_isolate = true; -} - -static void snapshot_set_state(Snapshot *s, SnapshotState state) { - SnapshotState old_state; - assert(s); - - old_state = s->state; - s->state = state; - - if (state != old_state) - log_unit_debug(UNIT(s), "Changed %s -> %s", snapshot_state_to_string(old_state), snapshot_state_to_string(state)); - - unit_notify(UNIT(s), state_translation_table[old_state], state_translation_table[state], true); -} - -static int snapshot_load(Unit *u) { - Snapshot *s = SNAPSHOT(u); - - assert(u); - assert(u->load_state == UNIT_STUB); - - /* Make sure that only snapshots created via snapshot_create() - * can be loaded */ - if (!u->transient && UNIT(s)->manager->n_reloading <= 0) - return -ENOENT; - - u->load_state = UNIT_LOADED; - return 0; -} - -static int snapshot_coldplug(Unit *u) { - Snapshot *s = SNAPSHOT(u); - - assert(s); - assert(s->state == SNAPSHOT_DEAD); - - if (s->deserialized_state != s->state) - snapshot_set_state(s, s->deserialized_state); - - return 0; -} - -static void snapshot_dump(Unit *u, FILE *f, const char *prefix) { - Snapshot *s = SNAPSHOT(u); - - assert(s); - assert(f); - - fprintf(f, - "%sSnapshot State: %s\n" - "%sClean Up: %s\n", - prefix, snapshot_state_to_string(s->state), - prefix, yes_no(s->cleanup)); -} - -static int snapshot_start(Unit *u) { - Snapshot *s = SNAPSHOT(u); - - assert(s); - assert(s->state == SNAPSHOT_DEAD); - - snapshot_set_state(s, SNAPSHOT_ACTIVE); - - if (s->cleanup) - unit_add_to_cleanup_queue(u); - - return 1; -} - -static int snapshot_stop(Unit *u) { - Snapshot *s = SNAPSHOT(u); - - assert(s); - assert(s->state == SNAPSHOT_ACTIVE); - - snapshot_set_state(s, SNAPSHOT_DEAD); - return 1; -} - -static int snapshot_serialize(Unit *u, FILE *f, FDSet *fds) { - Snapshot *s = SNAPSHOT(u); - Unit *other; - Iterator i; - - assert(s); - assert(f); - assert(fds); - - unit_serialize_item(u, f, "state", snapshot_state_to_string(s->state)); - unit_serialize_item(u, f, "cleanup", yes_no(s->cleanup)); - SET_FOREACH(other, u->dependencies[UNIT_WANTS], i) - unit_serialize_item(u, f, "wants", other->id); - - return 0; -} - -static int snapshot_deserialize_item(Unit *u, const char *key, const char *value, FDSet *fds) { - Snapshot *s = SNAPSHOT(u); - int r; - - assert(u); - assert(key); - assert(value); - assert(fds); - - if (streq(key, "state")) { - SnapshotState state; - - state = snapshot_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, "cleanup")) { - - r = parse_boolean(value); - if (r < 0) - log_unit_debug(u, "Failed to parse cleanup value: %s", value); - else - s->cleanup = r; - - } else if (streq(key, "wants")) { - - r = unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_WANTS, value, NULL, true); - if (r < 0) - return r; - } else - log_unit_debug(u, "Unknown serialization key: %s", key); - - return 0; -} - -_pure_ static UnitActiveState snapshot_active_state(Unit *u) { - assert(u); - - return state_translation_table[SNAPSHOT(u)->state]; -} - -_pure_ static const char *snapshot_sub_state_to_string(Unit *u) { - assert(u); - - return snapshot_state_to_string(SNAPSHOT(u)->state); -} - -int snapshot_create(Manager *m, const char *name, bool cleanup, sd_bus_error *e, Snapshot **_s) { - _cleanup_free_ char *n = NULL; - Unit *other, *u = NULL; - Iterator i; - int r; - const char *k; - - assert(m); - assert(_s); - - if (name) { - if (!unit_name_is_valid(name, UNIT_NAME_PLAIN)) - return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Unit name %s is not valid.", name); - - if (!endswith(name, ".snapshot")) - return sd_bus_error_setf(e, SD_BUS_ERROR_INVALID_ARGS, "Unit name %s lacks snapshot suffix.", name); - - if (manager_get_unit(m, name)) - return sd_bus_error_setf(e, BUS_ERROR_UNIT_EXISTS, "Snapshot %s exists already.", name); - - } else { - - for (;;) { - if (asprintf(&n, "snapshot-%u.snapshot", ++ m->n_snapshots) < 0) - return -ENOMEM; - - if (!manager_get_unit(m, n)) { - name = n; - break; - } - - free(n); - n = NULL; - } - } - - r = manager_load_unit_prepare(m, name, NULL, e, &u); - if (r < 0) - goto fail; - - u->transient = true; - manager_dispatch_load_queue(m); - assert(u->load_state == UNIT_LOADED); - - HASHMAP_FOREACH_KEY(other, k, m->units, i) { - - if (other->ignore_on_snapshot || - other->transient) - continue; - - if (k != other->id) - continue; - - if (UNIT_VTABLE(other)->check_snapshot) - if (!UNIT_VTABLE(other)->check_snapshot(other)) - continue; - - if (!UNIT_IS_ACTIVE_OR_ACTIVATING(unit_active_state(other))) - continue; - - r = unit_add_two_dependencies(u, UNIT_AFTER, UNIT_WANTS, other, true); - if (r < 0) - goto fail; - } - - SNAPSHOT(u)->cleanup = cleanup; - *_s = SNAPSHOT(u); - - log_unit_info(u, "Created snapshot."); - - return 0; - -fail: - if (u) - unit_add_to_cleanup_queue(u); - - return r; -} - -void snapshot_remove(Snapshot *s) { - assert(s); - - log_unit_info(UNIT(s), "Removing snapshot."); - - unit_add_to_cleanup_queue(UNIT(s)); -} - -static const char* const snapshot_state_table[_SNAPSHOT_STATE_MAX] = { - [SNAPSHOT_DEAD] = "dead", - [SNAPSHOT_ACTIVE] = "active" -}; - -DEFINE_STRING_TABLE_LOOKUP(snapshot_state, SnapshotState); - -const UnitVTable snapshot_vtable = { - .object_size = sizeof(Snapshot), - - .no_alias = true, - .no_instances = true, - .no_gc = true, - - .init = snapshot_init, - .load = snapshot_load, - - .coldplug = snapshot_coldplug, - - .dump = snapshot_dump, - - .start = snapshot_start, - .stop = snapshot_stop, - - .serialize = snapshot_serialize, - .deserialize_item = snapshot_deserialize_item, - - .active_state = snapshot_active_state, - .sub_state_to_string = snapshot_sub_state_to_string, - - .bus_interface = "org.freedesktop.systemd1.Snapshot", - .bus_vtable = bus_snapshot_vtable -}; diff --git a/src/core/snapshot.h b/src/core/snapshot.h deleted file mode 100644 index f2451b1193..0000000000 --- a/src/core/snapshot.h +++ /dev/null @@ -1,48 +0,0 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - -#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 Snapshot Snapshot; - - -typedef enum SnapshotState { - SNAPSHOT_DEAD, - SNAPSHOT_ACTIVE, - _SNAPSHOT_STATE_MAX, - _SNAPSHOT_STATE_INVALID = -1 -} SnapshotState; - -struct Snapshot { - Unit meta; - - SnapshotState state, deserialized_state; - - bool cleanup; -}; - -extern const UnitVTable snapshot_vtable; - -int snapshot_create(Manager *m, const char *name, bool cleanup, sd_bus_error *e, Snapshot **s); -void snapshot_remove(Snapshot *s); - -const char* snapshot_state_to_string(SnapshotState i) _const_; -SnapshotState snapshot_state_from_string(const char *s) _pure_; diff --git a/src/core/socket.c b/src/core/socket.c index 693cbc6080..0b1c4acfec 100644 --- a/src/core/socket.c +++ b/src/core/socket.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,37 +17,54 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <sys/stat.h> -#include <unistd.h> +#include <arpa/inet.h> #include <errno.h> #include <fcntl.h> -#include <sys/epoll.h> -#include <signal.h> -#include <arpa/inet.h> -#include <netinet/tcp.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 "sd-event.h" +#include "alloc-util.h" +#include "bus-error.h" +#include "bus-util.h" +#include "copy.h" +#include "dbus-socket.h" +#include "def.h" +#include "exit-status.h" +#include "fd-util.h" +#include "formats-util.h" +#include "io-util.h" +#include "label.h" #include "log.h" -#include "strv.h" +#include "missing.h" #include "mkdir.h" +#include "parse-util.h" #include "path-util.h" -#include "unit-name.h" -#include "unit-printf.h" -#include "missing.h" -#include "special.h" -#include "label.h" -#include "exit-status.h" -#include "def.h" -#include "smack-util.h" -#include "bus-util.h" -#include "bus-error.h" +#include "process-util.h" #include "selinux-util.h" -#include "dbus-socket.h" -#include "unit.h" -#include "formats-util.h" #include "signal-util.h" +#include "smack-util.h" #include "socket.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "unit-name.h" +#include "unit-printf.h" +#include "unit.h" +#include "user-util.h" +#include "in-addr-util.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, @@ -92,6 +107,9 @@ static void socket_init(Unit *u) { 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) { @@ -104,6 +122,14 @@ static void socket_unwatch_control_pid(Socket *s) { 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; @@ -114,6 +140,7 @@ void socket_free_ports(Socket *s) { sd_event_source_unref(p->event_source); + socket_cleanup_fd_list(p); safe_close(p->fd); free(p->path); free(p); @@ -122,60 +149,65 @@ void socket_free_ports(Socket *s) { 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); - free(s->tcp_congestion); - s->tcp_congestion = NULL; - - free(s->bind_to_device); - s->bind_to_device = NULL; + s->tcp_congestion = mfree(s->tcp_congestion); + s->bind_to_device = mfree(s->bind_to_device); - free(s->smack); - free(s->smack_ip_in); - free(s->smack_ip_out); + 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); - free(s->user); - free(s->group); + 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) { +static int socket_arm_timer(Socket *s, usec_t usec) { int r; assert(s); - if (s->timeout_usec <= 0) { - s->timer_event_source = sd_event_source_unref(s->timer_event_source); - return 0; - } - if (s->timer_event_source) { - r = sd_event_source_set_time(s->timer_event_source, now(CLOCK_MONOTONIC) + s->timeout_usec); + 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, - now(CLOCK_MONOTONIC) + s->timeout_usec, 0, + usec, 0, socket_dispatch_timer, s); if (r < 0) return r; @@ -214,7 +246,6 @@ int socket_instantiate_service(Socket *s) { if (r < 0) return r; - u->no_gc = true; unit_ref_set(&s->service, u); return unit_add_two_dependencies(UNIT(s), UNIT_BEFORE, UNIT_TRIGGERS, u, false); @@ -251,7 +282,7 @@ static int socket_add_mount_links(Socket *s) { if (p->type == SOCKET_SOCKET) path = socket_address_get_path(&p->address); - else if (p->type == SOCKET_FIFO || p->type == SOCKET_SPECIAL) + else if (IN_SET(p->type, SOCKET_FIFO, SOCKET_SPECIAL, SOCKET_USB_FUNCTION)) path = p->path; if (!path) @@ -274,18 +305,21 @@ static int socket_add_device_link(Socket *s) { return 0; t = strjoina("/sys/subsystem/net/devices/", s->bind_to_device); - return unit_add_node_link(UNIT(s), t, false); + 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 (UNIT(s)->manager->running_as == MANAGER_SYSTEM) { + 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; @@ -311,6 +345,25 @@ static int socket_add_extras(Socket *s) { 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)) { @@ -345,16 +398,14 @@ static int socket_add_extras(Socket *s) { if (r < 0) return r; - r = unit_add_default_slice(u, &s->cgroup_context); + r = unit_set_default_slice(u); if (r < 0) return r; } - if (u->default_dependencies) { - r = socket_add_default_dependencies(s); - if (r < 0) - return r; - } + r = socket_add_default_dependencies(s); + if (r < 0) + return r; return 0; } @@ -431,6 +482,40 @@ static int socket_verify(Socket *s) { 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; @@ -438,6 +523,10 @@ static int socket_load(Unit *u) { 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; @@ -452,6 +541,87 @@ static int socket_load(Unit *u) { 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) @@ -497,6 +667,8 @@ static void socket_dump(Unit *u, FILE *f, const char *prefix) { "%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), @@ -513,6 +685,8 @@ static void socket_dump(Unit *u, FILE *f, const char *prefix) { 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) @@ -601,8 +775,8 @@ static void socket_dump(Unit *u, FILE *f, const char *prefix) { if (!isempty(s->user) || !isempty(s->group)) fprintf(f, - "%sOwnerUser: %s\n" - "%sOwnerGroup: %s\n", + "%sSocketUser: %s\n" + "%sSocketGroup: %s\n", prefix, strna(s->user), prefix, strna(s->group)); @@ -633,7 +807,8 @@ static void socket_dump(Unit *u, FILE *f, const char *prefix) { int r; char *k = NULL; - if ((r = socket_address_print(&p->address, &k)) < 0) + r = socket_address_print(&p->address, &k); + if (r < 0) t = strerror(-r); else t = k; @@ -642,12 +817,20 @@ static void socket_dump(Unit *u, FILE *f, const char *prefix) { 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); @@ -682,16 +865,16 @@ static int instance_from_socket(int fd, unsigned nr, char **instance) { case AF_INET: { uint32_t - a = ntohl(local.in.sin_addr.s_addr), - b = ntohl(remote.in.sin_addr.s_addr); + 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, - ntohs(local.in.sin_port), + be16toh(local.in.sin_port), b >> 24, (b >> 16) & 0xFF, (b >> 8) & 0xFF, b & 0xFF, - ntohs(remote.in.sin_port)) < 0) + be16toh(remote.in.sin_port)) < 0) return -ENOMEM; break; @@ -712,9 +895,9 @@ static int instance_from_socket(int fd, unsigned nr, char **instance) { "%u-%u.%u.%u.%u:%u-%u.%u.%u.%u:%u", nr, a[0], a[1], a[2], a[3], - ntohs(local.in6.sin6_port), + be16toh(local.in6.sin6_port), b[0], b[1], b[2], b[3], - ntohs(remote.in6.sin6_port)) < 0) + be16toh(remote.in6.sin6_port)) < 0) return -ENOMEM; } else { char a[INET6_ADDRSTRLEN], b[INET6_ADDRSTRLEN]; @@ -723,9 +906,9 @@ static int instance_from_socket(int fd, unsigned nr, char **instance) { "%u-%s:%u-%s:%u", nr, inet_ntop(AF_INET6, &local.in6.sin6_addr, a, sizeof(a)), - ntohs(local.in6.sin6_port), + be16toh(local.in6.sin6_port), inet_ntop(AF_INET6, &remote.in6.sin6_addr, b, sizeof(b)), - ntohs(remote.in6.sin6_port)) < 0) + be16toh(remote.in6.sin6_port)) < 0) return -ENOMEM; } @@ -771,46 +954,45 @@ static void socket_close_fds(Socket *s) { assert(s); LIST_FOREACH(port, p, s->ports) { + bool was_open; - p->event_source = sd_event_source_unref(p->event_source); - - if (p->fd < 0) - continue; + 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 delete sockets in the file system - * before we create a new one, not after we stopped - * using one! */ + /* 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 (s->remove_on_stop) { - switch (p->type) { + if (!was_open || !s->remove_on_stop) + continue; - case SOCKET_FIFO: - unlink(p->path); - break; + switch (p->type) { - case SOCKET_MQUEUE: - mq_unlink(p->path); - break; + case SOCKET_FIFO: + (void) unlink(p->path); + break; - case SOCKET_SOCKET: - socket_address_unlink(&p->address); - break; + case SOCKET_MQUEUE: + (void) mq_unlink(p->path); + break; - default: - break; - } + case SOCKET_SOCKET: + (void) socket_address_unlink(&p->address); + break; + + default: + break; } } if (s->remove_on_stop) STRV_FOREACH(i, s->symlinks) - unlink(*i); + (void) unlink(*i); } static void socket_apply_socket_options(Socket *s, int fd) { @@ -839,7 +1021,7 @@ static void socket_apply_socket_options(Socket *s, int fd) { if (s->keep_alive_cnt) { int value = s->keep_alive_cnt; - if (setsockopt(fd, SOL_SOCKET, TCP_KEEPCNT, &value, sizeof(value)) < 0) + if (setsockopt(fd, SOL_TCP, TCP_KEEPCNT, &value, sizeof(value)) < 0) log_unit_warning_errno(UNIT(s), errno, "TCP_KEEPCNT failed: %m"); } @@ -851,8 +1033,14 @@ static void socket_apply_socket_options(Socket *s, int fd) { if (s->no_delay) { int b = s->no_delay; - if (setsockopt(fd, SOL_TCP, TCP_NODELAY, &b, sizeof(b)) < 0) - log_unit_warning_errno(UNIT(s), errno, "TCP_NODELAY failed: %m"); + + 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) { @@ -923,13 +1111,13 @@ static void socket_apply_socket_options(Socket *s, int fd) { log_unit_warning_errno(UNIT(s), errno, "TCP_CONGESTION failed: %m"); if (s->smack_ip_in) { - r = mac_smack_apply_ip_in_fd(fd, 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_ip_out_fd(fd, 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"); } @@ -943,49 +1131,49 @@ static void socket_apply_fifo_options(Socket *s, int fd) { if (s->pipe_size > 0) if (fcntl(fd, F_SETPIPE_SZ, s->pipe_size) < 0) - log_unit_warning_errno(UNIT(s), errno, "F_SETPIPE_SZ: %m"); + log_unit_warning_errno(UNIT(s), errno, "Setting pipe size failed, ignoring: %m"); if (s->smack) { - r = mac_smack_apply_fd(fd, s->smack); + r = mac_smack_apply_fd(fd, SMACK_ATTR_ACCESS, s->smack); if (r < 0) - log_unit_error_errno(UNIT(s), r, "mac_smack_apply_fd: %m"); + 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, - int *_fd) { + mode_t socket_mode) { - int fd = -1, r = 0; - struct stat st; + _cleanup_close_ int fd = -1; mode_t old_mask; + struct stat st; + int r; assert(path); - assert(_fd); mkdir_parents_label(path, directory_mode); r = mac_selinux_create_file_prepare(path, S_IFIFO); if (r < 0) - goto fail; + return r; /* Enforce the right access mode for the fifo */ old_mask = umask(~ socket_mode); /* Include the original umask in our mask */ - umask(~socket_mode | old_mask); + (void) umask(~socket_mode | old_mask); r = mkfifo(path, socket_mode); - umask(old_mask); + (void) umask(old_mask); if (r < 0 && errno != EEXIST) { r = -errno; goto fail; } - if ((fd = open(path, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW)) < 0) { + fd = open(path, O_RDWR | O_CLOEXEC | O_NOCTTY | O_NONBLOCK | O_NOFOLLOW); + if (fd < 0) { r = -errno; goto fail; } @@ -1001,53 +1189,64 @@ static int fifo_address_create( (st.st_mode & 0777) != (socket_mode & ~old_mask) || st.st_uid != getuid() || st.st_gid != getgid()) { - r = -EEXIST; goto fail; } - *_fd = fd; - return 0; + r = fd; + fd = -1; + + return r; fail: mac_selinux_create_file_clear(); - safe_close(fd); - return r; } -static int special_address_create( - const char *path, - int *_fd) { - - int fd = -1, r = 0; +static int special_address_create(const char *path, bool writable) { + _cleanup_close_ int fd = -1; struct stat st; + int r; assert(path); - assert(_fd); - fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NONBLOCK|O_NOFOLLOW); - if (fd < 0) { - r = -errno; - goto fail; - } + 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) { - r = -errno; - goto fail; - } + 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)) { - r = -EEXIST; - goto fail; - } + if (!S_ISREG(st.st_mode) && !S_ISCHR(st.st_mode)) + return -EEXIST; - *_fd = fd; - return 0; + r = fd; + fd = -1; -fail: - safe_close(fd); + 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; } @@ -1056,22 +1255,22 @@ static int mq_address_create( const char *path, mode_t mq_mode, long maxmsg, - long msgsize, - int *_fd) { + long msgsize) { - int fd = -1, r = 0; + _cleanup_close_ int fd = -1; struct stat st; mode_t old_mask; struct mq_attr _attr, *attr = NULL; + int r; assert(path); - assert(_fd); if (maxmsg > 0 && msgsize > 0) { - zero(_attr); - _attr.mq_flags = O_NONBLOCK; - _attr.mq_maxmsg = maxmsg; - _attr.mq_msgsize = msgsize; + _attr = (struct mq_attr) { + .mq_flags = O_NONBLOCK, + .mq_maxmsg = maxmsg, + .mq_msgsize = msgsize, + }; attr = &_attr; } @@ -1079,33 +1278,24 @@ static int mq_address_create( old_mask = umask(~ mq_mode); /* Include the original umask in our mask */ - umask(~mq_mode | old_mask); + (void) umask(~mq_mode | old_mask); fd = mq_open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_CREAT, mq_mode, attr); - umask(old_mask); + (void) umask(old_mask); - if (fd < 0) { - r = -errno; - goto fail; - } + if (fd < 0) + return -errno; - if (fstat(fd, &st) < 0) { - r = -errno; - goto fail; - } + if (fstat(fd, &st) < 0) + return -errno; if ((st.st_mode & 0777) != (mq_mode & ~old_mask) || st.st_uid != getuid() || - st.st_gid != getgid()) { + st.st_gid != getgid()) + return -EEXIST; - r = -EEXIST; - goto fail; - } + r = fd; + fd = -1; - *_fd = fd; - return 0; - -fail: - safe_close(fd); return r; } @@ -1125,11 +1315,112 @@ static int socket_symlink(Socket *s) { 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; - char *label = NULL; - bool know_label = false; assert(s); @@ -1138,39 +1429,36 @@ static int socket_open_fds(Socket *s) { if (p->fd >= 0) continue; - if (p->type == SOCKET_SOCKET) { + 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. */ - - if (s->selinux_context_from_net) { - /* Get it from the network label */ - - r = mac_selinux_get_our_label(&label); - if (r < 0 && r != -EOPNOTSUPP) - goto rollback; - - } else { - /* Get it from the executable we are about to start */ - - r = socket_instantiate_service(s); - if (r < 0) - goto rollback; - - if (UNIT_ISSET(s->service) && - SERVICE(UNIT_DEREF(s->service))->exec_command[SERVICE_EXEC_START]) { - r = mac_selinux_get_create_label_from_exe(SERVICE(UNIT_DEREF(s->service))->exec_command[SERVICE_EXEC_START]->path, &label); - if (r < 0 && r != -EPERM && r != -EOPNOTSUPP) - goto rollback; - } - } + /* 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, @@ -1189,49 +1477,75 @@ static int socket_open_fds(Socket *s) { p->fd = r; socket_apply_socket_options(s, p->fd); socket_symlink(s); + break; - } else if (p->type == SOCKET_SPECIAL) { + case SOCKET_SPECIAL: - r = special_address_create( - p->path, - &p->fd); - if (r < 0) + p->fd = special_address_create(p->path, s->writable); + if (p->fd < 0) { + r = p->fd; goto rollback; + } + break; - } else if (p->type == SOCKET_FIFO) { + case SOCKET_FIFO: - r = fifo_address_create( + p->fd = fifo_address_create( p->path, s->directory_mode, - s->socket_mode, - &p->fd); - if (r < 0) + s->socket_mode); + if (p->fd < 0) { + r = p->fd; goto rollback; + } socket_apply_fifo_options(s, p->fd); socket_symlink(s); + break; - } else if (p->type == SOCKET_MQUEUE) { + case SOCKET_MQUEUE: - r = mq_address_create( + p->fd = mq_address_create( p->path, s->socket_mode, s->mq_maxmsg, - s->mq_msgsize, - &p->fd); + 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; - } else + + r = usbffs_dispatch_eps(p); + if (r < 0) + goto rollback; + + break; + } + default: assert_not_reached("Unknown port type"); + } } - mac_selinux_free(label); return 0; rollback: socket_close_fds(s); - mac_selinux_free(label); - return r; } @@ -1285,6 +1599,34 @@ fail: 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); @@ -1338,7 +1680,9 @@ static int socket_coldplug(Unit *u) { if (s->deserialized_state == s->state) return 0; - if (IN_SET(s->deserialized_state, + 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, @@ -1349,14 +1693,11 @@ static int socket_coldplug(Unit *u) { SOCKET_FINAL_SIGTERM, SOCKET_FINAL_SIGKILL)) { - if (s->control_pid <= 0) - return -EBADMSG; - r = unit_watch_pid(UNIT(s), s->control_pid); if (r < 0) return r; - r = socket_arm_timer(s); + r = socket_arm_timer(s, usec_add(u->state_change_timestamp.monotonic, s->timeout_usec)); if (r < 0) return r; } @@ -1365,13 +1706,24 @@ static int socket_coldplug(Unit *u) { SOCKET_START_CHOWN, SOCKET_START_POST, SOCKET_LISTENING, - SOCKET_RUNNING, - SOCKET_STOP_PRE, - SOCKET_STOP_PRE_SIGTERM, - SOCKET_STOP_PRE_SIGKILL)) { - r = socket_open_fds(s); - if (r < 0) - return r; + 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) { @@ -1380,6 +1732,9 @@ static int socket_coldplug(Unit *u) { 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; } @@ -1389,10 +1744,10 @@ static int socket_spawn(Socket *s, ExecCommand *c, pid_t *_pid) { pid_t pid; int r; ExecParameters exec_params = { - .apply_permissions = true, - .apply_chroot = true, - .apply_tty_stdin = true, - .bus_endpoint_fd = -1, + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, }; assert(s); @@ -1407,19 +1762,23 @@ static int socket_spawn(Socket *s, ExecCommand *c, pid_t *_pid) { r = unit_setup_exec_runtime(UNIT(s)); if (r < 0) - goto fail; + return r; - r = socket_arm_timer(s); + r = unit_setup_dynamic_creds(UNIT(s)); if (r < 0) - goto fail; + 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) - goto fail; + return r; exec_params.argv = argv; exec_params.environment = UNIT(s)->manager->environment; - exec_params.confirm_spawn = UNIT(s)->manager->confirm_spawn; + 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; @@ -1430,28 +1789,25 @@ static int socket_spawn(Socket *s, ExecCommand *c, pid_t *_pid) { &s->exec_context, &exec_params, s->exec_runtime, + &s->dynamic_creds, &pid); if (r < 0) - goto fail; + return r; r = unit_watch_pid(UNIT(s), pid); if (r < 0) /* FIXME: we need to do something here */ - goto fail; + return r; *_pid = pid; return 0; - -fail: - s->timer_event_source = sd_event_source_unref(s->timer_event_source); - return r; } static int socket_chown(Socket *s, pid_t *_pid) { pid_t pid; int r; - r = socket_arm_timer(s); + r = socket_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); if (r < 0) goto fail; @@ -1534,15 +1890,19 @@ fail: static void socket_enter_dead(Socket *s, SocketResult f) { assert(s); - if (f != SOCKET_SUCCESS) + 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)); - socket_set_state(s, s->result != SOCKET_SUCCESS ? SOCKET_FAILED : SOCKET_DEAD); + unit_unref_uid_gid(UNIT(s), true); + + dynamic_creds_destroy(&s->dynamic_creds); } static void socket_enter_signal(Socket *s, SocketState state, SocketResult f); @@ -1551,7 +1911,7 @@ static void socket_enter_stop_post(Socket *s, SocketResult f) { int r; assert(s); - if (f != SOCKET_SUCCESS) + if (s->result == SOCKET_SUCCESS) s->result = f; socket_unwatch_control_pid(s); @@ -1579,7 +1939,7 @@ static void socket_enter_signal(Socket *s, SocketState state, SocketResult f) { assert(s); - if (f != SOCKET_SUCCESS) + if (s->result == SOCKET_SUCCESS) s->result = f; r = unit_kill_context( @@ -1594,7 +1954,7 @@ static void socket_enter_signal(Socket *s, SocketState state, SocketResult f) { goto fail; if (r > 0) { - r = socket_arm_timer(s); + r = socket_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); if (r < 0) goto fail; @@ -1623,7 +1983,7 @@ static void socket_enter_stop_pre(Socket *s, SocketResult f) { int r; assert(s); - if (f != SOCKET_SUCCESS) + if (s->result == SOCKET_SUCCESS) s->result = f; socket_unwatch_control_pid(s); @@ -1746,38 +2106,47 @@ 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_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _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 */ + /* 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) - safe_close(cfd); - else { - /* Flush all sockets by closing and reopening them */ - socket_close_fds(s); - - r = socket_open_fds(s); - if (r < 0) { - log_unit_warning_errno(UNIT(s), r, "Failed to listen on sockets: %m"); - socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); - return; - } + cfd = safe_close(cfd); + else + flush_ports(s); - r = socket_watch_fds(s); - if (r < 0) { - log_unit_warning_errno(UNIT(s), r, "Failed to watch sockets: %m"); - socket_enter_stop_pre(s, SOCKET_FAILURE_RESOURCES); - } - } + 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; } @@ -1801,7 +2170,7 @@ static void socket_enter_running(Socket *s, int cfd) { goto fail; } - r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT_DEREF(s->service), JOB_REPLACE, true, &error, NULL); + r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT_DEREF(s->service), JOB_REPLACE, &error, NULL); if (r < 0) goto fail; } @@ -1809,14 +2178,34 @@ static void socket_enter_running(Socket *s, int cfd) { 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)", s->n_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; @@ -1828,6 +2217,7 @@ static void socket_enter_running(Socket *s, int cfd) { /* 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; } @@ -1846,22 +2236,27 @@ static void socket_enter_running(Socket *s, int cfd) { service = SERVICE(UNIT_DEREF(s->service)); unit_ref_unset(&s->service); - s->n_accepted ++; - - UNIT(service)->no_gc = false; + 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; - s->n_connections ++; + cfd = -1; /* We passed ownership of the fd to the service now. Forget it here. */ + s->n_connections++; - r = manager_add_job(UNIT(s)->manager, JOB_START, UNIT(service), JOB_REPLACE, true, &error, NULL); - if (r < 0) + 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)); @@ -1908,6 +2303,7 @@ fail: static int socket_start(Unit *u) { Socket *s = SOCKET(u); + int r; assert(s); @@ -1952,11 +2348,20 @@ static int socket_start(Unit *u) { 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; } @@ -2036,6 +2441,8 @@ static int socket_serialize(Unit *u, FILE *f, FDSet *fds) { 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); @@ -2045,6 +2452,11 @@ static int socket_serialize(Unit *u, FILE *f, FDSet *fds) { 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); @@ -2099,18 +2511,13 @@ static int socket_deserialize_item(Unit *u, const char *key, const char *value, 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 { - + else LIST_FOREACH(port, p, s->ports) if (p->type == SOCKET_FIFO && - path_equal_or_files_same(p->path, value+skip)) + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); break; - - if (p) { - safe_close(p->fd); - p->fd = fdset_remove(fds, fd); - } - } + } } else if (streq(key, "special")) { int fd, skip = 0; @@ -2118,18 +2525,13 @@ static int socket_deserialize_item(Unit *u, const char *key, const char *value, 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 { - + else LIST_FOREACH(port, p, s->ports) if (p->type == SOCKET_SPECIAL && - path_equal_or_files_same(p->path, value+skip)) + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); break; - - if (p) { - safe_close(p->fd); - p->fd = fdset_remove(fds, fd); - } - } + } } else if (streq(key, "mqueue")) { int fd, skip = 0; @@ -2137,18 +2539,13 @@ static int socket_deserialize_item(Unit *u, const char *key, const char *value, 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 { - + else LIST_FOREACH(port, p, s->ports) if (p->type == SOCKET_MQUEUE && - streq(p->path, value+skip)) + streq(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); break; - - if (p) { - safe_close(p->fd); - p->fd = fdset_remove(fds, fd); - } - } + } } else if (streq(key, "socket")) { int fd, type, skip = 0; @@ -2156,17 +2553,12 @@ static int socket_deserialize_item(Unit *u, const char *key, const char *value, 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 { - + else LIST_FOREACH(port, p, s->ports) - if (socket_address_is(&p->address, value+skip, type)) + if (socket_address_is(&p->address, value+skip, type)) { + socket_port_take_fd(p, fds, fd); break; - - if (p) { - safe_close(p->fd); - p->fd = fdset_remove(fds, fd); - } - } + } } else if (streq(key, "netlink")) { int fd, skip = 0; @@ -2174,24 +2566,34 @@ static int socket_deserialize_item(Unit *u, const char *key, const char *value, 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 { + 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 (socket_address_is_netlink(&p->address, value+skip)) + if (p->type == SOCKET_USB_FUNCTION && + path_equal_or_files_same(p->path, value+skip)) { + socket_port_take_fd(p, fds, fd); break; + } - if (p) { - safe_close(p->fd); - p->fd = fdset_remove(fds, fd); - } - } } else log_unit_debug(UNIT(s), "Unknown serialization key: %s", key); return 0; } -static int socket_distribute_fds(Unit *u, FDSet *fds) { +static void socket_distribute_fds(Unit *u, FDSet *fds) { Socket *s = SOCKET(u); SocketPort *p; @@ -2215,8 +2617,6 @@ static int socket_distribute_fds(Unit *u, FDSet *fds) { } } } - - return 0; } _pure_ static UnitActiveState socket_active_state(Unit *u) { @@ -2267,6 +2667,9 @@ const char* socket_port_type_to_string(SocketPort *p) { case SOCKET_FIFO: return "FIFO"; + case SOCKET_USB_FUNCTION: + return "USBFunction"; + default: return NULL; } @@ -2298,7 +2701,6 @@ static int socket_dispatch_io(sd_event_source *source, int fd, uint32_t revents, 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; } @@ -2344,7 +2746,7 @@ static void socket_sigchld_event(Unit *u, pid_t pid, int code, int status) { s->control_pid = 0; - if (is_clean_exit(code, status, NULL)) + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) f = SOCKET_SUCCESS; else if (code == CLD_EXITED) f = SOCKET_FAILURE_EXIT_CODE; @@ -2366,7 +2768,7 @@ static void socket_sigchld_event(Unit *u, pid_t pid, int code, int status) { "Control process exited, code=%s status=%i", sigchld_code_to_string(code), status); - if (f != SOCKET_SUCCESS) + if (s->result == SOCKET_SUCCESS) s->result = f; if (s->control_command && @@ -2494,43 +2896,43 @@ static int socket_dispatch_timer(sd_event_source *source, usec_t usec, void *use return 0; } -int socket_collect_fds(Socket *s, int **fds, unsigned *n_fds) { - int *rfds; - unsigned rn_fds, k; +int socket_collect_fds(Socket *s, int **fds) { + int *rfds, k = 0, n = 0; SocketPort *p; assert(s); assert(fds); - assert(n_fds); /* Called from the service code for requesting our fds */ - rn_fds = 0; - LIST_FOREACH(port, p, s->ports) + LIST_FOREACH(port, p, s->ports) { if (p->fd >= 0) - rn_fds++; + n++; + n += p->n_auxiliary_fds; + } - if (rn_fds <= 0) { + if (n <= 0) { *fds = NULL; - *n_fds = 0; return 0; } - rfds = new(int, rn_fds); + rfds = new(int, n); if (!rfds) return -ENOMEM; - k = 0; - LIST_FOREACH(port, p, s->ports) + 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 == rn_fds); + assert(k == n); *fds = rfds; - *n_fds = rn_fds; - - return 0; + return n; } static void socket_reset_failed(Unit *u) { @@ -2544,23 +2946,6 @@ static void socket_reset_failed(Unit *u) { s->result = SOCKET_SUCCESS; } -static void socket_notify_service_dead(Socket *s, bool failed_permanent) { - assert(s); - - /* The service is dead. Dang! - * - * This is strictly for one-instance-for-all-connections - * services. */ - - if (s->state == SOCKET_RUNNING) { - log_unit_debug(UNIT(s), "Got notified about service death (failed permanently: %s)", yes_no(failed_permanent)); - if (failed_permanent) - socket_enter_stop_pre(s, SOCKET_FAILURE_SERVICE_FAILED_PERMANENT); - else - socket_enter_listening(s); - } -} - void socket_connection_unref(Socket *s) { assert(s); @@ -2577,34 +2962,39 @@ void socket_connection_unref(Socket *s) { static void socket_trigger_notify(Unit *u, Unit *other) { Socket *s = SOCKET(u); - Service *se; assert(u); assert(other); - /* Don't propagate state changes from the service if we are - already down or accepting connections */ - if ((s->state != SOCKET_RUNNING && - s->state != SOCKET_LISTENING) || - s->accept) + /* Filter out invocations with bogus state */ + if (other->load_state != UNIT_LOADED || other->type != UNIT_SERVICE) return; - if (other->load_state != UNIT_LOADED || - other->type != UNIT_SERVICE) + /* Don't propagate state changes from the service if we are already down */ + if (!IN_SET(s->state, SOCKET_RUNNING, SOCKET_LISTENING)) return; - se = SERVICE(other); + /* 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; + } - if (se->state == SERVICE_FAILED) - socket_notify_service_dead(s, se->result == SERVICE_FAILURE_START_LIMIT); + /* Don't propagate anything if there's still a job queued */ + if (other->job) + return; - if (se->state == SERVICE_DEAD || - se->state == SERVICE_FINAL_SIGTERM || - se->state == SERVICE_FINAL_SIGKILL || - se->state == SERVICE_AUTO_RESTART) - socket_notify_service_dead(s, false); + if (IN_SET(SERVICE(other)->state, + SERVICE_DEAD, SERVICE_FAILED, + SERVICE_FINAL_SIGTERM, SERVICE_FINAL_SIGKILL, + SERVICE_AUTO_RESTART)) + socket_enter_listening(s); - if (se->state == SERVICE_RUNNING) + if (SERVICE(other)->state == SERVICE_RUNNING) socket_set_state(s, SOCKET_RUNNING); } @@ -2612,37 +3002,44 @@ 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, uint64_t *timeout) { +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, timeout); + 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 const char* const socket_state_table[_SOCKET_STATE_MAX] = { - [SOCKET_DEAD] = "dead", - [SOCKET_START_PRE] = "start-pre", - [SOCKET_START_CHOWN] = "start-chown", - [SOCKET_START_POST] = "start-post", - [SOCKET_LISTENING] = "listening", - [SOCKET_RUNNING] = "running", - [SOCKET_STOP_PRE] = "stop-pre", - [SOCKET_STOP_PRE_SIGTERM] = "stop-pre-sigterm", - [SOCKET_STOP_PRE_SIGKILL] = "stop-pre-sigkill", - [SOCKET_STOP_POST] = "stop-post", - [SOCKET_FINAL_SIGTERM] = "final-sigterm", - [SOCKET_FINAL_SIGKILL] = "final-sigkill", - [SOCKET_FAILED] = "failed" -}; +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. */ -DEFINE_STRING_TABLE_LOOKUP(socket_state, SocketState); + 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", @@ -2661,7 +3058,9 @@ static const char* const socket_result_table[_SOCKET_RESULT_MAX] = { [SOCKET_FAILURE_EXIT_CODE] = "exit-code", [SOCKET_FAILURE_SIGNAL] = "signal", [SOCKET_FAILURE_CORE_DUMP] = "core-dump", - [SOCKET_FAILURE_SERVICE_FAILED_PERMANENT] = "service-failed-permanent" + [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); @@ -2672,6 +3071,7 @@ const UnitVTable socket_vtable = { .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" @@ -2709,7 +3109,8 @@ const UnitVTable socket_vtable = { .reset_failed = socket_reset_failed, - .bus_interface = "org.freedesktop.systemd1.Socket", + .control_pid = socket_control_pid, + .bus_vtable = bus_socket_vtable, .bus_set_property = bus_socket_set_property, .bus_commit_properties = bus_socket_commit_properties, @@ -2722,7 +3123,6 @@ const UnitVTable socket_vtable = { .finished_start_job = { [JOB_DONE] = "Listening on %s.", [JOB_FAILED] = "Failed to listen on %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", [JOB_TIMEOUT] = "Timed out starting %s.", }, .finished_stop_job = { diff --git a/src/core/socket.h b/src/core/socket.h index fa3ebdafa0..89f4664510 100644 --- a/src/core/socket.h +++ b/src/core/socket.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,28 +20,11 @@ ***/ typedef struct Socket Socket; +typedef struct SocketPeer SocketPeer; -#include "socket-util.h" #include "mount.h" #include "service.h" - -typedef enum SocketState { - SOCKET_DEAD, - SOCKET_START_PRE, - SOCKET_START_CHOWN, - SOCKET_START_POST, - SOCKET_LISTENING, - SOCKET_RUNNING, - SOCKET_STOP_PRE, - SOCKET_STOP_PRE_SIGTERM, - SOCKET_STOP_PRE_SIGKILL, - SOCKET_STOP_POST, - SOCKET_FINAL_SIGTERM, - SOCKET_FINAL_SIGKILL, - SOCKET_FAILED, - _SOCKET_STATE_MAX, - _SOCKET_STATE_INVALID = -1 -} SocketState; +#include "socket-util.h" typedef enum SocketExecCommand { SOCKET_EXEC_START_PRE, @@ -60,6 +41,7 @@ typedef enum SocketType { SOCKET_FIFO, SOCKET_SPECIAL, SOCKET_MQUEUE, + SOCKET_USB_FUNCTION, _SOCKET_FIFO_MAX, _SOCKET_FIFO_INVALID = -1 } SocketType; @@ -71,7 +53,9 @@ typedef enum SocketResult { SOCKET_FAILURE_EXIT_CODE, SOCKET_FAILURE_SIGNAL, SOCKET_FAILURE_CORE_DUMP, - SOCKET_FAILURE_SERVICE_FAILED_PERMANENT, + SOCKET_FAILURE_START_LIMIT_HIT, + SOCKET_FAILURE_TRIGGER_LIMIT_HIT, + SOCKET_FAILURE_SERVICE_START_LIMIT_HIT, _SOCKET_RESULT_MAX, _SOCKET_RESULT_INVALID = -1 } SocketResult; @@ -81,6 +65,8 @@ typedef struct SocketPort { SocketType type; int fd; + int *auxiliary_fds; + int n_auxiliary_fds; SocketAddress address; char *path; @@ -94,9 +80,12 @@ struct Socket { 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; @@ -109,7 +98,9 @@ struct Socket { 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 @@ -133,6 +124,9 @@ struct Socket { bool accept; bool remove_on_stop; + bool writable; + + int socket_protocol; /* Socket options */ bool keep_alive; @@ -168,20 +162,31 @@ struct Socket { 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, unsigned *n_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); -extern const UnitVTable socket_vtable; +int socket_instantiate_service(Socket *s); -const char* socket_state_to_string(SocketState i) _const_; -SocketState socket_state_from_string(const char *s) _pure_; +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_; @@ -190,5 +195,3 @@ 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_; - -int socket_instantiate_service(Socket *s); diff --git a/src/core/swap.c b/src/core/swap.c index 193c8c3767..b592abb9fb 100644 --- a/src/core/swap.c +++ b/src/core/swap.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,22 +18,30 @@ ***/ #include <errno.h> -#include <unistd.h> #include <sys/epoll.h> #include <sys/stat.h> -#include <libudev.h> +#include <unistd.h> -#include "unit.h" -#include "swap.h" -#include "unit-name.h" +#include "libudev.h" + +#include "alloc-util.h" #include "dbus-swap.h" -#include "special.h" +#include "escape.h" #include "exit-status.h" +#include "fd-util.h" +#include "formats-util.h" +#include "fstab-util.h" +#include "parse-util.h" #include "path-util.h" -#include "virt.h" +#include "process-util.h" +#include "special.h" +#include "string-table.h" +#include "string-util.h" +#include "swap.h" #include "udev-util.h" -#include "fstab-util.h" -#include "formats-util.h" +#include "unit-name.h" +#include "unit.h" +#include "virt.h" static const UnitActiveState state_translation_table[_SWAP_STATE_MAX] = { [SWAP_DEAD] = UNIT_INACTIVE, @@ -59,8 +65,7 @@ static void swap_unset_proc_swaps(Swap *s) { if (!s->from_proc_swaps) return; - free(s->parameters_proc_swaps.what); - s->parameters_proc_swaps.what = NULL; + s->parameters_proc_swaps.what = mfree(s->parameters_proc_swaps.what); s->from_proc_swaps = false; } @@ -87,8 +92,7 @@ static int swap_set_devnode(Swap *s, const char *devnode) { else hashmap_remove(swaps, s->devnode); - free(s->devnode); - s->devnode = NULL; + s->devnode = mfree(s->devnode); } if (devnode) { @@ -141,47 +145,42 @@ static void swap_done(Unit *u) { swap_unset_proc_swaps(s); swap_set_devnode(s, NULL); - free(s->what); - s->what = NULL; - - free(s->parameters_fragment.what); - s->parameters_fragment.what = NULL; - - free(s->parameters_fragment.options); - s->parameters_fragment.options = 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) { +static int swap_arm_timer(Swap *s, usec_t usec) { int r; assert(s); - if (s->timeout_usec <= 0) { - s->timer_event_source = sd_event_source_unref(s->timer_event_source); - return 0; - } - if (s->timer_event_source) { - r = sd_event_source_set_time(s->timer_event_source, now(CLOCK_MONOTONIC) + s->timeout_usec); + 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, - now(CLOCK_MONOTONIC) + s->timeout_usec, 0, + usec, 0, swap_dispatch_timer, s); if (r < 0) return r; @@ -201,7 +200,7 @@ static int swap_add_device_links(Swap *s) { return 0; if (is_device_path(s->what)) - return unit_add_node_link(UNIT(s), s->what, UNIT(s)->manager->running_as == MANAGER_SYSTEM); + 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 @@ -210,14 +209,25 @@ static int swap_add_device_links(Swap *s) { } static int swap_add_default_dependencies(Swap *s) { + int r; + assert(s); - if (UNIT(s)->manager->running_as != MANAGER_SYSTEM) + if (!UNIT(s)->default_dependencies) return 0; - if (detect_container(NULL) > 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); } @@ -326,15 +336,13 @@ static int swap_load(Unit *u) { if (r < 0) return r; - r = unit_add_default_slice(u, &s->cgroup_context); + r = unit_set_default_slice(u); if (r < 0) return r; - if (UNIT(s)->default_dependencies) { - r = swap_add_default_dependencies(s); - if (r < 0) - return r; - } + r = swap_add_default_dependencies(s); + if (r < 0) + return r; } return swap_verify(s); @@ -527,26 +535,29 @@ static int swap_coldplug(Unit *u) { if (new_state == s->state) return 0; - if (new_state == SWAP_ACTIVATING || - new_state == SWAP_ACTIVATING_SIGTERM || - new_state == SWAP_ACTIVATING_SIGKILL || - new_state == SWAP_ACTIVATING_DONE || - new_state == SWAP_DEACTIVATING || - new_state == SWAP_DEACTIVATING_SIGTERM || - new_state == SWAP_DEACTIVATING_SIGKILL) { - - if (s->control_pid <= 0) - return -EBADMSG; + 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); + 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; } @@ -600,10 +611,10 @@ static int swap_spawn(Swap *s, ExecCommand *c, pid_t *_pid) { pid_t pid; int r; ExecParameters exec_params = { - .apply_permissions = true, - .apply_chroot = true, - .apply_tty_stdin = true, - .bus_endpoint_fd = -1, + .flags = EXEC_APPLY_PERMISSIONS|EXEC_APPLY_CHROOT|EXEC_APPLY_TTY_STDIN, + .stdin_fd = -1, + .stdout_fd = -1, + .stderr_fd = -1, }; assert(s); @@ -620,12 +631,16 @@ static int swap_spawn(Swap *s, ExecCommand *c, pid_t *_pid) { if (r < 0) goto fail; - r = swap_arm_timer(s); + 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.confirm_spawn = UNIT(s)->manager->confirm_spawn; + 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; @@ -636,6 +651,7 @@ static int swap_spawn(Swap *s, ExecCommand *c, pid_t *_pid) { &s->exec_context, &exec_params, s->exec_runtime, + &s->dynamic_creds, &pid); if (r < 0) goto fail; @@ -657,21 +673,25 @@ fail: static void swap_enter_dead(Swap *s, SwapResult f) { assert(s); - if (f != SWAP_SUCCESS) + 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)); - swap_set_state(s, s->result != SWAP_SUCCESS ? SWAP_FAILED : SWAP_DEAD); + 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 (f != SWAP_SUCCESS) + if (s->result == SWAP_SUCCESS) s->result = f; swap_set_state(s, SWAP_ACTIVE); @@ -682,7 +702,7 @@ static void swap_enter_signal(Swap *s, SwapState state, SwapResult f) { assert(s); - if (f != SWAP_SUCCESS) + if (s->result == SWAP_SUCCESS) s->result = f; r = unit_kill_context( @@ -697,7 +717,7 @@ static void swap_enter_signal(Swap *s, SwapState state, SwapResult f) { goto fail; if (r > 0) { - r = swap_arm_timer(s); + r = swap_arm_timer(s, usec_add(now(CLOCK_MONOTONIC), s->timeout_usec)); if (r < 0) goto fail; @@ -806,6 +826,7 @@ fail: static int swap_start(Unit *u) { Swap *s = SWAP(u), *other; + int r; assert(s); @@ -824,7 +845,7 @@ static int swap_start(Unit *u) { assert(s->state == SWAP_DEAD || s->state == SWAP_FAILED); - if (detect_container(NULL) > 0) + if (detect_container() > 0) return -EPERM; /* If there's a job for another swap unit for the same node @@ -834,6 +855,16 @@ static int swap_start(Unit *u) { 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; @@ -857,7 +888,7 @@ static int swap_stop(Unit *u) { s->state == SWAP_ACTIVATING_DONE || s->state == SWAP_ACTIVE); - if (detect_container(NULL) > 0) + if (detect_container() > 0) return -EPERM; swap_enter_deactivating(s); @@ -961,7 +992,7 @@ static void swap_sigchld_event(Unit *u, pid_t pid, int code, int status) { s->control_pid = 0; - if (is_clean_exit(code, status, NULL)) + if (is_clean_exit(code, status, EXIT_CLEAN_COMMAND, NULL)) f = SWAP_SUCCESS; else if (code == CLD_EXITED) f = SWAP_FAILURE_EXIT_CODE; @@ -972,7 +1003,7 @@ static void swap_sigchld_event(Unit *u, pid_t pid, int code, int status) { else assert_not_reached("Unknown code"); - if (f != SWAP_SUCCESS) + if (s->result == SWAP_SUCCESS) s->result = f; if (s->control_command) { @@ -1162,6 +1193,7 @@ static int swap_dispatch_io(sd_event_source *source, int fd, uint32_t revents, v case SWAP_DEAD: case SWAP_FAILED: + (void) unit_acquire_invocation_id(UNIT(swap)); swap_enter_active(swap, SWAP_SUCCESS); break; @@ -1202,7 +1234,7 @@ static Unit *swap_following(Unit *u) { if (other->from_fragment) return UNIT(other); - /* Otherwise make everybody follow the unit that's named after + /* Otherwise, make everybody follow the unit that's named after * the swap device in the kernel */ if (streq_ptr(s->what, s->devnode)) @@ -1259,35 +1291,41 @@ static void swap_shutdown(Manager *m) { m->swap_event_source = sd_event_source_unref(m->swap_event_source); - if (m->proc_swaps) { - fclose(m->proc_swaps); - m->proc_swaps = NULL; - } + m->proc_swaps = safe_fclose(m->proc_swaps); - hashmap_free(m->swaps_by_devnode); - m->swaps_by_devnode = NULL; + m->swaps_by_devnode = hashmap_free(m->swaps_by_devnode); } -static int swap_enumerate(Manager *m) { +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) - return errno == ENOENT ? 0 : -errno; + 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) + 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) + 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"); } @@ -1296,11 +1334,10 @@ static int swap_enumerate(Manager *m) { if (r < 0) goto fail; - return 0; + return; fail: swap_shutdown(m); - return r; } int swap_process_device_new(Manager *m, struct udev_device *dev) { @@ -1380,17 +1417,21 @@ 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, uint64_t *timeout) { +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, timeout); + 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; } @@ -1404,25 +1445,18 @@ static bool swap_supported(void) { if (supported < 0) supported = access("/proc/swaps", F_OK) >= 0 && - detect_container(NULL) <= 0; + detect_container() <= 0; return supported; } -static const char* const swap_state_table[_SWAP_STATE_MAX] = { - [SWAP_DEAD] = "dead", - [SWAP_ACTIVATING] = "activating", - [SWAP_ACTIVATING_DONE] = "activating-done", - [SWAP_ACTIVE] = "active", - [SWAP_DEACTIVATING] = "deactivating", - [SWAP_ACTIVATING_SIGTERM] = "activating-sigterm", - [SWAP_ACTIVATING_SIGKILL] = "activating-sigkill", - [SWAP_DEACTIVATING_SIGTERM] = "deactivating-sigterm", - [SWAP_DEACTIVATING_SIGKILL] = "deactivating-sigkill", - [SWAP_FAILED] = "failed" -}; +static int swap_control_pid(Unit *u) { + Swap *s = SWAP(u); -DEFINE_STRING_TABLE_LOOKUP(swap_state, SwapState); + assert(s); + + return s->control_pid; +} static const char* const swap_exec_command_table[_SWAP_EXEC_COMMAND_MAX] = { [SWAP_EXEC_ACTIVATE] = "ExecActivate", @@ -1437,7 +1471,8 @@ static const char* const swap_result_table[_SWAP_RESULT_MAX] = { [SWAP_FAILURE_TIMEOUT] = "timeout", [SWAP_FAILURE_EXIT_CODE] = "exit-code", [SWAP_FAILURE_SIGNAL] = "signal", - [SWAP_FAILURE_CORE_DUMP] = "core-dump" + [SWAP_FAILURE_CORE_DUMP] = "core-dump", + [SWAP_FAILURE_START_LIMIT_HIT] = "start-limit-hit", }; DEFINE_STRING_TABLE_LOOKUP(swap_result, SwapResult); @@ -1448,6 +1483,7 @@ const UnitVTable swap_vtable = { .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" @@ -1455,9 +1491,6 @@ const UnitVTable swap_vtable = { "Install\0", .private_section = "Swap", - .no_alias = true, - .no_instances = true, - .init = swap_init, .load = swap_load, .done = swap_done, @@ -1485,7 +1518,8 @@ const UnitVTable swap_vtable = { .reset_failed = swap_reset_failed, - .bus_interface = "org.freedesktop.systemd1.Swap", + .control_pid = swap_control_pid, + .bus_vtable = bus_swap_vtable, .bus_set_property = bus_swap_set_property, .bus_commit_properties = bus_swap_commit_properties, @@ -1505,7 +1539,6 @@ const UnitVTable swap_vtable = { .finished_start_job = { [JOB_DONE] = "Activated swap %s.", [JOB_FAILED] = "Failed to activate swap %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", [JOB_TIMEOUT] = "Timed out activating swap %s.", }, .finished_stop_job = { diff --git a/src/core/swap.h b/src/core/swap.h index 9136b9abab..b0ef50f1e8 100644 --- a/src/core/swap.h +++ b/src/core/swap.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -22,26 +20,10 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <libudev.h> +#include "libudev.h" typedef struct Swap Swap; - -typedef enum SwapState { - SWAP_DEAD, - SWAP_ACTIVATING, /* /sbin/swapon is running, but the swap not yet enabled. */ - SWAP_ACTIVATING_DONE, /* /sbin/swapon is running, and the swap is done. */ - SWAP_ACTIVE, - SWAP_DEACTIVATING, - SWAP_ACTIVATING_SIGTERM, - SWAP_ACTIVATING_SIGKILL, - SWAP_DEACTIVATING_SIGTERM, - SWAP_DEACTIVATING_SIGKILL, - SWAP_FAILED, - _SWAP_STATE_MAX, - _SWAP_STATE_INVALID = -1 -} SwapState; - typedef enum SwapExecCommand { SWAP_EXEC_ACTIVATE, SWAP_EXEC_DEACTIVATE, @@ -56,6 +38,7 @@ typedef enum SwapResult { SWAP_FAILURE_EXIT_CODE, SWAP_FAILURE_SIGNAL, SWAP_FAILURE_CORE_DUMP, + SWAP_FAILURE_START_LIMIT_HIT, _SWAP_RESULT_MAX, _SWAP_RESULT_INVALID = -1 } SwapResult; @@ -99,6 +82,7 @@ struct Swap { CGroupContext cgroup_context; ExecRuntime *exec_runtime; + DynamicCreds dynamic_creds; SwapState state, deserialized_state; @@ -120,9 +104,6 @@ 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_state_to_string(SwapState i) _const_; -SwapState swap_state_from_string(const char *s) _pure_; - const char* swap_exec_command_to_string(SwapExecCommand i) _const_; SwapExecCommand swap_exec_command_from_string(const char *s) _pure_; diff --git a/src/core/system.conf b/src/core/system.conf index 231609033b..746572b7ff 100644 --- a/src/core/system.conf +++ b/src/core/system.conf @@ -17,9 +17,11 @@ #LogColor=yes #LogLocation=no #DumpCore=yes -#CrashShell=no #ShowStatus=yes -#CrashChVT=1 +#CrashChangeVT=no +#CrashShell=no +#CrashReboot=no +#CtrlAltDelBurstAction=reboot-force #CPUAffinity=1 2 #JoinControllers=cpu,cpuacct net_cls,net_prio #RuntimeWatchdogSec=0 @@ -33,12 +35,15 @@ #DefaultTimeoutStartSec=90s #DefaultTimeoutStopSec=90s #DefaultRestartSec=100ms -#DefaultStartLimitInterval=10s +#DefaultStartLimitIntervalSec=10s #DefaultStartLimitBurst=5 #DefaultEnvironment= #DefaultCPUAccounting=no +#DefaultIOAccounting=no #DefaultBlockIOAccounting=no #DefaultMemoryAccounting=no +#DefaultTasksAccounting=yes +#DefaultTasksMax=15% #DefaultLimitCPU= #DefaultLimitFSIZE= #DefaultLimitDATA= diff --git a/src/core/target.c b/src/core/target.c index 8817ef21c4..765c1f3fa4 100644 --- a/src/core/target.c +++ b/src/core/target.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,13 +17,13 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ - -#include "unit.h" -#include "target.h" -#include "log.h" #include "dbus-target.h" +#include "log.h" #include "special.h" +#include "string-util.h" #include "unit-name.h" +#include "unit.h" +#include "target.h" static const UnitActiveState state_translation_table[_TARGET_STATE_MAX] = { [TARGET_DEAD] = UNIT_INACTIVE, @@ -52,9 +50,7 @@ static int target_add_default_dependencies(Target *t) { static const UnitDependency deps[] = { UNIT_REQUIRES, - UNIT_REQUIRES_OVERRIDABLE, UNIT_REQUISITE, - UNIT_REQUISITE_OVERRIDABLE, UNIT_WANTS, UNIT_BINDS_TO, UNIT_PART_OF @@ -128,10 +124,15 @@ static void target_dump(Unit *u, FILE *f, const char *prefix) { 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; } @@ -192,13 +193,6 @@ _pure_ static const char *target_sub_state_to_string(Unit *u) { return target_state_to_string(TARGET(u)->state); } -static const char* const target_state_table[_TARGET_STATE_MAX] = { - [TARGET_DEAD] = "dead", - [TARGET_ACTIVE] = "active" -}; - -DEFINE_STRING_TABLE_LOOKUP(target_state, TargetState); - const UnitVTable target_vtable = { .object_size = sizeof(Target), @@ -221,13 +215,11 @@ const UnitVTable target_vtable = { .active_state = target_active_state, .sub_state_to_string = target_sub_state_to_string, - .bus_interface = "org.freedesktop.systemd1.Target", .bus_vtable = bus_target_vtable, .status_message_formats = { .finished_start_job = { [JOB_DONE] = "Reached target %s.", - [JOB_DEPENDENCY] = "Dependency failed for %s.", }, .finished_stop_job = { [JOB_DONE] = "Stopped target %s.", diff --git a/src/core/target.h b/src/core/target.h index 0a25ef469a..339aea154e 100644 --- a/src/core/target.h +++ b/src/core/target.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,14 +21,6 @@ typedef struct Target Target; - -typedef enum TargetState { - TARGET_DEAD, - TARGET_ACTIVE, - _TARGET_STATE_MAX, - _TARGET_STATE_INVALID = -1 -} TargetState; - struct Target { Unit meta; @@ -38,6 +28,3 @@ struct Target { }; extern const UnitVTable target_vtable; - -const char* target_state_to_string(TargetState i) _const_; -TargetState target_state_from_string(const char *s) _pure_; diff --git a/src/core/timer.c b/src/core/timer.c index 7f4a2eb716..9538059c13 100644 --- a/src/core/timer.c +++ b/src/core/timer.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,13 +19,21 @@ #include <errno.h> -#include "unit.h" -#include "unit-name.h" -#include "timer.h" +#include "alloc-util.h" +#include "bus-error.h" +#include "bus-util.h" #include "dbus-timer.h" +#include "fs-util.h" +#include "parse-util.h" +#include "random-util.h" #include "special.h" -#include "bus-util.h" -#include "bus-error.h" +#include "string-table.h" +#include "string-util.h" +#include "timer.h" +#include "unit-name.h" +#include "unit.h" +#include "user-util.h" +#include "virt.h" static const UnitActiveState state_translation_table[_TIMER_STATE_MAX] = { [TIMER_DEAD] = UNIT_INACTIVE, @@ -48,6 +54,7 @@ static void timer_init(Unit *u) { 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) { @@ -57,10 +64,7 @@ void timer_free_values(Timer *t) { while ((v = t->values)) { LIST_REMOVE(value, t->values, v); - - if (v->calendar_spec) - calendar_spec_free(v->calendar_spec); - + calendar_spec_free(v->calendar_spec); free(v); } } @@ -98,11 +102,14 @@ static int timer_add_default_dependencies(Timer *t) { 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 (UNIT(t)->manager->running_as == MANAGER_SYSTEM) { + 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; @@ -128,7 +135,7 @@ static int timer_setup_persistent(Timer *t) { if (!t->persistent) return 0; - if (UNIT(t)->manager->running_as == MANAGER_SYSTEM) { + if (MANAGER_IS_SYSTEM(UNIT(t)->manager)) { r = unit_require_mounts_for(UNIT(t), "/var/lib/systemd/timers"); if (r < 0) @@ -188,11 +195,9 @@ static int timer_load(Unit *u) { if (r < 0) return r; - if (u->default_dependencies) { - r = timer_add_default_dependencies(t); - if (r < 0) - return r; - } + r = timer_add_default_dependencies(t); + if (r < 0) + return r; } return timer_verify(t); @@ -212,13 +217,15 @@ static void timer_dump(Unit *u, FILE *f, const char *prefix) { "%sUnit: %s\n" "%sPersistent: %s\n" "%sWakeSystem: %s\n" - "%sAccuracy: %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, format_timespan(buf, sizeof(buf), t->accuracy_usec, 1), + prefix, yes_no(t->remain_after_elapse)); LIST_FOREACH(value, v, t->values) { @@ -270,13 +277,13 @@ static int timer_coldplug(Unit *u) { assert(t); assert(t->state == TIMER_DEAD); - if (t->deserialized_state != t->state) { + 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); - } + if (t->deserialized_state == TIMER_WAITING) + timer_enter_waiting(t, false); + else + timer_set_state(t, t->deserialized_state); return 0; } @@ -284,19 +291,36 @@ static int timer_coldplug(Unit *u) { static void timer_enter_dead(Timer *t, TimerResult f) { assert(t); - if (f != TIMER_SUCCESS) + 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); + a = now(clock_boottime_or_monotonic()); b = now(CLOCK_MONOTONIC); if (t + a > b) @@ -305,18 +329,51 @@ static usec_t monotonic_to_boottime(usec_t t) { 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 : CLOCK_MONOTONIC); + 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) { @@ -356,17 +413,21 @@ static void timer_enter_waiting(Timer *t, bool initial) { break; case TIMER_BOOT: - /* CLOCK_MONOTONIC equals the uptime on Linux */ - base = 0; - break; - + 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: - - base = UNIT_TRIGGER(UNIT(t))->inactive_exit_timestamp.monotonic; + leave_around = true; + base = trigger->inactive_exit_timestamp.monotonic; if (base <= 0) base = t->last_trigger.monotonic; @@ -377,8 +438,8 @@ static void timer_enter_waiting(Timer *t, bool initial) { break; case TIMER_UNIT_INACTIVE: - - base = UNIT_TRIGGER(UNIT(t))->inactive_enter_timestamp.monotonic; + leave_around = true; + base = trigger->inactive_enter_timestamp.monotonic; if (base <= 0) base = t->last_trigger.monotonic; @@ -414,14 +475,18 @@ static void timer_enter_waiting(Timer *t, bool initial) { if (!found_monotonic && !found_realtime) { log_unit_debug(UNIT(t), "Timer is elapsed."); - timer_set_state(t, TIMER_ELAPSED); + timer_enter_elapsed(t, leave_around); return; } if (found_monotonic) { char buf[FORMAT_TIMESPAN_MAX]; + usec_t left; - log_unit_debug(UNIT(t), "Monotonic timer elapses in %s.", format_timespan(buf, sizeof(buf), t->next_elapse_monotonic_or_boottime > ts_monotonic ? t->next_elapse_monotonic_or_boottime - ts_monotonic : 0, 0)); + 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); @@ -454,6 +519,9 @@ static void timer_enter_waiting(Timer *t, bool initial) { 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) { @@ -493,7 +561,8 @@ fail: } static void timer_enter_running(Timer *t) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; + Unit *trigger; int r; assert(t); @@ -502,15 +571,21 @@ static void timer_enter_running(Timer *t) { if (unit_stop_pending(UNIT(t))) return; - r = manager_add_job(UNIT(t)->manager, JOB_START, UNIT_TRIGGER(UNIT(t)), - JOB_REPLACE, true, &error, NULL); + 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, 0); + touch_file(t->stamp_path, true, t->last_trigger.realtime, UID_INVALID, GID_INVALID, MODE_INVALID); timer_set_state(t, TIMER_RUNNING); return; @@ -523,12 +598,27 @@ fail: 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); - if (UNIT_TRIGGER(u)->load_state != UNIT_LOADED) + 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; @@ -546,7 +636,7 @@ static int timer_start(Unit *u) { /* The timer has never run before, * make sure a stamp file exists. */ - touch_file(t->stamp_path, true, USEC_INFINITY, UID_INVALID, GID_INVALID, 0); + (void) touch_file(t->stamp_path, true, USEC_INFINITY, UID_INVALID, GID_INVALID, MODE_INVALID); } t->result = TIMER_SUCCESS; @@ -716,16 +806,6 @@ static void timer_time_change(Unit *u) { timer_enter_waiting(t, false); } -static const char* const timer_state_table[_TIMER_STATE_MAX] = { - [TIMER_DEAD] = "dead", - [TIMER_WAITING] = "waiting", - [TIMER_RUNNING] = "running", - [TIMER_ELAPSED] = "elapsed", - [TIMER_FAILED] = "failed" -}; - -DEFINE_STRING_TABLE_LOOKUP(timer_state, TimerState); - static const char* const timer_base_table[_TIMER_BASE_MAX] = { [TIMER_ACTIVE] = "OnActiveSec", [TIMER_BOOT] = "OnBootSec", @@ -739,7 +819,8 @@ 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_RESOURCES] = "resources", + [TIMER_FAILURE_START_LIMIT_HIT] = "start-limit-hit", }; DEFINE_STRING_TABLE_LOOKUP(timer_result, TimerResult); @@ -775,7 +856,6 @@ const UnitVTable timer_vtable = { .reset_failed = timer_reset_failed, .time_change = timer_time_change, - .bus_interface = "org.freedesktop.systemd1.Timer", .bus_vtable = bus_timer_vtable, .bus_set_property = bus_timer_set_property, diff --git a/src/core/timer.h b/src/core/timer.h index 9d919e4d3e..9c4b64f898 100644 --- a/src/core/timer.h +++ b/src/core/timer.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -25,16 +23,6 @@ typedef struct Timer Timer; #include "calendarspec.h" -typedef enum TimerState { - TIMER_DEAD, - TIMER_WAITING, - TIMER_RUNNING, - TIMER_ELAPSED, - TIMER_FAILED, - _TIMER_STATE_MAX, - _TIMER_STATE_INVALID = -1 -} TimerState; - typedef enum TimerBase { TIMER_ACTIVE, TIMER_BOOT, @@ -60,6 +48,7 @@ typedef struct TimerValue { typedef enum TimerResult { TIMER_SUCCESS, TIMER_FAILURE_RESOURCES, + TIMER_FAILURE_START_LIMIT_HIT, _TIMER_RESULT_MAX, _TIMER_RESULT_INVALID = -1 } TimerResult; @@ -68,6 +57,7 @@ struct Timer { Unit meta; usec_t accuracy_usec; + usec_t random_usec; LIST_HEAD(TimerValue, values); usec_t next_elapse_realtime; @@ -83,6 +73,7 @@ struct Timer { bool persistent; bool wake_system; + bool remain_after_elapse; char *stamp_path; }; @@ -91,9 +82,6 @@ void timer_free_values(Timer *t); extern const UnitVTable timer_vtable; -const char *timer_state_to_string(TimerState i) _const_; -TimerState timer_state_from_string(const char *s) _pure_; - const char *timer_base_to_string(TimerBase i) _const_; TimerBase timer_base_from_string(const char *s) _pure_; diff --git a/src/core/transaction.c b/src/core/transaction.c index 090103fbda..8370b864fb 100644 --- a/src/core/transaction.c +++ b/src/core/transaction.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,13 +17,15 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <unistd.h> #include <fcntl.h> +#include <unistd.h> +#include "alloc-util.h" #include "bus-common-errors.h" #include "bus-error.h" -#include "transaction.h" #include "terminal-util.h" +#include "transaction.h" +#include "dbus-unit.h" static void transaction_unlink_job(Transaction *tr, Job *j, bool delete_dependencies); @@ -98,9 +98,7 @@ static void transaction_merge_and_delete_job(Transaction *tr, Job *j, Job *other j->type = t; j->state = JOB_WAITING; - j->override = j->override || other->override; 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 */ @@ -375,17 +373,15 @@ static int transaction_verify_order_one(Transaction *tr, Job *j, Job *from, unsi delete = NULL; for (k = from; k; k = ((k->generation == generation && k->marker != k) ? k->marker : NULL)) { - /* logging for j not k here here to provide consistent narrative */ + /* 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)) { + 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 */ @@ -395,7 +391,8 @@ static int transaction_verify_order_one(Transaction *tr, Job *j, Job *from, unsi if (delete) { - /* logging for j not k here here to provide consistent narrative */ + 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)); @@ -403,7 +400,13 @@ static int transaction_verify_order_one(Transaction *tr, Job *j, Job *from, unsi "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)); - unit_status_printf(delete->unit, ANSI_HIGHLIGHT_RED_ON " SKIP " ANSI_HIGHLIGHT_OFF, + + 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; @@ -464,9 +467,11 @@ static int transaction_verify_order(Transaction *tr, unsigned *generation, sd_bu g = (*generation)++; - HASHMAP_FOREACH(j, tr->jobs, i) - if ((r = transaction_verify_order_one(tr, j, NULL, g, e)) < 0) + HASHMAP_FOREACH(j, tr->jobs, i) { + r = transaction_verify_order_one(tr, j, NULL, g, e); + if (r < 0) return r; + } return 0; } @@ -586,13 +591,16 @@ static int transaction_apply(Transaction *tr, Manager *m, JobMode mode) { 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); + job_finish_and_invalidate(j, JOB_CANCELED, false, false); } } @@ -736,15 +744,15 @@ int transaction_activate(Transaction *tr, Manager *m, JobMode mode, sd_bus_error if (m->idle_pipe[0] < 0 && m->idle_pipe[1] < 0 && m->idle_pipe[2] < 0 && m->idle_pipe[3] < 0) { - pipe2(m->idle_pipe, O_NONBLOCK|O_CLOEXEC); - pipe2(m->idle_pipe + 2, O_NONBLOCK|O_CLOEXEC); + (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 override, bool *is_new) { +static Job* transaction_add_one_job(Transaction *tr, JobType type, Unit *unit, bool *is_new) { Job *j, *f; assert(tr); @@ -773,7 +781,6 @@ static Job* transaction_add_one_job(Transaction *tr, JobType type, Unit *unit, b j->generation = 0; j->marker = NULL; j->matters_to_anchor = false; - j->override = override; j->irreversible = tr->irreversible; LIST_PREPEND(transaction, f, j); @@ -832,7 +839,6 @@ int transaction_add_job_and_dependencies( Unit *unit, Job *by, bool matters, - bool override, bool conflicts, bool ignore_requirements, bool ignore_order, @@ -852,7 +858,7 @@ int transaction_add_job_and_dependencies( * 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 (unit->manager->n_reloading > 0) + if (MANAGER_IS_RELOADING(unit->manager)) unit_coldplug(unit); /* log_debug("Pulling in %s/%s from %s/%s", */ @@ -863,30 +869,12 @@ int transaction_add_job_and_dependencies( 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 && unit->load_state == UNIT_ERROR) { - if (unit->load_error == -ENOENT || unit->manager->test_run) - return sd_bus_error_setf(e, BUS_ERROR_LOAD_FAILED, - "Unit %s failed to load: %s.", - unit->id, - strerror(-unit->load_error)); - else - return sd_bus_error_setf(e, BUS_ERROR_LOAD_FAILED, - "Unit %s failed to load: %s. " - "See system logs and 'systemctl status %s' for details.", - unit->id, - strerror(-unit->load_error), - unit->id); + if (type != JOB_STOP) { + r = bus_unit_check_load_state(unit, e); + if (r < 0) + return r; } - if (type != JOB_STOP && unit->load_state == UNIT_NOT_FOUND) - return sd_bus_error_setf(e, BUS_ERROR_LOAD_FAILED, - "Unit %s failed to load: %s.", - unit->id, strerror(-unit->load_error)); - - if (type != JOB_STOP && unit->load_state == UNIT_MASKED) - return sd_bus_error_setf(e, BUS_ERROR_UNIT_MASKED, - "Unit %s is masked.", unit->id); - 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.", @@ -894,7 +882,7 @@ int transaction_add_job_and_dependencies( /* First add the job. */ - ret = transaction_add_one_job(tr, type, unit, override, &is_new); + ret = transaction_add_one_job(tr, type, unit, &is_new); if (!ret) return -ENOMEM; @@ -917,7 +905,7 @@ int transaction_add_job_and_dependencies( * 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, override, false, false, ignore_order, e); + 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); @@ -930,9 +918,9 @@ int transaction_add_job_and_dependencies( /* 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, override, false, false, ignore_order, e); + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, false, false, ignore_order, e); if (r < 0) { - if (r != -EBADR) + if (r != -EBADR) /* job type not applicable */ goto fail; sd_bus_error_free(e); @@ -940,62 +928,41 @@ int transaction_add_job_and_dependencies( } SET_FOREACH(dep, ret->unit->dependencies[UNIT_BINDS_TO], i) { - r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, override, false, false, ignore_order, e); + r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, true, false, false, ignore_order, e); if (r < 0) { - if (r != -EBADR) + if (r != -EBADR) /* job type not applicable */ goto fail; sd_bus_error_free(e); } } - SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUIRES_OVERRIDABLE], i) { - r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, !override, override, false, false, ignore_order, e); - if (r < 0) { - log_unit_full(dep, - r == -EADDRNOTAVAIL ? 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_WANTS], i) { - r = transaction_add_job_and_dependencies(tr, JOB_START, dep, ret, false, false, false, false, ignore_order, e); + 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, - r == -EADDRNOTAVAIL ? LOG_DEBUG : LOG_WARNING, r, - "Cannot add dependency job, ignoring: %s", + 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, override, false, false, ignore_order, e); + r = transaction_add_job_and_dependencies(tr, JOB_VERIFY_ACTIVE, dep, ret, true, false, false, ignore_order, e); if (r < 0) { - if (r != -EBADR) + if (r != -EBADR) /* job type not applicable */ goto fail; sd_bus_error_free(e); } } - SET_FOREACH(dep, ret->unit->dependencies[UNIT_REQUISITE_OVERRIDABLE], i) { - r = transaction_add_job_and_dependencies(tr, JOB_VERIFY_ACTIVE, dep, ret, !override, override, false, false, ignore_order, e); - if (r < 0) { - log_unit_full(dep, - r == -EADDRNOTAVAIL ? 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_CONFLICTS], i) { - r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, true, override, true, false, ignore_order, e); + r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, true, true, false, ignore_order, e); if (r < 0) { - if (r != -EBADR) + if (r != -EBADR) /* job type not applicable */ goto fail; sd_bus_error_free(e); @@ -1003,7 +970,7 @@ int transaction_add_job_and_dependencies( } SET_FOREACH(dep, ret->unit->dependencies[UNIT_CONFLICTED_BY], i) { - r = transaction_add_job_and_dependencies(tr, JOB_STOP, dep, ret, false, override, false, false, ignore_order, e); + 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", @@ -1038,9 +1005,9 @@ int transaction_add_job_and_dependencies( if (nt == JOB_NOP) continue; - r = transaction_add_job_and_dependencies(tr, nt, dep, ret, true, override, false, false, ignore_order, e); + r = transaction_add_job_and_dependencies(tr, nt, dep, ret, true, false, false, ignore_order, e); if (r < 0) { - if (r != -EBADR) + if (r != -EBADR) /* job type not applicable */ goto fail; sd_bus_error_free(e); @@ -1051,7 +1018,13 @@ int transaction_add_job_and_dependencies( if (type == JOB_RELOAD) { SET_FOREACH(dep, ret->unit->dependencies[UNIT_PROPAGATES_RELOAD_TO], i) { - r = transaction_add_job_and_dependencies(tr, JOB_RELOAD, dep, ret, false, override, false, false, ignore_order, e); + 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", @@ -1096,7 +1069,7 @@ int transaction_add_isolate_jobs(Transaction *tr, Manager *m) { if (hashmap_get(tr->jobs, u)) continue; - r = transaction_add_job_and_dependencies(tr, JOB_STOP, u, tr->anchor_job, true, false, false, false, false, NULL); + 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"); } diff --git a/src/core/transaction.h b/src/core/transaction.h index d949b21b8d..6a3f927b0f 100644 --- a/src/core/transaction.h +++ b/src/core/transaction.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -23,10 +21,10 @@ typedef struct Transaction Transaction; -#include "unit.h" -#include "manager.h" -#include "job.h" #include "hashmap.h" +#include "job.h" +#include "manager.h" +#include "unit.h" struct Transaction { /* Jobs to be added */ @@ -44,7 +42,6 @@ int transaction_add_job_and_dependencies( Unit *unit, Job *by, bool matters, - bool override, bool conflicts, bool ignore_requirements, bool ignore_order, diff --git a/src/core/triggers.systemd.in b/src/core/triggers.systemd.in new file mode 100644 index 0000000000..0d8c303136 --- /dev/null +++ b/src/core/triggers.systemd.in @@ -0,0 +1,66 @@ +# -*- Mode: rpm-spec; indent-tabs-mode: nil -*- */ +# +# This file is part of systemd. +# +# Copyright 2015 Zbigniew Jędrzejewski-Szmek +# +# 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/>. + +# The contents of this are an example to be copied into systemd.spec. +# +# Minimum rpm version supported: 4.13.0 + +%transfiletriggerin -P 900900 -p <lua> -- @systemunitdir@ /etc/systemd/system +-- This script will run after any package is initially installed or +-- upgraded. We care about the case where a package is initially +-- installed, because other cases are covered by the *un scriptlets, +-- so sometimes we will reload needlessly. + +pid = posix.fork() +if pid == 0 then + assert(posix.exec("%{_bindir}/systemctl", "daemon-reload")) +elseif pid > 0 then + posix.wait(pid) +end + +%transfiletriggerun -p <lua> -- @systemunitdir@ /etc/systemd/system +-- On removal, we need to run daemon-reload after any units have been +-- removed. %transfiletriggerpostun would be ideal, but it does not get +-- executed for some reason. +-- On upgrade, we need to run daemon-reload after any new unit files +-- have been installed, but before %postun scripts in packages get +-- executed. %transfiletriggerun gets the right list of files +-- but it is invoked too early (before changes happen). +-- %filetriggerpostun happens at the right time, but it fires for +-- every package. +-- To execute the reload at the right time, we create a state +-- file in %transfiletriggerun and execute the daemon-reload in +-- the first %filetriggerpostun. + +posix.mkdir("%{_localstatedir}/lib") +posix.mkdir("%{_localstatedir}/lib/rpm-state") +posix.mkdir("%{_localstatedir}/lib/rpm-state/systemd") +io.open("%{_localstatedir}/lib/rpm-state/systemd/needs-reload", "w") + +%filetriggerpostun -P 1000100 -p <lua> -- @systemunitdir@ /etc/systemd/system +if posix.access("%{_localstatedir}/lib/rpm-state/systemd/needs-reload") then + posix.unlink("%{_localstatedir}/lib/rpm-state/systemd/needs-reload") + posix.rmdir("%{_localstatedir}/lib/rpm-state/systemd") + pid = posix.fork() + if pid == 0 then + assert(posix.exec("%{_bindir}/systemctl", "daemon-reload")) + elseif pid > 0 then + posix.wait(pid) + end +end diff --git a/src/core/umount.c b/src/core/umount.c index d59b5d0ffb..1e5459ed80 100644 --- a/src/core/umount.c +++ b/src/core/umount.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -21,23 +19,30 @@ #include <errno.h> #include <fcntl.h> +#include <linux/dm-ioctl.h> +#include <linux/loop.h> #include <string.h> #include <sys/mount.h> #include <sys/swap.h> -#include <linux/loop.h> -#include <linux/dm-ioctl.h> +#include "libudev.h" + +#include "alloc-util.h" +#include "escape.h" +#include "fd-util.h" +#include "fstab-util.h" #include "list.h" #include "mount-setup.h" -#include "umount.h" #include "path-util.h" +#include "string-util.h" +#include "udev-util.h" +#include "umount.h" #include "util.h" #include "virt.h" -#include "libudev.h" -#include "udev-util.h" typedef struct MountPoint { char *path; + char *options; dev_t devnum; LIST_FIELDS(struct MountPoint, mount_point); } MountPoint; @@ -71,7 +76,7 @@ static int mount_points_list_get(MountPoint **head) { return -errno; for (i = 1;; i++) { - _cleanup_free_ char *path = NULL; + _cleanup_free_ char *path = NULL, *options = NULL; char *p = NULL; MountPoint *m; int k; @@ -82,15 +87,15 @@ static int mount_points_list_get(MountPoint **head) { "%*s " /* (3) major:minor */ "%*s " /* (4) root */ "%ms " /* (5) mount point */ - "%*s" /* (6) mount options */ + "%*s" /* (6) mount flags */ "%*[^-]" /* (7) optional fields */ "- " /* (8) separator */ "%*s " /* (9) file system type */ "%*s" /* (10) mount source */ - "%*s" /* (11) mount options 2 */ + "%ms" /* (11) mount options */ "%*[^\n]", /* some rubbish at the end */ - &path); - if (k != 1) { + &path, &options); + if (k != 2) { if (k == EOF) break; @@ -125,6 +130,9 @@ static int mount_points_list_get(MountPoint **head) { } m->path = p; + m->options = options; + options = NULL; + LIST_PREPEND(mount_point, *head, m); } @@ -367,8 +375,16 @@ static int mount_points_list_umount(MountPoint **head, bool *changed, bool log_e /* If we are in a container, don't attempt to read-only mount anything as that brings no real benefits, but might confuse the host, as we remount - the superblock here, not the bind mound. */ - if (detect_container(NULL) <= 0) { + the superblock here, not the bind mount. */ + if (detect_container() <= 0) { + _cleanup_free_ char *options = NULL; + /* MS_REMOUNT requires that the data parameter + * should be the same from the original mount + * except for the desired changes. Since we want + * to remount read-only, we should filter out + * rw (and ro too, because it confuses the kernel) */ + (void) fstab_filter_options(m->options, "rw\0ro\0", NULL, NULL, &options); + /* We always try to remount directories * read-only first, before we go on and umount * them. @@ -385,7 +401,8 @@ static int mount_points_list_umount(MountPoint **head, bool *changed, bool log_e * alias read-only we hence should be * relatively safe regarding keeping the fs we * can otherwise not see dirty. */ - (void) mount(NULL, m->path, NULL, MS_REMOUNT|MS_RDONLY, NULL); + log_info("Remounting '%s' read-only with options '%s'.", m->path, options); + (void) mount(NULL, m->path, NULL, MS_REMOUNT|MS_RDONLY, options); } /* Skip / and /usr since we cannot unmount that @@ -395,6 +412,7 @@ static int mount_points_list_umount(MountPoint **head, bool *changed, bool log_e #ifndef HAVE_SPLIT_USR || path_equal(m->path, "/usr") #endif + || path_startswith(m->path, "/run/initramfs") ) continue; @@ -455,7 +473,7 @@ static int loopback_points_list_detach(MountPoint **head, bool *changed) { major(root_st.st_dev) != 0 && lstat(m->path, &loopback_st) >= 0 && root_st.st_dev == loopback_st.st_rdev) { - n_failed ++; + n_failed++; continue; } @@ -490,7 +508,7 @@ static int dm_points_list_detach(MountPoint **head, bool *changed) { if (k >= 0 && major(root_st.st_dev) != 0 && root_st.st_dev == m->devnum) { - n_failed ++; + n_failed++; continue; } diff --git a/src/core/umount.h b/src/core/umount.h index 8439ffe58f..4e2215a47d 100644 --- a/src/core/umount.h +++ b/src/core/umount.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/unit-printf.c b/src/core/unit-printf.c index 0889769d03..f11df42af3 100644 --- a/src/core/unit-printf.c +++ b/src/core/unit-printf.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,14 +17,17 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include "unit.h" +#include "alloc-util.h" +#include "cgroup-util.h" +#include "formats-util.h" +#include "macro.h" #include "specifier.h" +#include "string-util.h" #include "strv.h" #include "unit-name.h" #include "unit-printf.h" -#include "macro.h" -#include "cgroup-util.h" -#include "formats-util.h" +#include "unit.h" +#include "user-util.h" static int specifier_prefix_and_instance(char specifier, void *data, void *userdata, char **ret) { Unit *u = userdata; @@ -63,10 +64,7 @@ static int specifier_instance_unescaped(char specifier, void *data, void *userda assert(u); - if (!u->instance) - return -EINVAL; - - return unit_name_unescape(u->instance, ret); + return unit_name_unescape(strempty(u->instance), ret); } static int specifier_filename(char specifier, void *data, void *userdata, char **ret) { @@ -128,6 +126,8 @@ static int specifier_cgroup_slice(char specifier, void *data, void *userdata, ch n = unit_default_cgroup_path(slice); } else n = strdup(u->manager->cgroup_root); + if (!n) + return -ENOMEM; *ret = n; return 0; @@ -140,14 +140,9 @@ static int specifier_runtime(char specifier, void *data, void *userdata, char ** assert(u); - if (u->manager->running_as == MANAGER_SYSTEM) - e = "/run"; - else { - e = getenv("XDG_RUNTIME_DIR"); - if (!e) - return -EOPNOTSUPP; - } - + e = manager_get_runtime_prefix(u->manager); + if (!e) + return -EOPNOTSUPP; n = strdup(e); if (!n) return -ENOMEM; @@ -157,162 +152,43 @@ static int specifier_runtime(char specifier, void *data, void *userdata, char ** } static int specifier_user_name(char specifier, void *data, void *userdata, char **ret) { - char *printed = NULL; - Unit *u = userdata; - ExecContext *c; - int r = 0; - - assert(u); + char *t; - c = unit_get_exec_context(u); - if (!c) - return -EINVAL; + /* 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. */ - if (u->manager->running_as == MANAGER_SYSTEM) { - - /* We cannot use NSS from PID 1, hence try to make the - * best of it in that case, and fail if we can't help - * it */ - - if (!c->user || streq(c->user, "root") || streq(c->user, "0")) - printed = strdup(specifier == 'u' ? "root" : "0"); - else { - if (specifier == 'u') - printed = strdup(c->user); - else { - uid_t uid; - - r = parse_uid(c->user, &uid); - if (r < 0) - return -ENODATA; - - r = asprintf(&printed, UID_FMT, uid); - } - } - - } else { - _cleanup_free_ char *tmp = NULL; - const char *username = NULL; - uid_t uid; - - if (c->user) - username = c->user; - else - /* get USER env from env or our own uid */ - username = tmp = getusername_malloc(); - - /* fish username from passwd */ - r = get_user_creds(&username, &uid, NULL, NULL, NULL); - if (r < 0) - return r; - - if (specifier == 'u') - printed = strdup(username); - else - r = asprintf(&printed, UID_FMT, uid); - } - - if (r < 0 || !printed) + t = getusername_malloc(); + if (!t) return -ENOMEM; - *ret = printed; + *ret = t; return 0; } -static int specifier_user_home(char specifier, void *data, void *userdata, char **ret) { - Unit *u = userdata; - ExecContext *c; - char *n; - int r; - - assert(u); +static int specifier_user_id(char specifier, void *data, void *userdata, char **ret) { - c = unit_get_exec_context(u); - if (!c) - return -EOPNOTSUPP; - - if (u->manager->running_as == MANAGER_SYSTEM) { - - /* We cannot use NSS from PID 1, hence try to make the - * best of it if we can, but fail if we can't */ - - if (!c->user || streq(c->user, "root") || streq(c->user, "0")) - n = strdup("/root"); - else - return -EOPNOTSUPP; - - } else { - - /* return HOME if set, otherwise from passwd */ - if (!c || !c->user) { - r = get_home_dir(&n); - if (r < 0) - return r; - } else { - const char *username, *home; - - username = c->user; - r = get_user_creds(&username, NULL, NULL, &home, NULL); - if (r < 0) - return r; - - n = strdup(home); - } - } - - if (!n) + if (asprintf(ret, UID_FMT, getuid()) < 0) return -ENOMEM; - *ret = n; return 0; } -static int specifier_user_shell(char specifier, void *data, void *userdata, char **ret) { - Unit *u = userdata; - ExecContext *c; - char *n; - int r; - - assert(u); - - c = unit_get_exec_context(u); - if (!c) - return -EOPNOTSUPP; - - if (u->manager->running_as == MANAGER_SYSTEM) { - - /* We cannot use NSS from PID 1, hence try to make the - * best of it if we can, but fail if we can't */ - - if (!c->user || streq(c->user, "root") || streq(c->user, "0")) - n = strdup("/bin/sh"); - else - return -EOPNOTSUPP; - - } else { +static int specifier_user_home(char specifier, void *data, void *userdata, char **ret) { - /* return /bin/sh for root, otherwise the value from passwd */ - if (!c->user) { - r = get_shell(&n); - if (r < 0) - return r; - } else { - const char *username, *shell; + /* On PID 1 (which runs as root) this will not result in NSS, + * which is good. See above */ - username = c->user; - r = get_user_creds(&username, NULL, NULL, NULL, &shell); - if (r < 0) - return r; + return get_home_dir(ret); +} - n = strdup(shell); - } - } +static int specifier_user_shell(char specifier, void *data, void *userdata, char **ret) { - if (!n) - return -ENOMEM; + /* On PID 1 (which runs as root) this will not result in NSS, + * which is good. See above */ - *ret = n; - return 0; + return get_shell(ret); } int unit_name_printf(Unit *u, const char* format, char **ret) { @@ -352,10 +228,10 @@ int unit_full_printf(Unit *u, const char *format, char **ret) { * %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 configured user or running user - * %u the username of the configured user or running user - * %h the homedir of the configured user or running user - * %s the shell of the configured user or running user + * %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 @@ -375,7 +251,8 @@ int unit_full_printf(Unit *u, const char *format, char **ret) { { 'r', specifier_cgroup_slice, NULL }, { 'R', specifier_cgroup_root, NULL }, { 't', specifier_runtime, NULL }, - { 'U', specifier_user_name, NULL }, + + { 'U', specifier_user_id, NULL }, { 'u', specifier_user_name, NULL }, { 'h', specifier_user_home, NULL }, { 's', specifier_user_shell, NULL }, diff --git a/src/core/unit-printf.h b/src/core/unit-printf.h index 51acad63e9..4fc8531228 100644 --- a/src/core/unit-printf.h +++ b/src/core/unit-printf.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** diff --git a/src/core/unit.c b/src/core/unit.c index fac017c57d..690f7f7dd9 100644 --- a/src/core/unit.c +++ b/src/core/unit.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -20,41 +18,53 @@ ***/ #include <errno.h> -#include <string.h> #include <stdlib.h> -#include <unistd.h> +#include <string.h> #include <sys/stat.h> +#include <unistd.h> #include "sd-id128.h" #include "sd-messages.h" -#include "set.h" -#include "unit.h" -#include "macro.h" -#include "strv.h" -#include "path-util.h" -#include "load-fragment.h" -#include "load-dropin.h" -#include "log.h" -#include "unit-name.h" -#include "dbus-unit.h" -#include "special.h" -#include "cgroup-util.h" -#include "missing.h" -#include "mkdir.h" -#include "fileio-label.h" + +#include "alloc-util.h" #include "bus-common-errors.h" +#include "bus-util.h" +#include "cgroup-util.h" +#include "dbus-unit.h" #include "dbus.h" -#include "execute.h" #include "dropin.h" +#include "escape.h" +#include "execute.h" +#include "fileio-label.h" #include "formats-util.h" +#include "id128-util.h" +#include "load-dropin.h" +#include "load-fragment.h" +#include "log.h" +#include "macro.h" +#include "missing.h" +#include "mkdir.h" +#include "parse-util.h" +#include "path-util.h" #include "process-util.h" +#include "set.h" +#include "signal-util.h" +#include "special.h" +#include "stat-util.h" +#include "stdio-util.h" +#include "string-util.h" +#include "strv.h" +#include "umask-util.h" +#include "unit-name.h" +#include "unit.h" +#include "user-util.h" +#include "virt.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_SNAPSHOT] = &snapshot_vtable, [UNIT_DEVICE] = &device_vtable, [UNIT_MOUNT] = &mount_vtable, [UNIT_AUTOMOUNT] = &automount_vtable, @@ -89,7 +99,13 @@ Unit *unit_new(Manager *m, size_t size) { 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; @@ -99,7 +115,7 @@ bool unit_has_name(Unit *u, const char *name) { assert(u); assert(name); - return !!set_get(u->names, (char*) name); + return set_contains(u->names, (char*) name); } static void unit_init(Unit *u) { @@ -120,8 +136,13 @@ static void unit_init(Unit *u) { * 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); @@ -177,7 +198,7 @@ int unit_add_name(Unit *u, const char *text) { if (r < 0) return r; - if (i && unit_vtable[t]->no_instances) + if (i && !unit_type_may_template(t)) return -EINVAL; /* Ensure that this unit is either instanced or not instanced, @@ -186,7 +207,7 @@ int unit_add_name(Unit *u, const char *text) { if (u->type != _UNIT_TYPE_INVALID && !u->instance != !i) return -EINVAL; - if (unit_vtable[t]->no_alias && !set_isempty(u->names)) + if (!unit_type_may_alias(t) && !set_isempty(u->names)) return -EEXIST; if (hashmap_size(u->manager->units) >= MANAGER_MAX_NAMES) @@ -304,15 +325,15 @@ bool unit_check_gc(Unit *u) { if (state != UNIT_INACTIVE) return true; - if (UNIT_VTABLE(u)->no_gc) - return true; - if (u->no_gc) 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; @@ -353,7 +374,7 @@ void unit_add_to_gc_queue(Unit *u) { LIST_PREPEND(gc_queue, u->manager->gc_queue, u); u->in_gc_queue = true; - u->manager->n_in_gc_queue ++; + u->manager->n_in_gc_queue++; } void unit_add_to_dbus_queue(Unit *u) { @@ -404,17 +425,25 @@ static void unit_remove_transient(Unit *u) { return; if (u->fragment_path) - unlink(u->fragment_path); + (void) unlink(u->fragment_path); STRV_FOREACH(i, u->dropin_paths) { - _cleanup_free_ char *p = NULL; - int r; + _cleanup_free_ char *p = NULL, *pp = NULL; + + p = dirname_malloc(*i); /* Get the drop-in directory from the drop-in file */ + if (!p) + continue; - unlink(*i); + pp = dirname_malloc(p); /* Get the config directory from the drop-in directory */ + if (!pp) + continue; - r = path_get_parent(*i, &p); - if (r >= 0) - rmdir(p); + /* Only drop transient drop-ins */ + if (!path_equal(u->manager->lookup_paths.transient, pp)) + continue; + + (void) unlink(*i); + (void) rmdir(p); } } @@ -442,8 +471,7 @@ static void unit_free_requires_mounts_for(Unit *u) { } } - strv_free(u->requires_mounts_for); - u->requires_mounts_for = NULL; + u->requires_mounts_for = strv_free(u->requires_mounts_for); } static void unit_done(Unit *u) { @@ -474,18 +502,29 @@ void unit_free(Unit *u) { assert(u); - if (u->manager->n_reloading <= 0) + 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); @@ -521,12 +560,11 @@ void unit_free(Unit *u) { if (u->in_cgroup_queue) LIST_REMOVE(cgroup_queue, u->manager->cgroup_queue, u); - if (u->cgroup_path) { - hashmap_remove(u->manager->cgroup_unit, u->cgroup_path); - free(u->cgroup_path); - } + unit_release_cgroup(u); + + unit_unref_uid_gid(u, false); - manager_update_failed_units(u->manager, u, false); + (void) manager_update_failed_units(u->manager, u, false); set_remove(u->manager->startup_units, u); free(u->description); @@ -545,6 +583,8 @@ void unit_free(Unit *u) { condition_free_list(u->conditions); condition_free_list(u->asserts); + free(u->reboot_arg); + unit_ref_unset(&u->slice); while (u->refs) @@ -672,8 +712,7 @@ static void merge_dependencies(Unit *u, Unit *other, const char *other_id, UnitD /* The move cannot fail. The caller must have performed a reservation. */ assert_se(complete_move(&u->dependencies[d], &other->dependencies[d]) == 0); - set_free(other->dependencies[d]); - other->dependencies[d] = NULL; + other->dependencies[d] = set_free(other->dependencies[d]); } int unit_merge(Unit *u, Unit *other) { @@ -697,6 +736,9 @@ int unit_merge(Unit *u, Unit *other) { 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; @@ -753,9 +795,9 @@ int unit_merge(Unit *u, Unit *other) { } int unit_merge_by_name(Unit *u, const char *name) { + _cleanup_free_ char *s = NULL; Unit *other; int r; - _cleanup_free_ char *s = NULL; assert(u); assert(name); @@ -805,7 +847,7 @@ int unit_add_exec_dependencies(Unit *u, ExecContext *c) { return r; } - if (u->manager->running_as != MANAGER_SYSTEM) + if (!MANAGER_IS_SYSTEM(u->manager)) return 0; if (c->private_tmp) { @@ -857,6 +899,7 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { Iterator i; const char *prefix2; char + timestamp0[FORMAT_TIMESTAMP_MAX], timestamp1[FORMAT_TIMESTAMP_MAX], timestamp2[FORMAT_TIMESTAMP_MAX], timestamp3[FORMAT_TIMESTAMP_MAX], @@ -865,6 +908,7 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { Unit *following; _cleanup_set_free_ Set *following_set = NULL; int r; + const char *n; assert(u); assert(u->type >= 0); @@ -878,6 +922,7 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { "%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" @@ -895,6 +940,7 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { 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)), @@ -911,6 +957,10 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { 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); @@ -935,7 +985,7 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { STRV_FOREACH(j, u->dropin_paths) fprintf(f, "%s\tDropIn Path: %s\n", prefix, *j); - if (u->job_timeout > 0) + 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 != FAILURE_ACTION_NONE) @@ -986,15 +1036,13 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { "%s\tRefuseManualStop: %s\n" "%s\tDefaultDependencies: %s\n" "%s\tOnFailureJobMode: %s\n" - "%s\tIgnoreOnIsolate: %s\n" - "%s\tIgnoreOnSnapshot: %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), - prefix, yes_no(u->ignore_on_snapshot)); + prefix, yes_no(u->ignore_on_isolate)); if (UNIT_VTABLE(u)->dump) UNIT_VTABLE(u)->dump(u, f, prefix2); @@ -1006,13 +1054,14 @@ void unit_dump(Unit *u, FILE *f, const char *prefix) { 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 */ @@ -1092,9 +1141,7 @@ static int unit_add_target_dependencies(Unit *u) { static const UnitDependency deps[] = { UNIT_REQUIRED_BY, - UNIT_REQUIRED_BY_OVERRIDABLE, UNIT_REQUISITE_OF, - UNIT_REQUISITE_OF_OVERRIDABLE, UNIT_WANTED_BY, UNIT_BOUND_BY }; @@ -1119,16 +1166,16 @@ static int unit_add_target_dependencies(Unit *u) { static int unit_add_slice_dependencies(Unit *u) { assert(u); - if (!unit_get_cgroup_context(u)) + if (!UNIT_HAS_CGROUP_CONTEXT(u)) return 0; if (UNIT_ISSET(u->slice)) - return unit_add_two_dependencies(u, UNIT_AFTER, UNIT_WANTS, UNIT_DEREF(u->slice), true); + return unit_add_two_dependencies(u, UNIT_AFTER, UNIT_REQUIRES, UNIT_DEREF(u->slice), true); - if (streq(u->id, SPECIAL_ROOT_SLICE)) + if (unit_has_name(u, SPECIAL_ROOT_SLICE)) return 0; - return unit_add_two_dependencies_by_name(u, UNIT_AFTER, UNIT_WANTS, SPECIAL_ROOT_SLICE, NULL, true); + 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) { @@ -1141,13 +1188,23 @@ static int unit_add_mount_dependencies(Unit *u) { char prefix[strlen(*i) + 1]; PATH_FOREACH_PREFIX_MORE(prefix, *i) { + _cleanup_free_ char *p = NULL; Unit *m; - r = manager_get_unit_by_path(u->manager, prefix, ".mount", &m); + r = unit_name_from_path(prefix, ".mount", &p); if (r < 0) return r; - if (r == 0) + + 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; @@ -1171,15 +1228,21 @@ static int unit_add_mount_dependencies(Unit *u) { 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 == (unsigned long) -1 && - c->startup_blockio_weight == (unsigned long) -1) + 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); } @@ -1199,6 +1262,17 @@ int unit_load(Unit *u) { 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) @@ -1317,43 +1391,35 @@ static bool unit_assert_test(Unit *u) { return u->assert_result; } -_pure_ static const char* unit_get_status_message_format(Unit *u, JobType t) { - const UnitStatusMessageFormats *format_table; - - assert(u); - assert(t >= 0); - assert(t < _JOB_TYPE_MAX); - - if (t != JOB_START && t != JOB_STOP) - return NULL; - - format_table = &UNIT_VTABLE(u)->status_message_formats; - if (!format_table) - return NULL; - - return format_table->starting_stopping[t == JOB_STOP]; +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_try_harder(Unit *u, JobType t) { +_pure_ static const char* unit_get_status_message_format(Unit *u, JobType t) { const char *format; + const UnitStatusMessageFormats *format_table; assert(u); - assert(t >= 0); - assert(t < _JOB_TYPE_MAX); - - format = unit_get_status_message_format(u, t); - if (format) - return format; + 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 if (t == JOB_RELOAD) + else return "Reloading %s."; - - return NULL; } static void unit_status_print_starting_stopping(Unit *u, JobType t) { @@ -1361,12 +1427,11 @@ static void unit_status_print_starting_stopping(Unit *u, JobType t) { assert(u); - /* We only print status messages for selected units on - * selected operations. */ + /* 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); - if (!format) - return; DISABLE_WARNING_FORMAT_NONLITERAL; unit_status_printf(u, "", format); @@ -1380,7 +1445,7 @@ static void unit_status_log_starting_stopping_reloading(Unit *u, JobType t) { assert(u); - if (t != JOB_START && t != JOB_STOP && t != JOB_RELOAD) + if (!IN_SET(t, JOB_START, JOB_STOP, JOB_RELOAD)) return; if (log_on_console()) @@ -1388,12 +1453,10 @@ static void unit_status_log_starting_stopping_reloading(Unit *u, JobType t) { /* We log status messages for all units and all operations. */ - format = unit_get_status_message_format_try_harder(u, t); - if (!format) - return; + format = unit_get_status_message_format(u, t); DISABLE_WARNING_FORMAT_NONLITERAL; - snprintf(buf, sizeof(buf), format, unit_description(u)); + xsprintf(buf, format, unit_description(u)); REENABLE_WARNING; mid = t == JOB_START ? SD_MESSAGE_UNIT_STARTING : @@ -1413,23 +1476,44 @@ static void unit_status_log_starting_stopping_reloading(Unit *u, JobType t) { 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 failure_action(u->manager, u->start_limit_action, u->reboot_arg); +} + /* 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 + * -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; - int r; assert(u); - if (u->load_state != UNIT_LOADED) - return -EINVAL; - /* 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 @@ -1438,6 +1522,10 @@ int unit_start(Unit *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, @@ -1455,6 +1543,15 @@ int unit_start(Unit *u) { 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) { @@ -1462,9 +1559,6 @@ int unit_start(Unit *u) { return unit_start(following); } - if (!unit_supported(u)) - return -EOPNOTSUPP; - /* If it is stopped, but we cannot start it, then fail */ if (!UNIT_VTABLE(u)->start) return -EBADR; @@ -1477,19 +1571,18 @@ int unit_start(Unit *u) { unit_add_to_dbus_queue(u); - r = UNIT_VTABLE(u)->start(u); - if (r <= 0) - return r; - - /* Log if the start function actually did something */ - unit_status_log_starting_stopping_reloading(u, JOB_START); - unit_status_print_starting_stopping(u, JOB_START); - return r; + 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; } @@ -1508,7 +1601,6 @@ bool unit_can_isolate(Unit *u) { int unit_stop(Unit *u) { UnitActiveState state; Unit *following; - int r; assert(u); @@ -1527,13 +1619,7 @@ int unit_stop(Unit *u) { unit_add_to_dbus_queue(u); - r = UNIT_VTABLE(u)->stop(u); - if (r <= 0) - return r; - - unit_status_log_starting_stopping_reloading(u, JOB_STOP); - unit_status_print_starting_stopping(u, JOB_STOP); - return r; + return UNIT_VTABLE(u)->stop(u); } /* Errors: @@ -1544,7 +1630,6 @@ int unit_stop(Unit *u) { int unit_reload(Unit *u) { UnitActiveState state; Unit *following; - int r; assert(u); @@ -1571,12 +1656,7 @@ int unit_reload(Unit *u) { unit_add_to_dbus_queue(u); - r = UNIT_VTABLE(u)->reload(u); - if (r <= 0) - return r; - - unit_status_log_starting_stopping_reloading(u, JOB_RELOAD); - return r; + return UNIT_VTABLE(u)->reload(u); } bool unit_can_reload(Unit *u) { @@ -1593,11 +1673,11 @@ bool unit_can_reload(Unit *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_REQUIRED_BY_OVERRIDABLE, UNIT_REQUISITE_OF, - UNIT_REQUISITE_OF_OVERRIDABLE, UNIT_WANTED_BY, UNIT_BOUND_BY, }; @@ -1623,7 +1703,7 @@ static void unit_check_unneeded(Unit *u) { if (unit_active_or_pending(other)) return; - /* If stopping a unit fails continously we might enter a stop + /* 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)) { @@ -1634,12 +1714,13 @@ static void unit_check_unneeded(Unit *u) { 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, true, NULL, NULL); + 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: %m"); + 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; @@ -1667,7 +1748,7 @@ static void unit_check_binds_to(Unit *u) { if (!stop) return; - /* If stopping a unit fails continously we might enter a stop + /* 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)) { @@ -1679,9 +1760,9 @@ static void unit_check_binds_to(Unit *u) { 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, true, NULL, NULL); + 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: %m"); + 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) { @@ -1694,30 +1775,25 @@ static void retroactively_start_dependencies(Unit *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, true, NULL, NULL); + 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, true, NULL, NULL); - - SET_FOREACH(other, u->dependencies[UNIT_REQUIRES_OVERRIDABLE], 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, false, NULL, NULL); + 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, false, NULL, NULL); + 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, true, NULL, NULL); + 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, true, NULL, NULL); + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, NULL, NULL); } static void retroactively_stop_dependencies(Unit *u) { @@ -1730,7 +1806,7 @@ static void retroactively_stop_dependencies(Unit *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, true, NULL, NULL); + manager_add_job(u->manager, JOB_STOP, other, JOB_REPLACE, NULL, NULL); } static void check_unneeded_dependencies(Unit *u) { @@ -1744,18 +1820,12 @@ static void check_unneeded_dependencies(Unit *u) { 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_REQUIRES_OVERRIDABLE], 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_REQUISITE_OVERRIDABLE], 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); @@ -1775,7 +1845,7 @@ void unit_start_on_failure(Unit *u) { 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, true, NULL, NULL); + 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"); } @@ -1809,28 +1879,26 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su m = u->manager; /* Update timestamps for state changes */ - if (m->n_reloading <= 0) { - dual_timestamp ts; - - dual_timestamp_get(&ts); + 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 = ts; + 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 = ts; + 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 = ts; + 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 = ts; + u->active_exit_timestamp = u->state_change_timestamp; } /* Keep track of failed units */ - manager_update_failed_units(u->manager, u, ns == UNIT_FAILED); + (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_destroy_cgroup_if_empty(u); + 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 @@ -1841,13 +1909,13 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su 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 --; + 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 ++; + m->n_on_console++; } } @@ -1871,27 +1939,28 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su case JOB_VERIFY_ACTIVE: if (UNIT_IS_ACTIVE_OR_RELOADING(ns)) - job_finish_and_invalidate(u->job, JOB_DONE, true); + 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); + 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); + 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); + job_finish_and_invalidate(u->job, ns == UNIT_FAILED ? JOB_FAILED : JOB_DONE, true, false); } } @@ -1902,10 +1971,10 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su case JOB_TRY_RESTART: if (UNIT_IS_INACTIVE_OR_FAILED(ns)) - job_finish_and_invalidate(u->job, JOB_DONE, true); + 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); + job_finish_and_invalidate(u->job, JOB_FAILED, true, false); } break; @@ -1917,7 +1986,7 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su } else unexpected = true; - if (m->n_reloading <= 0) { + if (!MANAGER_IS_RELOADING(m)) { /* If this state change happened without being * requested by a job, then let's retroactively start @@ -1954,7 +2023,7 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su if (u->type == UNIT_SERVICE && !UNIT_IS_ACTIVE_OR_RELOADING(os) && - m->n_reloading <= 0) { + !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; @@ -1971,7 +2040,7 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su if (u->type == UNIT_SERVICE && UNIT_IS_INACTIVE_OR_FAILED(ns) && !UNIT_IS_INACTIVE_OR_FAILED(os) && - m->n_reloading <= 0) { + !MANAGER_IS_RELOADING(m)) { /* Hmm, if there was no start record written * write it now, so that we always have a nice @@ -1992,7 +2061,7 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su manager_recheck_journal(m); unit_trigger_notify(u); - if (u->manager->n_reloading <= 0) { + if (!MANAGER_IS_RELOADING(u->manager)) { /* Maybe we finished startup and are now ready for * being stopped because unneeded? */ unit_check_unneeded(u); @@ -2026,16 +2095,16 @@ int unit_watch_pid(Unit *u, pid_t pid) { if (r < 0) return r; - r = hashmap_put(u->manager->watch_pids1, LONG_TO_PTR(pid), u); + 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, LONG_TO_PTR(pid), u); + r = hashmap_put(u->manager->watch_pids2, PID_TO_PTR(pid), u); } - q = set_put(u->pids, LONG_TO_PTR(pid)); + q = set_put(u->pids, PID_TO_PTR(pid)); if (q < 0) return q; @@ -2046,81 +2115,18 @@ void unit_unwatch_pid(Unit *u, pid_t pid) { assert(u); assert(pid >= 1); - hashmap_remove_value(u->manager->watch_pids1, LONG_TO_PTR(pid), u); - hashmap_remove_value(u->manager->watch_pids2, LONG_TO_PTR(pid), u); - set_remove(u->pids, LONG_TO_PTR(pid)); + (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_LONG(set_first(u->pids))); + unit_unwatch_pid(u, PTR_TO_PID(set_first(u->pids))); - set_free(u->pids); - u->pids = NULL; -} - -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); - - /* Adds all PIDs from a specific cgroup path to the set of PIDs we watch. */ - - r = cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, path, &f); - if (r >= 0) { - 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; - - } else if (ret >= 0) - ret = r; - - r = cg_enumerate_subgroups(SYSTEMD_CGROUP_CONTROLLER, path, &d); - if (r >= 0) { - 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; - - } else if (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 */ - - if (!u->cgroup_path) - return -ENOENT; - - return unit_watch_pids_in_path(u, u->cgroup_path); + u->pids = set_free(u->pids); } void unit_tidy_watch_pids(Unit *u, pid_t except1, pid_t except2) { @@ -2132,7 +2138,7 @@ void unit_tidy_watch_pids(Unit *u, pid_t except1, pid_t except2) { /* Cleans dead PIDs from our list */ SET_FOREACH(e, u->pids, i) { - pid_t pid = PTR_TO_LONG(e); + pid_t pid = PTR_TO_PID(e); if (pid == except1 || pid == except2) continue; @@ -2159,6 +2165,7 @@ bool unit_job_is_applicable(Unit *u, JobType j) { return unit_can_start(u); case JOB_RELOAD: + case JOB_TRY_RELOAD: return unit_can_reload(u); case JOB_RELOAD_OR_START: @@ -2186,16 +2193,12 @@ int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_referen static const UnitDependency inverse_table[_UNIT_DEPENDENCY_MAX] = { [UNIT_REQUIRES] = UNIT_REQUIRED_BY, - [UNIT_REQUIRES_OVERRIDABLE] = UNIT_REQUIRED_BY_OVERRIDABLE, [UNIT_WANTS] = UNIT_WANTED_BY, [UNIT_REQUISITE] = UNIT_REQUISITE_OF, - [UNIT_REQUISITE_OVERRIDABLE] = UNIT_REQUISITE_OF_OVERRIDABLE, [UNIT_BINDS_TO] = UNIT_BOUND_BY, [UNIT_PART_OF] = UNIT_CONSISTS_OF, [UNIT_REQUIRED_BY] = UNIT_REQUIRES, - [UNIT_REQUIRED_BY_OVERRIDABLE] = UNIT_REQUIRES_OVERRIDABLE, [UNIT_REQUISITE_OF] = UNIT_REQUISITE, - [UNIT_REQUISITE_OF_OVERRIDABLE] = UNIT_REQUISITE_OVERRIDABLE, [UNIT_WANTED_BY] = UNIT_WANTS, [UNIT_BOUND_BY] = UNIT_BINDS_TO, [UNIT_CONSISTS_OF] = UNIT_PART_OF, @@ -2229,6 +2232,11 @@ int unit_add_dependency(Unit *u, UnitDependency d, Unit *other, bool add_referen 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; @@ -2374,47 +2382,9 @@ int unit_add_two_dependencies_by_name(Unit *u, UnitDependency d, UnitDependency return unit_add_two_dependencies(u, d, e, other, add_reference); } -int unit_add_dependency_by_name_inverse(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(other, d, u, add_reference); -} - -int unit_add_two_dependencies_by_name_inverse(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(other, d, e, u, add_reference); -} - int set_unit_path(const char *p) { /* This is mostly for debug purposes */ - if (setenv("SYSTEMD_UNIT_PATH", p, 0) < 0) + if (setenv("SYSTEMD_UNIT_PATH", p, 1) < 0) return -errno; return 0; @@ -2429,39 +2399,60 @@ char *unit_dbus_path(Unit *u) { return unit_dbus_path_from_name(u->id); } -char *unit_default_cgroup_path(Unit *u) { - _cleanup_free_ char *escaped = NULL, *slice = NULL; - int r; +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); - if (unit_has_name(u, SPECIAL_ROOT_SLICE)) - return strdup(u->manager->cgroup_root); + /* 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_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; - } + if (!UNIT_HAS_CGROUP_CONTEXT(u)) + return -EOPNOTSUPP; - escaped = cg_escape(u->id); - if (!escaped) - return NULL; + if (u->type == UNIT_SLICE) + return -EINVAL; - if (slice) - return strjoin(u->manager->cgroup_root, "/", slice, "/", escaped, NULL); - else - return strjoin(u->manager->cgroup_root, "/", escaped, NULL); + 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_add_default_slice(Unit *u, CGroupContext *c) { +int unit_set_default_slice(Unit *u) { _cleanup_free_ char *b = NULL; const char *slice_name; Unit *slice; int r; assert(u); - assert(c); if (UNIT_ISSET(u->slice)) return 0; @@ -2483,7 +2474,7 @@ int unit_add_default_slice(Unit *u, CGroupContext *c) { if (!escaped) return -ENOMEM; - if (u->manager->running_as == MANAGER_SYSTEM) + if (MANAGER_IS_SYSTEM(u->manager)) b = strjoin("system-", escaped, ".slice", NULL); else b = strappend(escaped, ".slice"); @@ -2493,7 +2484,7 @@ int unit_add_default_slice(Unit *u, CGroupContext *c) { slice_name = b; } else slice_name = - u->manager->running_as == MANAGER_SYSTEM + MANAGER_IS_SYSTEM(u->manager) && !unit_has_name(u, SPECIAL_INIT_SCOPE) ? SPECIAL_SYSTEM_SLICE : SPECIAL_ROOT_SLICE; @@ -2501,8 +2492,7 @@ int unit_add_default_slice(Unit *u, CGroupContext *c) { if (r < 0) return r; - unit_ref_set(&u->slice, slice); - return 0; + return unit_set_slice(u, slice); } const char *unit_slice_name(Unit *u) { @@ -2533,14 +2523,70 @@ int unit_load_related_unit(Unit *u, const char *type, Unit **_found) { 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. */ - return hashmap_put(u->manager->watch_bus, name, u); + 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) { @@ -2548,6 +2594,7 @@ void unit_unwatch_bus_name(Unit *u, const char *name) { 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) { @@ -2578,10 +2625,13 @@ int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { } } + 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); @@ -2592,21 +2642,34 @@ int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { 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, "cpuacct-usage-base", "%" PRIu64, u->cpuacct_usage_base); + + 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, fds); + job_serialize(u->job, f); } if (u->nop_job) { fprintf(f, "job\n"); - job_serialize(u->nop_job, f, fds); + job_serialize(u->nop_job, f); } } @@ -2615,65 +2678,78 @@ int unit_serialize(Unit *u, FILE *f, FDSet *fds, bool serialize_jobs) { return 0; } -void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *format, ...) { - va_list ap; - +int unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value) { assert(u); assert(f); assert(key); - assert(format); + + if (!value) + return 0; fputs(key, f); fputc('=', f); - - va_start(ap, format); - vfprintf(f, format, ap); - va_end(ap); - + fputs(value, f); fputc('\n', f); + + return 1; } -void 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) { + _cleanup_free_ char *c = NULL; + assert(u); assert(f); assert(key); - assert(value); - fprintf(f, "%s=%s\n", key, value); + if (!value) + return 0; + + c = cescape(value); + if (!c) + return -ENOMEM; + + fputs(key, f); + fputc('=', f); + fputs(c, f); + fputc('\n', f); + + return 1; } -static int unit_set_cgroup_path(Unit *u, const char *path) { - _cleanup_free_ char *p = NULL; - int r; +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 (path) { - p = strdup(path); - if (!p) - return -ENOMEM; - } else - p = NULL; - - if (streq_ptr(u->cgroup_path, p)) + if (fd < 0) return 0; - if (p) { - r = hashmap_put(u->manager->cgroup_unit, p, u); - if (r < 0) - return r; - } + copy = fdset_put_dup(fds, fd); + if (copy < 0) + return copy; - if (u->cgroup_path) { - log_unit_debug(u, "Changing cgroup path from %s to %s.", u->cgroup_path, strna(p)); - hashmap_remove(u->manager->cgroup_unit, u->cgroup_path); - free(u->cgroup_path); - } + 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; - u->cgroup_path = p; - p = NULL; + assert(u); + assert(f); + assert(key); + assert(format); - return 0; + 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) { @@ -2704,7 +2780,7 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { /* End marker */ if (isempty(l)) - return 0; + break; k = strcspn(l, "="); @@ -2723,7 +2799,7 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { if (!j) return log_oom(); - r = job_deserialize(j, f, fds); + r = job_deserialize(j, f); if (r < 0) { job_free(j); return r; @@ -2744,6 +2820,9 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { } 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; @@ -2792,11 +2871,19 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { continue; - } else if (streq(l, "cpuacct-usage-base")) { + } else if (STR_IN_SET(l, "cpu-usage-base", "cpuacct-usage-base")) { - r = safe_atou64(v, &u->cpuacct_usage_base); + r = safe_atou64(v, &u->cpu_usage_base); if (r < 0) - log_unit_debug(u, "Failed to parse CPU usage %s, ignoring.", v); + 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; @@ -2806,6 +2893,8 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { 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; @@ -2817,6 +2906,47 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { 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)) { @@ -2837,9 +2967,18 @@ int unit_deserialize(Unit *u, FILE *f, FDSet *fds) { 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) { +int unit_add_node_link(Unit *u, const char *what, bool wants, UnitDependency dep) { Unit *device; _cleanup_free_ char *e = NULL; int r; @@ -2866,7 +3005,9 @@ int unit_add_node_link(Unit *u, const char *what, bool wants) { if (r < 0) return r; - r = unit_add_two_dependencies(u, UNIT_AFTER, u->manager->running_as == MANAGER_SYSTEM ? UNIT_BINDS_TO : UNIT_WANTS, device, true); + r = unit_add_two_dependencies(u, UNIT_AFTER, + MANAGER_IS_SYSTEM(u->manager) ? dep : UNIT_WANTS, + device, true); if (r < 0) return r; @@ -2880,7 +3021,8 @@ int unit_add_node_link(Unit *u, const char *what, bool wants) { } int unit_coldplug(Unit *u) { - int r; + int r = 0, q; + char **i; assert(u); @@ -2891,81 +3033,69 @@ int unit_coldplug(Unit *u) { 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) { - r = UNIT_VTABLE(u)->coldplug(u); - if (r < 0) - return r; + q = UNIT_VTABLE(u)->coldplug(u); + if (q < 0 && r >= 0) + r = q; } if (u->job) { - r = job_coldplug(u->job); - if (r < 0) - return r; + q = job_coldplug(u->job); + if (q < 0 && r >= 0) + r = q; } - return 0; + return r; } -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; +static bool fragment_mtime_newer(const char *path, usec_t mtime) { + struct stat st; + + if (!path) + return false; + + if (stat(path, &st) < 0) + /* What, cannot access this anymore? */ + return true; + + if (mtime > 0) + /* For non-empty files check the mtime */ + return timespec_load(&st.st_mtim) > mtime; + else if (!null_or_empty(&st)) + /* For masked files check if they are still so */ + return true; + + return false; } bool unit_need_daemon_reload(Unit *u) { _cleanup_strv_free_ char **t = NULL; char **path; - struct stat st; - unsigned loaded_cnt, current_cnt; assert(u); - if (u->fragment_path) { - zero(st); - if (stat(u->fragment_path, &st) < 0) - /* What, cannot access this anymore? */ - return true; - - if (u->fragment_mtime > 0 && - timespec_load(&st.st_mtim) != u->fragment_mtime) - return true; - } - - if (u->source_path) { - zero(st); - if (stat(u->source_path, &st) < 0) - return true; + if (fragment_mtime_newer(u->fragment_path, u->fragment_mtime)) + return true; - if (u->source_mtime > 0 && - timespec_load(&st.st_mtim) != u->source_mtime) - return true; - } + if (fragment_mtime_newer(u->source_path, u->source_mtime)) + return true; (void) unit_find_dropin_paths(u, &t); - loaded_cnt = strv_length(t); - current_cnt = strv_length(u->dropin_paths); - - if (loaded_cnt == current_cnt) { - if (loaded_cnt == 0) - return false; - - if (strv_overlap(u->dropin_paths, t)) { - STRV_FOREACH(path, u->dropin_paths) { - zero(st); - if (stat(*path, &st) < 0) - return true; - - if (u->dropin_mtime > 0 && - timespec_load(&st.st_mtim) > u->dropin_mtime) - return true; - } + if (!strv_equal(u->dropin_paths, t)) + return true; - return false; - } else + STRV_FOREACH(path, u->dropin_paths) + if (fragment_mtime_newer(*path, u->dropin_mtime)) return true; - } else - return true; + + return false; } void unit_reset_failed(Unit *u) { @@ -2973,6 +3103,9 @@ void unit_reset_failed(Unit *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) { @@ -3030,8 +3163,7 @@ bool unit_active_or_pending(Unit *u) { int unit_kill(Unit *u, KillWho w, int signo, sd_bus_error *error) { assert(u); assert(w >= 0 && w < _KILL_WHO_MAX); - assert(signo > 0); - assert(signo < _NSIG); + assert(SIGNAL_VALID(signo)); if (!UNIT_VTABLE(u)->kill) return -EOPNOTSUPP; @@ -3049,13 +3181,13 @@ static Set *unit_pid_set(pid_t main_pid, pid_t control_pid) { /* Exclude the main/control pids from being killed via the cgroup */ if (main_pid > 0) { - r = set_put(pid_set, LONG_TO_PTR(main_pid)); + r = set_put(pid_set, PID_TO_PTR(main_pid)); if (r < 0) goto fail; } if (control_pid > 0) { - r = set_put(pid_set, LONG_TO_PTR(control_pid)); + r = set_put(pid_set, PID_TO_PTR(control_pid)); if (r < 0) goto fail; } @@ -3076,32 +3208,39 @@ int unit_kill_common( sd_bus_error *error) { int r = 0; + bool killed = false; - if (who == KILL_MAIN && main_pid <= 0) { + 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 + else if (main_pid == 0) return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No main process to kill"); } - if (who == KILL_CONTROL && control_pid <= 0) { + 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 + else if (control_pid == 0) return sd_bus_error_set_const(error, BUS_ERROR_NO_SUCH_PROCESS, "No control process to kill"); } - if (who == KILL_CONTROL || who == KILL_ALL) - if (control_pid > 0) + 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 (who == KILL_MAIN || who == KILL_ALL) - if (main_pid > 0) + 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 (who == KILL_ALL && u->cgroup_path) { + if (IN_SET(who, KILL_ALL, KILL_ALL_FAIL) && u->cgroup_path) { _cleanup_set_free_ Set *pid_set = NULL; int q; @@ -3110,11 +3249,16 @@ int unit_kill_common( if (!pid_set) return -ENOMEM; - q = cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, signo, false, true, false, pid_set); + 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; } @@ -3130,12 +3274,19 @@ int unit_following_set(Unit *u, Set **s) { } UnitFileState unit_get_unit_file_state(Unit *u) { + int r; + assert(u); - if (u->unit_file_state < 0 && u->fragment_path) - u->unit_file_state = unit_file_get_state( - u->manager->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, - NULL, basename(u->fragment_path)); + 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; } @@ -3145,8 +3296,9 @@ int unit_get_unit_file_preset(Unit *u) { if (u->unit_file_preset < 0 && u->fragment_path) u->unit_file_preset = unit_file_query_preset( - u->manager->running_as == MANAGER_SYSTEM ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, - NULL, basename(u->fragment_path)); + u->manager->unit_file_scope, + NULL, + basename(u->fragment_path)); return u->unit_file_preset; } @@ -3169,10 +3321,41 @@ void unit_ref_unset(UnitRef *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; @@ -3195,7 +3378,7 @@ int unit_patch_contexts(Unit *u) { return -ENOMEM; } - if (u->manager->running_as == MANAGER_USER && + if (MANAGER_IS_USER(u->manager) && !ec->working_directory) { r = get_home_dir(&ec->working_directory); @@ -3207,7 +3390,7 @@ int unit_patch_contexts(Unit *u) { ec->working_directory_missing_ok = true; } - if (u->manager->running_as == MANAGER_USER && + if (MANAGER_IS_USER(u->manager) && (ec->syscall_whitelist || !set_isempty(ec->syscall_filter) || !set_isempty(ec->syscall_archs) || @@ -3216,7 +3399,30 @@ int unit_patch_contexts(Unit *u) { ec->no_new_privileges = true; if (ec->private_devices) - ec->capability_bounding_set_drop |= (uint64_t) 1ULL << (uint64_t) CAP_MKNOD; + ec->capability_bounding_set &= ~(UINT64_C(1) << CAP_MKNOD); + + 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); @@ -3285,71 +3491,65 @@ ExecRuntime *unit_get_exec_runtime(Unit *u) { return *(ExecRuntime**) ((uint8_t*) u + offset); } -static int unit_drop_in_dir(Unit *u, UnitSetPropertiesMode mode, bool transient, char **dir) { - if (u->manager->running_as == MANAGER_USER) { - int r; - - if (mode == UNIT_PERSISTENT && !transient) - r = user_config_home(dir); - else - r = user_runtime_dir(dir); +static const char* unit_drop_in_dir(Unit *u, UnitSetPropertiesMode mode) { + assert(u); - if (r == 0) - return -ENOENT; - return r; - } + if (!IN_SET(mode, UNIT_RUNTIME, UNIT_PERSISTENT)) + return NULL; - if (mode == UNIT_PERSISTENT && !transient) - *dir = strdup("/etc/systemd/system"); - else - *dir = strdup("/run/systemd/system"); - if (!*dir) - return -ENOMEM; + if (u->transient) /* Redirect drop-ins for transient units always into the transient directory. */ + return u->manager->lookup_paths.transient; - return 0; -} + if (mode == UNIT_RUNTIME) + return u->manager->lookup_paths.runtime_control; -static int unit_drop_in_file(Unit *u, - UnitSetPropertiesMode mode, const char *name, char **p, char **q) { - _cleanup_free_ char *dir = NULL; - int r; + if (mode == UNIT_PERSISTENT) + return u->manager->lookup_paths.persistent_control; - assert(u); - - r = unit_drop_in_dir(u, mode, u->transient, &dir); - if (r < 0) - return r; - - return drop_in_file(dir, u->id, 50, name, p, q); + return NULL; } int unit_write_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data) { - - _cleanup_free_ char *dir = NULL, *p = NULL, *q = NULL; + _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; - r = unit_drop_in_dir(u, mode, u->transient, &dir); - if (r < 0) - return r; + 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 = write_drop_in(dir, u->id, 50, name, data); + r = drop_in_file(dir, u->id, 50, name, &p, &q); if (r < 0) return r; - r = drop_in_file(dir, u->id, 50, name, &p, &q); + (void) mkdir_p(p, 0755); + r = write_string_file_atomic_label(q, wrapped); if (r < 0) return r; - r = strv_extend(&u->dropin_paths, q); + r = strv_push(&u->dropin_paths, q); if (r < 0) return r; + q = NULL; - strv_sort(u->dropin_paths); strv_uniq(u->dropin_paths); u->dropin_mtime = now(CLOCK_REALTIME); @@ -3380,7 +3580,7 @@ int unit_write_drop_in_format(Unit *u, UnitSetPropertiesMode mode, const char *n } int unit_write_drop_in_private(Unit *u, UnitSetPropertiesMode mode, const char *name, const char *data) { - _cleanup_free_ char *ndata = NULL; + const char *ndata; assert(u); assert(name); @@ -3392,9 +3592,7 @@ int unit_write_drop_in_private(Unit *u, UnitSetPropertiesMode mode, const char * if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) return 0; - ndata = strjoin("[", UNIT_VTABLE(u)->private_section, "]\n", data, NULL); - if (!ndata) - return -ENOMEM; + ndata = strjoina("[", UNIT_VTABLE(u)->private_section, "]\n", data); return unit_write_drop_in(u, mode, name, ndata); } @@ -3421,63 +3619,89 @@ int unit_write_drop_in_private_format(Unit *u, UnitSetPropertiesMode mode, const return unit_write_drop_in_private(u, mode, name, p); } -int unit_remove_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name) { - _cleanup_free_ char *p = NULL, *q = NULL; - int r; +int unit_make_transient(Unit *u) { + FILE *f; + char *path; assert(u); - if (!IN_SET(mode, UNIT_PERSISTENT, UNIT_RUNTIME)) - return 0; + if (!UNIT_VTABLE(u)->can_transient) + return -EOPNOTSUPP; - r = unit_drop_in_file(u, mode, name, &p, &q); - if (r < 0) - return r; + path = strjoin(u->manager->lookup_paths.transient, "/", u->id, NULL); + if (!path) + return -ENOMEM; - if (unlink(q) < 0) - r = errno == ENOENT ? 0 : -errno; - else - r = 1; + /* 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. */ - rmdir(p); - return r; -} + RUN_WITH_UMASK(0022) { + f = fopen(path, "we"); + if (!f) { + free(path); + return -errno; + } + } -int unit_make_transient(Unit *u) { - int r; + if (u->transient_file) + fclose(u->transient_file); + u->transient_file = f; - assert(u); + 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; - free(u->fragment_path); - u->fragment_path = NULL; + unit_add_to_dbus_queue(u); + unit_add_to_gc_queue(u); - if (u->manager->running_as == MANAGER_USER) { - _cleanup_free_ char *c = NULL; + fputs("# This is a transient unit file, created programmatically via the systemd API. Do not edit.\n", + u->transient_file); - r = user_runtime_dir(&c); - if (r < 0) - return r; - if (r == 0) - return -ENOENT; + return 0; +} - u->fragment_path = strjoin(c, "/", u->id, NULL); - if (!u->fragment_path) - return -ENOMEM; +static void log_kill(pid_t pid, int sig, void *userdata) { + _cleanup_free_ char *comm = NULL; - mkdir_p(c, 0755); - } else { - u->fragment_path = strappend("/run/systemd/system/", u->id); - if (!u->fragment_path) - return -ENOMEM; + (void) get_process_comm(pid, &comm); - mkdir_p("/run/systemd/system", 0755); - } + /* 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); - return write_string_file_atomic_label(u->fragment_path, "# Transient stub"); + 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( @@ -3488,62 +3712,69 @@ int unit_kill_context( pid_t control_pid, bool main_pid_alien) { - int sig, wait_for_exit = false, r; + 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; - switch (k) { - case KILL_KILL: - sig = SIGKILL; - break; - case KILL_ABORT: - sig = SIGABRT; - break; - case KILL_TERMINATE: - sig = c->kill_signal; - break; - default: - assert_not_reached("KillOperation unknown"); - } + 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) { - r = kill_and_sigcont(main_pid, sig); + 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; - get_process_comm(main_pid, &comm); + (void) get_process_comm(main_pid, &comm); - log_unit_warning_errno(u, r, "Failed to kill main process " PID_FMT " (%s): %m", main_pid, strna(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 (c->send_sighup && k != KILL_KILL) - kill(main_pid, SIGHUP); + if (r != -ESRCH && send_sighup) + (void) kill(main_pid, SIGHUP); } } if (control_pid > 0) { - r = kill_and_sigcont(control_pid, sig); + 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; - get_process_comm(control_pid, &comm); + (void) get_process_comm(control_pid, &comm); - log_unit_warning_errno(u, r, "Failed to kill control process " PID_FMT " (%s): %m", control_pid, strna(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 (c->send_sighup && k != KILL_KILL) - kill(control_pid, SIGHUP); + if (r != -ESRCH && send_sighup) + (void) kill(control_pid, SIGHUP); } } - if ((c->kill_mode == KILL_CONTROL_GROUP || (c->kill_mode == KILL_MIXED && k == KILL_KILL)) && u->cgroup_path) { + 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 */ @@ -3551,30 +3782,47 @@ int unit_kill_context( if (!pid_set) return -ENOMEM; - r = cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, sig, true, true, false, pid_set); + 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: %m"); - } else if (r > 0) { + log_unit_warning_errno(u, r, "Failed to kill control group %s, ignoring: %m", u->cgroup_path); - /* FIXME: For now, we will not wait for the - * cgroup members to die, simply because - * cgroup notification is unreliable. It - * doesn't work at all in containers, and - * outside of containers it can be confused - * easily by leaving directories in the - * cgroup. */ + } else if (r > 0) { - /* wait_for_exit = true; */ + /* 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 (c->send_sighup && k != KILL_KILL) { + 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, false, true, false, pid_set); + cg_kill_recursive(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, + SIGHUP, + CGROUP_IGNORE_SELF, + pid_set, + NULL, NULL); } } } @@ -3683,6 +3931,26 @@ int unit_setup_exec_runtime(Unit *u) { 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; @@ -3741,13 +4009,233 @@ int unit_fail_if_symlink(Unit *u, const char* where) { return -ELOOP; } -static const char* const unit_active_state_table[_UNIT_ACTIVE_STATE_MAX] = { - [UNIT_ACTIVE] = "active", - [UNIT_RELOADING] = "reloading", - [UNIT_INACTIVE] = "inactive", - [UNIT_FAILED] = "failed", - [UNIT_ACTIVATING] = "activating", - [UNIT_DEACTIVATING] = "deactivating" -}; +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; +} -DEFINE_STRING_TABLE_LOOKUP(unit_active_state, UnitActiveState); +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; +} diff --git a/src/core/unit.h b/src/core/unit.h index 9491ef64f9..a8dd3e602c 100644 --- a/src/core/unit.h +++ b/src/core/unit.h @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - #pragma once /*** @@ -27,29 +25,18 @@ typedef struct Unit Unit; typedef struct UnitVTable UnitVTable; -typedef enum UnitActiveState UnitActiveState; typedef struct UnitRef UnitRef; typedef struct UnitStatusMessageFormats UnitStatusMessageFormats; -#include "list.h" #include "condition.h" +#include "failure-action.h" #include "install.h" +#include "list.h" #include "unit-name.h" -#include "failure-action.h" - -enum UnitActiveState { - UNIT_ACTIVE, - UNIT_RELOADING, - UNIT_INACTIVE, - UNIT_FAILED, - UNIT_ACTIVATING, - UNIT_DEACTIVATING, - _UNIT_ACTIVE_STATE_MAX, - _UNIT_ACTIVE_STATE_INVALID = -1 -}; typedef enum KillOperation { KILL_TERMINATE, + KILL_TERMINATE_AND_LOG, KILL_KILL, KILL_ABORT, _KILL_OPERATION_MAX, @@ -109,12 +96,22 @@ struct Unit { 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; FailureAction job_timeout_action; @@ -130,6 +127,10 @@ struct Unit { 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; @@ -158,36 +159,58 @@ struct Unit { /* 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; + FailureAction 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 cpuacct.usage cgroup counter was at the time the unit was started */ - nsec_t cpuacct_usage_base; + /* 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; - CGroupControllerMask cgroup_realized_mask; - CGroupControllerMask cgroup_subtree_mask; - CGroupControllerMask cgroup_members_mask; + 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; @@ -206,9 +229,6 @@ struct Unit { /* Ignore this unit when isolating */ bool ignore_on_isolate; - /* Ignore this unit when snapshotting */ - bool ignore_on_snapshot; - /* Did the last condition check succeed? */ bool condition_result; bool assert_result; @@ -232,8 +252,13 @@ struct Unit { 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 { @@ -248,17 +273,16 @@ typedef enum UnitSetPropertiesMode { UNIT_PERSISTENT = 2, } UnitSetPropertiesMode; -#include "socket.h" +#include "automount.h" #include "busname.h" -#include "target.h" -#include "snapshot.h" #include "device.h" -#include "automount.h" -#include "swap.h" -#include "timer.h" -#include "slice.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 */ @@ -281,6 +305,10 @@ struct UnitVTable { * 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; @@ -325,7 +353,7 @@ struct UnitVTable { 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 */ - int (*distribute_fds)(Unit *u, FDSet *fds); + 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 */ @@ -346,9 +374,6 @@ struct UnitVTable { * shall release its runtime resources */ void (*release_resources)(Unit *u); - /* Return true when this unit is suitable for snapshotting */ - bool (*check_snapshot)(Unit *u); - /* Invoked on every child that died */ void (*sigchld_event)(Unit *u, pid_t pid, int code, int status); @@ -362,8 +387,7 @@ struct UnitVTable { /* 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. */ + /* 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 */ @@ -385,14 +409,21 @@ struct UnitVTable { /* Called whenever CLOCK_REALTIME made a jump */ void (*time_change)(Unit *u); - int (*get_timeout)(Unit *u, uint64_t *timeout); + /* 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. */ - int (*enumerate)(Manager *m); + void (*enumerate)(Manager *m); /* Type specific cleanups. */ void (*shutdown)(Manager *m); @@ -401,24 +432,12 @@ struct UnitVTable { * of this type will immediately fail. */ bool (*supported)(void); - /* The interface name */ - const char *bus_interface; - /* The bus vtable */ const sd_bus_vtable *bus_vtable; /* The strings to print in status messages */ UnitStatusMessageFormats status_message_formats; - /* Can units of this type have multiple names? */ - bool no_alias:1; - - /* Instances make no sense for this type */ - bool no_instances:1; - - /* Exclude from automatic gc */ - bool no_gc:1; - /* True if transient units of this type are OK */ bool can_transient:1; }; @@ -439,13 +458,16 @@ extern const UnitVTable * const unit_vtable[_UNIT_TYPE_MAX]; /* 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(SNAPSHOT, Snapshot); DEFINE_CAST(DEVICE, Device); DEFINE_CAST(MOUNT, Mount); DEFINE_CAST(AUTOMOUNT, Automount); @@ -466,9 +488,6 @@ int unit_add_two_dependencies(Unit *u, UnitDependency d, UnitDependency e, Unit 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_dependency_by_name_inverse(Unit *u, UnitDependency d, const char *name, const char *filename, bool add_reference); -int unit_add_two_dependencies_by_name_inverse(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); @@ -490,7 +509,8 @@ int unit_load_fragment_and_dropin(Unit *u); int unit_load_fragment_and_dropin_optional(Unit *u); int unit_load(Unit *unit); -int unit_add_default_slice(Unit *u, CGroupContext *c); +int unit_set_slice(Unit *u, Unit *slice); +int unit_set_default_slice(Unit *u); const char *unit_description(Unit *u) _pure_; @@ -517,11 +537,11 @@ void unit_notify(Unit *u, UnitActiveState os, UnitActiveState ns, bool reload_su int unit_watch_pid(Unit *u, pid_t pid); void unit_unwatch_pid(Unit *u, pid_t pid); -int unit_watch_all_pids(Unit *u); 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); @@ -530,20 +550,26 @@ 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); -void unit_serialize_item_format(Unit *u, FILE *f, const char *key, const char *value, ...) _printf_(4,5); -void unit_serialize_item(Unit *u, FILE *f, const char *key, const char *value); int unit_deserialize(Unit *u, FILE *f, FDSet *fds); -int unit_add_node_link(Unit *u, const char *what, bool wants); +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); @@ -560,8 +586,6 @@ bool unit_active_or_pending(Unit *u); int unit_add_default_target_dependency(Unit *u, Unit *target); -char *unit_default_cgroup_path(Unit *u); - void unit_start_on_failure(Unit *u); void unit_trigger_notify(Unit *u); @@ -583,6 +607,7 @@ 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); @@ -590,8 +615,6 @@ int unit_write_drop_in_format(Unit *u, UnitSetPropertiesMode mode, const char *n 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_remove_drop_in(Unit *u, UnitSetPropertiesMode mode, const char *name); - 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); @@ -600,6 +623,11 @@ 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); } @@ -607,15 +635,28 @@ static inline bool unit_supported(Unit *u) { void unit_warn_if_dir_nonempty(Unit *u, const char* where); int unit_fail_if_symlink(Unit *u, const char* where); -const char *unit_active_state_to_string(UnitActiveState i) _const_; -UnitActiveState unit_active_state_from_string(const char *s) _pure_; +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, ...) \ ({ \ - Unit *_u = (unit); \ - _u ? log_object_internal(level, error, __FILE__, __LINE__, __func__, _u->manager->unit_log_field, _u->id, ##__VA_ARGS__) : \ + 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__); \ }) diff --git a/src/core/user.conf b/src/core/user.conf index 87c8164378..b427f1ef6d 100644 --- a/src/core/user.conf +++ b/src/core/user.conf @@ -23,7 +23,7 @@ #DefaultTimeoutStartSec=90s #DefaultTimeoutStopSec=90s #DefaultRestartSec=100ms -#DefaultStartLimitInterval=10s +#DefaultStartLimitIntervalSec=10s #DefaultStartLimitBurst=5 #DefaultEnvironment= #DefaultLimitCPU= |