diff options
Diffstat (limited to 'src/nspawn/nspawn.c')
-rw-r--r-- | src/nspawn/nspawn.c | 4554 |
1 files changed, 1905 insertions, 2649 deletions
diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 3428109da4..14af51fc0e 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -1,5 +1,3 @@ -/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ - /*** This file is part of systemd. @@ -19,95 +17,99 @@ along with systemd; If not, see <http://www.gnu.org/licenses/>. ***/ -#include <signal.h> -#include <sched.h> -#include <unistd.h> -#include <sys/types.h> -#include <sys/mount.h> -#include <stdlib.h> -#include <string.h> -#include <stdio.h> +#ifdef HAVE_BLKID +#include <blkid/blkid.h> +#endif #include <errno.h> -#include <sys/prctl.h> #include <getopt.h> #include <grp.h> -#include <linux/fs.h> -#include <sys/socket.h> -#include <linux/netlink.h> -#include <net/if.h> -#include <linux/veth.h> -#include <sys/personality.h> #include <linux/loop.h> -#include <sys/file.h> - +#include <pwd.h> +#include <sched.h> #ifdef HAVE_SELINUX #include <selinux/selinux.h> #endif - -#ifdef HAVE_SECCOMP -#include <seccomp.h> -#endif - -#ifdef HAVE_BLKID -#include <blkid/blkid.h> -#endif +#include <signal.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <sys/file.h> +#include <sys/mount.h> +#include <sys/personality.h> +#include <sys/prctl.h> +#include <sys/types.h> +#include <unistd.h> #include "sd-daemon.h" -#include "sd-bus.h" #include "sd-id128.h" -#include "sd-netlink.h" -#include "random-util.h" -#include "log.h" -#include "util.h" -#include "mkdir.h" -#include "rm-rf.h" -#include "macro.h" -#include "missing.h" + +#include "alloc-util.h" +#include "barrier.h" +#include "base-filesystem.h" +#include "blkid-util.h" +#include "btrfs-util.h" +#include "cap-list.h" +#include "capability-util.h" #include "cgroup-util.h" -#include "strv.h" -#include "path-util.h" -#include "loopback-setup.h" +#include "copy.h" #include "dev-setup.h" +#include "env-util.h" +#include "fd-util.h" #include "fdset.h" -#include "build.h" #include "fileio.h" -#include "bus-util.h" -#include "bus-error.h" -#include "ptyfwd.h" -#include "env-util.h" -#include "netlink-util.h" -#include "udev-util.h" -#include "blkid-util.h" +#include "formats-util.h" +#include "fs-util.h" #include "gpt.h" -#include "siphash24.h" -#include "copy.h" -#include "base-filesystem.h" -#include "barrier.h" -#include "event-util.h" -#include "capability.h" -#include "cap-list.h" -#include "btrfs-util.h" +#include "hostname-util.h" +#include "id128-util.h" +#include "log.h" +#include "loopback-setup.h" #include "machine-image.h" -#include "list.h" -#include "in-addr-util.h" -#include "firewall-util.h" -#include "local-addresses.h" -#include "formats-util.h" +#include "macro.h" +#include "missing.h" +#include "mkdir.h" +#include "mount-util.h" +#include "netlink-util.h" +#include "nspawn-cgroup.h" +#include "nspawn-expose-ports.h" +#include "nspawn-mount.h" +#include "nspawn-network.h" +#include "nspawn-patch-uid.h" +#include "nspawn-register.h" +#include "nspawn-seccomp.h" +#include "nspawn-settings.h" +#include "nspawn-setuid.h" +#include "nspawn-stub-pid1.h" +#include "parse-util.h" +#include "path-util.h" #include "process-util.h" -#include "terminal-util.h" -#include "hostname-util.h" +#include "ptyfwd.h" +#include "random-util.h" +#include "raw-clone.h" +#include "rm-rf.h" +#include "selinux-util.h" #include "signal-util.h" +#include "socket-util.h" +#include "stat-util.h" +#include "stdio-util.h" +#include "string-util.h" +#include "strv.h" +#include "terminal-util.h" +#include "udev-util.h" +#include "umask-util.h" +#include "user-util.h" +#include "util.h" -#ifdef HAVE_SECCOMP -#include "seccomp-util.h" -#endif +/* Note that devpts's gid= parameter parses GIDs as signed values, hence we stay away from the upper half of the 32bit + * UID range here. We leave a bit of room at the lower end and a lot of room at the upper end, so that other subsystems + * may have their own allocation ranges too. */ +#define UID_SHIFT_PICK_MIN ((uid_t) UINT32_C(0x00080000)) +#define UID_SHIFT_PICK_MAX ((uid_t) UINT32_C(0x6FFF0000)) -typedef struct ExposePort { - int protocol; - uint16_t host_port; - uint16_t container_port; - LIST_FIELDS(struct ExposePort, ports); -} ExposePort; +/* nspawn is listening on the socket at the path in the constant nspawn_notify_socket_path + * nspawn_notify_socket_path is relative to the container + * the init process in the container pid can send messages to nspawn following the sd_notify(3) protocol */ +#define NSPAWN_NOTIFY_SOCKET_PATH "/run/systemd/nspawn/notify" typedef enum ContainerStatus { CONTAINER_TERMINATED, @@ -121,30 +123,9 @@ typedef enum LinkJournal { LINK_GUEST } LinkJournal; -typedef enum Volatile { - VOLATILE_NO, - VOLATILE_YES, - VOLATILE_STATE, -} Volatile; - -typedef enum CustomMountType { - CUSTOM_MOUNT_BIND, - CUSTOM_MOUNT_TMPFS, - CUSTOM_MOUNT_OVERLAY, -} CustomMountType; - -typedef struct CustomMount { - CustomMountType type; - bool read_only; - char *source; /* for overlayfs this is the upper directory */ - char *destination; - char *options; - char *work_dir; - char **lower; -} CustomMount; - static char *arg_directory = NULL; static char *arg_template = NULL; +static char *arg_chdir = NULL; static char *arg_user = NULL; static sd_id128_t arg_uuid = {}; static char *arg_machine = NULL; @@ -153,11 +134,13 @@ static const char *arg_selinux_apifs_context = NULL; static const char *arg_slice = NULL; static bool arg_private_network = false; static bool arg_read_only = false; -static bool arg_boot = false; +static StartMode arg_start_mode = START_PID1; static bool arg_ephemeral = false; static LinkJournal arg_link_journal = LINK_AUTO; static bool arg_link_journal_try = false; -static uint64_t arg_retain = +static uint64_t arg_caps_retain = + (1ULL << CAP_AUDIT_CONTROL) | + (1ULL << CAP_AUDIT_WRITE) | (1ULL << CAP_CHOWN) | (1ULL << CAP_DAC_OVERRIDE) | (1ULL << CAP_DAC_READ_SEARCH) | @@ -167,43 +150,51 @@ static uint64_t arg_retain = (1ULL << CAP_KILL) | (1ULL << CAP_LEASE) | (1ULL << CAP_LINUX_IMMUTABLE) | + (1ULL << CAP_MKNOD) | (1ULL << CAP_NET_BIND_SERVICE) | (1ULL << CAP_NET_BROADCAST) | (1ULL << CAP_NET_RAW) | - (1ULL << CAP_SETGID) | (1ULL << CAP_SETFCAP) | + (1ULL << CAP_SETGID) | (1ULL << CAP_SETPCAP) | (1ULL << CAP_SETUID) | (1ULL << CAP_SYS_ADMIN) | + (1ULL << CAP_SYS_BOOT) | (1ULL << CAP_SYS_CHROOT) | (1ULL << CAP_SYS_NICE) | (1ULL << CAP_SYS_PTRACE) | - (1ULL << CAP_SYS_TTY_CONFIG) | (1ULL << CAP_SYS_RESOURCE) | - (1ULL << CAP_SYS_BOOT) | - (1ULL << CAP_AUDIT_WRITE) | - (1ULL << CAP_AUDIT_CONTROL) | - (1ULL << CAP_MKNOD); + (1ULL << CAP_SYS_TTY_CONFIG); static CustomMount *arg_custom_mounts = NULL; static unsigned arg_n_custom_mounts = 0; static char **arg_setenv = NULL; static bool arg_quiet = false; -static bool arg_share_system = false; static bool arg_register = true; static bool arg_keep_unit = false; static char **arg_network_interfaces = NULL; static char **arg_network_macvlan = NULL; static char **arg_network_ipvlan = NULL; static bool arg_network_veth = false; -static const char *arg_network_bridge = NULL; +static char **arg_network_veth_extra = NULL; +static char *arg_network_bridge = NULL; +static char *arg_network_zone = NULL; static unsigned long arg_personality = PERSONALITY_INVALID; static char *arg_image = NULL; -static Volatile arg_volatile = VOLATILE_NO; +static VolatileMode arg_volatile_mode = VOLATILE_NO; static ExposePort *arg_expose_ports = NULL; static char **arg_property = NULL; +static UserNamespaceMode arg_userns_mode = USER_NAMESPACE_NO; static uid_t arg_uid_shift = UID_INVALID, arg_uid_range = 0x10000U; -static bool arg_userns = false; +static bool arg_userns_chown = false; static int arg_kill_signal = 0; +static CGroupUnified arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_UNKNOWN; +static SettingsMask arg_settings_mask = 0; +static int arg_settings_trusted = -1; +static char **arg_parameters = NULL; +static const char *arg_container_service_name = "systemd-nspawn"; +static bool arg_notify_ready = false; +static bool arg_use_cgns = true; +static unsigned long arg_clone_ns_flags = CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS; static void help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" @@ -217,14 +208,18 @@ static void help(void) { " -x --ephemeral Run container with snapshot of root directory, and\n" " remove it after exit\n" " -i --image=PATH File system device or disk image for the container\n" + " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n" " -b --boot Boot up full system (i.e. invoke init)\n" + " --chdir=PATH Set working directory in the container\n" " -u --user=USER Run the command under specified user or uid\n" " -M --machine=NAME Set the machine name for the container\n" " --uuid=UUID Set a specific machine UUID for the container\n" " -S --slice=SLICE Place the container in the specified slice\n" " --property=NAME=VALUE Set scope unit property\n" + " -U --private-users=pick Run within user namespace, autoselect UID/GID range\n" " --private-users[=UIDBASE[:NUIDS]]\n" - " Run within user namespace\n" + " Similar, but with user configured UID/GID range\n" + " --private-user-chown Adjust OS tree ownership to private UID/GID range\n" " --private-network Disable network in container\n" " --network-interface=INTERFACE\n" " Assign an existing network interface to the\n" @@ -235,12 +230,16 @@ static void help(void) { " --network-ipvlan=INTERFACE\n" " Create a ipvlan network interface based on an\n" " existing network interface to the container\n" - " -n --network-veth Add a virtual ethernet connection between host\n" + " -n --network-veth Add a virtual Ethernet connection between host\n" " and container\n" + " --network-veth-extra=HOSTIF[:CONTAINERIF]\n" + " Add an additional virtual Ethernet link between\n" + " host and container\n" " --network-bridge=INTERFACE\n" - " Add a virtual ethernet connection between host\n" - " and container and add it to an existing bridge on\n" - " the host\n" + " Add a virtual Ethernet connection to the container\n" + " and attach it to an existing bridge on the host\n" + " --network-zone=NAME Similar, but attach the new interface to an\n" + " an automatically managed bridge interface\n" " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n" " Expose a container IP port on the host\n" " -Z --selinux-context=SECLABEL\n" @@ -253,83 +252,31 @@ static void help(void) { " capability\n" " --drop-capability=CAP Drop the specified capability from the default set\n" " --kill-signal=SIGNAL Select signal to use for shutting down PID 1\n" - " --link-journal=MODE Link up guest journal, one of no, auto, guest, host,\n" - " try-guest, try-host\n" + " --link-journal=MODE Link up guest journal, one of no, auto, guest, \n" + " host, try-guest, try-host\n" " -j Equivalent to --link-journal=try-guest\n" " --read-only Mount the root directory read-only\n" - " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" + " --bind=PATH[:PATH[:OPTIONS]]\n" + " Bind mount a file or directory from the host into\n" " the container\n" - " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n" + " --bind-ro=PATH[:PATH[:OPTIONS]\n" + " Similar, but creates a read-only bind mount\n" " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n" " --overlay=PATH[:PATH...]:PATH\n" " Create an overlay mount from the host to \n" " the container\n" " --overlay-ro=PATH[:PATH...]:PATH\n" " Similar, but creates a read-only overlay mount\n" - " --setenv=NAME=VALUE Pass an environment variable to PID 1\n" - " --share-system Share system namespaces with host\n" + " -E --setenv=NAME=VALUE Pass an environment variable to PID 1\n" " --register=BOOLEAN Register container as machine\n" " --keep-unit Do not register a scope for the machine, reuse\n" " the service unit nspawn is running in\n" " --volatile[=MODE] Run the system in volatile mode\n" + " --settings=BOOLEAN Load additional settings from .nspawn file\n" + " --notify-ready=BOOLEAN Receive notifications from the child init process\n" , program_invocation_short_name); } -static CustomMount* custom_mount_add(CustomMountType t) { - CustomMount *c, *ret; - - c = realloc(arg_custom_mounts, (arg_n_custom_mounts + 1) * sizeof(CustomMount)); - if (!c) - return NULL; - - arg_custom_mounts = c; - ret = arg_custom_mounts + arg_n_custom_mounts; - arg_n_custom_mounts++; - - *ret = (CustomMount) { .type = t }; - - return ret; -} - -static void custom_mount_free_all(void) { - unsigned i; - - for (i = 0; i < arg_n_custom_mounts; i++) { - CustomMount *m = &arg_custom_mounts[i]; - - free(m->source); - free(m->destination); - free(m->options); - - if (m->work_dir) { - (void) rm_rf(m->work_dir, REMOVE_ROOT|REMOVE_PHYSICAL); - free(m->work_dir); - } - - strv_free(m->lower); - } - - free(arg_custom_mounts); - arg_custom_mounts = NULL; - arg_n_custom_mounts = 0; -} - -static int custom_mount_compare(const void *a, const void *b) { - const CustomMount *x = a, *y = b; - int r; - - r = path_compare(x->destination, y->destination); - if (r != 0) - return r; - - if (x->type < y->type) - return -1; - if (x->type > y->type) - return 1; - - return 0; -} - static int custom_mounts_prepare(void) { unsigned i; int r; @@ -341,9 +288,15 @@ static int custom_mounts_prepare(void) { for (i = 0; i < arg_n_custom_mounts; i++) { CustomMount *m = &arg_custom_mounts[i]; - if (arg_userns && arg_uid_shift == UID_INVALID && path_equal(m->destination, "/")) { - log_error("--private-users with automatic UID shift may not be combined with custom root mounts."); - return -EINVAL; + if (path_equal(m->destination, "/") && arg_userns_mode != USER_NAMESPACE_NO) { + + if (arg_userns_chown) { + log_error("--private-users-chown may not be combined with custom root mounts."); + return -EINVAL; + } else if (arg_uid_shift == UID_INVALID) { + log_error("--private-users with automatic UID shift may not be combined with custom root mounts."); + return -EINVAL; + } } if (m->type != CUSTOM_MOUNT_OVERLAY) @@ -363,27 +316,68 @@ static int custom_mounts_prepare(void) { return 0; } -static int set_sanitized_path(char **b, const char *path) { - char *p; +static int detect_unified_cgroup_hierarchy(const char *directory) { + const char *e; + int r, all_unified, systemd_unified; - assert(b); - assert(path); - - p = canonicalize_file_name(path); - if (!p) { - if (errno != ENOENT) - return -errno; + /* Allow the user to control whether the unified hierarchy is used */ + e = getenv("UNIFIED_CGROUP_HIERARCHY"); + if (e) { + r = parse_boolean(e); + if (r < 0) + return log_error_errno(r, "Failed to parse $UNIFIED_CGROUP_HIERARCHY."); + if (r > 0) + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL; + else + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE; - p = path_make_absolute_cwd(path); - if (!p) - return -ENOMEM; + return 0; } - free(*b); - *b = path_kill_slashes(p); + 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, + "Failed to determine whether the unified cgroups hierarchy is used: %m"); + + /* Otherwise inherit the default from the host system */ + if (all_unified > 0) { + /* Unified cgroup hierarchy support was added in 230. Unfortunately the detection + * routine only detects 231, so we'll have a false negative here for 230. */ + r = systemd_installation_has_version(directory, 230); + if (r < 0) + return log_error_errno(r, "Failed to determine systemd version in container: %m"); + if (r > 0) + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL; + else + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE; + } else if (systemd_unified > 0) { + /* Mixed cgroup hierarchy support was added in 232 */ + r = systemd_installation_has_version(directory, 232); + if (r < 0) + return log_error_errno(r, "Failed to determine systemd version in container: %m"); + if (r > 0) + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD; + else + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE; + } else + arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE; + return 0; } +static void parse_share_ns_env(const char *name, unsigned long ns_flag) { + int r; + + r = getenv_bool(name); + if (r == -ENXIO) + return; + if (r < 0) + log_warning_errno(r, "Failed to parse %s from environment, defaulting to false.", name); + arg_clone_ns_flags = (arg_clone_ns_flags & ~ns_flag) | (r > 0 ? 0 : ns_flag); +} + static int parse_argv(int argc, char *argv[]) { enum { @@ -399,7 +393,6 @@ static int parse_argv(int argc, char *argv[]) { ARG_TMPFS, ARG_OVERLAY, ARG_OVERLAY_RO, - ARG_SETENV, ARG_SHARE_SYSTEM, ARG_REGISTER, ARG_KEEP_UNIT, @@ -407,64 +400,79 @@ static int parse_argv(int argc, char *argv[]) { ARG_NETWORK_MACVLAN, ARG_NETWORK_IPVLAN, ARG_NETWORK_BRIDGE, + ARG_NETWORK_ZONE, + ARG_NETWORK_VETH_EXTRA, ARG_PERSONALITY, ARG_VOLATILE, ARG_TEMPLATE, ARG_PROPERTY, ARG_PRIVATE_USERS, ARG_KILL_SIGNAL, + ARG_SETTINGS, + ARG_CHDIR, + ARG_PRIVATE_USERS_CHOWN, + ARG_NOTIFY_READY, }; static const struct option options[] = { - { "help", no_argument, NULL, 'h' }, - { "version", no_argument, NULL, ARG_VERSION }, - { "directory", required_argument, NULL, 'D' }, - { "template", required_argument, NULL, ARG_TEMPLATE }, - { "ephemeral", no_argument, NULL, 'x' }, - { "user", required_argument, NULL, 'u' }, - { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, - { "boot", no_argument, NULL, 'b' }, - { "uuid", required_argument, NULL, ARG_UUID }, - { "read-only", no_argument, NULL, ARG_READ_ONLY }, - { "capability", required_argument, NULL, ARG_CAPABILITY }, - { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY }, - { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, - { "bind", required_argument, NULL, ARG_BIND }, - { "bind-ro", required_argument, NULL, ARG_BIND_RO }, - { "tmpfs", required_argument, NULL, ARG_TMPFS }, - { "overlay", required_argument, NULL, ARG_OVERLAY }, - { "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO }, - { "machine", required_argument, NULL, 'M' }, - { "slice", required_argument, NULL, 'S' }, - { "setenv", required_argument, NULL, ARG_SETENV }, - { "selinux-context", required_argument, NULL, 'Z' }, - { "selinux-apifs-context", required_argument, NULL, 'L' }, - { "quiet", no_argument, NULL, 'q' }, - { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, - { "register", required_argument, NULL, ARG_REGISTER }, - { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, - { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, - { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN }, - { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN }, - { "network-veth", no_argument, NULL, 'n' }, - { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE }, - { "personality", required_argument, NULL, ARG_PERSONALITY }, - { "image", required_argument, NULL, 'i' }, - { "volatile", optional_argument, NULL, ARG_VOLATILE }, - { "port", required_argument, NULL, 'p' }, - { "property", required_argument, NULL, ARG_PROPERTY }, - { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS }, - { "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL }, + { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, + { "directory", required_argument, NULL, 'D' }, + { "template", required_argument, NULL, ARG_TEMPLATE }, + { "ephemeral", no_argument, NULL, 'x' }, + { "user", required_argument, NULL, 'u' }, + { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, + { "as-pid2", no_argument, NULL, 'a' }, + { "boot", no_argument, NULL, 'b' }, + { "uuid", required_argument, NULL, ARG_UUID }, + { "read-only", no_argument, NULL, ARG_READ_ONLY }, + { "capability", required_argument, NULL, ARG_CAPABILITY }, + { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY }, + { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, + { "bind", required_argument, NULL, ARG_BIND }, + { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "tmpfs", required_argument, NULL, ARG_TMPFS }, + { "overlay", required_argument, NULL, ARG_OVERLAY }, + { "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO }, + { "machine", required_argument, NULL, 'M' }, + { "slice", required_argument, NULL, 'S' }, + { "setenv", required_argument, NULL, 'E' }, + { "selinux-context", required_argument, NULL, 'Z' }, + { "selinux-apifs-context", required_argument, NULL, 'L' }, + { "quiet", no_argument, NULL, 'q' }, + { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, /* not documented */ + { "register", required_argument, NULL, ARG_REGISTER }, + { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, + { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, + { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN }, + { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN }, + { "network-veth", no_argument, NULL, 'n' }, + { "network-veth-extra", required_argument, NULL, ARG_NETWORK_VETH_EXTRA }, + { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE }, + { "network-zone", required_argument, NULL, ARG_NETWORK_ZONE }, + { "personality", required_argument, NULL, ARG_PERSONALITY }, + { "image", required_argument, NULL, 'i' }, + { "volatile", optional_argument, NULL, ARG_VOLATILE }, + { "port", required_argument, NULL, 'p' }, + { "property", required_argument, NULL, ARG_PROPERTY }, + { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS }, + { "private-users-chown", optional_argument, NULL, ARG_PRIVATE_USERS_CHOWN }, + { "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL }, + { "settings", required_argument, NULL, ARG_SETTINGS }, + { "chdir", required_argument, NULL, ARG_CHDIR }, + { "notify-ready", required_argument, NULL, ARG_NOTIFY_READY }, {} }; int c, r; + const char *p, *e; uint64_t plus = 0, minus = 0; + bool mask_all_settings = false, mask_no_settings = false; assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:xp:n", options, NULL)) >= 0) + while ((c = getopt_long(argc, argv, "+hD:u:abL:M:jS:Z:qi:xp:nU", options, NULL)) >= 0) switch (c) { @@ -473,29 +481,24 @@ static int parse_argv(int argc, char *argv[]) { return 0; case ARG_VERSION: - puts(PACKAGE_STRING); - puts(SYSTEMD_FEATURES); - return 0; + return version(); case 'D': - r = set_sanitized_path(&arg_directory, optarg); + r = parse_path_argument_and_warn(optarg, false, &arg_directory); if (r < 0) - return log_error_errno(r, "Invalid root directory: %m"); - + return r; break; case ARG_TEMPLATE: - r = set_sanitized_path(&arg_template, optarg); + r = parse_path_argument_and_warn(optarg, false, &arg_template); if (r < 0) - return log_error_errno(r, "Invalid template directory: %m"); - + return r; break; case 'i': - r = set_sanitized_path(&arg_image, optarg); + r = parse_path_argument_and_warn(optarg, false, &arg_image); if (r < 0) - return log_error_errno(r, "Invalid image path: %m"); - + return r; break; case 'x': @@ -503,38 +506,98 @@ static int parse_argv(int argc, char *argv[]) { break; case 'u': - free(arg_user); - arg_user = strdup(optarg); - if (!arg_user) + r = free_and_strdup(&arg_user, optarg); + if (r < 0) + return log_oom(); + + arg_settings_mask |= SETTING_USER; + break; + + case ARG_NETWORK_ZONE: { + char *j; + + j = strappend("vz-", optarg); + if (!j) return log_oom(); + if (!ifname_valid(j)) { + log_error("Network zone name not valid: %s", j); + free(j); + return -EINVAL; + } + + free(arg_network_zone); + arg_network_zone = j; + + arg_network_veth = true; + arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; break; + } case ARG_NETWORK_BRIDGE: - arg_network_bridge = optarg; + + if (!ifname_valid(optarg)) { + log_error("Bridge interface name not valid: %s", optarg); + return -EINVAL; + } + + r = free_and_strdup(&arg_network_bridge, optarg); + if (r < 0) + return log_oom(); /* fall through */ case 'n': arg_network_veth = true; arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; + break; + + case ARG_NETWORK_VETH_EXTRA: + r = veth_extra_parse(&arg_network_veth_extra, optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse --network-veth-extra= parameter: %s", optarg); + + arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; break; case ARG_NETWORK_INTERFACE: + + if (!ifname_valid(optarg)) { + log_error("Network interface name not valid: %s", optarg); + return -EINVAL; + } + if (strv_extend(&arg_network_interfaces, optarg) < 0) return log_oom(); arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; break; case ARG_NETWORK_MACVLAN: + + if (!ifname_valid(optarg)) { + log_error("MACVLAN network interface name not valid: %s", optarg); + return -EINVAL; + } + if (strv_extend(&arg_network_macvlan, optarg) < 0) return log_oom(); arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; break; case ARG_NETWORK_IPVLAN: + + if (!ifname_valid(optarg)) { + log_error("IPVLAN network interface name not valid: %s", optarg); + return -EINVAL; + } + if (strv_extend(&arg_network_ipvlan, optarg) < 0) return log_oom(); @@ -542,18 +605,40 @@ static int parse_argv(int argc, char *argv[]) { case ARG_PRIVATE_NETWORK: arg_private_network = true; + arg_settings_mask |= SETTING_NETWORK; break; case 'b': - arg_boot = true; + if (arg_start_mode == START_PID2) { + log_error("--boot and --as-pid2 may not be combined."); + return -EINVAL; + } + + arg_start_mode = START_BOOT; + arg_settings_mask |= SETTING_START_MODE; + break; + + case 'a': + if (arg_start_mode == START_BOOT) { + log_error("--boot and --as-pid2 may not be combined."); + return -EINVAL; + } + + arg_start_mode = START_PID2; + arg_settings_mask |= SETTING_START_MODE; break; case ARG_UUID: r = sd_id128_from_string(optarg, &arg_uuid); - if (r < 0) { - log_error("Invalid UUID: %s", optarg); - return r; + if (r < 0) + return log_error_errno(r, "Invalid UUID: %s", optarg); + + if (sd_id128_is_null(arg_uuid)) { + log_error("Machine UUID may not be all zeroes."); + return -EINVAL; } + + arg_settings_mask |= SETTING_MACHINE_ID; break; case 'S': @@ -561,10 +646,9 @@ static int parse_argv(int argc, char *argv[]) { break; case 'M': - if (isempty(optarg)) { - free(arg_machine); - arg_machine = NULL; - } else { + if (isempty(optarg)) + arg_machine = mfree(arg_machine); + else { if (!machine_name_is_valid(optarg)) { log_error("Invalid machine name: %s", optarg); return -EINVAL; @@ -587,19 +671,21 @@ static int parse_argv(int argc, char *argv[]) { case ARG_READ_ONLY: arg_read_only = true; + arg_settings_mask |= SETTING_READ_ONLY; break; case ARG_CAPABILITY: case ARG_DROP_CAPABILITY: { - const char *state, *word; - size_t length; + p = optarg; + for (;;) { + _cleanup_free_ char *t = NULL; - FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) { - _cleanup_free_ char *t; + r = extract_first_word(&p, &t, ",", 0); + if (r < 0) + return log_error_errno(r, "Failed to parse capability %s.", t); - t = strndup(word, length); - if (!t) - return log_oom(); + if (r == 0) + break; if (streq(t, "all")) { if (c == ARG_CAPABILITY) @@ -622,6 +708,7 @@ static int parse_argv(int argc, char *argv[]) { } } + arg_settings_mask |= SETTING_CAPABILITY; break; } @@ -657,74 +744,21 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_BIND: - case ARG_BIND_RO: { - _cleanup_free_ char *source = NULL, *destination = NULL; - CustomMount *m; - char *e; - - e = strchr(optarg, ':'); - if (e) { - source = strndup(optarg, e - optarg); - destination = strdup(e + 1); - } else { - source = strdup(optarg); - destination = strdup(optarg); - } - - if (!source || !destination) - return log_oom(); - - if (!path_is_absolute(source) || !path_is_absolute(destination)) { - log_error("Invalid bind mount specification: %s", optarg); - return -EINVAL; - } - - m = custom_mount_add(CUSTOM_MOUNT_BIND); - if (!m) - return log_oom(); - - m->source = source; - m->destination = destination; - m->read_only = c == ARG_BIND_RO; - - source = destination = NULL; + case ARG_BIND_RO: + r = bind_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg, c == ARG_BIND_RO); + if (r < 0) + return log_error_errno(r, "Failed to parse --bind(-ro)= argument %s: %m", optarg); + arg_settings_mask |= SETTING_CUSTOM_MOUNTS; break; - } - - case ARG_TMPFS: { - _cleanup_free_ char *path = NULL, *opts = NULL; - CustomMount *m; - char *e; - e = strchr(optarg, ':'); - if (e) { - path = strndup(optarg, e - optarg); - opts = strdup(e + 1); - } else { - path = strdup(optarg); - opts = strdup("mode=0755"); - } - - if (!path || !opts) - return log_oom(); - - if (!path_is_absolute(path)) { - log_error("Invalid tmpfs specification: %s", optarg); - return -EINVAL; - } - - m = custom_mount_add(CUSTOM_MOUNT_TMPFS); - if (!m) - return log_oom(); - - m->destination = path; - m->options = opts; - - path = opts = NULL; + case ARG_TMPFS: + r = tmpfs_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse --tmpfs= argument %s: %m", optarg); + arg_settings_mask |= SETTING_CUSTOM_MOUNTS; break; - } case ARG_OVERLAY: case ARG_OVERLAY_RO: { @@ -734,9 +768,13 @@ static int parse_argv(int argc, char *argv[]) { unsigned n = 0; char **i; - lower = strv_split(optarg, ":"); - if (!lower) + r = strv_split_extract(&lower, optarg, ":", EXTRACT_DONT_COALESCE_SEPARATORS); + if (r == -ENOMEM) return log_oom(); + else if (r < 0) { + log_error("Invalid overlay specification: %s", optarg); + return r; + } STRV_FOREACH(i, lower) { if (!path_is_absolute(*i)) { @@ -771,7 +809,7 @@ static int parse_argv(int argc, char *argv[]) { lower[n - 2] = NULL; } - m = custom_mount_add(CUSTOM_MOUNT_OVERLAY); + m = custom_mount_add(&arg_custom_mounts, &arg_n_custom_mounts, CUSTOM_MOUNT_OVERLAY); if (!m) return log_oom(); @@ -783,10 +821,11 @@ static int parse_argv(int argc, char *argv[]) { upper = destination = NULL; lower = NULL; + arg_settings_mask |= SETTING_CUSTOM_MOUNTS; break; } - case ARG_SETENV: { + case 'E': { char **n; if (!env_assignment_is_valid(optarg)) { @@ -800,6 +839,8 @@ static int parse_argv(int argc, char *argv[]) { strv_free(arg_setenv); arg_setenv = n; + + arg_settings_mask |= SETTING_ENVIRONMENT; break; } @@ -808,7 +849,9 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_SHARE_SYSTEM: - arg_share_system = true; + /* We don't officially support this anymore, except for compat reasons. People should use the + * $SYSTEMD_NSPAWN_SHARE_* environment variables instead. */ + arg_clone_ns_flags = 0; break; case ARG_REGISTER: @@ -833,85 +876,36 @@ static int parse_argv(int argc, char *argv[]) { return -EINVAL; } + arg_settings_mask |= SETTING_PERSONALITY; break; case ARG_VOLATILE: if (!optarg) - arg_volatile = VOLATILE_YES; - else { - r = parse_boolean(optarg); - if (r < 0) { - if (streq(optarg, "state")) - arg_volatile = VOLATILE_STATE; - else { - log_error("Failed to parse --volatile= argument: %s", optarg); - return r; - } - } else - arg_volatile = r ? VOLATILE_YES : VOLATILE_NO; - } - - break; - - case 'p': { - const char *split, *e; - uint16_t container_port, host_port; - int protocol; - ExposePort *p; - - if ((e = startswith(optarg, "tcp:"))) - protocol = IPPROTO_TCP; - else if ((e = startswith(optarg, "udp:"))) - protocol = IPPROTO_UDP; + arg_volatile_mode = VOLATILE_YES; else { - e = optarg; - protocol = IPPROTO_TCP; - } - - split = strchr(e, ':'); - if (split) { - char v[split - e + 1]; - - memcpy(v, e, split - e); - v[split - e] = 0; + VolatileMode m; - r = safe_atou16(v, &host_port); - if (r < 0 || host_port <= 0) { - log_error("Failed to parse host port: %s", optarg); + m = volatile_mode_from_string(optarg); + if (m < 0) { + log_error("Failed to parse --volatile= argument: %s", optarg); return -EINVAL; - } - - r = safe_atou16(split + 1, &container_port); - } else { - r = safe_atou16(e, &container_port); - host_port = container_port; - } - - if (r < 0 || container_port <= 0) { - log_error("Failed to parse host port: %s", optarg); - return -EINVAL; - } - - LIST_FOREACH(ports, p, arg_expose_ports) { - if (p->protocol == protocol && p->host_port == host_port) { - log_error("Duplicate port specification: %s", optarg); - return -EINVAL; - } + } else + arg_volatile_mode = m; } - p = new(ExposePort, 1); - if (!p) - return log_oom(); - - p->protocol = protocol; - p->host_port = host_port; - p->container_port = container_port; + arg_settings_mask |= SETTING_VOLATILE_MODE; + break; - LIST_PREPEND(ports, arg_expose_ports, p); + case 'p': + r = expose_port_parse(&arg_expose_ports, optarg); + if (r == -EEXIST) + return log_error_errno(r, "Duplicate port specification: %s", optarg); + if (r < 0) + return log_error_errno(r, "Failed to parse host port %s: %m", optarg); + arg_settings_mask |= SETTING_EXPOSE_PORTS; break; - } case ARG_PROPERTY: if (strv_extend(&arg_property, optarg) < 0) @@ -919,11 +913,36 @@ static int parse_argv(int argc, char *argv[]) { break; - case ARG_PRIVATE_USERS: - if (optarg) { + case ARG_PRIVATE_USERS: { + int boolean = -1; + + if (!optarg) + boolean = true; + else if (!in_charset(optarg, DIGITS)) + /* do *not* parse numbers as booleans */ + boolean = parse_boolean(optarg); + + if (boolean == false) { + /* no: User namespacing off */ + arg_userns_mode = USER_NAMESPACE_NO; + arg_uid_shift = UID_INVALID; + arg_uid_range = UINT32_C(0x10000); + } else if (boolean == true) { + /* yes: User namespacing on, UID range is read from root dir */ + arg_userns_mode = USER_NAMESPACE_FIXED; + arg_uid_shift = UID_INVALID; + arg_uid_range = UINT32_C(0x10000); + } else if (streq(optarg, "pick")) { + /* pick: User namespacing on, UID range is picked randomly */ + arg_userns_mode = USER_NAMESPACE_PICK; + arg_uid_shift = UID_INVALID; + arg_uid_range = UINT32_C(0x10000); + } else { _cleanup_free_ char *buffer = NULL; const char *range, *shift; + /* anything else: User namespacing on, UID range is explicitly configured */ + range = strchr(optarg, ':'); if (range) { buffer = strndup(optarg, range - optarg); @@ -932,20 +951,43 @@ static int parse_argv(int argc, char *argv[]) { shift = buffer; range++; - if (safe_atou32(range, &arg_uid_range) < 0 || arg_uid_range <= 0) { - log_error("Failed to parse UID range: %s", range); - return -EINVAL; - } + r = safe_atou32(range, &arg_uid_range); + if (r < 0) + return log_error_errno(r, "Failed to parse UID range \"%s\": %m", range); } else shift = optarg; - if (parse_uid(shift, &arg_uid_shift) < 0) { - log_error("Failed to parse UID: %s", optarg); - return -EINVAL; - } + r = parse_uid(shift, &arg_uid_shift); + if (r < 0) + return log_error_errno(r, "Failed to parse UID \"%s\": %m", optarg); + + arg_userns_mode = USER_NAMESPACE_FIXED; + } + + if (arg_uid_range <= 0) { + log_error("UID range cannot be 0."); + return -EINVAL; + } + + arg_settings_mask |= SETTING_USERNS; + break; + } + + case 'U': + if (userns_supported()) { + arg_userns_mode = USER_NAMESPACE_PICK; + arg_uid_shift = UID_INVALID; + arg_uid_range = UINT32_C(0x10000); + + arg_settings_mask |= SETTING_USERNS; } - arg_userns = true; + break; + + case ARG_PRIVATE_USERS_CHOWN: + arg_userns_chown = true; + + arg_settings_mask |= SETTING_USERNS; break; case ARG_KILL_SIGNAL: @@ -955,6 +997,65 @@ static int parse_argv(int argc, char *argv[]) { return -EINVAL; } + arg_settings_mask |= SETTING_KILL_SIGNAL; + break; + + case ARG_SETTINGS: + + /* no → do not read files + * yes → read files, do not override cmdline, trust only subset + * override → read files, override cmdline, trust only subset + * trusted → read files, do not override cmdline, trust all + */ + + r = parse_boolean(optarg); + if (r < 0) { + if (streq(optarg, "trusted")) { + mask_all_settings = false; + mask_no_settings = false; + arg_settings_trusted = true; + + } else if (streq(optarg, "override")) { + mask_all_settings = false; + mask_no_settings = true; + arg_settings_trusted = -1; + } else + return log_error_errno(r, "Failed to parse --settings= argument: %s", optarg); + } else if (r > 0) { + /* yes */ + mask_all_settings = false; + mask_no_settings = false; + arg_settings_trusted = -1; + } else { + /* no */ + mask_all_settings = true; + mask_no_settings = false; + arg_settings_trusted = false; + } + + break; + + case ARG_CHDIR: + if (!path_is_absolute(optarg)) { + log_error("Working directory %s is not an absolute path.", optarg); + return -EINVAL; + } + + r = free_and_strdup(&arg_chdir, optarg); + if (r < 0) + return log_oom(); + + arg_settings_mask |= SETTING_WORKING_DIRECTORY; + break; + + case ARG_NOTIFY_READY: + r = parse_boolean(optarg); + if (r < 0) { + log_error("%s is not a valid notify mode. Valid modes are: yes, no, and ready.", optarg); + return -EINVAL; + } + arg_notify_ready = r; + arg_settings_mask |= SETTING_NOTIFY_READY; break; case '?': @@ -964,14 +1065,23 @@ static int parse_argv(int argc, char *argv[]) { assert_not_reached("Unhandled option"); } - if (arg_share_system) - arg_register = false; + parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_IPC", CLONE_NEWIPC); + parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_PID", CLONE_NEWPID); + parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_UTS", CLONE_NEWUTS); + parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_SYSTEM", CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS); - if (arg_boot && arg_share_system) { - log_error("--boot and --share-system may not be combined."); - return -EINVAL; + if (!(arg_clone_ns_flags & CLONE_NEWPID) || + !(arg_clone_ns_flags & CLONE_NEWUTS)) { + arg_register = false; + if (arg_start_mode != START_PID1) { + log_error("--boot cannot be used without namespacing."); + return -EINVAL; + } } + if (arg_userns_mode == USER_NAMESPACE_PICK) + arg_userns_chown = true; + if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) { log_error("--keep-unit may not be used when invoked from a user session."); return -EINVAL; @@ -1007,421 +1117,73 @@ static int parse_argv(int argc, char *argv[]) { return -EINVAL; } - if (arg_volatile != VOLATILE_NO && arg_read_only) { - log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy."); - return -EINVAL; + if (arg_userns_mode != USER_NAMESPACE_NO && !userns_supported()) { + log_error("--private-users= is not supported, kernel compiled without user namespace support."); + return -EOPNOTSUPP; } - if (arg_expose_ports && !arg_private_network) { - log_error("Cannot use --port= without private networking."); + if (arg_userns_chown && arg_read_only) { + log_error("--read-only and --private-users-chown may not be combined."); return -EINVAL; } - if (arg_userns && access("/proc/self/uid_map", F_OK) < 0) - return log_error_errno(EOPNOTSUPP, "--private-users= is not supported, kernel compiled without user namespace support."); - - arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; - - if (arg_boot && arg_kill_signal <= 0) - arg_kill_signal = SIGRTMIN+3; - - return 1; -} - -static int tmpfs_patch_options(const char *options, char **ret) { - char *buf = NULL; - - if (arg_userns && arg_uid_shift != 0) { - assert(arg_uid_shift != UID_INVALID); - - if (options) - (void) asprintf(&buf, "%s,uid=" UID_FMT ",gid=" UID_FMT, options, arg_uid_shift, arg_uid_shift); - else - (void) asprintf(&buf, "uid=" UID_FMT ",gid=" UID_FMT, arg_uid_shift, arg_uid_shift); - if (!buf) - return -ENOMEM; - - options = buf; - } - -#ifdef HAVE_SELINUX - if (arg_selinux_apifs_context) { - char *t; - - if (options) - t = strjoin(options, ",context=\"", arg_selinux_apifs_context, "\"", NULL); - else - t = strjoin("context=\"", arg_selinux_apifs_context, "\"", NULL); - if (!t) { - free(buf); - return -ENOMEM; - } - - free(buf); - buf = t; + if (arg_network_bridge && arg_network_zone) { + log_error("--network-bridge= and --network-zone= may not be combined."); + return -EINVAL; } -#endif - - *ret = buf; - return !!buf; -} - -static int mount_all(const char *dest, bool userns) { - - typedef struct MountPoint { - const char *what; - const char *where; - const char *type; - const char *options; - unsigned long flags; - bool fatal; - bool userns; - } MountPoint; - - static const MountPoint mount_table[] = { - { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true }, - { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true, true }, /* Bind mount first */ - { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, true, true }, /* Then, make it r/o */ - { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false }, - { "tmpfs", "/sys/fs/cgroup", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, true, false }, - { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true, false }, - { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false }, - { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false }, - { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true, false }, -#ifdef HAVE_SELINUX - { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false, false }, /* Bind mount first */ - { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, false, false }, /* Then, make it r/o */ -#endif - }; - - unsigned k; - int r; - for (k = 0; k < ELEMENTSOF(mount_table); k++) { - _cleanup_free_ char *where = NULL, *options = NULL; - const char *o; - - if (userns != mount_table[k].userns) - continue; - - where = prefix_root(dest, mount_table[k].where); - if (!where) + if (argc > optind) { + arg_parameters = strv_copy(argv + optind); + if (!arg_parameters) return log_oom(); - r = path_is_mount_point(where, AT_SYMLINK_FOLLOW); - if (r < 0 && r != -ENOENT) - return log_error_errno(r, "Failed to detect whether %s is a mount point: %m", where); - - /* Skip this entry if it is not a remount. */ - if (mount_table[k].what && r > 0) - continue; - - r = mkdir_p(where, 0755); - if (r < 0) { - if (mount_table[k].fatal) - return log_error_errno(r, "Failed to create directory %s: %m", where); - - log_warning_errno(r, "Failed to create directory %s: %m", where); - continue; - } - - o = mount_table[k].options; - if (streq_ptr(mount_table[k].type, "tmpfs")) { - r = tmpfs_patch_options(o, &options); - if (r < 0) - return log_oom(); - if (r > 0) - o = options; - } - - if (mount(mount_table[k].what, - where, - mount_table[k].type, - mount_table[k].flags, - o) < 0) { - - if (mount_table[k].fatal) - return log_error_errno(errno, "mount(%s) failed: %m", where); - - log_warning_errno(errno, "mount(%s) failed, ignoring: %m", where); - } + arg_settings_mask |= SETTING_START_MODE; } - return 0; -} - -static int mount_bind(const char *dest, CustomMount *m) { - struct stat source_st, dest_st; - const char *where; - int r; - - assert(m); + /* Load all settings from .nspawn files */ + if (mask_no_settings) + arg_settings_mask = 0; - if (stat(m->source, &source_st) < 0) - return log_error_errno(errno, "Failed to stat %s: %m", m->source); - - where = prefix_roota(dest, m->destination); - - if (stat(where, &dest_st) >= 0) { - if (S_ISDIR(source_st.st_mode) && !S_ISDIR(dest_st.st_mode)) { - log_error("Cannot bind mount directory %s on file %s.", m->source, where); - return -EINVAL; - } + /* Don't load any settings from .nspawn files */ + if (mask_all_settings) + arg_settings_mask = _SETTINGS_MASK_ALL; - if (!S_ISDIR(source_st.st_mode) && S_ISDIR(dest_st.st_mode)) { - log_error("Cannot bind mount file %s on directory %s.", m->source, where); - return -EINVAL; - } + arg_caps_retain = (arg_caps_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; - } else if (errno == ENOENT) { - r = mkdir_parents_label(where, 0755); - if (r < 0) - return log_error_errno(r, "Failed to make parents of %s: %m", where); - } else { - log_error_errno(errno, "Failed to stat %s: %m", where); - return -errno; - } + e = getenv("SYSTEMD_NSPAWN_CONTAINER_SERVICE"); + if (e) + arg_container_service_name = e; - /* Create the mount point. Any non-directory file can be - * mounted on any non-directory file (regular, fifo, socket, - * char, block). - */ - if (S_ISDIR(source_st.st_mode)) - r = mkdir_label(where, 0755); - else - r = touch(where); - if (r < 0 && r != -EEXIST) - return log_error_errno(r, "Failed to create mount point %s: %m", where); - - if (mount(m->source, where, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "mount(%s) failed: %m", where); - - if (m->read_only) { - r = bind_remount_recursive(where, true); - if (r < 0) - return log_error_errno(r, "Read-only bind mount failed: %m"); - } - - return 0; -} - -static int mount_tmpfs(const char *dest, CustomMount *m) { - const char *where, *options; - _cleanup_free_ char *buf = NULL; - int r; - - assert(dest); - assert(m); - - where = prefix_roota(dest, m->destination); - - r = mkdir_p_label(where, 0755); - if (r < 0 && r != -EEXIST) - return log_error_errno(r, "Creating mount point for tmpfs %s failed: %m", where); - - r = tmpfs_patch_options(m->options, &buf); + r = getenv_bool("SYSTEMD_NSPAWN_USE_CGNS"); if (r < 0) - return log_oom(); - options = r > 0 ? buf : m->options; - - if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, options) < 0) - return log_error_errno(errno, "tmpfs mount to %s failed: %m", where); - - return 0; -} - -static int mount_overlay(const char *dest, CustomMount *m) { - _cleanup_free_ char *lower = NULL; - const char *where, *options; - int r; - - assert(dest); - assert(m); - - where = prefix_roota(dest, m->destination); - - r = mkdir_label(where, 0755); - if (r < 0 && r != -EEXIST) - return log_error_errno(r, "Creating mount point for overlay %s failed: %m", where); - - (void) mkdir_p_label(m->source, 0755); - - strv_reverse(m->lower); - lower = strv_join(m->lower, ":"); - strv_reverse(m->lower); - if (!lower) - return log_oom(); - - if (m->read_only) - options = strjoina("lowerdir=", m->source, ":", lower); - else { - assert(m->work_dir); - (void) mkdir_label(m->work_dir, 0700); - - options = strjoina("lowerdir=", lower, ",upperdir=", m->source, ",workdir=", m->work_dir); - } - - if (mount("overlay", where, "overlay", m->read_only ? MS_RDONLY : 0, options) < 0) - return log_error_errno(errno, "overlay mount to %s failed: %m", where); + arg_use_cgns = cg_ns_supported(); + else + arg_use_cgns = r; - return 0; + return 1; } -static int mount_custom(const char *dest) { - unsigned i; - int r; - - assert(dest); - - for (i = 0; i < arg_n_custom_mounts; i++) { - CustomMount *m = &arg_custom_mounts[i]; - - switch (m->type) { - - case CUSTOM_MOUNT_BIND: - r = mount_bind(dest, m); - break; - - case CUSTOM_MOUNT_TMPFS: - r = mount_tmpfs(dest, m); - break; - - case CUSTOM_MOUNT_OVERLAY: - r = mount_overlay(dest, m); - break; - - default: - assert_not_reached("Unknown custom mount type"); - } +static int verify_arguments(void) { - if (r < 0) - return r; + if (arg_volatile_mode != VOLATILE_NO && arg_read_only) { + log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy."); + return -EINVAL; } - return 0; -} - -static int mount_cgroup_hierarchy(const char *dest, const char *controller, const char *hierarchy, bool read_only) { - char *to; - int r; - - to = strjoina(dest, "/sys/fs/cgroup/", hierarchy); - - r = path_is_mount_point(to, 0); - if (r < 0 && r != -ENOENT) - return log_error_errno(r, "Failed to determine if %s is mounted already: %m", to); - if (r > 0) - return 0; - - mkdir_p(to, 0755); - - /* The superblock mount options of the mount point need to be - * identical to the hosts', and hence writable... */ - if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, controller) < 0) - return log_error_errno(errno, "Failed to mount to %s: %m", to); - - /* ... hence let's only make the bind mount read-only, not the - * superblock. */ - if (read_only) { - if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) - return log_error_errno(errno, "Failed to remount %s read-only: %m", to); + if (arg_expose_ports && !arg_private_network) { + log_error("Cannot use --port= without private networking."); + return -EINVAL; } - return 1; -} - -static int mount_cgroup(const char *dest) { - _cleanup_set_free_free_ Set *controllers = NULL; - const char *cgroup_root; - int r; - - controllers = set_new(&string_hash_ops); - if (!controllers) - return log_oom(); - - r = cg_kernel_controllers(controllers); - if (r < 0) - return log_error_errno(r, "Failed to determine cgroup controllers: %m"); - - for (;;) { - _cleanup_free_ char *controller = NULL, *origin = NULL, *combined = NULL; - controller = set_steal_first(controllers); - if (!controller) - break; - - origin = prefix_root("/sys/fs/cgroup/", controller); - if (!origin) - return log_oom(); - - r = readlink_malloc(origin, &combined); - if (r == -EINVAL) { - /* Not a symbolic link, but directly a single cgroup hierarchy */ - - r = mount_cgroup_hierarchy(dest, controller, controller, true); - if (r < 0) - return r; - - } else if (r < 0) - return log_error_errno(r, "Failed to read link %s: %m", origin); - else { - _cleanup_free_ char *target = NULL; - - target = prefix_root(dest, origin); - if (!target) - return log_oom(); - - /* A symbolic link, a combination of controllers in one hierarchy */ - - if (!filename_is_valid(combined)) { - log_warning("Ignoring invalid combined hierarchy %s.", combined); - continue; - } - - r = mount_cgroup_hierarchy(dest, combined, combined, true); - if (r < 0) - return r; - - r = symlink_idempotent(combined, target); - if (r == -EINVAL) { - log_error("Invalid existing symlink for combined hierarchy"); - return r; - } - if (r < 0) - return log_error_errno(r, "Failed to create symlink for combined hierarchy: %m"); - } +#ifndef HAVE_LIBIPTC + if (arg_expose_ports) { + log_error("--port= is not supported, compiled without libiptc support."); + return -EOPNOTSUPP; } +#endif - r = mount_cgroup_hierarchy(dest, "name=systemd,xattr", "systemd", false); - if (r < 0) - return r; - - cgroup_root = prefix_roota(dest, "/sys/fs/cgroup"); - if (mount(NULL, cgroup_root, NULL, MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755") < 0) - return log_error_errno(errno, "Failed to remount %s read-only: %m", cgroup_root); - - return 0; -} - -static int mount_systemd_cgroup_writable(const char *dest) { - _cleanup_free_ char *own_cgroup_path = NULL; - const char *systemd_root, *systemd_own; - int r; - - assert(dest); - - r = cg_pid_get_path(NULL, 0, &own_cgroup_path); - if (r < 0) - return log_error_errno(r, "Failed to determine our own cgroup path: %m"); - - /* Make our own cgroup a (writable) bind mount */ - systemd_own = strjoina(dest, "/sys/fs/cgroup/systemd", own_cgroup_path); - if (mount(systemd_own, systemd_own, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "Failed to turn %s into a bind mount: %m", own_cgroup_path); - - /* And then remount the systemd cgroup root read-only */ - systemd_root = prefix_roota(dest, "/sys/fs/cgroup/systemd"); - if (mount(NULL, systemd_root, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) - return log_error_errno(errno, "Failed to mount cgroup root read-only: %m"); + if (arg_start_mode == START_BOOT && arg_kill_signal <= 0) + arg_kill_signal = SIGRTMIN+3; return 0; } @@ -1429,7 +1191,7 @@ static int mount_systemd_cgroup_writable(const char *dest) { static int userns_lchown(const char *p, uid_t uid, gid_t gid) { assert(p); - if (!arg_userns) + if (arg_userns_mode == USER_NAMESPACE_NO) return 0; if (uid == UID_INVALID && gid == GID_INVALID) @@ -1479,7 +1241,13 @@ static int setup_timezone(const char *dest) { /* Fix the timezone, if possible */ r = readlink_malloc("/etc/localtime", &p); if (r < 0) { - log_warning("/etc/localtime is not a symlink, not updating container timezone."); + log_warning("host's /etc/localtime is not a symlink, not updating container timezone."); + /* to handle warning, delete /etc/localtime and replace it + * with a symbolic link to a time zone data file. + * + * Example: + * ln -s /usr/share/zoneinfo/UTC /etc/localtime + */ return 0; } @@ -1504,7 +1272,7 @@ static int setup_timezone(const char *dest) { } check = strjoina("/usr/share/zoneinfo/", z); - check = prefix_root(dest, check); + check = prefix_roota(dest, check); if (laccess(check, F_OK) < 0) { log_warning("Timezone %s does not exist in container, not updating container timezone.", z); return 0; @@ -1541,155 +1309,44 @@ static int setup_resolv_conf(const char *dest) { /* Fix resolv.conf, if possible */ where = prefix_roota(dest, "/etc/resolv.conf"); + if (access("/usr/lib/systemd/resolv.conf", F_OK) >= 0) { + /* resolved is enabled on the host. In this, case bind mount its static resolv.conf file into the + * container, so that the container can use the host's resolver. Given that network namespacing is + * disabled it's only natural of the container also uses the host's resolver. It also has the big + * advantage that the container will be able to follow the host's DNS server configuration changes + * transparently. */ + + r = mount_verbose(LOG_WARNING, "/usr/lib/systemd/resolv.conf", where, NULL, MS_BIND, NULL); + if (r >= 0) + return mount_verbose(LOG_ERR, NULL, where, NULL, + MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL); + } + + /* If that didn't work, let's copy the file */ r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0); if (r < 0) { - /* If the file already exists as symlink, let's - * suppress the warning, under the assumption that - * resolved or something similar runs inside and the - * symlink points there. + /* If the file already exists as symlink, let's suppress the warning, under the assumption that + * resolved or something similar runs inside and the symlink points there. * - * If the disk image is read-only, there's also no - * point in complaining. + * If the disk image is read-only, there's also no point in complaining. */ log_full_errno(IN_SET(r, -ELOOP, -EROFS) ? LOG_DEBUG : LOG_WARNING, r, - "Failed to copy /etc/resolv.conf to %s: %m", where); + "Failed to copy /etc/resolv.conf to %s, ignoring: %m", where); return 0; } r = userns_lchown(where, 0, 0); if (r < 0) - log_warning_errno(r, "Failed to chown /etc/resolv.conf: %m"); - - return 0; -} - -static int setup_volatile_state(const char *directory) { - _cleanup_free_ char *buf = NULL; - const char *p, *options; - int r; - - assert(directory); - - if (arg_volatile != VOLATILE_STATE) - return 0; - - /* --volatile=state means we simply overmount /var - with a tmpfs, and the rest read-only. */ - - r = bind_remount_recursive(directory, true); - if (r < 0) - return log_error_errno(r, "Failed to remount %s read-only: %m", directory); - - p = prefix_roota(directory, "/var"); - r = mkdir(p, 0755); - if (r < 0 && errno != EEXIST) - return log_error_errno(errno, "Failed to create %s: %m", directory); - - options = "mode=755"; - r = tmpfs_patch_options(options, &buf); - if (r < 0) - return log_oom(); - if (r > 0) - options = buf; - - if (mount("tmpfs", p, "tmpfs", MS_STRICTATIME, options) < 0) - return log_error_errno(errno, "Failed to mount tmpfs to /var: %m"); - - return 0; -} - -static int setup_volatile(const char *directory) { - bool tmpfs_mounted = false, bind_mounted = false; - char template[] = "/tmp/nspawn-volatile-XXXXXX"; - _cleanup_free_ char *buf = NULL; - const char *f, *t, *options; - int r; - - assert(directory); - - if (arg_volatile != VOLATILE_YES) - return 0; - - /* --volatile=yes means we mount a tmpfs to the root dir, and - the original /usr to use inside it, and that read-only. */ - - if (!mkdtemp(template)) - return log_error_errno(errno, "Failed to create temporary directory: %m"); - - options = "mode=755"; - r = tmpfs_patch_options(options, &buf); - if (r < 0) - return log_oom(); - if (r > 0) - options = buf; - - if (mount("tmpfs", template, "tmpfs", MS_STRICTATIME, options) < 0) { - r = log_error_errno(errno, "Failed to mount tmpfs for root directory: %m"); - goto fail; - } - - tmpfs_mounted = true; - - f = prefix_roota(directory, "/usr"); - t = prefix_roota(template, "/usr"); - - r = mkdir(t, 0755); - if (r < 0 && errno != EEXIST) { - r = log_error_errno(errno, "Failed to create %s: %m", t); - goto fail; - } - - if (mount(f, t, NULL, MS_BIND|MS_REC, NULL) < 0) { - r = log_error_errno(errno, "Failed to create /usr bind mount: %m"); - goto fail; - } - - bind_mounted = true; - - r = bind_remount_recursive(t, true); - if (r < 0) { - log_error_errno(r, "Failed to remount %s read-only: %m", t); - goto fail; - } - - if (mount(template, directory, NULL, MS_MOVE, NULL) < 0) { - r = log_error_errno(errno, "Failed to move root mount: %m"); - goto fail; - } - - (void) rmdir(template); + log_warning_errno(r, "Failed to chown /etc/resolv.conf, ignoring: %m"); return 0; - -fail: - if (bind_mounted) - (void) umount(t); - - if (tmpfs_mounted) - (void) umount(template); - (void) rmdir(template); - return r; -} - -static char* id128_format_as_uuid(sd_id128_t id, char s[37]) { - assert(s); - - snprintf(s, 37, - "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", - SD_ID128_FORMAT_VAL(id)); - - return s; } static int setup_boot_id(const char *dest) { + sd_id128_t rnd = SD_ID128_NULL; const char *from, *to; - sd_id128_t rnd = {}; - char as_uuid[37]; int r; - if (arg_share_system) - return 0; - /* Generate a new randomized boot ID, so that each boot-up of * the container gets a new one */ @@ -1700,18 +1357,16 @@ static int setup_boot_id(const char *dest) { if (r < 0) return log_error_errno(r, "Failed to generate random boot id: %m"); - id128_format_as_uuid(rnd, as_uuid); - - r = write_string_file(from, as_uuid, WRITE_STRING_FILE_CREATE); + r = id128_write(from, ID128_UUID, rnd, false); if (r < 0) return log_error_errno(r, "Failed to write boot id: %m"); - if (mount(from, to, NULL, MS_BIND, NULL) < 0) - r = log_error_errno(errno, "Failed to bind mount boot id: %m"); - else if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL) < 0) - log_warning_errno(errno, "Failed to make boot id read-only: %m"); + r = mount_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL); + if (r >= 0) + r = mount_verbose(LOG_ERR, NULL, to, NULL, + MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL); - unlink(from); + (void) unlink(from); return r; } @@ -1757,6 +1412,12 @@ static int copy_devnodes(const char *dest) { } else { if (mknod(to, st.st_mode, st.st_rdev) < 0) { + /* + * This is some sort of protection too against + * recursive userns chown on shared /dev/ + */ + if (errno == EEXIST) + log_notice("%s/dev/ should be an empty directory", dest); if (errno != EPERM) return log_error_errno(errno, "mknod(%s) failed: %m", to); @@ -1765,8 +1426,9 @@ static int copy_devnodes(const char *dest) { r = touch(to); if (r < 0) return log_error_errno(r, "touch (%s) failed: %m", to); - if (mount(from, to, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "Both mknod and bind mount (%s) failed: %m", to); + r = mount_verbose(LOG_DEBUG, from, to, NULL, MS_BIND, NULL); + if (r < 0) + return log_error_errno(r, "Both mknod and bind mount (%s) failed: %m", to); } r = userns_lchown(to, 0, 0); @@ -1781,19 +1443,18 @@ static int copy_devnodes(const char *dest) { static int setup_pts(const char *dest) { _cleanup_free_ char *options = NULL; const char *p; + int r; #ifdef HAVE_SELINUX if (arg_selinux_apifs_context) (void) asprintf(&options, - "newinstance,ptmxmode=0666,mode=620,uid=" UID_FMT ",gid=" GID_FMT ",context=\"%s\"", - arg_uid_shift, + "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"", arg_uid_shift + TTY_GID, arg_selinux_apifs_context); else #endif (void) asprintf(&options, - "newinstance,ptmxmode=0666,mode=620,uid=" UID_FMT ",gid=" GID_FMT, - arg_uid_shift, + "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT, arg_uid_shift + TTY_GID); if (!options) @@ -1803,22 +1464,26 @@ static int setup_pts(const char *dest) { p = prefix_roota(dest, "/dev/pts"); if (mkdir(p, 0755) < 0) return log_error_errno(errno, "Failed to create /dev/pts: %m"); - if (mount("devpts", p, "devpts", MS_NOSUID|MS_NOEXEC, options) < 0) - return log_error_errno(errno, "Failed to mount /dev/pts: %m"); - if (userns_lchown(p, 0, 0) < 0) - return log_error_errno(errno, "Failed to chown /dev/pts: %m"); + r = mount_verbose(LOG_ERR, "devpts", p, "devpts", MS_NOSUID|MS_NOEXEC, options); + if (r < 0) + return r; + r = userns_lchown(p, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to chown /dev/pts: %m"); /* Create /dev/ptmx symlink */ p = prefix_roota(dest, "/dev/ptmx"); if (symlink("pts/ptmx", p) < 0) return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m"); - if (userns_lchown(p, 0, 0) < 0) - return log_error_errno(errno, "Failed to chown /dev/ptmx: %m"); + r = userns_lchown(p, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to chown /dev/ptmx: %m"); /* And fix /dev/pts/ptmx ownership */ p = prefix_roota(dest, "/dev/pts/ptmx"); - if (userns_lchown(p, 0, 0) < 0) - return log_error_errno(errno, "Failed to chown /dev/pts/ptmx: %m"); + r = userns_lchown(p, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to chown /dev/pts/ptmx: %m"); return 0; } @@ -1846,25 +1511,13 @@ static int setup_dev_console(const char *dest, const char *console) { if (r < 0) return log_error_errno(r, "touch() for /dev/console failed: %m"); - if (mount(console, to, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "Bind mount for /dev/console failed: %m"); - - return 0; + return mount_verbose(LOG_ERR, console, to, NULL, MS_BIND, NULL); } static int setup_kmsg(const char *dest, int kmsg_socket) { const char *from, *to; _cleanup_umask_ mode_t u; - int fd, k; - 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 fd, r; assert(kmsg_socket >= 0); @@ -1882,28 +1535,21 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { if (mkfifo(from, 0600) < 0) return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m"); - if (mount(from, to, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "Bind mount for /proc/kmsg failed: %m"); + r = mount_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL); + if (r < 0) + return r; fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC); if (fd < 0) return log_error_errno(errno, "Failed to open fifo: %m"); - 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), &fd, sizeof(int)); - - mh.msg_controllen = cmsg->cmsg_len; - /* Store away the fd in the socket, so that it stays open as * long as we run the child */ - k = sendmsg(kmsg_socket, &mh, MSG_NOSIGNAL); + r = send_one_fd(kmsg_socket, fd, 0); safe_close(fd); - if (k < 0) - return log_error_errno(errno, "Failed to send FIFO fd: %m"); + if (r < 0) + return log_error_errno(r, "Failed to send FIFO fd: %m"); /* And now make the FIFO unavailable as /run/kmsg... */ (void) unlink(from); @@ -1911,132 +1557,6 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { return 0; } -static int send_rtnl(int send_fd) { - 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; - _cleanup_close_ int fd = -1; - ssize_t k; - - assert(send_fd >= 0); - - if (!arg_expose_ports) - return 0; - - fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, NETLINK_ROUTE); - if (fd < 0) - return log_error_errno(errno, "Failed to allocate container netlink: %m"); - - 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), &fd, sizeof(int)); - - mh.msg_controllen = cmsg->cmsg_len; - - /* Store away the fd in the socket, so that it stays open as - * long as we run the child */ - k = sendmsg(send_fd, &mh, MSG_NOSIGNAL); - if (k < 0) - return log_error_errno(errno, "Failed to send netlink fd: %m"); - - return 0; -} - -static int flush_ports(union in_addr_union *exposed) { - ExposePort *p; - int r, af = AF_INET; - - assert(exposed); - - if (!arg_expose_ports) - return 0; - - if (in_addr_is_null(af, exposed)) - return 0; - - log_debug("Lost IP address."); - - LIST_FOREACH(ports, p, arg_expose_ports) { - r = fw_add_local_dnat(false, - af, - p->protocol, - NULL, - NULL, 0, - NULL, 0, - p->host_port, - exposed, - p->container_port, - NULL); - if (r < 0) - log_warning_errno(r, "Failed to modify firewall: %m"); - } - - *exposed = IN_ADDR_NULL; - return 0; -} - -static int expose_ports(sd_netlink *rtnl, union in_addr_union *exposed) { - _cleanup_free_ struct local_address *addresses = NULL; - _cleanup_free_ char *pretty = NULL; - union in_addr_union new_exposed; - ExposePort *p; - bool add; - int af = AF_INET, r; - - assert(exposed); - - /* Invoked each time an address is added or removed inside the - * container */ - - if (!arg_expose_ports) - return 0; - - r = local_addresses(rtnl, 0, af, &addresses); - if (r < 0) - return log_error_errno(r, "Failed to enumerate local addresses: %m"); - - add = r > 0 && - addresses[0].family == af && - addresses[0].scope < RT_SCOPE_LINK; - - if (!add) - return flush_ports(exposed); - - new_exposed = addresses[0].address; - if (in_addr_equal(af, exposed, &new_exposed)) - return 0; - - in_addr_to_string(af, &new_exposed, &pretty); - log_debug("New container IP is %s.", strna(pretty)); - - LIST_FOREACH(ports, p, arg_expose_ports) { - - r = fw_add_local_dnat(true, - af, - p->protocol, - NULL, - NULL, 0, - NULL, 0, - p->host_port, - &new_exposed, - p->container_port, - in_addr_is_null(af, exposed) ? NULL : exposed); - if (r < 0) - log_warning_errno(r, "Failed to modify firewall: %m"); - } - - *exposed = new_exposed; - return 0; -} - static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) { union in_addr_union *exposed = userdata; @@ -2044,68 +1564,13 @@ static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *user assert(m); assert(exposed); - expose_ports(rtnl, exposed); - return 0; -} - -static int watch_rtnl(sd_event *event, int recv_fd, union in_addr_union *exposed, sd_netlink **ret) { - 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; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - int fd, r; - ssize_t k; - - assert(event); - assert(recv_fd >= 0); - assert(ret); - - if (!arg_expose_ports) - return 0; - - k = recvmsg(recv_fd, &mh, MSG_NOSIGNAL); - if (k < 0) - return log_error_errno(errno, "Failed to recv netlink fd: %m"); - - cmsg = CMSG_FIRSTHDR(&mh); - assert(cmsg->cmsg_level == SOL_SOCKET); - assert(cmsg->cmsg_type == SCM_RIGHTS); - assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int))); - memcpy(&fd, CMSG_DATA(cmsg), sizeof(int)); - - r = sd_netlink_open_fd(&rtnl, fd); - if (r < 0) { - safe_close(fd); - return log_error_errno(r, "Failed to create rtnl object: %m"); - } - - r = sd_netlink_add_match(rtnl, RTM_NEWADDR, on_address_change, exposed); - if (r < 0) - return log_error_errno(r, "Failed to subscribe to RTM_NEWADDR messages: %m"); - - r = sd_netlink_add_match(rtnl, RTM_DELADDR, on_address_change, exposed); - if (r < 0) - return log_error_errno(r, "Failed to subscribe to RTM_DELADDR messages: %m"); - - r = sd_netlink_attach_event(rtnl, event, 0); - if (r < 0) - return log_error_errno(r, "Failed to add to even loop: %m"); - - *ret = rtnl; - rtnl = NULL; - + expose_port_execute(rtnl, arg_expose_ports, exposed); return 0; } static int setup_hostname(void) { - if (arg_share_system) + if ((arg_clone_ns_flags & CLONE_NEWUTS) == 0) return 0; if (sethostname_idempotent(arg_machine) < 0) @@ -2115,48 +1580,34 @@ static int setup_hostname(void) { } static int setup_journal(const char *directory) { - sd_id128_t machine_id, this_id; - _cleanup_free_ char *b = NULL, *d = NULL; - const char *etc_machine_id, *p, *q; - char *id; + sd_id128_t this_id; + _cleanup_free_ char *d = NULL; + const char *p, *q; + bool try; + char id[33]; int r; /* Don't link journals in ephemeral mode */ if (arg_ephemeral) return 0; - etc_machine_id = prefix_roota(directory, "/etc/machine-id"); - - r = read_one_line_file(etc_machine_id, &b); - if (r == -ENOENT && arg_link_journal == LINK_AUTO) - return 0; - else if (r < 0) - return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id); - - id = strstrip(b); - if (isempty(id) && arg_link_journal == LINK_AUTO) + if (arg_link_journal == LINK_NO) return 0; - /* Verify validity */ - r = sd_id128_from_string(id, &machine_id); - if (r < 0) - return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id); + try = arg_link_journal_try || arg_link_journal == LINK_AUTO; r = sd_id128_get_machine(&this_id); if (r < 0) return log_error_errno(r, "Failed to retrieve machine ID: %m"); - if (sd_id128_equal(machine_id, this_id)) { - log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR, - "Host and machine ids are equal (%s): refusing to link journals", id); - if (arg_link_journal == LINK_AUTO) + if (sd_id128_equal(arg_uuid, this_id)) { + log_full(try ? LOG_WARNING : LOG_ERR, + "Host and machine ids are equal (%s): refusing to link journals", sd_id128_to_string(arg_uuid, id)); + if (try) return 0; return -EEXIST; } - if (arg_link_journal == LINK_NO) - return 0; - r = userns_mkdir(directory, "/var", 0755, 0, 0); if (r < 0) return log_error_errno(r, "Failed to create /var: %m"); @@ -2169,25 +1620,25 @@ static int setup_journal(const char *directory) { if (r < 0) return log_error_errno(r, "Failed to create /var/log/journal: %m"); + (void) sd_id128_to_string(arg_uuid, id); + p = strjoina("/var/log/journal/", id); q = prefix_roota(directory, p); if (path_is_mount_point(p, 0) > 0) { - if (arg_link_journal != LINK_AUTO) { - log_error("%s: already a mount point, refusing to use for journal", p); - return -EEXIST; - } + if (try) + return 0; - return 0; + log_error("%s: already a mount point, refusing to use for journal", p); + return -EEXIST; } if (path_is_mount_point(q, 0) > 0) { - if (arg_link_journal != LINK_AUTO) { - log_error("%s: already a mount point, refusing to use for journal", q); - return -EEXIST; - } + if (try) + return 0; - return 0; + log_error("%s: already a mount point, refusing to use for journal", q); + return -EEXIST; } r = readlink_and_make_absolute(p, &d); @@ -2198,7 +1649,7 @@ static int setup_journal(const char *directory) { r = userns_mkdir(directory, p, 0755, 0, 0); if (r < 0) - log_warning_errno(errno, "Failed to create directory %s: %m", q); + log_warning_errno(r, "Failed to create directory %s: %m", q); return 0; } @@ -2212,46 +1663,38 @@ static int setup_journal(const char *directory) { if (errno == ENOTDIR) { log_error("%s already exists and is neither a symlink nor a directory", p); return r; - } else { - log_error_errno(errno, "Failed to remove %s: %m", p); - return -errno; - } + } else + return log_error_errno(errno, "Failed to remove %s: %m", p); } - } else if (r != -ENOENT) { - log_error_errno(errno, "readlink(%s) failed: %m", p); - return r; - } + } else if (r != -ENOENT) + return log_error_errno(r, "readlink(%s) failed: %m", p); if (arg_link_journal == LINK_GUEST) { if (symlink(q, p) < 0) { - if (arg_link_journal_try) { + if (try) { log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p); return 0; - } else { - log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p); - return -errno; - } + } else + return log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p); } r = userns_mkdir(directory, p, 0755, 0, 0); if (r < 0) - log_warning_errno(errno, "Failed to create directory %s: %m", q); + log_warning_errno(r, "Failed to create directory %s: %m", q); return 0; } if (arg_link_journal == LINK_HOST) { - /* don't create parents here -- if the host doesn't have + /* don't create parents here — if the host doesn't have * permanent journal set up, don't force it here */ - r = mkdir(p, 0755); - if (r < 0) { - if (arg_link_journal_try) { + + if (mkdir(p, 0755) < 0 && errno != EEXIST) { + if (try) { log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p); return 0; - } else { - log_error_errno(errno, "Failed to create %s: %m", p); - return r; - } + } else + return log_error_errno(errno, "Failed to create %s: %m", p); } } else if (access(p, F_OK) < 0) @@ -2261,240 +1704,25 @@ static int setup_journal(const char *directory) { log_warning("%s is not empty, proceeding anyway.", q); r = userns_mkdir(directory, p, 0755, 0, 0); - if (r < 0) { - log_error_errno(errno, "Failed to create %s: %m", q); - return r; - } + if (r < 0) + return log_error_errno(r, "Failed to create %s: %m", q); - if (mount(p, q, NULL, MS_BIND, NULL) < 0) + r = mount_verbose(LOG_DEBUG, p, q, NULL, MS_BIND, NULL); + if (r < 0) return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m"); return 0; } static int drop_capabilities(void) { - return capability_bounding_set_drop(~arg_retain, false); -} - -static int register_machine(pid_t pid, int local_ifindex) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_bus_flush_close_unref_ sd_bus *bus = NULL; - int r; - - if (!arg_register) - return 0; - - r = sd_bus_default_system(&bus); - if (r < 0) - return log_error_errno(r, "Failed to open system bus: %m"); - - if (arg_keep_unit) { - r = sd_bus_call_method( - bus, - "org.freedesktop.machine1", - "/org/freedesktop/machine1", - "org.freedesktop.machine1.Manager", - "RegisterMachineWithNetwork", - &error, - NULL, - "sayssusai", - arg_machine, - SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), - "nspawn", - "container", - (uint32_t) pid, - strempty(arg_directory), - local_ifindex > 0 ? 1 : 0, local_ifindex); - } else { - _cleanup_bus_message_unref_ sd_bus_message *m = NULL; - char **i; - unsigned j; - - r = sd_bus_message_new_method_call( - bus, - &m, - "org.freedesktop.machine1", - "/org/freedesktop/machine1", - "org.freedesktop.machine1.Manager", - "CreateMachineWithNetwork"); - if (r < 0) - return bus_log_create_error(r); - - r = sd_bus_message_append( - m, - "sayssusai", - arg_machine, - SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), - "nspawn", - "container", - (uint32_t) pid, - strempty(arg_directory), - local_ifindex > 0 ? 1 : 0, local_ifindex); - if (r < 0) - return bus_log_create_error(r); - - r = sd_bus_message_open_container(m, 'a', "(sv)"); - if (r < 0) - return bus_log_create_error(r); - - if (!isempty(arg_slice)) { - r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice); - if (r < 0) - return bus_log_create_error(r); - } - - r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict"); - if (r < 0) - return bus_log_create_error(r); - - /* If you make changes here, also make sure to update - * systemd-nspawn@.service, to keep the device - * policies in sync regardless if we are run with or - * without the --keep-unit switch. */ - r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 9, - /* Allow the container to - * access and create the API - * device nodes, so that - * PrivateDevices= in the - * container can work - * fine */ - "/dev/null", "rwm", - "/dev/zero", "rwm", - "/dev/full", "rwm", - "/dev/random", "rwm", - "/dev/urandom", "rwm", - "/dev/tty", "rwm", - "/dev/net/tun", "rwm", - /* Allow the container - * access to ptys. However, - * do not permit the - * container to ever create - * these device nodes. */ - "/dev/pts/ptmx", "rw", - "char-pts", "rw"); - if (r < 0) - return bus_log_create_error(r); - - for (j = 0; j < arg_n_custom_mounts; j++) { - CustomMount *cm = &arg_custom_mounts[j]; - - if (cm->type != CUSTOM_MOUNT_BIND) - continue; - - r = is_device_node(cm->source); - if (r < 0) - return log_error_errno(r, "Failed to stat %s: %m", cm->source); - - if (r) { - r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 1, - cm->source, cm->read_only ? "r" : "rw"); - if (r < 0) - return log_error_errno(r, "Failed to append message arguments: %m"); - } - } - - if (arg_kill_signal != 0) { - r = sd_bus_message_append(m, "(sv)", "KillSignal", "i", arg_kill_signal); - if (r < 0) - return bus_log_create_error(r); - - r = sd_bus_message_append(m, "(sv)", "KillMode", "s", "mixed"); - if (r < 0) - return bus_log_create_error(r); - } - - STRV_FOREACH(i, arg_property) { - r = sd_bus_message_open_container(m, 'r', "sv"); - if (r < 0) - return bus_log_create_error(r); - - r = bus_append_unit_property_assignment(m, *i); - if (r < 0) - return r; - - r = sd_bus_message_close_container(m); - if (r < 0) - return bus_log_create_error(r); - } - - r = sd_bus_message_close_container(m); - if (r < 0) - return bus_log_create_error(r); - - r = sd_bus_call(bus, m, 0, &error, NULL); - } - - if (r < 0) { - log_error("Failed to register machine: %s", bus_error_message(&error, r)); - return r; - } - - return 0; -} - -static int terminate_machine(pid_t pid) { - _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; - _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; - _cleanup_bus_flush_close_unref_ sd_bus *bus = NULL; - const char *path; - int r; - - if (!arg_register) - return 0; - - /* If we are reusing the unit, then just exit, systemd will do - * the right thing when we exit. */ - if (arg_keep_unit) - return 0; - - r = sd_bus_default_system(&bus); - if (r < 0) - return log_error_errno(r, "Failed to open system bus: %m"); - - r = sd_bus_call_method( - bus, - "org.freedesktop.machine1", - "/org/freedesktop/machine1", - "org.freedesktop.machine1.Manager", - "GetMachineByPID", - &error, - &reply, - "u", - (uint32_t) pid); - if (r < 0) { - /* Note that the machine might already have been - * cleaned up automatically, hence don't consider it a - * failure if we cannot get the machine object. */ - log_debug("Failed to get machine: %s", bus_error_message(&error, r)); - return 0; - } - - r = sd_bus_message_read(reply, "o", &path); - if (r < 0) - return bus_log_parse_error(r); - - r = sd_bus_call_method( - bus, - "org.freedesktop.machine1", - path, - "org.freedesktop.machine1.Machine", - "Terminate", - &error, - NULL, - NULL); - if (r < 0) { - log_debug("Failed to terminate machine: %s", bus_error_message(&error, r)); - return 0; - } - - return 0; + return capability_bounding_set_drop(arg_caps_retain, false); } static int reset_audit_loginuid(void) { _cleanup_free_ char *p = NULL; int r; - if (arg_share_system) + if ((arg_clone_ns_flags & CLONE_NEWPID) == 0) return 0; r = read_one_line_file("/proc/self/loginuid", &p); @@ -2522,547 +1750,40 @@ static int reset_audit_loginuid(void) { return 0; } -#define HOST_HASH_KEY SD_ID128_MAKE(1a,37,6f,c7,46,ec,45,0b,ad,a3,d5,31,06,60,5d,b1) -#define CONTAINER_HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2) -#define MACVLAN_HASH_KEY SD_ID128_MAKE(00,13,6d,bc,66,83,44,81,bb,0c,f9,51,1f,24,a6,6f) - -static int generate_mac(struct ether_addr *mac, sd_id128_t hash_key, uint64_t idx) { - uint8_t result[8]; - size_t l, sz; - uint8_t *v, *i; - int r; - - l = strlen(arg_machine); - sz = sizeof(sd_id128_t) + l; - if (idx > 0) - sz += sizeof(idx); - - v = alloca(sz); - - /* fetch some persistent data unique to the host */ - r = sd_id128_get_machine((sd_id128_t*) v); - if (r < 0) - return r; - - /* combine with some data unique (on this host) to this - * container instance */ - i = mempcpy(v + sizeof(sd_id128_t), arg_machine, l); - if (idx > 0) { - idx = htole64(idx); - memcpy(i, &idx, sizeof(idx)); - } - - /* Let's hash the host machine ID plus the container name. We - * use a fixed, but originally randomly created hash key here. */ - siphash24(result, v, sz, hash_key.bytes); - - assert_cc(ETH_ALEN <= sizeof(result)); - memcpy(mac->ether_addr_octet, result, ETH_ALEN); - - /* see eth_random_addr in the kernel */ - mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */ - mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */ - - return 0; -} - -static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ], int *ifi) { - _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - struct ether_addr mac_host, mac_container; - int r, i; - - if (!arg_private_network) - return 0; - - if (!arg_network_veth) - return 0; - - /* Use two different interface name prefixes depending whether - * we are in bridge mode or not. */ - snprintf(iface_name, IFNAMSIZ - 1, "%s-%s", - arg_network_bridge ? "vb" : "ve", arg_machine); - - r = generate_mac(&mac_container, CONTAINER_HASH_KEY, 0); - if (r < 0) - return log_error_errno(r, "Failed to generate predictable MAC address for container side: %m"); - - r = generate_mac(&mac_host, HOST_HASH_KEY, 0); - if (r < 0) - return log_error_errno(r, "Failed to generate predictable MAC address for host side: %m"); - - r = sd_netlink_open(&rtnl); - if (r < 0) - return log_error_errno(r, "Failed to connect to netlink: %m"); - - r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); - if (r < 0) - return log_error_errno(r, "Failed to allocate netlink message: %m"); - - r = sd_netlink_message_append_string(m, IFLA_IFNAME, iface_name); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface name: %m"); - - r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac_host); - if (r < 0) - return log_error_errno(r, "Failed to add netlink MAC address: %m"); - - r = sd_netlink_message_open_container(m, IFLA_LINKINFO); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "veth"); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_open_container(m, VETH_INFO_PEER); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_append_string(m, IFLA_IFNAME, "host0"); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface name: %m"); - - r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac_container); - if (r < 0) - return log_error_errno(r, "Failed to add netlink MAC address: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid); - if (r < 0) - return log_error_errno(r, "Failed to add netlink namespace field: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_call(rtnl, m, 0, NULL); - if (r < 0) - return log_error_errno(r, "Failed to add new veth interfaces (host0, %s): %m", iface_name); - - i = (int) if_nametoindex(iface_name); - if (i <= 0) - return log_error_errno(errno, "Failed to resolve interface %s: %m", iface_name); - - *ifi = i; - - return 0; -} - -static int setup_bridge(const char veth_name[], int *ifi) { - _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - int r, bridge; - - if (!arg_private_network) - return 0; - - if (!arg_network_veth) - return 0; - - if (!arg_network_bridge) - return 0; - - bridge = (int) if_nametoindex(arg_network_bridge); - if (bridge <= 0) - return log_error_errno(errno, "Failed to resolve interface %s: %m", arg_network_bridge); - - *ifi = bridge; - - r = sd_netlink_open(&rtnl); - if (r < 0) - return log_error_errno(r, "Failed to connect to netlink: %m"); - - r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0); - if (r < 0) - return log_error_errno(r, "Failed to allocate netlink message: %m"); - - r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP); - if (r < 0) - return log_error_errno(r, "Failed to set IFF_UP flag: %m"); - - r = sd_netlink_message_append_string(m, IFLA_IFNAME, veth_name); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface name field: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_MASTER, bridge); - if (r < 0) - return log_error_errno(r, "Failed to add netlink master field: %m"); - - r = sd_netlink_call(rtnl, m, 0, NULL); - if (r < 0) - return log_error_errno(r, "Failed to add veth interface to bridge: %m"); - - return 0; -} - -static int parse_interface(struct udev *udev, const char *name) { - _cleanup_udev_device_unref_ struct udev_device *d = NULL; - char ifi_str[2 + DECIMAL_STR_MAX(int)]; - int ifi; - - ifi = (int) if_nametoindex(name); - if (ifi <= 0) - return log_error_errno(errno, "Failed to resolve interface %s: %m", name); - - sprintf(ifi_str, "n%i", ifi); - d = udev_device_new_from_device_id(udev, ifi_str); - if (!d) - return log_error_errno(errno, "Failed to get udev device for interface %s: %m", name); - - if (udev_device_get_is_initialized(d) <= 0) { - log_error("Network interface %s is not initialized yet.", name); - return -EBUSY; - } - - return ifi; -} - -static int move_network_interfaces(pid_t pid) { - _cleanup_udev_unref_ struct udev *udev = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - char **i; - int r; - - if (!arg_private_network) - return 0; - - if (strv_isempty(arg_network_interfaces)) - return 0; - - r = sd_netlink_open(&rtnl); - if (r < 0) - return log_error_errno(r, "Failed to connect to netlink: %m"); - - udev = udev_new(); - if (!udev) { - log_error("Failed to connect to udev."); - return -ENOMEM; - } - - STRV_FOREACH(i, arg_network_interfaces) { - _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL; - int ifi; - - ifi = parse_interface(udev, *i); - if (ifi < 0) - return ifi; - - r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, ifi); - if (r < 0) - return log_error_errno(r, "Failed to allocate netlink message: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid); - if (r < 0) - return log_error_errno(r, "Failed to append namespace PID to netlink message: %m"); - - r = sd_netlink_call(rtnl, m, 0, NULL); - if (r < 0) - return log_error_errno(r, "Failed to move interface %s to namespace: %m", *i); - } - - return 0; -} - -static int setup_macvlan(pid_t pid) { - _cleanup_udev_unref_ struct udev *udev = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - unsigned idx = 0; - char **i; - int r; - - if (!arg_private_network) - return 0; - - if (strv_isempty(arg_network_macvlan)) - return 0; - - r = sd_netlink_open(&rtnl); - if (r < 0) - return log_error_errno(r, "Failed to connect to netlink: %m"); - - udev = udev_new(); - if (!udev) { - log_error("Failed to connect to udev."); - return -ENOMEM; - } - - STRV_FOREACH(i, arg_network_macvlan) { - _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL; - _cleanup_free_ char *n = NULL; - struct ether_addr mac; - int ifi; - - ifi = parse_interface(udev, *i); - if (ifi < 0) - return ifi; - - r = generate_mac(&mac, MACVLAN_HASH_KEY, idx++); - if (r < 0) - return log_error_errno(r, "Failed to create MACVLAN MAC address: %m"); - - r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); - if (r < 0) - return log_error_errno(r, "Failed to allocate netlink message: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_LINK, ifi); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface index: %m"); - - n = strappend("mv-", *i); - if (!n) - return log_oom(); - - strshorten(n, IFNAMSIZ-1); - - r = sd_netlink_message_append_string(m, IFLA_IFNAME, n); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface name: %m"); - - r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac); - if (r < 0) - return log_error_errno(r, "Failed to add netlink MAC address: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid); - if (r < 0) - return log_error_errno(r, "Failed to add netlink namespace field: %m"); - - r = sd_netlink_message_open_container(m, IFLA_LINKINFO); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "macvlan"); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE); - if (r < 0) - return log_error_errno(r, "Failed to append macvlan mode: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_call(rtnl, m, 0, NULL); - if (r < 0) - return log_error_errno(r, "Failed to add new macvlan interfaces: %m"); - } - - return 0; -} - -static int setup_ipvlan(pid_t pid) { - _cleanup_udev_unref_ struct udev *udev = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - char **i; - int r; - - if (!arg_private_network) - return 0; - - if (strv_isempty(arg_network_ipvlan)) - return 0; - - r = sd_netlink_open(&rtnl); - if (r < 0) - return log_error_errno(r, "Failed to connect to netlink: %m"); - - udev = udev_new(); - if (!udev) { - log_error("Failed to connect to udev."); - return -ENOMEM; - } - - STRV_FOREACH(i, arg_network_ipvlan) { - _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL; - _cleanup_free_ char *n = NULL; - int ifi; - - ifi = parse_interface(udev, *i); - if (ifi < 0) - return ifi; - - r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); - if (r < 0) - return log_error_errno(r, "Failed to allocate netlink message: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_LINK, ifi); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface index: %m"); - - n = strappend("iv-", *i); - if (!n) - return log_oom(); - - strshorten(n, IFNAMSIZ-1); - - r = sd_netlink_message_append_string(m, IFLA_IFNAME, n); - if (r < 0) - return log_error_errno(r, "Failed to add netlink interface name: %m"); - - r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid); - if (r < 0) - return log_error_errno(r, "Failed to add netlink namespace field: %m"); - - r = sd_netlink_message_open_container(m, IFLA_LINKINFO); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "ipvlan"); - if (r < 0) - return log_error_errno(r, "Failed to open netlink container: %m"); - - r = sd_netlink_message_append_u16(m, IFLA_IPVLAN_MODE, IPVLAN_MODE_L2); - if (r < 0) - return log_error_errno(r, "Failed to add ipvlan mode: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_message_close_container(m); - if (r < 0) - return log_error_errno(r, "Failed to close netlink container: %m"); - - r = sd_netlink_call(rtnl, m, 0, NULL); - if (r < 0) - return log_error_errno(r, "Failed to add new ipvlan interfaces: %m"); - } - - return 0; -} - -static int setup_seccomp(void) { - -#ifdef HAVE_SECCOMP - static const struct { - uint64_t capability; - int syscall_num; - } blacklist[] = { - { CAP_SYS_RAWIO, SCMP_SYS(iopl) }, - { CAP_SYS_RAWIO, SCMP_SYS(ioperm) }, - { CAP_SYS_BOOT, SCMP_SYS(kexec_load) }, - { CAP_SYS_ADMIN, SCMP_SYS(swapon) }, - { CAP_SYS_ADMIN, SCMP_SYS(swapoff) }, - { CAP_SYS_ADMIN, SCMP_SYS(open_by_handle_at) }, - { CAP_SYS_MODULE, SCMP_SYS(init_module) }, - { CAP_SYS_MODULE, SCMP_SYS(finit_module) }, - { CAP_SYS_MODULE, SCMP_SYS(delete_module) }, - { CAP_SYSLOG, SCMP_SYS(syslog) }, - }; - - scmp_filter_ctx seccomp; - unsigned i; - int r; - - seccomp = seccomp_init(SCMP_ACT_ALLOW); - if (!seccomp) - return log_oom(); - - r = seccomp_add_secondary_archs(seccomp); - if (r < 0) { - log_error_errno(r, "Failed to add secondary archs to seccomp filter: %m"); - goto finish; - } - - for (i = 0; i < ELEMENTSOF(blacklist); i++) { - if (arg_retain & (1ULL << blacklist[i].capability)) - continue; - - r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), blacklist[i].syscall_num, 0); - if (r == -EFAULT) - continue; /* unknown syscall */ - if (r < 0) { - log_error_errno(r, "Failed to block syscall: %m"); - goto finish; - } - } - - - /* - Audit is broken in containers, much of the userspace audit - hookup will fail if running inside a container. We don't - care and just turn off creation of audit sockets. - - This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail - with EAFNOSUPPORT which audit userspace uses as indication - that audit is disabled in the kernel. - */ - - r = seccomp_rule_add( - seccomp, - SCMP_ACT_ERRNO(EAFNOSUPPORT), - SCMP_SYS(socket), - 2, - SCMP_A0(SCMP_CMP_EQ, AF_NETLINK), - SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT)); - if (r < 0) { - log_error_errno(r, "Failed to add audit seccomp rule: %m"); - goto finish; - } - - r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); - if (r < 0) { - log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m"); - goto finish; - } - - r = seccomp_load(seccomp); - if (r == -EINVAL) { - log_debug_errno(r, "Kernel is probably not configured with CONFIG_SECCOMP. Disabling seccomp audit filter: %m"); - r = 0; - goto finish; - } - if (r < 0) { - log_error_errno(r, "Failed to install seccomp audit filter: %m"); - goto finish; - } - -finish: - seccomp_release(seccomp); - return r; -#else - return 0; -#endif - -} static int setup_propagate(const char *root) { const char *p, *q; + int r; (void) mkdir_p("/run/systemd/nspawn/", 0755); (void) mkdir_p("/run/systemd/nspawn/propagate", 0600); p = strjoina("/run/systemd/nspawn/propagate/", arg_machine); (void) mkdir_p(p, 0600); - if (userns_mkdir(root, "/run/systemd", 0755, 0, 0) < 0) - return log_error_errno(errno, "Failed to create /run/systemd: %m"); + r = userns_mkdir(root, "/run/systemd", 0755, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to create /run/systemd: %m"); - if (userns_mkdir(root, "/run/systemd/nspawn", 0755, 0, 0) < 0) - return log_error_errno(errno, "Failed to create /run/systemd/nspawn: %m"); + r = userns_mkdir(root, "/run/systemd/nspawn", 0755, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to create /run/systemd/nspawn: %m"); - if (userns_mkdir(root, "/run/systemd/nspawn/incoming", 0600, 0, 0) < 0) - return log_error_errno(errno, "Failed to create /run/systemd/nspawn/incoming: %m"); + r = userns_mkdir(root, "/run/systemd/nspawn/incoming", 0600, 0, 0); + if (r < 0) + return log_error_errno(r, "Failed to create /run/systemd/nspawn/incoming: %m"); q = prefix_roota(root, "/run/systemd/nspawn/incoming"); - if (mount(p, q, NULL, MS_BIND, NULL) < 0) - return log_error_errno(errno, "Failed to install propagation bind mount."); + r = mount_verbose(LOG_ERR, p, q, NULL, MS_BIND, NULL); + if (r < 0) + return r; - if (mount(NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0) - return log_error_errno(errno, "Failed to make propagation mount read-only"); + r = mount_verbose(LOG_ERR, NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + if (r < 0) + return r; - return 0; + /* machined will MS_MOVE into that directory, and that's only + * supported for non-shared mounts. */ + return mount_verbose(LOG_ERR, NULL, q, NULL, MS_SLAVE, NULL); } static int setup_image(char **device_path, int *loop_nr) { @@ -3103,7 +1824,7 @@ static int setup_image(char **device_path, int *loop_nr) { } if (!S_ISREG(st.st_mode)) { - log_error_errno(errno, "%s is not a regular file or block device: %m", arg_image); + log_error("%s is not a regular file or block device.", arg_image); return -EINVAL; } @@ -3154,17 +1875,18 @@ static int dissect_image( char **root_device, bool *root_device_rw, char **home_device, bool *home_device_rw, char **srv_device, bool *srv_device_rw, + char **esp_device, bool *secondary) { #ifdef HAVE_BLKID - int home_nr = -1, srv_nr = -1; + int home_nr = -1, srv_nr = -1, esp_nr = -1; #ifdef GPT_ROOT_NATIVE int root_nr = -1; #endif #ifdef GPT_ROOT_SECONDARY int secondary_root_nr = -1; #endif - _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *generic = NULL; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *esp = NULL, *generic = NULL; _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; _cleanup_udev_device_unref_ struct udev_device *d = NULL; _cleanup_blkid_free_probe_ blkid_probe b = NULL; @@ -3182,6 +1904,7 @@ static int dissect_image( assert(root_device); assert(home_device); assert(srv_device); + assert(esp_device); assert(secondary); assert(arg_image); @@ -3195,8 +1918,7 @@ static int dissect_image( if (errno == 0) return log_oom(); - log_error_errno(errno, "Failed to set device on blkid probe: %m"); - return -errno; + return log_error_errno(errno, "Failed to set device on blkid probe: %m"); } blkid_probe_enable_partitions(b, 1); @@ -3212,8 +1934,7 @@ static int dissect_image( } else if (r != 0) { if (errno == 0) errno = EIO; - log_error_errno(errno, "Failed to probe: %m"); - return -errno; + return log_error_errno(errno, "Failed to probe: %m"); } (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); @@ -3336,8 +2057,7 @@ static int dissect_image( if (!errno) errno = ENOMEM; - log_error_errno(errno, "Failed to get partition device of %s: %m", arg_image); - return -errno; + return log_error_errno(errno, "Failed to get partition device of %s: %m", arg_image); } qn = udev_device_get_devnum(q); @@ -3398,6 +2118,16 @@ static int dissect_image( r = free_and_strdup(&srv, node); if (r < 0) return log_oom(); + } else if (sd_id128_equal(type_id, GPT_ESP)) { + + if (esp && nr >= esp_nr) + continue; + + esp_nr = nr; + + r = free_and_strdup(&esp, node); + if (r < 0) + return log_oom(); } #ifdef GPT_ROOT_NATIVE else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { @@ -3515,6 +2245,11 @@ static int dissect_image( *srv_device_rw = srv_rw; } + if (esp) { + *esp_device = esp; + esp = NULL; + } + return 0; #else log_error("--image= is not supported, compiled without blkid support."); @@ -3544,8 +2279,7 @@ static int mount_device(const char *what, const char *where, const char *directo if (!b) { if (errno == 0) return log_oom(); - log_error_errno(errno, "Failed to allocate prober for %s: %m", what); - return -errno; + return log_error_errno(errno, "Failed to allocate prober for %s: %m", what); } blkid_probe_enable_superblocks(b, 1); @@ -3559,8 +2293,7 @@ static int mount_device(const char *what, const char *where, const char *directo } else if (r != 0) { if (errno == 0) errno = EIO; - log_error_errno(errno, "Failed to probe %s: %m", what); - return -errno; + return log_error_errno(errno, "Failed to probe %s: %m", what); } errno = 0; @@ -3576,21 +2309,78 @@ static int mount_device(const char *what, const char *where, const char *directo return -EOPNOTSUPP; } - if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) - return log_error_errno(errno, "Failed to mount %s: %m", what); - - return 0; + return mount_verbose(LOG_ERR, what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL); #else log_error("--image= is not supported, compiled without blkid support."); return -EOPNOTSUPP; #endif } +static int setup_machine_id(const char *directory) { + const char *etc_machine_id; + sd_id128_t id; + int r; + + /* If the UUID in the container is already set, then that's what counts, and we use. If it isn't set, and the + * caller passed --uuid=, then we'll pass it in the $container_uuid env var to PID 1 of the container. The + * assumption is that PID 1 will then write it to /etc/machine-id to make it persistent. If --uuid= is not + * passed we generate a random UUID, and pass it via $container_uuid. In effect this means that /etc/machine-id + * in the container and our idea of the container UUID will always be in sync (at least if PID 1 in the + * container behaves nicely). */ + + etc_machine_id = prefix_roota(directory, "/etc/machine-id"); + + r = id128_read(etc_machine_id, ID128_PLAIN, &id); + if (r < 0) { + if (!IN_SET(r, -ENOENT, -ENOMEDIUM)) /* If the file is missing or empty, we don't mind */ + return log_error_errno(r, "Failed to read machine ID from container image: %m"); + + if (sd_id128_is_null(arg_uuid)) { + r = sd_id128_randomize(&arg_uuid); + if (r < 0) + return log_error_errno(r, "Failed to acquire randomized machine UUID: %m"); + } + } else { + if (sd_id128_is_null(id)) { + log_error("Machine ID in container image is zero, refusing."); + return -EINVAL; + } + + arg_uuid = id; + } + + return 0; +} + +static int recursive_chown(const char *directory, uid_t shift, uid_t range) { + int r; + + assert(directory); + + if (arg_userns_mode == USER_NAMESPACE_NO || !arg_userns_chown) + return 0; + + r = path_patch_uid(directory, arg_uid_shift, arg_uid_range); + if (r == -EOPNOTSUPP) + return log_error_errno(r, "Automatic UID/GID adjusting is only supported for UID/GID ranges starting at multiples of 2^16 with a range of 2^16."); + if (r == -EBADE) + return log_error_errno(r, "Upper 16 bits of root directory UID and GID do not match."); + if (r < 0) + return log_error_errno(r, "Failed to adjust UID/GID shift of OS tree: %m"); + if (r == 0) + log_debug("Root directory of image is already owned by the right UID/GID range, skipping recursive chown operation."); + else + log_debug("Patched directory tree to match UID/GID range."); + + return r; +} + static int mount_devices( const char *where, const char *root_device, bool root_device_rw, const char *home_device, bool home_device_rw, - const char *srv_device, bool srv_device_rw) { + const char *srv_device, bool srv_device_rw, + const char *esp_device) { int r; assert(where); @@ -3613,6 +2403,27 @@ static int mount_devices( return log_error_errno(r, "Failed to mount server data directory: %m"); } + if (esp_device) { + const char *mp, *x; + + /* Mount the ESP to /efi if it exists and is empty. If it doesn't exist, use /boot instead. */ + + mp = "/efi"; + x = strjoina(arg_directory, mp); + r = dir_is_empty(x); + if (r == -ENOENT) { + mp = "/boot"; + x = strjoina(arg_directory, mp); + r = dir_is_empty(x); + } + + if (r > 0) { + r = mount_device(esp_device, arg_directory, mp, true); + if (r < 0) + return log_error_errno(r, "Failed to mount ESP: %m"); + } + } + return 0; } @@ -3641,247 +2452,6 @@ static void loop_remove(int nr, int *image_fd) { log_debug_errno(errno, "Failed to remove loop %d: %m", nr); } -static int spawn_getent(const char *database, const char *key, pid_t *rpid) { - int pipe_fds[2]; - pid_t pid; - - assert(database); - assert(key); - assert(rpid); - - if (pipe2(pipe_fds, O_CLOEXEC) < 0) - return log_error_errno(errno, "Failed to allocate pipe: %m"); - - pid = fork(); - if (pid < 0) - return log_error_errno(errno, "Failed to fork getent child: %m"); - else if (pid == 0) { - int nullfd; - char *empty_env = NULL; - - if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0) - _exit(EXIT_FAILURE); - - if (pipe_fds[0] > 2) - safe_close(pipe_fds[0]); - if (pipe_fds[1] > 2) - safe_close(pipe_fds[1]); - - nullfd = open("/dev/null", O_RDWR); - if (nullfd < 0) - _exit(EXIT_FAILURE); - - if (dup3(nullfd, STDIN_FILENO, 0) < 0) - _exit(EXIT_FAILURE); - - if (dup3(nullfd, STDERR_FILENO, 0) < 0) - _exit(EXIT_FAILURE); - - if (nullfd > 2) - safe_close(nullfd); - - (void) reset_all_signal_handlers(); - (void) reset_signal_mask(); - close_all_fds(NULL, 0); - - execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env); - execle("/bin/getent", "getent", database, key, NULL, &empty_env); - _exit(EXIT_FAILURE); - } - - pipe_fds[1] = safe_close(pipe_fds[1]); - - *rpid = pid; - - return pipe_fds[0]; -} - -static int change_uid_gid(char **_home) { - char line[LINE_MAX], *x, *u, *g, *h; - const char *word, *state; - _cleanup_free_ uid_t *uids = NULL; - _cleanup_free_ char *home = NULL; - _cleanup_fclose_ FILE *f = NULL; - _cleanup_close_ int fd = -1; - unsigned n_uids = 0; - size_t sz = 0, l; - uid_t uid; - gid_t gid; - pid_t pid; - int r; - - assert(_home); - - if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) { - /* Reset everything fully to 0, just in case */ - - r = reset_uid_gid(); - if (r < 0) - return log_error_errno(r, "Failed to become root: %m"); - - *_home = NULL; - return 0; - } - - /* First, get user credentials */ - fd = spawn_getent("passwd", arg_user, &pid); - if (fd < 0) - return fd; - - f = fdopen(fd, "r"); - if (!f) - return log_oom(); - fd = -1; - - if (!fgets(line, sizeof(line), f)) { - - if (!ferror(f)) { - log_error("Failed to resolve user %s.", arg_user); - return -ESRCH; - } - - log_error_errno(errno, "Failed to read from getent: %m"); - return -errno; - } - - truncate_nl(line); - - wait_for_terminate_and_warn("getent passwd", pid, true); - - x = strchr(line, ':'); - if (!x) { - log_error("/etc/passwd entry has invalid user field."); - return -EIO; - } - - u = strchr(x+1, ':'); - if (!u) { - log_error("/etc/passwd entry has invalid password field."); - return -EIO; - } - - u++; - g = strchr(u, ':'); - if (!g) { - log_error("/etc/passwd entry has invalid UID field."); - return -EIO; - } - - *g = 0; - g++; - x = strchr(g, ':'); - if (!x) { - log_error("/etc/passwd entry has invalid GID field."); - return -EIO; - } - - *x = 0; - h = strchr(x+1, ':'); - if (!h) { - log_error("/etc/passwd entry has invalid GECOS field."); - return -EIO; - } - - h++; - x = strchr(h, ':'); - if (!x) { - log_error("/etc/passwd entry has invalid home directory field."); - return -EIO; - } - - *x = 0; - - r = parse_uid(u, &uid); - if (r < 0) { - log_error("Failed to parse UID of user."); - return -EIO; - } - - r = parse_gid(g, &gid); - if (r < 0) { - log_error("Failed to parse GID of user."); - return -EIO; - } - - home = strdup(h); - if (!home) - return log_oom(); - - /* Second, get group memberships */ - fd = spawn_getent("initgroups", arg_user, &pid); - if (fd < 0) - return fd; - - fclose(f); - f = fdopen(fd, "r"); - if (!f) - return log_oom(); - fd = -1; - - if (!fgets(line, sizeof(line), f)) { - if (!ferror(f)) { - log_error("Failed to resolve user %s.", arg_user); - return -ESRCH; - } - - log_error_errno(errno, "Failed to read from getent: %m"); - return -errno; - } - - truncate_nl(line); - - wait_for_terminate_and_warn("getent initgroups", pid, true); - - /* Skip over the username and subsequent separator whitespace */ - x = line; - x += strcspn(x, WHITESPACE); - x += strspn(x, WHITESPACE); - - FOREACH_WORD(word, l, x, state) { - char c[l+1]; - - memcpy(c, word, l); - c[l] = 0; - - if (!GREEDY_REALLOC(uids, sz, n_uids+1)) - return log_oom(); - - r = parse_uid(c, &uids[n_uids++]); - if (r < 0) { - log_error("Failed to parse group data from getent."); - return -EIO; - } - } - - r = mkdir_parents(home, 0775); - if (r < 0) - return log_error_errno(r, "Failed to make home root directory: %m"); - - r = mkdir_safe(home, 0755, uid, gid); - if (r < 0 && r != -EEXIST) - return log_error_errno(r, "Failed to make home directory: %m"); - - (void) fchown(STDIN_FILENO, uid, gid); - (void) fchown(STDOUT_FILENO, uid, gid); - (void) fchown(STDERR_FILENO, uid, gid); - - if (setgroups(n_uids, uids) < 0) - return log_error_errno(errno, "Failed to set auxiliary groups: %m"); - - if (setresgid(gid, gid, gid) < 0) - return log_error_errno(errno, "setregid() failed: %m"); - - if (setresuid(uid, uid, uid) < 0) - return log_error_errno(errno, "setreuid() failed: %m"); - - if (_home) { - *_home = home; - home = NULL; - } - - return 0; -} - /* * Return values: * < 0 : wait_for_terminate() failed to get the state of the @@ -3910,10 +2480,9 @@ static int wait_for_container(pid_t pid, ContainerStatus *container) { switch (status.si_code) { case CLD_EXITED: - if (status.si_status == 0) { + if (status.si_status == 0) log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s exited successfully.", arg_machine); - - } else + else log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s failed with error code %i.", arg_machine, status.si_status); *container = CONTAINER_TERMINATED; @@ -3921,13 +2490,11 @@ static int wait_for_container(pid_t pid, ContainerStatus *container) { case CLD_KILLED: if (status.si_status == SIGINT) { - log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s has been shut down.", arg_machine); *container = CONTAINER_TERMINATED; return 0; } else if (status.si_status == SIGHUP) { - log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s is being rebooted.", arg_machine); *container = CONTAINER_REBOOTED; return 0; @@ -3943,16 +2510,12 @@ static int wait_for_container(pid_t pid, ContainerStatus *container) { log_error("Container %s failed due to unknown reason.", arg_machine); return -EIO; } - - return r; } -static void nop_handler(int sig) {} - static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) { pid_t pid; - pid = PTR_TO_UINT32(userdata); + pid = PTR_TO_PID(userdata); if (pid > 0) { if (kill(pid, arg_kill_signal) >= 0) { log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination."); @@ -3968,6 +2531,17 @@ static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo static int determine_names(void) { int r; + if (arg_template && !arg_directory && arg_machine) { + + /* If --template= was specified then we should not + * search for a machine, but instead create a new one + * in /var/lib/machine. */ + + arg_directory = strjoin("/var/lib/machines/", arg_machine, NULL); + if (!arg_directory) + return log_oom(); + } + if (!arg_image && !arg_directory) { if (arg_machine) { _cleanup_(image_unrefp) Image *i = NULL; @@ -3981,9 +2555,9 @@ static int determine_names(void) { } if (i->type == IMAGE_RAW) - r = set_sanitized_path(&arg_image, i->path); + r = free_and_strdup(&arg_image, i->path); else - r = set_sanitized_path(&arg_directory, i->path); + r = free_and_strdup(&arg_directory, i->path); if (r < 0) return log_error_errno(r, "Invalid image directory: %m"); @@ -4007,7 +2581,7 @@ static int determine_names(void) { if (!arg_machine) return log_oom(); - hostname_cleanup(arg_machine, false); + hostname_cleanup(arg_machine); if (!machine_name_is_valid(arg_machine)) { log_error("Failed to determine machine name automatically, please use -M."); return -EINVAL; @@ -4035,7 +2609,7 @@ static int determine_names(void) { static int determine_uid_shift(const char *directory) { int r; - if (!arg_userns) { + if (arg_userns_mode == USER_NAMESPACE_NO) { arg_uid_shift = 0; return 0; } @@ -4062,7 +2636,6 @@ static int determine_uid_shift(const char *directory) { return -EINVAL; } - log_info("Using user namespaces with base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range); return 0; } @@ -4072,15 +2645,14 @@ static int inner_child( bool secondary, int kmsg_socket, int rtnl_socket, - FDSet *fds, - int argc, - char *argv[]) { + FDSet *fds) { _cleanup_free_ char *home = NULL; - unsigned n_env = 2; + char as_uuid[37]; + unsigned n_env = 1; const char *envp[] = { "PATH=" DEFAULT_PATH_SPLIT_USR, - "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ + NULL, /* container */ NULL, /* TERM */ NULL, /* HOME */ NULL, /* USER */ @@ -4088,6 +2660,7 @@ static int inner_child( NULL, /* container_uuid */ NULL, /* LISTEN_FDS */ NULL, /* LISTEN_PID */ + NULL, /* NOTIFY_SOCKET */ NULL }; @@ -4098,7 +2671,9 @@ static int inner_child( assert(directory); assert(kmsg_socket >= 0); - if (arg_userns) { + cg_unified_flush(); + + if (arg_userns_mode != USER_NAMESPACE_NO) { /* Tell the parent, that it now can write the UID map. */ (void) barrier_place(barrier); /* #1 */ @@ -4109,7 +2684,18 @@ static int inner_child( } } - r = mount_all(NULL, true); + r = mount_all(NULL, + arg_userns_mode != USER_NAMESPACE_NO, + true, + arg_private_network, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context); + + if (r < 0) + return r; + + r = mount_sysfs(NULL); if (r < 0) return r; @@ -4120,9 +2706,25 @@ static int inner_child( return -ESRCH; } - r = mount_systemd_cgroup_writable(""); - if (r < 0) - return r; + if (arg_use_cgns && cg_ns_supported()) { + r = unshare(CLONE_NEWCGROUP); + if (r < 0) + return log_error_errno(errno, "Failed to unshare cgroup namespace"); + r = mount_cgroups( + "", + arg_unified_cgroup_hierarchy, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context, + true); + if (r < 0) + return r; + } else { + r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy); + if (r < 0) + return r; + } r = reset_uid_gid(); if (r < 0) @@ -4145,13 +2747,16 @@ static int inner_child( if (arg_private_network) loopback_setup(); - r = send_rtnl(rtnl_socket); - if (r < 0) - return r; - rtnl_socket = safe_close(rtnl_socket); + if (arg_expose_ports) { + r = expose_port_send_rtnl(rtnl_socket); + if (r < 0) + return r; + rtnl_socket = safe_close(rtnl_socket); + } - if (drop_capabilities() < 0) - return log_error_errno(errno, "drop_capabilities() failed: %m"); + r = drop_capabilities(); + if (r < 0) + return log_error_errno(r, "drop_capabilities() failed: %m"); setup_hostname(); @@ -4165,29 +2770,30 @@ static int inner_child( #ifdef HAVE_SELINUX if (arg_selinux_context) - if (setexeccon((security_context_t) arg_selinux_context) < 0) + if (setexeccon(arg_selinux_context) < 0) return log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context); #endif - r = change_uid_gid(&home); + r = change_uid_gid(arg_user, &home); if (r < 0) return r; + /* LXC sets container=lxc, so follow the scheme here */ + envp[n_env++] = strjoina("container=", arg_container_service_name); + envp[n_env] = strv_find_prefix(environ, "TERM="); if (envp[n_env]) - n_env ++; + n_env++; if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) return log_oom(); - if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) { - char as_uuid[37]; + assert(!sd_id128_is_null(arg_uuid)); - if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) - return log_oom(); - } + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_to_uuid_string(arg_uuid, as_uuid)) < 0) + return log_oom(); if (fdset_size(fds) > 0) { r = fdset_cloexec(fds, false); @@ -4198,6 +2804,8 @@ static int inner_child( (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) return log_oom(); } + if (asprintf((char **)(envp + n_env++), "NOTIFY_SOCKET=%s", NSPAWN_NOTIFY_SOCKET_PATH) < 0) + return log_oom(); env_use = strv_env_merge(2, envp, arg_setenv); if (!env_use) @@ -4211,6 +2819,16 @@ static int inner_child( return -ESRCH; } + if (arg_chdir) + if (chdir(arg_chdir) < 0) + return log_error_errno(errno, "Failed to change to specified working directory %s: %m", arg_chdir); + + if (arg_start_mode == START_PID2) { + r = stub_pid1(); + if (r < 0) + return r; + } + /* Now, explicitly close the log, so that we * then can close all remaining fds. Closing * the log explicitly first has the benefit @@ -4222,15 +2840,16 @@ static int inner_child( log_close(); (void) fdset_close_others(fds); - if (arg_boot) { + if (arg_start_mode == START_BOOT) { char **a; size_t m; /* Automatically search for the init system */ - m = 1 + argc - optind; - a = newa(char*, m + 1); - memcpy(a + 1, argv + optind, m * sizeof(char*)); + m = strv_length(arg_parameters); + a = newa(char*, m + 2); + memcpy_safe(a + 1, arg_parameters, m * sizeof(char*)); + a[1 + m] = NULL; a[0] = (char*) "/usr/lib/systemd/systemd"; execve(a[0], a, env_use); @@ -4240,16 +2859,51 @@ static int inner_child( a[0] = (char*) "/sbin/init"; execve(a[0], a, env_use); - } else if (argc > optind) - execvpe(argv[optind], argv + optind, env_use); + } else if (!strv_isempty(arg_parameters)) + execvpe(arg_parameters[0], arg_parameters, env_use); else { - chdir(home ? home : "/root"); + if (!arg_chdir) + /* If we cannot change the directory, we'll end up in /, that is expected. */ + (void) chdir(home ?: "/root"); + execle("/bin/bash", "-bash", NULL, env_use); execle("/bin/sh", "-sh", NULL, env_use); } + r = -errno; (void) log_open(); - return log_error_errno(errno, "execv() failed: %m"); + return log_error_errno(r, "execv() failed: %m"); +} + +static int setup_sd_notify_child(void) { + static const int one = 1; + int fd = -1; + union sockaddr_union sa = { + .sa.sa_family = AF_UNIX, + }; + int r; + + 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"); + + (void) mkdir_parents(NSPAWN_NOTIFY_SOCKET_PATH, 0755); + (void) unlink(NSPAWN_NOTIFY_SOCKET_PATH); + + strncpy(sa.un.sun_path, NSPAWN_NOTIFY_SOCKET_PATH, sizeof(sa.un.sun_path)-1); + r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)); + if (r < 0) { + safe_close(fd); + return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path); + } + + r = setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one)); + if (r < 0) { + safe_close(fd); + return log_error_errno(errno, "SO_PASSCRED failed: %m"); + } + + return fd; } static int outer_child( @@ -4259,26 +2913,32 @@ static int outer_child( const char *root_device, bool root_device_rw, const char *home_device, bool home_device_rw, const char *srv_device, bool srv_device_rw, + const char *esp_device, bool interactive, bool secondary, int pid_socket, + int uuid_socket, + int notify_socket, int kmsg_socket, int rtnl_socket, int uid_shift_socket, - FDSet *fds, - int argc, - char *argv[]) { + FDSet *fds) { pid_t pid; ssize_t l; int r; + _cleanup_close_ int fd = -1; assert(barrier); assert(directory); assert(console); assert(pid_socket >= 0); + assert(uuid_socket >= 0); + assert(notify_socket >= 0); assert(kmsg_socket >= 0); + cg_unified_flush(); + if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) return log_error_errno(errno, "PR_SET_PDEATHSIG failed: %m"); @@ -4309,13 +2969,15 @@ static int outer_child( /* Mark everything as slave, so that we still * receive mounts from the real root, but don't * propagate mounts to the real root. */ - if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) - return log_error_errno(errno, "MS_SLAVE|MS_REC failed: %m"); + r = mount_verbose(LOG_ERR, NULL, "/", NULL, MS_SLAVE|MS_REC, NULL); + if (r < 0) + return r; r = mount_devices(directory, root_device, root_device_rw, home_device, home_device_rw, - srv_device, srv_device_rw); + srv_device, srv_device_rw, + esp_device); if (r < 0) return r; @@ -4323,7 +2985,12 @@ static int outer_child( if (r < 0) return r; - if (arg_userns) { + r = detect_unified_cgroup_hierarchy(directory); + if (r < 0) + return r; + + if (arg_userns_mode != USER_NAMESPACE_NO) { + /* Let the parent know which UID shift we read from the image */ l = send(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), MSG_NOSIGNAL); if (l < 0) return log_error_errno(errno, "Failed to send UID shift: %m"); @@ -4331,17 +2998,60 @@ static int outer_child( log_error("Short write while sending UID shift."); return -EIO; } + + if (arg_userns_mode == USER_NAMESPACE_PICK) { + /* When we are supposed to pick the UID shift, the parent will check now whether the UID shift + * we just read from the image is available. If yes, it will send the UID shift back to us, if + * not it will pick a different one, and send it back to us. */ + + l = recv(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), 0); + if (l < 0) + return log_error_errno(errno, "Failed to recv UID shift: %m"); + if (l != sizeof(arg_uid_shift)) { + log_error("Short read while receiving UID shift."); + return -EIO; + } + } + + log_info("Selected user namespace base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range); } /* Turn directory into bind mount */ - if (mount(directory, directory, NULL, MS_BIND|MS_REC, NULL) < 0) - return log_error_errno(errno, "Failed to make bind mount: %m"); + r = mount_verbose(LOG_ERR, directory, directory, NULL, MS_BIND|MS_REC, NULL); + if (r < 0) + return r; + + /* Mark everything as shared so our mounts get propagated down. This is + * required to make new bind mounts available in systemd services + * inside the containter that create a new mount namespace. + * See https://github.com/systemd/systemd/issues/3860 + * Further submounts (such as /dev) done after this will inherit the + * shared propagation mode.*/ + r = mount_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL); + if (r < 0) + return r; + + r = recursive_chown(directory, arg_uid_shift, arg_uid_range); + if (r < 0) + return r; - r = setup_volatile(directory); + r = setup_volatile( + directory, + arg_volatile_mode, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_context); if (r < 0) return r; - r = setup_volatile_state(directory); + r = setup_volatile_state( + directory, + arg_volatile_mode, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_context); if (r < 0) return r; @@ -4350,21 +3060,29 @@ static int outer_child( return r; if (arg_read_only) { - r = bind_remount_recursive(directory, true); + r = bind_remount_recursive(directory, true, NULL); if (r < 0) return log_error_errno(r, "Failed to make tree read-only: %m"); } - r = mount_all(directory, false); + r = mount_all(directory, + arg_userns_mode != USER_NAMESPACE_NO, + false, + arg_private_network, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context); if (r < 0) return r; - if (copy_devnodes(directory) < 0) + r = copy_devnodes(directory); + if (r < 0) return r; dev_setup(directory, arg_uid_shift, arg_uid_shift); - if (setup_pts(directory) < 0) + r = setup_pts(directory); + if (r < 0) return r; r = setup_propagate(directory); @@ -4375,7 +3093,7 @@ static int outer_child( if (r < 0) return r; - r = setup_seccomp(); + r = setup_seccomp(arg_caps_retain); if (r < 0) return r; @@ -4387,39 +3105,63 @@ static int outer_child( if (r < 0) return r; - r = setup_journal(directory); + r = setup_machine_id(directory); if (r < 0) return r; - r = mount_custom(directory); + r = setup_journal(directory); if (r < 0) return r; - r = mount_cgroup(directory); + r = mount_custom( + directory, + arg_custom_mounts, + arg_n_custom_mounts, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context); if (r < 0) return r; + if (!arg_use_cgns || !cg_ns_supported()) { + r = mount_cgroups( + directory, + arg_unified_cgroup_hierarchy, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context, + false); + if (r < 0) + return r; + } + r = mount_move_root(directory); if (r < 0) return log_error_errno(r, "Failed to move root directory: %m"); + fd = setup_sd_notify_child(); + if (fd < 0) + return fd; + pid = raw_clone(SIGCHLD|CLONE_NEWNS| - (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS) | + arg_clone_ns_flags | (arg_private_network ? CLONE_NEWNET : 0) | - (arg_userns ? CLONE_NEWUSER : 0), - NULL); + (arg_userns_mode != USER_NAMESPACE_NO ? CLONE_NEWUSER : 0)); if (pid < 0) return log_error_errno(errno, "Failed to fork inner child: %m"); - if (pid == 0) { pid_socket = safe_close(pid_socket); + uuid_socket = safe_close(uuid_socket); + notify_socket = safe_close(notify_socket); uid_shift_socket = safe_close(uid_shift_socket); /* The inner child has all namespaces that are * requested, so that we all are owned by the user if * user namespaces are turned on. */ - r = inner_child(barrier, directory, secondary, kmsg_socket, rtnl_socket, fds, argc, argv); + r = inner_child(barrier, directory, secondary, kmsg_socket, rtnl_socket, fds); if (r < 0) _exit(EXIT_FAILURE); @@ -4434,11 +3176,82 @@ static int outer_child( return -EIO; } + l = send(uuid_socket, &arg_uuid, sizeof(arg_uuid), MSG_NOSIGNAL); + if (l < 0) + return log_error_errno(errno, "Failed to send machine ID: %m"); + if (l != sizeof(arg_uuid)) { + log_error("Short write while sending machine ID."); + return -EIO; + } + + l = send_one_fd(notify_socket, fd, 0); + if (l < 0) + return log_error_errno(errno, "Failed to send notify fd: %m"); + pid_socket = safe_close(pid_socket); + uuid_socket = safe_close(uuid_socket); + notify_socket = safe_close(notify_socket); + kmsg_socket = safe_close(kmsg_socket); + rtnl_socket = safe_close(rtnl_socket); return 0; } +static int uid_shift_pick(uid_t *shift, LockFile *ret_lock_file) { + unsigned n_tries = 100; + uid_t candidate; + int r; + + assert(shift); + assert(ret_lock_file); + assert(arg_userns_mode == USER_NAMESPACE_PICK); + assert(arg_uid_range == 0x10000U); + + candidate = *shift; + + (void) mkdir("/run/systemd/nspawn-uid", 0755); + + for (;;) { + char lock_path[strlen("/run/systemd/nspawn-uid/") + DECIMAL_STR_MAX(uid_t) + 1]; + _cleanup_release_lock_file_ LockFile lf = LOCK_FILE_INIT; + + if (--n_tries <= 0) + return -EBUSY; + + if (candidate < UID_SHIFT_PICK_MIN || candidate > UID_SHIFT_PICK_MAX) + goto next; + if ((candidate & UINT32_C(0xFFFF)) != 0) + goto next; + + xsprintf(lock_path, "/run/systemd/nspawn-uid/" UID_FMT, candidate); + r = make_lock_file(lock_path, LOCK_EX|LOCK_NB, &lf); + if (r == -EBUSY) /* Range already taken by another nspawn instance */ + goto next; + if (r < 0) + return r; + + /* Make some superficial checks whether the range is currently known in the user database */ + if (getpwuid(candidate)) + goto next; + if (getpwuid(candidate + UINT32_C(0xFFFE))) + goto next; + if (getgrgid(candidate)) + goto next; + if (getgrgid(candidate + UINT32_C(0xFFFE))) + goto next; + + *ret_lock_file = lf; + lf = (struct LockFile) LOCK_FILE_INIT; + *shift = candidate; + return 0; + + next: + random_bytes(&candidate, sizeof(candidate)); + candidate = (candidate % (UID_SHIFT_PICK_MAX - UID_SHIFT_PICK_MIN)) + UID_SHIFT_PICK_MIN; + candidate &= (uid_t) UINT32_C(0xFFFF0000); + } +} + static int setup_uid_map(pid_t pid) { char uid_map[strlen("/proc//uid_map") + DECIMAL_STR_MAX(uid_t) + 1], line[DECIMAL_STR_MAX(uid_t)*3+3+1]; int r; @@ -4460,63 +3273,791 @@ static int setup_uid_map(pid_t pid) { return 0; } -static int chown_cgroup(pid_t pid) { - _cleanup_free_ char *path = NULL, *fs = NULL; - _cleanup_close_ int fd = -1; - const char *fn; +static int nspawn_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) { + char buf[NOTIFY_BUFFER_MAX+1]; + char *p = NULL; + 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; + ssize_t n; + pid_t inner_child_pid; + _cleanup_strv_free_ char **tags = NULL; + + assert(userdata); + + inner_child_pid = PTR_TO_PID(userdata); + + if (revents != EPOLLIN) { + log_warning("Got unexpected poll event for notify fd."); + return 0; + } + + n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); + if (n < 0) { + if (errno == EAGAIN || errno == EINTR) + return 0; + + return log_warning_errno(errno, "Couldn't read notification socket: %m"); + } + cmsg_close_all(&msghdr); + + CMSG_FOREACH(cmsg, &msghdr) { + if (cmsg->cmsg_level == SOL_SOCKET && + cmsg->cmsg_type == SCM_CREDENTIALS && + cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) { + + ucred = (struct ucred*) CMSG_DATA(cmsg); + } + } + + if (!ucred || ucred->pid != inner_child_pid) { + 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."); + return 0; + } + + buf[n] = 0; + tags = strv_split(buf, "\n\r"); + if (!tags) + return log_oom(); + + if (strv_find(tags, "READY=1")) + sd_notifyf(false, "READY=1\n"); + + p = strv_find_startswith(tags, "STATUS="); + if (p) + sd_notifyf(false, "STATUS=Container running: %s", p); + + return 0; +} + +static int setup_sd_notify_parent(sd_event *event, int fd, pid_t *inner_child_pid) { int r; + sd_event_source *notify_event_source; - r = cg_pid_get_path(NULL, pid, &path); + r = sd_event_add_io(event, ¬ify_event_source, fd, EPOLLIN, nspawn_dispatch_notify_fd, inner_child_pid); if (r < 0) - return log_error_errno(r, "Failed to get container cgroup path: %m"); + return log_error_errno(r, "Failed to allocate notify event source: %m"); + + (void) sd_event_source_set_description(notify_event_source, "nspawn-notify"); + + return 0; +} + +static int load_settings(void) { + _cleanup_(settings_freep) Settings *settings = NULL; + _cleanup_fclose_ FILE *f = NULL; + _cleanup_free_ char *p = NULL; + const char *fn, *i; + int r; + + /* If all settings are masked, there's no point in looking for + * the settings file */ + if ((arg_settings_mask & _SETTINGS_MASK_ALL) == _SETTINGS_MASK_ALL) + return 0; + + fn = strjoina(arg_machine, ".nspawn"); + + /* We first look in the admin's directories in /etc and /run */ + FOREACH_STRING(i, "/etc/systemd/nspawn", "/run/systemd/nspawn") { + _cleanup_free_ char *j = NULL; + + j = strjoin(i, "/", fn, NULL); + if (!j) + return log_oom(); + + f = fopen(j, "re"); + if (f) { + p = j; + j = NULL; + + /* By default, we trust configuration from /etc and /run */ + if (arg_settings_trusted < 0) + arg_settings_trusted = true; - r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, path, NULL, &fs); + break; + } + + if (errno != ENOENT) + return log_error_errno(errno, "Failed to open %s: %m", j); + } + + if (!f) { + /* After that, let's look for a file next to the + * actual image we shall boot. */ + + if (arg_image) { + p = file_in_same_dir(arg_image, fn); + if (!p) + return log_oom(); + } else if (arg_directory) { + p = file_in_same_dir(arg_directory, fn); + if (!p) + return log_oom(); + } + + if (p) { + f = fopen(p, "re"); + if (!f && errno != ENOENT) + return log_error_errno(errno, "Failed to open %s: %m", p); + + /* By default, we do not trust configuration from /var/lib/machines */ + if (arg_settings_trusted < 0) + arg_settings_trusted = false; + } + } + + if (!f) + return 0; + + log_debug("Settings are trusted: %s", yes_no(arg_settings_trusted)); + + r = settings_load(f, p, &settings); if (r < 0) - return log_error_errno(r, "Failed to get file system path for container cgroup: %m"); + return r; - fd = open(fs, O_RDONLY|O_CLOEXEC|O_DIRECTORY); - if (fd < 0) - return log_error_errno(errno, "Failed to open %s: %m", fs); + /* Copy over bits from the settings, unless they have been + * explicitly masked by command line switches. */ + + if ((arg_settings_mask & SETTING_START_MODE) == 0 && + settings->start_mode >= 0) { + arg_start_mode = settings->start_mode; + + strv_free(arg_parameters); + arg_parameters = settings->parameters; + settings->parameters = NULL; + } + + if ((arg_settings_mask & SETTING_WORKING_DIRECTORY) == 0 && + settings->working_directory) { + free(arg_chdir); + arg_chdir = settings->working_directory; + settings->working_directory = NULL; + } + + if ((arg_settings_mask & SETTING_ENVIRONMENT) == 0 && + settings->environment) { + strv_free(arg_setenv); + arg_setenv = settings->environment; + settings->environment = NULL; + } + + if ((arg_settings_mask & SETTING_USER) == 0 && + settings->user) { + free(arg_user); + arg_user = settings->user; + settings->user = NULL; + } + + if ((arg_settings_mask & SETTING_CAPABILITY) == 0) { + uint64_t plus; + + plus = settings->capability; + if (settings_private_network(settings)) + plus |= (1ULL << CAP_NET_ADMIN); + + if (!arg_settings_trusted && plus != 0) { + if (settings->capability != 0) + log_warning("Ignoring Capability= setting, file %s is not trusted.", p); + } else + arg_caps_retain |= plus; + + arg_caps_retain &= ~settings->drop_capability; + } + + if ((arg_settings_mask & SETTING_KILL_SIGNAL) == 0 && + settings->kill_signal > 0) + arg_kill_signal = settings->kill_signal; + + if ((arg_settings_mask & SETTING_PERSONALITY) == 0 && + settings->personality != PERSONALITY_INVALID) + arg_personality = settings->personality; + + if ((arg_settings_mask & SETTING_MACHINE_ID) == 0 && + !sd_id128_is_null(settings->machine_id)) { + + if (!arg_settings_trusted) + log_warning("Ignoring MachineID= setting, file %s is not trusted.", p); + else + arg_uuid = settings->machine_id; + } + + if ((arg_settings_mask & SETTING_READ_ONLY) == 0 && + settings->read_only >= 0) + arg_read_only = settings->read_only; + + if ((arg_settings_mask & SETTING_VOLATILE_MODE) == 0 && + settings->volatile_mode != _VOLATILE_MODE_INVALID) + arg_volatile_mode = settings->volatile_mode; + + if ((arg_settings_mask & SETTING_CUSTOM_MOUNTS) == 0 && + settings->n_custom_mounts > 0) { + + if (!arg_settings_trusted) + log_warning("Ignoring TemporaryFileSystem=, Bind= and BindReadOnly= settings, file %s is not trusted.", p); + else { + custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts); + arg_custom_mounts = settings->custom_mounts; + arg_n_custom_mounts = settings->n_custom_mounts; + + settings->custom_mounts = NULL; + settings->n_custom_mounts = 0; + } + } + + if ((arg_settings_mask & SETTING_NETWORK) == 0 && + (settings->private_network >= 0 || + settings->network_veth >= 0 || + settings->network_bridge || + settings->network_zone || + settings->network_interfaces || + settings->network_macvlan || + settings->network_ipvlan || + settings->network_veth_extra)) { + + if (!arg_settings_trusted) + log_warning("Ignoring network settings, file %s is not trusted.", p); + else { + arg_network_veth = settings_network_veth(settings); + arg_private_network = settings_private_network(settings); - FOREACH_STRING(fn, ".", "tasks", "notify_on_release", "cgroup.procs", "cgroup.clone_children") - if (fchownat(fd, fn, arg_uid_shift, arg_uid_shift, 0) < 0) - log_warning_errno(errno, "Failed to chown() cgroup file %s, ignoring: %m", fn); + strv_free(arg_network_interfaces); + arg_network_interfaces = settings->network_interfaces; + settings->network_interfaces = NULL; + + strv_free(arg_network_macvlan); + arg_network_macvlan = settings->network_macvlan; + settings->network_macvlan = NULL; + + strv_free(arg_network_ipvlan); + arg_network_ipvlan = settings->network_ipvlan; + settings->network_ipvlan = NULL; + + strv_free(arg_network_veth_extra); + arg_network_veth_extra = settings->network_veth_extra; + settings->network_veth_extra = NULL; + + free(arg_network_bridge); + arg_network_bridge = settings->network_bridge; + settings->network_bridge = NULL; + + free(arg_network_zone); + arg_network_zone = settings->network_zone; + settings->network_zone = NULL; + } + } + + if ((arg_settings_mask & SETTING_EXPOSE_PORTS) == 0 && + settings->expose_ports) { + + if (!arg_settings_trusted) + log_warning("Ignoring Port= setting, file %s is not trusted.", p); + else { + expose_port_free_all(arg_expose_ports); + arg_expose_ports = settings->expose_ports; + settings->expose_ports = NULL; + } + } + + if ((arg_settings_mask & SETTING_USERNS) == 0 && + settings->userns_mode != _USER_NAMESPACE_MODE_INVALID) { + + if (!arg_settings_trusted) + log_warning("Ignoring PrivateUsers= and PrivateUsersChown= settings, file %s is not trusted.", p); + else { + arg_userns_mode = settings->userns_mode; + arg_uid_shift = settings->uid_shift; + arg_uid_range = settings->uid_range; + arg_userns_chown = settings->userns_chown; + } + } + + if ((arg_settings_mask & SETTING_NOTIFY_READY) == 0) + arg_notify_ready = settings->notify_ready; return 0; } +static int run(int master, + const char* console, + const char *root_device, bool root_device_rw, + const char *home_device, bool home_device_rw, + const char *srv_device, bool srv_device_rw, + const char *esp_device, + bool interactive, + bool secondary, + FDSet *fds, + char veth_name[IFNAMSIZ], bool *veth_created, + union in_addr_union *exposed, + pid_t *pid, int *ret) { + + static const struct sigaction sa = { + .sa_handler = nop_signal_handler, + .sa_flags = SA_NOCLDSTOP, + }; + + _cleanup_release_lock_file_ LockFile uid_shift_lock = LOCK_FILE_INIT; + _cleanup_close_ int etc_passwd_lock = -1; + _cleanup_close_pair_ int + kmsg_socket_pair[2] = { -1, -1 }, + rtnl_socket_pair[2] = { -1, -1 }, + pid_socket_pair[2] = { -1, -1 }, + uuid_socket_pair[2] = { -1, -1 }, + notify_socket_pair[2] = { -1, -1 }, + uid_shift_socket_pair[2] = { -1, -1 }; + _cleanup_close_ int notify_socket= -1; + _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; + _cleanup_(sd_event_unrefp) sd_event *event = NULL; + _cleanup_(pty_forward_freep) PTYForward *forward = NULL; + _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL; + ContainerStatus container_status = 0; + char last_char = 0; + int ifi = 0, r; + ssize_t l; + sigset_t mask_chld; + + assert_se(sigemptyset(&mask_chld) == 0); + assert_se(sigaddset(&mask_chld, SIGCHLD) == 0); + + if (arg_userns_mode == USER_NAMESPACE_PICK) { + /* When we shall pick the UID/GID range, let's first lock /etc/passwd, so that we can safely + * check with getpwuid() if the specific user already exists. Note that /etc might be + * read-only, in which case this will fail with EROFS. But that's really OK, as in that case we + * can be reasonably sure that no users are going to be added. Note that getpwuid() checks are + * really just an extra safety net. We kinda assume that the UID range we allocate from is + * really ours. */ + + etc_passwd_lock = take_etc_passwd_lock(NULL); + if (etc_passwd_lock < 0 && etc_passwd_lock != -EROFS) + return log_error_errno(etc_passwd_lock, "Failed to take /etc/passwd lock: %m"); + } + + r = barrier_create(&barrier); + if (r < 0) + return log_error_errno(r, "Cannot initialize IPC barrier: %m"); + + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) + return log_error_errno(errno, "Failed to create kmsg socket pair: %m"); + + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) + return log_error_errno(errno, "Failed to create rtnl socket pair: %m"); + + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, pid_socket_pair) < 0) + return log_error_errno(errno, "Failed to create pid socket pair: %m"); + + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uuid_socket_pair) < 0) + return log_error_errno(errno, "Failed to create id socket pair: %m"); + + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, notify_socket_pair) < 0) + return log_error_errno(errno, "Failed to create notify socket pair: %m"); + + if (arg_userns_mode != USER_NAMESPACE_NO) + if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0) + return log_error_errno(errno, "Failed to create uid shift socket pair: %m"); + + /* Child can be killed before execv(), so handle SIGCHLD in order to interrupt + * parent's blocking calls and give it a chance to call wait() and terminate. */ + r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL); + if (r < 0) + return log_error_errno(errno, "Failed to change the signal mask: %m"); + + r = sigaction(SIGCHLD, &sa, NULL); + if (r < 0) + return log_error_errno(errno, "Failed to install SIGCHLD handler: %m"); + + *pid = raw_clone(SIGCHLD|CLONE_NEWNS); + if (*pid < 0) + return log_error_errno(errno, "clone() failed%s: %m", + errno == EINVAL ? + ", do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in)" : ""); + + if (*pid == 0) { + /* The outer child only has a file system namespace. */ + barrier_set_role(&barrier, BARRIER_CHILD); + + master = safe_close(master); + + kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); + rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); + pid_socket_pair[0] = safe_close(pid_socket_pair[0]); + uuid_socket_pair[0] = safe_close(uuid_socket_pair[0]); + notify_socket_pair[0] = safe_close(notify_socket_pair[0]); + uid_shift_socket_pair[0] = safe_close(uid_shift_socket_pair[0]); + + (void) reset_all_signal_handlers(); + (void) reset_signal_mask(); + + r = outer_child(&barrier, + arg_directory, + console, + root_device, root_device_rw, + home_device, home_device_rw, + srv_device, srv_device_rw, + esp_device, + interactive, + secondary, + pid_socket_pair[1], + uuid_socket_pair[1], + notify_socket_pair[1], + kmsg_socket_pair[1], + rtnl_socket_pair[1], + uid_shift_socket_pair[1], + fds); + if (r < 0) + _exit(EXIT_FAILURE); + + _exit(EXIT_SUCCESS); + } + + barrier_set_role(&barrier, BARRIER_PARENT); + + fds = fdset_free(fds); + + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); + rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]); + pid_socket_pair[1] = safe_close(pid_socket_pair[1]); + uuid_socket_pair[1] = safe_close(uuid_socket_pair[1]); + notify_socket_pair[1] = safe_close(notify_socket_pair[1]); + uid_shift_socket_pair[1] = safe_close(uid_shift_socket_pair[1]); + + if (arg_userns_mode != USER_NAMESPACE_NO) { + /* The child just let us know the UID shift it might have read from the image. */ + l = recv(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, 0); + if (l < 0) + return log_error_errno(errno, "Failed to read UID shift: %m"); + + if (l != sizeof arg_uid_shift) { + log_error("Short read while reading UID shift."); + return -EIO; + } + + if (arg_userns_mode == USER_NAMESPACE_PICK) { + /* If we are supposed to pick the UID shift, let's try to use the shift read from the + * image, but if that's already in use, pick a new one, and report back to the child, + * which one we now picked. */ + + r = uid_shift_pick(&arg_uid_shift, &uid_shift_lock); + if (r < 0) + return log_error_errno(r, "Failed to pick suitable UID/GID range: %m"); + + l = send(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, MSG_NOSIGNAL); + if (l < 0) + return log_error_errno(errno, "Failed to send UID shift: %m"); + if (l != sizeof arg_uid_shift) { + log_error("Short write while writing UID shift."); + return -EIO; + } + } + } + + /* Wait for the outer child. */ + r = wait_for_terminate_and_warn("namespace helper", *pid, NULL); + if (r != 0) + return r < 0 ? r : -EIO; + + /* And now retrieve the PID of the inner child. */ + l = recv(pid_socket_pair[0], pid, sizeof *pid, 0); + if (l < 0) + return log_error_errno(errno, "Failed to read inner child PID: %m"); + if (l != sizeof *pid) { + log_error("Short read while reading inner child PID."); + return -EIO; + } + + /* We also retrieve container UUID in case it was generated by outer child */ + l = recv(uuid_socket_pair[0], &arg_uuid, sizeof arg_uuid, 0); + if (l < 0) + return log_error_errno(errno, "Failed to read container machine ID: %m"); + if (l != sizeof(arg_uuid)) { + log_error("Short read while reading container machined ID."); + return -EIO; + } + + /* We also retrieve the socket used for notifications generated by outer child */ + notify_socket = receive_one_fd(notify_socket_pair[0], 0); + if (notify_socket < 0) + return log_error_errno(notify_socket, + "Failed to receive notification socket from the outer child: %m"); + + log_debug("Init process invoked as PID "PID_FMT, *pid); + + if (arg_userns_mode != USER_NAMESPACE_NO) { + if (!barrier_place_and_sync(&barrier)) { /* #1 */ + log_error("Child died too early."); + return -ESRCH; + } + + r = setup_uid_map(*pid); + if (r < 0) + return r; + + (void) barrier_place(&barrier); /* #2 */ + } + + if (arg_private_network) { + + r = move_network_interfaces(*pid, arg_network_interfaces); + if (r < 0) + return r; + + if (arg_network_veth) { + r = setup_veth(arg_machine, *pid, veth_name, + arg_network_bridge || arg_network_zone); + if (r < 0) + return r; + else if (r > 0) + ifi = r; + + if (arg_network_bridge) { + /* Add the interface to a bridge */ + r = setup_bridge(veth_name, arg_network_bridge, false); + if (r < 0) + return r; + if (r > 0) + ifi = r; + } else if (arg_network_zone) { + /* Add the interface to a bridge, possibly creating it */ + r = setup_bridge(veth_name, arg_network_zone, true); + if (r < 0) + return r; + if (r > 0) + ifi = r; + } + } + + r = setup_veth_extra(arg_machine, *pid, arg_network_veth_extra); + if (r < 0) + return r; + + /* We created the primary and extra veth links now; let's remember this, so that we know to + remove them later on. Note that we don't bother with removing veth links that were created + here when their setup failed half-way, because in that case the kernel should be able to + remove them on its own, since they cannot be referenced by anything yet. */ + *veth_created = true; + + r = setup_macvlan(arg_machine, *pid, arg_network_macvlan); + if (r < 0) + return r; + + r = setup_ipvlan(arg_machine, *pid, arg_network_ipvlan); + if (r < 0) + return r; + } + + if (arg_register) { + r = register_machine( + arg_machine, + *pid, + arg_directory, + arg_uuid, + ifi, + arg_slice, + arg_custom_mounts, arg_n_custom_mounts, + arg_kill_signal, + arg_property, + arg_keep_unit, + arg_container_service_name); + if (r < 0) + return r; + } + + r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift); + if (r < 0) + return r; + + if (arg_keep_unit) { + r = create_subcgroup(*pid, arg_unified_cgroup_hierarchy); + if (r < 0) + return r; + } + + r = chown_cgroup(*pid, arg_uid_shift); + if (r < 0) + return r; + + /* Notify the child that the parent is ready with all + * its setup (including cgroup-ification), and that + * the child can now hand over control to the code to + * run inside the container. */ + (void) barrier_place(&barrier); /* #3 */ + + /* Block SIGCHLD here, before notifying child. + * process_pty() will handle it with the other signals. */ + assert_se(sigprocmask(SIG_BLOCK, &mask_chld, NULL) >= 0); + + /* Reset signal to default */ + r = default_signals(SIGCHLD, -1); + if (r < 0) + return log_error_errno(r, "Failed to reset SIGCHLD: %m"); + + r = sd_event_new(&event); + if (r < 0) + return log_error_errno(r, "Failed to get default event source: %m"); + + r = setup_sd_notify_parent(event, notify_socket, PID_TO_PTR(*pid)); + if (r < 0) + return r; + + /* Let the child know that we are ready and wait that the child is completely ready now. */ + if (!barrier_place_and_sync(&barrier)) { /* #4 */ + log_error("Child died too early."); + return -ESRCH; + } + + /* At this point we have made use of the UID we picked, and thus nss-mymachines + * will make them appear in getpwuid(), thus we can release the /etc/passwd lock. */ + etc_passwd_lock = safe_close(etc_passwd_lock); + + sd_notifyf(false, + "STATUS=Container running.\n" + "X_NSPAWN_LEADER_PID=" PID_FMT, *pid); + if (!arg_notify_ready) + sd_notify(false, "READY=1\n"); + + if (arg_kill_signal > 0) { + /* Try to kill the init system on SIGINT or SIGTERM */ + sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, PID_TO_PTR(*pid)); + sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, PID_TO_PTR(*pid)); + } else { + /* Immediately exit */ + sd_event_add_signal(event, NULL, SIGINT, NULL, NULL); + sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL); + } + + /* simply exit on sigchld */ + sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL); + + if (arg_expose_ports) { + r = expose_port_watch_rtnl(event, rtnl_socket_pair[0], on_address_change, exposed, &rtnl); + if (r < 0) + return r; + + (void) expose_port_execute(rtnl, arg_expose_ports, exposed); + } + + rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); + + r = pty_forward_new(event, master, + PTY_FORWARD_IGNORE_VHANGUP | (interactive ? 0 : PTY_FORWARD_READ_ONLY), + &forward); + if (r < 0) + return log_error_errno(r, "Failed to create PTY forwarder: %m"); + + r = sd_event_loop(event); + if (r < 0) + return log_error_errno(r, "Failed to run event loop: %m"); + + pty_forward_get_last_char(forward, &last_char); + + forward = pty_forward_free(forward); + + if (!arg_quiet && last_char != '\n') + putc('\n', stdout); + + /* Kill if it is not dead yet anyway */ + if (arg_register && !arg_keep_unit) + terminate_machine(*pid); + + /* Normally redundant, but better safe than sorry */ + kill(*pid, SIGKILL); + + r = wait_for_container(*pid, &container_status); + *pid = 0; + + if (r < 0) + /* We failed to wait for the container, or the container exited abnormally. */ + return r; + if (r > 0 || container_status == CONTAINER_TERMINATED) { + /* r > 0 → The container exited with a non-zero status. + * As a special case, we need to replace 133 with a different value, + * because 133 is special-cased in the service file to reboot the container. + * otherwise → The container exited with zero status and a reboot was not requested. + */ + if (r == 133) + r = EXIT_FAILURE; /* replace 133 with the general failure code */ + *ret = r; + return 0; /* finito */ + } + + /* CONTAINER_REBOOTED, loop again */ + + if (arg_keep_unit) { + /* Special handling if we are running as a service: instead of simply + * restarting the machine we want to restart the entire service, so let's + * inform systemd about this with the special exit code 133. The service + * file uses RestartForceExitStatus=133 so that this results in a full + * nspawn restart. This is necessary since we might have cgroup parameters + * set we want to have flushed out. */ + *ret = 0; + return 133; + } + + expose_port_flush(arg_expose_ports, exposed); + + (void) remove_veth_links(veth_name, arg_network_veth_extra); + *veth_created = false; + return 1; /* loop again */ +} + int main(int argc, char *argv[]) { - _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *console = NULL; + _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *esp_device = NULL, *console = NULL; bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; _cleanup_close_ int master = -1, image_fd = -1; _cleanup_fdset_free_ FDSet *fds = NULL; - int r, n_fd_passed, loop_nr = -1; - char veth_name[IFNAMSIZ]; + int r, n_fd_passed, loop_nr = -1, ret = EXIT_FAILURE; + char veth_name[IFNAMSIZ] = ""; bool secondary = false, remove_subvol = false; - sigset_t mask_chld; pid_t pid = 0; - int ret = EXIT_SUCCESS; union in_addr_union exposed = {}; _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT; - bool interactive; + bool interactive, veth_created = false; log_parse_environment(); log_open(); + /* Make sure rename_process() in the stub init process can work */ + saved_argv = argv; + saved_argc = argc; + r = parse_argv(argc, argv); if (r <= 0) goto finish; - r = determine_names(); - if (r < 0) - goto finish; - if (geteuid() != 0) { log_error("Need to be root."); r = -EPERM; goto finish; } + r = determine_names(); + if (r < 0) + goto finish; + + r = load_settings(); + if (r < 0) + goto finish; + + r = verify_arguments(); + if (r < 0) + goto finish; n_fd_passed = sd_listen_fds(false); if (n_fd_passed > 0) { @@ -4565,7 +4106,7 @@ int main(int argc, char *argv[]) { goto finish; } - r = btrfs_subvol_snapshot(arg_directory, np, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE); + r = btrfs_subvol_snapshot(arg_directory, np, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA); if (r < 0) { log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory); goto finish; @@ -4585,11 +4126,11 @@ int main(int argc, char *argv[]) { } if (r < 0) { log_error_errno(r, "Failed to lock %s: %m", arg_directory); - return r; + goto finish; } if (arg_template) { - r = btrfs_subvol_snapshot(arg_template, arg_directory, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE); + r = btrfs_subvol_snapshot(arg_template, arg_directory, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA); if (r == -EEXIST) { if (!arg_quiet) log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template); @@ -4603,7 +4144,7 @@ int main(int argc, char *argv[]) { } } - if (arg_boot) { + if (arg_start_mode == START_BOOT) { if (path_is_os_tree(arg_directory) <= 0) { log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory); r = -EINVAL; @@ -4612,10 +4153,9 @@ int main(int argc, char *argv[]) { } else { const char *p; - p = strjoina(arg_directory, - argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/"); - if (access(p, F_OK) < 0) { - log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory); + p = strjoina(arg_directory, "/usr/"); + if (laccess(p, F_OK) < 0) { + log_error("Directory %s doesn't look like it has an OS tree. Refusing.", arg_directory); r = -EINVAL; goto finish; } @@ -4659,6 +4199,7 @@ int main(int argc, char *argv[]) { &root_device, &root_device_rw, &home_device, &home_device_rw, &srv_device, &srv_device_rw, + &esp_device, &secondary); if (r < 0) goto finish; @@ -4684,6 +4225,12 @@ int main(int argc, char *argv[]) { goto finish; } + if (arg_selinux_apifs_context) { + r = mac_selinux_apply(console, arg_selinux_apifs_context); + if (r < 0) + goto finish; + } + if (unlockpt(master) < 0) { r = log_error_errno(errno, "Failed to unlock tty: %m"); goto finish; @@ -4695,319 +4242,25 @@ int main(int argc, char *argv[]) { assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1) >= 0); - assert_se(sigemptyset(&mask_chld) == 0); - assert_se(sigaddset(&mask_chld, SIGCHLD) == 0); - if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) { r = log_error_errno(errno, "Failed to become subreaper: %m"); goto finish; } for (;;) { - _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }, rtnl_socket_pair[2] = { -1, -1 }, pid_socket_pair[2] = { -1, -1 }, - uid_shift_socket_pair[2] = { -1, -1 }; - ContainerStatus container_status; - _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; - static const struct sigaction sa = { - .sa_handler = nop_handler, - .sa_flags = SA_NOCLDSTOP, - }; - int ifi = 0; - ssize_t l; - _cleanup_event_unref_ sd_event *event = NULL; - _cleanup_(pty_forward_freep) PTYForward *forward = NULL; - _cleanup_netlink_unref_ sd_netlink *rtnl = NULL; - char last_char = 0; - - r = barrier_create(&barrier); - if (r < 0) { - log_error_errno(r, "Cannot initialize IPC barrier: %m"); - goto finish; - } - - if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { - r = log_error_errno(errno, "Failed to create kmsg socket pair: %m"); - goto finish; - } - - if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) { - r = log_error_errno(errno, "Failed to create rtnl socket pair: %m"); - goto finish; - } - - if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, pid_socket_pair) < 0) { - r = log_error_errno(errno, "Failed to create pid socket pair: %m"); - goto finish; - } - - if (arg_userns) - if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0) { - r = log_error_errno(errno, "Failed to create uid shift socket pair: %m"); - goto finish; - } - - /* Child can be killed before execv(), so handle SIGCHLD - * in order to interrupt parent's blocking calls and - * give it a chance to call wait() and terminate. */ - r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL); - if (r < 0) { - r = log_error_errno(errno, "Failed to change the signal mask: %m"); - goto finish; - } - - r = sigaction(SIGCHLD, &sa, NULL); - if (r < 0) { - r = log_error_errno(errno, "Failed to install SIGCHLD handler: %m"); - goto finish; - } - - pid = raw_clone(SIGCHLD|CLONE_NEWNS, NULL); - if (pid < 0) { - if (errno == EINVAL) - r = log_error_errno(errno, "clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); - else - r = log_error_errno(errno, "clone() failed: %m"); - - goto finish; - } - - if (pid == 0) { - /* The outer child only has a file system namespace. */ - barrier_set_role(&barrier, BARRIER_CHILD); - - master = safe_close(master); - - kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); - rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); - pid_socket_pair[0] = safe_close(pid_socket_pair[0]); - uid_shift_socket_pair[0] = safe_close(uid_shift_socket_pair[0]); - - (void) reset_all_signal_handlers(); - (void) reset_signal_mask(); - - r = outer_child(&barrier, - arg_directory, - console, - root_device, root_device_rw, - home_device, home_device_rw, - srv_device, srv_device_rw, - interactive, - secondary, - pid_socket_pair[1], - kmsg_socket_pair[1], - rtnl_socket_pair[1], - uid_shift_socket_pair[1], - fds, - argc, argv); - if (r < 0) - _exit(EXIT_FAILURE); - - _exit(EXIT_SUCCESS); - } - - barrier_set_role(&barrier, BARRIER_PARENT); - - fdset_free(fds); - fds = NULL; - - kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); - rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]); - pid_socket_pair[1] = safe_close(pid_socket_pair[1]); - - /* Wait for the outer child. */ - r = wait_for_terminate_and_warn("namespace helper", pid, NULL); - if (r < 0) - goto finish; - if (r != 0) { - r = -EIO; - goto finish; - } - pid = 0; - - /* And now retrieve the PID of the inner child. */ - l = recv(pid_socket_pair[0], &pid, sizeof(pid), 0); - if (l < 0) { - r = log_error_errno(errno, "Failed to read inner child PID: %m"); - goto finish; - } - if (l != sizeof(pid)) { - log_error("Short read while reading inner child PID: %m"); - r = EIO; - goto finish; - } - - log_debug("Init process invoked as PID " PID_FMT, pid); - - if (arg_userns) { - if (!barrier_place_and_sync(&barrier)) { /* #1 */ - log_error("Child died too early."); - r = -ESRCH; - goto finish; - } - - l = recv(uid_shift_socket_pair[0], &arg_uid_shift, sizeof(arg_uid_shift), 0); - if (l < 0) { - r = log_error_errno(errno, "Failed to read UID shift: %m"); - goto finish; - } - if (l != sizeof(arg_uid_shift)) { - log_error("Short read while reading UID shift: %m"); - r = EIO; - goto finish; - } - - r = setup_uid_map(pid); - if (r < 0) - goto finish; - - (void) barrier_place(&barrier); /* #2 */ - } - - r = move_network_interfaces(pid); - if (r < 0) - goto finish; - - r = setup_veth(pid, veth_name, &ifi); - if (r < 0) - goto finish; - - r = setup_bridge(veth_name, &ifi); - if (r < 0) - goto finish; - - r = setup_macvlan(pid); - if (r < 0) - goto finish; - - r = setup_ipvlan(pid); - if (r < 0) - goto finish; - - r = register_machine(pid, ifi); - if (r < 0) - goto finish; - - r = chown_cgroup(pid); - if (r < 0) - goto finish; - - /* Notify the child that the parent is ready with all - * its setup (including cgroup-ification), and that - * the child can now hand over control to the code to - * run inside the container. */ - (void) barrier_place(&barrier); /* #3 */ - - /* Block SIGCHLD here, before notifying child. - * process_pty() will handle it with the other signals. */ - assert_se(sigprocmask(SIG_BLOCK, &mask_chld, NULL) >= 0); - - /* Reset signal to default */ - r = default_signals(SIGCHLD, -1); - if (r < 0) { - log_error_errno(r, "Failed to reset SIGCHLD: %m"); - goto finish; - } - - /* Let the child know that we are ready and wait that the child is completely ready now. */ - if (!barrier_place_and_sync(&barrier)) { /* #5 */ - log_error("Client died too early."); - r = -ESRCH; - goto finish; - } - - sd_notifyf(false, - "READY=1\n" - "STATUS=Container running.\n" - "X_NSPAWN_LEADER_PID=" PID_FMT, pid); - - r = sd_event_new(&event); - if (r < 0) { - log_error_errno(r, "Failed to get default event source: %m"); - goto finish; - } - - if (arg_kill_signal > 0) { - /* Try to kill the init system on SIGINT or SIGTERM */ - sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, UINT32_TO_PTR(pid)); - sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, UINT32_TO_PTR(pid)); - } else { - /* Immediately exit */ - sd_event_add_signal(event, NULL, SIGINT, NULL, NULL); - sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL); - } - - /* simply exit on sigchld */ - sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL); - - if (arg_expose_ports) { - r = watch_rtnl(event, rtnl_socket_pair[0], &exposed, &rtnl); - if (r < 0) - goto finish; - - (void) expose_ports(rtnl, &exposed); - } - - rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); - - r = pty_forward_new(event, master, true, !interactive, &forward); - if (r < 0) { - log_error_errno(r, "Failed to create PTY forwarder: %m"); - goto finish; - } - - r = sd_event_loop(event); - if (r < 0) { - log_error_errno(r, "Failed to run event loop: %m"); - goto finish; - } - - pty_forward_get_last_char(forward, &last_char); - - forward = pty_forward_free(forward); - - if (!arg_quiet && last_char != '\n') - putc('\n', stdout); - - /* Kill if it is not dead yet anyway */ - terminate_machine(pid); - - /* Normally redundant, but better safe than sorry */ - kill(pid, SIGKILL); - - r = wait_for_container(pid, &container_status); - pid = 0; - - if (r < 0) - /* We failed to wait for the container, or the - * container exited abnormally */ - goto finish; - else if (r > 0 || container_status == CONTAINER_TERMINATED){ - /* The container exited with a non-zero - * status, or with zero status and no reboot - * was requested. */ - ret = r; + r = run(master, + console, + root_device, root_device_rw, + home_device, home_device_rw, + srv_device, srv_device_rw, + esp_device, + interactive, secondary, + fds, + veth_name, &veth_created, + &exposed, + &pid, &ret); + if (r <= 0) break; - } - - /* CONTAINER_REBOOTED, loop again */ - - if (arg_keep_unit) { - /* Special handling if we are running as a - * service: instead of simply restarting the - * machine we want to restart the entire - * service, so let's inform systemd about this - * with the special exit code 133. The service - * file uses RestartForceExitStatus=133 so - * that this results in a full nspawn - * restart. This is necessary since we might - * have cgroup parameters set we want to have - * flushed out. */ - ret = 133; - r = 0; - break; - } - - flush_ports(&exposed); } finish: @@ -5020,14 +4273,14 @@ finish: /* Try to flush whatever is still queued in the pty */ if (master >= 0) - (void) copy_bytes(master, STDOUT_FILENO, (off_t) -1, false); + (void) copy_bytes(master, STDOUT_FILENO, (uint64_t) -1, false); loop_remove(loop_nr, &image_fd); if (remove_subvol && arg_directory) { int k; - k = btrfs_subvol_remove(arg_directory, true); + k = btrfs_subvol_remove(arg_directory, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA); if (k < 0) log_warning_errno(k, "Cannot remove subvolume '%s', ignoring: %m", arg_directory); } @@ -5039,24 +4292,27 @@ finish: (void) rm_rf(p, REMOVE_ROOT); } + expose_port_flush(arg_expose_ports, &exposed); + + if (veth_created) + (void) remove_veth_links(veth_name, arg_network_veth_extra); + (void) remove_bridge(arg_network_zone); + free(arg_directory); free(arg_template); free(arg_image); free(arg_machine); free(arg_user); + free(arg_chdir); strv_free(arg_setenv); + free(arg_network_bridge); strv_free(arg_network_interfaces); strv_free(arg_network_macvlan); strv_free(arg_network_ipvlan); - custom_mount_free_all(); - - flush_ports(&exposed); - - while (arg_expose_ports) { - ExposePort *p = arg_expose_ports; - LIST_REMOVE(ports, arg_expose_ports, p); - free(p); - } + strv_free(arg_network_veth_extra); + strv_free(arg_parameters); + custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts); + expose_port_free_all(arg_expose_ports); return r < 0 ? EXIT_FAILURE : ret; } |