1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2010 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
25 #include <sys/types.h>
26 #include <sys/syscall.h>
27 #include <sys/mount.h>
33 #include <sys/prctl.h>
34 #include <sys/capability.h>
37 #include <sys/signalfd.h>
41 #include <sys/socket.h>
42 #include <linux/netlink.h>
44 #include "sd-daemon.h"
53 #include "cgroup-util.h"
55 #include "path-util.h"
56 #include "loopback-setup.h"
57 #include "dev-setup.h"
62 #include "bus-error.h"
69 typedef enum LinkJournal {
76 static char *arg_directory = NULL;
77 static char *arg_user = NULL;
78 static sd_id128_t arg_uuid = {};
79 static char *arg_machine = NULL;
80 static const char *arg_slice = NULL;
81 static bool arg_private_network = false;
82 static bool arg_read_only = false;
83 static bool arg_boot = false;
84 static LinkJournal arg_link_journal = LINK_AUTO;
85 static uint64_t arg_retain =
87 (1ULL << CAP_DAC_OVERRIDE) |
88 (1ULL << CAP_DAC_READ_SEARCH) |
89 (1ULL << CAP_FOWNER) |
90 (1ULL << CAP_FSETID) |
91 (1ULL << CAP_IPC_OWNER) |
94 (1ULL << CAP_LINUX_IMMUTABLE) |
95 (1ULL << CAP_NET_BIND_SERVICE) |
96 (1ULL << CAP_NET_BROADCAST) |
97 (1ULL << CAP_NET_RAW) |
98 (1ULL << CAP_SETGID) |
99 (1ULL << CAP_SETFCAP) |
100 (1ULL << CAP_SETPCAP) |
101 (1ULL << CAP_SETUID) |
102 (1ULL << CAP_SYS_ADMIN) |
103 (1ULL << CAP_SYS_CHROOT) |
104 (1ULL << CAP_SYS_NICE) |
105 (1ULL << CAP_SYS_PTRACE) |
106 (1ULL << CAP_SYS_TTY_CONFIG) |
107 (1ULL << CAP_SYS_RESOURCE) |
108 (1ULL << CAP_SYS_BOOT) |
109 (1ULL << CAP_AUDIT_WRITE) |
110 (1ULL << CAP_AUDIT_CONTROL);
111 static char **arg_bind = NULL;
112 static char **arg_bind_ro = NULL;
114 static int help(void) {
116 printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
117 "Spawn a minimal namespace container for debugging, testing and building.\n\n"
118 " -h --help Show this help\n"
119 " --version Print version string\n"
120 " -D --directory=NAME Root directory for the container\n"
121 " -b --boot Boot up full system (i.e. invoke init)\n"
122 " -u --user=USER Run the command under specified user or uid\n"
123 " --uuid=UUID Set a specific machine UUID for the container\n"
124 " -M --machine=NAME Set the machine name for the container\n"
125 " -S --slice=SLICE Place the container in the specified slice\n"
126 " --private-network Disable network in container\n"
127 " --read-only Mount the root directory read-only\n"
128 " --capability=CAP In addition to the default, retain specified\n"
130 " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n"
131 " -j Equivalent to --link-journal=host\n"
132 " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n"
134 " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n",
135 program_invocation_short_name);
140 static int parse_argv(int argc, char *argv[]) {
153 static const struct option options[] = {
154 { "help", no_argument, NULL, 'h' },
155 { "version", no_argument, NULL, ARG_VERSION },
156 { "directory", required_argument, NULL, 'D' },
157 { "user", required_argument, NULL, 'u' },
158 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
159 { "boot", no_argument, NULL, 'b' },
160 { "uuid", required_argument, NULL, ARG_UUID },
161 { "read-only", no_argument, NULL, ARG_READ_ONLY },
162 { "capability", required_argument, NULL, ARG_CAPABILITY },
163 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
164 { "bind", required_argument, NULL, ARG_BIND },
165 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
166 { "machine", required_argument, NULL, 'M' },
167 { "slice", required_argument, NULL, 'S' },
176 while ((c = getopt_long(argc, argv, "+hD:u:bM:jS:", options, NULL)) >= 0) {
184 puts(PACKAGE_STRING);
185 puts(SYSTEMD_FEATURES);
190 arg_directory = canonicalize_file_name(optarg);
191 if (!arg_directory) {
192 log_error("Failed to canonicalize root directory.");
200 arg_user = strdup(optarg);
206 case ARG_PRIVATE_NETWORK:
207 arg_private_network = true;
215 r = sd_id128_from_string(optarg, &arg_uuid);
217 log_error("Invalid UUID: %s", optarg);
223 arg_slice = strdup(optarg);
230 if (!hostname_is_valid(optarg)) {
231 log_error("Invalid machine name: %s", optarg);
236 arg_machine = strdup(optarg);
243 arg_read_only = true;
246 case ARG_CAPABILITY: {
250 FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
254 t = strndup(word, length);
258 if (cap_from_name(t, &cap) < 0) {
259 log_error("Failed to parse capability %s.", t);
265 arg_retain |= 1ULL << (uint64_t) cap;
272 arg_link_journal = LINK_GUEST;
275 case ARG_LINK_JOURNAL:
276 if (streq(optarg, "auto"))
277 arg_link_journal = LINK_AUTO;
278 else if (streq(optarg, "no"))
279 arg_link_journal = LINK_NO;
280 else if (streq(optarg, "guest"))
281 arg_link_journal = LINK_GUEST;
282 else if (streq(optarg, "host"))
283 arg_link_journal = LINK_HOST;
285 log_error("Failed to parse link journal mode %s", optarg);
293 _cleanup_free_ char *a = NULL, *b = NULL;
297 x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
299 e = strchr(optarg, ':');
301 a = strndup(optarg, e - optarg);
311 if (!path_is_absolute(a) || !path_is_absolute(b)) {
312 log_error("Invalid bind mount specification: %s", optarg);
316 r = strv_extend(x, a);
320 r = strv_extend(x, b);
331 assert_not_reached("Unhandled option");
338 static int mount_all(const char *dest) {
340 typedef struct MountPoint {
349 static const MountPoint mount_table[] = {
350 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
351 { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */
352 { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
353 { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
354 { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
355 { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true },
356 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
357 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
359 { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
360 { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
367 for (k = 0; k < ELEMENTSOF(mount_table); k++) {
368 _cleanup_free_ char *where = NULL;
371 where = strjoin(dest, "/", mount_table[k].where, NULL);
375 t = path_is_mount_point(where, true);
377 log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
385 /* Skip this entry if it is not a remount. */
386 if (mount_table[k].what && t > 0)
389 mkdir_p(where, 0755);
391 if (mount(mount_table[k].what,
394 mount_table[k].flags,
395 mount_table[k].options) < 0 &&
396 mount_table[k].fatal) {
398 log_error("mount(%s) failed: %m", where);
408 static int mount_binds(const char *dest, char **l, unsigned long flags) {
411 STRV_FOREACH_PAIR(x, y, l) {
412 _cleanup_free_ char *where = NULL;
413 struct stat source_st, dest_st;
415 if (stat(*x, &source_st) < 0) {
416 log_error("failed to stat %s: %m", *x);
420 where = strjoin(dest, "/", *y, NULL);
424 if (stat(where, &dest_st) == 0) {
425 if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) {
426 log_error("The file types of %s and %s do not match. Refusing bind mount",
431 /* Create the mount point, but be conservative -- refuse to create block
432 * and char devices. */
433 if (S_ISDIR(source_st.st_mode))
434 mkdir_p_label(where, 0755);
435 else if (S_ISFIFO(source_st.st_mode))
437 else if (S_ISSOCK(source_st.st_mode))
438 mknod(where, 0644 | S_IFSOCK, 0);
439 else if (S_ISREG(source_st.st_mode))
442 log_error("Refusing to create mountpoint for file: %s", *x);
447 if (mount(*x, where, "bind", MS_BIND, NULL) < 0) {
448 log_error("mount(%s) failed: %m", where);
452 if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) {
453 log_error("mount(%s) failed: %m", where);
461 static int setup_timezone(const char *dest) {
462 _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
468 /* Fix the timezone, if possible */
469 r = readlink_malloc("/etc/localtime", &p);
471 log_warning("/etc/localtime is not a symlink, not updating container timezone.");
475 z = path_startswith(p, "../usr/share/zoneinfo/");
477 z = path_startswith(p, "/usr/share/zoneinfo/");
479 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
483 where = strappend(dest, "/etc/localtime");
487 r = readlink_malloc(where, &q);
489 y = path_startswith(q, "../usr/share/zoneinfo/");
491 y = path_startswith(q, "/usr/share/zoneinfo/");
494 /* Already pointing to the right place? Then do nothing .. */
495 if (y && streq(y, z))
499 check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
503 if (access(check, F_OK) < 0) {
504 log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
508 what = strappend("../usr/share/zoneinfo/", z);
513 if (symlink(what, where) < 0) {
514 log_error("Failed to correct timezone of container: %m");
521 static int setup_resolv_conf(const char *dest) {
522 char _cleanup_free_ *where = NULL;
526 if (arg_private_network)
529 /* Fix resolv.conf, if possible */
530 where = strappend(dest, "/etc/resolv.conf");
534 /* We don't really care for the results of this really. If it
535 * fails, it fails, but meh... */
536 copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW);
541 static int setup_boot_id(const char *dest) {
542 _cleanup_free_ char *from = NULL, *to = NULL;
549 /* Generate a new randomized boot ID, so that each boot-up of
550 * the container gets a new one */
552 from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
553 to = strappend(dest, "/proc/sys/kernel/random/boot_id");
557 r = sd_id128_randomize(&rnd);
559 log_error("Failed to generate random boot id: %s", strerror(-r));
563 snprintf(as_uuid, sizeof(as_uuid),
564 "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
565 SD_ID128_FORMAT_VAL(rnd));
566 char_array_0(as_uuid);
568 r = write_string_file(from, as_uuid);
570 log_error("Failed to write boot id: %s", strerror(-r));
574 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
575 log_error("Failed to bind mount boot id: %m");
577 } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL))
578 log_warning("Failed to make boot id read-only: %m");
584 static int copy_devnodes(const char *dest) {
586 static const char devnodes[] =
596 _cleanup_umask_ mode_t u;
602 NULSTR_FOREACH(d, devnodes) {
604 _cleanup_free_ char *from = NULL, *to = NULL;
606 asprintf(&from, "/dev/%s", d);
607 asprintf(&to, "%s/dev/%s", dest, d);
618 if (stat(from, &st) < 0) {
620 if (errno != ENOENT) {
621 log_error("Failed to stat %s: %m", from);
626 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
628 log_error("%s is not a char or block device, cannot copy", from);
632 } else if (mknod(to, st.st_mode, st.st_rdev) < 0) {
634 log_error("mknod(%s) failed: %m", dest);
643 static int setup_ptmx(const char *dest) {
644 _cleanup_free_ char *p = NULL;
646 p = strappend(dest, "/dev/ptmx");
650 if (symlink("pts/ptmx", p) < 0) {
651 log_error("Failed to create /dev/ptmx symlink: %m");
658 static int setup_dev_console(const char *dest, const char *console) {
660 _cleanup_free_ char *to = NULL;
662 _cleanup_umask_ mode_t u;
669 if (stat(console, &st) < 0) {
670 log_error("Failed to stat %s: %m", console);
673 } else if (!S_ISCHR(st.st_mode)) {
674 log_error("/dev/console is not a char device");
678 r = chmod_and_chown(console, 0600, 0, 0);
680 log_error("Failed to correct access mode for TTY: %s", strerror(-r));
684 if (asprintf(&to, "%s/dev/console", dest) < 0)
687 /* We need to bind mount the right tty to /dev/console since
688 * ptys can only exist on pts file systems. To have something
689 * to bind mount things on we create a device node first, that
690 * has the right major/minor (note that the major minor
691 * doesn't actually matter here, since we mount it over
694 if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
695 log_error("mknod() for /dev/console failed: %m");
699 if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
700 log_error("Bind mount for /dev/console failed: %m");
707 static int setup_kmsg(const char *dest, int kmsg_socket) {
708 _cleanup_free_ char *from = NULL, *to = NULL;
710 _cleanup_umask_ mode_t u;
712 struct cmsghdr cmsghdr;
713 uint8_t buf[CMSG_SPACE(sizeof(int))];
716 .msg_control = &control,
717 .msg_controllen = sizeof(control),
719 struct cmsghdr *cmsg;
722 assert(kmsg_socket >= 0);
726 /* We create the kmsg FIFO as /dev/kmsg, but immediately
727 * delete it after bind mounting it to /proc/kmsg. While FIFOs
728 * on the reading side behave very similar to /proc/kmsg,
729 * their writing side behaves differently from /dev/kmsg in
730 * that writing blocks when nothing is reading. In order to
731 * avoid any problems with containers deadlocking due to this
732 * we simply make /dev/kmsg unavailable to the container. */
733 if (asprintf(&from, "%s/dev/kmsg", dest) < 0 ||
734 asprintf(&to, "%s/proc/kmsg", dest) < 0)
737 if (mkfifo(from, 0600) < 0) {
738 log_error("mkfifo() for /dev/kmsg failed: %m");
742 r = chmod_and_chown(from, 0600, 0, 0);
744 log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
748 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
749 log_error("Bind mount for /proc/kmsg failed: %m");
753 fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
755 log_error("Failed to open fifo: %m");
759 cmsg = CMSG_FIRSTHDR(&mh);
760 cmsg->cmsg_level = SOL_SOCKET;
761 cmsg->cmsg_type = SCM_RIGHTS;
762 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
763 memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
765 mh.msg_controllen = cmsg->cmsg_len;
767 /* Store away the fd in the socket, so that it stays open as
768 * long as we run the child */
769 k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL);
770 close_nointr_nofail(fd);
773 log_error("Failed to send FIFO fd: %m");
777 /* And now make the FIFO unavailable as /dev/kmsg... */
782 static int setup_hostname(void) {
784 if (sethostname(arg_machine, strlen(arg_machine)) < 0)
790 static int setup_journal(const char *directory) {
791 sd_id128_t machine_id;
792 _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL;
796 if (arg_link_journal == LINK_NO)
799 p = strappend(directory, "/etc/machine-id");
803 r = read_one_line_file(p, &b);
804 if (r == -ENOENT && arg_link_journal == LINK_AUTO)
807 log_error("Failed to read machine ID from %s: %s", p, strerror(-r));
812 if (isempty(id) && arg_link_journal == LINK_AUTO)
815 /* Verify validity */
816 r = sd_id128_from_string(id, &machine_id);
818 log_error("Failed to parse machine ID from %s: %s", p, strerror(-r));
823 p = strappend("/var/log/journal/", id);
824 q = strjoin(directory, "/var/log/journal/", id, NULL);
828 if (path_is_mount_point(p, false) > 0) {
829 if (arg_link_journal != LINK_AUTO) {
830 log_error("%s: already a mount point, refusing to use for journal", p);
837 if (path_is_mount_point(q, false) > 0) {
838 if (arg_link_journal != LINK_AUTO) {
839 log_error("%s: already a mount point, refusing to use for journal", q);
846 r = readlink_and_make_absolute(p, &d);
848 if ((arg_link_journal == LINK_GUEST ||
849 arg_link_journal == LINK_AUTO) &&
852 r = mkdir_p(q, 0755);
854 log_warning("failed to create directory %s: %m", q);
859 log_error("Failed to remove symlink %s: %m", p);
862 } else if (r == -EINVAL) {
864 if (arg_link_journal == LINK_GUEST &&
867 if (errno == ENOTDIR) {
868 log_error("%s already exists and is neither a symlink nor a directory", p);
871 log_error("Failed to remove %s: %m", p);
875 } else if (r != -ENOENT) {
876 log_error("readlink(%s) failed: %m", p);
880 if (arg_link_journal == LINK_GUEST) {
882 if (symlink(q, p) < 0) {
883 log_error("Failed to symlink %s to %s: %m", q, p);
887 r = mkdir_p(q, 0755);
889 log_warning("failed to create directory %s: %m", q);
893 if (arg_link_journal == LINK_HOST) {
894 r = mkdir_p(p, 0755);
896 log_error("Failed to create %s: %m", p);
900 } else if (access(p, F_OK) < 0)
903 if (dir_is_empty(q) == 0) {
904 log_error("%s not empty.", q);
908 r = mkdir_p(q, 0755);
910 log_error("Failed to create %s: %m", q);
914 if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
915 log_error("Failed to bind mount journal from host into guest: %m");
922 static int drop_capabilities(void) {
923 return capability_bounding_set_drop(~arg_retain, false);
926 static int register_machine(void) {
927 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
928 _cleanup_bus_unref_ sd_bus *bus = NULL;
931 r = sd_bus_open_system(&bus);
933 log_error("Failed to open system bus: %s", strerror(-r));
937 r = sd_bus_call_method(
939 "org.freedesktop.machine1",
940 "/org/freedesktop/machine1",
941 "org.freedesktop.machine1.Manager",
947 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
951 strempty(arg_directory),
952 !isempty(arg_slice), "Slice", "s", arg_slice);
954 log_error("Failed to register machine: %s", bus_error_message(&error, r));
961 static int terminate_machine(pid_t pid) {
962 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
963 _cleanup_bus_message_unref_ sd_bus_message *reply = NULL;
964 _cleanup_bus_unref_ sd_bus *bus = NULL;
968 r = sd_bus_open_system(&bus);
970 log_error("Failed to open system bus: %s", strerror(-r));
974 r = sd_bus_call_method(
976 "org.freedesktop.machine1",
977 "/org/freedesktop/machine1",
978 "org.freedesktop.machine1.Manager",
985 /* Note that the machine might already have been
986 * cleaned up automatically, hence don't consider it a
987 * failure if we cannot get the machine object. */
988 log_debug("Failed to get machine: %s", bus_error_message(&error, r));
992 r = sd_bus_message_read(reply, "o", &path);
994 return bus_log_parse_error(r);
996 r = sd_bus_call_method(
998 "org.freedesktop.machine1",
1000 "org.freedesktop.machine1.Machine",
1006 log_debug("Failed to terminate machine: %s", bus_error_message(&error, r));
1013 static bool audit_enabled(void) {
1016 fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT);
1018 close_nointr_nofail(fd);
1024 int main(int argc, char *argv[]) {
1026 int r = EXIT_FAILURE, k;
1027 _cleanup_close_ int master = -1;
1029 const char *console = NULL;
1031 _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 };
1032 _cleanup_fdset_free_ FDSet *fds = NULL;
1034 log_parse_environment();
1037 k = parse_argv(argc, argv);
1045 if (arg_directory) {
1048 p = path_make_absolute_cwd(arg_directory);
1049 free(arg_directory);
1052 arg_directory = get_current_dir_name();
1054 if (!arg_directory) {
1055 log_error("Failed to determine path, please use -D.");
1059 path_kill_slashes(arg_directory);
1062 arg_machine = strdup(path_get_file_name(arg_directory));
1068 hostname_cleanup(arg_machine, false);
1069 if (isempty(arg_machine)) {
1070 log_error("Failed to determine machine name automatically, please use -M.");
1075 if (geteuid() != 0) {
1076 log_error("Need to be root.");
1080 if (sd_booted() <= 0) {
1081 log_error("Not running on a systemd system.");
1085 if (arg_boot && audit_enabled()) {
1086 log_warning("The kernel auditing subsystem is known to be incompatible with containers.\n"
1087 "Please make sure to turn off auditing with 'audit=0' on the kernel command\n"
1088 "line before using systemd-nspawn. Sleeping for 5s...\n");
1092 if (path_equal(arg_directory, "/")) {
1093 log_error("Spawning container on root directory not supported.");
1097 if (path_is_os_tree(arg_directory) <= 0) {
1098 log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory);
1103 n_fd_passed = sd_listen_fds(false);
1104 if (n_fd_passed > 0) {
1105 k = fdset_new_listen_fds(&fds, false);
1107 log_error("Failed to collect file descriptors: %s", strerror(-k));
1111 fdset_close_others(fds);
1114 master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
1116 log_error("Failed to acquire pseudo tty: %m");
1120 console = ptsname(master);
1122 log_error("Failed to determine tty name: %m");
1126 log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_directory);
1128 if (unlockpt(master) < 0) {
1129 log_error("Failed to unlock tty: %m");
1133 if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
1134 log_error("Failed to create kmsg socket pair.");
1138 sd_notify(0, "READY=1");
1140 assert_se(sigemptyset(&mask) == 0);
1141 sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
1142 assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
1147 pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
1149 if (errno == EINVAL)
1150 log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
1152 log_error("clone() failed: %m");
1159 const char *home = NULL;
1160 uid_t uid = (uid_t) -1;
1161 gid_t gid = (gid_t) -1;
1163 const char *envp[] = {
1164 "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
1165 "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
1170 NULL, /* container_uuid */
1171 NULL, /* LISTEN_FDS */
1172 NULL, /* LISTEN_PID */
1176 envp[n_env] = strv_find_prefix(environ, "TERM=");
1180 close_nointr_nofail(master);
1183 close_nointr(STDIN_FILENO);
1184 close_nointr(STDOUT_FILENO);
1185 close_nointr(STDERR_FILENO);
1187 close_nointr_nofail(kmsg_socket_pair[0]);
1188 kmsg_socket_pair[0] = -1;
1190 reset_all_signal_handlers();
1192 assert_se(sigemptyset(&mask) == 0);
1193 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
1195 k = open_terminal(console, O_RDWR);
1196 if (k != STDIN_FILENO) {
1198 close_nointr_nofail(k);
1202 log_error("Failed to open console: %s", strerror(-k));
1206 if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
1207 dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
1208 log_error("Failed to duplicate console: %m");
1213 log_error("setsid() failed: %m");
1217 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
1218 log_error("PR_SET_PDEATHSIG failed: %m");
1222 r = register_machine();
1226 /* Mark everything as slave, so that we still
1227 * receive mounts from the real root, but don't
1228 * propagate mounts to the real root. */
1229 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
1230 log_error("MS_SLAVE|MS_REC failed: %m");
1234 /* Turn directory into bind mount */
1235 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
1236 log_error("Failed to make bind mount.");
1241 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) {
1242 log_error("Failed to make read-only.");
1246 if (mount_all(arg_directory) < 0)
1249 if (copy_devnodes(arg_directory) < 0)
1252 if (setup_ptmx(arg_directory) < 0)
1255 dev_setup(arg_directory);
1257 if (setup_dev_console(arg_directory, console) < 0)
1260 if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
1263 close_nointr_nofail(kmsg_socket_pair[1]);
1264 kmsg_socket_pair[1] = -1;
1266 if (setup_boot_id(arg_directory) < 0)
1269 if (setup_timezone(arg_directory) < 0)
1272 if (setup_resolv_conf(arg_directory) < 0)
1275 if (setup_journal(arg_directory) < 0)
1278 if (mount_binds(arg_directory, arg_bind, 0) < 0)
1281 if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
1284 if (chdir(arg_directory) < 0) {
1285 log_error("chdir(%s) failed: %m", arg_directory);
1289 if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
1290 log_error("mount(MS_MOVE) failed: %m");
1294 if (chroot(".") < 0) {
1295 log_error("chroot() failed: %m");
1299 if (chdir("/") < 0) {
1300 log_error("chdir() failed: %m");
1308 if (drop_capabilities() < 0) {
1309 log_error("drop_capabilities() failed: %m");
1315 /* Note that this resolves user names
1316 * inside the container, and hence
1317 * accesses the NSS modules from the
1318 * container and not the host. This is
1321 if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) {
1322 log_error("get_user_creds() failed: %m");
1326 if (mkdir_parents_label(home, 0775) < 0) {
1327 log_error("mkdir_parents_label() failed: %m");
1331 if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
1332 log_error("mkdir_safe_label() failed: %m");
1336 if (initgroups((const char*)arg_user, gid) < 0) {
1337 log_error("initgroups() failed: %m");
1341 if (setresgid(gid, gid, gid) < 0) {
1342 log_error("setregid() failed: %m");
1346 if (setresuid(uid, uid, uid) < 0) {
1347 log_error("setreuid() failed: %m");
1351 /* Reset everything fully to 0, just in case */
1353 if (setgroups(0, NULL) < 0) {
1354 log_error("setgroups() failed: %m");
1358 if (setresgid(0, 0, 0) < 0) {
1359 log_error("setregid() failed: %m");
1363 if (setresuid(0, 0, 0) < 0) {
1364 log_error("setreuid() failed: %m");
1369 if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
1370 (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
1371 (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
1376 if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) {
1377 if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) {
1383 if (fdset_size(fds) > 0) {
1384 k = fdset_cloexec(fds, false);
1386 log_error("Failed to unset O_CLOEXEC for file descriptors.");
1390 if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
1391 (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) {
1403 /* Automatically search for the init system */
1405 l = 1 + argc - optind;
1406 a = newa(char*, l + 1);
1407 memcpy(a + 1, argv + optind, l * sizeof(char*));
1409 a[0] = (char*) "/usr/lib/systemd/systemd";
1410 execve(a[0], a, (char**) envp);
1412 a[0] = (char*) "/lib/systemd/systemd";
1413 execve(a[0], a, (char**) envp);
1415 a[0] = (char*) "/sbin/init";
1416 execve(a[0], a, (char**) envp);
1417 } else if (argc > optind)
1418 execvpe(argv[optind], argv + optind, (char**) envp);
1420 chdir(home ? home : "/root");
1421 execle("/bin/bash", "-bash", NULL, (char**) envp);
1424 log_error("execv() failed: %m");
1427 _exit(EXIT_FAILURE);
1433 k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3);
1441 /* Kill if it is not dead yet anyway */
1442 terminate_machine(pid);
1444 /* Redundant, but better safe than sorry */
1447 k = wait_for_terminate(pid, &status);
1455 if (status.si_code == CLD_EXITED) {
1456 r = status.si_status;
1457 if (status.si_status != 0) {
1458 log_error("Container %s failed with error code %i.", arg_machine, status.si_status);
1462 log_debug("Container %s exited successfully.", arg_machine);
1464 } else if (status.si_code == CLD_KILLED &&
1465 status.si_status == SIGINT) {
1466 log_info("Container %s has been shut down.", arg_machine);
1469 } else if (status.si_code == CLD_KILLED &&
1470 status.si_status == SIGHUP) {
1471 log_info("Container %s is being rebooted.", arg_machine);
1473 } else if (status.si_code == CLD_KILLED ||
1474 status.si_code == CLD_DUMPED) {
1476 log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
1480 log_error("Container %s failed due to unknown reason.", arg_machine);
1490 free(arg_directory);