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 <linux/veth.h>
45 #include <sys/personality.h>
46 #include <linux/loop.h>
49 #include <selinux/selinux.h>
57 #include <blkid/blkid.h>
60 #include "sd-daemon.h"
70 #include "cgroup-util.h"
72 #include "path-util.h"
73 #include "loopback-setup.h"
74 #include "dev-setup.h"
79 #include "bus-error.h"
81 #include "bus-kernel.h"
84 #include "rtnl-util.h"
85 #include "udev-util.h"
86 #include "blkid-util.h"
88 #include "siphash24.h"
90 #include "base-filesystem.h"
94 #include "seccomp-util.h"
97 typedef enum ContainerStatus {
102 typedef enum LinkJournal {
109 typedef enum Volatile {
115 static char *arg_directory = NULL;
116 static char *arg_user = NULL;
117 static sd_id128_t arg_uuid = {};
118 static char *arg_machine = NULL;
119 static const char *arg_selinux_context = NULL;
120 static const char *arg_selinux_apifs_context = NULL;
121 static const char *arg_slice = NULL;
122 static bool arg_private_network = false;
123 static bool arg_read_only = false;
124 static bool arg_boot = false;
125 static LinkJournal arg_link_journal = LINK_AUTO;
126 static uint64_t arg_retain =
127 (1ULL << CAP_CHOWN) |
128 (1ULL << CAP_DAC_OVERRIDE) |
129 (1ULL << CAP_DAC_READ_SEARCH) |
130 (1ULL << CAP_FOWNER) |
131 (1ULL << CAP_FSETID) |
132 (1ULL << CAP_IPC_OWNER) |
134 (1ULL << CAP_LEASE) |
135 (1ULL << CAP_LINUX_IMMUTABLE) |
136 (1ULL << CAP_NET_BIND_SERVICE) |
137 (1ULL << CAP_NET_BROADCAST) |
138 (1ULL << CAP_NET_RAW) |
139 (1ULL << CAP_SETGID) |
140 (1ULL << CAP_SETFCAP) |
141 (1ULL << CAP_SETPCAP) |
142 (1ULL << CAP_SETUID) |
143 (1ULL << CAP_SYS_ADMIN) |
144 (1ULL << CAP_SYS_CHROOT) |
145 (1ULL << CAP_SYS_NICE) |
146 (1ULL << CAP_SYS_PTRACE) |
147 (1ULL << CAP_SYS_TTY_CONFIG) |
148 (1ULL << CAP_SYS_RESOURCE) |
149 (1ULL << CAP_SYS_BOOT) |
150 (1ULL << CAP_AUDIT_WRITE) |
151 (1ULL << CAP_AUDIT_CONTROL) |
153 static char **arg_bind = NULL;
154 static char **arg_bind_ro = NULL;
155 static char **arg_tmpfs = NULL;
156 static char **arg_setenv = NULL;
157 static bool arg_quiet = false;
158 static bool arg_share_system = false;
159 static bool arg_register = true;
160 static bool arg_keep_unit = false;
161 static char **arg_network_interfaces = NULL;
162 static char **arg_network_macvlan = NULL;
163 static bool arg_network_veth = false;
164 static const char *arg_network_bridge = NULL;
165 static unsigned long arg_personality = 0xffffffffLU;
166 static const char *arg_image = NULL;
167 static Volatile arg_volatile = VOLATILE_NO;
169 static void help(void) {
170 printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
171 "Spawn a minimal namespace container for debugging, testing and building.\n\n"
172 " -h --help Show this help\n"
173 " --version Print version string\n"
174 " -q --quiet Do not show status information\n"
175 " -D --directory=PATH Root directory for the container\n"
176 " -i --image=PATH File system device or image for the container\n"
177 " -b --boot Boot up full system (i.e. invoke init)\n"
178 " -u --user=USER Run the command under specified user or uid\n"
179 " -M --machine=NAME Set the machine name for the container\n"
180 " --uuid=UUID Set a specific machine UUID for the container\n"
181 " -S --slice=SLICE Place the container in the specified slice\n"
182 " --private-network Disable network in container\n"
183 " --network-interface=INTERFACE\n"
184 " Assign an existing network interface to the\n"
186 " --network-macvlan=INTERFACE\n"
187 " Create a macvlan network interface based on an\n"
188 " existing network interface to the container\n"
189 " --network-veth Add a virtual ethernet connection between host\n"
191 " --network-bridge=INTERFACE\n"
192 " Add a virtual ethernet connection between host\n"
193 " and container and add it to an existing bridge on\n"
195 " -Z --selinux-context=SECLABEL\n"
196 " Set the SELinux security context to be used by\n"
197 " processes in the container\n"
198 " -L --selinux-apifs-context=SECLABEL\n"
199 " Set the SELinux security context to be used by\n"
200 " API/tmpfs file systems in the container\n"
201 " --capability=CAP In addition to the default, retain specified\n"
203 " --drop-capability=CAP Drop the specified capability from the default set\n"
204 " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n"
205 " -j Equivalent to --link-journal=host\n"
206 " --read-only Mount the root directory read-only\n"
207 " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n"
209 " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n"
210 " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
211 " --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
212 " --share-system Share system namespaces with host\n"
213 " --register=BOOLEAN Register container as machine\n"
214 " --keep-unit Do not register a scope for the machine, reuse\n"
215 " the service unit nspawn is running in\n"
216 " --volatile[=MODE] Run the system in volatile mode\n",
217 program_invocation_short_name);
220 static int parse_argv(int argc, char *argv[]) {
237 ARG_NETWORK_INTERFACE,
245 static const struct option options[] = {
246 { "help", no_argument, NULL, 'h' },
247 { "version", no_argument, NULL, ARG_VERSION },
248 { "directory", required_argument, NULL, 'D' },
249 { "user", required_argument, NULL, 'u' },
250 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
251 { "boot", no_argument, NULL, 'b' },
252 { "uuid", required_argument, NULL, ARG_UUID },
253 { "read-only", no_argument, NULL, ARG_READ_ONLY },
254 { "capability", required_argument, NULL, ARG_CAPABILITY },
255 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
256 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
257 { "bind", required_argument, NULL, ARG_BIND },
258 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
259 { "tmpfs", required_argument, NULL, ARG_TMPFS },
260 { "machine", required_argument, NULL, 'M' },
261 { "slice", required_argument, NULL, 'S' },
262 { "setenv", required_argument, NULL, ARG_SETENV },
263 { "selinux-context", required_argument, NULL, 'Z' },
264 { "selinux-apifs-context", required_argument, NULL, 'L' },
265 { "quiet", no_argument, NULL, 'q' },
266 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM },
267 { "register", required_argument, NULL, ARG_REGISTER },
268 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
269 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
270 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
271 { "network-veth", no_argument, NULL, ARG_NETWORK_VETH },
272 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
273 { "personality", required_argument, NULL, ARG_PERSONALITY },
274 { "image", required_argument, NULL, 'i' },
275 { "volatile", optional_argument, NULL, ARG_VOLATILE },
280 uint64_t plus = 0, minus = 0;
285 while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:", options, NULL)) >= 0)
294 puts(PACKAGE_STRING);
295 puts(SYSTEMD_FEATURES);
300 arg_directory = canonicalize_file_name(optarg);
301 if (!arg_directory) {
302 log_error("Invalid root directory: %m");
314 arg_user = strdup(optarg);
320 case ARG_NETWORK_BRIDGE:
321 arg_network_bridge = optarg;
325 case ARG_NETWORK_VETH:
326 arg_network_veth = true;
327 arg_private_network = true;
330 case ARG_NETWORK_INTERFACE:
331 if (strv_extend(&arg_network_interfaces, optarg) < 0)
334 arg_private_network = true;
337 case ARG_NETWORK_MACVLAN:
338 if (strv_extend(&arg_network_macvlan, optarg) < 0)
343 case ARG_PRIVATE_NETWORK:
344 arg_private_network = true;
352 r = sd_id128_from_string(optarg, &arg_uuid);
354 log_error("Invalid UUID: %s", optarg);
364 if (isempty(optarg)) {
369 if (!hostname_is_valid(optarg)) {
370 log_error("Invalid machine name: %s", optarg);
375 arg_machine = strdup(optarg);
383 arg_selinux_context = optarg;
387 arg_selinux_apifs_context = optarg;
391 arg_read_only = true;
395 case ARG_DROP_CAPABILITY: {
396 const char *state, *word;
399 FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
400 _cleanup_free_ char *t;
403 t = strndup(word, length);
407 if (streq(t, "all")) {
408 if (c == ARG_CAPABILITY)
409 plus = (uint64_t) -1;
411 minus = (uint64_t) -1;
413 if (cap_from_name(t, &cap) < 0) {
414 log_error("Failed to parse capability %s.", t);
418 if (c == ARG_CAPABILITY)
419 plus |= 1ULL << (uint64_t) cap;
421 minus |= 1ULL << (uint64_t) cap;
429 arg_link_journal = LINK_GUEST;
432 case ARG_LINK_JOURNAL:
433 if (streq(optarg, "auto"))
434 arg_link_journal = LINK_AUTO;
435 else if (streq(optarg, "no"))
436 arg_link_journal = LINK_NO;
437 else if (streq(optarg, "guest"))
438 arg_link_journal = LINK_GUEST;
439 else if (streq(optarg, "host"))
440 arg_link_journal = LINK_HOST;
442 log_error("Failed to parse link journal mode %s", optarg);
450 _cleanup_free_ char *a = NULL, *b = NULL;
454 x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
456 e = strchr(optarg, ':');
458 a = strndup(optarg, e - optarg);
468 if (!path_is_absolute(a) || !path_is_absolute(b)) {
469 log_error("Invalid bind mount specification: %s", optarg);
473 r = strv_extend(x, a);
477 r = strv_extend(x, b);
485 _cleanup_free_ char *a = NULL, *b = NULL;
488 e = strchr(optarg, ':');
490 a = strndup(optarg, e - optarg);
494 b = strdup("mode=0755");
500 if (!path_is_absolute(a)) {
501 log_error("Invalid tmpfs specification: %s", optarg);
505 r = strv_push(&arg_tmpfs, a);
511 r = strv_push(&arg_tmpfs, b);
523 if (!env_assignment_is_valid(optarg)) {
524 log_error("Environment variable assignment '%s' is not valid.", optarg);
528 n = strv_env_set(arg_setenv, optarg);
532 strv_free(arg_setenv);
541 case ARG_SHARE_SYSTEM:
542 arg_share_system = true;
546 r = parse_boolean(optarg);
548 log_error("Failed to parse --register= argument: %s", optarg);
556 arg_keep_unit = true;
559 case ARG_PERSONALITY:
561 arg_personality = personality_from_string(optarg);
562 if (arg_personality == 0xffffffffLU) {
563 log_error("Unknown or unsupported personality '%s'.", optarg);
572 arg_volatile = VOLATILE_YES;
574 r = parse_boolean(optarg);
576 if (streq(optarg, "state"))
577 arg_volatile = VOLATILE_STATE;
579 log_error("Failed to parse --volatile= argument: %s", optarg);
583 arg_volatile = r ? VOLATILE_YES : VOLATILE_NO;
592 assert_not_reached("Unhandled option");
595 if (arg_share_system)
596 arg_register = false;
598 if (arg_boot && arg_share_system) {
599 log_error("--boot and --share-system may not be combined.");
603 if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) {
604 log_error("--keep-unit may not be used when invoked from a user session.");
608 if (arg_directory && arg_image) {
609 log_error("--directory= and --image= may not be combined.");
613 if (arg_volatile != VOLATILE_NO && arg_read_only) {
614 log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy.");
618 arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus;
623 static int mount_all(const char *dest) {
625 typedef struct MountPoint {
634 static const MountPoint mount_table[] = {
635 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
636 { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */
637 { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
638 { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
639 { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
640 { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true },
641 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
642 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
644 { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
645 { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
652 for (k = 0; k < ELEMENTSOF(mount_table); k++) {
653 _cleanup_free_ char *where = NULL;
655 _cleanup_free_ char *options = NULL;
660 where = strjoin(dest, "/", mount_table[k].where, NULL);
664 t = path_is_mount_point(where, true);
666 log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
674 /* Skip this entry if it is not a remount. */
675 if (mount_table[k].what && t > 0)
678 mkdir_p(where, 0755);
681 if (arg_selinux_apifs_context &&
682 (streq_ptr(mount_table[k].what, "tmpfs") || streq_ptr(mount_table[k].what, "devpts"))) {
683 options = strjoin(mount_table[k].options, ",context=\"", arg_selinux_apifs_context, "\"", NULL);
690 o = mount_table[k].options;
693 if (mount(mount_table[k].what,
696 mount_table[k].flags,
698 mount_table[k].fatal) {
700 log_error("mount(%s) failed: %m", where);
710 static int mount_binds(const char *dest, char **l, bool ro) {
713 STRV_FOREACH_PAIR(x, y, l) {
714 _cleanup_free_ char *where = NULL;
715 struct stat source_st, dest_st;
718 if (stat(*x, &source_st) < 0) {
719 log_error("Failed to stat %s: %m", *x);
723 where = strappend(dest, *y);
727 r = stat(where, &dest_st);
729 if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) {
730 log_error("The file types of %s and %s do not match. Refusing bind mount", *x, where);
733 } else if (errno == ENOENT) {
734 r = mkdir_parents_label(where, 0755);
736 log_error("Failed to bind mount %s: %s", *x, strerror(-r));
740 log_error("Failed to bind mount %s: %m", *x);
744 /* Create the mount point, but be conservative -- refuse to create block
745 * and char devices. */
746 if (S_ISDIR(source_st.st_mode))
747 mkdir_label(where, 0755);
748 else if (S_ISFIFO(source_st.st_mode))
750 else if (S_ISSOCK(source_st.st_mode))
751 mknod(where, 0644 | S_IFSOCK, 0);
752 else if (S_ISREG(source_st.st_mode))
755 log_error("Refusing to create mountpoint for file: %s", *x);
759 if (mount(*x, where, "bind", MS_BIND, NULL) < 0) {
760 log_error("mount(%s) failed: %m", where);
765 r = bind_remount_recursive(where, true);
767 log_error("Read-Only bind mount failed: %s", strerror(-r));
776 static int mount_tmpfs(const char *dest) {
779 STRV_FOREACH_PAIR(i, o, arg_tmpfs) {
780 _cleanup_free_ char *where = NULL;
782 where = strappend(dest, *i);
786 mkdir_label(where, 0755);
788 if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, *o) < 0) {
789 log_error("tmpfs mount to %s failed: %m", where);
797 static int setup_timezone(const char *dest) {
798 _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
804 /* Fix the timezone, if possible */
805 r = readlink_malloc("/etc/localtime", &p);
807 log_warning("/etc/localtime is not a symlink, not updating container timezone.");
811 z = path_startswith(p, "../usr/share/zoneinfo/");
813 z = path_startswith(p, "/usr/share/zoneinfo/");
815 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
819 where = strappend(dest, "/etc/localtime");
823 r = readlink_malloc(where, &q);
825 y = path_startswith(q, "../usr/share/zoneinfo/");
827 y = path_startswith(q, "/usr/share/zoneinfo/");
829 /* Already pointing to the right place? Then do nothing .. */
830 if (y && streq(y, z))
834 check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
838 if (access(check, F_OK) < 0) {
839 log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
843 what = strappend("../usr/share/zoneinfo/", z);
847 mkdir_parents(where, 0755);
850 if (symlink(what, where) < 0) {
851 log_error("Failed to correct timezone of container: %m");
858 static int setup_resolv_conf(const char *dest) {
859 _cleanup_free_ char *where = NULL;
863 if (arg_private_network)
866 /* Fix resolv.conf, if possible */
867 where = strappend(dest, "/etc/resolv.conf");
871 /* We don't really care for the results of this really. If it
872 * fails, it fails, but meh... */
873 mkdir_parents(where, 0755);
874 copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644);
879 static int setup_volatile_state(const char *directory) {
885 if (arg_volatile != VOLATILE_STATE)
888 /* --volatile=state means we simply overmount /var
889 with a tmpfs, and the rest read-only. */
891 r = bind_remount_recursive(directory, true);
893 log_error("Failed to remount %s read-only: %s", directory, strerror(-r));
897 p = strappenda(directory, "/var");
900 if (mount("tmpfs", p, "tmpfs", MS_STRICTATIME, "mode=755") < 0) {
901 log_error("Failed to mount tmpfs to /var: %m");
908 static int setup_volatile(const char *directory) {
909 bool tmpfs_mounted = false, bind_mounted = false;
910 char template[] = "/tmp/nspawn-volatile-XXXXXX";
916 if (arg_volatile != VOLATILE_YES)
919 /* --volatile=yes means we mount a tmpfs to the root dir, and
920 the original /usr to use inside it, and that read-only. */
922 if (!mkdtemp(template)) {
923 log_error("Failed to create temporary directory: %m");
927 if (mount("tmpfs", template, "tmpfs", MS_STRICTATIME, "mode=755") < 0) {
928 log_error("Failed to mount tmpfs for root directory: %m");
933 tmpfs_mounted = true;
935 f = strappenda(directory, "/usr");
936 t = strappenda(template, "/usr");
939 if (mount(f, t, "bind", MS_BIND|MS_REC, NULL) < 0) {
940 log_error("Failed to create /usr bind mount: %m");
947 r = bind_remount_recursive(t, true);
949 log_error("Failed to remount %s read-only: %s", t, strerror(-r));
953 if (mount(template, directory, NULL, MS_MOVE, NULL) < 0) {
954 log_error("Failed to move root mount: %m");
972 static char* id128_format_as_uuid(sd_id128_t id, char s[37]) {
975 "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
976 SD_ID128_FORMAT_VAL(id));
981 static int setup_boot_id(const char *dest) {
982 _cleanup_free_ char *from = NULL, *to = NULL;
989 if (arg_share_system)
992 /* Generate a new randomized boot ID, so that each boot-up of
993 * the container gets a new one */
995 from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
996 to = strappend(dest, "/proc/sys/kernel/random/boot_id");
1000 r = sd_id128_randomize(&rnd);
1002 log_error("Failed to generate random boot id: %s", strerror(-r));
1006 id128_format_as_uuid(rnd, as_uuid);
1008 r = write_string_file(from, as_uuid);
1010 log_error("Failed to write boot id: %s", strerror(-r));
1014 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
1015 log_error("Failed to bind mount boot id: %m");
1017 } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL))
1018 log_warning("Failed to make boot id read-only: %m");
1024 static int copy_devnodes(const char *dest) {
1026 static const char devnodes[] =
1036 _cleanup_umask_ mode_t u;
1042 NULSTR_FOREACH(d, devnodes) {
1043 _cleanup_free_ char *from = NULL, *to = NULL;
1046 from = strappend("/dev/", d);
1047 to = strjoin(dest, "/dev/", d, NULL);
1051 if (stat(from, &st) < 0) {
1053 if (errno != ENOENT) {
1054 log_error("Failed to stat %s: %m", from);
1058 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
1060 log_error("%s is not a char or block device, cannot copy", from);
1063 } else if (mknod(to, st.st_mode, st.st_rdev) < 0) {
1065 log_error("mknod(%s) failed: %m", dest);
1073 static int setup_ptmx(const char *dest) {
1074 _cleanup_free_ char *p = NULL;
1076 p = strappend(dest, "/dev/ptmx");
1080 if (symlink("pts/ptmx", p) < 0) {
1081 log_error("Failed to create /dev/ptmx symlink: %m");
1088 static int setup_dev_console(const char *dest, const char *console) {
1089 _cleanup_umask_ mode_t u;
1099 if (stat("/dev/null", &st) < 0) {
1100 log_error("Failed to stat /dev/null: %m");
1104 r = chmod_and_chown(console, 0600, 0, 0);
1106 log_error("Failed to correct access mode for TTY: %s", strerror(-r));
1110 /* We need to bind mount the right tty to /dev/console since
1111 * ptys can only exist on pts file systems. To have something
1112 * to bind mount things on we create a device node first, and
1113 * use /dev/null for that since we the cgroups device policy
1114 * allows us to create that freely, while we cannot create
1115 * /dev/console. (Note that the major minor doesn't actually
1116 * matter here, since we mount it over anyway). */
1118 to = strappenda(dest, "/dev/console");
1119 if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
1120 log_error("mknod() for /dev/console failed: %m");
1124 if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
1125 log_error("Bind mount for /dev/console failed: %m");
1132 static int setup_kmsg(const char *dest, int kmsg_socket) {
1133 _cleanup_free_ char *from = NULL, *to = NULL;
1135 _cleanup_umask_ mode_t u;
1137 struct cmsghdr cmsghdr;
1138 uint8_t buf[CMSG_SPACE(sizeof(int))];
1140 struct msghdr mh = {
1141 .msg_control = &control,
1142 .msg_controllen = sizeof(control),
1144 struct cmsghdr *cmsg;
1147 assert(kmsg_socket >= 0);
1151 /* We create the kmsg FIFO as /dev/kmsg, but immediately
1152 * delete it after bind mounting it to /proc/kmsg. While FIFOs
1153 * on the reading side behave very similar to /proc/kmsg,
1154 * their writing side behaves differently from /dev/kmsg in
1155 * that writing blocks when nothing is reading. In order to
1156 * avoid any problems with containers deadlocking due to this
1157 * we simply make /dev/kmsg unavailable to the container. */
1158 if (asprintf(&from, "%s/dev/kmsg", dest) < 0 ||
1159 asprintf(&to, "%s/proc/kmsg", dest) < 0)
1162 if (mkfifo(from, 0600) < 0) {
1163 log_error("mkfifo() for /dev/kmsg failed: %m");
1167 r = chmod_and_chown(from, 0600, 0, 0);
1169 log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
1173 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
1174 log_error("Bind mount for /proc/kmsg failed: %m");
1178 fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
1180 log_error("Failed to open fifo: %m");
1184 cmsg = CMSG_FIRSTHDR(&mh);
1185 cmsg->cmsg_level = SOL_SOCKET;
1186 cmsg->cmsg_type = SCM_RIGHTS;
1187 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
1188 memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
1190 mh.msg_controllen = cmsg->cmsg_len;
1192 /* Store away the fd in the socket, so that it stays open as
1193 * long as we run the child */
1194 k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL);
1198 log_error("Failed to send FIFO fd: %m");
1202 /* And now make the FIFO unavailable as /dev/kmsg... */
1207 static int setup_hostname(void) {
1209 if (arg_share_system)
1212 if (sethostname(arg_machine, strlen(arg_machine)) < 0)
1218 static int setup_journal(const char *directory) {
1219 sd_id128_t machine_id, this_id;
1220 _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL;
1224 p = strappend(directory, "/etc/machine-id");
1228 r = read_one_line_file(p, &b);
1229 if (r == -ENOENT && arg_link_journal == LINK_AUTO)
1232 log_error("Failed to read machine ID from %s: %s", p, strerror(-r));
1237 if (isempty(id) && arg_link_journal == LINK_AUTO)
1240 /* Verify validity */
1241 r = sd_id128_from_string(id, &machine_id);
1243 log_error("Failed to parse machine ID from %s: %s", p, strerror(-r));
1247 r = sd_id128_get_machine(&this_id);
1249 log_error("Failed to retrieve machine ID: %s", strerror(-r));
1253 if (sd_id128_equal(machine_id, this_id)) {
1254 log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
1255 "Host and machine ids are equal (%s): refusing to link journals", id);
1256 if (arg_link_journal == LINK_AUTO)
1262 if (arg_link_journal == LINK_NO)
1266 p = strappend("/var/log/journal/", id);
1267 q = strjoin(directory, "/var/log/journal/", id, NULL);
1271 if (path_is_mount_point(p, false) > 0) {
1272 if (arg_link_journal != LINK_AUTO) {
1273 log_error("%s: already a mount point, refusing to use for journal", p);
1280 if (path_is_mount_point(q, false) > 0) {
1281 if (arg_link_journal != LINK_AUTO) {
1282 log_error("%s: already a mount point, refusing to use for journal", q);
1289 r = readlink_and_make_absolute(p, &d);
1291 if ((arg_link_journal == LINK_GUEST ||
1292 arg_link_journal == LINK_AUTO) &&
1295 r = mkdir_p(q, 0755);
1297 log_warning("failed to create directory %s: %m", q);
1301 if (unlink(p) < 0) {
1302 log_error("Failed to remove symlink %s: %m", p);
1305 } else if (r == -EINVAL) {
1307 if (arg_link_journal == LINK_GUEST &&
1310 if (errno == ENOTDIR) {
1311 log_error("%s already exists and is neither a symlink nor a directory", p);
1314 log_error("Failed to remove %s: %m", p);
1318 } else if (r != -ENOENT) {
1319 log_error("readlink(%s) failed: %m", p);
1323 if (arg_link_journal == LINK_GUEST) {
1325 if (symlink(q, p) < 0) {
1326 log_error("Failed to symlink %s to %s: %m", q, p);
1330 r = mkdir_p(q, 0755);
1332 log_warning("failed to create directory %s: %m", q);
1336 if (arg_link_journal == LINK_HOST) {
1337 r = mkdir_p(p, 0755);
1339 log_error("Failed to create %s: %m", p);
1343 } else if (access(p, F_OK) < 0)
1346 if (dir_is_empty(q) == 0)
1347 log_warning("%s is not empty, proceeding anyway.", q);
1349 r = mkdir_p(q, 0755);
1351 log_error("Failed to create %s: %m", q);
1355 if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
1356 log_error("Failed to bind mount journal from host into guest: %m");
1363 static int setup_kdbus(const char *dest, const char *path) {
1369 p = strappenda(dest, "/dev/kdbus");
1370 if (mkdir(p, 0755) < 0) {
1371 log_error("Failed to create kdbus path: %m");
1375 if (mount(path, p, "bind", MS_BIND, NULL) < 0) {
1376 log_error("Failed to mount kdbus domain path: %m");
1383 static int drop_capabilities(void) {
1384 return capability_bounding_set_drop(~arg_retain, false);
1387 static int register_machine(pid_t pid, int local_ifindex) {
1388 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1389 _cleanup_bus_unref_ sd_bus *bus = NULL;
1395 r = sd_bus_default_system(&bus);
1397 log_error("Failed to open system bus: %s", strerror(-r));
1401 if (arg_keep_unit) {
1402 r = sd_bus_call_method(
1404 "org.freedesktop.machine1",
1405 "/org/freedesktop/machine1",
1406 "org.freedesktop.machine1.Manager",
1407 "RegisterMachineWithNetwork",
1412 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1416 strempty(arg_directory),
1417 local_ifindex > 0 ? 1 : 0, local_ifindex);
1419 _cleanup_bus_message_unref_ sd_bus_message *m = NULL;
1421 r = sd_bus_message_new_method_call(
1424 "org.freedesktop.machine1",
1425 "/org/freedesktop/machine1",
1426 "org.freedesktop.machine1.Manager",
1427 "CreateMachineWithNetwork");
1429 log_error("Failed to create message: %s", strerror(-r));
1433 r = sd_bus_message_append(
1437 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1441 strempty(arg_directory),
1442 local_ifindex > 0 ? 1 : 0, local_ifindex);
1444 log_error("Failed to append message arguments: %s", strerror(-r));
1448 r = sd_bus_message_open_container(m, 'a', "(sv)");
1450 log_error("Failed to open container: %s", strerror(-r));
1454 if (!isempty(arg_slice)) {
1455 r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice);
1457 log_error("Failed to append slice: %s", strerror(-r));
1462 r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict");
1464 log_error("Failed to add device policy: %s", strerror(-r));
1468 r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10,
1469 /* Allow the container to
1470 * access and create the API
1471 * device nodes, so that
1472 * PrivateDevices= in the
1473 * container can work
1478 "/dev/random", "rwm",
1479 "/dev/urandom", "rwm",
1481 /* Allow the container
1482 * access to ptys. However,
1484 * container to ever create
1485 * these device nodes. */
1486 "/dev/pts/ptmx", "rw",
1488 /* Allow the container
1489 * access to all kdbus
1490 * devices. Again, the
1491 * container cannot create
1492 * these nodes, only use
1493 * them. We use a pretty
1494 * open match here, so that
1495 * the kernel API can still
1498 "char-kdbus/*", "rw");
1500 log_error("Failed to add device whitelist: %s", strerror(-r));
1504 r = sd_bus_message_close_container(m);
1506 log_error("Failed to close container: %s", strerror(-r));
1510 r = sd_bus_call(bus, m, 0, &error, NULL);
1514 log_error("Failed to register machine: %s", bus_error_message(&error, r));
1521 static int terminate_machine(pid_t pid) {
1522 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1523 _cleanup_bus_message_unref_ sd_bus_message *reply = NULL;
1524 _cleanup_bus_unref_ sd_bus *bus = NULL;
1531 r = sd_bus_default_system(&bus);
1533 log_error("Failed to open system bus: %s", strerror(-r));
1537 r = sd_bus_call_method(
1539 "org.freedesktop.machine1",
1540 "/org/freedesktop/machine1",
1541 "org.freedesktop.machine1.Manager",
1548 /* Note that the machine might already have been
1549 * cleaned up automatically, hence don't consider it a
1550 * failure if we cannot get the machine object. */
1551 log_debug("Failed to get machine: %s", bus_error_message(&error, r));
1555 r = sd_bus_message_read(reply, "o", &path);
1557 return bus_log_parse_error(r);
1559 r = sd_bus_call_method(
1561 "org.freedesktop.machine1",
1563 "org.freedesktop.machine1.Machine",
1569 log_debug("Failed to terminate machine: %s", bus_error_message(&error, r));
1576 static int reset_audit_loginuid(void) {
1577 _cleanup_free_ char *p = NULL;
1580 if (arg_share_system)
1583 r = read_one_line_file("/proc/self/loginuid", &p);
1587 log_error("Failed to read /proc/self/loginuid: %s", strerror(-r));
1591 /* Already reset? */
1592 if (streq(p, "4294967295"))
1595 r = write_string_file("/proc/self/loginuid", "4294967295");
1597 log_error("Failed to reset audit login UID. This probably means that your kernel is too\n"
1598 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
1599 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
1600 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
1601 "using systemd-nspawn. Sleeping for 5s... (%s)\n", strerror(-r));
1609 #define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2)
1611 static int get_mac(struct ether_addr *mac) {
1618 l = strlen(arg_machine);
1619 sz = sizeof(sd_id128_t) + l;
1622 /* fetch some persistent data unique to the host */
1623 r = sd_id128_get_machine((sd_id128_t*) v);
1627 /* combine with some data unique (on this host) to this
1628 * container instance */
1629 memcpy(v + sizeof(sd_id128_t), arg_machine, l);
1631 /* Let's hash the host machine ID plus the container name. We
1632 * use a fixed, but originally randomly created hash key here. */
1633 siphash24(result, v, sz, HASH_KEY.bytes);
1635 assert_cc(ETH_ALEN <= sizeof(result));
1636 memcpy(mac->ether_addr_octet, result, ETH_ALEN);
1638 /* see eth_random_addr in the kernel */
1639 mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */
1640 mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */
1645 static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ], int *ifi) {
1646 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1647 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1648 struct ether_addr mac;
1651 if (!arg_private_network)
1654 if (!arg_network_veth)
1657 /* Use two different interface name prefixes depending whether
1658 * we are in bridge mode or not. */
1659 snprintf(iface_name, IFNAMSIZ, "%s-%s",
1660 arg_network_bridge ? "vb" : "ve", arg_machine);
1664 log_error("Failed to generate predictable MAC address for host0");
1668 r = sd_rtnl_open(&rtnl, 0);
1670 log_error("Failed to connect to netlink: %s", strerror(-r));
1674 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
1676 log_error("Failed to allocate netlink message: %s", strerror(-r));
1680 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name);
1682 log_error("Failed to add netlink interface name: %s", strerror(-r));
1686 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
1688 log_error("Failed to open netlink container: %s", strerror(-r));
1692 r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth");
1694 log_error("Failed to open netlink container: %s", strerror(-r));
1698 r = sd_rtnl_message_open_container(m, VETH_INFO_PEER);
1700 log_error("Failed to open netlink container: %s", strerror(-r));
1704 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0");
1706 log_error("Failed to add netlink interface name: %s", strerror(-r));
1710 r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac);
1712 log_error("Failed to add netlink MAC address: %s", strerror(-r));
1716 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1718 log_error("Failed to add netlink namespace field: %s", strerror(-r));
1722 r = sd_rtnl_message_close_container(m);
1724 log_error("Failed to close netlink container: %s", strerror(-r));
1728 r = sd_rtnl_message_close_container(m);
1730 log_error("Failed to close netlink container: %s", strerror(-r));
1734 r = sd_rtnl_message_close_container(m);
1736 log_error("Failed to close netlink container: %s", strerror(-r));
1740 r = sd_rtnl_call(rtnl, m, 0, NULL);
1742 log_error("Failed to add new veth interfaces: %s", strerror(-r));
1746 i = (int) if_nametoindex(iface_name);
1748 log_error("Failed to resolve interface %s: %m", iface_name);
1757 static int setup_bridge(const char veth_name[], int *ifi) {
1758 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1759 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1762 if (!arg_private_network)
1765 if (!arg_network_veth)
1768 if (!arg_network_bridge)
1771 bridge = (int) if_nametoindex(arg_network_bridge);
1773 log_error("Failed to resolve interface %s: %m", arg_network_bridge);
1779 r = sd_rtnl_open(&rtnl, 0);
1781 log_error("Failed to connect to netlink: %s", strerror(-r));
1785 r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0);
1787 log_error("Failed to allocate netlink message: %s", strerror(-r));
1791 r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP);
1793 log_error("Failed to set IFF_UP flag: %s", strerror(-r));
1797 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name);
1799 log_error("Failed to add netlink interface name field: %s", strerror(-r));
1803 r = sd_rtnl_message_append_u32(m, IFLA_MASTER, bridge);
1805 log_error("Failed to add netlink master field: %s", strerror(-r));
1809 r = sd_rtnl_call(rtnl, m, 0, NULL);
1811 log_error("Failed to add veth interface to bridge: %s", strerror(-r));
1818 static int parse_interface(struct udev *udev, const char *name) {
1819 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
1820 char ifi_str[2 + DECIMAL_STR_MAX(int)];
1823 ifi = (int) if_nametoindex(name);
1825 log_error("Failed to resolve interface %s: %m", name);
1829 sprintf(ifi_str, "n%i", ifi);
1830 d = udev_device_new_from_device_id(udev, ifi_str);
1832 log_error("Failed to get udev device for interface %s: %m", name);
1836 if (udev_device_get_is_initialized(d) <= 0) {
1837 log_error("Network interface %s is not initialized yet.", name);
1844 static int move_network_interfaces(pid_t pid) {
1845 _cleanup_udev_unref_ struct udev *udev = NULL;
1846 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1850 if (!arg_private_network)
1853 if (strv_isempty(arg_network_interfaces))
1856 r = sd_rtnl_open(&rtnl, 0);
1858 log_error("Failed to connect to netlink: %s", strerror(-r));
1864 log_error("Failed to connect to udev.");
1868 STRV_FOREACH(i, arg_network_interfaces) {
1869 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1872 ifi = parse_interface(udev, *i);
1876 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, ifi);
1878 log_error("Failed to allocate netlink message: %s", strerror(-r));
1882 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1884 log_error("Failed to append namespace PID to netlink message: %s", strerror(-r));
1888 r = sd_rtnl_call(rtnl, m, 0, NULL);
1890 log_error("Failed to move interface %s to namespace: %s", *i, strerror(-r));
1898 static int setup_macvlan(pid_t pid) {
1899 _cleanup_udev_unref_ struct udev *udev = NULL;
1900 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1904 if (!arg_private_network)
1907 if (strv_isempty(arg_network_macvlan))
1910 r = sd_rtnl_open(&rtnl, 0);
1912 log_error("Failed to connect to netlink: %s", strerror(-r));
1918 log_error("Failed to connect to udev.");
1922 STRV_FOREACH(i, arg_network_macvlan) {
1923 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1924 _cleanup_free_ char *n = NULL;
1927 ifi = parse_interface(udev, *i);
1931 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
1933 log_error("Failed to allocate netlink message: %s", strerror(-r));
1937 r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi);
1939 log_error("Failed to add netlink interface index: %s", strerror(-r));
1943 n = strappend("mv-", *i);
1947 strshorten(n, IFNAMSIZ-1);
1949 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n);
1951 log_error("Failed to add netlink interface name: %s", strerror(-r));
1955 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1957 log_error("Failed to add netlink namespace field: %s", strerror(-r));
1961 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
1963 log_error("Failed to open netlink container: %s", strerror(-r));
1967 r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "macvlan");
1969 log_error("Failed to open netlink container: %s", strerror(-r));
1973 r = sd_rtnl_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE);
1975 log_error("Failed to append macvlan mode: %s", strerror(-r));
1979 r = sd_rtnl_message_close_container(m);
1981 log_error("Failed to close netlink container: %s", strerror(-r));
1985 r = sd_rtnl_message_close_container(m);
1987 log_error("Failed to close netlink container: %s", strerror(-r));
1991 r = sd_rtnl_call(rtnl, m, 0, NULL);
1993 log_error("Failed to add new macvlan interfaces: %s", strerror(-r));
2001 static int setup_seccomp(void) {
2004 static const int blacklist[] = {
2005 SCMP_SYS(kexec_load),
2006 SCMP_SYS(open_by_handle_at),
2007 SCMP_SYS(init_module),
2008 SCMP_SYS(finit_module),
2009 SCMP_SYS(delete_module),
2016 scmp_filter_ctx seccomp;
2020 seccomp = seccomp_init(SCMP_ACT_ALLOW);
2024 r = seccomp_add_secondary_archs(seccomp);
2026 log_error("Failed to add secondary archs to seccomp filter: %s", strerror(-r));
2030 for (i = 0; i < ELEMENTSOF(blacklist); i++) {
2031 r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), blacklist[i], 0);
2033 continue; /* unknown syscall */
2035 log_error("Failed to block syscall: %s", strerror(-r));
2041 Audit is broken in containers, much of the userspace audit
2042 hookup will fail if running inside a container. We don't
2043 care and just turn off creation of audit sockets.
2045 This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail
2046 with EAFNOSUPPORT which audit userspace uses as indication
2047 that audit is disabled in the kernel.
2050 r = seccomp_rule_add(
2052 SCMP_ACT_ERRNO(EAFNOSUPPORT),
2055 SCMP_A0(SCMP_CMP_EQ, AF_NETLINK),
2056 SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT));
2058 log_error("Failed to add audit seccomp rule: %s", strerror(-r));
2062 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
2064 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
2068 r = seccomp_load(seccomp);
2070 log_error("Failed to install seccomp audit filter: %s", strerror(-r));
2073 seccomp_release(seccomp);
2081 static int setup_image(char **device_path, int *loop_nr) {
2082 struct loop_info64 info = {
2083 .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN
2085 _cleanup_close_ int fd = -1, control = -1, loop = -1;
2086 _cleanup_free_ char* loopdev = NULL;
2090 assert(device_path);
2093 fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
2095 log_error("Failed to open %s: %m", arg_image);
2099 if (fstat(fd, &st) < 0) {
2100 log_error("Failed to stat %s: %m", arg_image);
2104 if (S_ISBLK(st.st_mode)) {
2107 p = strdup(arg_image);
2121 if (!S_ISREG(st.st_mode)) {
2122 log_error("%s is not a regular file or block device: %m", arg_image);
2126 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
2128 log_error("Failed to open /dev/loop-control: %m");
2132 nr = ioctl(control, LOOP_CTL_GET_FREE);
2134 log_error("Failed to allocate loop device: %m");
2138 if (asprintf(&loopdev, "/dev/loop%i", nr) < 0)
2141 loop = open(loopdev, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
2143 log_error("Failed to open loop device %s: %m", loopdev);
2147 if (ioctl(loop, LOOP_SET_FD, fd) < 0) {
2148 log_error("Failed to set loopback file descriptor on %s: %m", loopdev);
2153 info.lo_flags |= LO_FLAGS_READ_ONLY;
2155 if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) {
2156 log_error("Failed to set loopback settings on %s: %m", loopdev);
2160 *device_path = loopdev;
2171 static int dissect_image(
2173 char **root_device, bool *root_device_rw,
2174 char **home_device, bool *home_device_rw,
2175 char **srv_device, bool *srv_device_rw,
2179 int home_nr = -1, root_nr = -1, secondary_root_nr = -1, srv_nr = -1;
2180 _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL;
2181 _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL;
2182 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
2183 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
2184 _cleanup_udev_unref_ struct udev *udev = NULL;
2185 struct udev_list_entry *first, *item;
2186 bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true;
2187 const char *pttype = NULL;
2193 assert(root_device);
2194 assert(home_device);
2198 b = blkid_new_probe();
2203 r = blkid_probe_set_device(b, fd, 0, 0);
2208 log_error("Failed to set device on blkid probe: %m");
2212 blkid_probe_enable_partitions(b, 1);
2213 blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
2216 r = blkid_do_safeprobe(b);
2217 if (r == -2 || r == 1) {
2218 log_error("Failed to identify any partition table on %s.\n"
2219 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
2221 } else if (r != 0) {
2224 log_error("Failed to probe: %m");
2228 blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
2229 if (!streq_ptr(pttype, "gpt")) {
2230 log_error("Image %s does not carry a GUID Partition Table.\n"
2231 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
2236 pl = blkid_probe_get_partitions(b);
2241 log_error("Failed to list partitions of %s", arg_image);
2249 if (fstat(fd, &st) < 0) {
2250 log_error("Failed to stat block device: %m");
2254 d = udev_device_new_from_devnum(udev, 'b', st.st_rdev);
2258 e = udev_enumerate_new(udev);
2262 r = udev_enumerate_add_match_parent(e, d);
2266 r = udev_enumerate_scan_devices(e);
2268 log_error("Failed to scan for partition devices of %s: %s", arg_image, strerror(-r));
2272 first = udev_enumerate_get_list_entry(e);
2273 udev_list_entry_foreach(item, first) {
2274 _cleanup_udev_device_unref_ struct udev_device *q;
2275 const char *stype, *node;
2276 unsigned long long flags;
2283 q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));
2288 log_error("Failed to get partition device of %s: %m", arg_image);
2292 qn = udev_device_get_devnum(q);
2296 if (st.st_rdev == qn)
2299 node = udev_device_get_devnode(q);
2303 pp = blkid_partlist_devno_to_partition(pl, qn);
2307 flags = blkid_partition_get_flags(pp);
2308 if (flags & GPT_FLAG_NO_AUTO)
2311 nr = blkid_partition_get_partno(pp);
2315 stype = blkid_partition_get_type_string(pp);
2319 if (sd_id128_from_string(stype, &type_id) < 0)
2322 if (sd_id128_equal(type_id, GPT_HOME)) {
2324 if (home && nr >= home_nr)
2328 home_rw = !(flags & GPT_FLAG_READ_ONLY);
2331 home = strdup(node);
2334 } else if (sd_id128_equal(type_id, GPT_SRV)) {
2336 if (srv && nr >= srv_nr)
2340 srv_rw = !(flags & GPT_FLAG_READ_ONLY);
2347 #ifdef GPT_ROOT_NATIVE
2348 else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) {
2350 if (root && nr >= root_nr)
2354 root_rw = !(flags & GPT_FLAG_READ_ONLY);
2357 root = strdup(node);
2362 #ifdef GPT_ROOT_SECONDARY
2363 else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) {
2365 if (secondary_root && nr >= secondary_root_nr)
2368 secondary_root_nr = nr;
2369 secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY);
2372 free(secondary_root);
2373 secondary_root = strdup(node);
2374 if (!secondary_root)
2380 if (!root && !secondary_root) {
2381 log_error("Failed to identify root partition in disk image %s.\n"
2382 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
2387 *root_device = root;
2390 *root_device_rw = root_rw;
2392 } else if (secondary_root) {
2393 *root_device = secondary_root;
2394 secondary_root = NULL;
2396 *root_device_rw = secondary_root_rw;
2401 *home_device = home;
2404 *home_device_rw = home_rw;
2411 *srv_device_rw = srv_rw;
2416 log_error("--image= is not supported, compiled without blkid support.");
2421 static int mount_device(const char *what, const char *where, const char *directory, bool rw) {
2423 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
2424 const char *fstype, *p;
2434 p = strappenda(where, directory);
2439 b = blkid_new_probe_from_filename(what);
2443 log_error("Failed to allocate prober for %s: %m", what);
2447 blkid_probe_enable_superblocks(b, 1);
2448 blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE);
2451 r = blkid_do_safeprobe(b);
2452 if (r == -1 || r == 1) {
2453 log_error("Cannot determine file system type of %s", what);
2455 } else if (r != 0) {
2458 log_error("Failed to probe %s: %m", what);
2463 if (blkid_probe_lookup_value(b, "TYPE", &fstype, NULL) < 0) {
2466 log_error("Failed to determine file system type of %s", what);
2470 if (streq(fstype, "crypto_LUKS")) {
2471 log_error("nspawn currently does not support LUKS disk images.");
2475 if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) {
2476 log_error("Failed to mount %s: %m", what);
2482 log_error("--image= is not supported, compiled without blkid support.");
2487 static int mount_devices(
2489 const char *root_device, bool root_device_rw,
2490 const char *home_device, bool home_device_rw,
2491 const char *srv_device, bool srv_device_rw) {
2497 r = mount_device(root_device, arg_directory, NULL, root_device_rw);
2499 log_error("Failed to mount root directory: %s", strerror(-r));
2505 r = mount_device(home_device, arg_directory, "/home", home_device_rw);
2507 log_error("Failed to mount home directory: %s", strerror(-r));
2513 r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw);
2515 log_error("Failed to mount server data directory: %s", strerror(-r));
2523 static void loop_remove(int nr, int *image_fd) {
2524 _cleanup_close_ int control = -1;
2529 if (image_fd && *image_fd >= 0) {
2530 ioctl(*image_fd, LOOP_CLR_FD);
2531 *image_fd = safe_close(*image_fd);
2534 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
2538 ioctl(control, LOOP_CTL_REMOVE, nr);
2541 static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
2549 if (pipe2(pipe_fds, O_CLOEXEC) < 0) {
2550 log_error("Failed to allocate pipe: %m");
2556 log_error("Failed to fork getent child: %m");
2558 } else if (pid == 0) {
2560 char *empty_env = NULL;
2562 if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0)
2563 _exit(EXIT_FAILURE);
2565 if (pipe_fds[0] > 2)
2566 safe_close(pipe_fds[0]);
2567 if (pipe_fds[1] > 2)
2568 safe_close(pipe_fds[1]);
2570 nullfd = open("/dev/null", O_RDWR);
2572 _exit(EXIT_FAILURE);
2574 if (dup3(nullfd, STDIN_FILENO, 0) < 0)
2575 _exit(EXIT_FAILURE);
2577 if (dup3(nullfd, STDERR_FILENO, 0) < 0)
2578 _exit(EXIT_FAILURE);
2583 reset_all_signal_handlers();
2584 close_all_fds(NULL, 0);
2586 execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env);
2587 execle("/bin/getent", "getent", database, key, NULL, &empty_env);
2588 _exit(EXIT_FAILURE);
2591 pipe_fds[1] = safe_close(pipe_fds[1]);
2598 static int change_uid_gid(char **_home) {
2599 char line[LINE_MAX], *x, *u, *g, *h;
2600 const char *word, *state;
2601 _cleanup_free_ uid_t *uids = NULL;
2602 _cleanup_free_ char *home = NULL;
2603 _cleanup_fclose_ FILE *f = NULL;
2604 _cleanup_close_ int fd = -1;
2605 unsigned n_uids = 0;
2614 if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) {
2615 /* Reset everything fully to 0, just in case */
2617 if (setgroups(0, NULL) < 0) {
2618 log_error("setgroups() failed: %m");
2622 if (setresgid(0, 0, 0) < 0) {
2623 log_error("setregid() failed: %m");
2627 if (setresuid(0, 0, 0) < 0) {
2628 log_error("setreuid() failed: %m");
2636 /* First, get user credentials */
2637 fd = spawn_getent("passwd", arg_user, &pid);
2641 f = fdopen(fd, "r");
2646 if (!fgets(line, sizeof(line), f)) {
2649 log_error("Failed to resolve user %s.", arg_user);
2653 log_error("Failed to read from getent: %m");
2659 wait_for_terminate_and_warn("getent passwd", pid);
2661 x = strchr(line, ':');
2663 log_error("/etc/passwd entry has invalid user field.");
2667 u = strchr(x+1, ':');
2669 log_error("/etc/passwd entry has invalid password field.");
2676 log_error("/etc/passwd entry has invalid UID field.");
2684 log_error("/etc/passwd entry has invalid GID field.");
2689 h = strchr(x+1, ':');
2691 log_error("/etc/passwd entry has invalid GECOS field.");
2698 log_error("/etc/passwd entry has invalid home directory field.");
2704 r = parse_uid(u, &uid);
2706 log_error("Failed to parse UID of user.");
2710 r = parse_gid(g, &gid);
2712 log_error("Failed to parse GID of user.");
2720 /* Second, get group memberships */
2721 fd = spawn_getent("initgroups", arg_user, &pid);
2726 f = fdopen(fd, "r");
2731 if (!fgets(line, sizeof(line), f)) {
2733 log_error("Failed to resolve user %s.", arg_user);
2737 log_error("Failed to read from getent: %m");
2743 wait_for_terminate_and_warn("getent initgroups", pid);
2745 /* Skip over the username and subsequent separator whitespace */
2747 x += strcspn(x, WHITESPACE);
2748 x += strspn(x, WHITESPACE);
2750 FOREACH_WORD(word, l, x, state) {
2756 if (!GREEDY_REALLOC(uids, sz, n_uids+1))
2759 r = parse_uid(c, &uids[n_uids++]);
2761 log_error("Failed to parse group data from getent.");
2766 r = mkdir_parents(home, 0775);
2768 log_error("Failed to make home root directory: %s", strerror(-r));
2772 r = mkdir_safe(home, 0755, uid, gid);
2773 if (r < 0 && r != -EEXIST) {
2774 log_error("Failed to make home directory: %s", strerror(-r));
2778 fchown(STDIN_FILENO, uid, gid);
2779 fchown(STDOUT_FILENO, uid, gid);
2780 fchown(STDERR_FILENO, uid, gid);
2782 if (setgroups(n_uids, uids) < 0) {
2783 log_error("Failed to set auxiliary groups: %m");
2787 if (setresgid(gid, gid, gid) < 0) {
2788 log_error("setregid() failed: %m");
2792 if (setresuid(uid, uid, uid) < 0) {
2793 log_error("setreuid() failed: %m");
2807 * < 0 : wait_for_terminate() failed to get the state of the
2808 * container, the container was terminated by a signal, or
2809 * failed for an unknown reason. No change is made to the
2810 * container argument.
2811 * > 0 : The program executed in the container terminated with an
2812 * error. The exit code of the program executed in the
2813 * container is returned. No change is made to the container
2815 * 0 : The container is being rebooted, has been shut down or exited
2816 * successfully. The container argument has been set to either
2817 * CONTAINER_TERMINATED or CONTAINER_REBOOTED.
2819 * That is, success is indicated by a return value of zero, and an
2820 * error is indicated by a non-zero value.
2822 static int wait_for_container(pid_t pid, ContainerStatus *container) {
2826 r = wait_for_terminate(pid, &status);
2828 log_warning("Failed to wait for container: %s", strerror(-r));
2832 switch (status.si_code) {
2834 r = status.si_status;
2837 log_debug("Container %s exited successfully.",
2840 *container = CONTAINER_TERMINATED;
2842 log_error("Container %s failed with error code %i.",
2843 arg_machine, status.si_status);
2848 if (status.si_status == SIGINT) {
2850 log_info("Container %s has been shut down.",
2853 *container = CONTAINER_TERMINATED;
2856 } else if (status.si_status == SIGHUP) {
2858 log_info("Container %s is being rebooted.",
2861 *container = CONTAINER_REBOOTED;
2865 /* CLD_KILLED fallthrough */
2868 log_error("Container %s terminated by signal %s.",
2869 arg_machine, signal_to_string(status.si_status));
2874 log_error("Container %s failed due to unknown reason.",
2883 static void nop_handler(int sig) {}
2885 int main(int argc, char *argv[]) {
2887 _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL;
2888 bool root_device_rw = true, home_device_rw = true, srv_device_rw = true;
2889 _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1;
2890 _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 };
2891 _cleanup_fdset_free_ FDSet *fds = NULL;
2892 int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1;
2893 const char *console = NULL;
2894 char veth_name[IFNAMSIZ];
2895 bool secondary = false;
2896 sigset_t mask, mask_chld;
2899 log_parse_environment();
2902 k = parse_argv(argc, argv);
2911 if (arg_directory) {
2914 p = path_make_absolute_cwd(arg_directory);
2915 free(arg_directory);
2918 arg_directory = get_current_dir_name();
2920 if (!arg_directory) {
2921 log_error("Failed to determine path, please use -D.");
2924 path_kill_slashes(arg_directory);
2928 arg_machine = strdup(basename(arg_image ? arg_image : arg_directory));
2934 hostname_cleanup(arg_machine, false);
2935 if (isempty(arg_machine)) {
2936 log_error("Failed to determine machine name automatically, please use -M.");
2941 if (geteuid() != 0) {
2942 log_error("Need to be root.");
2946 if (sd_booted() <= 0) {
2947 log_error("Not running on a systemd system.");
2952 n_fd_passed = sd_listen_fds(false);
2953 if (n_fd_passed > 0) {
2954 k = fdset_new_listen_fds(&fds, false);
2956 log_error("Failed to collect file descriptors: %s", strerror(-k));
2960 fdset_close_others(fds);
2963 if (arg_directory) {
2964 if (path_equal(arg_directory, "/")) {
2965 log_error("Spawning container on root directory not supported.");
2970 if (path_is_os_tree(arg_directory) <= 0) {
2971 log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory);
2977 p = strappenda(arg_directory,
2978 argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/");
2979 if (access(p, F_OK) < 0) {
2980 log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory);
2986 char template[] = "/tmp/nspawn-root-XXXXXX";
2988 if (!mkdtemp(template)) {
2989 log_error("Failed to create temporary directory: %m");
2994 arg_directory = strdup(template);
2995 if (!arg_directory) {
3000 image_fd = setup_image(&device_path, &loop_nr);
3006 r = dissect_image(image_fd,
3007 &root_device, &root_device_rw,
3008 &home_device, &home_device_rw,
3009 &srv_device, &srv_device_rw,
3015 master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
3017 log_error("Failed to acquire pseudo tty: %m");
3021 console = ptsname(master);
3023 log_error("Failed to determine tty name: %m");
3028 log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.",
3029 arg_machine, arg_image ? arg_image : arg_directory);
3031 if (unlockpt(master) < 0) {
3032 log_error("Failed to unlock tty: %m");
3036 if (access("/dev/kdbus/control", F_OK) >= 0) {
3038 if (arg_share_system) {
3039 kdbus_domain = strdup("/dev/kdbus");
3040 if (!kdbus_domain) {
3047 ns = strappenda("machine-", arg_machine);
3048 kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain);
3050 log_debug("Failed to create kdbus domain: %s", strerror(-r));
3052 log_debug("Successfully created kdbus domain as %s", kdbus_domain);
3056 if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
3057 log_error("Failed to create kmsg socket pair: %m");
3061 sd_notify(0, "READY=1");
3063 assert_se(sigemptyset(&mask) == 0);
3064 assert_se(sigemptyset(&mask_chld) == 0);
3065 sigaddset(&mask_chld, SIGCHLD);
3066 sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
3067 assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
3070 ContainerStatus container_status;
3071 _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
3072 struct sigaction sa = {
3073 .sa_handler = nop_handler,
3074 .sa_flags = SA_NOCLDSTOP,
3077 r = barrier_create(&barrier);
3079 log_error("Cannot initialize IPC barrier: %s", strerror(-r));
3083 /* Child can be killed before execv(), so handle SIGCHLD
3084 * in order to interrupt parent's blocking calls and
3085 * give it a chance to call wait() and terminate. */
3086 r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
3088 log_error("Failed to change the signal mask: %m");
3092 r = sigaction(SIGCHLD, &sa, NULL);
3094 log_error("Failed to install SIGCHLD handler: %m");
3098 pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWNS|
3099 (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)|
3100 (arg_private_network ? CLONE_NEWNET : 0), NULL);
3102 if (errno == EINVAL)
3103 log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
3105 log_error("clone() failed: %m");
3113 _cleanup_free_ char *home = NULL;
3115 const char *envp[] = {
3116 "PATH=" DEFAULT_PATH_SPLIT_USR,
3117 "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
3122 NULL, /* container_uuid */
3123 NULL, /* LISTEN_FDS */
3124 NULL, /* LISTEN_PID */
3129 barrier_set_role(&barrier, BARRIER_CHILD);
3131 envp[n_env] = strv_find_prefix(environ, "TERM=");
3135 master = safe_close(master);
3137 close_nointr(STDIN_FILENO);
3138 close_nointr(STDOUT_FILENO);
3139 close_nointr(STDERR_FILENO);
3141 kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
3143 reset_all_signal_handlers();
3145 assert_se(sigemptyset(&mask) == 0);
3146 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
3148 k = open_terminal(console, O_RDWR);
3149 if (k != STDIN_FILENO) {
3155 log_error("Failed to open console: %s", strerror(-k));
3156 _exit(EXIT_FAILURE);
3159 if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
3160 dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
3161 log_error("Failed to duplicate console: %m");
3162 _exit(EXIT_FAILURE);
3166 log_error("setsid() failed: %m");
3167 _exit(EXIT_FAILURE);
3170 if (reset_audit_loginuid() < 0)
3171 _exit(EXIT_FAILURE);
3173 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
3174 log_error("PR_SET_PDEATHSIG failed: %m");
3175 _exit(EXIT_FAILURE);
3178 /* Mark everything as slave, so that we still
3179 * receive mounts from the real root, but don't
3180 * propagate mounts to the real root. */
3181 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
3182 log_error("MS_SLAVE|MS_REC failed: %m");
3183 _exit(EXIT_FAILURE);
3186 if (mount_devices(arg_directory,
3187 root_device, root_device_rw,
3188 home_device, home_device_rw,
3189 srv_device, srv_device_rw) < 0)
3190 _exit(EXIT_FAILURE);
3192 /* Turn directory into bind mount */
3193 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
3194 log_error("Failed to make bind mount: %m");
3195 _exit(EXIT_FAILURE);
3198 r = setup_volatile(arg_directory);
3200 _exit(EXIT_FAILURE);
3202 if (setup_volatile_state(arg_directory) < 0)
3203 _exit(EXIT_FAILURE);
3205 r = base_filesystem_create(arg_directory);
3207 _exit(EXIT_FAILURE);
3209 if (arg_read_only) {
3210 k = bind_remount_recursive(arg_directory, true);
3212 log_error("Failed to make tree read-only: %s", strerror(-k));
3213 _exit(EXIT_FAILURE);
3217 if (mount_all(arg_directory) < 0)
3218 _exit(EXIT_FAILURE);
3220 if (copy_devnodes(arg_directory) < 0)
3221 _exit(EXIT_FAILURE);
3223 if (setup_ptmx(arg_directory) < 0)
3224 _exit(EXIT_FAILURE);
3226 dev_setup(arg_directory);
3228 if (setup_seccomp() < 0)
3229 _exit(EXIT_FAILURE);
3231 if (setup_dev_console(arg_directory, console) < 0)
3232 _exit(EXIT_FAILURE);
3234 if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
3235 _exit(EXIT_FAILURE);
3237 kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
3239 if (setup_boot_id(arg_directory) < 0)
3240 _exit(EXIT_FAILURE);
3242 if (setup_timezone(arg_directory) < 0)
3243 _exit(EXIT_FAILURE);
3245 if (setup_resolv_conf(arg_directory) < 0)
3246 _exit(EXIT_FAILURE);
3248 if (setup_journal(arg_directory) < 0)
3249 _exit(EXIT_FAILURE);
3251 if (mount_binds(arg_directory, arg_bind, false) < 0)
3252 _exit(EXIT_FAILURE);
3254 if (mount_binds(arg_directory, arg_bind_ro, true) < 0)
3255 _exit(EXIT_FAILURE);
3257 if (mount_tmpfs(arg_directory) < 0)
3258 _exit(EXIT_FAILURE);
3260 if (setup_kdbus(arg_directory, kdbus_domain) < 0)
3261 _exit(EXIT_FAILURE);
3263 /* Tell the parent that we are ready, and that
3264 * it can cgroupify us to that we lack access
3265 * to certain devices and resources. */
3266 barrier_place(&barrier);
3268 if (chdir(arg_directory) < 0) {
3269 log_error("chdir(%s) failed: %m", arg_directory);
3270 _exit(EXIT_FAILURE);
3273 if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
3274 log_error("mount(MS_MOVE) failed: %m");
3275 _exit(EXIT_FAILURE);
3278 if (chroot(".") < 0) {
3279 log_error("chroot() failed: %m");
3280 _exit(EXIT_FAILURE);
3283 if (chdir("/") < 0) {
3284 log_error("chdir() failed: %m");
3285 _exit(EXIT_FAILURE);
3290 if (arg_private_network)
3293 if (drop_capabilities() < 0) {
3294 log_error("drop_capabilities() failed: %m");
3295 _exit(EXIT_FAILURE);
3298 r = change_uid_gid(&home);
3300 _exit(EXIT_FAILURE);
3302 if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
3303 (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
3304 (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
3306 _exit(EXIT_FAILURE);
3309 if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) {
3312 if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) {
3314 _exit(EXIT_FAILURE);
3318 if (fdset_size(fds) > 0) {
3319 k = fdset_cloexec(fds, false);
3321 log_error("Failed to unset O_CLOEXEC for file descriptors.");
3322 _exit(EXIT_FAILURE);
3325 if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
3326 (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) {
3328 _exit(EXIT_FAILURE);
3334 if (arg_personality != 0xffffffffLU) {
3335 if (personality(arg_personality) < 0) {
3336 log_error("personality() failed: %m");
3337 _exit(EXIT_FAILURE);
3339 } else if (secondary) {
3340 if (personality(PER_LINUX32) < 0) {
3341 log_error("personality() failed: %m");
3342 _exit(EXIT_FAILURE);
3347 if (arg_selinux_context)
3348 if (setexeccon((security_context_t) arg_selinux_context) < 0) {
3349 log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context);
3350 _exit(EXIT_FAILURE);
3354 if (!strv_isempty(arg_setenv)) {
3357 n = strv_env_merge(2, envp, arg_setenv);
3360 _exit(EXIT_FAILURE);
3365 env_use = (char**) envp;
3367 /* Wait until the parent is ready with the setup, too... */
3368 if (!barrier_place_and_sync(&barrier))
3369 _exit(EXIT_FAILURE);
3375 /* Automatically search for the init system */
3377 l = 1 + argc - optind;
3378 a = newa(char*, l + 1);
3379 memcpy(a + 1, argv + optind, l * sizeof(char*));
3381 a[0] = (char*) "/usr/lib/systemd/systemd";
3382 execve(a[0], a, env_use);
3384 a[0] = (char*) "/lib/systemd/systemd";
3385 execve(a[0], a, env_use);
3387 a[0] = (char*) "/sbin/init";
3388 execve(a[0], a, env_use);
3389 } else if (argc > optind)
3390 execvpe(argv[optind], argv + optind, env_use);
3392 chdir(home ? home : "/root");
3393 execle("/bin/bash", "-bash", NULL, env_use);
3394 execle("/bin/sh", "-sh", NULL, env_use);
3397 log_error("execv() failed: %m");
3398 _exit(EXIT_FAILURE);
3401 barrier_set_role(&barrier, BARRIER_PARENT);
3405 /* wait for child-setup to be done */
3406 if (barrier_place_and_sync(&barrier)) {
3409 r = move_network_interfaces(pid);
3413 r = setup_veth(pid, veth_name, &ifi);
3417 r = setup_bridge(veth_name, &ifi);
3421 r = setup_macvlan(pid);
3425 r = register_machine(pid, ifi);
3429 /* Block SIGCHLD here, before notifying child.
3430 * process_pty() will handle it with the other signals. */
3431 r = sigprocmask(SIG_BLOCK, &mask_chld, NULL);
3435 /* Reset signal to default */
3436 r = default_signals(SIGCHLD, -1);
3440 /* Notify the child that the parent is ready with all
3441 * its setup, and that the child can now hand over
3442 * control to the code to run inside the container. */
3443 barrier_place(&barrier);
3445 k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3);
3454 /* Kill if it is not dead yet anyway */
3455 terminate_machine(pid);
3458 /* Normally redundant, but better safe than sorry */
3461 r = wait_for_container(pid, &container_status);
3465 /* We failed to wait for the container, or the
3466 * container exited abnormally */
3469 } else if (r > 0 || container_status == CONTAINER_TERMINATED)
3470 /* The container exited with a non-zero
3471 * status, or with zero status and no reboot
3475 /* CONTAINER_REBOOTED, loop again */
3477 if (arg_keep_unit) {
3478 /* Special handling if we are running as a
3479 * service: instead of simply restarting the
3480 * machine we want to restart the entire
3481 * service, so let's inform systemd about this
3482 * with the special exit code 133. The service
3483 * file uses RestartForceExitStatus=133 so
3484 * that this results in a full nspawn
3485 * restart. This is necessary since we might
3486 * have cgroup parameters set we want to have
3494 loop_remove(loop_nr, &image_fd);
3499 free(arg_directory);
3502 strv_free(arg_setenv);
3503 strv_free(arg_network_interfaces);
3504 strv_free(arg_network_macvlan);
3505 strv_free(arg_bind);
3506 strv_free(arg_bind_ro);
3507 strv_free(arg_tmpfs);