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>
36 #include <sys/signalfd.h>
40 #include <sys/socket.h>
41 #include <linux/netlink.h>
43 #include <linux/veth.h>
44 #include <sys/personality.h>
45 #include <linux/loop.h>
50 #include <selinux/selinux.h>
58 #include <blkid/blkid.h>
61 #include "sd-daemon.h"
71 #include "cgroup-util.h"
73 #include "path-util.h"
74 #include "loopback-setup.h"
75 #include "dev-setup.h"
80 #include "bus-error.h"
82 #include "bus-kernel.h"
85 #include "rtnl-util.h"
86 #include "udev-util.h"
87 #include "blkid-util.h"
89 #include "siphash24.h"
91 #include "base-filesystem.h"
93 #include "event-util.h"
94 #include "capability.h"
96 #include "btrfs-util.h"
97 #include "machine-image.h"
99 #include "in-addr-util.h"
101 #include "local-addresses.h"
104 #include "seccomp-util.h"
107 typedef struct ExposePort {
110 uint16_t container_port;
111 LIST_FIELDS(struct ExposePort, ports);
114 typedef enum ContainerStatus {
115 CONTAINER_TERMINATED,
119 typedef enum LinkJournal {
126 typedef enum Volatile {
132 static char *arg_directory = NULL;
133 static char *arg_template = NULL;
134 static char *arg_user = NULL;
135 static sd_id128_t arg_uuid = {};
136 static char *arg_machine = NULL;
137 static const char *arg_selinux_context = NULL;
138 static const char *arg_selinux_apifs_context = NULL;
139 static const char *arg_slice = NULL;
140 static bool arg_private_network = false;
141 static bool arg_read_only = false;
142 static bool arg_boot = false;
143 static bool arg_ephemeral = false;
144 static LinkJournal arg_link_journal = LINK_AUTO;
145 static bool arg_link_journal_try = false;
146 static uint64_t arg_retain =
147 (1ULL << CAP_CHOWN) |
148 (1ULL << CAP_DAC_OVERRIDE) |
149 (1ULL << CAP_DAC_READ_SEARCH) |
150 (1ULL << CAP_FOWNER) |
151 (1ULL << CAP_FSETID) |
152 (1ULL << CAP_IPC_OWNER) |
154 (1ULL << CAP_LEASE) |
155 (1ULL << CAP_LINUX_IMMUTABLE) |
156 (1ULL << CAP_NET_BIND_SERVICE) |
157 (1ULL << CAP_NET_BROADCAST) |
158 (1ULL << CAP_NET_RAW) |
159 (1ULL << CAP_SETGID) |
160 (1ULL << CAP_SETFCAP) |
161 (1ULL << CAP_SETPCAP) |
162 (1ULL << CAP_SETUID) |
163 (1ULL << CAP_SYS_ADMIN) |
164 (1ULL << CAP_SYS_CHROOT) |
165 (1ULL << CAP_SYS_NICE) |
166 (1ULL << CAP_SYS_PTRACE) |
167 (1ULL << CAP_SYS_TTY_CONFIG) |
168 (1ULL << CAP_SYS_RESOURCE) |
169 (1ULL << CAP_SYS_BOOT) |
170 (1ULL << CAP_AUDIT_WRITE) |
171 (1ULL << CAP_AUDIT_CONTROL) |
173 static char **arg_bind = NULL;
174 static char **arg_bind_ro = NULL;
175 static char **arg_tmpfs = NULL;
176 static char **arg_setenv = NULL;
177 static bool arg_quiet = false;
178 static bool arg_share_system = false;
179 static bool arg_register = true;
180 static bool arg_keep_unit = false;
181 static char **arg_network_interfaces = NULL;
182 static char **arg_network_macvlan = NULL;
183 static char **arg_network_ipvlan = NULL;
184 static bool arg_network_veth = false;
185 static const char *arg_network_bridge = NULL;
186 static unsigned long arg_personality = 0xffffffffLU;
187 static char *arg_image = NULL;
188 static Volatile arg_volatile = VOLATILE_NO;
189 static ExposePort *arg_expose_ports = NULL;
191 static void help(void) {
192 printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
193 "Spawn a minimal namespace container for debugging, testing and building.\n\n"
194 " -h --help Show this help\n"
195 " --version Print version string\n"
196 " -q --quiet Do not show status information\n"
197 " -D --directory=PATH Root directory for the container\n"
198 " --template=PATH Initialize root directory from template directory,\n"
200 " -x --ephemeral Run container with snapshot of root directory, and\n"
201 " remove it after exit\n"
202 " -i --image=PATH File system device or disk image for the container\n"
203 " -b --boot Boot up full system (i.e. invoke init)\n"
204 " -u --user=USER Run the command under specified user or uid\n"
205 " -M --machine=NAME Set the machine name for the container\n"
206 " --uuid=UUID Set a specific machine UUID for the container\n"
207 " -S --slice=SLICE Place the container in the specified slice\n"
208 " --private-network Disable network in container\n"
209 " --network-interface=INTERFACE\n"
210 " Assign an existing network interface to the\n"
212 " --network-macvlan=INTERFACE\n"
213 " Create a macvlan network interface based on an\n"
214 " existing network interface to the container\n"
215 " --network-ipvlan=INTERFACE\n"
216 " Create a ipvlan network interface based on an\n"
217 " existing network interface to the container\n"
218 " -n --network-veth Add a virtual ethernet connection between host\n"
220 " --network-bridge=INTERFACE\n"
221 " Add a virtual ethernet connection between host\n"
222 " and container and add it to an existing bridge on\n"
224 " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n"
225 " Expose a container IP port on the host\n"
226 " -Z --selinux-context=SECLABEL\n"
227 " Set the SELinux security context to be used by\n"
228 " processes in the container\n"
229 " -L --selinux-apifs-context=SECLABEL\n"
230 " Set the SELinux security context to be used by\n"
231 " API/tmpfs file systems in the container\n"
232 " --capability=CAP In addition to the default, retain specified\n"
234 " --drop-capability=CAP Drop the specified capability from the default set\n"
235 " --link-journal=MODE Link up guest journal, one of no, auto, guest, host,\n"
236 " try-guest, try-host\n"
237 " -j Equivalent to --link-journal=try-guest\n"
238 " --read-only Mount the root directory read-only\n"
239 " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n"
241 " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n"
242 " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
243 " --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
244 " --share-system Share system namespaces with host\n"
245 " --register=BOOLEAN Register container as machine\n"
246 " --keep-unit Do not register a scope for the machine, reuse\n"
247 " the service unit nspawn is running in\n"
248 " --volatile[=MODE] Run the system in volatile mode\n"
249 , program_invocation_short_name);
252 static int set_sanitized_path(char **b, const char *path) {
258 p = canonicalize_file_name(path);
263 p = path_make_absolute_cwd(path);
269 *b = path_kill_slashes(p);
273 static int parse_argv(int argc, char *argv[]) {
290 ARG_NETWORK_INTERFACE,
299 static const struct option options[] = {
300 { "help", no_argument, NULL, 'h' },
301 { "version", no_argument, NULL, ARG_VERSION },
302 { "directory", required_argument, NULL, 'D' },
303 { "template", required_argument, NULL, ARG_TEMPLATE },
304 { "ephemeral", no_argument, NULL, 'x' },
305 { "user", required_argument, NULL, 'u' },
306 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
307 { "boot", no_argument, NULL, 'b' },
308 { "uuid", required_argument, NULL, ARG_UUID },
309 { "read-only", no_argument, NULL, ARG_READ_ONLY },
310 { "capability", required_argument, NULL, ARG_CAPABILITY },
311 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
312 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
313 { "bind", required_argument, NULL, ARG_BIND },
314 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
315 { "tmpfs", required_argument, NULL, ARG_TMPFS },
316 { "machine", required_argument, NULL, 'M' },
317 { "slice", required_argument, NULL, 'S' },
318 { "setenv", required_argument, NULL, ARG_SETENV },
319 { "selinux-context", required_argument, NULL, 'Z' },
320 { "selinux-apifs-context", required_argument, NULL, 'L' },
321 { "quiet", no_argument, NULL, 'q' },
322 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM },
323 { "register", required_argument, NULL, ARG_REGISTER },
324 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
325 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
326 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
327 { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN },
328 { "network-veth", no_argument, NULL, 'n' },
329 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
330 { "personality", required_argument, NULL, ARG_PERSONALITY },
331 { "image", required_argument, NULL, 'i' },
332 { "volatile", optional_argument, NULL, ARG_VOLATILE },
333 { "port", required_argument, NULL, 'p' },
338 uint64_t plus = 0, minus = 0;
343 while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:xp:n", options, NULL)) >= 0)
352 puts(PACKAGE_STRING);
353 puts(SYSTEMD_FEATURES);
357 r = set_sanitized_path(&arg_directory, optarg);
359 return log_error_errno(r, "Invalid root directory: %m");
364 r = set_sanitized_path(&arg_template, optarg);
366 return log_error_errno(r, "Invalid template directory: %m");
371 r = set_sanitized_path(&arg_image, optarg);
373 return log_error_errno(r, "Invalid image path: %m");
378 arg_ephemeral = true;
383 arg_user = strdup(optarg);
389 case ARG_NETWORK_BRIDGE:
390 arg_network_bridge = optarg;
395 arg_network_veth = true;
396 arg_private_network = true;
399 case ARG_NETWORK_INTERFACE:
400 if (strv_extend(&arg_network_interfaces, optarg) < 0)
403 arg_private_network = true;
406 case ARG_NETWORK_MACVLAN:
407 if (strv_extend(&arg_network_macvlan, optarg) < 0)
410 arg_private_network = true;
413 case ARG_NETWORK_IPVLAN:
414 if (strv_extend(&arg_network_ipvlan, optarg) < 0)
419 case ARG_PRIVATE_NETWORK:
420 arg_private_network = true;
428 r = sd_id128_from_string(optarg, &arg_uuid);
430 log_error("Invalid UUID: %s", optarg);
440 if (isempty(optarg)) {
444 if (!machine_name_is_valid(optarg)) {
445 log_error("Invalid machine name: %s", optarg);
449 r = free_and_strdup(&arg_machine, optarg);
457 arg_selinux_context = optarg;
461 arg_selinux_apifs_context = optarg;
465 arg_read_only = true;
469 case ARG_DROP_CAPABILITY: {
470 const char *state, *word;
473 FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
474 _cleanup_free_ char *t;
476 t = strndup(word, length);
480 if (streq(t, "all")) {
481 if (c == ARG_CAPABILITY)
482 plus = (uint64_t) -1;
484 minus = (uint64_t) -1;
488 cap = capability_from_name(t);
490 log_error("Failed to parse capability %s.", t);
494 if (c == ARG_CAPABILITY)
495 plus |= 1ULL << (uint64_t) cap;
497 minus |= 1ULL << (uint64_t) cap;
505 arg_link_journal = LINK_GUEST;
506 arg_link_journal_try = true;
509 case ARG_LINK_JOURNAL:
510 if (streq(optarg, "auto")) {
511 arg_link_journal = LINK_AUTO;
512 arg_link_journal_try = false;
513 } else if (streq(optarg, "no")) {
514 arg_link_journal = LINK_NO;
515 arg_link_journal_try = false;
516 } else if (streq(optarg, "guest")) {
517 arg_link_journal = LINK_GUEST;
518 arg_link_journal_try = false;
519 } else if (streq(optarg, "host")) {
520 arg_link_journal = LINK_HOST;
521 arg_link_journal_try = false;
522 } else if (streq(optarg, "try-guest")) {
523 arg_link_journal = LINK_GUEST;
524 arg_link_journal_try = true;
525 } else if (streq(optarg, "try-host")) {
526 arg_link_journal = LINK_HOST;
527 arg_link_journal_try = true;
529 log_error("Failed to parse link journal mode %s", optarg);
537 _cleanup_free_ char *a = NULL, *b = NULL;
541 x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
543 e = strchr(optarg, ':');
545 a = strndup(optarg, e - optarg);
555 if (!path_is_absolute(a) || !path_is_absolute(b)) {
556 log_error("Invalid bind mount specification: %s", optarg);
560 r = strv_extend(x, a);
564 r = strv_extend(x, b);
572 _cleanup_free_ char *a = NULL, *b = NULL;
575 e = strchr(optarg, ':');
577 a = strndup(optarg, e - optarg);
581 b = strdup("mode=0755");
587 if (!path_is_absolute(a)) {
588 log_error("Invalid tmpfs specification: %s", optarg);
592 r = strv_push(&arg_tmpfs, a);
598 r = strv_push(&arg_tmpfs, b);
610 if (!env_assignment_is_valid(optarg)) {
611 log_error("Environment variable assignment '%s' is not valid.", optarg);
615 n = strv_env_set(arg_setenv, optarg);
619 strv_free(arg_setenv);
628 case ARG_SHARE_SYSTEM:
629 arg_share_system = true;
633 r = parse_boolean(optarg);
635 log_error("Failed to parse --register= argument: %s", optarg);
643 arg_keep_unit = true;
646 case ARG_PERSONALITY:
648 arg_personality = personality_from_string(optarg);
649 if (arg_personality == 0xffffffffLU) {
650 log_error("Unknown or unsupported personality '%s'.", optarg);
659 arg_volatile = VOLATILE_YES;
661 r = parse_boolean(optarg);
663 if (streq(optarg, "state"))
664 arg_volatile = VOLATILE_STATE;
666 log_error("Failed to parse --volatile= argument: %s", optarg);
670 arg_volatile = r ? VOLATILE_YES : VOLATILE_NO;
676 const char *split, *e;
677 uint16_t container_port, host_port;
681 if ((e = startswith(optarg, "tcp:")))
682 protocol = IPPROTO_TCP;
683 else if ((e = startswith(optarg, "udp:")))
684 protocol = IPPROTO_UDP;
687 protocol = IPPROTO_TCP;
690 split = strchr(e, ':');
692 char v[split - e + 1];
694 memcpy(v, e, split - e);
697 r = safe_atou16(v, &host_port);
698 if (r < 0 || host_port <= 0) {
699 log_error("Failed to parse host port: %s", optarg);
703 r = safe_atou16(split + 1, &container_port);
705 r = safe_atou16(e, &container_port);
706 host_port = container_port;
709 if (r < 0 || container_port <= 0) {
710 log_error("Failed to parse host port: %s", optarg);
714 LIST_FOREACH(ports, p, arg_expose_ports) {
715 if (p->protocol == protocol && p->host_port == host_port) {
716 log_error("Duplicate port specification: %s", optarg);
721 p = new(ExposePort, 1);
725 p->protocol = protocol;
726 p->host_port = host_port;
727 p->container_port = container_port;
729 LIST_PREPEND(ports, arg_expose_ports, p);
738 assert_not_reached("Unhandled option");
741 if (arg_share_system)
742 arg_register = false;
744 if (arg_boot && arg_share_system) {
745 log_error("--boot and --share-system may not be combined.");
749 if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) {
750 log_error("--keep-unit may not be used when invoked from a user session.");
754 if (arg_directory && arg_image) {
755 log_error("--directory= and --image= may not be combined.");
759 if (arg_template && arg_image) {
760 log_error("--template= and --image= may not be combined.");
764 if (arg_template && !(arg_directory || arg_machine)) {
765 log_error("--template= needs --directory= or --machine=.");
769 if (arg_ephemeral && arg_template) {
770 log_error("--ephemeral and --template= may not be combined.");
774 if (arg_ephemeral && arg_image) {
775 log_error("--ephemeral and --image= may not be combined.");
779 if (arg_ephemeral && !IN_SET(arg_link_journal, LINK_NO, LINK_AUTO)) {
780 log_error("--ephemeral and --link-journal= may not be combined.");
784 if (arg_volatile != VOLATILE_NO && arg_read_only) {
785 log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy.");
789 if (arg_expose_ports && !arg_private_network) {
790 log_error("Cannot use --port= without private networking.");
794 arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus;
799 static int mount_all(const char *dest) {
801 typedef struct MountPoint {
810 static const MountPoint mount_table[] = {
811 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
812 { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */
813 { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
814 { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
815 { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
816 { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true },
817 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
818 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
819 { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true },
821 { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
822 { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
829 for (k = 0; k < ELEMENTSOF(mount_table); k++) {
830 _cleanup_free_ char *where = NULL;
832 _cleanup_free_ char *options = NULL;
837 where = strjoin(dest, "/", mount_table[k].where, NULL);
841 t = path_is_mount_point(where, true);
843 log_error_errno(t, "Failed to detect whether %s is a mount point: %m", where);
851 /* Skip this entry if it is not a remount. */
852 if (mount_table[k].what && t > 0)
855 t = mkdir_p(where, 0755);
857 if (mount_table[k].fatal) {
858 log_error_errno(t, "Failed to create directory %s: %m", where);
863 log_warning_errno(t, "Failed to create directory %s: %m", where);
869 if (arg_selinux_apifs_context &&
870 (streq_ptr(mount_table[k].what, "tmpfs") || streq_ptr(mount_table[k].what, "devpts"))) {
871 options = strjoin(mount_table[k].options, ",context=\"", arg_selinux_apifs_context, "\"", NULL);
878 o = mount_table[k].options;
881 if (mount(mount_table[k].what,
884 mount_table[k].flags,
887 if (mount_table[k].fatal) {
888 log_error_errno(errno, "mount(%s) failed: %m", where);
893 log_warning_errno(errno, "mount(%s) failed: %m", where);
900 static int mount_binds(const char *dest, char **l, bool ro) {
903 STRV_FOREACH_PAIR(x, y, l) {
904 _cleanup_free_ char *where = NULL;
905 struct stat source_st, dest_st;
908 if (stat(*x, &source_st) < 0)
909 return log_error_errno(errno, "Failed to stat %s: %m", *x);
911 where = strappend(dest, *y);
915 r = stat(where, &dest_st);
917 if (S_ISDIR(source_st.st_mode) && !S_ISDIR(dest_st.st_mode)) {
918 log_error("Cannot bind mount directory %s on file %s.", *x, where);
921 if (!S_ISDIR(source_st.st_mode) && S_ISDIR(dest_st.st_mode)) {
922 log_error("Cannot bind mount file %s on directory %s.", *x, where);
925 } else if (errno == ENOENT) {
926 r = mkdir_parents_label(where, 0755);
928 return log_error_errno(r, "Failed to bind mount %s: %m", *x);
930 log_error_errno(errno, "Failed to bind mount %s: %m", *x);
934 /* Create the mount point. Any non-directory file can be
935 * mounted on any non-directory file (regular, fifo, socket,
938 if (S_ISDIR(source_st.st_mode)) {
939 r = mkdir_label(where, 0755);
940 if (r < 0 && errno != EEXIST)
941 return log_error_errno(r, "Failed to create mount point %s: %m", where);
945 return log_error_errno(r, "Failed to create mount point %s: %m", where);
948 if (mount(*x, where, "bind", MS_BIND, NULL) < 0)
949 return log_error_errno(errno, "mount(%s) failed: %m", where);
952 r = bind_remount_recursive(where, true);
954 return log_error_errno(r, "Read-Only bind mount failed: %m");
961 static int mount_cgroup_hierarchy(const char *dest, const char *controller, const char *hierarchy, bool read_only) {
965 to = strappenda(dest, "/sys/fs/cgroup/", hierarchy);
967 r = path_is_mount_point(to, false);
969 return log_error_errno(r, "Failed to determine if %s is mounted already: %m", to);
975 if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV|(read_only ? MS_RDONLY : 0), controller) < 0)
976 return log_error_errno(errno, "Failed to mount to %s: %m", to);
981 static int mount_cgroup(const char *dest) {
982 _cleanup_set_free_free_ Set *controllers = NULL;
983 _cleanup_free_ char *own_cgroup_path = NULL;
984 const char *cgroup_root, *systemd_root, *systemd_own;
987 controllers = set_new(&string_hash_ops);
991 r = cg_kernel_controllers(controllers);
993 return log_error_errno(r, "Failed to determine cgroup controllers: %m");
995 r = cg_pid_get_path(NULL, 0, &own_cgroup_path);
997 return log_error_errno(r, "Failed to determine our own cgroup path: %m");
999 cgroup_root = strappenda(dest, "/sys/fs/cgroup");
1000 if (mount("tmpfs", cgroup_root, "tmpfs", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, "mode=755") < 0)
1001 return log_error_errno(errno, "Failed to mount tmpfs to /sys/fs/cgroup: %m");
1004 _cleanup_free_ char *controller = NULL, *origin = NULL, *combined = NULL;
1006 controller = set_steal_first(controllers);
1010 origin = strappend("/sys/fs/cgroup/", controller);
1014 r = readlink_malloc(origin, &combined);
1016 /* Not a symbolic link, but directly a single cgroup hierarchy */
1018 r = mount_cgroup_hierarchy(dest, controller, controller, true);
1023 return log_error_errno(r, "Failed to read link %s: %m", origin);
1025 _cleanup_free_ char *target = NULL;
1027 target = strjoin(dest, "/sys/fs/cgroup/", controller, NULL);
1031 /* A symbolic link, a combination of controllers in one hierarchy */
1033 if (!filename_is_valid(combined)) {
1034 log_warning("Ignoring invalid combined hierarchy %s.", combined);
1038 r = mount_cgroup_hierarchy(dest, combined, combined, true);
1042 if (symlink(combined, target) < 0)
1043 return log_error_errno(errno, "Failed to create symlink for combined hierarchy: %m");
1047 r = mount_cgroup_hierarchy(dest, "name=systemd", "systemd", false);
1051 /* Make our own cgroup a (writable) bind mount */
1052 systemd_own = strappenda(dest, "/sys/fs/cgroup/systemd", own_cgroup_path);
1053 if (mount(systemd_own, systemd_own, NULL, MS_BIND, NULL) < 0)
1054 return log_error_errno(errno, "Failed to turn %s into a bind mount: %m", own_cgroup_path);
1056 /* And then remount the systemd cgroup root read-only */
1057 systemd_root = strappenda(dest, "/sys/fs/cgroup/systemd");
1058 if (mount(NULL, systemd_root, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0)
1059 return log_error_errno(errno, "Failed to mount cgroup root read-only: %m");
1061 if (mount(NULL, cgroup_root, NULL, MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755") < 0)
1062 return log_error_errno(errno, "Failed to remount %s read-only: %m", cgroup_root);
1067 static int mount_tmpfs(const char *dest) {
1070 STRV_FOREACH_PAIR(i, o, arg_tmpfs) {
1071 _cleanup_free_ char *where = NULL;
1074 where = strappend(dest, *i);
1078 r = mkdir_label(where, 0755);
1079 if (r < 0 && r != -EEXIST)
1080 return log_error_errno(r, "Creating mount point for tmpfs %s failed: %m", where);
1082 if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, *o) < 0)
1083 return log_error_errno(errno, "tmpfs mount to %s failed: %m", where);
1089 static int setup_timezone(const char *dest) {
1090 _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
1096 /* Fix the timezone, if possible */
1097 r = readlink_malloc("/etc/localtime", &p);
1099 log_warning("/etc/localtime is not a symlink, not updating container timezone.");
1103 z = path_startswith(p, "../usr/share/zoneinfo/");
1105 z = path_startswith(p, "/usr/share/zoneinfo/");
1107 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
1111 where = strappend(dest, "/etc/localtime");
1115 r = readlink_malloc(where, &q);
1117 y = path_startswith(q, "../usr/share/zoneinfo/");
1119 y = path_startswith(q, "/usr/share/zoneinfo/");
1121 /* Already pointing to the right place? Then do nothing .. */
1122 if (y && streq(y, z))
1126 check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
1130 if (access(check, F_OK) < 0) {
1131 log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
1135 what = strappend("../usr/share/zoneinfo/", z);
1139 r = mkdir_parents(where, 0755);
1141 log_error_errno(r, "Failed to create directory for timezone info %s in container: %m", where);
1147 if (r < 0 && errno != ENOENT) {
1148 log_error_errno(errno, "Failed to remove existing timezone info %s in container: %m", where);
1153 if (symlink(what, where) < 0) {
1154 log_error_errno(errno, "Failed to correct timezone of container: %m");
1161 static int setup_resolv_conf(const char *dest) {
1162 _cleanup_free_ char *where = NULL;
1167 if (arg_private_network)
1170 /* Fix resolv.conf, if possible */
1171 where = strappend(dest, "/etc/resolv.conf");
1175 /* We don't really care for the results of this really. If it
1176 * fails, it fails, but meh... */
1177 r = mkdir_parents(where, 0755);
1179 log_warning_errno(r, "Failed to create parent directory for resolv.conf %s: %m", where);
1184 r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0);
1186 log_warning_errno(r, "Failed to copy /etc/resolv.conf to %s: %m", where);
1194 static int setup_volatile_state(const char *directory) {
1200 if (arg_volatile != VOLATILE_STATE)
1203 /* --volatile=state means we simply overmount /var
1204 with a tmpfs, and the rest read-only. */
1206 r = bind_remount_recursive(directory, true);
1208 return log_error_errno(r, "Failed to remount %s read-only: %m", directory);
1210 p = strappenda(directory, "/var");
1212 if (r < 0 && errno != EEXIST)
1213 return log_error_errno(errno, "Failed to create %s: %m", directory);
1215 if (mount("tmpfs", p, "tmpfs", MS_STRICTATIME, "mode=755") < 0)
1216 return log_error_errno(errno, "Failed to mount tmpfs to /var: %m");
1221 static int setup_volatile(const char *directory) {
1222 bool tmpfs_mounted = false, bind_mounted = false;
1223 char template[] = "/tmp/nspawn-volatile-XXXXXX";
1229 if (arg_volatile != VOLATILE_YES)
1232 /* --volatile=yes means we mount a tmpfs to the root dir, and
1233 the original /usr to use inside it, and that read-only. */
1235 if (!mkdtemp(template))
1236 return log_error_errno(errno, "Failed to create temporary directory: %m");
1238 if (mount("tmpfs", template, "tmpfs", MS_STRICTATIME, "mode=755") < 0) {
1239 log_error_errno(errno, "Failed to mount tmpfs for root directory: %m");
1244 tmpfs_mounted = true;
1246 f = strappenda(directory, "/usr");
1247 t = strappenda(template, "/usr");
1250 if (r < 0 && errno != EEXIST) {
1251 log_error_errno(errno, "Failed to create %s: %m", t);
1256 if (mount(f, t, "bind", MS_BIND|MS_REC, NULL) < 0) {
1257 log_error_errno(errno, "Failed to create /usr bind mount: %m");
1262 bind_mounted = true;
1264 r = bind_remount_recursive(t, true);
1266 log_error_errno(r, "Failed to remount %s read-only: %m", t);
1270 if (mount(template, directory, NULL, MS_MOVE, NULL) < 0) {
1271 log_error_errno(errno, "Failed to move root mount: %m");
1289 static char* id128_format_as_uuid(sd_id128_t id, char s[37]) {
1292 "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
1293 SD_ID128_FORMAT_VAL(id));
1298 static int setup_boot_id(const char *dest) {
1299 _cleanup_free_ char *from = NULL, *to = NULL;
1300 sd_id128_t rnd = {};
1306 if (arg_share_system)
1309 /* Generate a new randomized boot ID, so that each boot-up of
1310 * the container gets a new one */
1312 from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
1313 to = strappend(dest, "/proc/sys/kernel/random/boot_id");
1317 r = sd_id128_randomize(&rnd);
1319 return log_error_errno(r, "Failed to generate random boot id: %m");
1321 id128_format_as_uuid(rnd, as_uuid);
1323 r = write_string_file(from, as_uuid);
1325 return log_error_errno(r, "Failed to write boot id: %m");
1327 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
1328 log_error_errno(errno, "Failed to bind mount boot id: %m");
1330 } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL))
1331 log_warning_errno(errno, "Failed to make boot id read-only: %m");
1337 static int copy_devnodes(const char *dest) {
1339 static const char devnodes[] =
1350 _cleanup_umask_ mode_t u;
1356 NULSTR_FOREACH(d, devnodes) {
1357 _cleanup_free_ char *from = NULL, *to = NULL;
1360 from = strappend("/dev/", d);
1361 to = strjoin(dest, "/dev/", d, NULL);
1365 if (stat(from, &st) < 0) {
1367 if (errno != ENOENT)
1368 return log_error_errno(errno, "Failed to stat %s: %m", from);
1370 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
1372 log_error("%s is not a char or block device, cannot copy", from);
1376 r = mkdir_parents(to, 0775);
1378 log_error_errno(r, "Failed to create parent directory of %s: %m", to);
1382 if (mknod(to, st.st_mode, st.st_rdev) < 0)
1383 return log_error_errno(errno, "mknod(%s) failed: %m", to);
1390 static int setup_ptmx(const char *dest) {
1391 _cleanup_free_ char *p = NULL;
1393 p = strappend(dest, "/dev/ptmx");
1397 if (symlink("pts/ptmx", p) < 0)
1398 return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
1403 static int setup_dev_console(const char *dest, const char *console) {
1404 _cleanup_umask_ mode_t u;
1414 if (stat("/dev/null", &st) < 0)
1415 return log_error_errno(errno, "Failed to stat /dev/null: %m");
1417 r = chmod_and_chown(console, 0600, 0, 0);
1419 return log_error_errno(r, "Failed to correct access mode for TTY: %m");
1421 /* We need to bind mount the right tty to /dev/console since
1422 * ptys can only exist on pts file systems. To have something
1423 * to bind mount things on we create a device node first, and
1424 * use /dev/null for that since we the cgroups device policy
1425 * allows us to create that freely, while we cannot create
1426 * /dev/console. (Note that the major minor doesn't actually
1427 * matter here, since we mount it over anyway). */
1429 to = strappenda(dest, "/dev/console");
1430 if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0)
1431 return log_error_errno(errno, "mknod() for /dev/console failed: %m");
1433 if (mount(console, to, "bind", MS_BIND, NULL) < 0)
1434 return log_error_errno(errno, "Bind mount for /dev/console failed: %m");
1439 static int setup_kmsg(const char *dest, int kmsg_socket) {
1440 _cleanup_free_ char *from = NULL, *to = NULL;
1441 _cleanup_umask_ mode_t u;
1444 struct cmsghdr cmsghdr;
1445 uint8_t buf[CMSG_SPACE(sizeof(int))];
1447 struct msghdr mh = {
1448 .msg_control = &control,
1449 .msg_controllen = sizeof(control),
1451 struct cmsghdr *cmsg;
1454 assert(kmsg_socket >= 0);
1458 /* We create the kmsg FIFO as /dev/kmsg, but immediately
1459 * delete it after bind mounting it to /proc/kmsg. While FIFOs
1460 * on the reading side behave very similar to /proc/kmsg,
1461 * their writing side behaves differently from /dev/kmsg in
1462 * that writing blocks when nothing is reading. In order to
1463 * avoid any problems with containers deadlocking due to this
1464 * we simply make /dev/kmsg unavailable to the container. */
1465 if (asprintf(&from, "%s/dev/kmsg", dest) < 0 ||
1466 asprintf(&to, "%s/proc/kmsg", dest) < 0)
1469 if (mkfifo(from, 0600) < 0)
1470 return log_error_errno(errno, "mkfifo() for /dev/kmsg failed: %m");
1472 r = chmod_and_chown(from, 0600, 0, 0);
1474 return log_error_errno(r, "Failed to correct access mode for /dev/kmsg: %m");
1476 if (mount(from, to, "bind", MS_BIND, NULL) < 0)
1477 return log_error_errno(errno, "Bind mount for /proc/kmsg failed: %m");
1479 fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
1481 return log_error_errno(errno, "Failed to open fifo: %m");
1483 cmsg = CMSG_FIRSTHDR(&mh);
1484 cmsg->cmsg_level = SOL_SOCKET;
1485 cmsg->cmsg_type = SCM_RIGHTS;
1486 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
1487 memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
1489 mh.msg_controllen = cmsg->cmsg_len;
1491 /* Store away the fd in the socket, so that it stays open as
1492 * long as we run the child */
1493 k = sendmsg(kmsg_socket, &mh, MSG_NOSIGNAL);
1497 return log_error_errno(errno, "Failed to send FIFO fd: %m");
1499 /* And now make the FIFO unavailable as /dev/kmsg... */
1504 static int send_rtnl(int send_fd) {
1506 struct cmsghdr cmsghdr;
1507 uint8_t buf[CMSG_SPACE(sizeof(int))];
1509 struct msghdr mh = {
1510 .msg_control = &control,
1511 .msg_controllen = sizeof(control),
1513 struct cmsghdr *cmsg;
1514 _cleanup_close_ int fd = -1;
1517 assert(send_fd >= 0);
1519 if (!arg_expose_ports)
1522 fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, NETLINK_ROUTE);
1524 return log_error_errno(errno, "failed to allocate container netlink: %m");
1526 cmsg = CMSG_FIRSTHDR(&mh);
1527 cmsg->cmsg_level = SOL_SOCKET;
1528 cmsg->cmsg_type = SCM_RIGHTS;
1529 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
1530 memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
1532 mh.msg_controllen = cmsg->cmsg_len;
1534 /* Store away the fd in the socket, so that it stays open as
1535 * long as we run the child */
1536 k = sendmsg(send_fd, &mh, MSG_NOSIGNAL);
1538 return log_error_errno(errno, "Failed to send netlink fd: %m");
1543 static int flush_ports(union in_addr_union *exposed) {
1545 int r, af = AF_INET;
1549 if (!arg_expose_ports)
1552 if (in_addr_is_null(af, exposed))
1555 log_debug("Lost IP address.");
1557 LIST_FOREACH(ports, p, arg_expose_ports) {
1558 r = fw_add_local_dnat(false,
1569 log_warning_errno(r, "Failed to modify firewall: %m");
1572 *exposed = IN_ADDR_NULL;
1576 static int expose_ports(sd_rtnl *rtnl, union in_addr_union *exposed) {
1577 _cleanup_free_ struct local_address *addresses = NULL;
1578 _cleanup_free_ char *pretty = NULL;
1579 union in_addr_union new_exposed;
1582 int af = AF_INET, r;
1586 /* Invoked each time an address is added or removed inside the
1589 if (!arg_expose_ports)
1592 r = local_addresses(rtnl, 0, af, &addresses);
1594 return log_error_errno(r, "Failed to enumerate local addresses: %m");
1597 addresses[0].family == af &&
1598 addresses[0].scope < RT_SCOPE_LINK;
1601 return flush_ports(exposed);
1603 new_exposed = addresses[0].address;
1604 if (in_addr_equal(af, exposed, &new_exposed))
1607 in_addr_to_string(af, &new_exposed, &pretty);
1608 log_debug("New container IP is %s.", strna(pretty));
1610 LIST_FOREACH(ports, p, arg_expose_ports) {
1612 r = fw_add_local_dnat(true,
1621 in_addr_is_null(af, exposed) ? NULL : exposed);
1623 log_warning_errno(r, "Failed to modify firewall: %m");
1626 *exposed = new_exposed;
1630 static int on_address_change(sd_rtnl *rtnl, sd_rtnl_message *m, void *userdata) {
1631 union in_addr_union *exposed = userdata;
1637 expose_ports(rtnl, exposed);
1641 static int watch_rtnl(sd_event *event, int recv_fd, union in_addr_union *exposed, sd_rtnl **ret) {
1643 struct cmsghdr cmsghdr;
1644 uint8_t buf[CMSG_SPACE(sizeof(int))];
1646 struct msghdr mh = {
1647 .msg_control = &control,
1648 .msg_controllen = sizeof(control),
1650 struct cmsghdr *cmsg;
1651 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1656 assert(recv_fd >= 0);
1659 if (!arg_expose_ports)
1662 k = recvmsg(recv_fd, &mh, MSG_NOSIGNAL);
1664 return log_error_errno(errno, "Failed to recv netlink fd: %m");
1666 cmsg = CMSG_FIRSTHDR(&mh);
1667 assert(cmsg->cmsg_level == SOL_SOCKET);
1668 assert(cmsg->cmsg_type == SCM_RIGHTS);
1669 assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int)));
1670 memcpy(&fd, CMSG_DATA(cmsg), sizeof(int));
1672 r = sd_rtnl_open_fd(&rtnl, fd, 1, RTNLGRP_IPV4_IFADDR);
1675 return log_error_errno(r, "Failed to create rtnl object: %m");
1678 r = sd_rtnl_add_match(rtnl, RTM_NEWADDR, on_address_change, exposed);
1680 return log_error_errno(r, "Failed to subscribe to RTM_NEWADDR messages: %m");
1682 r = sd_rtnl_add_match(rtnl, RTM_DELADDR, on_address_change, exposed);
1684 return log_error_errno(r, "Failed to subscribe to RTM_DELADDR messages: %m");
1686 r = sd_rtnl_attach_event(rtnl, event, 0);
1688 return log_error_errno(r, "Failed to add to even loop: %m");
1696 static int setup_hostname(void) {
1698 if (arg_share_system)
1701 if (sethostname_idempotent(arg_machine) < 0)
1707 static int setup_journal(const char *directory) {
1708 sd_id128_t machine_id, this_id;
1709 _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL;
1713 /* Don't link journals in ephemeral mode */
1717 p = strappend(directory, "/etc/machine-id");
1721 r = read_one_line_file(p, &b);
1722 if (r == -ENOENT && arg_link_journal == LINK_AUTO)
1725 return log_error_errno(r, "Failed to read machine ID from %s: %m", p);
1728 if (isempty(id) && arg_link_journal == LINK_AUTO)
1731 /* Verify validity */
1732 r = sd_id128_from_string(id, &machine_id);
1734 return log_error_errno(r, "Failed to parse machine ID from %s: %m", p);
1736 r = sd_id128_get_machine(&this_id);
1738 return log_error_errno(r, "Failed to retrieve machine ID: %m");
1740 if (sd_id128_equal(machine_id, this_id)) {
1741 log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
1742 "Host and machine ids are equal (%s): refusing to link journals", id);
1743 if (arg_link_journal == LINK_AUTO)
1748 if (arg_link_journal == LINK_NO)
1752 p = strappend("/var/log/journal/", id);
1753 q = strjoin(directory, "/var/log/journal/", id, NULL);
1757 if (path_is_mount_point(p, false) > 0) {
1758 if (arg_link_journal != LINK_AUTO) {
1759 log_error("%s: already a mount point, refusing to use for journal", p);
1766 if (path_is_mount_point(q, false) > 0) {
1767 if (arg_link_journal != LINK_AUTO) {
1768 log_error("%s: already a mount point, refusing to use for journal", q);
1775 r = readlink_and_make_absolute(p, &d);
1777 if ((arg_link_journal == LINK_GUEST ||
1778 arg_link_journal == LINK_AUTO) &&
1781 r = mkdir_p(q, 0755);
1783 log_warning_errno(errno, "Failed to create directory %s: %m", q);
1788 return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
1789 } else if (r == -EINVAL) {
1791 if (arg_link_journal == LINK_GUEST &&
1794 if (errno == ENOTDIR) {
1795 log_error("%s already exists and is neither a symlink nor a directory", p);
1798 log_error_errno(errno, "Failed to remove %s: %m", p);
1802 } else if (r != -ENOENT) {
1803 log_error_errno(errno, "readlink(%s) failed: %m", p);
1807 if (arg_link_journal == LINK_GUEST) {
1809 if (symlink(q, p) < 0) {
1810 if (arg_link_journal_try) {
1811 log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
1814 log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
1819 r = mkdir_p(q, 0755);
1821 log_warning_errno(errno, "Failed to create directory %s: %m", q);
1825 if (arg_link_journal == LINK_HOST) {
1826 /* don't create parents here -- if the host doesn't have
1827 * permanent journal set up, don't force it here */
1830 if (arg_link_journal_try) {
1831 log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p);
1834 log_error_errno(errno, "Failed to create %s: %m", p);
1839 } else if (access(p, F_OK) < 0)
1842 if (dir_is_empty(q) == 0)
1843 log_warning("%s is not empty, proceeding anyway.", q);
1845 r = mkdir_p(q, 0755);
1847 log_error_errno(errno, "Failed to create %s: %m", q);
1851 if (mount(p, q, "bind", MS_BIND, NULL) < 0)
1852 return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
1857 static int drop_capabilities(void) {
1858 return capability_bounding_set_drop(~arg_retain, false);
1861 static int register_machine(pid_t pid, int local_ifindex) {
1862 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1863 _cleanup_bus_close_unref_ sd_bus *bus = NULL;
1869 r = sd_bus_default_system(&bus);
1871 return log_error_errno(r, "Failed to open system bus: %m");
1873 if (arg_keep_unit) {
1874 r = sd_bus_call_method(
1876 "org.freedesktop.machine1",
1877 "/org/freedesktop/machine1",
1878 "org.freedesktop.machine1.Manager",
1879 "RegisterMachineWithNetwork",
1884 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1888 strempty(arg_directory),
1889 local_ifindex > 0 ? 1 : 0, local_ifindex);
1891 _cleanup_bus_message_unref_ sd_bus_message *m = NULL;
1893 r = sd_bus_message_new_method_call(
1896 "org.freedesktop.machine1",
1897 "/org/freedesktop/machine1",
1898 "org.freedesktop.machine1.Manager",
1899 "CreateMachineWithNetwork");
1901 return log_error_errno(r, "Failed to create message: %m");
1903 r = sd_bus_message_append(
1907 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1911 strempty(arg_directory),
1912 local_ifindex > 0 ? 1 : 0, local_ifindex);
1914 return log_error_errno(r, "Failed to append message arguments: %m");
1916 r = sd_bus_message_open_container(m, 'a', "(sv)");
1918 return log_error_errno(r, "Failed to open container: %m");
1920 if (!isempty(arg_slice)) {
1921 r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice);
1923 return log_error_errno(r, "Failed to append slice: %m");
1926 r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict");
1928 return log_error_errno(r, "Failed to add device policy: %m");
1930 r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 9,
1931 /* Allow the container to
1932 * access and create the API
1933 * device nodes, so that
1934 * PrivateDevices= in the
1935 * container can work
1940 "/dev/random", "rwm",
1941 "/dev/urandom", "rwm",
1943 "/dev/net/tun", "rwm",
1944 /* Allow the container
1945 * access to ptys. However,
1947 * container to ever create
1948 * these device nodes. */
1949 "/dev/pts/ptmx", "rw",
1952 return log_error_errno(r, "Failed to add device whitelist: %m");
1954 r = sd_bus_message_close_container(m);
1956 return log_error_errno(r, "Failed to close container: %m");
1958 r = sd_bus_call(bus, m, 0, &error, NULL);
1962 log_error("Failed to register machine: %s", bus_error_message(&error, r));
1969 static int terminate_machine(pid_t pid) {
1970 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1971 _cleanup_bus_message_unref_ sd_bus_message *reply = NULL;
1972 _cleanup_bus_close_unref_ sd_bus *bus = NULL;
1979 r = sd_bus_default_system(&bus);
1981 return log_error_errno(r, "Failed to open system bus: %m");
1983 r = sd_bus_call_method(
1985 "org.freedesktop.machine1",
1986 "/org/freedesktop/machine1",
1987 "org.freedesktop.machine1.Manager",
1994 /* Note that the machine might already have been
1995 * cleaned up automatically, hence don't consider it a
1996 * failure if we cannot get the machine object. */
1997 log_debug("Failed to get machine: %s", bus_error_message(&error, r));
2001 r = sd_bus_message_read(reply, "o", &path);
2003 return bus_log_parse_error(r);
2005 r = sd_bus_call_method(
2007 "org.freedesktop.machine1",
2009 "org.freedesktop.machine1.Machine",
2015 log_debug("Failed to terminate machine: %s", bus_error_message(&error, r));
2022 static int reset_audit_loginuid(void) {
2023 _cleanup_free_ char *p = NULL;
2026 if (arg_share_system)
2029 r = read_one_line_file("/proc/self/loginuid", &p);
2033 return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
2035 /* Already reset? */
2036 if (streq(p, "4294967295"))
2039 r = write_string_file("/proc/self/loginuid", "4294967295");
2041 log_error("Failed to reset audit login UID. This probably means that your kernel is too\n"
2042 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
2043 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
2044 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
2045 "using systemd-nspawn. Sleeping for 5s... (%s)\n", strerror(-r));
2053 #define HOST_HASH_KEY SD_ID128_MAKE(1a,37,6f,c7,46,ec,45,0b,ad,a3,d5,31,06,60,5d,b1)
2054 #define CONTAINER_HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2)
2055 #define MACVLAN_HASH_KEY SD_ID128_MAKE(00,13,6d,bc,66,83,44,81,bb,0c,f9,51,1f,24,a6,6f)
2057 static int generate_mac(struct ether_addr *mac, sd_id128_t hash_key, uint64_t idx) {
2063 l = strlen(arg_machine);
2064 sz = sizeof(sd_id128_t) + l;
2070 /* fetch some persistent data unique to the host */
2071 r = sd_id128_get_machine((sd_id128_t*) v);
2075 /* combine with some data unique (on this host) to this
2076 * container instance */
2077 i = mempcpy(v + sizeof(sd_id128_t), arg_machine, l);
2080 memcpy(i, &idx, sizeof(idx));
2083 /* Let's hash the host machine ID plus the container name. We
2084 * use a fixed, but originally randomly created hash key here. */
2085 siphash24(result, v, sz, hash_key.bytes);
2087 assert_cc(ETH_ALEN <= sizeof(result));
2088 memcpy(mac->ether_addr_octet, result, ETH_ALEN);
2090 /* see eth_random_addr in the kernel */
2091 mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */
2092 mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */
2097 static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ], int *ifi) {
2098 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
2099 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
2100 struct ether_addr mac_host, mac_container;
2103 if (!arg_private_network)
2106 if (!arg_network_veth)
2109 /* Use two different interface name prefixes depending whether
2110 * we are in bridge mode or not. */
2111 snprintf(iface_name, IFNAMSIZ - 1, "%s-%s",
2112 arg_network_bridge ? "vb" : "ve", arg_machine);
2114 r = generate_mac(&mac_container, CONTAINER_HASH_KEY, 0);
2116 return log_error_errno(r, "Failed to generate predictable MAC address for container side: %m");
2118 r = generate_mac(&mac_host, HOST_HASH_KEY, 0);
2120 return log_error_errno(r, "Failed to generate predictable MAC address for host side: %m");
2122 r = sd_rtnl_open(&rtnl, 0);
2124 return log_error_errno(r, "Failed to connect to netlink: %m");
2126 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
2128 return log_error_errno(r, "Failed to allocate netlink message: %m");
2130 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name);
2132 return log_error_errno(r, "Failed to add netlink interface name: %m");
2134 r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac_host);
2136 return log_error_errno(r, "Failed to add netlink MAC address: %m");
2138 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
2140 return log_error_errno(r, "Failed to open netlink container: %m");
2142 r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth");
2144 return log_error_errno(r, "Failed to open netlink container: %m");
2146 r = sd_rtnl_message_open_container(m, VETH_INFO_PEER);
2148 return log_error_errno(r, "Failed to open netlink container: %m");
2150 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0");
2152 return log_error_errno(r, "Failed to add netlink interface name: %m");
2154 r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac_container);
2156 return log_error_errno(r, "Failed to add netlink MAC address: %m");
2158 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
2160 return log_error_errno(r, "Failed to add netlink namespace field: %m");
2162 r = sd_rtnl_message_close_container(m);
2164 return log_error_errno(r, "Failed to close netlink container: %m");
2166 r = sd_rtnl_message_close_container(m);
2168 return log_error_errno(r, "Failed to close netlink container: %m");
2170 r = sd_rtnl_message_close_container(m);
2172 return log_error_errno(r, "Failed to close netlink container: %m");
2174 r = sd_rtnl_call(rtnl, m, 0, NULL);
2176 return log_error_errno(r, "Failed to add new veth interfaces: %m");
2178 i = (int) if_nametoindex(iface_name);
2180 return log_error_errno(errno, "Failed to resolve interface %s: %m", iface_name);
2187 static int setup_bridge(const char veth_name[], int *ifi) {
2188 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
2189 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
2192 if (!arg_private_network)
2195 if (!arg_network_veth)
2198 if (!arg_network_bridge)
2201 bridge = (int) if_nametoindex(arg_network_bridge);
2203 return log_error_errno(errno, "Failed to resolve interface %s: %m", arg_network_bridge);
2207 r = sd_rtnl_open(&rtnl, 0);
2209 return log_error_errno(r, "Failed to connect to netlink: %m");
2211 r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0);
2213 return log_error_errno(r, "Failed to allocate netlink message: %m");
2215 r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP);
2217 return log_error_errno(r, "Failed to set IFF_UP flag: %m");
2219 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name);
2221 return log_error_errno(r, "Failed to add netlink interface name field: %m");
2223 r = sd_rtnl_message_append_u32(m, IFLA_MASTER, bridge);
2225 return log_error_errno(r, "Failed to add netlink master field: %m");
2227 r = sd_rtnl_call(rtnl, m, 0, NULL);
2229 return log_error_errno(r, "Failed to add veth interface to bridge: %m");
2234 static int parse_interface(struct udev *udev, const char *name) {
2235 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
2236 char ifi_str[2 + DECIMAL_STR_MAX(int)];
2239 ifi = (int) if_nametoindex(name);
2241 return log_error_errno(errno, "Failed to resolve interface %s: %m", name);
2243 sprintf(ifi_str, "n%i", ifi);
2244 d = udev_device_new_from_device_id(udev, ifi_str);
2246 return log_error_errno(errno, "Failed to get udev device for interface %s: %m", name);
2248 if (udev_device_get_is_initialized(d) <= 0) {
2249 log_error("Network interface %s is not initialized yet.", name);
2256 static int move_network_interfaces(pid_t pid) {
2257 _cleanup_udev_unref_ struct udev *udev = NULL;
2258 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
2262 if (!arg_private_network)
2265 if (strv_isempty(arg_network_interfaces))
2268 r = sd_rtnl_open(&rtnl, 0);
2270 return log_error_errno(r, "Failed to connect to netlink: %m");
2274 log_error("Failed to connect to udev.");
2278 STRV_FOREACH(i, arg_network_interfaces) {
2279 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
2282 ifi = parse_interface(udev, *i);
2286 r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, ifi);
2288 return log_error_errno(r, "Failed to allocate netlink message: %m");
2290 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
2292 return log_error_errno(r, "Failed to append namespace PID to netlink message: %m");
2294 r = sd_rtnl_call(rtnl, m, 0, NULL);
2296 return log_error_errno(r, "Failed to move interface %s to namespace: %m", *i);
2302 static int setup_macvlan(pid_t pid) {
2303 _cleanup_udev_unref_ struct udev *udev = NULL;
2304 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
2309 if (!arg_private_network)
2312 if (strv_isempty(arg_network_macvlan))
2315 r = sd_rtnl_open(&rtnl, 0);
2317 return log_error_errno(r, "Failed to connect to netlink: %m");
2321 log_error("Failed to connect to udev.");
2325 STRV_FOREACH(i, arg_network_macvlan) {
2326 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
2327 _cleanup_free_ char *n = NULL;
2328 struct ether_addr mac;
2331 ifi = parse_interface(udev, *i);
2335 r = generate_mac(&mac, MACVLAN_HASH_KEY, idx++);
2337 return log_error_errno(r, "Failed to create MACVLAN MAC address: %m");
2339 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
2341 return log_error_errno(r, "Failed to allocate netlink message: %m");
2343 r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi);
2345 return log_error_errno(r, "Failed to add netlink interface index: %m");
2347 n = strappend("mv-", *i);
2351 strshorten(n, IFNAMSIZ-1);
2353 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n);
2355 return log_error_errno(r, "Failed to add netlink interface name: %m");
2357 r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac);
2359 return log_error_errno(r, "Failed to add netlink MAC address: %m");
2361 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
2363 return log_error_errno(r, "Failed to add netlink namespace field: %m");
2365 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
2367 return log_error_errno(r, "Failed to open netlink container: %m");
2369 r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "macvlan");
2371 return log_error_errno(r, "Failed to open netlink container: %m");
2373 r = sd_rtnl_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE);
2375 return log_error_errno(r, "Failed to append macvlan mode: %m");
2377 r = sd_rtnl_message_close_container(m);
2379 return log_error_errno(r, "Failed to close netlink container: %m");
2381 r = sd_rtnl_message_close_container(m);
2383 return log_error_errno(r, "Failed to close netlink container: %m");
2385 r = sd_rtnl_call(rtnl, m, 0, NULL);
2387 return log_error_errno(r, "Failed to add new macvlan interfaces: %m");
2393 static int setup_ipvlan(pid_t pid) {
2394 _cleanup_udev_unref_ struct udev *udev = NULL;
2395 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
2399 if (!arg_private_network)
2402 if (strv_isempty(arg_network_ipvlan))
2405 r = sd_rtnl_open(&rtnl, 0);
2407 return log_error_errno(r, "Failed to connect to netlink: %m");
2411 log_error("Failed to connect to udev.");
2415 STRV_FOREACH(i, arg_network_ipvlan) {
2416 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
2417 _cleanup_free_ char *n = NULL;
2420 ifi = parse_interface(udev, *i);
2424 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
2426 return log_error_errno(r, "Failed to allocate netlink message: %m");
2428 r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi);
2430 return log_error_errno(r, "Failed to add netlink interface index: %m");
2432 n = strappend("iv-", *i);
2436 strshorten(n, IFNAMSIZ-1);
2438 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n);
2440 return log_error_errno(r, "Failed to add netlink interface name: %m");
2442 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
2444 return log_error_errno(r, "Failed to add netlink namespace field: %m");
2446 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
2448 return log_error_errno(r, "Failed to open netlink container: %m");
2450 r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "ipvlan");
2452 return log_error_errno(r, "Failed to open netlink container: %m");
2454 r = sd_rtnl_message_append_u16(m, IFLA_IPVLAN_MODE, IPVLAN_MODE_L2);
2456 return log_error_errno(r, "Failed to add ipvlan mode: %m");
2458 r = sd_rtnl_message_close_container(m);
2460 return log_error_errno(r, "Failed to close netlink container: %m");
2462 r = sd_rtnl_message_close_container(m);
2464 return log_error_errno(r, "Failed to close netlink container: %m");
2466 r = sd_rtnl_call(rtnl, m, 0, NULL);
2468 return log_error_errno(r, "Failed to add new ipvlan interfaces: %m");
2474 static int setup_seccomp(void) {
2477 static const int blacklist[] = {
2478 SCMP_SYS(kexec_load),
2479 SCMP_SYS(open_by_handle_at),
2480 SCMP_SYS(init_module),
2481 SCMP_SYS(finit_module),
2482 SCMP_SYS(delete_module),
2489 scmp_filter_ctx seccomp;
2493 seccomp = seccomp_init(SCMP_ACT_ALLOW);
2497 r = seccomp_add_secondary_archs(seccomp);
2499 log_error_errno(r, "Failed to add secondary archs to seccomp filter: %m");
2503 for (i = 0; i < ELEMENTSOF(blacklist); i++) {
2504 r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), blacklist[i], 0);
2506 continue; /* unknown syscall */
2508 log_error_errno(r, "Failed to block syscall: %m");
2514 Audit is broken in containers, much of the userspace audit
2515 hookup will fail if running inside a container. We don't
2516 care and just turn off creation of audit sockets.
2518 This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail
2519 with EAFNOSUPPORT which audit userspace uses as indication
2520 that audit is disabled in the kernel.
2523 r = seccomp_rule_add(
2525 SCMP_ACT_ERRNO(EAFNOSUPPORT),
2528 SCMP_A0(SCMP_CMP_EQ, AF_NETLINK),
2529 SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT));
2531 log_error_errno(r, "Failed to add audit seccomp rule: %m");
2535 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
2537 log_error_errno(r, "Failed to unset NO_NEW_PRIVS: %m");
2541 r = seccomp_load(seccomp);
2543 log_error_errno(r, "Failed to install seccomp audit filter: %m");
2546 seccomp_release(seccomp);
2554 static int setup_propagate(const char *root) {
2557 (void) mkdir_p("/run/systemd/nspawn/", 0755);
2558 (void) mkdir_p("/run/systemd/nspawn/propagate", 0600);
2559 p = strappenda("/run/systemd/nspawn/propagate/", arg_machine);
2560 (void) mkdir_p(p, 0600);
2562 q = strappenda(root, "/run/systemd/nspawn/incoming");
2563 mkdir_parents(q, 0755);
2566 if (mount(p, q, NULL, MS_BIND, NULL) < 0)
2567 return log_error_errno(errno, "Failed to install propagation bind mount.");
2569 if (mount(NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0)
2570 return log_error_errno(errno, "Failed to make propagation mount read-only");
2575 static int setup_image(char **device_path, int *loop_nr) {
2576 struct loop_info64 info = {
2577 .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN
2579 _cleanup_close_ int fd = -1, control = -1, loop = -1;
2580 _cleanup_free_ char* loopdev = NULL;
2584 assert(device_path);
2588 fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
2590 return log_error_errno(errno, "Failed to open %s: %m", arg_image);
2592 if (fstat(fd, &st) < 0)
2593 return log_error_errno(errno, "Failed to stat %s: %m", arg_image);
2595 if (S_ISBLK(st.st_mode)) {
2598 p = strdup(arg_image);
2612 if (!S_ISREG(st.st_mode)) {
2613 log_error_errno(errno, "%s is not a regular file or block device: %m", arg_image);
2617 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
2619 return log_error_errno(errno, "Failed to open /dev/loop-control: %m");
2621 nr = ioctl(control, LOOP_CTL_GET_FREE);
2623 return log_error_errno(errno, "Failed to allocate loop device: %m");
2625 if (asprintf(&loopdev, "/dev/loop%i", nr) < 0)
2628 loop = open(loopdev, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
2630 return log_error_errno(errno, "Failed to open loop device %s: %m", loopdev);
2632 if (ioctl(loop, LOOP_SET_FD, fd) < 0)
2633 return log_error_errno(errno, "Failed to set loopback file descriptor on %s: %m", loopdev);
2636 info.lo_flags |= LO_FLAGS_READ_ONLY;
2638 if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0)
2639 return log_error_errno(errno, "Failed to set loopback settings on %s: %m", loopdev);
2641 *device_path = loopdev;
2652 #define PARTITION_TABLE_BLURB \
2653 "Note that the disk image needs to either contain only a single MBR partition of\n" \
2654 "type 0x83 that is marked bootable, or a sinlge GPT partition of type" \
2655 "0FC63DAF-8483-4772-8E79-3D69D8477DE4 or follow\n" \
2656 " http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/\n" \
2657 "to be bootable with systemd-nspawn."
2659 static int dissect_image(
2661 char **root_device, bool *root_device_rw,
2662 char **home_device, bool *home_device_rw,
2663 char **srv_device, bool *srv_device_rw,
2667 int home_nr = -1, srv_nr = -1;
2668 #ifdef GPT_ROOT_NATIVE
2671 #ifdef GPT_ROOT_SECONDARY
2672 int secondary_root_nr = -1;
2674 _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *generic = NULL;
2675 _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL;
2676 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
2677 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
2678 _cleanup_udev_unref_ struct udev *udev = NULL;
2679 struct udev_list_entry *first, *item;
2680 bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true, generic_rw = true;
2681 bool is_gpt, is_mbr, multiple_generic = false;
2682 const char *pttype = NULL;
2689 assert(root_device);
2690 assert(home_device);
2695 b = blkid_new_probe();
2700 r = blkid_probe_set_device(b, fd, 0, 0);
2705 log_error_errno(errno, "Failed to set device on blkid probe: %m");
2709 blkid_probe_enable_partitions(b, 1);
2710 blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
2713 r = blkid_do_safeprobe(b);
2714 if (r == -2 || r == 1) {
2715 log_error("Failed to identify any partition table on\n"
2717 PARTITION_TABLE_BLURB, arg_image);
2719 } else if (r != 0) {
2722 log_error_errno(errno, "Failed to probe: %m");
2726 blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
2728 is_gpt = streq_ptr(pttype, "gpt");
2729 is_mbr = streq_ptr(pttype, "dos");
2731 if (!is_gpt && !is_mbr) {
2732 log_error("No GPT or MBR partition table discovered on\n"
2734 PARTITION_TABLE_BLURB, arg_image);
2739 pl = blkid_probe_get_partitions(b);
2744 log_error("Failed to list partitions of %s", arg_image);
2752 if (fstat(fd, &st) < 0)
2753 return log_error_errno(errno, "Failed to stat block device: %m");
2755 d = udev_device_new_from_devnum(udev, 'b', st.st_rdev);
2763 log_error("Kernel partitions never appeared.");
2767 e = udev_enumerate_new(udev);
2771 r = udev_enumerate_add_match_parent(e, d);
2775 r = udev_enumerate_scan_devices(e);
2777 return log_error_errno(r, "Failed to scan for partition devices of %s: %m", arg_image);
2779 /* Count the partitions enumerated by the kernel */
2781 first = udev_enumerate_get_list_entry(e);
2782 udev_list_entry_foreach(item, first)
2785 /* Count the partitions enumerated by blkid */
2786 m = blkid_partlist_numof_partitions(pl);
2790 log_error("blkid and kernel partition list do not match.");
2796 /* The kernel has probed fewer partitions than
2797 * blkid? Maybe the kernel prober is still
2798 * running or it got EBUSY because udev
2799 * already opened the device. Let's reprobe
2800 * the device, which is a synchronous call
2801 * that waits until probing is complete. */
2803 for (j = 0; j < 20; j++) {
2805 r = ioctl(fd, BLKRRPART, 0);
2808 if (r >= 0 || r != -EBUSY)
2811 /* If something else has the device
2812 * open, such as an udev rule, the
2813 * ioctl will return EBUSY. Since
2814 * there's no way to wait until it
2815 * isn't busy anymore, let's just wait
2816 * a bit, and try again.
2818 * This is really something they
2819 * should fix in the kernel! */
2821 usleep(50 * USEC_PER_MSEC);
2825 return log_error_errno(r, "Failed to reread partition table: %m");
2828 e = udev_enumerate_unref(e);
2831 first = udev_enumerate_get_list_entry(e);
2832 udev_list_entry_foreach(item, first) {
2833 _cleanup_udev_device_unref_ struct udev_device *q;
2835 unsigned long long flags;
2841 q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));
2846 log_error_errno(errno, "Failed to get partition device of %s: %m", arg_image);
2850 qn = udev_device_get_devnum(q);
2854 if (st.st_rdev == qn)
2857 node = udev_device_get_devnode(q);
2861 pp = blkid_partlist_devno_to_partition(pl, qn);
2865 flags = blkid_partition_get_flags(pp);
2867 nr = blkid_partition_get_partno(pp);
2875 if (flags & GPT_FLAG_NO_AUTO)
2878 stype = blkid_partition_get_type_string(pp);
2882 if (sd_id128_from_string(stype, &type_id) < 0)
2885 if (sd_id128_equal(type_id, GPT_HOME)) {
2887 if (home && nr >= home_nr)
2891 home_rw = !(flags & GPT_FLAG_READ_ONLY);
2893 r = free_and_strdup(&home, node);
2897 } else if (sd_id128_equal(type_id, GPT_SRV)) {
2899 if (srv && nr >= srv_nr)
2903 srv_rw = !(flags & GPT_FLAG_READ_ONLY);
2905 r = free_and_strdup(&srv, node);
2909 #ifdef GPT_ROOT_NATIVE
2910 else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) {
2912 if (root && nr >= root_nr)
2916 root_rw = !(flags & GPT_FLAG_READ_ONLY);
2918 r = free_and_strdup(&root, node);
2923 #ifdef GPT_ROOT_SECONDARY
2924 else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) {
2926 if (secondary_root && nr >= secondary_root_nr)
2929 secondary_root_nr = nr;
2930 secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY);
2932 r = free_and_strdup(&secondary_root, node);
2937 else if (sd_id128_equal(type_id, GPT_LINUX_GENERIC)) {
2940 multiple_generic = true;
2942 generic_rw = !(flags & GPT_FLAG_READ_ONLY);
2944 r = free_and_strdup(&generic, node);
2950 } else if (is_mbr) {
2953 if (flags != 0x80) /* Bootable flag */
2956 type = blkid_partition_get_type(pp);
2957 if (type != 0x83) /* Linux partition */
2961 multiple_generic = true;
2965 r = free_and_strdup(&root, node);
2973 *root_device = root;
2976 *root_device_rw = root_rw;
2978 } else if (secondary_root) {
2979 *root_device = secondary_root;
2980 secondary_root = NULL;
2982 *root_device_rw = secondary_root_rw;
2984 } else if (generic) {
2986 /* There were no partitions with precise meanings
2987 * around, but we found generic partitions. In this
2988 * case, if there's only one, we can go ahead and boot
2989 * it, otherwise we bail out, because we really cannot
2990 * make any sense of it. */
2992 if (multiple_generic) {
2993 log_error("Identified multiple bootable Linux partitions on\n"
2995 PARTITION_TABLE_BLURB, arg_image);
2999 *root_device = generic;
3002 *root_device_rw = generic_rw;
3005 log_error("Failed to identify root partition in disk image\n"
3007 PARTITION_TABLE_BLURB, arg_image);
3012 *home_device = home;
3015 *home_device_rw = home_rw;
3022 *srv_device_rw = srv_rw;
3027 log_error("--image= is not supported, compiled without blkid support.");
3032 static int mount_device(const char *what, const char *where, const char *directory, bool rw) {
3034 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
3035 const char *fstype, *p;
3045 p = strappenda(where, directory);
3050 b = blkid_new_probe_from_filename(what);
3054 log_error_errno(errno, "Failed to allocate prober for %s: %m", what);
3058 blkid_probe_enable_superblocks(b, 1);
3059 blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE);
3062 r = blkid_do_safeprobe(b);
3063 if (r == -1 || r == 1) {
3064 log_error("Cannot determine file system type of %s", what);
3066 } else if (r != 0) {
3069 log_error_errno(errno, "Failed to probe %s: %m", what);
3074 if (blkid_probe_lookup_value(b, "TYPE", &fstype, NULL) < 0) {
3077 log_error("Failed to determine file system type of %s", what);
3081 if (streq(fstype, "crypto_LUKS")) {
3082 log_error("nspawn currently does not support LUKS disk images.");
3086 if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0)
3087 return log_error_errno(errno, "Failed to mount %s: %m", what);
3091 log_error("--image= is not supported, compiled without blkid support.");
3096 static int mount_devices(
3098 const char *root_device, bool root_device_rw,
3099 const char *home_device, bool home_device_rw,
3100 const char *srv_device, bool srv_device_rw) {
3106 r = mount_device(root_device, arg_directory, NULL, root_device_rw);
3108 return log_error_errno(r, "Failed to mount root directory: %m");
3112 r = mount_device(home_device, arg_directory, "/home", home_device_rw);
3114 return log_error_errno(r, "Failed to mount home directory: %m");
3118 r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw);
3120 return log_error_errno(r, "Failed to mount server data directory: %m");
3126 static void loop_remove(int nr, int *image_fd) {
3127 _cleanup_close_ int control = -1;
3133 if (image_fd && *image_fd >= 0) {
3134 r = ioctl(*image_fd, LOOP_CLR_FD);
3136 log_debug_errno(errno, "Failed to close loop image: %m");
3137 *image_fd = safe_close(*image_fd);
3140 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
3142 log_warning_errno(errno, "Failed to open /dev/loop-control: %m");
3146 r = ioctl(control, LOOP_CTL_REMOVE, nr);
3148 log_debug_errno(errno, "Failed to remove loop %d: %m", nr);
3151 static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
3159 if (pipe2(pipe_fds, O_CLOEXEC) < 0)
3160 return log_error_errno(errno, "Failed to allocate pipe: %m");
3164 return log_error_errno(errno, "Failed to fork getent child: %m");
3165 else if (pid == 0) {
3167 char *empty_env = NULL;
3169 if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0)
3170 _exit(EXIT_FAILURE);
3172 if (pipe_fds[0] > 2)
3173 safe_close(pipe_fds[0]);
3174 if (pipe_fds[1] > 2)
3175 safe_close(pipe_fds[1]);
3177 nullfd = open("/dev/null", O_RDWR);
3179 _exit(EXIT_FAILURE);
3181 if (dup3(nullfd, STDIN_FILENO, 0) < 0)
3182 _exit(EXIT_FAILURE);
3184 if (dup3(nullfd, STDERR_FILENO, 0) < 0)
3185 _exit(EXIT_FAILURE);
3190 reset_all_signal_handlers();
3191 close_all_fds(NULL, 0);
3193 execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env);
3194 execle("/bin/getent", "getent", database, key, NULL, &empty_env);
3195 _exit(EXIT_FAILURE);
3198 pipe_fds[1] = safe_close(pipe_fds[1]);
3205 static int change_uid_gid(char **_home) {
3206 char line[LINE_MAX], *x, *u, *g, *h;
3207 const char *word, *state;
3208 _cleanup_free_ uid_t *uids = NULL;
3209 _cleanup_free_ char *home = NULL;
3210 _cleanup_fclose_ FILE *f = NULL;
3211 _cleanup_close_ int fd = -1;
3212 unsigned n_uids = 0;
3221 if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) {
3222 /* Reset everything fully to 0, just in case */
3224 if (setgroups(0, NULL) < 0)
3225 return log_error_errno(errno, "setgroups() failed: %m");
3227 if (setresgid(0, 0, 0) < 0)
3228 return log_error_errno(errno, "setregid() failed: %m");
3230 if (setresuid(0, 0, 0) < 0)
3231 return log_error_errno(errno, "setreuid() failed: %m");
3237 /* First, get user credentials */
3238 fd = spawn_getent("passwd", arg_user, &pid);
3242 f = fdopen(fd, "r");
3247 if (!fgets(line, sizeof(line), f)) {
3250 log_error("Failed to resolve user %s.", arg_user);
3254 log_error_errno(errno, "Failed to read from getent: %m");
3260 wait_for_terminate_and_warn("getent passwd", pid, true);
3262 x = strchr(line, ':');
3264 log_error("/etc/passwd entry has invalid user field.");
3268 u = strchr(x+1, ':');
3270 log_error("/etc/passwd entry has invalid password field.");
3277 log_error("/etc/passwd entry has invalid UID field.");
3285 log_error("/etc/passwd entry has invalid GID field.");
3290 h = strchr(x+1, ':');
3292 log_error("/etc/passwd entry has invalid GECOS field.");
3299 log_error("/etc/passwd entry has invalid home directory field.");
3305 r = parse_uid(u, &uid);
3307 log_error("Failed to parse UID of user.");
3311 r = parse_gid(g, &gid);
3313 log_error("Failed to parse GID of user.");
3321 /* Second, get group memberships */
3322 fd = spawn_getent("initgroups", arg_user, &pid);
3327 f = fdopen(fd, "r");
3332 if (!fgets(line, sizeof(line), f)) {
3334 log_error("Failed to resolve user %s.", arg_user);
3338 log_error_errno(errno, "Failed to read from getent: %m");
3344 wait_for_terminate_and_warn("getent initgroups", pid, true);
3346 /* Skip over the username and subsequent separator whitespace */
3348 x += strcspn(x, WHITESPACE);
3349 x += strspn(x, WHITESPACE);
3351 FOREACH_WORD(word, l, x, state) {
3357 if (!GREEDY_REALLOC(uids, sz, n_uids+1))
3360 r = parse_uid(c, &uids[n_uids++]);
3362 log_error("Failed to parse group data from getent.");
3367 r = mkdir_parents(home, 0775);
3369 return log_error_errno(r, "Failed to make home root directory: %m");
3371 r = mkdir_safe(home, 0755, uid, gid);
3372 if (r < 0 && r != -EEXIST)
3373 return log_error_errno(r, "Failed to make home directory: %m");
3375 fchown(STDIN_FILENO, uid, gid);
3376 fchown(STDOUT_FILENO, uid, gid);
3377 fchown(STDERR_FILENO, uid, gid);
3379 if (setgroups(n_uids, uids) < 0)
3380 return log_error_errno(errno, "Failed to set auxiliary groups: %m");
3382 if (setresgid(gid, gid, gid) < 0)
3383 return log_error_errno(errno, "setregid() failed: %m");
3385 if (setresuid(uid, uid, uid) < 0)
3386 return log_error_errno(errno, "setreuid() failed: %m");
3398 * < 0 : wait_for_terminate() failed to get the state of the
3399 * container, the container was terminated by a signal, or
3400 * failed for an unknown reason. No change is made to the
3401 * container argument.
3402 * > 0 : The program executed in the container terminated with an
3403 * error. The exit code of the program executed in the
3404 * container is returned. The container argument has been set
3405 * to CONTAINER_TERMINATED.
3406 * 0 : The container is being rebooted, has been shut down or exited
3407 * successfully. The container argument has been set to either
3408 * CONTAINER_TERMINATED or CONTAINER_REBOOTED.
3410 * That is, success is indicated by a return value of zero, and an
3411 * error is indicated by a non-zero value.
3413 static int wait_for_container(pid_t pid, ContainerStatus *container) {
3417 r = wait_for_terminate(pid, &status);
3419 return log_warning_errno(r, "Failed to wait for container: %m");
3421 switch (status.si_code) {
3424 if (status.si_status == 0) {
3425 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s exited successfully.", arg_machine);
3428 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s failed with error code %i.", arg_machine, status.si_status);
3430 *container = CONTAINER_TERMINATED;
3431 return status.si_status;
3434 if (status.si_status == SIGINT) {
3436 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s has been shut down.", arg_machine);
3437 *container = CONTAINER_TERMINATED;
3440 } else if (status.si_status == SIGHUP) {
3442 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s is being rebooted.", arg_machine);
3443 *container = CONTAINER_REBOOTED;
3447 /* CLD_KILLED fallthrough */
3450 log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
3454 log_error("Container %s failed due to unknown reason.", arg_machine);
3461 static void nop_handler(int sig) {}
3463 static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
3466 pid = PTR_TO_UINT32(userdata);
3468 if (kill(pid, SIGRTMIN+3) >= 0) {
3469 log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
3470 sd_event_source_set_userdata(s, NULL);
3475 sd_event_exit(sd_event_source_get_event(s), 0);
3479 static int determine_names(void) {
3482 if (!arg_image && !arg_directory) {
3484 _cleanup_(image_unrefp) Image *i = NULL;
3486 r = image_find(arg_machine, &i);
3488 return log_error_errno(r, "Failed to find image for machine '%s': %m", arg_machine);
3490 log_error("No image for machine '%s': %m", arg_machine);
3494 if (i->type == IMAGE_RAW)
3495 r = set_sanitized_path(&arg_image, i->path);
3497 r = set_sanitized_path(&arg_directory, i->path);
3499 return log_error_errno(r, "Invalid image directory: %m");
3501 arg_read_only = arg_read_only || i->read_only;
3503 arg_directory = get_current_dir_name();
3505 if (!arg_directory && !arg_machine) {
3506 log_error("Failed to determine path, please use -D or -i.");
3512 if (arg_directory && path_equal(arg_directory, "/"))
3513 arg_machine = gethostname_malloc();
3515 arg_machine = strdup(basename(arg_image ?: arg_directory));
3520 hostname_cleanup(arg_machine, false);
3521 if (!machine_name_is_valid(arg_machine)) {
3522 log_error("Failed to determine machine name automatically, please use -M.");
3526 if (arg_ephemeral) {
3529 /* Add a random suffix when this is an
3530 * ephemeral machine, so that we can run many
3531 * instances at once without manually having
3532 * to specify -M each time. */
3534 if (asprintf(&b, "%s-%016" PRIx64, arg_machine, random_u64()) < 0)
3545 int main(int argc, char *argv[]) {
3547 _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *console = NULL;
3548 bool root_device_rw = true, home_device_rw = true, srv_device_rw = true;
3549 _cleanup_close_ int master = -1, image_fd = -1;
3550 _cleanup_fdset_free_ FDSet *fds = NULL;
3551 int r, n_fd_passed, loop_nr = -1;
3552 char veth_name[IFNAMSIZ];
3553 bool secondary = false, remove_subvol = false;
3554 sigset_t mask, mask_chld;
3556 int ret = EXIT_SUCCESS;
3557 union in_addr_union exposed = {};
3558 _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT;
3560 log_parse_environment();
3563 r = parse_argv(argc, argv);
3567 r = determine_names();
3571 if (geteuid() != 0) {
3572 log_error("Need to be root.");
3577 if (sd_booted() <= 0) {
3578 log_error("Not running on a systemd system.");
3584 n_fd_passed = sd_listen_fds(false);
3585 if (n_fd_passed > 0) {
3586 r = fdset_new_listen_fds(&fds, false);
3588 log_error_errno(r, "Failed to collect file descriptors: %m");
3592 fdset_close_others(fds);
3595 if (arg_directory) {
3598 if (path_equal(arg_directory, "/") && !arg_ephemeral) {
3599 log_error("Spawning container on root directory is not supported. Consider using --ephemeral.");
3604 if (arg_ephemeral) {
3605 _cleanup_release_lock_file_ LockFile original_lock = LOCK_FILE_INIT;
3608 /* If the specified path is a mount point we
3609 * generate the new snapshot immediately
3610 * inside it under a random name. However if
3611 * the specified is not a mount point we
3612 * create the new snapshot in the parent
3613 * directory, just next to it. */
3614 r = path_is_mount_point(arg_directory, false);
3616 log_error_errno(r, "Failed to determine whether directory %s is mount point: %m", arg_directory);
3620 r = tempfn_random_child(arg_directory, &np);
3622 r = tempfn_random(arg_directory, &np);
3624 log_error_errno(r, "Failed to generate name for snapshot: %m");
3628 r = image_path_lock(np, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3630 log_error_errno(r, "Failed to lock %s: %m", np);
3634 r = btrfs_subvol_snapshot(arg_directory, np, arg_read_only, true);
3637 log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory);
3641 free(arg_directory);
3644 remove_subvol = true;
3647 r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3649 log_error_errno(r, "Directory tree %s is currently busy.", arg_directory);
3653 log_error_errno(r, "Failed to lock %s: %m", arg_directory);
3658 r = btrfs_subvol_snapshot(arg_template, arg_directory, arg_read_only, true);
3661 log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template);
3663 log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template);
3667 log_info("Populated %s from template %s.", arg_directory, arg_template);
3673 if (path_is_os_tree(arg_directory) <= 0) {
3674 log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory);
3681 p = strappenda(arg_directory,
3682 argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/");
3683 if (access(p, F_OK) < 0) {
3684 log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory);
3691 char template[] = "/tmp/nspawn-root-XXXXXX";
3694 assert(!arg_template);
3696 r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3698 r = log_error_errno(r, "Disk image %s is currently busy.", arg_image);
3702 r = log_error_errno(r, "Failed to create image lock: %m");
3706 if (!mkdtemp(template)) {
3707 log_error_errno(errno, "Failed to create temporary directory: %m");
3712 arg_directory = strdup(template);
3713 if (!arg_directory) {
3718 image_fd = setup_image(&device_path, &loop_nr);
3724 r = dissect_image(image_fd,
3725 &root_device, &root_device_rw,
3726 &home_device, &home_device_rw,
3727 &srv_device, &srv_device_rw,
3733 master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
3735 r = log_error_errno(errno, "Failed to acquire pseudo tty: %m");
3739 r = ptsname_malloc(master, &console);
3741 r = log_error_errno(r, "Failed to determine tty name: %m");
3746 log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.",
3747 arg_machine, arg_image ?: arg_directory);
3749 if (unlockpt(master) < 0) {
3750 r = log_error_errno(errno, "Failed to unlock tty: %m");
3754 assert_se(sigemptyset(&mask) == 0);
3755 sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
3756 assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
3758 assert_se(sigemptyset(&mask_chld) == 0);
3759 assert_se(sigaddset(&mask_chld, SIGCHLD) == 0);
3762 _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }, rtnl_socket_pair[2] = { -1, -1 };
3763 ContainerStatus container_status;
3764 _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
3765 struct sigaction sa = {
3766 .sa_handler = nop_handler,
3767 .sa_flags = SA_NOCLDSTOP,
3770 r = barrier_create(&barrier);
3772 log_error_errno(r, "Cannot initialize IPC barrier: %m");
3776 if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
3777 r = log_error_errno(errno, "Failed to create kmsg socket pair: %m");
3781 if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) {
3782 r = log_error_errno(errno, "Failed to create rtnl socket pair: %m");
3786 /* Child can be killed before execv(), so handle SIGCHLD
3787 * in order to interrupt parent's blocking calls and
3788 * give it a chance to call wait() and terminate. */
3789 r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
3791 r = log_error_errno(errno, "Failed to change the signal mask: %m");
3795 r = sigaction(SIGCHLD, &sa, NULL);
3797 r = log_error_errno(errno, "Failed to install SIGCHLD handler: %m");
3801 pid = raw_clone(SIGCHLD|CLONE_NEWNS|
3802 (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)|
3803 (arg_private_network ? CLONE_NEWNET : 0), NULL);
3805 if (errno == EINVAL)
3806 r = log_error_errno(errno, "clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
3808 r = log_error_errno(errno, "clone() failed: %m");
3815 _cleanup_free_ char *home = NULL;
3817 const char *envp[] = {
3818 "PATH=" DEFAULT_PATH_SPLIT_USR,
3819 "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
3824 NULL, /* container_uuid */
3825 NULL, /* LISTEN_FDS */
3826 NULL, /* LISTEN_PID */
3831 barrier_set_role(&barrier, BARRIER_CHILD);
3833 envp[n_env] = strv_find_prefix(environ, "TERM=");
3837 master = safe_close(master);
3839 close_nointr(STDIN_FILENO);
3840 close_nointr(STDOUT_FILENO);
3841 close_nointr(STDERR_FILENO);
3843 kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
3844 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
3846 reset_all_signal_handlers();
3847 reset_signal_mask();
3849 r = open_terminal(console, O_RDWR);
3850 if (r != STDIN_FILENO) {
3856 log_error_errno(r, "Failed to open console: %m");
3857 _exit(EXIT_FAILURE);
3860 if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
3861 dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
3862 log_error_errno(errno, "Failed to duplicate console: %m");
3863 _exit(EXIT_FAILURE);
3867 log_error_errno(errno, "setsid() failed: %m");
3868 _exit(EXIT_FAILURE);
3871 if (reset_audit_loginuid() < 0)
3872 _exit(EXIT_FAILURE);
3874 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
3875 log_error_errno(errno, "PR_SET_PDEATHSIG failed: %m");
3876 _exit(EXIT_FAILURE);
3879 /* Mark everything as slave, so that we still
3880 * receive mounts from the real root, but don't
3881 * propagate mounts to the real root. */
3882 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
3883 log_error_errno(errno, "MS_SLAVE|MS_REC failed: %m");
3884 _exit(EXIT_FAILURE);
3887 if (mount_devices(arg_directory,
3888 root_device, root_device_rw,
3889 home_device, home_device_rw,
3890 srv_device, srv_device_rw) < 0)
3891 _exit(EXIT_FAILURE);
3893 /* Turn directory into bind mount */
3894 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
3895 log_error_errno(errno, "Failed to make bind mount: %m");
3896 _exit(EXIT_FAILURE);
3899 r = setup_volatile(arg_directory);
3901 _exit(EXIT_FAILURE);
3903 if (setup_volatile_state(arg_directory) < 0)
3904 _exit(EXIT_FAILURE);
3906 r = base_filesystem_create(arg_directory);
3908 _exit(EXIT_FAILURE);
3910 if (arg_read_only) {
3911 r = bind_remount_recursive(arg_directory, true);
3913 log_error_errno(r, "Failed to make tree read-only: %m");
3914 _exit(EXIT_FAILURE);
3918 if (mount_all(arg_directory) < 0)
3919 _exit(EXIT_FAILURE);
3921 if (copy_devnodes(arg_directory) < 0)
3922 _exit(EXIT_FAILURE);
3924 if (setup_ptmx(arg_directory) < 0)
3925 _exit(EXIT_FAILURE);
3927 dev_setup(arg_directory);
3929 if (setup_propagate(arg_directory) < 0)
3930 _exit(EXIT_FAILURE);
3932 if (setup_seccomp() < 0)
3933 _exit(EXIT_FAILURE);
3935 if (setup_dev_console(arg_directory, console) < 0)
3936 _exit(EXIT_FAILURE);
3938 if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
3939 _exit(EXIT_FAILURE);
3940 kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
3942 if (send_rtnl(rtnl_socket_pair[1]) < 0)
3943 _exit(EXIT_FAILURE);
3944 rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
3946 /* Tell the parent that we are ready, and that
3947 * it can cgroupify us to that we lack access
3948 * to certain devices and resources. */
3949 (void) barrier_place(&barrier);
3951 if (setup_boot_id(arg_directory) < 0)
3952 _exit(EXIT_FAILURE);
3954 if (setup_timezone(arg_directory) < 0)
3955 _exit(EXIT_FAILURE);
3957 if (setup_resolv_conf(arg_directory) < 0)
3958 _exit(EXIT_FAILURE);
3960 if (setup_journal(arg_directory) < 0)
3961 _exit(EXIT_FAILURE);
3963 if (mount_binds(arg_directory, arg_bind, false) < 0)
3964 _exit(EXIT_FAILURE);
3966 if (mount_binds(arg_directory, arg_bind_ro, true) < 0)
3967 _exit(EXIT_FAILURE);
3969 if (mount_tmpfs(arg_directory) < 0)
3970 _exit(EXIT_FAILURE);
3972 /* Wait until we are cgroup-ified, so that we
3973 * can mount the right cgroup path writable */
3974 (void) barrier_sync_next(&barrier);
3976 if (mount_cgroup(arg_directory) < 0)
3977 _exit(EXIT_FAILURE);
3979 if (chdir(arg_directory) < 0) {
3980 log_error_errno(errno, "chdir(%s) failed: %m", arg_directory);
3981 _exit(EXIT_FAILURE);
3984 if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
3985 log_error_errno(errno, "mount(MS_MOVE) failed: %m");
3986 _exit(EXIT_FAILURE);
3989 if (chroot(".") < 0) {
3990 log_error_errno(errno, "chroot() failed: %m");
3991 _exit(EXIT_FAILURE);
3994 if (chdir("/") < 0) {
3995 log_error_errno(errno, "chdir() failed: %m");
3996 _exit(EXIT_FAILURE);
4001 if (arg_private_network)
4004 if (drop_capabilities() < 0) {
4005 log_error_errno(errno, "drop_capabilities() failed: %m");
4006 _exit(EXIT_FAILURE);
4009 r = change_uid_gid(&home);
4011 _exit(EXIT_FAILURE);
4013 if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
4014 (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
4015 (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
4017 _exit(EXIT_FAILURE);
4020 if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) {
4023 if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) {
4025 _exit(EXIT_FAILURE);
4029 if (fdset_size(fds) > 0) {
4030 r = fdset_cloexec(fds, false);
4032 log_error_errno(r, "Failed to unset O_CLOEXEC for file descriptors.");
4033 _exit(EXIT_FAILURE);
4036 if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
4037 (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) {
4039 _exit(EXIT_FAILURE);
4045 if (arg_personality != 0xffffffffLU) {
4046 if (personality(arg_personality) < 0) {
4047 log_error_errno(errno, "personality() failed: %m");
4048 _exit(EXIT_FAILURE);
4050 } else if (secondary) {
4051 if (personality(PER_LINUX32) < 0) {
4052 log_error_errno(errno, "personality() failed: %m");
4053 _exit(EXIT_FAILURE);
4058 if (arg_selinux_context)
4059 if (setexeccon((security_context_t) arg_selinux_context) < 0) {
4060 log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context);
4061 _exit(EXIT_FAILURE);
4065 if (!strv_isempty(arg_setenv)) {
4068 n = strv_env_merge(2, envp, arg_setenv);
4071 _exit(EXIT_FAILURE);
4076 env_use = (char**) envp;
4078 /* Wait until the parent is ready with the setup, too... */
4079 if (!barrier_place_and_sync(&barrier))
4080 _exit(EXIT_FAILURE);
4086 /* Automatically search for the init system */
4088 l = 1 + argc - optind;
4089 a = newa(char*, l + 1);
4090 memcpy(a + 1, argv + optind, l * sizeof(char*));
4092 a[0] = (char*) "/usr/lib/systemd/systemd";
4093 execve(a[0], a, env_use);
4095 a[0] = (char*) "/lib/systemd/systemd";
4096 execve(a[0], a, env_use);
4098 a[0] = (char*) "/sbin/init";
4099 execve(a[0], a, env_use);
4100 } else if (argc > optind)
4101 execvpe(argv[optind], argv + optind, env_use);
4103 chdir(home ? home : "/root");
4104 execle("/bin/bash", "-bash", NULL, env_use);
4105 execle("/bin/sh", "-sh", NULL, env_use);
4108 log_error_errno(errno, "execv() failed: %m");
4109 _exit(EXIT_FAILURE);
4112 barrier_set_role(&barrier, BARRIER_PARENT);
4116 kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
4117 rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
4119 /* Wait for the most basic Child-setup to be done,
4120 * before we add hardware to it, and place it in a
4122 if (barrier_sync_next(&barrier)) {
4125 r = move_network_interfaces(pid);
4129 r = setup_veth(pid, veth_name, &ifi);
4133 r = setup_bridge(veth_name, &ifi);
4137 r = setup_macvlan(pid);
4141 r = setup_ipvlan(pid);
4145 r = register_machine(pid, ifi);
4149 /* Block SIGCHLD here, before notifying child.
4150 * process_pty() will handle it with the other signals. */
4151 r = sigprocmask(SIG_BLOCK, &mask_chld, NULL);
4155 /* Reset signal to default */
4156 r = default_signals(SIGCHLD, -1);
4160 /* Notify the child that the parent is ready with all
4161 * its setup, and that the child can now hand over
4162 * control to the code to run inside the container. */
4163 (void) barrier_place(&barrier);
4165 /* And wait that the child is completely ready now. */
4166 if (barrier_place_and_sync(&barrier)) {
4167 _cleanup_event_unref_ sd_event *event = NULL;
4168 _cleanup_(pty_forward_freep) PTYForward *forward = NULL;
4169 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
4174 "STATUS=Container running.\n"
4175 "X_NSPAWN_LEADER_PID=" PID_FMT, pid);
4177 r = sd_event_new(&event);
4179 log_error_errno(r, "Failed to get default event source: %m");
4184 /* Try to kill the init system on SIGINT or SIGTERM */
4185 sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, UINT32_TO_PTR(pid));
4186 sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, UINT32_TO_PTR(pid));
4188 /* Immediately exit */
4189 sd_event_add_signal(event, NULL, SIGINT, NULL, NULL);
4190 sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL);
4193 /* simply exit on sigchld */
4194 sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL);
4196 if (arg_expose_ports) {
4197 r = watch_rtnl(event, rtnl_socket_pair[0], &exposed, &rtnl);
4201 (void) expose_ports(rtnl, &exposed);
4204 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
4206 r = pty_forward_new(event, master, true, &forward);
4208 log_error_errno(r, "Failed to create PTY forwarder: %m");
4212 r = sd_event_loop(event);
4214 log_error_errno(r, "Failed to run event loop: %m");
4218 pty_forward_get_last_char(forward, &last_char);
4220 forward = pty_forward_free(forward);
4222 if (!arg_quiet && last_char != '\n')
4225 /* Kill if it is not dead yet anyway */
4226 terminate_machine(pid);
4230 /* Normally redundant, but better safe than sorry */
4233 r = wait_for_container(pid, &container_status);
4237 /* We failed to wait for the container, or the
4238 * container exited abnormally */
4240 else if (r > 0 || container_status == CONTAINER_TERMINATED){
4241 /* The container exited with a non-zero
4242 * status, or with zero status and no reboot
4248 /* CONTAINER_REBOOTED, loop again */
4250 if (arg_keep_unit) {
4251 /* Special handling if we are running as a
4252 * service: instead of simply restarting the
4253 * machine we want to restart the entire
4254 * service, so let's inform systemd about this
4255 * with the special exit code 133. The service
4256 * file uses RestartForceExitStatus=133 so
4257 * that this results in a full nspawn
4258 * restart. This is necessary since we might
4259 * have cgroup parameters set we want to have
4266 flush_ports(&exposed);
4272 "STATUS=Terminating...");
4274 loop_remove(loop_nr, &image_fd);
4279 if (remove_subvol && arg_directory) {
4282 k = btrfs_subvol_remove(arg_directory);
4284 log_warning_errno(k, "Cannot remove subvolume '%s', ignoring: %m", arg_directory);
4290 p = strappenda("/run/systemd/nspawn/propagate/", arg_machine);
4291 (void) rm_rf(p, false, true, false);
4294 free(arg_directory);
4299 strv_free(arg_setenv);
4300 strv_free(arg_network_interfaces);
4301 strv_free(arg_network_macvlan);
4302 strv_free(arg_network_ipvlan);
4303 strv_free(arg_bind);
4304 strv_free(arg_bind_ro);
4305 strv_free(arg_tmpfs);
4307 flush_ports(&exposed);
4309 while (arg_expose_ports) {
4310 ExposePort *p = arg_expose_ports;
4311 LIST_REMOVE(ports, arg_expose_ports, p);
4315 return r < 0 ? EXIT_FAILURE : ret;