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>
43 #include <sys/eventfd.h>
45 #include <linux/veth.h>
46 #include <sys/personality.h>
47 #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"
91 #include "seccomp-util.h"
94 typedef enum LinkJournal {
101 static char *arg_directory = NULL;
102 static char *arg_user = NULL;
103 static sd_id128_t arg_uuid = {};
104 static char *arg_machine = NULL;
105 static const char *arg_selinux_context = NULL;
106 static const char *arg_selinux_apifs_context = NULL;
107 static const char *arg_slice = NULL;
108 static bool arg_private_network = false;
109 static bool arg_read_only = false;
110 static bool arg_boot = false;
111 static LinkJournal arg_link_journal = LINK_AUTO;
112 static uint64_t arg_retain =
113 (1ULL << CAP_CHOWN) |
114 (1ULL << CAP_DAC_OVERRIDE) |
115 (1ULL << CAP_DAC_READ_SEARCH) |
116 (1ULL << CAP_FOWNER) |
117 (1ULL << CAP_FSETID) |
118 (1ULL << CAP_IPC_OWNER) |
120 (1ULL << CAP_LEASE) |
121 (1ULL << CAP_LINUX_IMMUTABLE) |
122 (1ULL << CAP_NET_BIND_SERVICE) |
123 (1ULL << CAP_NET_BROADCAST) |
124 (1ULL << CAP_NET_RAW) |
125 (1ULL << CAP_SETGID) |
126 (1ULL << CAP_SETFCAP) |
127 (1ULL << CAP_SETPCAP) |
128 (1ULL << CAP_SETUID) |
129 (1ULL << CAP_SYS_ADMIN) |
130 (1ULL << CAP_SYS_CHROOT) |
131 (1ULL << CAP_SYS_NICE) |
132 (1ULL << CAP_SYS_PTRACE) |
133 (1ULL << CAP_SYS_TTY_CONFIG) |
134 (1ULL << CAP_SYS_RESOURCE) |
135 (1ULL << CAP_SYS_BOOT) |
136 (1ULL << CAP_AUDIT_WRITE) |
137 (1ULL << CAP_AUDIT_CONTROL) |
139 static char **arg_bind = NULL;
140 static char **arg_bind_ro = NULL;
141 static char **arg_setenv = NULL;
142 static bool arg_quiet = false;
143 static bool arg_share_system = false;
144 static bool arg_register = true;
145 static bool arg_keep_unit = false;
146 static char **arg_network_interfaces = NULL;
147 static char **arg_network_macvlan = NULL;
148 static bool arg_network_veth = false;
149 static const char *arg_network_bridge = NULL;
150 static unsigned long arg_personality = 0xffffffffLU;
151 static const char *arg_image = NULL;
153 static int help(void) {
155 printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
156 "Spawn a minimal namespace container for debugging, testing and building.\n\n"
157 " -h --help Show this help\n"
158 " --version Print version string\n"
159 " -q --quiet Do not show status information\n"
160 " -D --directory=PATH Root directory for the container\n"
161 " -i --image=PATH File system device or image for the container\n"
162 " -b --boot Boot up full system (i.e. invoke init)\n"
163 " -u --user=USER Run the command under specified user or uid\n"
164 " -M --machine=NAME Set the machine name for the container\n"
165 " --uuid=UUID Set a specific machine UUID for the container\n"
166 " -S --slice=SLICE Place the container in the specified slice\n"
167 " --private-network Disable network in container\n"
168 " --network-interface=INTERFACE\n"
169 " Assign an existing network interface to the\n"
171 " --network-macvlan=INTERFACE\n"
172 " Create a macvlan network interface based on an\n"
173 " existing network interface to the container\n"
174 " --network-veth Add a virtual ethernet connection between host\n"
176 " --network-bridge=INTERFACE\n"
177 " Add a virtual ethernet connection between host\n"
178 " and container and add it to an existing bridge on\n"
180 " -Z --selinux-context=SECLABEL\n"
181 " Set the SELinux security context to be used by\n"
182 " processes in the container\n"
183 " -L --selinux-apifs-context=SECLABEL\n"
184 " Set the SELinux security context to be used by\n"
185 " API/tmpfs file systems in the container\n"
186 " --capability=CAP In addition to the default, retain specified\n"
188 " --drop-capability=CAP Drop the specified capability from the default set\n"
189 " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n"
190 " -j Equivalent to --link-journal=host\n"
191 " --read-only Mount the root directory read-only\n"
192 " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n"
194 " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n"
195 " --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
196 " --share-system Share system namespaces with host\n"
197 " --register=BOOLEAN Register container as machine\n"
198 " --keep-unit Do not register a scope for the machine, reuse\n"
199 " the service unit nspawn is running in\n",
200 program_invocation_short_name);
205 static int parse_argv(int argc, char *argv[]) {
221 ARG_NETWORK_INTERFACE,
228 static const struct option options[] = {
229 { "help", no_argument, NULL, 'h' },
230 { "version", no_argument, NULL, ARG_VERSION },
231 { "directory", required_argument, NULL, 'D' },
232 { "user", required_argument, NULL, 'u' },
233 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
234 { "boot", no_argument, NULL, 'b' },
235 { "uuid", required_argument, NULL, ARG_UUID },
236 { "read-only", no_argument, NULL, ARG_READ_ONLY },
237 { "capability", required_argument, NULL, ARG_CAPABILITY },
238 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
239 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
240 { "bind", required_argument, NULL, ARG_BIND },
241 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
242 { "machine", required_argument, NULL, 'M' },
243 { "slice", required_argument, NULL, 'S' },
244 { "setenv", required_argument, NULL, ARG_SETENV },
245 { "selinux-context", required_argument, NULL, 'Z' },
246 { "selinux-apifs-context", required_argument, NULL, 'L' },
247 { "quiet", no_argument, NULL, 'q' },
248 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM },
249 { "register", required_argument, NULL, ARG_REGISTER },
250 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
251 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
252 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
253 { "network-veth", no_argument, NULL, ARG_NETWORK_VETH },
254 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
255 { "personality", required_argument, NULL, ARG_PERSONALITY },
256 { "image", required_argument, NULL, 'i' },
261 uint64_t plus = 0, minus = 0;
266 while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:", options, NULL)) >= 0) {
274 puts(PACKAGE_STRING);
275 puts(SYSTEMD_FEATURES);
280 arg_directory = canonicalize_file_name(optarg);
281 if (!arg_directory) {
282 log_error("Invalid root directory: %m");
294 arg_user = strdup(optarg);
300 case ARG_NETWORK_BRIDGE:
301 arg_network_bridge = optarg;
305 case ARG_NETWORK_VETH:
306 arg_network_veth = true;
307 arg_private_network = true;
310 case ARG_NETWORK_INTERFACE:
311 if (strv_extend(&arg_network_interfaces, optarg) < 0)
314 arg_private_network = true;
317 case ARG_NETWORK_MACVLAN:
318 if (strv_extend(&arg_network_macvlan, optarg) < 0)
323 case ARG_PRIVATE_NETWORK:
324 arg_private_network = true;
332 r = sd_id128_from_string(optarg, &arg_uuid);
334 log_error("Invalid UUID: %s", optarg);
344 if (isempty(optarg)) {
349 if (!hostname_is_valid(optarg)) {
350 log_error("Invalid machine name: %s", optarg);
355 arg_machine = strdup(optarg);
363 arg_selinux_context = optarg;
367 arg_selinux_apifs_context = optarg;
371 arg_read_only = true;
375 case ARG_DROP_CAPABILITY: {
379 FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
380 _cleanup_free_ char *t;
383 t = strndup(word, length);
387 if (streq(t, "all")) {
388 if (c == ARG_CAPABILITY)
389 plus = (uint64_t) -1;
391 minus = (uint64_t) -1;
393 if (cap_from_name(t, &cap) < 0) {
394 log_error("Failed to parse capability %s.", t);
398 if (c == ARG_CAPABILITY)
399 plus |= 1ULL << (uint64_t) cap;
401 minus |= 1ULL << (uint64_t) cap;
409 arg_link_journal = LINK_GUEST;
412 case ARG_LINK_JOURNAL:
413 if (streq(optarg, "auto"))
414 arg_link_journal = LINK_AUTO;
415 else if (streq(optarg, "no"))
416 arg_link_journal = LINK_NO;
417 else if (streq(optarg, "guest"))
418 arg_link_journal = LINK_GUEST;
419 else if (streq(optarg, "host"))
420 arg_link_journal = LINK_HOST;
422 log_error("Failed to parse link journal mode %s", optarg);
430 _cleanup_free_ char *a = NULL, *b = NULL;
434 x = c == ARG_BIND ? &arg_bind : &arg_bind_ro;
436 e = strchr(optarg, ':');
438 a = strndup(optarg, e - optarg);
448 if (!path_is_absolute(a) || !path_is_absolute(b)) {
449 log_error("Invalid bind mount specification: %s", optarg);
453 r = strv_extend(x, a);
457 r = strv_extend(x, b);
467 if (!env_assignment_is_valid(optarg)) {
468 log_error("Environment variable assignment '%s' is not valid.", optarg);
472 n = strv_env_set(arg_setenv, optarg);
476 strv_free(arg_setenv);
485 case ARG_SHARE_SYSTEM:
486 arg_share_system = true;
490 r = parse_boolean(optarg);
492 log_error("Failed to parse --register= argument: %s", optarg);
500 arg_keep_unit = true;
503 case ARG_PERSONALITY:
505 arg_personality = personality_from_string(optarg);
506 if (arg_personality == 0xffffffffLU) {
507 log_error("Unknown or unsupported personality '%s'.", optarg);
517 assert_not_reached("Unhandled option");
521 if (arg_share_system)
522 arg_register = false;
524 if (arg_boot && arg_share_system) {
525 log_error("--boot and --share-system may not be combined.");
529 if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) {
530 log_error("--keep-unit may not be used when invoked from a user session.");
534 if (arg_directory && arg_image) {
535 log_error("--directory= and --image= may not be combined.");
539 arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus;
544 static int mount_all(const char *dest) {
546 typedef struct MountPoint {
555 static const MountPoint mount_table[] = {
556 { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
557 { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */
558 { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
559 { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
560 { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
561 { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true },
562 { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
563 { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
565 { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
566 { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
573 for (k = 0; k < ELEMENTSOF(mount_table); k++) {
574 _cleanup_free_ char *where = NULL;
576 _cleanup_free_ char *options = NULL;
581 where = strjoin(dest, "/", mount_table[k].where, NULL);
585 t = path_is_mount_point(where, true);
587 log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
595 /* Skip this entry if it is not a remount. */
596 if (mount_table[k].what && t > 0)
599 mkdir_p(where, 0755);
602 if (arg_selinux_apifs_context &&
603 (streq_ptr(mount_table[k].what, "tmpfs") || streq_ptr(mount_table[k].what, "devpts"))) {
604 options = strjoin(mount_table[k].options, ",context=\"", arg_selinux_apifs_context, "\"", NULL);
611 o = mount_table[k].options;
614 if (mount(mount_table[k].what,
617 mount_table[k].flags,
619 mount_table[k].fatal) {
621 log_error("mount(%s) failed: %m", where);
631 static int mount_binds(const char *dest, char **l, unsigned long flags) {
634 STRV_FOREACH_PAIR(x, y, l) {
636 struct stat source_st, dest_st;
639 if (stat(*x, &source_st) < 0) {
640 log_error("Failed to stat %s: %m", *x);
644 where = strappenda(dest, *y);
645 r = stat(where, &dest_st);
647 if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) {
648 log_error("The file types of %s and %s do not match. Refusing bind mount",
652 } else if (errno == ENOENT) {
653 r = mkdir_parents_label(where, 0755);
655 log_error("Failed to bind mount %s: %s", *x, strerror(-r));
659 log_error("Failed to bind mount %s: %s", *x, strerror(errno));
662 /* Create the mount point, but be conservative -- refuse to create block
663 * and char devices. */
664 if (S_ISDIR(source_st.st_mode))
665 mkdir_label(where, 0755);
666 else if (S_ISFIFO(source_st.st_mode))
668 else if (S_ISSOCK(source_st.st_mode))
669 mknod(where, 0644 | S_IFSOCK, 0);
670 else if (S_ISREG(source_st.st_mode))
673 log_error("Refusing to create mountpoint for file: %s", *x);
677 if (mount(*x, where, "bind", MS_BIND, NULL) < 0) {
678 log_error("mount(%s) failed: %m", where);
682 if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) {
683 log_error("mount(%s) failed: %m", where);
691 static int setup_timezone(const char *dest) {
692 _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
698 /* Fix the timezone, if possible */
699 r = readlink_malloc("/etc/localtime", &p);
701 log_warning("/etc/localtime is not a symlink, not updating container timezone.");
705 z = path_startswith(p, "../usr/share/zoneinfo/");
707 z = path_startswith(p, "/usr/share/zoneinfo/");
709 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
713 where = strappend(dest, "/etc/localtime");
717 r = readlink_malloc(where, &q);
719 y = path_startswith(q, "../usr/share/zoneinfo/");
721 y = path_startswith(q, "/usr/share/zoneinfo/");
724 /* Already pointing to the right place? Then do nothing .. */
725 if (y && streq(y, z))
729 check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
733 if (access(check, F_OK) < 0) {
734 log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
738 what = strappend("../usr/share/zoneinfo/", z);
743 if (symlink(what, where) < 0) {
744 log_error("Failed to correct timezone of container: %m");
751 static int setup_resolv_conf(const char *dest) {
752 char _cleanup_free_ *where = NULL;
756 if (arg_private_network)
759 /* Fix resolv.conf, if possible */
760 where = strappend(dest, "/etc/resolv.conf");
764 /* We don't really care for the results of this really. If it
765 * fails, it fails, but meh... */
766 copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW);
771 static int setup_boot_id(const char *dest) {
772 _cleanup_free_ char *from = NULL, *to = NULL;
779 if (arg_share_system)
782 /* Generate a new randomized boot ID, so that each boot-up of
783 * the container gets a new one */
785 from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
786 to = strappend(dest, "/proc/sys/kernel/random/boot_id");
790 r = sd_id128_randomize(&rnd);
792 log_error("Failed to generate random boot id: %s", strerror(-r));
796 snprintf(as_uuid, sizeof(as_uuid),
797 "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
798 SD_ID128_FORMAT_VAL(rnd));
799 char_array_0(as_uuid);
801 r = write_string_file(from, as_uuid);
803 log_error("Failed to write boot id: %s", strerror(-r));
807 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
808 log_error("Failed to bind mount boot id: %m");
810 } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL))
811 log_warning("Failed to make boot id read-only: %m");
817 static int copy_devnodes(const char *dest) {
819 static const char devnodes[] =
829 _cleanup_umask_ mode_t u;
835 NULSTR_FOREACH(d, devnodes) {
836 _cleanup_free_ char *from = NULL, *to = NULL;
839 from = strappend("/dev/", d);
840 to = strjoin(dest, "/dev/", d, NULL);
844 if (stat(from, &st) < 0) {
846 if (errno != ENOENT) {
847 log_error("Failed to stat %s: %m", from);
851 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
853 log_error("%s is not a char or block device, cannot copy", from);
856 } else if (mknod(to, st.st_mode, st.st_rdev) < 0) {
858 log_error("mknod(%s) failed: %m", dest);
866 static int setup_ptmx(const char *dest) {
867 _cleanup_free_ char *p = NULL;
869 p = strappend(dest, "/dev/ptmx");
873 if (symlink("pts/ptmx", p) < 0) {
874 log_error("Failed to create /dev/ptmx symlink: %m");
881 static int setup_dev_console(const char *dest, const char *console) {
882 _cleanup_umask_ mode_t u;
892 if (stat("/dev/null", &st) < 0) {
893 log_error("Failed to stat /dev/null: %m");
897 r = chmod_and_chown(console, 0600, 0, 0);
899 log_error("Failed to correct access mode for TTY: %s", strerror(-r));
903 /* We need to bind mount the right tty to /dev/console since
904 * ptys can only exist on pts file systems. To have something
905 * to bind mount things on we create a device node first, and
906 * use /dev/null for that since we the cgroups device policy
907 * allows us to create that freely, while we cannot create
908 * /dev/console. (Note that the major minor doesn't actually
909 * matter here, since we mount it over anyway). */
911 to = strappenda(dest, "/dev/console");
912 if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
913 log_error("mknod() for /dev/console failed: %m");
917 if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
918 log_error("Bind mount for /dev/console failed: %m");
925 static int setup_kmsg(const char *dest, int kmsg_socket) {
926 _cleanup_free_ char *from = NULL, *to = NULL;
928 _cleanup_umask_ mode_t u;
930 struct cmsghdr cmsghdr;
931 uint8_t buf[CMSG_SPACE(sizeof(int))];
934 .msg_control = &control,
935 .msg_controllen = sizeof(control),
937 struct cmsghdr *cmsg;
940 assert(kmsg_socket >= 0);
944 /* We create the kmsg FIFO as /dev/kmsg, but immediately
945 * delete it after bind mounting it to /proc/kmsg. While FIFOs
946 * on the reading side behave very similar to /proc/kmsg,
947 * their writing side behaves differently from /dev/kmsg in
948 * that writing blocks when nothing is reading. In order to
949 * avoid any problems with containers deadlocking due to this
950 * we simply make /dev/kmsg unavailable to the container. */
951 if (asprintf(&from, "%s/dev/kmsg", dest) < 0 ||
952 asprintf(&to, "%s/proc/kmsg", dest) < 0)
955 if (mkfifo(from, 0600) < 0) {
956 log_error("mkfifo() for /dev/kmsg failed: %m");
960 r = chmod_and_chown(from, 0600, 0, 0);
962 log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
966 if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
967 log_error("Bind mount for /proc/kmsg failed: %m");
971 fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
973 log_error("Failed to open fifo: %m");
977 cmsg = CMSG_FIRSTHDR(&mh);
978 cmsg->cmsg_level = SOL_SOCKET;
979 cmsg->cmsg_type = SCM_RIGHTS;
980 cmsg->cmsg_len = CMSG_LEN(sizeof(int));
981 memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
983 mh.msg_controllen = cmsg->cmsg_len;
985 /* Store away the fd in the socket, so that it stays open as
986 * long as we run the child */
987 k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL);
988 close_nointr_nofail(fd);
991 log_error("Failed to send FIFO fd: %m");
995 /* And now make the FIFO unavailable as /dev/kmsg... */
1000 static int setup_hostname(void) {
1002 if (arg_share_system)
1005 if (sethostname(arg_machine, strlen(arg_machine)) < 0)
1011 static int setup_journal(const char *directory) {
1012 sd_id128_t machine_id, this_id;
1013 _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL;
1017 p = strappend(directory, "/etc/machine-id");
1021 r = read_one_line_file(p, &b);
1022 if (r == -ENOENT && arg_link_journal == LINK_AUTO)
1025 log_error("Failed to read machine ID from %s: %s", p, strerror(-r));
1030 if (isempty(id) && arg_link_journal == LINK_AUTO)
1033 /* Verify validity */
1034 r = sd_id128_from_string(id, &machine_id);
1036 log_error("Failed to parse machine ID from %s: %s", p, strerror(-r));
1040 r = sd_id128_get_machine(&this_id);
1042 log_error("Failed to retrieve machine ID: %s", strerror(-r));
1046 if (sd_id128_equal(machine_id, this_id)) {
1047 log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
1048 "Host and machine ids are equal (%s): refusing to link journals", id);
1049 if (arg_link_journal == LINK_AUTO)
1055 if (arg_link_journal == LINK_NO)
1059 p = strappend("/var/log/journal/", id);
1060 q = strjoin(directory, "/var/log/journal/", id, NULL);
1064 if (path_is_mount_point(p, false) > 0) {
1065 if (arg_link_journal != LINK_AUTO) {
1066 log_error("%s: already a mount point, refusing to use for journal", p);
1073 if (path_is_mount_point(q, false) > 0) {
1074 if (arg_link_journal != LINK_AUTO) {
1075 log_error("%s: already a mount point, refusing to use for journal", q);
1082 r = readlink_and_make_absolute(p, &d);
1084 if ((arg_link_journal == LINK_GUEST ||
1085 arg_link_journal == LINK_AUTO) &&
1088 r = mkdir_p(q, 0755);
1090 log_warning("failed to create directory %s: %m", q);
1094 if (unlink(p) < 0) {
1095 log_error("Failed to remove symlink %s: %m", p);
1098 } else if (r == -EINVAL) {
1100 if (arg_link_journal == LINK_GUEST &&
1103 if (errno == ENOTDIR) {
1104 log_error("%s already exists and is neither a symlink nor a directory", p);
1107 log_error("Failed to remove %s: %m", p);
1111 } else if (r != -ENOENT) {
1112 log_error("readlink(%s) failed: %m", p);
1116 if (arg_link_journal == LINK_GUEST) {
1118 if (symlink(q, p) < 0) {
1119 log_error("Failed to symlink %s to %s: %m", q, p);
1123 r = mkdir_p(q, 0755);
1125 log_warning("failed to create directory %s: %m", q);
1129 if (arg_link_journal == LINK_HOST) {
1130 r = mkdir_p(p, 0755);
1132 log_error("Failed to create %s: %m", p);
1136 } else if (access(p, F_OK) < 0)
1139 if (dir_is_empty(q) == 0) {
1140 log_error("%s not empty.", q);
1144 r = mkdir_p(q, 0755);
1146 log_error("Failed to create %s: %m", q);
1150 if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
1151 log_error("Failed to bind mount journal from host into guest: %m");
1158 static int setup_kdbus(const char *dest, const char *path) {
1164 p = strappenda(dest, "/dev/kdbus");
1165 if (mkdir(p, 0755) < 0) {
1166 log_error("Failed to create kdbus path: %m");
1170 if (mount(path, p, "bind", MS_BIND, NULL) < 0) {
1171 log_error("Failed to mount kdbus domain path: %m");
1178 static int drop_capabilities(void) {
1179 return capability_bounding_set_drop(~arg_retain, false);
1182 static int register_machine(pid_t pid) {
1183 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1184 _cleanup_bus_unref_ sd_bus *bus = NULL;
1190 r = sd_bus_default_system(&bus);
1192 log_error("Failed to open system bus: %s", strerror(-r));
1196 if (arg_keep_unit) {
1197 r = sd_bus_call_method(
1199 "org.freedesktop.machine1",
1200 "/org/freedesktop/machine1",
1201 "org.freedesktop.machine1.Manager",
1207 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1211 strempty(arg_directory));
1213 _cleanup_bus_message_unref_ sd_bus_message *m = NULL;
1215 r = sd_bus_message_new_method_call(
1218 "org.freedesktop.machine1",
1219 "/org/freedesktop/machine1",
1220 "org.freedesktop.machine1.Manager",
1223 log_error("Failed to create message: %s", strerror(-r));
1227 r = sd_bus_message_append(
1231 SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
1235 strempty(arg_directory));
1237 log_error("Failed to append message arguments: %s", strerror(-r));
1241 r = sd_bus_message_open_container(m, 'a', "(sv)");
1243 log_error("Failed to open container: %s", strerror(-r));
1247 if (!isempty(arg_slice)) {
1248 r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice);
1250 log_error("Failed to append slice: %s", strerror(-r));
1255 r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict");
1257 log_error("Failed to add device policy: %s", strerror(-r));
1261 r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 8,
1262 /* Allow the container to
1263 * access and create the API
1264 * device nodes, so that
1265 * PrivateDevices= in the
1266 * container can work
1271 "/dev/random", "rwm",
1272 "/dev/urandom", "rwm",
1274 /* Allow the container
1275 * access to ptys. However,
1277 * container to ever create
1278 * these device nodes. */
1279 "/dev/pts/ptmx", "rw",
1282 log_error("Failed to add device whitelist: %s", strerror(-r));
1286 r = sd_bus_message_close_container(m);
1288 log_error("Failed to close container: %s", strerror(-r));
1292 r = sd_bus_call(bus, m, 0, &error, NULL);
1296 log_error("Failed to register machine: %s", bus_error_message(&error, r));
1303 static int terminate_machine(pid_t pid) {
1304 _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
1305 _cleanup_bus_message_unref_ sd_bus_message *reply = NULL;
1306 _cleanup_bus_unref_ sd_bus *bus = NULL;
1313 r = sd_bus_default_system(&bus);
1315 log_error("Failed to open system bus: %s", strerror(-r));
1319 r = sd_bus_call_method(
1321 "org.freedesktop.machine1",
1322 "/org/freedesktop/machine1",
1323 "org.freedesktop.machine1.Manager",
1330 /* Note that the machine might already have been
1331 * cleaned up automatically, hence don't consider it a
1332 * failure if we cannot get the machine object. */
1333 log_debug("Failed to get machine: %s", bus_error_message(&error, r));
1337 r = sd_bus_message_read(reply, "o", &path);
1339 return bus_log_parse_error(r);
1341 r = sd_bus_call_method(
1343 "org.freedesktop.machine1",
1345 "org.freedesktop.machine1.Machine",
1351 log_debug("Failed to terminate machine: %s", bus_error_message(&error, r));
1358 static int reset_audit_loginuid(void) {
1359 _cleanup_free_ char *p = NULL;
1362 if (arg_share_system)
1365 r = read_one_line_file("/proc/self/loginuid", &p);
1369 log_error("Failed to read /proc/self/loginuid: %s", strerror(-r));
1373 /* Already reset? */
1374 if (streq(p, "4294967295"))
1377 r = write_string_file("/proc/self/loginuid", "4294967295");
1379 log_error("Failed to reset audit login UID. This probably means that your kernel is too\n"
1380 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
1381 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
1382 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
1383 "using systemd-nspawn. Sleeping for 5s... (%s)\n", strerror(-r));
1391 static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) {
1392 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1393 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1396 if (!arg_private_network)
1399 if (!arg_network_veth)
1402 /* Use two different interface name prefixes depending whether
1403 * we are in bridge mode or not. */
1404 if (arg_network_bridge)
1405 memcpy(iface_name, "vb-", 3);
1407 memcpy(iface_name, "ve-", 3);
1409 strncpy(iface_name+3, arg_machine, IFNAMSIZ - 3);
1411 r = sd_rtnl_open(&rtnl, 0);
1413 log_error("Failed to connect to netlink: %s", strerror(-r));
1417 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
1419 log_error("Failed to allocate netlink message: %s", strerror(-r));
1423 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name);
1425 log_error("Failed to add netlink interface name: %s", strerror(-r));
1429 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
1431 log_error("Failed to open netlink container: %s", strerror(-r));
1435 r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth");
1437 log_error("Failed to append netlink kind: %s", strerror(-r));
1441 r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA);
1443 log_error("Failed to open netlink container: %s", strerror(-r));
1447 r = sd_rtnl_message_open_container(m, VETH_INFO_PEER);
1449 log_error("Failed to open netlink container: %s", strerror(-r));
1453 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0");
1455 log_error("Failed to add netlink interface name: %s", strerror(-r));
1459 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1461 log_error("Failed to add netlink namespace field: %s", strerror(-r));
1465 r = sd_rtnl_message_close_container(m);
1467 log_error("Failed to close netlink container: %s", strerror(-r));
1471 r = sd_rtnl_message_close_container(m);
1473 log_error("Failed to close netlink container: %s", strerror(-r));
1477 r = sd_rtnl_message_close_container(m);
1479 log_error("Failed to close netlink container: %s", strerror(-r));
1483 r = sd_rtnl_call(rtnl, m, 0, NULL);
1485 log_error("Failed to add new veth interfaces: %s", strerror(-r));
1492 static int setup_bridge(const char veth_name[]) {
1493 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1494 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1497 if (!arg_private_network)
1500 if (!arg_network_veth)
1503 if (!arg_network_bridge)
1506 bridge = (int) if_nametoindex(arg_network_bridge);
1508 log_error("Failed to resolve interface %s: %m", arg_network_bridge);
1512 r = sd_rtnl_open(&rtnl, 0);
1514 log_error("Failed to connect to netlink: %s", strerror(-r));
1518 r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0);
1520 log_error("Failed to allocate netlink message: %s", strerror(-r));
1524 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name);
1526 log_error("Failed to add netlink interface name field: %s", strerror(-r));
1530 r = sd_rtnl_message_append_u32(m, IFLA_MASTER, bridge);
1532 log_error("Failed to add netlink master field: %s", strerror(-r));
1536 r = sd_rtnl_call(rtnl, m, 0, NULL);
1538 log_error("Failed to add veth interface to bridge: %s", strerror(-r));
1545 static int parse_interface(struct udev *udev, const char *name) {
1546 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
1547 char ifi_str[2 + DECIMAL_STR_MAX(int)];
1550 ifi = (int) if_nametoindex(name);
1552 log_error("Failed to resolve interface %s: %m", name);
1556 sprintf(ifi_str, "n%i", ifi);
1557 d = udev_device_new_from_device_id(udev, ifi_str);
1559 log_error("Failed to get udev device for interface %s: %m", name);
1563 if (udev_device_get_is_initialized(d) <= 0) {
1564 log_error("Network interface %s is not initialized yet.", name);
1571 static int move_network_interfaces(pid_t pid) {
1572 _cleanup_udev_unref_ struct udev *udev = NULL;
1573 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1577 if (!arg_private_network)
1580 if (strv_isempty(arg_network_interfaces))
1583 r = sd_rtnl_open(&rtnl, 0);
1585 log_error("Failed to connect to netlink: %s", strerror(-r));
1591 log_error("Failed to connect to udev.");
1595 STRV_FOREACH(i, arg_network_interfaces) {
1596 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1599 ifi = parse_interface(udev, *i);
1603 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, ifi);
1605 log_error("Failed to allocate netlink message: %s", strerror(-r));
1609 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1611 log_error("Failed to append namespace PID to netlink message: %s", strerror(-r));
1615 r = sd_rtnl_call(rtnl, m, 0, NULL);
1617 log_error("Failed to move interface %s to namespace: %s", *i, strerror(-r));
1625 static int setup_macvlan(pid_t pid) {
1626 _cleanup_udev_unref_ struct udev *udev = NULL;
1627 _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
1631 if (!arg_private_network)
1634 if (strv_isempty(arg_network_macvlan))
1637 r = sd_rtnl_open(&rtnl, 0);
1639 log_error("Failed to connect to netlink: %s", strerror(-r));
1645 log_error("Failed to connect to udev.");
1649 STRV_FOREACH(i, arg_network_macvlan) {
1650 _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
1651 _cleanup_free_ char *n = NULL;
1654 ifi = parse_interface(udev, *i);
1658 r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
1660 log_error("Failed to allocate netlink message: %s", strerror(-r));
1664 r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi);
1666 log_error("Failed to add netlink interface index: %s", strerror(-r));
1670 n = strappend("mv-", *i);
1674 strshorten(n, IFNAMSIZ-1);
1676 r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n);
1678 log_error("Failed to add netlink interface name: %s", strerror(-r));
1682 r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
1684 log_error("Failed to add netlink namespace field: %s", strerror(-r));
1688 r = sd_rtnl_message_open_container(m, IFLA_LINKINFO);
1690 log_error("Failed to open netlink container: %s", strerror(-r));
1694 r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "macvlan");
1696 log_error("Failed to append netlink kind: %s", strerror(-r));
1700 r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA);
1702 log_error("Failed to open netlink container: %s", strerror(-r));
1706 r = sd_rtnl_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE);
1708 log_error("Failed to append macvlan mode: %s", strerror(-r));
1712 r = sd_rtnl_message_close_container(m);
1714 log_error("Failed to close netlink container: %s", strerror(-r));
1718 r = sd_rtnl_message_close_container(m);
1720 log_error("Failed to close netlink container: %s", strerror(-r));
1724 r = sd_rtnl_call(rtnl, m, 0, NULL);
1726 log_error("Failed to add new macvlan interfaces: %s", strerror(-r));
1734 static int audit_still_doesnt_work_in_containers(void) {
1737 scmp_filter_ctx seccomp;
1741 Audit is broken in containers, much of the userspace audit
1742 hookup will fail if running inside a container. We don't
1743 care and just turn off creation of audit sockets.
1745 This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail
1746 with EAFNOSUPPORT which audit userspace uses as indication
1747 that audit is disabled in the kernel.
1750 seccomp = seccomp_init(SCMP_ACT_ALLOW);
1754 r = seccomp_add_secondary_archs(seccomp);
1756 log_error("Failed to add secondary archs to seccomp filter: %s", strerror(-r));
1760 r = seccomp_rule_add(
1762 SCMP_ACT_ERRNO(EAFNOSUPPORT),
1765 SCMP_A0(SCMP_CMP_EQ, AF_NETLINK),
1766 SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT));
1768 log_error("Failed to add audit seccomp rule: %s", strerror(-r));
1772 r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
1774 log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r));
1778 r = seccomp_load(seccomp);
1780 log_error("Failed to install seccomp audit filter: %s", strerror(-r));
1783 seccomp_release(seccomp);
1791 static int setup_image(char **device_path, int *loop_nr) {
1792 struct loop_info64 info = {
1793 .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN
1795 _cleanup_close_ int fd = -1, control = -1, loop = -1;
1796 _cleanup_free_ char* loopdev = NULL;
1800 assert(device_path);
1803 fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
1805 log_error("Failed to open %s: %m", arg_image);
1809 if (fstat(fd, &st) < 0) {
1810 log_error("Failed to stat %s: %m", arg_image);
1814 if (S_ISBLK(st.st_mode)) {
1817 p = strdup(arg_image);
1831 if (!S_ISREG(st.st_mode)) {
1832 log_error("%s is not a regular file or block device: %m", arg_image);
1836 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
1838 log_error("Failed to open /dev/loop-control: %m");
1842 nr = ioctl(control, LOOP_CTL_GET_FREE);
1844 log_error("Failed to allocate loop device: %m");
1848 if (asprintf(&loopdev, "/dev/loop%i", nr) < 0)
1851 loop = open(loopdev, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY);
1853 log_error("Failed to open loop device %s: %m", loopdev);
1857 if (ioctl(loop, LOOP_SET_FD, fd) < 0) {
1858 log_error("Failed to set loopback file descriptor on %s: %m", loopdev);
1863 info.lo_flags |= LO_FLAGS_READ_ONLY;
1865 if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) {
1866 log_error("Failed to set loopback settings on %s: %m", loopdev);
1870 *device_path = loopdev;
1881 static int dissect_image(
1889 int home_nr = -1, root_nr = -1, secondary_root_nr = -1, srv_nr = -1;
1890 _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL;
1891 _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL;
1892 _cleanup_udev_device_unref_ struct udev_device *d = NULL;
1893 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
1894 _cleanup_udev_unref_ struct udev *udev = NULL;
1895 struct udev_list_entry *first, *item;
1896 const char *pttype = NULL;
1902 assert(root_device);
1903 assert(home_device);
1907 b = blkid_new_probe();
1912 r = blkid_probe_set_device(b, fd, 0, 0);
1917 log_error("Failed to set device on blkid probe: %m");
1921 blkid_probe_enable_partitions(b, 1);
1922 blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
1925 r = blkid_do_safeprobe(b);
1926 if (r == -2 || r == 1) {
1927 log_error("Failed to identify any partition table on %s.\n"
1928 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
1930 } else if (r != 0) {
1933 log_error("Failed to probe: %m");
1937 blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
1938 if (!streq_ptr(pttype, "gpt")) {
1939 log_error("Image %s does not carry a GUID Partition Table.\n"
1940 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
1945 pl = blkid_probe_get_partitions(b);
1950 log_error("Failed to list partitions of %s", arg_image);
1958 if (fstat(fd, &st) < 0) {
1959 log_error("Failed to stat block device: %m");
1963 d = udev_device_new_from_devnum(udev, 'b', st.st_rdev);
1967 e = udev_enumerate_new(udev);
1971 r = udev_enumerate_add_match_parent(e, d);
1975 r = udev_enumerate_scan_devices(e);
1977 log_error("Failed to scan for partition devices of %s: %s", arg_image, strerror(-r));
1981 first = udev_enumerate_get_list_entry(e);
1982 udev_list_entry_foreach(item, first) {
1983 _cleanup_udev_device_unref_ struct udev_device *q;
1984 const char *stype, *node;
1991 q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));
1996 log_error("Failed to get partition device of %s: %m", arg_image);
2000 qn = udev_device_get_devnum(q);
2004 if (st.st_rdev == qn)
2007 node = udev_device_get_devnode(q);
2011 pp = blkid_partlist_devno_to_partition(pl, qn);
2015 nr = blkid_partition_get_partno(pp);
2019 stype = blkid_partition_get_type_string(pp);
2023 if (sd_id128_from_string(stype, &type_id) < 0)
2026 if (sd_id128_equal(type_id, GPT_HOME)) {
2028 if (home && nr >= home_nr)
2033 home = strdup(node);
2036 } else if (sd_id128_equal(type_id, GPT_SRV)) {
2038 if (srv && nr >= srv_nr)
2047 #ifdef GPT_ROOT_NATIVE
2048 else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) {
2050 if (root && nr >= root_nr)
2055 root = strdup(node);
2060 #ifdef GPT_ROOT_SECONDARY
2061 else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) {
2063 if (secondary_root && nr >= secondary_root_nr)
2066 secondary_root_nr = nr;
2067 free(secondary_root);
2068 secondary_root = strdup(node);
2069 if (!secondary_root)
2075 if (!root && !secondary_root) {
2076 log_error("Failed to identify root partition in disk image %s.\n"
2077 "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image);
2082 *root_device = root;
2085 } else if (secondary_root) {
2086 *root_device = secondary_root;
2087 secondary_root = NULL;
2092 *home_device = home;
2103 log_error("--image= is not supported, compiled without blkid support.");
2108 static int mount_device(const char *what, const char *where, const char *directory) {
2110 _cleanup_blkid_free_probe_ blkid_probe b = NULL;
2111 const char *fstype, *p;
2118 p = strappenda(where, directory);
2123 b = blkid_new_probe_from_filename(what);
2127 log_error("Failed to allocate prober for %s: %m", what);
2131 blkid_probe_enable_superblocks(b, 1);
2132 blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE);
2135 r = blkid_do_safeprobe(b);
2136 if (r == -1 || r == 1) {
2137 log_error("Cannot determine file system type of %s", what);
2139 } else if (r != 0) {
2142 log_error("Failed to probe %s: %m", what);
2147 if (blkid_probe_lookup_value(b, "TYPE", &fstype, NULL) < 0) {
2150 log_error("Failed to determine file system type of %s", what);
2154 if (streq(fstype, "crypto_LUKS")) {
2155 log_error("nspawn currently does not support LUKS disk images.");
2159 if (mount(what, p, fstype, arg_read_only ? MS_NODEV|MS_RDONLY : 0, NULL) < 0) {
2160 log_error("Failed to mount %s: %m", what);
2166 log_error("--image= is not supported, compiled without blkid support.");
2171 static int mount_devices(const char *where, const char *root_device, const char *home_device, const char *srv_device) {
2177 r = mount_device(root_device, arg_directory, NULL);
2179 log_error("Failed to mount root directory: %s", strerror(-r));
2185 r = mount_device(home_device, arg_directory, "/home");
2187 log_error("Failed to mount home directory: %s", strerror(-r));
2193 r = mount_device(srv_device, arg_directory, "/srv");
2195 log_error("Failed to mount server data directory: %s", strerror(-r));
2203 static void loop_remove(int nr, int *image_fd) {
2204 _cleanup_close_ int control = -1;
2209 if (image_fd && *image_fd >= 0) {
2210 ioctl(*image_fd, LOOP_CLR_FD);
2211 close_nointr_nofail(*image_fd);
2215 control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
2219 ioctl(control, LOOP_CTL_REMOVE, nr);
2222 static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
2230 if (pipe2(pipe_fds, O_CLOEXEC) < 0) {
2231 log_error("Failed to allocate pipe: %m");
2237 log_error("Failed to fork getent child: %m");
2239 } else if (pid == 0) {
2241 char *empty_env = NULL;
2243 if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0)
2244 _exit(EXIT_FAILURE);
2246 if (pipe_fds[0] > 2)
2247 close_nointr_nofail(pipe_fds[0]);
2248 if (pipe_fds[1] > 2)
2249 close_nointr_nofail(pipe_fds[1]);
2251 nullfd = open("/dev/null", O_RDWR);
2253 _exit(EXIT_FAILURE);
2255 if (dup3(nullfd, STDIN_FILENO, 0) < 0)
2256 _exit(EXIT_FAILURE);
2258 if (dup3(nullfd, STDERR_FILENO, 0) < 0)
2259 _exit(EXIT_FAILURE);
2262 close_nointr_nofail(nullfd);
2264 reset_all_signal_handlers();
2265 close_all_fds(NULL, 0);
2267 execle("/usr/bin/getent", "getenv", database, key, NULL, &empty_env);
2268 execle("/usr/getent", "getenv", database, key, NULL, &empty_env);
2269 _exit(EXIT_FAILURE);
2272 close_nointr_nofail(pipe_fds[1]);
2280 static int change_uid_gid(char **_home) {
2282 _cleanup_strv_free_ char **passwd = NULL;
2283 char line[LINE_MAX], *w, *x, *state, *u, *g, *h;
2284 _cleanup_free_ uid_t *uids = NULL;
2285 _cleanup_free_ char *home = NULL;
2286 _cleanup_fclose_ FILE *f = NULL;
2287 _cleanup_close_ int fd = -1;
2288 unsigned n_uids = 0;
2297 if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) {
2298 /* Reset everything fully to 0, just in case */
2300 if (setgroups(0, NULL) < 0) {
2301 log_error("setgroups() failed: %m");
2305 if (setresgid(0, 0, 0) < 0) {
2306 log_error("setregid() failed: %m");
2310 if (setresuid(0, 0, 0) < 0) {
2311 log_error("setreuid() failed: %m");
2319 /* First, get user credentials */
2320 fd = spawn_getent("passwd", arg_user, &pid);
2324 f = fdopen(fd, "r");
2329 if (!fgets(line, sizeof(line), f)) {
2332 log_error("Failed to resolve user %s.", arg_user);
2336 log_error("Failed to read from getent: %m");
2342 wait_for_terminate_and_warn("getent passwd", pid);
2344 x = strchr(line, ':');
2346 log_error("/etc/passwd entry has invalid user field.");
2350 u = strchr(x+1, ':');
2352 log_error("/etc/passwd entry has invalid password field.");
2359 log_error("/etc/passwd entry has invalid UID field.");
2367 log_error("/etc/passwd entry has invalid GID field.");
2372 h = strchr(x+1, ':');
2374 log_error("/etc/passwd entry has invalid GECOS field.");
2381 log_error("/etc/passwd entry has invalid home directory field.");
2387 r = parse_uid(u, &uid);
2389 log_error("Failed to parse UID of user.");
2393 r = parse_gid(g, &gid);
2395 log_error("Failed to parse GID of user.");
2403 /* Second, get group memberships */
2404 fd = spawn_getent("initgroups", arg_user, &pid);
2409 f = fdopen(fd, "r");
2414 if (!fgets(line, sizeof(line), f)) {
2416 log_error("Failed to resolve user %s.", arg_user);
2420 log_error("Failed to read from getent: %m");
2426 wait_for_terminate_and_warn("getent initgroups", pid);
2428 /* Skip over the username and subsequent separator whitespace */
2430 x += strcspn(x, WHITESPACE);
2431 x += strspn(x, WHITESPACE);
2433 FOREACH_WORD(w, l, x, state) {
2439 if (!GREEDY_REALLOC(uids, sz, n_uids+1))
2442 r = parse_uid(c, &uids[n_uids++]);
2444 log_error("Failed to parse group data from getent.");
2449 r = mkdir_parents(home, 0775);
2451 log_error("Failed to make home root directory: %s", strerror(-r));
2455 r = mkdir_safe(home, 0755, uid, gid);
2457 log_error("Failed to make home directory: %s", strerror(-r));
2461 fchown(STDIN_FILENO, uid, gid);
2462 fchown(STDOUT_FILENO, uid, gid);
2463 fchown(STDERR_FILENO, uid, gid);
2465 if (setgroups(n_uids, uids) < 0) {
2466 log_error("Failed to set auxiliary groups: %m");
2470 if (setresgid(gid, gid, gid) < 0) {
2471 log_error("setregid() failed: %m");
2475 if (setresuid(uid, uid, uid) < 0) {
2476 log_error("setreuid() failed: %m");
2488 int main(int argc, char *argv[]) {
2490 _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL;
2491 _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1;
2492 _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 };
2493 _cleanup_fdset_free_ FDSet *fds = NULL;
2494 int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1;
2495 const char *console = NULL;
2496 char veth_name[IFNAMSIZ];
2497 bool secondary = false;
2501 log_parse_environment();
2504 k = parse_argv(argc, argv);
2513 if (arg_directory) {
2516 p = path_make_absolute_cwd(arg_directory);
2517 free(arg_directory);
2520 arg_directory = get_current_dir_name();
2522 if (!arg_directory) {
2523 log_error("Failed to determine path, please use -D.");
2526 path_kill_slashes(arg_directory);
2530 arg_machine = strdup(basename(arg_image ? arg_image : arg_directory));
2536 hostname_cleanup(arg_machine, false);
2537 if (isempty(arg_machine)) {
2538 log_error("Failed to determine machine name automatically, please use -M.");
2543 if (geteuid() != 0) {
2544 log_error("Need to be root.");
2548 if (sd_booted() <= 0) {
2549 log_error("Not running on a systemd system.");
2554 n_fd_passed = sd_listen_fds(false);
2555 if (n_fd_passed > 0) {
2556 k = fdset_new_listen_fds(&fds, false);
2558 log_error("Failed to collect file descriptors: %s", strerror(-k));
2562 fdset_close_others(fds);
2565 if (arg_directory) {
2566 if (path_equal(arg_directory, "/")) {
2567 log_error("Spawning container on root directory not supported.");
2572 if (path_is_os_tree(arg_directory) <= 0) {
2573 log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory);
2579 p = strappenda(arg_directory,
2580 argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/");
2581 if (access(p, F_OK) < 0) {
2582 log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory);
2588 char template[] = "/tmp/nspawn-root-XXXXXX";
2590 if (!mkdtemp(template)) {
2591 log_error("Failed to create temporary directory: %m");
2596 arg_directory = strdup(template);
2597 if (!arg_directory) {
2602 image_fd = setup_image(&device_path, &loop_nr);
2608 r = dissect_image(image_fd, &root_device, &home_device, &srv_device, &secondary);
2613 master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
2615 log_error("Failed to acquire pseudo tty: %m");
2619 console = ptsname(master);
2621 log_error("Failed to determine tty name: %m");
2626 log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_image ? arg_image : arg_directory);
2628 if (unlockpt(master) < 0) {
2629 log_error("Failed to unlock tty: %m");
2633 if (access("/dev/kdbus/control", F_OK) >= 0) {
2635 if (arg_share_system) {
2636 kdbus_domain = strdup("/dev/kdbus");
2637 if (!kdbus_domain) {
2644 ns = strappenda("machine-", arg_machine);
2645 kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain);
2647 log_debug("Failed to create kdbus domain: %s", strerror(-r));
2649 log_debug("Successfully created kdbus domain as %s", kdbus_domain);
2653 if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
2654 log_error("Failed to create kmsg socket pair: %m");
2658 sd_notify(0, "READY=1");
2660 assert_se(sigemptyset(&mask) == 0);
2661 sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
2662 assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
2665 int parent_ready_fd = -1, child_ready_fd = -1;
2669 parent_ready_fd = eventfd(0, EFD_CLOEXEC);
2670 if (parent_ready_fd < 0) {
2671 log_error("Failed to create event fd: %m");
2675 child_ready_fd = eventfd(0, EFD_CLOEXEC);
2676 if (child_ready_fd < 0) {
2677 log_error("Failed to create event fd: %m");
2681 pid = syscall(__NR_clone,
2682 SIGCHLD|CLONE_NEWNS|
2683 (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)|
2684 (arg_private_network ? CLONE_NEWNET : 0), NULL);
2686 if (errno == EINVAL)
2687 log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
2689 log_error("clone() failed: %m");
2696 _cleanup_free_ char *home = NULL;
2698 const char *envp[] = {
2699 "PATH=" DEFAULT_PATH_SPLIT_USR,
2700 "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
2705 NULL, /* container_uuid */
2706 NULL, /* LISTEN_FDS */
2707 NULL, /* LISTEN_PID */
2712 envp[n_env] = strv_find_prefix(environ, "TERM=");
2716 close_nointr_nofail(master);
2719 close_nointr(STDIN_FILENO);
2720 close_nointr(STDOUT_FILENO);
2721 close_nointr(STDERR_FILENO);
2723 close_nointr_nofail(kmsg_socket_pair[0]);
2724 kmsg_socket_pair[0] = -1;
2726 reset_all_signal_handlers();
2728 assert_se(sigemptyset(&mask) == 0);
2729 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
2731 k = open_terminal(console, O_RDWR);
2732 if (k != STDIN_FILENO) {
2734 close_nointr_nofail(k);
2738 log_error("Failed to open console: %s", strerror(-k));
2742 if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
2743 dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
2744 log_error("Failed to duplicate console: %m");
2749 log_error("setsid() failed: %m");
2753 if (reset_audit_loginuid() < 0)
2756 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
2757 log_error("PR_SET_PDEATHSIG failed: %m");
2761 /* Mark everything as slave, so that we still
2762 * receive mounts from the real root, but don't
2763 * propagate mounts to the real root. */
2764 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
2765 log_error("MS_SLAVE|MS_REC failed: %m");
2769 if (mount_devices(arg_directory, root_device, home_device, srv_device) < 0)
2772 /* Turn directory into bind mount */
2773 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
2774 log_error("Failed to make bind mount.");
2779 if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) {
2780 log_error("Failed to make read-only.");
2784 if (mount_all(arg_directory) < 0)
2787 if (copy_devnodes(arg_directory) < 0)
2790 if (setup_ptmx(arg_directory) < 0)
2793 dev_setup(arg_directory);
2795 if (audit_still_doesnt_work_in_containers() < 0)
2798 if (setup_dev_console(arg_directory, console) < 0)
2801 if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
2804 close_nointr_nofail(kmsg_socket_pair[1]);
2805 kmsg_socket_pair[1] = -1;
2807 if (setup_boot_id(arg_directory) < 0)
2810 if (setup_timezone(arg_directory) < 0)
2813 if (setup_resolv_conf(arg_directory) < 0)
2816 if (setup_journal(arg_directory) < 0)
2819 if (mount_binds(arg_directory, arg_bind, 0) < 0)
2822 if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
2825 if (setup_kdbus(arg_directory, kdbus_domain) < 0)
2828 /* Tell the parent that we are ready, and that
2829 * it can cgroupify us to that we lack access
2830 * to certain devices and resources. */
2831 eventfd_write(child_ready_fd, 1);
2832 close_nointr_nofail(child_ready_fd);
2833 child_ready_fd = -1;
2835 if (chdir(arg_directory) < 0) {
2836 log_error("chdir(%s) failed: %m", arg_directory);
2840 if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
2841 log_error("mount(MS_MOVE) failed: %m");
2845 if (chroot(".") < 0) {
2846 log_error("chroot() failed: %m");
2850 if (chdir("/") < 0) {
2851 log_error("chdir() failed: %m");
2857 if (arg_private_network)
2860 if (drop_capabilities() < 0) {
2861 log_error("drop_capabilities() failed: %m");
2865 r = change_uid_gid(&home);
2869 if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
2870 (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
2871 (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
2876 if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) {
2877 if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) {
2883 if (fdset_size(fds) > 0) {
2884 k = fdset_cloexec(fds, false);
2886 log_error("Failed to unset O_CLOEXEC for file descriptors.");
2890 if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
2891 (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) {
2899 if (arg_personality != 0xffffffffLU) {
2900 if (personality(arg_personality) < 0) {
2901 log_error("personality() failed: %m");
2904 } else if (secondary) {
2905 if (personality(PER_LINUX32) < 0) {
2906 log_error("personality() failed: %m");
2912 if (arg_selinux_context)
2913 if (setexeccon((security_context_t) arg_selinux_context) < 0) {
2914 log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context);
2919 if (!strv_isempty(arg_setenv)) {
2922 n = strv_env_merge(2, envp, arg_setenv);
2930 env_use = (char**) envp;
2932 /* Wait until the parent is ready with the setup, too... */
2933 eventfd_read(parent_ready_fd, &x);
2934 close_nointr_nofail(parent_ready_fd);
2935 parent_ready_fd = -1;
2941 /* Automatically search for the init system */
2943 l = 1 + argc - optind;
2944 a = newa(char*, l + 1);
2945 memcpy(a + 1, argv + optind, l * sizeof(char*));
2947 a[0] = (char*) "/usr/lib/systemd/systemd";
2948 execve(a[0], a, env_use);
2950 a[0] = (char*) "/lib/systemd/systemd";
2951 execve(a[0], a, env_use);
2953 a[0] = (char*) "/sbin/init";
2954 execve(a[0], a, env_use);
2955 } else if (argc > optind)
2956 execvpe(argv[optind], argv + optind, env_use);
2958 chdir(home ? home : "/root");
2959 execle("/bin/bash", "-bash", NULL, env_use);
2960 execle("/bin/sh", "-sh", NULL, env_use);
2963 log_error("execv() failed: %m");
2966 _exit(EXIT_FAILURE);
2972 /* Wait until the child reported that it is ready with
2973 * all it needs to do with priviliges. After we got
2974 * the notification we can make the process join its
2975 * cgroup which might limit what it can do */
2976 eventfd_read(child_ready_fd, &x);
2978 r = register_machine(pid);
2982 r = move_network_interfaces(pid);
2986 r = setup_veth(pid, veth_name);
2990 r = setup_bridge(veth_name);
2994 r = setup_macvlan(pid);
2998 /* Notify the child that the parent is ready with all
2999 * its setup, and thtat the child can now hand over
3000 * control to the code to run inside the container. */
3001 eventfd_write(parent_ready_fd, 1);
3003 k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3);
3012 /* Kill if it is not dead yet anyway */
3013 terminate_machine(pid);
3015 /* Redundant, but better safe than sorry */
3018 k = wait_for_terminate(pid, &status);
3026 if (status.si_code == CLD_EXITED) {
3027 r = status.si_status;
3028 if (status.si_status != 0) {
3029 log_error("Container %s failed with error code %i.", arg_machine, status.si_status);
3034 log_debug("Container %s exited successfully.", arg_machine);
3036 } else if (status.si_code == CLD_KILLED &&
3037 status.si_status == SIGINT) {
3040 log_info("Container %s has been shut down.", arg_machine);
3043 } else if (status.si_code == CLD_KILLED &&
3044 status.si_status == SIGHUP) {
3047 log_info("Container %s is being rebooted.", arg_machine);
3049 } else if (status.si_code == CLD_KILLED ||
3050 status.si_code == CLD_DUMPED) {
3052 log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
3056 log_error("Container %s failed due to unknown reason.", arg_machine);
3063 loop_remove(loop_nr, &image_fd);
3068 free(arg_directory);
3071 strv_free(arg_setenv);
3072 strv_free(arg_network_interfaces);
3073 strv_free(arg_network_macvlan);
3074 strv_free(arg_bind);
3075 strv_free(arg_bind_ro);