]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/nspawn/nspawn.c
projects / thirdparty / systemd.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nspawn: turn on watchdog logic for nspawn too
[thirdparty/systemd.git] / src / nspawn / nspawn.c
CommitLineData
53e1b683 1/* SPDX-License-Identifier: LGPL-2.1+ */
88213476
LP		 2/***
3 This file is part of systemd.
4
5 Copyright 2010 Lennart Poettering
6
7 systemd is free software; you can redistribute it and/or modify it
5430f7f2
LP		 8 under the terms of the GNU Lesser General Public License as published by
9 the Free Software Foundation; either version 2.1 of the License, or
88213476
LP		 10 (at your option) any later version.
11
12 systemd is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
5430f7f2 15 Lesser General Public License for more details.
88213476 16
5430f7f2 17 You should have received a copy of the GNU Lesser General Public License
88213476
LP		 18 along with systemd; If not, see <http://www.gnu.org/licenses/>.
19***/
20
349cc4a5 21#if HAVE_BLKID
6b5cf3ea 22#include <blkid.h>
8fe0087e 23#endif
88213476 24#include <errno.h>
88213476 25#include <getopt.h>
0e7ac751 26#include <grp.h>
1b9e5b12 27#include <linux/loop.h>
0e7ac751 28#include <pwd.h>
8fe0087e 29#include <sched.h>
349cc4a5 30#if HAVE_SELINUX
8fe0087e 31#include <selinux/selinux.h>
1b9e5b12 32#endif
8fe0087e
LP		 33#include <signal.h>
34#include <stdio.h>
35#include <stdlib.h>
36#include <string.h>
37#include <sys/file.h>
38#include <sys/mount.h>
39#include <sys/personality.h>
40#include <sys/prctl.h>
41#include <sys/types.h>
6916b164 42#include <sys/wait.h>
8fe0087e 43#include <unistd.h>
1b9e5b12 44
b053cd5f 45#include "sd-bus.h"
1f0cd86b 46#include "sd-daemon.h"
1f0cd86b 47#include "sd-id128.h"
8fe0087e 48
b5efdb8a 49#include "alloc-util.h"
8fe0087e
LP		 50#include "barrier.h"
51#include "base-filesystem.h"
52#include "blkid-util.h"
53#include "btrfs-util.h"
b053cd5f 54#include "bus-util.h"
8fe0087e 55#include "cap-list.h"
430f0182 56#include "capability-util.h"
04d391da 57#include "cgroup-util.h"
8fe0087e 58#include "copy.h"
4fc9982c 59#include "dev-setup.h"
2d845785 60#include "dissect-image.h"
8fe0087e 61#include "env-util.h"
3ffd4af2 62#include "fd-util.h"
842f3b0f 63#include "fdset.h"
a5c32cff 64#include "fileio.h"
f97b34a6 65#include "format-util.h"
f4f15635 66#include "fs-util.h"
1b9e5b12 67#include "gpt.h"
4623e8e6 68#include "hexdecoct.h"
8fe0087e 69#include "hostname-util.h"
910fd145 70#include "id128-util.h"
8fe0087e 71#include "log.h"
2d845785 72#include "loop-util.h"
8fe0087e 73#include "loopback-setup.h"
1b9cebf6 74#include "machine-image.h"
8fe0087e
LP		 75#include "macro.h"
76#include "missing.h"
77#include "mkdir.h"
4349cd7c 78#include "mount-util.h"
8fe0087e 79#include "netlink-util.h"
07630cea 80#include "nspawn-cgroup.h"
3603efde 81#include "nspawn-def.h"
07630cea
LP		 82#include "nspawn-expose-ports.h"
83#include "nspawn-mount.h"
84#include "nspawn-network.h"
7336138e 85#include "nspawn-patch-uid.h"
07630cea 86#include "nspawn-register.h"
910fd145 87#include "nspawn-seccomp.h"
07630cea
LP		 88#include "nspawn-settings.h"
89#include "nspawn-setuid.h"
7732f92b 90#include "nspawn-stub-pid1.h"
6bedfcbb 91#include "parse-util.h"
8fe0087e 92#include "path-util.h"
0b452006 93#include "process-util.h"
8fe0087e
LP		 94#include "ptyfwd.h"
95#include "random-util.h"
8869a0b4 96#include "raw-clone.h"
8fe0087e 97#include "rm-rf.h"
68b02049 98#include "selinux-util.h"
8fe0087e 99#include "signal-util.h"
2583fbea 100#include "socket-util.h"
8fcde012 101#include "stat-util.h"
15a5e950 102#include "stdio-util.h"
07630cea 103#include "string-util.h"
8fe0087e
LP		 104#include "strv.h"
105#include "terminal-util.h"
106#include "udev-util.h"
affb60b1 107#include "umask-util.h"
b1d4f8e1 108#include "user-util.h"
8fe0087e 109#include "util.h"
e9642be2 110
62b1e758
YW		 111#if HAVE_SPLIT_USR
112#define STATIC_RESOLV_CONF "/lib/systemd/resolv.conf"
113#else
114#define STATIC_RESOLV_CONF "/usr/lib/systemd/resolv.conf"
115#endif
116
9c1e04d0
AP		 117/* nspawn is listening on the socket at the path in the constant nspawn_notify_socket_path
118 * nspawn_notify_socket_path is relative to the container
119 * the init process in the container pid can send messages to nspawn following the sd_notify(3) protocol */
120#define NSPAWN_NOTIFY_SOCKET_PATH "/run/systemd/nspawn/notify"
0e7ac751 121
2a49b612
ZJS		 122#define EXIT_FORCE_RESTART 133
123
113cea80
DH		 124typedef enum ContainerStatus {
125 CONTAINER_TERMINATED,
126 CONTAINER_REBOOTED
127} ContainerStatus;
128
57fb9fb5
LP		 129typedef enum LinkJournal {
130 LINK_NO,
131 LINK_AUTO,
132 LINK_HOST,
133 LINK_GUEST
134} LinkJournal;
88213476
LP		 135
136static char *arg_directory = NULL;
ec16945e 137static char *arg_template = NULL;
5f932eb9 138static char *arg_chdir = NULL;
b53ede69
PW		 139static char *arg_pivot_root_new = NULL;
140static char *arg_pivot_root_old = NULL;
687d0825 141static char *arg_user = NULL;
9444b1f2 142static sd_id128_t arg_uuid = {};
7027ff61 143static char *arg_machine = NULL;
c74e630d
LP		 144static const char *arg_selinux_context = NULL;
145static const char *arg_selinux_apifs_context = NULL;
9444b1f2 146static const char *arg_slice = NULL;
ff01d048 147static bool arg_private_network = false;
bc2f673e 148static bool arg_read_only = false;
7732f92b 149static StartMode arg_start_mode = START_PID1;
ec16945e 150static bool arg_ephemeral = false;
57fb9fb5 151static LinkJournal arg_link_journal = LINK_AUTO;
574edc90 152static bool arg_link_journal_try = false;
520e0d54 153static uint64_t arg_caps_retain =
50b52222
LP		 154 (1ULL << CAP_AUDIT_CONTROL) |
155 (1ULL << CAP_AUDIT_WRITE) |
5076f0cc
LP		 156 (1ULL << CAP_CHOWN) |
157 (1ULL << CAP_DAC_OVERRIDE) |
158 (1ULL << CAP_DAC_READ_SEARCH) |
159 (1ULL << CAP_FOWNER) |
160 (1ULL << CAP_FSETID) |
161 (1ULL << CAP_IPC_OWNER) |
162 (1ULL << CAP_KILL) |
163 (1ULL << CAP_LEASE) |
164 (1ULL << CAP_LINUX_IMMUTABLE) |
50b52222 165 (1ULL << CAP_MKNOD) |
5076f0cc
LP		 166 (1ULL << CAP_NET_BIND_SERVICE) |
167 (1ULL << CAP_NET_BROADCAST) |
168 (1ULL << CAP_NET_RAW) |
5076f0cc 169 (1ULL << CAP_SETFCAP) |
50b52222 170 (1ULL << CAP_SETGID) |
5076f0cc
LP		 171 (1ULL << CAP_SETPCAP) |
172 (1ULL << CAP_SETUID) |
173 (1ULL << CAP_SYS_ADMIN) |
50b52222 174 (1ULL << CAP_SYS_BOOT) |
5076f0cc
LP		 175 (1ULL << CAP_SYS_CHROOT) |
176 (1ULL << CAP_SYS_NICE) |
177 (1ULL << CAP_SYS_PTRACE) |
d87be9b0 178 (1ULL << CAP_SYS_RESOURCE) |
50b52222 179 (1ULL << CAP_SYS_TTY_CONFIG);
5a8af538
LP		 180static CustomMount *arg_custom_mounts = NULL;
181static unsigned arg_n_custom_mounts = 0;
f4889f65 182static char **arg_setenv = NULL;
284c0b91 183static bool arg_quiet = false;
eb91eb18 184static bool arg_register = true;
89f7c846 185static bool arg_keep_unit = false;
aa28aefe 186static char **arg_network_interfaces = NULL;
c74e630d 187static char **arg_network_macvlan = NULL;
4bbfe7ad 188static char **arg_network_ipvlan = NULL;
69c79d3c 189static bool arg_network_veth = false;
f6d6bad1 190static char **arg_network_veth_extra = NULL;
f757855e 191static char *arg_network_bridge = NULL;
22b28dfd 192static char *arg_network_zone = NULL;
050f7277 193static unsigned long arg_personality = PERSONALITY_INVALID;
ec16945e 194static char *arg_image = NULL;
f757855e 195static VolatileMode arg_volatile_mode = VOLATILE_NO;
6d0b55c2 196static ExposePort *arg_expose_ports = NULL;
f36933fe 197static char **arg_property = NULL;
0de7acce 198static UserNamespaceMode arg_userns_mode = USER_NAMESPACE_NO;
6dac160c 199static uid_t arg_uid_shift = UID_INVALID, arg_uid_range = 0x10000U;
0de7acce 200static bool arg_userns_chown = false;
c6c8f6e2 201static int arg_kill_signal = 0;
5da38d07 202static CGroupUnified arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_UNKNOWN;
f757855e
LP		 203static SettingsMask arg_settings_mask = 0;
204static int arg_settings_trusted = -1;
205static char **arg_parameters = NULL;
6aadfa4c 206static const char *arg_container_service_name = "systemd-nspawn";
9c1e04d0 207static bool arg_notify_ready = false;
5a8ff0e6 208static bool arg_use_cgns = true;
0c582db0 209static unsigned long arg_clone_ns_flags = CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS;
4f086aab 210static MountSettingsMask arg_mount_settings = MOUNT_APPLY_APIVFS_RO;
4623e8e6
LP		 211static void *arg_root_hash = NULL;
212static size_t arg_root_hash_size = 0;
960e4569
LP		 213static char **arg_syscall_whitelist = NULL;
214static char **arg_syscall_blacklist = NULL;
88213476 215
601185b4 216static void help(void) {
88213476
LP		 217 printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
218 "Spawn a minimal namespace container for debugging, testing and building.\n\n"
a8828ed9
DW		 219 " -h --help Show this help\n"
220 " --version Print version string\n"
69c79d3c 221 " -q --quiet Do not show status information\n"
1b9e5b12 222 " -D --directory=PATH Root directory for the container\n"
ec16945e
LP		 223 " --template=PATH Initialize root directory from template directory,\n"
224 " if missing\n"
225 " -x --ephemeral Run container with snapshot of root directory, and\n"
226 " remove it after exit\n"
227 " -i --image=PATH File system device or disk image for the container\n"
4623e8e6 228 " --root-hash=HASH Specify verity root hash\n"
7732f92b 229 " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n"
a8828ed9 230 " -b --boot Boot up full system (i.e. invoke init)\n"
5f932eb9 231 " --chdir=PATH Set working directory in the container\n"
b53ede69
PW		 232 " --pivot-root=PATH[:PATH]\n"
233 " Pivot root to given directory in the container\n"
a8828ed9 234 " -u --user=USER Run the command under specified user or uid\n"
a8828ed9 235 " -M --machine=NAME Set the machine name for the container\n"
69c79d3c 236 " --uuid=UUID Set a specific machine UUID for the container\n"
a8828ed9 237 " -S --slice=SLICE Place the container in the specified slice\n"
f36933fe 238 " --property=NAME=VALUE Set scope unit property\n"
90b4a64d 239 " -U --private-users=pick Run within user namespace, autoselect UID/GID range\n"
03cfe0d5 240 " --private-users[=UIDBASE[:NUIDS]]\n"
90b4a64d 241 " Similar, but with user configured UID/GID range\n"
24597ee0 242 " --private-users-chown Adjust OS tree ownership to private UID/GID range\n"
69c79d3c
LP		 243 " --private-network Disable network in container\n"
244 " --network-interface=INTERFACE\n"
245 " Assign an existing network interface to the\n"
246 " container\n"
c74e630d
LP		 247 " --network-macvlan=INTERFACE\n"
248 " Create a macvlan network interface based on an\n"
249 " existing network interface to the container\n"
4bbfe7ad
TG		 250 " --network-ipvlan=INTERFACE\n"
251 " Create a ipvlan network interface based on an\n"
252 " existing network interface to the container\n"
a8eaaee7 253 " -n --network-veth Add a virtual Ethernet connection between host\n"
69c79d3c 254 " and container\n"
f6d6bad1
LP		 255 " --network-veth-extra=HOSTIF[:CONTAINERIF]\n"
256 " Add an additional virtual Ethernet link between\n"
257 " host and container\n"
ab046dde 258 " --network-bridge=INTERFACE\n"
90b4a64d
ZJS		 259 " Add a virtual Ethernet connection to the container\n"
260 " and attach it to an existing bridge on the host\n"
261 " --network-zone=NAME Similar, but attach the new interface to an\n"
262 " an automatically managed bridge interface\n"
6d0b55c2 263 " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n"
ab5e3a1b 264 " Expose a container IP port on the host\n"
82adf6af
LP		 265 " -Z --selinux-context=SECLABEL\n"
266 " Set the SELinux security context to be used by\n"
267 " processes in the container\n"
268 " -L --selinux-apifs-context=SECLABEL\n"
269 " Set the SELinux security context to be used by\n"
270 " API/tmpfs file systems in the container\n"
a8828ed9
DW		 271 " --capability=CAP In addition to the default, retain specified\n"
272 " capability\n"
273 " --drop-capability=CAP Drop the specified capability from the default set\n"
960e4569
LP		 274 " --system-call-filter=LIST|~LIST\n"
275 " Permit/prohibit specific system calls\n"
c6c8f6e2 276 " --kill-signal=SIGNAL Select signal to use for shutting down PID 1\n"
2b26a728
LP		 277 " --link-journal=MODE Link up guest journal, one of no, auto, guest, \n"
278 " host, try-guest, try-host\n"
574edc90 279 " -j Equivalent to --link-journal=try-guest\n"
69c79d3c 280 " --read-only Mount the root directory read-only\n"
5e5bfa6e
EY		 281 " --bind=PATH[:PATH[:OPTIONS]]\n"
282 " Bind mount a file or directory from the host into\n"
a8828ed9 283 " the container\n"
5e5bfa6e
EY		 284 " --bind-ro=PATH[:PATH[:OPTIONS]\n"
285 " Similar, but creates a read-only bind mount\n"
06c17c39 286 " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
5a8af538
LP		 287 " --overlay=PATH[:PATH...]:PATH\n"
288 " Create an overlay mount from the host to \n"
289 " the container\n"
290 " --overlay-ro=PATH[:PATH...]:PATH\n"
291 " Similar, but creates a read-only overlay mount\n"
a5f1cb3b 292 " -E --setenv=NAME=VALUE Pass an environment variable to PID 1\n"
eb91eb18 293 " --register=BOOLEAN Register container as machine\n"
89f7c846 294 " --keep-unit Do not register a scope for the machine, reuse\n"
4d9f07b4 295 " the service unit nspawn is running in\n"
6d0b55c2 296 " --volatile[=MODE] Run the system in volatile mode\n"
f757855e 297 " --settings=BOOLEAN Load additional settings from .nspawn file\n"
90b4a64d 298 " --notify-ready=BOOLEAN Receive notifications from the child init process\n"
6d0b55c2 299 , program_invocation_short_name);
88213476
LP		 300}
301
86c0dd4a 302static int custom_mount_check_all(void) {
5a8af538 303 unsigned i;
5a8af538 304
5a8af538
LP		 305 for (i = 0; i < arg_n_custom_mounts; i++) {
306 CustomMount *m = &arg_custom_mounts[i];
307
0de7acce 308 if (path_equal(m->destination, "/") && arg_userns_mode != USER_NAMESPACE_NO) {
0e7ac751
LP		 309
310 if (arg_userns_chown) {
311 log_error("--private-users-chown may not be combined with custom root mounts.");
312 return -EINVAL;
313 } else if (arg_uid_shift == UID_INVALID) {
314 log_error("--private-users with automatic UID shift may not be combined with custom root mounts.");
315 return -EINVAL;
316 }
825d5287 317 }
5a8af538
LP		 318 }
319
320 return 0;
321}
322
8199d554 323static int detect_unified_cgroup_hierarchy_from_environment(void) {
efdb0237 324 const char *e;
415fc41c 325 int r;
5da38d07 326
efdb0237
LP		 327 /* Allow the user to control whether the unified hierarchy is used */
328 e = getenv("UNIFIED_CGROUP_HIERARCHY");
329 if (e) {
330 r = parse_boolean(e);
331 if (r < 0)
332 return log_error_errno(r, "Failed to parse $UNIFIED_CGROUP_HIERARCHY.");
5da38d07
TH		 333 if (r > 0)
334 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
335 else
336 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237
LP		 337 }
338
8199d554
LP		 339 return 0;
340}
341
342static int detect_unified_cgroup_hierarchy_from_image(const char *directory) {
343 int r;
344
345 /* Let's inherit the mode to use from the host system, but let's take into consideration what systemd in the
346 * image actually supports. */
b4cccbc1
LP		 347 r = cg_all_unified();
348 if (r < 0)
349 return log_error_errno(r, "Failed to determine whether we are in all unified mode.");
350 if (r > 0) {
a8725a06
ZJS		 351 /* Unified cgroup hierarchy support was added in 230. Unfortunately the detection
352 * routine only detects 231, so we'll have a false negative here for 230. */
353 r = systemd_installation_has_version(directory, 230);
354 if (r < 0)
355 return log_error_errno(r, "Failed to determine systemd version in container: %m");
356 if (r > 0)
357 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
358 else
359 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
c22800e4 360 } else if (cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER) > 0) {
2977724b
TH		 361 /* Mixed cgroup hierarchy support was added in 233 */
362 r = systemd_installation_has_version(directory, 233);
0fd9563f
ZJS		 363 if (r < 0)
364 return log_error_errno(r, "Failed to determine systemd version in container: %m");
365 if (r > 0)
366 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD;
367 else
368 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
369 } else
5da38d07 370 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237 371
8199d554
LP		 372 log_debug("Using %s hierarchy for container.",
373 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_NONE ? "legacy" :
374 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_SYSTEMD ? "hybrid" : "unified");
375
efdb0237
LP		 376 return 0;
377}
378
0c582db0
LB		 379static void parse_share_ns_env(const char *name, unsigned long ns_flag) {
380 int r;
381
382 r = getenv_bool(name);
383 if (r == -ENXIO)
384 return;
385 if (r < 0)
386 log_warning_errno(r, "Failed to parse %s from environment, defaulting to false.", name);
387 arg_clone_ns_flags = (arg_clone_ns_flags & ~ns_flag) | (r > 0 ? 0 : ns_flag);
388}
389
4f086aab
SU		 390static void parse_mount_settings_env(void) {
391 int r;
392 const char *e;
393
394 e = getenv("SYSTEMD_NSPAWN_API_VFS_WRITABLE");
395 if (!e)
396 return;
397
398 if (streq(e, "network")) {
399 arg_mount_settings |= MOUNT_APPLY_APIVFS_RO|MOUNT_APPLY_APIVFS_NETNS;
400 return;
401 }
402
403 r = parse_boolean(e);
404 if (r < 0) {
405 log_warning_errno(r, "Failed to parse SYSTEMD_NSPAWN_API_VFS_WRITABLE from environment, ignoring.");
406 return;
ab8ee0f2 407 }
4f086aab 408
ab8ee0f2
ZJS		 409 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_RO, r == 0);
410 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_NETNS, false);
4f086aab
SU		 411}
412
88213476
LP		 413static int parse_argv(int argc, char *argv[]) {
414
a41fe3a2 415 enum {
acbeb427
ZJS		 416 ARG_VERSION = 0x100,
417 ARG_PRIVATE_NETWORK,
bc2f673e 418 ARG_UUID,
5076f0cc 419 ARG_READ_ONLY,
57fb9fb5 420 ARG_CAPABILITY,
420c7379 421 ARG_DROP_CAPABILITY,
17fe0523
LP		 422 ARG_LINK_JOURNAL,
423 ARG_BIND,
f4889f65 424 ARG_BIND_RO,
06c17c39 425 ARG_TMPFS,
5a8af538
LP		 426 ARG_OVERLAY,
427 ARG_OVERLAY_RO,
eb91eb18 428 ARG_SHARE_SYSTEM,
89f7c846 429 ARG_REGISTER,
aa28aefe 430 ARG_KEEP_UNIT,
69c79d3c 431 ARG_NETWORK_INTERFACE,
c74e630d 432 ARG_NETWORK_MACVLAN,
4bbfe7ad 433 ARG_NETWORK_IPVLAN,
ab046dde 434 ARG_NETWORK_BRIDGE,
22b28dfd 435 ARG_NETWORK_ZONE,
f6d6bad1 436 ARG_NETWORK_VETH_EXTRA,
6afc95b7 437 ARG_PERSONALITY,
4d9f07b4 438 ARG_VOLATILE,
ec16945e 439 ARG_TEMPLATE,
f36933fe 440 ARG_PROPERTY,
6dac160c 441 ARG_PRIVATE_USERS,
c6c8f6e2 442 ARG_KILL_SIGNAL,
f757855e 443 ARG_SETTINGS,
5f932eb9 444 ARG_CHDIR,
b53ede69 445 ARG_PIVOT_ROOT,
7336138e 446 ARG_PRIVATE_USERS_CHOWN,
9c1e04d0 447 ARG_NOTIFY_READY,
4623e8e6 448 ARG_ROOT_HASH,
960e4569 449 ARG_SYSTEM_CALL_FILTER,
a41fe3a2
LP		 450 };
451
88213476 452 static const struct option options[] = {
27eb8e90
ZJS		 453 { "help", no_argument, NULL, 'h' },
454 { "version", no_argument, NULL, ARG_VERSION },
455 { "directory", required_argument, NULL, 'D' },
456 { "template", required_argument, NULL, ARG_TEMPLATE },
457 { "ephemeral", no_argument, NULL, 'x' },
458 { "user", required_argument, NULL, 'u' },
459 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
460 { "as-pid2", no_argument, NULL, 'a' },
461 { "boot", no_argument, NULL, 'b' },
462 { "uuid", required_argument, NULL, ARG_UUID },
463 { "read-only", no_argument, NULL, ARG_READ_ONLY },
464 { "capability", required_argument, NULL, ARG_CAPABILITY },
465 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
466 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
467 { "bind", required_argument, NULL, ARG_BIND },
468 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
469 { "tmpfs", required_argument, NULL, ARG_TMPFS },
470 { "overlay", required_argument, NULL, ARG_OVERLAY },
471 { "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO },
472 { "machine", required_argument, NULL, 'M' },
473 { "slice", required_argument, NULL, 'S' },
474 { "setenv", required_argument, NULL, 'E' },
475 { "selinux-context", required_argument, NULL, 'Z' },
476 { "selinux-apifs-context", required_argument, NULL, 'L' },
477 { "quiet", no_argument, NULL, 'q' },
478 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, /* not documented */
479 { "register", required_argument, NULL, ARG_REGISTER },
480 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
481 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
482 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
483 { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN },
484 { "network-veth", no_argument, NULL, 'n' },
485 { "network-veth-extra", required_argument, NULL, ARG_NETWORK_VETH_EXTRA },
486 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
487 { "network-zone", required_argument, NULL, ARG_NETWORK_ZONE },
488 { "personality", required_argument, NULL, ARG_PERSONALITY },
489 { "image", required_argument, NULL, 'i' },
490 { "volatile", optional_argument, NULL, ARG_VOLATILE },
491 { "port", required_argument, NULL, 'p' },
492 { "property", required_argument, NULL, ARG_PROPERTY },
493 { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS },
494 { "private-users-chown", optional_argument, NULL, ARG_PRIVATE_USERS_CHOWN },
495 { "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL },
496 { "settings", required_argument, NULL, ARG_SETTINGS },
497 { "chdir", required_argument, NULL, ARG_CHDIR },
b53ede69 498 { "pivot-root", required_argument, NULL, ARG_PIVOT_ROOT },
27eb8e90 499 { "notify-ready", required_argument, NULL, ARG_NOTIFY_READY },
4623e8e6 500 { "root-hash", required_argument, NULL, ARG_ROOT_HASH },
960e4569 501 { "system-call-filter", required_argument, NULL, ARG_SYSTEM_CALL_FILTER },
eb9da376 502 {}
88213476
LP		 503 };
504
9444b1f2 505 int c, r;
6aadfa4c 506 const char *p, *e;
a42c8b54 507 uint64_t plus = 0, minus = 0;
f757855e 508 bool mask_all_settings = false, mask_no_settings = false;
88213476
LP		 509
510 assert(argc >= 0);
511 assert(argv);
512
2e1f244e 513 while ((c = getopt_long(argc, argv, "+hD:u:abL:M:jS:Z:qi:xp:nUE:", options, NULL)) >= 0)
88213476
LP		 514
515 switch (c) {
516
517 case 'h':
601185b4
ZJS		 518 help();
519 return 0;
88213476 520
acbeb427 521 case ARG_VERSION:
3f6fd1ba 522 return version();
acbeb427 523
88213476 524 case 'D':
0f03c2a4 525 r = parse_path_argument_and_warn(optarg, false, &arg_directory);
ec16945e 526 if (r < 0)
0f03c2a4 527 return r;
ec16945e
LP		 528 break;
529
530 case ARG_TEMPLATE:
0f03c2a4 531 r = parse_path_argument_and_warn(optarg, false, &arg_template);
ec16945e 532 if (r < 0)
0f03c2a4 533 return r;
88213476
LP		 534 break;
535
1b9e5b12 536 case 'i':
0f03c2a4 537 r = parse_path_argument_and_warn(optarg, false, &arg_image);
ec16945e 538 if (r < 0)
0f03c2a4 539 return r;
ec16945e
LP		 540 break;
541
542 case 'x':
543 arg_ephemeral = true;
1b9e5b12
LP		 544 break;
545
687d0825 546 case 'u':
2fc09a9c
DM		 547 r = free_and_strdup(&arg_user, optarg);
548 if (r < 0)
7027ff61 549 return log_oom();
687d0825 550
f757855e 551 arg_settings_mask |= SETTING_USER;
687d0825
MV		 552 break;
553
22b28dfd
LP		 554 case ARG_NETWORK_ZONE: {
555 char *j;
556
557 j = strappend("vz-", optarg);
558 if (!j)
559 return log_oom();
560
561 if (!ifname_valid(j)) {
562 log_error("Network zone name not valid: %s", j);
563 free(j);
564 return -EINVAL;
565 }
566
567 free(arg_network_zone);
568 arg_network_zone = j;
569
570 arg_network_veth = true;
571 arg_private_network = true;
572 arg_settings_mask |= SETTING_NETWORK;
573 break;
574 }
575
ab046dde 576 case ARG_NETWORK_BRIDGE:
ef76dff2
LP		 577
578 if (!ifname_valid(optarg)) {
579 log_error("Bridge interface name not valid: %s", optarg);
580 return -EINVAL;
581 }
582
f757855e
LP		 583 r = free_and_strdup(&arg_network_bridge, optarg);
584 if (r < 0)
585 return log_oom();
ab046dde 586
4831981d 587 _fallthrough_;
0dfaa006 588 case 'n':
69c79d3c
LP		 589 arg_network_veth = true;
590 arg_private_network = true;
f757855e 591 arg_settings_mask |= SETTING_NETWORK;
69c79d3c
LP		 592 break;
593
f6d6bad1
LP		 594 case ARG_NETWORK_VETH_EXTRA:
595 r = veth_extra_parse(&arg_network_veth_extra, optarg);
596 if (r < 0)
597 return log_error_errno(r, "Failed to parse --network-veth-extra= parameter: %s", optarg);
598
599 arg_private_network = true;
600 arg_settings_mask |= SETTING_NETWORK;
601 break;
602
aa28aefe 603 case ARG_NETWORK_INTERFACE:
ef76dff2
LP		 604
605 if (!ifname_valid(optarg)) {
606 log_error("Network interface name not valid: %s", optarg);
607 return -EINVAL;
608 }
609
c74e630d
LP		 610 if (strv_extend(&arg_network_interfaces, optarg) < 0)
611 return log_oom();
612
613 arg_private_network = true;
f757855e 614 arg_settings_mask |= SETTING_NETWORK;
c74e630d
LP		 615 break;
616
617 case ARG_NETWORK_MACVLAN:
ef76dff2
LP		 618
619 if (!ifname_valid(optarg)) {
620 log_error("MACVLAN network interface name not valid: %s", optarg);
621 return -EINVAL;
622 }
623
c74e630d 624 if (strv_extend(&arg_network_macvlan, optarg) < 0)
aa28aefe
LP		 625 return log_oom();
626
4bbfe7ad 627 arg_private_network = true;
f757855e 628 arg_settings_mask |= SETTING_NETWORK;
4bbfe7ad
TG		 629 break;
630
631 case ARG_NETWORK_IPVLAN:
ef76dff2
LP		 632
633 if (!ifname_valid(optarg)) {
634 log_error("IPVLAN network interface name not valid: %s", optarg);
635 return -EINVAL;
636 }
637
4bbfe7ad
TG		 638 if (strv_extend(&arg_network_ipvlan, optarg) < 0)
639 return log_oom();
640
4831981d 641 _fallthrough_;
ff01d048
LP		 642 case ARG_PRIVATE_NETWORK:
643 arg_private_network = true;
f757855e 644 arg_settings_mask |= SETTING_NETWORK;
a41fe3a2
LP		 645 break;
646
0f0dbc46 647 case 'b':
7732f92b
LP		 648 if (arg_start_mode == START_PID2) {
649 log_error("--boot and --as-pid2 may not be combined.");
650 return -EINVAL;
651 }
652
653 arg_start_mode = START_BOOT;
654 arg_settings_mask |= SETTING_START_MODE;
655 break;
656
657 case 'a':
658 if (arg_start_mode == START_BOOT) {
659 log_error("--boot and --as-pid2 may not be combined.");
660 return -EINVAL;
661 }
662
663 arg_start_mode = START_PID2;
664 arg_settings_mask |= SETTING_START_MODE;
0f0dbc46
LP		 665 break;
666
144f0fc0 667 case ARG_UUID:
9444b1f2 668 r = sd_id128_from_string(optarg, &arg_uuid);
317feb4d
LP		 669 if (r < 0)
670 return log_error_errno(r, "Invalid UUID: %s", optarg);
671
672 if (sd_id128_is_null(arg_uuid)) {
673 log_error("Machine UUID may not be all zeroes.");
674 return -EINVAL;
aa96c6cb 675 }
f757855e
LP		 676
677 arg_settings_mask |= SETTING_MACHINE_ID;
9444b1f2 678 break;
aa96c6cb 679
9444b1f2 680 case 'S':
c74e630d 681 arg_slice = optarg;
144f0fc0
LP		 682 break;
683
7027ff61 684 case 'M':
c1521918 685 if (isempty(optarg))
97b11eed 686 arg_machine = mfree(arg_machine);
c1521918 687 else {
0c3c4284 688 if (!machine_name_is_valid(optarg)) {
eb91eb18
LP		 689 log_error("Invalid machine name: %s", optarg);
690 return -EINVAL;
691 }
7027ff61 692
0c3c4284
LP		 693 r = free_and_strdup(&arg_machine, optarg);
694 if (r < 0)
eb91eb18 695 return log_oom();
eb91eb18 696 }
9ce6d1b3 697 break;
7027ff61 698
82adf6af
LP		 699 case 'Z':
700 arg_selinux_context = optarg;
a8828ed9
DW		 701 break;
702
82adf6af
LP		 703 case 'L':
704 arg_selinux_apifs_context = optarg;
a8828ed9
DW		 705 break;
706
bc2f673e
LP		 707 case ARG_READ_ONLY:
708 arg_read_only = true;
f757855e 709 arg_settings_mask |= SETTING_READ_ONLY;
bc2f673e
LP		 710 break;
711
420c7379
LP		 712 case ARG_CAPABILITY:
713 case ARG_DROP_CAPABILITY: {
6cbe4ed1 714 p = optarg;
9ed794a3 715 for (;;) {
6cbe4ed1 716 _cleanup_free_ char *t = NULL;
5076f0cc 717
6cbe4ed1
SS		 718 r = extract_first_word(&p, &t, ",", 0);
719 if (r < 0)
720 return log_error_errno(r, "Failed to parse capability %s.", t);
5076f0cc 721
6cbe4ed1
SS		 722 if (r == 0)
723 break;
5076f0cc 724
39ed67d1
LP		 725 if (streq(t, "all")) {
726 if (c == ARG_CAPABILITY)
a42c8b54 727 plus = (uint64_t) -1;
39ed67d1 728 else
a42c8b54 729 minus = (uint64_t) -1;
39ed67d1 730 } else {
2822da4f
LP		 731 int cap;
732
733 cap = capability_from_name(t);
734 if (cap < 0) {
39ed67d1
LP		 735 log_error("Failed to parse capability %s.", t);
736 return -EINVAL;
737 }
738
739 if (c == ARG_CAPABILITY)
a42c8b54 740 plus |= 1ULL << (uint64_t) cap;
39ed67d1 741 else
a42c8b54 742 minus |= 1ULL << (uint64_t) cap;
5076f0cc 743 }
5076f0cc
LP		 744 }
745
f757855e 746 arg_settings_mask |= SETTING_CAPABILITY;
5076f0cc
LP		 747 break;
748 }
749
57fb9fb5
LP		 750 case 'j':
751 arg_link_journal = LINK_GUEST;
574edc90 752 arg_link_journal_try = true;
57fb9fb5
LP		 753 break;
754
755 case ARG_LINK_JOURNAL:
53e438e3 756 if (streq(optarg, "auto")) {
57fb9fb5 757 arg_link_journal = LINK_AUTO;
53e438e3
LP		 758 arg_link_journal_try = false;
759 } else if (streq(optarg, "no")) {
57fb9fb5 760 arg_link_journal = LINK_NO;
53e438e3
LP		 761 arg_link_journal_try = false;
762 } else if (streq(optarg, "guest")) {
57fb9fb5 763 arg_link_journal = LINK_GUEST;
53e438e3
LP		 764 arg_link_journal_try = false;
765 } else if (streq(optarg, "host")) {
57fb9fb5 766 arg_link_journal = LINK_HOST;
53e438e3
LP		 767 arg_link_journal_try = false;
768 } else if (streq(optarg, "try-guest")) {
574edc90
MP		 769 arg_link_journal = LINK_GUEST;
770 arg_link_journal_try = true;
771 } else if (streq(optarg, "try-host")) {
772 arg_link_journal = LINK_HOST;
773 arg_link_journal_try = true;
774 } else {
57fb9fb5
LP		 775 log_error("Failed to parse link journal mode %s", optarg);
776 return -EINVAL;
777 }
778
779 break;
780
17fe0523 781 case ARG_BIND:
f757855e
LP		 782 case ARG_BIND_RO:
783 r = bind_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg, c == ARG_BIND_RO);
784 if (r < 0)
785 return log_error_errno(r, "Failed to parse --bind(-ro)= argument %s: %m", optarg);
17fe0523 786
f757855e 787 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
17fe0523 788 break;
06c17c39 789
f757855e
LP		 790 case ARG_TMPFS:
791 r = tmpfs_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg);
792 if (r < 0)
793 return log_error_errno(r, "Failed to parse --tmpfs= argument %s: %m", optarg);
5a8af538 794
f757855e 795 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
5a8af538 796 break;
5a8af538
LP		 797
798 case ARG_OVERLAY:
ad85779a
LP		 799 case ARG_OVERLAY_RO:
800 r = overlay_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg, c == ARG_OVERLAY_RO);
801 if (r == -EADDRNOTAVAIL)
802 return log_error_errno(r, "--overlay(-ro)= needs at least two colon-separated directories specified.");
803 if (r < 0)
804 return log_error_errno(r, "Failed to parse --overlay(-ro)= argument %s: %m", optarg);
06c17c39 805
f757855e 806 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
06c17c39 807 break;
06c17c39 808
a5f1cb3b 809 case 'E': {
f4889f65
LP		 810 char **n;
811
812 if (!env_assignment_is_valid(optarg)) {
813 log_error("Environment variable assignment '%s' is not valid.", optarg);
814 return -EINVAL;
815 }
816
817 n = strv_env_set(arg_setenv, optarg);
818 if (!n)
819 return log_oom();
820
821 strv_free(arg_setenv);
822 arg_setenv = n;
f757855e
LP		 823
824 arg_settings_mask |= SETTING_ENVIRONMENT;
f4889f65
LP		 825 break;
826 }
827
284c0b91
LP		 828 case 'q':
829 arg_quiet = true;
830 break;
831
8a96d94e 832 case ARG_SHARE_SYSTEM:
a6b5216c 833 /* We don't officially support this anymore, except for compat reasons. People should use the
0c582db0
LB		 834 * $SYSTEMD_NSPAWN_SHARE_* environment variables instead. */
835 arg_clone_ns_flags = 0;
8a96d94e
LP		 836 break;
837
eb91eb18
LP		 838 case ARG_REGISTER:
839 r = parse_boolean(optarg);
840 if (r < 0) {
841 log_error("Failed to parse --register= argument: %s", optarg);
842 return r;
843 }
844
845 arg_register = r;
846 break;
847
89f7c846
LP		 848 case ARG_KEEP_UNIT:
849 arg_keep_unit = true;
850 break;
851
6afc95b7
LP		 852 case ARG_PERSONALITY:
853
ac45f971 854 arg_personality = personality_from_string(optarg);
050f7277 855 if (arg_personality == PERSONALITY_INVALID) {
6afc95b7
LP		 856 log_error("Unknown or unsupported personality '%s'.", optarg);
857 return -EINVAL;
858 }
859
f757855e 860 arg_settings_mask |= SETTING_PERSONALITY;
6afc95b7
LP		 861 break;
862
4d9f07b4
LP		 863 case ARG_VOLATILE:
864
865 if (!optarg)
f757855e 866 arg_volatile_mode = VOLATILE_YES;
4d9f07b4 867 else {
f757855e 868 VolatileMode m;
4d9f07b4 869
f757855e
LP		 870 m = volatile_mode_from_string(optarg);
871 if (m < 0) {
872 log_error("Failed to parse --volatile= argument: %s", optarg);
6d0b55c2 873 return -EINVAL;
f757855e
LP		 874 } else
875 arg_volatile_mode = m;
6d0b55c2
LP		 876 }
877
f757855e
LP		 878 arg_settings_mask |= SETTING_VOLATILE_MODE;
879 break;
6d0b55c2 880
f757855e
LP		 881 case 'p':
882 r = expose_port_parse(&arg_expose_ports, optarg);
883 if (r == -EEXIST)
884 return log_error_errno(r, "Duplicate port specification: %s", optarg);
885 if (r < 0)
886 return log_error_errno(r, "Failed to parse host port %s: %m", optarg);
6d0b55c2 887
f757855e 888 arg_settings_mask |= SETTING_EXPOSE_PORTS;
6d0b55c2 889 break;
6d0b55c2 890
f36933fe
LP		 891 case ARG_PROPERTY:
892 if (strv_extend(&arg_property, optarg) < 0)
893 return log_oom();
894
895 break;
896
ae209204
ZJS		 897 case ARG_PRIVATE_USERS: {
898 int boolean = -1;
0de7acce 899
ae209204
ZJS		 900 if (!optarg)
901 boolean = true;
902 else if (!in_charset(optarg, DIGITS))
903 /* do *not* parse numbers as booleans */
904 boolean = parse_boolean(optarg);
905
906 if (boolean == false) {
0de7acce
LP		 907 /* no: User namespacing off */
908 arg_userns_mode = USER_NAMESPACE_NO;
909 arg_uid_shift = UID_INVALID;
910 arg_uid_range = UINT32_C(0x10000);
ae209204 911 } else if (boolean == true) {
0de7acce
LP		 912 /* yes: User namespacing on, UID range is read from root dir */
913 arg_userns_mode = USER_NAMESPACE_FIXED;
914 arg_uid_shift = UID_INVALID;
915 arg_uid_range = UINT32_C(0x10000);
916 } else if (streq(optarg, "pick")) {
917 /* pick: User namespacing on, UID range is picked randomly */
918 arg_userns_mode = USER_NAMESPACE_PICK;
919 arg_uid_shift = UID_INVALID;
920 arg_uid_range = UINT32_C(0x10000);
921 } else {
6c2058b3 922 _cleanup_free_ char *buffer = NULL;
6dac160c
LP		 923 const char *range, *shift;
924
0de7acce
LP		 925 /* anything else: User namespacing on, UID range is explicitly configured */
926
6dac160c
LP		 927 range = strchr(optarg, ':');
928 if (range) {
6c2058b3
ZJS		 929 buffer = strndup(optarg, range - optarg);
930 if (!buffer)
931 return log_oom();
932 shift = buffer;
6dac160c
LP		 933
934 range++;
bfd292ec
ZJS		 935 r = safe_atou32(range, &arg_uid_range);
936 if (r < 0)
be715731 937 return log_error_errno(r, "Failed to parse UID range \"%s\": %m", range);
6dac160c
LP		 938 } else
939 shift = optarg;
940
be715731
ZJS		 941 r = parse_uid(shift, &arg_uid_shift);
942 if (r < 0)
943 return log_error_errno(r, "Failed to parse UID \"%s\": %m", optarg);
0de7acce
LP		 944
945 arg_userns_mode = USER_NAMESPACE_FIXED;
6dac160c
LP		 946 }
947
be715731
ZJS		 948 if (arg_uid_range <= 0) {
949 log_error("UID range cannot be 0.");
950 return -EINVAL;
951 }
952
0de7acce 953 arg_settings_mask |= SETTING_USERNS;
6dac160c 954 break;
ae209204 955 }
6dac160c 956
0de7acce 957 case 'U':
ccabee0d
LP		 958 if (userns_supported()) {
959 arg_userns_mode = USER_NAMESPACE_PICK;
960 arg_uid_shift = UID_INVALID;
961 arg_uid_range = UINT32_C(0x10000);
962
963 arg_settings_mask |= SETTING_USERNS;
6dac160c
LP		 964 }
965
7336138e
LP		 966 break;
967
0de7acce 968 case ARG_PRIVATE_USERS_CHOWN:
19aac838 969 arg_userns_chown = true;
0de7acce
LP		 970
971 arg_settings_mask |= SETTING_USERNS;
6dac160c
LP		 972 break;
973
c6c8f6e2
LP		 974 case ARG_KILL_SIGNAL:
975 arg_kill_signal = signal_from_string_try_harder(optarg);
976 if (arg_kill_signal < 0) {
977 log_error("Cannot parse signal: %s", optarg);
978 return -EINVAL;
979 }
980
f757855e
LP		 981 arg_settings_mask |= SETTING_KILL_SIGNAL;
982 break;
983
984 case ARG_SETTINGS:
985
986 /* no → do not read files
987 * yes → read files, do not override cmdline, trust only subset
988 * override → read files, override cmdline, trust only subset
989 * trusted → read files, do not override cmdline, trust all
990 */
991
992 r = parse_boolean(optarg);
993 if (r < 0) {
994 if (streq(optarg, "trusted")) {
995 mask_all_settings = false;
996 mask_no_settings = false;
997 arg_settings_trusted = true;
998
999 } else if (streq(optarg, "override")) {
1000 mask_all_settings = false;
1001 mask_no_settings = true;
1002 arg_settings_trusted = -1;
1003 } else
1004 return log_error_errno(r, "Failed to parse --settings= argument: %s", optarg);
1005 } else if (r > 0) {
1006 /* yes */
1007 mask_all_settings = false;
1008 mask_no_settings = false;
1009 arg_settings_trusted = -1;
1010 } else {
1011 /* no */
1012 mask_all_settings = true;
1013 mask_no_settings = false;
1014 arg_settings_trusted = false;
1015 }
1016
c6c8f6e2
LP		 1017 break;
1018
5f932eb9
LP		 1019 case ARG_CHDIR:
1020 if (!path_is_absolute(optarg)) {
1021 log_error("Working directory %s is not an absolute path.", optarg);
1022 return -EINVAL;
1023 }
1024
1025 r = free_and_strdup(&arg_chdir, optarg);
1026 if (r < 0)
1027 return log_oom();
1028
1029 arg_settings_mask |= SETTING_WORKING_DIRECTORY;
1030 break;
1031
b53ede69
PW		 1032 case ARG_PIVOT_ROOT:
1033 r = pivot_root_parse(&arg_pivot_root_new, &arg_pivot_root_old, optarg);
1034 if (r < 0)
1035 return log_error_errno(r, "Failed to parse --pivot-root= argument %s: %m", optarg);
1036
1037 arg_settings_mask |= SETTING_PIVOT_ROOT;
1038 break;
1039
9c1e04d0
AP		 1040 case ARG_NOTIFY_READY:
1041 r = parse_boolean(optarg);
1042 if (r < 0) {
1043 log_error("%s is not a valid notify mode. Valid modes are: yes, no, and ready.", optarg);
1044 return -EINVAL;
1045 }
1046 arg_notify_ready = r;
1047 arg_settings_mask |= SETTING_NOTIFY_READY;
1048 break;
1049
4623e8e6
LP		 1050 case ARG_ROOT_HASH: {
1051 void *k;
1052 size_t l;
1053
1054 r = unhexmem(optarg, strlen(optarg), &k, &l);
1055 if (r < 0)
1056 return log_error_errno(r, "Failed to parse root hash: %s", optarg);
1057 if (l < sizeof(sd_id128_t)) {
1058 log_error("Root hash must be at least 128bit long: %s", optarg);
1059 free(k);
1060 return -EINVAL;
1061 }
1062
1063 free(arg_root_hash);
1064 arg_root_hash = k;
1065 arg_root_hash_size = l;
1066 break;
1067 }
1068
960e4569
LP		 1069 case ARG_SYSTEM_CALL_FILTER: {
1070 bool negative;
1071 const char *items;
1072
1073 negative = optarg[0] == '~';
1074 items = negative ? optarg + 1 : optarg;
1075
1076 for (;;) {
1077 _cleanup_free_ char *word = NULL;
1078
1079 r = extract_first_word(&items, &word, NULL, 0);
1080 if (r == 0)
1081 break;
1082 if (r == -ENOMEM)
1083 return log_oom();
1084 if (r < 0)
1085 return log_error_errno(r, "Failed to parse system call filter: %m");
1086
1087 if (negative)
1088 r = strv_extend(&arg_syscall_blacklist, word);
1089 else
1090 r = strv_extend(&arg_syscall_whitelist, word);
1091 if (r < 0)
1092 return log_oom();
1093 }
1094
1095 arg_settings_mask |= SETTING_SYSCALL_FILTER;
1096 break;
1097 }
1098
88213476
LP		 1099 case '?':
1100 return -EINVAL;
1101
1102 default:
eb9da376 1103 assert_not_reached("Unhandled option");
88213476 1104 }
88213476 1105
0c582db0
LB		 1106 parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_IPC", CLONE_NEWIPC);
1107 parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_PID", CLONE_NEWPID);
1108 parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_UTS", CLONE_NEWUTS);
1109 parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_SYSTEM", CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS);
a6b5216c 1110
4f086aab
SU		 1111 if (arg_userns_mode != USER_NAMESPACE_NO)
1112 arg_mount_settings |= MOUNT_USE_USERNS;
1113
1114 if (arg_private_network)
1115 arg_mount_settings |= MOUNT_APPLY_APIVFS_NETNS;
1116
1117 parse_mount_settings_env();
1118
48a8d337
LB		 1119 if (!(arg_clone_ns_flags & CLONE_NEWPID) ||
1120 !(arg_clone_ns_flags & CLONE_NEWUTS)) {
eb91eb18 1121 arg_register = false;
0c582db0
LB		 1122 if (arg_start_mode != START_PID1) {
1123 log_error("--boot cannot be used without namespacing.");
1124 return -EINVAL;
1125 }
1126 }
eb91eb18 1127
0de7acce 1128 if (arg_userns_mode == USER_NAMESPACE_PICK)
0e7ac751
LP		 1129 arg_userns_chown = true;
1130
cd2dfc6f 1131 if (arg_keep_unit && arg_register && cg_pid_get_owner_uid(0, NULL) >= 0) {
8d9c2bca
AJ		 1132 /* Save the user from accidentally registering either user-$SESSION.scope or user@.service.
1133 * The latter is not technically a user session, but we don't need to labour the point. */
cd2dfc6f 1134 log_error("--keep-unit --register=yes may not be used when invoked from a user session.");
89f7c846
LP		 1135 return -EINVAL;
1136 }
1137
1b9e5b12
LP		 1138 if (arg_directory && arg_image) {
1139 log_error("--directory= and --image= may not be combined.");
1140 return -EINVAL;
1141 }
1142
ec16945e
LP		 1143 if (arg_template && arg_image) {
1144 log_error("--template= and --image= may not be combined.");
1145 return -EINVAL;
1146 }
1147
8cd328d8
LP		 1148 if (arg_ephemeral && arg_template && !arg_directory) {
1149 /* User asked for ephemeral execution but specified --template= instead of --directory=. Semantically
1150 * such an invocation makes some sense, see https://github.com/systemd/systemd/issues/3667. Let's
1151 * accept this here, and silently make "--ephemeral --template=" equivalent to "--ephemeral
1152 * --directory=". */
1153
1154 arg_directory = arg_template;
1155 arg_template = NULL;
1156 }
1157
ec16945e
LP		 1158 if (arg_template && !(arg_directory || arg_machine)) {
1159 log_error("--template= needs --directory= or --machine=.");
1160 return -EINVAL;
1161 }
1162
1163 if (arg_ephemeral && arg_template) {
1164 log_error("--ephemeral and --template= may not be combined.");
1165 return -EINVAL;
1166 }
1167
df9a75e4
LP		 1168 if (arg_ephemeral && !IN_SET(arg_link_journal, LINK_NO, LINK_AUTO)) {
1169 log_error("--ephemeral and --link-journal= may not be combined.");
1170 return -EINVAL;
1171 }
1172
ccabee0d 1173 if (arg_userns_mode != USER_NAMESPACE_NO && !userns_supported()) {
7336138e
LP		 1174 log_error("--private-users= is not supported, kernel compiled without user namespace support.");
1175 return -EOPNOTSUPP;
1176 }
1177
1178 if (arg_userns_chown && arg_read_only) {
1179 log_error("--read-only and --private-users-chown may not be combined.");
1180 return -EINVAL;
1181 }
f757855e 1182
22b28dfd
LP		 1183 if (arg_network_bridge && arg_network_zone) {
1184 log_error("--network-bridge= and --network-zone= may not be combined.");
1185 return -EINVAL;
1186 }
1187
f757855e
LP		 1188 if (argc > optind) {
1189 arg_parameters = strv_copy(argv + optind);
1190 if (!arg_parameters)
1191 return log_oom();
1192
7732f92b 1193 arg_settings_mask |= SETTING_START_MODE;
f757855e
LP		 1194 }
1195
1196 /* Load all settings from .nspawn files */
1197 if (mask_no_settings)
1198 arg_settings_mask = 0;
1199
1200 /* Don't load any settings from .nspawn files */
1201 if (mask_all_settings)
1202 arg_settings_mask = _SETTINGS_MASK_ALL;
1203
520e0d54 1204 arg_caps_retain = (arg_caps_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus;
f757855e 1205
399e391f
ZJS		 1206 r = cg_unified_flush();
1207 if (r < 0)
1208 return log_error_errno(r, "Failed to determine whether the unified cgroups hierarchy is used: %m");
1209
6aadfa4c
ILG		 1210 e = getenv("SYSTEMD_NSPAWN_CONTAINER_SERVICE");
1211 if (e)
1212 arg_container_service_name = e;
1213
5a8ff0e6
CB		 1214 r = getenv_bool("SYSTEMD_NSPAWN_USE_CGNS");
1215 if (r < 0)
1216 arg_use_cgns = cg_ns_supported();
1217 else
1218 arg_use_cgns = r;
1219
86c0dd4a
LP		 1220 r = custom_mount_check_all();
1221 if (r < 0)
1222 return r;
1223
f757855e
LP		 1224 return 1;
1225}
1226
1227static int verify_arguments(void) {
4f086aab
SU		 1228 if (arg_userns_mode != USER_NAMESPACE_NO && (arg_mount_settings & MOUNT_APPLY_APIVFS_NETNS) && !arg_private_network) {
1229 log_error("Invalid namespacing settings. Mounting sysfs with --private-users requires --private-network.");
1230 return -EINVAL;
1231 }
1232
1233 if (arg_userns_mode != USER_NAMESPACE_NO && !(arg_mount_settings & MOUNT_APPLY_APIVFS_RO)) {
1234 log_error("Cannot combine --private-users with read-write mounts.");
1235 return -EINVAL;
1236 }
f757855e
LP		 1237
1238 if (arg_volatile_mode != VOLATILE_NO && arg_read_only) {
4d9f07b4
LP		 1239 log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy.");
1240 return -EINVAL;
1241 }
1242
6d0b55c2
LP		 1243 if (arg_expose_ports && !arg_private_network) {
1244 log_error("Cannot use --port= without private networking.");
1245 return -EINVAL;
1246 }
1247
349cc4a5 1248#if ! HAVE_LIBIPTC
1c1ea217
EV		 1249 if (arg_expose_ports) {
1250 log_error("--port= is not supported, compiled without libiptc support.");
1251 return -EOPNOTSUPP;
1252 }
1253#endif
1254
7732f92b 1255 if (arg_start_mode == START_BOOT && arg_kill_signal <= 0)
c6c8f6e2
LP		 1256 arg_kill_signal = SIGRTMIN+3;
1257
f757855e 1258 return 0;
88213476
LP		 1259}
1260
03cfe0d5
LP		 1261static int userns_lchown(const char *p, uid_t uid, gid_t gid) {
1262 assert(p);
1263
0de7acce 1264 if (arg_userns_mode == USER_NAMESPACE_NO)
03cfe0d5
LP		 1265 return 0;
1266
1267 if (uid == UID_INVALID && gid == GID_INVALID)
1268 return 0;
1269
1270 if (uid != UID_INVALID) {
1271 uid += arg_uid_shift;
1272
1273 if (uid < arg_uid_shift || uid >= arg_uid_shift + arg_uid_range)
1274 return -EOVERFLOW;
1275 }
1276
1277 if (gid != GID_INVALID) {
1278 gid += (gid_t) arg_uid_shift;
1279
1280 if (gid < (gid_t) arg_uid_shift || gid >= (gid_t) (arg_uid_shift + arg_uid_range))
1281 return -EOVERFLOW;
1282 }
1283
1284 if (lchown(p, uid, gid) < 0)
1285 return -errno;
b12afc8c
LP		 1286
1287 return 0;
1288}
1289
03cfe0d5
LP		 1290static int userns_mkdir(const char *root, const char *path, mode_t mode, uid_t uid, gid_t gid) {
1291 const char *q;
1292
1293 q = prefix_roota(root, path);
1294 if (mkdir(q, mode) < 0) {
1295 if (errno == EEXIST)
1296 return 0;
1297 return -errno;
1298 }
1299
1300 return userns_lchown(q, uid, gid);
1301}
1302
e58a1277 1303static int setup_timezone(const char *dest) {
03cfe0d5
LP		 1304 _cleanup_free_ char *p = NULL, *q = NULL;
1305 const char *where, *check, *what;
d4036145
LP		 1306 char *z, *y;
1307 int r;
f8440af5 1308
e58a1277
LP		 1309 assert(dest);
1310
1311 /* Fix the timezone, if possible */
d4036145
LP		 1312 r = readlink_malloc("/etc/localtime", &p);
1313 if (r < 0) {
0b493a02
MP		 1314 log_warning("host's /etc/localtime is not a symlink, not updating container timezone.");
1315 /* to handle warning, delete /etc/localtime and replace it
d23a0044 1316 * with a symbolic link to a time zone data file.
0b493a02
MP		 1317 *
1318 * Example:
21dc0227 1319 * ln -s /usr/share/zoneinfo/UTC /etc/localtime
0b493a02 1320 */
d4036145
LP		 1321 return 0;
1322 }
1323
1324 z = path_startswith(p, "../usr/share/zoneinfo/");
1325 if (!z)
1326 z = path_startswith(p, "/usr/share/zoneinfo/");
1327 if (!z) {
1328 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
1329 return 0;
1330 }
1331
03cfe0d5 1332 where = prefix_roota(dest, "/etc/localtime");
d4036145
LP		 1333 r = readlink_malloc(where, &q);
1334 if (r >= 0) {
1335 y = path_startswith(q, "../usr/share/zoneinfo/");
1336 if (!y)
1337 y = path_startswith(q, "/usr/share/zoneinfo/");
4d1c38b8 1338
d4036145
LP		 1339 /* Already pointing to the right place? Then do nothing .. */
1340 if (y && streq(y, z))
1341 return 0;
1342 }
1343
03cfe0d5 1344 check = strjoina("/usr/share/zoneinfo/", z);
61e741ed 1345 check = prefix_roota(dest, check);
03cfe0d5 1346 if (laccess(check, F_OK) < 0) {
d4036145
LP		 1347 log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
1348 return 0;
1349 }
68fb0892 1350
8ccf7e9e
LP		 1351 if (unlink(where) < 0 && errno != ENOENT) {
1352 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING, /* Don't complain on read-only images */
1353 errno,
1354 "Failed to remove existing timezone info %s in container, ignoring: %m", where);
79d80fc1
TG		 1355 return 0;
1356 }
4d9f07b4 1357
03cfe0d5 1358 what = strjoina("../usr/share/zoneinfo/", z);
d4036145 1359 if (symlink(what, where) < 0) {
8ccf7e9e
LP		 1360 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING,
1361 errno,
1362 "Failed to correct timezone of container, ignoring: %m");
d4036145
LP		 1363 return 0;
1364 }
e58a1277 1365
03cfe0d5
LP		 1366 r = userns_lchown(where, 0, 0);
1367 if (r < 0)
1368 return log_warning_errno(r, "Failed to chown /etc/localtime: %m");
1369
e58a1277 1370 return 0;
88213476
LP		 1371}
1372
7357272e 1373static int resolved_listening(void) {
b053cd5f 1374 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
7357272e 1375 _cleanup_free_ char *dns_stub_listener_mode = NULL;
b053cd5f
LP		 1376 int r;
1377
7357272e 1378 /* Check if resolved is listening */
b053cd5f
LP		 1379
1380 r = sd_bus_open_system(&bus);
1381 if (r < 0)
1382 return r;
1383
7357272e
DM		 1384 r = bus_name_has_owner(bus, "org.freedesktop.resolve1", NULL);
1385 if (r <= 0)
1386 return r;
1387
1388 r = sd_bus_get_property_string(bus,
1389 "org.freedesktop.resolve1",
1390 "/org/freedesktop/resolve1",
1391 "org.freedesktop.resolve1.Manager",
1392 "DNSStubListener",
1393 NULL,
1394 &dns_stub_listener_mode);
1395 if (r < 0)
1396 return r;
1397
1398 return STR_IN_SET(dns_stub_listener_mode, "udp", "yes");
b053cd5f
LP		 1399}
1400
2547bb41 1401static int setup_resolv_conf(const char *dest) {
87447ae4
LP		 1402 _cleanup_free_ char *resolved = NULL, *etc = NULL;
1403 const char *where;
1404 int r, found;
2547bb41
LP		 1405
1406 assert(dest);
1407
1408 if (arg_private_network)
1409 return 0;
1410
87447ae4
LP		 1411 r = chase_symlinks("/etc", dest, CHASE_PREFIX_ROOT, &etc);
1412 if (r < 0) {
1413 log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
1414 return 0;
1415 }
1416
1417 where = strjoina(etc, "/resolv.conf");
1418 found = chase_symlinks(where, dest, CHASE_NONEXISTENT, &resolved);
1419 if (found < 0) {
1420 log_warning_errno(found, "Failed to resolve /etc/resolv.conf path in container, ignoring: %m");
1421 return 0;
1422 }
79d80fc1 1423
62b1e758 1424 if (access(STATIC_RESOLV_CONF, F_OK) >= 0 &&
7357272e 1425 resolved_listening() > 0) {
87447ae4 1426
3539724c
LP		 1427 /* resolved is enabled on the host. In this, case bind mount its static resolv.conf file into the
1428 * container, so that the container can use the host's resolver. Given that network namespacing is
1429 * disabled it's only natural of the container also uses the host's resolver. It also has the big
1430 * advantage that the container will be able to follow the host's DNS server configuration changes
1431 * transparently. */
1432
87447ae4
LP		 1433 if (found == 0) /* missing? */
1434 (void) touch(resolved);
5367354d 1435
62b1e758 1436 r = mount_verbose(LOG_DEBUG, STATIC_RESOLV_CONF, resolved, NULL, MS_BIND, NULL);
60e76d48 1437 if (r >= 0)
87447ae4 1438 return mount_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
3539724c
LP		 1439 }
1440
1441 /* If that didn't work, let's copy the file */
1c876927 1442 r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0, COPY_REFLINK);
79d80fc1 1443 if (r < 0) {
3539724c
LP		 1444 /* If the file already exists as symlink, let's suppress the warning, under the assumption that
1445 * resolved or something similar runs inside and the symlink points there.
68a313c5 1446 *
3539724c 1447 * If the disk image is read-only, there's also no point in complaining.
68a313c5 1448 */
87447ae4 1449 log_full_errno(IN_SET(r, -ELOOP, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
3539724c 1450 "Failed to copy /etc/resolv.conf to %s, ignoring: %m", where);
79d80fc1
TG		 1451 return 0;
1452 }
2547bb41 1453
03cfe0d5
LP		 1454 r = userns_lchown(where, 0, 0);
1455 if (r < 0)
3539724c 1456 log_warning_errno(r, "Failed to chown /etc/resolv.conf, ignoring: %m");
03cfe0d5 1457
2547bb41
LP		 1458 return 0;
1459}
1460
04bc4a3f 1461static int setup_boot_id(const char *dest) {
3bbaff3e 1462 sd_id128_t rnd = SD_ID128_NULL;
03cfe0d5 1463 const char *from, *to;
04bc4a3f
LP		 1464 int r;
1465
04bc4a3f
LP		 1466 /* Generate a new randomized boot ID, so that each boot-up of
1467 * the container gets a new one */
1468
03cfe0d5
LP		 1469 from = prefix_roota(dest, "/run/proc-sys-kernel-random-boot-id");
1470 to = prefix_roota(dest, "/proc/sys/kernel/random/boot_id");
04bc4a3f
LP		 1471
1472 r = sd_id128_randomize(&rnd);
f647962d
MS		 1473 if (r < 0)
1474 return log_error_errno(r, "Failed to generate random boot id: %m");
04bc4a3f 1475
15b1248a 1476 r = id128_write(from, ID128_UUID, rnd, false);
f647962d
MS		 1477 if (r < 0)
1478 return log_error_errno(r, "Failed to write boot id: %m");
04bc4a3f 1479
60e76d48
ZJS		 1480 r = mount_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL);
1481 if (r >= 0)
1482 r = mount_verbose(LOG_ERR, NULL, to, NULL,
1483 MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
04bc4a3f 1484
3bbaff3e 1485 (void) unlink(from);
04bc4a3f
LP		 1486 return r;
1487}
1488
e58a1277 1489static int copy_devnodes(const char *dest) {
88213476
LP		 1490
1491 static const char devnodes[] =
1492 "null\0"
1493 "zero\0"
1494 "full\0"
1495 "random\0"
1496 "urandom\0"
85614d66
TG		 1497 "tty\0"
1498 "net/tun\0";
88213476
LP		 1499
1500 const char *d;
e58a1277 1501 int r = 0;
7fd1b19b 1502 _cleanup_umask_ mode_t u;
a258bf26
LP		 1503
1504 assert(dest);
124640f1
LP		 1505
1506 u = umask(0000);
88213476 1507
03cfe0d5
LP		 1508 /* Create /dev/net, so that we can create /dev/net/tun in it */
1509 if (userns_mkdir(dest, "/dev/net", 0755, 0, 0) < 0)
1510 return log_error_errno(r, "Failed to create /dev/net directory: %m");
1511
88213476 1512 NULSTR_FOREACH(d, devnodes) {
7fd1b19b 1513 _cleanup_free_ char *from = NULL, *to = NULL;
7f112f50 1514 struct stat st;
88213476 1515
7f112f50 1516 from = strappend("/dev/", d);
03cfe0d5 1517 to = prefix_root(dest, from);
88213476
LP		 1518
1519 if (stat(from, &st) < 0) {
1520
4a62c710
MS		 1521 if (errno != ENOENT)
1522 return log_error_errno(errno, "Failed to stat %s: %m", from);
88213476 1523
a258bf26 1524 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
88213476 1525
03cfe0d5 1526 log_error("%s is not a char or block device, cannot copy.", from);
7f112f50 1527 return -EIO;
a258bf26 1528
85614d66 1529 } else {
81f5049b 1530 if (mknod(to, st.st_mode, st.st_rdev) < 0) {
8dbf71ec 1531 /* Explicitly warn the user when /dev is already populated. */
41eb4362 1532 if (errno == EEXIST)
8dbf71ec 1533 log_notice("%s/dev is pre-mounted and pre-populated. If a pre-mounted /dev is provided it needs to be an unpopulated file system.", dest);
81f5049b
AC		 1534 if (errno != EPERM)
1535 return log_error_errno(errno, "mknod(%s) failed: %m", to);
1536
1537 /* Some systems abusively restrict mknod but
1538 * allow bind mounts. */
1539 r = touch(to);
1540 if (r < 0)
1541 return log_error_errno(r, "touch (%s) failed: %m", to);
60e76d48
ZJS		 1542 r = mount_verbose(LOG_DEBUG, from, to, NULL, MS_BIND, NULL);
1543 if (r < 0)
1544 return log_error_errno(r, "Both mknod and bind mount (%s) failed: %m", to);
81f5049b 1545 }
6278cf60 1546
03cfe0d5
LP		 1547 r = userns_lchown(to, 0, 0);
1548 if (r < 0)
1549 return log_error_errno(r, "chown() of device node %s failed: %m", to);
88213476 1550 }
88213476
LP		 1551 }
1552
e58a1277
LP		 1553 return r;
1554}
88213476 1555
03cfe0d5
LP		 1556static int setup_pts(const char *dest) {
1557 _cleanup_free_ char *options = NULL;
1558 const char *p;
709f6e46 1559 int r;
03cfe0d5 1560
349cc4a5 1561#if HAVE_SELINUX
03cfe0d5
LP		 1562 if (arg_selinux_apifs_context)
1563 (void) asprintf(&options,
3dce8915 1564 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"",
03cfe0d5
LP		 1565 arg_uid_shift + TTY_GID,
1566 arg_selinux_apifs_context);
1567 else
1568#endif
1569 (void) asprintf(&options,
3dce8915 1570 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT,
03cfe0d5 1571 arg_uid_shift + TTY_GID);
f2d88580 1572
03cfe0d5 1573 if (!options)
f2d88580
LP		 1574 return log_oom();
1575
03cfe0d5 1576 /* Mount /dev/pts itself */
cc9fce65 1577 p = prefix_roota(dest, "/dev/pts");
03cfe0d5
LP		 1578 if (mkdir(p, 0755) < 0)
1579 return log_error_errno(errno, "Failed to create /dev/pts: %m");
60e76d48
ZJS		 1580 r = mount_verbose(LOG_ERR, "devpts", p, "devpts", MS_NOSUID|MS_NOEXEC, options);
1581 if (r < 0)
1582 return r;
709f6e46
MS		 1583 r = userns_lchown(p, 0, 0);
1584 if (r < 0)
1585 return log_error_errno(r, "Failed to chown /dev/pts: %m");
03cfe0d5
LP		 1586
1587 /* Create /dev/ptmx symlink */
1588 p = prefix_roota(dest, "/dev/ptmx");
4a62c710
MS		 1589 if (symlink("pts/ptmx", p) < 0)
1590 return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
709f6e46
MS		 1591 r = userns_lchown(p, 0, 0);
1592 if (r < 0)
1593 return log_error_errno(r, "Failed to chown /dev/ptmx: %m");
f2d88580 1594
03cfe0d5
LP		 1595 /* And fix /dev/pts/ptmx ownership */
1596 p = prefix_roota(dest, "/dev/pts/ptmx");
709f6e46
MS		 1597 r = userns_lchown(p, 0, 0);
1598 if (r < 0)
1599 return log_error_errno(r, "Failed to chown /dev/pts/ptmx: %m");
6278cf60 1600
f2d88580
LP		 1601 return 0;
1602}
1603
e58a1277 1604static int setup_dev_console(const char *dest, const char *console) {
eb0f0863
LP		 1605 _cleanup_umask_ mode_t u;
1606 const char *to;
e58a1277 1607 int r;
e58a1277
LP		 1608
1609 assert(dest);
1610 assert(console);
1611
1612 u = umask(0000);
1613
03cfe0d5 1614 r = chmod_and_chown(console, 0600, arg_uid_shift, arg_uid_shift);
f647962d
MS		 1615 if (r < 0)
1616 return log_error_errno(r, "Failed to correct access mode for TTY: %m");
88213476 1617
a258bf26
LP		 1618 /* We need to bind mount the right tty to /dev/console since
1619 * ptys can only exist on pts file systems. To have something
81f5049b 1620 * to bind mount things on we create a empty regular file. */
a258bf26 1621
03cfe0d5 1622 to = prefix_roota(dest, "/dev/console");
81f5049b
AC		 1623 r = touch(to);
1624 if (r < 0)
1625 return log_error_errno(r, "touch() for /dev/console failed: %m");
a258bf26 1626
60e76d48 1627 return mount_verbose(LOG_ERR, console, to, NULL, MS_BIND, NULL);
e58a1277
LP		 1628}
1629
8e5430c4
LP		 1630static int setup_keyring(void) {
1631 key_serial_t keyring;
1632
1633 /* Allocate a new session keyring for the container. This makes sure the keyring of the session systemd-nspawn
1634 * was invoked from doesn't leak into the container. Note that by default we block keyctl() and request_key()
1635 * anyway via seccomp so doing this operation isn't strictly necessary, but in case people explicitly whitelist
1636 * these system calls let's make sure we don't leak anything into the container. */
1637
1638 keyring = keyctl(KEYCTL_JOIN_SESSION_KEYRING, 0, 0, 0, 0);
1639 if (keyring == -1) {
1640 if (errno == ENOSYS)
1641 log_debug_errno(errno, "Kernel keyring not supported, ignoring.");
1642 else if (IN_SET(errno, EACCES, EPERM))
1643 log_debug_errno(errno, "Kernel keyring access prohibited, ignoring.");
1644 else
1645 return log_error_errno(errno, "Setting up kernel keyring failed: %m");
1646 }
1647
1648 return 0;
1649}
1650
e58a1277 1651static int setup_kmsg(const char *dest, int kmsg_socket) {
03cfe0d5 1652 const char *from, *to;
7fd1b19b 1653 _cleanup_umask_ mode_t u;
d9603714 1654 int fd, r;
e58a1277 1655
e58a1277 1656 assert(kmsg_socket >= 0);
a258bf26 1657
e58a1277 1658 u = umask(0000);
a258bf26 1659
03cfe0d5 1660 /* We create the kmsg FIFO as /run/kmsg, but immediately
f1e5dfe2
LP		 1661 * delete it after bind mounting it to /proc/kmsg. While FIFOs
1662 * on the reading side behave very similar to /proc/kmsg,
1663 * their writing side behaves differently from /dev/kmsg in
1664 * that writing blocks when nothing is reading. In order to
1665 * avoid any problems with containers deadlocking due to this
1666 * we simply make /dev/kmsg unavailable to the container. */
03cfe0d5
LP		 1667 from = prefix_roota(dest, "/run/kmsg");
1668 to = prefix_roota(dest, "/proc/kmsg");
e58a1277 1669
4a62c710 1670 if (mkfifo(from, 0600) < 0)
03cfe0d5 1671 return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m");
60e76d48
ZJS		 1672 r = mount_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL);
1673 if (r < 0)
1674 return r;
e58a1277
LP		 1675
1676 fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
4a62c710
MS		 1677 if (fd < 0)
1678 return log_error_errno(errno, "Failed to open fifo: %m");
e58a1277 1679
e58a1277
LP		 1680 /* Store away the fd in the socket, so that it stays open as
1681 * long as we run the child */
3ee897d6 1682 r = send_one_fd(kmsg_socket, fd, 0);
03e334a1 1683 safe_close(fd);
e58a1277 1684
d9603714
DH		 1685 if (r < 0)
1686 return log_error_errno(r, "Failed to send FIFO fd: %m");
a258bf26 1687
03cfe0d5
LP		 1688 /* And now make the FIFO unavailable as /run/kmsg... */
1689 (void) unlink(from);
1690
25ea79fe 1691 return 0;
88213476
LP		 1692}
1693
1c4baffc 1694static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) {
6d0b55c2
LP		 1695 union in_addr_union *exposed = userdata;
1696
1697 assert(rtnl);
1698 assert(m);
1699 assert(exposed);
1700
7a8f6325 1701 expose_port_execute(rtnl, arg_expose_ports, exposed);
6d0b55c2
LP		 1702 return 0;
1703}
1704
3a74cea5 1705static int setup_hostname(void) {
3a74cea5 1706
0c582db0 1707 if ((arg_clone_ns_flags & CLONE_NEWUTS) == 0)
eb91eb18
LP		 1708 return 0;
1709
605f81a8 1710 if (sethostname_idempotent(arg_machine) < 0)
7027ff61 1711 return -errno;
3a74cea5 1712
7027ff61 1713 return 0;
3a74cea5
LP		 1714}
1715
57fb9fb5 1716static int setup_journal(const char *directory) {
e01ff70a 1717 sd_id128_t this_id;
0f5e1382 1718 _cleanup_free_ char *d = NULL;
e01ff70a 1719 const char *p, *q;
8054d749 1720 bool try;
e01ff70a 1721 char id[33];
57fb9fb5
LP		 1722 int r;
1723
df9a75e4
LP		 1724 /* Don't link journals in ephemeral mode */
1725 if (arg_ephemeral)
1726 return 0;
1727
8054d749
LP		 1728 if (arg_link_journal == LINK_NO)
1729 return 0;
1730
1731 try = arg_link_journal_try || arg_link_journal == LINK_AUTO;
1732
4d680aee 1733 r = sd_id128_get_machine(&this_id);
f647962d
MS		 1734 if (r < 0)
1735 return log_error_errno(r, "Failed to retrieve machine ID: %m");
4d680aee 1736
e01ff70a 1737 if (sd_id128_equal(arg_uuid, this_id)) {
8054d749 1738 log_full(try ? LOG_WARNING : LOG_ERR,
e192a281 1739 "Host and machine ids are equal (%s): refusing to link journals", sd_id128_to_string(arg_uuid, id));
8054d749 1740 if (try)
4d680aee 1741 return 0;
df9a75e4 1742 return -EEXIST;
4d680aee
ZJS		 1743 }
1744
03cfe0d5
LP		 1745 r = userns_mkdir(directory, "/var", 0755, 0, 0);
1746 if (r < 0)
1747 return log_error_errno(r, "Failed to create /var: %m");
1748
1749 r = userns_mkdir(directory, "/var/log", 0755, 0, 0);
1750 if (r < 0)
1751 return log_error_errno(r, "Failed to create /var/log: %m");
1752
1753 r = userns_mkdir(directory, "/var/log/journal", 0755, 0, 0);
1754 if (r < 0)
1755 return log_error_errno(r, "Failed to create /var/log/journal: %m");
1756
e01ff70a
MS		 1757 (void) sd_id128_to_string(arg_uuid, id);
1758
03cfe0d5
LP		 1759 p = strjoina("/var/log/journal/", id);
1760 q = prefix_roota(directory, p);
27407a01 1761
e1873695 1762 if (path_is_mount_point(p, NULL, 0) > 0) {
8054d749
LP		 1763 if (try)
1764 return 0;
27407a01 1765
8054d749
LP		 1766 log_error("%s: already a mount point, refusing to use for journal", p);
1767 return -EEXIST;
57fb9fb5
LP		 1768 }
1769
e1873695 1770 if (path_is_mount_point(q, NULL, 0) > 0) {
8054d749
LP		 1771 if (try)
1772 return 0;
57fb9fb5 1773
8054d749
LP		 1774 log_error("%s: already a mount point, refusing to use for journal", q);
1775 return -EEXIST;
57fb9fb5
LP		 1776 }
1777
1778 r = readlink_and_make_absolute(p, &d);
1779 if (r >= 0) {
3742095b 1780 if (IN_SET(arg_link_journal, LINK_GUEST, LINK_AUTO) &&
57fb9fb5
LP		 1781 path_equal(d, q)) {
1782
03cfe0d5 1783 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 1784 if (r < 0)
709f6e46 1785 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 1786 return 0;
57fb9fb5
LP		 1787 }
1788
4a62c710
MS		 1789 if (unlink(p) < 0)
1790 return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
57fb9fb5
LP		 1791 } else if (r == -EINVAL) {
1792
1793 if (arg_link_journal == LINK_GUEST &&
1794 rmdir(p) < 0) {
1795
27407a01
ZJS		 1796 if (errno == ENOTDIR) {
1797 log_error("%s already exists and is neither a symlink nor a directory", p);
1798 return r;
4314d33f
MS		 1799 } else
1800 return log_error_errno(errno, "Failed to remove %s: %m", p);
57fb9fb5 1801 }
4314d33f
MS		 1802 } else if (r != -ENOENT)
1803 return log_error_errno(r, "readlink(%s) failed: %m", p);
57fb9fb5
LP		 1804
1805 if (arg_link_journal == LINK_GUEST) {
1806
1807 if (symlink(q, p) < 0) {
8054d749 1808 if (try) {
56f64d95 1809 log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
574edc90 1810 return 0;
4314d33f
MS		 1811 } else
1812 return log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
57fb9fb5
LP		 1813 }
1814
03cfe0d5 1815 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 1816 if (r < 0)
709f6e46 1817 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 1818 return 0;
57fb9fb5
LP		 1819 }
1820
1821 if (arg_link_journal == LINK_HOST) {
ccddd104 1822 /* don't create parents here — if the host doesn't have
574edc90 1823 * permanent journal set up, don't force it here */
ba8e6c4d
LP		 1824
1825 if (mkdir(p, 0755) < 0 && errno != EEXIST) {
8054d749 1826 if (try) {
56f64d95 1827 log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p);
574edc90 1828 return 0;
4314d33f
MS		 1829 } else
1830 return log_error_errno(errno, "Failed to create %s: %m", p);
57fb9fb5
LP		 1831 }
1832
27407a01
ZJS		 1833 } else if (access(p, F_OK) < 0)
1834 return 0;
57fb9fb5 1835
cdb2b9d0
LP		 1836 if (dir_is_empty(q) == 0)
1837 log_warning("%s is not empty, proceeding anyway.", q);
1838
03cfe0d5 1839 r = userns_mkdir(directory, p, 0755, 0, 0);
709f6e46
MS		 1840 if (r < 0)
1841 return log_error_errno(r, "Failed to create %s: %m", q);
57fb9fb5 1842
60e76d48
ZJS		 1843 r = mount_verbose(LOG_DEBUG, p, q, NULL, MS_BIND, NULL);
1844 if (r < 0)
4a62c710 1845 return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
57fb9fb5 1846
27407a01 1847 return 0;
57fb9fb5
LP		 1848}
1849
88213476 1850static int drop_capabilities(void) {
520e0d54 1851 return capability_bounding_set_drop(arg_caps_retain, false);
88213476
LP		 1852}
1853
db999e0f
LP		 1854static int reset_audit_loginuid(void) {
1855 _cleanup_free_ char *p = NULL;
1856 int r;
1857
0c582db0 1858 if ((arg_clone_ns_flags & CLONE_NEWPID) == 0)
db999e0f
LP		 1859 return 0;
1860
1861 r = read_one_line_file("/proc/self/loginuid", &p);
13e8ceb8 1862 if (r == -ENOENT)
db999e0f 1863 return 0;
f647962d
MS		 1864 if (r < 0)
1865 return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
db999e0f
LP		 1866
1867 /* Already reset? */
1868 if (streq(p, "4294967295"))
1869 return 0;
1870
ad118bda 1871 r = write_string_file("/proc/self/loginuid", "4294967295", 0);
db999e0f 1872 if (r < 0) {
10a87006
LP		 1873 log_error_errno(r,
1874 "Failed to reset audit login UID. This probably means that your kernel is too\n"
1875 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
1876 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
1877 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
1878 "using systemd-nspawn. Sleeping for 5s... (%m)");
77b6e194 1879
db999e0f 1880 sleep(5);
77b6e194 1881 }
db999e0f
LP		 1882
1883 return 0;
77b6e194
LP		 1884}
1885
24fb1112 1886
785890ac
LP		 1887static int setup_propagate(const char *root) {
1888 const char *p, *q;
709f6e46 1889 int r;
785890ac
LP		 1890
1891 (void) mkdir_p("/run/systemd/nspawn/", 0755);
1892 (void) mkdir_p("/run/systemd/nspawn/propagate", 0600);
63c372cb 1893 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
785890ac
LP		 1894 (void) mkdir_p(p, 0600);
1895
709f6e46
MS		 1896 r = userns_mkdir(root, "/run/systemd", 0755, 0, 0);
1897 if (r < 0)
1898 return log_error_errno(r, "Failed to create /run/systemd: %m");
03cfe0d5 1899
709f6e46
MS		 1900 r = userns_mkdir(root, "/run/systemd/nspawn", 0755, 0, 0);
1901 if (r < 0)
1902 return log_error_errno(r, "Failed to create /run/systemd/nspawn: %m");
03cfe0d5 1903
709f6e46
MS		 1904 r = userns_mkdir(root, "/run/systemd/nspawn/incoming", 0600, 0, 0);
1905 if (r < 0)
1906 return log_error_errno(r, "Failed to create /run/systemd/nspawn/incoming: %m");
785890ac 1907
03cfe0d5 1908 q = prefix_roota(root, "/run/systemd/nspawn/incoming");
60e76d48
ZJS		 1909 r = mount_verbose(LOG_ERR, p, q, NULL, MS_BIND, NULL);
1910 if (r < 0)
1911 return r;
785890ac 1912
60e76d48
ZJS		 1913 r = mount_verbose(LOG_ERR, NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
1914 if (r < 0)
1915 return r;
785890ac 1916
19caffac
AC		 1917 /* machined will MS_MOVE into that directory, and that's only
1918 * supported for non-shared mounts. */
60e76d48 1919 return mount_verbose(LOG_ERR, NULL, q, NULL, MS_SLAVE, NULL);
785890ac
LP		 1920}
1921
317feb4d 1922static int setup_machine_id(const char *directory) {
691675ba
LP		 1923 const char *etc_machine_id;
1924 sd_id128_t id;
3bbaff3e 1925 int r;
e01ff70a 1926
317feb4d
LP		 1927 /* If the UUID in the container is already set, then that's what counts, and we use. If it isn't set, and the
1928 * caller passed --uuid=, then we'll pass it in the $container_uuid env var to PID 1 of the container. The
1929 * assumption is that PID 1 will then write it to /etc/machine-id to make it persistent. If --uuid= is not
1930 * passed we generate a random UUID, and pass it via $container_uuid. In effect this means that /etc/machine-id
1931 * in the container and our idea of the container UUID will always be in sync (at least if PID 1 in the
1932 * container behaves nicely). */
1933
e01ff70a
MS		 1934 etc_machine_id = prefix_roota(directory, "/etc/machine-id");
1935
691675ba 1936 r = id128_read(etc_machine_id, ID128_PLAIN, &id);
317feb4d
LP		 1937 if (r < 0) {
1938 if (!IN_SET(r, -ENOENT, -ENOMEDIUM)) /* If the file is missing or empty, we don't mind */
1939 return log_error_errno(r, "Failed to read machine ID from container image: %m");
691675ba 1940
317feb4d
LP		 1941 if (sd_id128_is_null(arg_uuid)) {
1942 r = sd_id128_randomize(&arg_uuid);
1943 if (r < 0)
1944 return log_error_errno(r, "Failed to acquire randomized machine UUID: %m");
1945 }
1946 } else {
1947 if (sd_id128_is_null(id)) {
1948 log_error("Machine ID in container image is zero, refusing.");
1949 return -EINVAL;
1950 }
e01ff70a 1951
317feb4d
LP		 1952 arg_uuid = id;
1953 }
691675ba 1954
e01ff70a
MS		 1955 return 0;
1956}
1957
7336138e
LP		 1958static int recursive_chown(const char *directory, uid_t shift, uid_t range) {
1959 int r;
1960
1961 assert(directory);
1962
0de7acce 1963 if (arg_userns_mode == USER_NAMESPACE_NO || !arg_userns_chown)
7336138e
LP		 1964 return 0;
1965
1966 r = path_patch_uid(directory, arg_uid_shift, arg_uid_range);
1967 if (r == -EOPNOTSUPP)
1968 return log_error_errno(r, "Automatic UID/GID adjusting is only supported for UID/GID ranges starting at multiples of 2^16 with a range of 2^16.");
1969 if (r == -EBADE)
1970 return log_error_errno(r, "Upper 16 bits of root directory UID and GID do not match.");
1971 if (r < 0)
1972 return log_error_errno(r, "Failed to adjust UID/GID shift of OS tree: %m");
1973 if (r == 0)
1974 log_debug("Root directory of image is already owned by the right UID/GID range, skipping recursive chown operation.");
1975 else
1976 log_debug("Patched directory tree to match UID/GID range.");
1977
1978 return r;
1979}
1980
113cea80 1981/*
6d416b9c
LS		 1982 * Return values:
1983 * < 0 : wait_for_terminate() failed to get the state of the
1984 * container, the container was terminated by a signal, or
1985 * failed for an unknown reason. No change is made to the
1986 * container argument.
1987 * > 0 : The program executed in the container terminated with an
1988 * error. The exit code of the program executed in the
919699ec
LP		 1989 * container is returned. The container argument has been set
1990 * to CONTAINER_TERMINATED.
6d416b9c
LS		 1991 * 0 : The container is being rebooted, has been shut down or exited
1992 * successfully. The container argument has been set to either
1993 * CONTAINER_TERMINATED or CONTAINER_REBOOTED.
113cea80 1994 *
6d416b9c
LS		 1995 * That is, success is indicated by a return value of zero, and an
1996 * error is indicated by a non-zero value.
113cea80
DH		 1997 */
1998static int wait_for_container(pid_t pid, ContainerStatus *container) {
113cea80 1999 siginfo_t status;
919699ec 2000 int r;
113cea80
DH		 2001
2002 r = wait_for_terminate(pid, &status);
f647962d
MS		 2003 if (r < 0)
2004 return log_warning_errno(r, "Failed to wait for container: %m");
113cea80
DH		 2005
2006 switch (status.si_code) {
fddbb89c 2007
113cea80 2008 case CLD_EXITED:
b5a2179b 2009 if (status.si_status == 0)
919699ec 2010 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s exited successfully.", arg_machine);
b5a2179b 2011 else
919699ec 2012 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s failed with error code %i.", arg_machine, status.si_status);
fddbb89c 2013
919699ec
LP		 2014 *container = CONTAINER_TERMINATED;
2015 return status.si_status;
113cea80
DH		 2016
2017 case CLD_KILLED:
2018 if (status.si_status == SIGINT) {
919699ec 2019 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s has been shut down.", arg_machine);
113cea80 2020 *container = CONTAINER_TERMINATED;
919699ec
LP		 2021 return 0;
2022
113cea80 2023 } else if (status.si_status == SIGHUP) {
919699ec 2024 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s is being rebooted.", arg_machine);
113cea80 2025 *container = CONTAINER_REBOOTED;
919699ec 2026 return 0;
113cea80 2027 }
919699ec 2028
4831981d 2029 _fallthrough_;
113cea80 2030 case CLD_DUMPED:
fddbb89c 2031 log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
919699ec 2032 return -EIO;
113cea80
DH		 2033
2034 default:
fddbb89c 2035 log_error("Container %s failed due to unknown reason.", arg_machine);
919699ec 2036 return -EIO;
113cea80 2037 }
113cea80
DH		 2038}
2039
023fb90b
LP		 2040static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
2041 pid_t pid;
2042
4a0b58c4 2043 pid = PTR_TO_PID(userdata);
023fb90b 2044 if (pid > 0) {
c6c8f6e2 2045 if (kill(pid, arg_kill_signal) >= 0) {
023fb90b
LP		 2046 log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
2047 sd_event_source_set_userdata(s, NULL);
2048 return 0;
2049 }
2050 }
2051
2052 sd_event_exit(sd_event_source_get_event(s), 0);
2053 return 0;
2054}
2055
6916b164 2056static int on_sigchld(sd_event_source *s, const struct signalfd_siginfo *ssi, void *userdata) {
abdb9b08
LP		 2057 pid_t pid;
2058
2059 assert(s);
2060 assert(ssi);
2061
2062 pid = PTR_TO_PID(userdata);
2063
6916b164
AU		 2064 for (;;) {
2065 siginfo_t si = {};
abdb9b08 2066
6916b164
AU		 2067 if (waitid(P_ALL, 0, &si, WNOHANG|WNOWAIT|WEXITED) < 0)
2068 return log_error_errno(errno, "Failed to waitid(): %m");
2069 if (si.si_pid == 0) /* No pending children. */
2070 break;
abdb9b08 2071 if (si.si_pid == pid) {
6916b164
AU		 2072 /* The main process we care for has exited. Return from
2073 * signal handler but leave the zombie. */
2074 sd_event_exit(sd_event_source_get_event(s), 0);
2075 break;
2076 }
abdb9b08 2077
6916b164
AU		 2078 /* Reap all other children. */
2079 (void) waitid(P_PID, si.si_pid, &si, WNOHANG|WEXITED);
2080 }
2081
2082 return 0;
2083}
2084
abdb9b08
LP		 2085static int on_request_stop(sd_bus_message *m, void *userdata, sd_bus_error *error) {
2086 pid_t pid;
2087
2088 assert(m);
2089
2090 pid = PTR_TO_PID(userdata);
2091
2092 if (arg_kill_signal > 0) {
2093 log_info("Container termination requested. Attempting to halt container.");
2094 (void) kill(pid, arg_kill_signal);
2095 } else {
2096 log_info("Container termination requested. Exiting.");
2097 sd_event_exit(sd_bus_get_event(sd_bus_message_get_bus(m)), 0);
2098 }
2099
2100 return 0;
2101}
2102
ec16945e 2103static int determine_names(void) {
1b9cebf6 2104 int r;
ec16945e 2105
c1521918
LP		 2106 if (arg_template && !arg_directory && arg_machine) {
2107
2108 /* If --template= was specified then we should not
2109 * search for a machine, but instead create a new one
2110 * in /var/lib/machine. */
2111
605405c6 2112 arg_directory = strjoin("/var/lib/machines/", arg_machine);
c1521918
LP		 2113 if (!arg_directory)
2114 return log_oom();
2115 }
2116
ec16945e 2117 if (!arg_image && !arg_directory) {
1b9cebf6
LP		 2118 if (arg_machine) {
2119 _cleanup_(image_unrefp) Image *i = NULL;
2120
2121 r = image_find(arg_machine, &i);
2122 if (r < 0)
2123 return log_error_errno(r, "Failed to find image for machine '%s': %m", arg_machine);
0f3be6ca 2124 if (r == 0) {
35bca925 2125 log_error("No image for machine '%s'.", arg_machine);
1b9cebf6
LP		 2126 return -ENOENT;
2127 }
2128
eb38edce 2129 if (IN_SET(i->type, IMAGE_RAW, IMAGE_BLOCK))
0f03c2a4 2130 r = free_and_strdup(&arg_image, i->path);
1b9cebf6 2131 else
0f03c2a4 2132 r = free_and_strdup(&arg_directory, i->path);
1b9cebf6 2133 if (r < 0)
0f3be6ca 2134 return log_oom();
1b9cebf6 2135
aee327b8
LP		 2136 if (!arg_ephemeral)
2137 arg_read_only = arg_read_only || i->read_only;
1b9cebf6 2138 } else
ec16945e
LP		 2139 arg_directory = get_current_dir_name();
2140
0f3be6ca 2141 if (!arg_directory && !arg_image) {
1b9cebf6 2142 log_error("Failed to determine path, please use -D or -i.");
ec16945e
LP		 2143 return -EINVAL;
2144 }
2145 }
2146
2147 if (!arg_machine) {
4827ab48 2148
b9ba4dab
LP		 2149 if (arg_directory && path_equal(arg_directory, "/"))
2150 arg_machine = gethostname_malloc();
4827ab48
LP		 2151 else {
2152 if (arg_image) {
2153 char *e;
2154
2155 arg_machine = strdup(basename(arg_image));
2156
2157 /* Truncate suffix if there is one */
2158 e = endswith(arg_machine, ".raw");
2159 if (e)
2160 *e = 0;
2161 } else
2162 arg_machine = strdup(basename(arg_directory));
2163 }
ec16945e
LP		 2164 if (!arg_machine)
2165 return log_oom();
2166
ae691c1d 2167 hostname_cleanup(arg_machine);
ec16945e
LP		 2168 if (!machine_name_is_valid(arg_machine)) {
2169 log_error("Failed to determine machine name automatically, please use -M.");
2170 return -EINVAL;
2171 }
b9ba4dab
LP		 2172
2173 if (arg_ephemeral) {
2174 char *b;
2175
2176 /* Add a random suffix when this is an
2177 * ephemeral machine, so that we can run many
2178 * instances at once without manually having
2179 * to specify -M each time. */
2180
2181 if (asprintf(&b, "%s-%016" PRIx64, arg_machine, random_u64()) < 0)
2182 return log_oom();
2183
2184 free(arg_machine);
2185 arg_machine = b;
2186 }
ec16945e
LP		 2187 }
2188
2189 return 0;
2190}
2191
8d4aa2bb 2192static int chase_symlinks_and_update(char **p, unsigned flags) {
3f342ec4
LP		 2193 char *chased;
2194 int r;
2195
2196 assert(p);
2197
2198 if (!*p)
2199 return 0;
2200
8d4aa2bb 2201 r = chase_symlinks(*p, NULL, flags, &chased);
3f342ec4
LP		 2202 if (r < 0)
2203 return log_error_errno(r, "Failed to resolve path %s: %m", *p);
2204
2205 free(*p);
2206 *p = chased;
2207
2208 return 0;
2209}
2210
03cfe0d5 2211static int determine_uid_shift(const char *directory) {
6dac160c
LP		 2212 int r;
2213
0de7acce 2214 if (arg_userns_mode == USER_NAMESPACE_NO) {
03cfe0d5 2215 arg_uid_shift = 0;
6dac160c 2216 return 0;
03cfe0d5 2217 }
6dac160c
LP		 2218
2219 if (arg_uid_shift == UID_INVALID) {
2220 struct stat st;
2221
03cfe0d5 2222 r = stat(directory, &st);
6dac160c 2223 if (r < 0)
03cfe0d5 2224 return log_error_errno(errno, "Failed to determine UID base of %s: %m", directory);
6dac160c
LP		 2225
2226 arg_uid_shift = st.st_uid & UINT32_C(0xffff0000);
2227
2228 if (arg_uid_shift != (st.st_gid & UINT32_C(0xffff0000))) {
03cfe0d5 2229 log_error("UID and GID base of %s don't match.", directory);
6dac160c
LP		 2230 return -EINVAL;
2231 }
2232
2233 arg_uid_range = UINT32_C(0x10000);
2234 }
2235
2236 if (arg_uid_shift > (uid_t) -1 - arg_uid_range) {
2237 log_error("UID base too high for UID range.");
2238 return -EINVAL;
2239 }
2240
6dac160c
LP		 2241 return 0;
2242}
2243
03cfe0d5
LP		 2244static int inner_child(
2245 Barrier *barrier,
2246 const char *directory,
2247 bool secondary,
2248 int kmsg_socket,
2249 int rtnl_socket,
f757855e 2250 FDSet *fds) {
69c79d3c 2251
03cfe0d5 2252 _cleanup_free_ char *home = NULL;
e01ff70a 2253 char as_uuid[37];
6aadfa4c 2254 unsigned n_env = 1;
03cfe0d5
LP		 2255 const char *envp[] = {
2256 "PATH=" DEFAULT_PATH_SPLIT_USR,
6aadfa4c 2257 NULL, /* container */
03cfe0d5
LP		 2258 NULL, /* TERM */
2259 NULL, /* HOME */
2260 NULL, /* USER */
2261 NULL, /* LOGNAME */
2262 NULL, /* container_uuid */
2263 NULL, /* LISTEN_FDS */
2264 NULL, /* LISTEN_PID */
9c1e04d0 2265 NULL, /* NOTIFY_SOCKET */
03cfe0d5
LP		 2266 NULL
2267 };
1a68e1e5 2268 const char *exec_target;
88213476 2269
2371271c 2270 _cleanup_strv_free_ char **env_use = NULL;
03cfe0d5 2271 int r;
88213476 2272
03cfe0d5
LP		 2273 assert(barrier);
2274 assert(directory);
2275 assert(kmsg_socket >= 0);
88213476 2276
0de7acce 2277 if (arg_userns_mode != USER_NAMESPACE_NO) {
03cfe0d5
LP		 2278 /* Tell the parent, that it now can write the UID map. */
2279 (void) barrier_place(barrier); /* #1 */
7027ff61 2280
03cfe0d5
LP		 2281 /* Wait until the parent wrote the UID map */
2282 if (!barrier_place_and_sync(barrier)) { /* #2 */
2283 log_error("Parent died too early");
2284 return -ESRCH;
2285 }
88213476
LP		 2286 }
2287
6d66bd3b
EV		 2288 r = reset_uid_gid();
2289 if (r < 0)
2290 return log_error_errno(r, "Couldn't become new root: %m");
2291
0de7acce 2292 r = mount_all(NULL,
4f086aab 2293 arg_mount_settings | MOUNT_IN_USERNS,
0de7acce
LP		 2294 arg_uid_shift,
2295 arg_uid_range,
2296 arg_selinux_apifs_context);
2297
03cfe0d5
LP		 2298 if (r < 0)
2299 return r;
2300
4f086aab 2301 r = mount_sysfs(NULL, arg_mount_settings);
d8fc6a00
LP		 2302 if (r < 0)
2303 return r;
2304
03cfe0d5
LP		 2305 /* Wait until we are cgroup-ified, so that we
2306 * can mount the right cgroup path writable */
2307 if (!barrier_place_and_sync(barrier)) { /* #3 */
2308 log_error("Parent died too early");
2309 return -ESRCH;
88213476
LP		 2310 }
2311
5a8ff0e6 2312 if (arg_use_cgns && cg_ns_supported()) {
0996ef00
CB		 2313 r = unshare(CLONE_NEWCGROUP);
2314 if (r < 0)
2315 return log_error_errno(errno, "Failed to unshare cgroup namespace");
2316 r = mount_cgroups(
2317 "",
2318 arg_unified_cgroup_hierarchy,
2319 arg_userns_mode != USER_NAMESPACE_NO,
2320 arg_uid_shift,
2321 arg_uid_range,
5a8ff0e6 2322 arg_selinux_apifs_context,
ada54120 2323 true);
0996ef00
CB		 2324 if (r < 0)
2325 return r;
2326 } else {
2327 r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy);
2328 if (r < 0)
2329 return r;
2330 }
ec16945e 2331
03cfe0d5
LP		 2332 r = setup_boot_id(NULL);
2333 if (r < 0)
2334 return r;
ec16945e 2335
03cfe0d5
LP		 2336 r = setup_kmsg(NULL, kmsg_socket);
2337 if (r < 0)
2338 return r;
2339 kmsg_socket = safe_close(kmsg_socket);
ec16945e 2340
03cfe0d5 2341 umask(0022);
30535c16 2342
03cfe0d5
LP		 2343 if (setsid() < 0)
2344 return log_error_errno(errno, "setsid() failed: %m");
2345
2346 if (arg_private_network)
2347 loopback_setup();
2348
7a8f6325
LP		 2349 if (arg_expose_ports) {
2350 r = expose_port_send_rtnl(rtnl_socket);
2351 if (r < 0)
2352 return r;
2353 rtnl_socket = safe_close(rtnl_socket);
2354 }
03cfe0d5 2355
709f6e46
MS		 2356 r = drop_capabilities();
2357 if (r < 0)
2358 return log_error_errno(r, "drop_capabilities() failed: %m");
03cfe0d5
LP		 2359
2360 setup_hostname();
2361
050f7277 2362 if (arg_personality != PERSONALITY_INVALID) {
21022b9d
LP		 2363 r = safe_personality(arg_personality);
2364 if (r < 0)
2365 return log_error_errno(r, "personality() failed: %m");
03cfe0d5 2366 } else if (secondary) {
21022b9d
LP		 2367 r = safe_personality(PER_LINUX32);
2368 if (r < 0)
2369 return log_error_errno(r, "personality() failed: %m");
03cfe0d5
LP		 2370 }
2371
349cc4a5 2372#if HAVE_SELINUX
03cfe0d5 2373 if (arg_selinux_context)
2ed96880 2374 if (setexeccon(arg_selinux_context) < 0)
03cfe0d5
LP		 2375 return log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context);
2376#endif
2377
ee645080 2378 r = change_uid_gid(arg_user, &home);
03cfe0d5
LP		 2379 if (r < 0)
2380 return r;
2381
6aadfa4c
ILG		 2382 /* LXC sets container=lxc, so follow the scheme here */
2383 envp[n_env++] = strjoina("container=", arg_container_service_name);
2384
03cfe0d5
LP		 2385 envp[n_env] = strv_find_prefix(environ, "TERM=");
2386 if (envp[n_env])
313cefa1 2387 n_env++;
03cfe0d5
LP		 2388
2389 if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
2390 (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
2391 (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0))
2392 return log_oom();
2393
3bbaff3e 2394 assert(!sd_id128_is_null(arg_uuid));
03cfe0d5 2395
691675ba 2396 if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_to_uuid_string(arg_uuid, as_uuid)) < 0)
e01ff70a 2397 return log_oom();
03cfe0d5
LP		 2398
2399 if (fdset_size(fds) > 0) {
2400 r = fdset_cloexec(fds, false);
2401 if (r < 0)
2402 return log_error_errno(r, "Failed to unset O_CLOEXEC for file descriptors.");
2403
2404 if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", fdset_size(fds)) < 0) ||
2405 (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0))
2406 return log_oom();
2407 }
9c1e04d0
AP		 2408 if (asprintf((char **)(envp + n_env++), "NOTIFY_SOCKET=%s", NSPAWN_NOTIFY_SOCKET_PATH) < 0)
2409 return log_oom();
03cfe0d5 2410
2371271c
TG		 2411 env_use = strv_env_merge(2, envp, arg_setenv);
2412 if (!env_use)
2413 return log_oom();
03cfe0d5
LP		 2414
2415 /* Let the parent know that we are ready and
2416 * wait until the parent is ready with the
2417 * setup, too... */
2418 if (!barrier_place_and_sync(barrier)) { /* #4 */
2419 log_error("Parent died too early");
2420 return -ESRCH;
2421 }
2422
5f932eb9
LP		 2423 if (arg_chdir)
2424 if (chdir(arg_chdir) < 0)
2425 return log_error_errno(errno, "Failed to change to specified working directory %s: %m", arg_chdir);
2426
7732f92b 2427 if (arg_start_mode == START_PID2) {
75bf701f 2428 r = stub_pid1(arg_uuid);
7732f92b
LP		 2429 if (r < 0)
2430 return r;
2431 }
2432
03cfe0d5
LP		 2433 /* Now, explicitly close the log, so that we
2434 * then can close all remaining fds. Closing
2435 * the log explicitly first has the benefit
2436 * that the logging subsystem knows about it,
2437 * and is thus ready to be reopened should we
2438 * need it again. Note that the other fds
2439 * closed here are at least the locking and
2440 * barrier fds. */
2441 log_close();
2442 (void) fdset_close_others(fds);
2443
7732f92b 2444 if (arg_start_mode == START_BOOT) {
03cfe0d5
LP		 2445 char **a;
2446 size_t m;
2447
2448 /* Automatically search for the init system */
2449
75f32f04
ZJS		 2450 m = strv_length(arg_parameters);
2451 a = newa(char*, m + 2);
2452 memcpy_safe(a + 1, arg_parameters, m * sizeof(char*));
2453 a[1 + m] = NULL;
03cfe0d5 2454
ced58da7 2455 a[0] = (char*) "/usr/lib/systemd/systemd";
03cfe0d5
LP		 2456 execve(a[0], a, env_use);
2457
ced58da7 2458 a[0] = (char*) "/lib/systemd/systemd";
03cfe0d5
LP		 2459 execve(a[0], a, env_use);
2460
ced58da7 2461 a[0] = (char*) "/sbin/init";
03cfe0d5 2462 execve(a[0], a, env_use);
ced58da7
LP		 2463
2464 exec_target = "/usr/lib/systemd/systemd, /lib/systemd/systemd, /sbin/init";
1a68e1e5
PW		 2465 } else if (!strv_isempty(arg_parameters)) {
2466 exec_target = arg_parameters[0];
f757855e 2467 execvpe(arg_parameters[0], arg_parameters, env_use);
1a68e1e5 2468 } else {
5f932eb9 2469 if (!arg_chdir)
d929b0f9
ZJS		 2470 /* If we cannot change the directory, we'll end up in /, that is expected. */
2471 (void) chdir(home ?: "/root");
5f932eb9 2472
03cfe0d5
LP		 2473 execle("/bin/bash", "-bash", NULL, env_use);
2474 execle("/bin/sh", "-sh", NULL, env_use);
ced58da7
LP		 2475
2476 exec_target = "/bin/bash, /bin/sh";
03cfe0d5
LP		 2477 }
2478
35607a8d 2479 r = -errno;
03cfe0d5 2480 (void) log_open();
1a68e1e5 2481 return log_error_errno(r, "execv(%s) failed: %m", exec_target);
03cfe0d5
LP		 2482}
2483
9c1e04d0
AP		 2484static int setup_sd_notify_child(void) {
2485 static const int one = 1;
2486 int fd = -1;
2487 union sockaddr_union sa = {
2488 .sa.sa_family = AF_UNIX,
2489 };
2490 int r;
2491
2492 fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
2493 if (fd < 0)
2494 return log_error_errno(errno, "Failed to allocate notification socket: %m");
2495
2496 (void) mkdir_parents(NSPAWN_NOTIFY_SOCKET_PATH, 0755);
2497 (void) unlink(NSPAWN_NOTIFY_SOCKET_PATH);
2498
2499 strncpy(sa.un.sun_path, NSPAWN_NOTIFY_SOCKET_PATH, sizeof(sa.un.sun_path)-1);
2500 r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un));
2501 if (r < 0) {
2502 safe_close(fd);
2503 return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path);
2504 }
2505
adc7d9f0
EV		 2506 r = userns_lchown(NSPAWN_NOTIFY_SOCKET_PATH, 0, 0);
2507 if (r < 0) {
2508 safe_close(fd);
2509 return log_error_errno(r, "Failed to chown " NSPAWN_NOTIFY_SOCKET_PATH ": %m");
2510 }
2511
9c1e04d0
AP		 2512 r = setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
2513 if (r < 0) {
2514 safe_close(fd);
2515 return log_error_errno(errno, "SO_PASSCRED failed: %m");
2516 }
2517
2518 return fd;
2519}
2520
03cfe0d5
LP		 2521static int outer_child(
2522 Barrier *barrier,
2523 const char *directory,
2524 const char *console,
2d845785 2525 DissectedImage *dissected_image,
03cfe0d5
LP		 2526 bool interactive,
2527 bool secondary,
2528 int pid_socket,
e01ff70a 2529 int uuid_socket,
9c1e04d0 2530 int notify_socket,
03cfe0d5
LP		 2531 int kmsg_socket,
2532 int rtnl_socket,
825d5287 2533 int uid_shift_socket,
8199d554 2534 int unified_cgroup_hierarchy_socket,
f757855e 2535 FDSet *fds) {
03cfe0d5
LP		 2536
2537 pid_t pid;
2538 ssize_t l;
2539 int r;
9c1e04d0 2540 _cleanup_close_ int fd = -1;
03cfe0d5
LP		 2541
2542 assert(barrier);
2543 assert(directory);
2544 assert(console);
2545 assert(pid_socket >= 0);
e01ff70a 2546 assert(uuid_socket >= 0);
9c1e04d0 2547 assert(notify_socket >= 0);
03cfe0d5
LP		 2548 assert(kmsg_socket >= 0);
2549
2550 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
2551 return log_error_errno(errno, "PR_SET_PDEATHSIG failed: %m");
2552
2553 if (interactive) {
2554 close_nointr(STDIN_FILENO);
2555 close_nointr(STDOUT_FILENO);
2556 close_nointr(STDERR_FILENO);
2557
2558 r = open_terminal(console, O_RDWR);
2559 if (r != STDIN_FILENO) {
2560 if (r >= 0) {
2561 safe_close(r);
2562 r = -EINVAL;
2563 }
2564
2565 return log_error_errno(r, "Failed to open console: %m");
2566 }
2567
2568 if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
2569 dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
2570 return log_error_errno(errno, "Failed to duplicate console: %m");
2571 }
2572
2573 r = reset_audit_loginuid();
2574 if (r < 0)
2575 return r;
2576
2577 /* Mark everything as slave, so that we still
2578 * receive mounts from the real root, but don't
2579 * propagate mounts to the real root. */
60e76d48
ZJS		 2580 r = mount_verbose(LOG_ERR, NULL, "/", NULL, MS_SLAVE|MS_REC, NULL);
2581 if (r < 0)
2582 return r;
03cfe0d5 2583
2d845785 2584 if (dissected_image) {
2d3a5a73
LP		 2585 /* If we are operating on a disk image, then mount its root directory now, but leave out the rest. We
2586 * can read the UID shift from it if we need to. Further down we'll mount the rest, but then with the
2587 * uid shift known. That way we can mount VFAT file systems shifted to the right place right away. This
2588 * makes sure ESP partitions and userns are compatible. */
2589
2590 r = dissected_image_mount(dissected_image, directory, arg_uid_shift,
2591 DISSECT_IMAGE_MOUNT_ROOT_ONLY|DISSECT_IMAGE_DISCARD_ON_LOOP|(arg_read_only ? DISSECT_IMAGE_READ_ONLY : 0));
2d845785
LP		 2592 if (r < 0)
2593 return r;
2594 }
03cfe0d5 2595
391567f4
LP		 2596 r = determine_uid_shift(directory);
2597 if (r < 0)
2598 return r;
2599
0de7acce 2600 if (arg_userns_mode != USER_NAMESPACE_NO) {
0e7ac751 2601 /* Let the parent know which UID shift we read from the image */
825d5287
RM		 2602 l = send(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), MSG_NOSIGNAL);
2603 if (l < 0)
2604 return log_error_errno(errno, "Failed to send UID shift: %m");
2605 if (l != sizeof(arg_uid_shift)) {
2606 log_error("Short write while sending UID shift.");
2607 return -EIO;
2608 }
0e7ac751 2609
0de7acce 2610 if (arg_userns_mode == USER_NAMESPACE_PICK) {
0e7ac751
LP		 2611 /* When we are supposed to pick the UID shift, the parent will check now whether the UID shift
2612 * we just read from the image is available. If yes, it will send the UID shift back to us, if
2613 * not it will pick a different one, and send it back to us. */
2614
2615 l = recv(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), 0);
2616 if (l < 0)
2617 return log_error_errno(errno, "Failed to recv UID shift: %m");
2618 if (l != sizeof(arg_uid_shift)) {
595bfe7d 2619 log_error("Short read while receiving UID shift.");
0e7ac751
LP		 2620 return -EIO;
2621 }
2622 }
2623
2624 log_info("Selected user namespace base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range);
825d5287
RM		 2625 }
2626
2d3a5a73
LP		 2627 if (dissected_image) {
2628 /* Now we know the uid shift, let's now mount everything else that might be in the image. */
2629 r = dissected_image_mount(dissected_image, directory, arg_uid_shift,
2630 DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY|DISSECT_IMAGE_DISCARD_ON_LOOP|(arg_read_only ? DISSECT_IMAGE_READ_ONLY : 0));
2631 if (r < 0)
2632 return r;
2633 }
2634
8199d554
LP		 2635 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
2636 /* OK, we don't know yet which cgroup mode to use yet. Let's figure it out, and tell the parent. */
2637
2638 r = detect_unified_cgroup_hierarchy_from_image(directory);
2639 if (r < 0)
2640 return r;
2641
2642 l = send(unified_cgroup_hierarchy_socket, &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), MSG_NOSIGNAL);
2643 if (l < 0)
2644 return log_error_errno(errno, "Failed to send cgroup mode: %m");
2645 if (l != sizeof(arg_unified_cgroup_hierarchy)) {
2646 log_error("Short write while sending cgroup mode: %m");
2647 return -EIO;
2648 }
2649
2650 unified_cgroup_hierarchy_socket = safe_close(unified_cgroup_hierarchy_socket);
2651 }
2652
03cfe0d5 2653 /* Turn directory into bind mount */
60e76d48
ZJS		 2654 r = mount_verbose(LOG_ERR, directory, directory, NULL, MS_BIND|MS_REC, NULL);
2655 if (r < 0)
2656 return r;
03cfe0d5 2657
b53ede69
PW		 2658 r = setup_pivot_root(
2659 directory,
2660 arg_pivot_root_new,
2661 arg_pivot_root_old);
2662 if (r < 0)
2663 return r;
2664
0de7acce
LP		 2665 r = setup_volatile(
2666 directory,
2667 arg_volatile_mode,
2668 arg_userns_mode != USER_NAMESPACE_NO,
2669 arg_uid_shift,
2670 arg_uid_range,
2671 arg_selinux_context);
03cfe0d5
LP		 2672 if (r < 0)
2673 return r;
2674
0de7acce
LP		 2675 r = setup_volatile_state(
2676 directory,
2677 arg_volatile_mode,
2678 arg_userns_mode != USER_NAMESPACE_NO,
2679 arg_uid_shift,
2680 arg_uid_range,
2681 arg_selinux_context);
03cfe0d5
LP		 2682 if (r < 0)
2683 return r;
2684
4ad14eff
LP		 2685 /* Mark everything as shared so our mounts get propagated down. This is
2686 * required to make new bind mounts available in systemd services
2687 * inside the containter that create a new mount namespace.
2688 * See https://github.com/systemd/systemd/issues/3860
2689 * Further submounts (such as /dev) done after this will inherit the
13e785f7 2690 * shared propagation mode. */
4ad14eff
LP		 2691 r = mount_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL);
2692 if (r < 0)
2693 return r;
2694
2695 r = recursive_chown(directory, arg_uid_shift, arg_uid_range);
2696 if (r < 0)
2697 return r;
2698
03cfe0d5
LP		 2699 r = base_filesystem_create(directory, arg_uid_shift, (gid_t) arg_uid_shift);
2700 if (r < 0)
2701 return r;
2702
03cfe0d5 2703 if (arg_read_only) {
6b7c9f8b 2704 r = bind_remount_recursive(directory, true, NULL);
03cfe0d5
LP		 2705 if (r < 0)
2706 return log_error_errno(r, "Failed to make tree read-only: %m");
2707 }
2708
0de7acce 2709 r = mount_all(directory,
4f086aab 2710 arg_mount_settings,
0de7acce
LP		 2711 arg_uid_shift,
2712 arg_uid_range,
2713 arg_selinux_apifs_context);
03cfe0d5
LP		 2714 if (r < 0)
2715 return r;
2716
07fa00f9
LP		 2717 r = copy_devnodes(directory);
2718 if (r < 0)
03cfe0d5
LP		 2719 return r;
2720
2721 dev_setup(directory, arg_uid_shift, arg_uid_shift);
2722
07fa00f9
LP		 2723 r = setup_pts(directory);
2724 if (r < 0)
03cfe0d5
LP		 2725 return r;
2726
2727 r = setup_propagate(directory);
2728 if (r < 0)
2729 return r;
2730
2731 r = setup_dev_console(directory, console);
2732 if (r < 0)
2733 return r;
2734
8e5430c4
LP		 2735 r = setup_keyring();
2736 if (r < 0)
2737 return r;
2738
960e4569 2739 r = setup_seccomp(arg_caps_retain, arg_syscall_whitelist, arg_syscall_blacklist);
03cfe0d5
LP		 2740 if (r < 0)
2741 return r;
2742
2743 r = setup_timezone(directory);
2744 if (r < 0)
2745 return r;
2746
2747 r = setup_resolv_conf(directory);
2748 if (r < 0)
2749 return r;
2750
e01ff70a
MS		 2751 r = setup_machine_id(directory);
2752 if (r < 0)
2753 return r;
2754
03cfe0d5
LP		 2755 r = setup_journal(directory);
2756 if (r < 0)
2757 return r;
2758
0de7acce
LP		 2759 r = mount_custom(
2760 directory,
2761 arg_custom_mounts,
2762 arg_n_custom_mounts,
2763 arg_userns_mode != USER_NAMESPACE_NO,
2764 arg_uid_shift,
2765 arg_uid_range,
2766 arg_selinux_apifs_context);
03cfe0d5
LP		 2767 if (r < 0)
2768 return r;
2769
5a8ff0e6 2770 if (!arg_use_cgns || !cg_ns_supported()) {
0996ef00
CB		 2771 r = mount_cgroups(
2772 directory,
2773 arg_unified_cgroup_hierarchy,
2774 arg_userns_mode != USER_NAMESPACE_NO,
2775 arg_uid_shift,
2776 arg_uid_range,
5a8ff0e6 2777 arg_selinux_apifs_context,
ada54120 2778 false);
0996ef00
CB		 2779 if (r < 0)
2780 return r;
2781 }
03cfe0d5
LP		 2782
2783 r = mount_move_root(directory);
2784 if (r < 0)
2785 return log_error_errno(r, "Failed to move root directory: %m");
2786
9c1e04d0
AP		 2787 fd = setup_sd_notify_child();
2788 if (fd < 0)
2789 return fd;
2790
03cfe0d5 2791 pid = raw_clone(SIGCHLD|CLONE_NEWNS|
0c582db0 2792 arg_clone_ns_flags |
03cfe0d5 2793 (arg_private_network ? CLONE_NEWNET : 0) |
8869a0b4 2794 (arg_userns_mode != USER_NAMESPACE_NO ? CLONE_NEWUSER : 0));
03cfe0d5
LP		 2795 if (pid < 0)
2796 return log_error_errno(errno, "Failed to fork inner child: %m");
03cfe0d5
LP		 2797 if (pid == 0) {
2798 pid_socket = safe_close(pid_socket);
e01ff70a 2799 uuid_socket = safe_close(uuid_socket);
9c1e04d0 2800 notify_socket = safe_close(notify_socket);
825d5287 2801 uid_shift_socket = safe_close(uid_shift_socket);
03cfe0d5
LP		 2802
2803 /* The inner child has all namespaces that are
2804 * requested, so that we all are owned by the user if
2805 * user namespaces are turned on. */
2806
f757855e 2807 r = inner_child(barrier, directory, secondary, kmsg_socket, rtnl_socket, fds);
03cfe0d5
LP		 2808 if (r < 0)
2809 _exit(EXIT_FAILURE);
2810
2811 _exit(EXIT_SUCCESS);
2812 }
2813
2814 l = send(pid_socket, &pid, sizeof(pid), MSG_NOSIGNAL);
2815 if (l < 0)
2816 return log_error_errno(errno, "Failed to send PID: %m");
2817 if (l != sizeof(pid)) {
2818 log_error("Short write while sending PID.");
2819 return -EIO;
2820 }
2821
e01ff70a
MS		 2822 l = send(uuid_socket, &arg_uuid, sizeof(arg_uuid), MSG_NOSIGNAL);
2823 if (l < 0)
2824 return log_error_errno(errno, "Failed to send machine ID: %m");
2825 if (l != sizeof(arg_uuid)) {
2826 log_error("Short write while sending machine ID.");
2827 return -EIO;
2828 }
2829
9c1e04d0
AP		 2830 l = send_one_fd(notify_socket, fd, 0);
2831 if (l < 0)
2832 return log_error_errno(errno, "Failed to send notify fd: %m");
2833
03cfe0d5 2834 pid_socket = safe_close(pid_socket);
e01ff70a 2835 uuid_socket = safe_close(uuid_socket);
9c1e04d0 2836 notify_socket = safe_close(notify_socket);
327e26d6
KN		 2837 kmsg_socket = safe_close(kmsg_socket);
2838 rtnl_socket = safe_close(rtnl_socket);
03cfe0d5
LP		 2839
2840 return 0;
2841}
2842
0e7ac751 2843static int uid_shift_pick(uid_t *shift, LockFile *ret_lock_file) {
d381c8a6 2844 bool tried_hashed = false;
0e7ac751
LP		 2845 unsigned n_tries = 100;
2846 uid_t candidate;
2847 int r;
2848
2849 assert(shift);
2850 assert(ret_lock_file);
0de7acce 2851 assert(arg_userns_mode == USER_NAMESPACE_PICK);
0e7ac751
LP		 2852 assert(arg_uid_range == 0x10000U);
2853
2854 candidate = *shift;
2855
2856 (void) mkdir("/run/systemd/nspawn-uid", 0755);
2857
2858 for (;;) {
2859 char lock_path[strlen("/run/systemd/nspawn-uid/") + DECIMAL_STR_MAX(uid_t) + 1];
2860 _cleanup_release_lock_file_ LockFile lf = LOCK_FILE_INIT;
2861
2862 if (--n_tries <= 0)
2863 return -EBUSY;
2864
87d5e4f2 2865 if (candidate < CONTAINER_UID_BASE_MIN || candidate > CONTAINER_UID_BASE_MAX)
0e7ac751
LP		 2866 goto next;
2867 if ((candidate & UINT32_C(0xFFFF)) != 0)
2868 goto next;
2869
2870 xsprintf(lock_path, "/run/systemd/nspawn-uid/" UID_FMT, candidate);
2871 r = make_lock_file(lock_path, LOCK_EX|LOCK_NB, &lf);
2872 if (r == -EBUSY) /* Range already taken by another nspawn instance */
2873 goto next;
2874 if (r < 0)
2875 return r;
2876
2877 /* Make some superficial checks whether the range is currently known in the user database */
2878 if (getpwuid(candidate))
2879 goto next;
2880 if (getpwuid(candidate + UINT32_C(0xFFFE)))
2881 goto next;
2882 if (getgrgid(candidate))
2883 goto next;
2884 if (getgrgid(candidate + UINT32_C(0xFFFE)))
2885 goto next;
2886
2887 *ret_lock_file = lf;
2888 lf = (struct LockFile) LOCK_FILE_INIT;
2889 *shift = candidate;
2890 return 0;
2891
2892 next:
d381c8a6
LP		 2893 if (arg_machine && !tried_hashed) {
2894 /* Try to hash the base from the container name */
2895
2896 static const uint8_t hash_key[] = {
2897 0xe1, 0x56, 0xe0, 0xf0, 0x4a, 0xf0, 0x41, 0xaf,
2898 0x96, 0x41, 0xcf, 0x41, 0x33, 0x94, 0xff, 0x72
2899 };
2900
2901 candidate = (uid_t) siphash24(arg_machine, strlen(arg_machine), hash_key);
2902
2903 tried_hashed = true;
2904 } else
2905 random_bytes(&candidate, sizeof(candidate));
2906
87d5e4f2 2907 candidate = (candidate % (CONTAINER_UID_BASE_MAX - CONTAINER_UID_BASE_MIN)) + CONTAINER_UID_BASE_MIN;
0e7ac751
LP		 2908 candidate &= (uid_t) UINT32_C(0xFFFF0000);
2909 }
2910}
2911
03cfe0d5
LP		 2912static int setup_uid_map(pid_t pid) {
2913 char uid_map[strlen("/proc//uid_map") + DECIMAL_STR_MAX(uid_t) + 1], line[DECIMAL_STR_MAX(uid_t)*3+3+1];
2914 int r;
2915
2916 assert(pid > 1);
2917
2918 xsprintf(uid_map, "/proc/" PID_FMT "/uid_map", pid);
2919 xsprintf(line, UID_FMT " " UID_FMT " " UID_FMT "\n", 0, arg_uid_shift, arg_uid_range);
ad118bda 2920 r = write_string_file(uid_map, line, 0);
03cfe0d5
LP		 2921 if (r < 0)
2922 return log_error_errno(r, "Failed to write UID map: %m");
2923
2924 /* We always assign the same UID and GID ranges */
2925 xsprintf(uid_map, "/proc/" PID_FMT "/gid_map", pid);
ad118bda 2926 r = write_string_file(uid_map, line, 0);
03cfe0d5
LP		 2927 if (r < 0)
2928 return log_error_errno(r, "Failed to write GID map: %m");
2929
2930 return 0;
2931}
2932
9c1e04d0 2933static int nspawn_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) {
9c1e04d0
AP		 2934 char buf[NOTIFY_BUFFER_MAX+1];
2935 char *p = NULL;
2936 struct iovec iovec = {
2937 .iov_base = buf,
2938 .iov_len = sizeof(buf)-1,
2939 };
2940 union {
2941 struct cmsghdr cmsghdr;
2942 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
2943 CMSG_SPACE(sizeof(int) * NOTIFY_FD_MAX)];
2944 } control = {};
2945 struct msghdr msghdr = {
2946 .msg_iov = &iovec,
2947 .msg_iovlen = 1,
2948 .msg_control = &control,
2949 .msg_controllen = sizeof(control),
2950 };
2951 struct cmsghdr *cmsg;
2952 struct ucred *ucred = NULL;
2953 ssize_t n;
2954 pid_t inner_child_pid;
2955 _cleanup_strv_free_ char **tags = NULL;
2956
2957 assert(userdata);
2958
2959 inner_child_pid = PTR_TO_PID(userdata);
2960
2961 if (revents != EPOLLIN) {
2962 log_warning("Got unexpected poll event for notify fd.");
2963 return 0;
2964 }
2965
2966 n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
2967 if (n < 0) {
3742095b 2968 if (IN_SET(errno, EAGAIN, EINTR))
9c1e04d0
AP		 2969 return 0;
2970
2971 return log_warning_errno(errno, "Couldn't read notification socket: %m");
2972 }
2973 cmsg_close_all(&msghdr);
2974
2975 CMSG_FOREACH(cmsg, &msghdr) {
2976 if (cmsg->cmsg_level == SOL_SOCKET &&
2977 cmsg->cmsg_type == SCM_CREDENTIALS &&
2978 cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred))) {
2979
2980 ucred = (struct ucred*) CMSG_DATA(cmsg);
2981 }
2982 }
2983
2984 if (!ucred || ucred->pid != inner_child_pid) {
8cb57430 2985 log_debug("Received notify message without valid credentials. Ignoring.");
9c1e04d0
AP		 2986 return 0;
2987 }
2988
2989 if ((size_t) n >= sizeof(buf)) {
2990 log_warning("Received notify message exceeded maximum size. Ignoring.");
2991 return 0;
2992 }
2993
2994 buf[n] = 0;
2995 tags = strv_split(buf, "\n\r");
2996 if (!tags)
2997 return log_oom();
2998
2999 if (strv_find(tags, "READY=1"))
3000 sd_notifyf(false, "READY=1\n");
3001
3002 p = strv_find_startswith(tags, "STATUS=");
3003 if (p)
3004 sd_notifyf(false, "STATUS=Container running: %s", p);
3005
3006 return 0;
3007}
3008
5773024d 3009static int setup_sd_notify_parent(sd_event *event, int fd, pid_t *inner_child_pid, sd_event_source **notify_event_source) {
9c1e04d0 3010 int r;
9c1e04d0 3011
5773024d 3012 r = sd_event_add_io(event, notify_event_source, fd, EPOLLIN, nspawn_dispatch_notify_fd, inner_child_pid);
9c1e04d0
AP		 3013 if (r < 0)
3014 return log_error_errno(r, "Failed to allocate notify event source: %m");
3015
5773024d 3016 (void) sd_event_source_set_description(*notify_event_source, "nspawn-notify");
9c1e04d0
AP		 3017
3018 return 0;
3019}
3020
f757855e
LP		 3021static int load_settings(void) {
3022 _cleanup_(settings_freep) Settings *settings = NULL;
3023 _cleanup_fclose_ FILE *f = NULL;
3024 _cleanup_free_ char *p = NULL;
3025 const char *fn, *i;
3026 int r;
3027
3028 /* If all settings are masked, there's no point in looking for
3029 * the settings file */
3030 if ((arg_settings_mask & _SETTINGS_MASK_ALL) == _SETTINGS_MASK_ALL)
3031 return 0;
3032
3033 fn = strjoina(arg_machine, ".nspawn");
3034
3035 /* We first look in the admin's directories in /etc and /run */
3036 FOREACH_STRING(i, "/etc/systemd/nspawn", "/run/systemd/nspawn") {
3037 _cleanup_free_ char *j = NULL;
3038
605405c6 3039 j = strjoin(i, "/", fn);
f757855e
LP		 3040 if (!j)
3041 return log_oom();
3042
3043 f = fopen(j, "re");
3044 if (f) {
3045 p = j;
3046 j = NULL;
3047
b938cb90 3048 /* By default, we trust configuration from /etc and /run */
f757855e
LP		 3049 if (arg_settings_trusted < 0)
3050 arg_settings_trusted = true;
3051
3052 break;
3053 }
3054
3055 if (errno != ENOENT)
3056 return log_error_errno(errno, "Failed to open %s: %m", j);
3057 }
3058
3059 if (!f) {
3060 /* After that, let's look for a file next to the
3061 * actual image we shall boot. */
3062
3063 if (arg_image) {
3064 p = file_in_same_dir(arg_image, fn);
3065 if (!p)
3066 return log_oom();
3067 } else if (arg_directory) {
3068 p = file_in_same_dir(arg_directory, fn);
3069 if (!p)
3070 return log_oom();
3071 }
3072
3073 if (p) {
3074 f = fopen(p, "re");
3075 if (!f && errno != ENOENT)
3076 return log_error_errno(errno, "Failed to open %s: %m", p);
3077
b938cb90 3078 /* By default, we do not trust configuration from /var/lib/machines */
f757855e
LP		 3079 if (arg_settings_trusted < 0)
3080 arg_settings_trusted = false;
3081 }
3082 }
3083
3084 if (!f)
3085 return 0;
3086
3087 log_debug("Settings are trusted: %s", yes_no(arg_settings_trusted));
3088
3089 r = settings_load(f, p, &settings);
3090 if (r < 0)
3091 return r;
3092
3093 /* Copy over bits from the settings, unless they have been
3094 * explicitly masked by command line switches. */
3095
7732f92b
LP		 3096 if ((arg_settings_mask & SETTING_START_MODE) == 0 &&
3097 settings->start_mode >= 0) {
3098 arg_start_mode = settings->start_mode;
f757855e
LP		 3099
3100 strv_free(arg_parameters);
3101 arg_parameters = settings->parameters;
3102 settings->parameters = NULL;
3103 }
3104
b53ede69
PW		 3105 if ((arg_settings_mask & SETTING_PIVOT_ROOT) == 0 &&
3106 settings->pivot_root_new) {
3107 free_and_replace(arg_pivot_root_new, settings->pivot_root_new);
3108 free_and_replace(arg_pivot_root_old, settings->pivot_root_old);
3109 }
3110
5f932eb9
LP		 3111 if ((arg_settings_mask & SETTING_WORKING_DIRECTORY) == 0 &&
3112 settings->working_directory) {
3113 free(arg_chdir);
3114 arg_chdir = settings->working_directory;
3115 settings->working_directory = NULL;
3116 }
3117
f757855e
LP		 3118 if ((arg_settings_mask & SETTING_ENVIRONMENT) == 0 &&
3119 settings->environment) {
3120 strv_free(arg_setenv);
3121 arg_setenv = settings->environment;
3122 settings->environment = NULL;
3123 }
3124
3125 if ((arg_settings_mask & SETTING_USER) == 0 &&
3126 settings->user) {
3127 free(arg_user);
3128 arg_user = settings->user;
3129 settings->user = NULL;
3130 }
3131
3132 if ((arg_settings_mask & SETTING_CAPABILITY) == 0) {
0e265674 3133 uint64_t plus;
f757855e 3134
0e265674
LP		 3135 plus = settings->capability;
3136 if (settings_private_network(settings))
3137 plus |= (1ULL << CAP_NET_ADMIN);
3138
3139 if (!arg_settings_trusted && plus != 0) {
3140 if (settings->capability != 0)
3141 log_warning("Ignoring Capability= setting, file %s is not trusted.", p);
3142 } else
520e0d54 3143 arg_caps_retain |= plus;
f757855e 3144
520e0d54 3145 arg_caps_retain &= ~settings->drop_capability;
f757855e
LP		 3146 }
3147
3148 if ((arg_settings_mask & SETTING_KILL_SIGNAL) == 0 &&
3149 settings->kill_signal > 0)
3150 arg_kill_signal = settings->kill_signal;
3151
3152 if ((arg_settings_mask & SETTING_PERSONALITY) == 0 &&
3153 settings->personality != PERSONALITY_INVALID)
3154 arg_personality = settings->personality;
3155
3156 if ((arg_settings_mask & SETTING_MACHINE_ID) == 0 &&
3157 !sd_id128_is_null(settings->machine_id)) {
3158
3159 if (!arg_settings_trusted)
3160 log_warning("Ignoring MachineID= setting, file %s is not trusted.", p);
3161 else
3162 arg_uuid = settings->machine_id;
3163 }
3164
3165 if ((arg_settings_mask & SETTING_READ_ONLY) == 0 &&
3166 settings->read_only >= 0)
3167 arg_read_only = settings->read_only;
3168
3169 if ((arg_settings_mask & SETTING_VOLATILE_MODE) == 0 &&
3170 settings->volatile_mode != _VOLATILE_MODE_INVALID)
3171 arg_volatile_mode = settings->volatile_mode;
3172
3173 if ((arg_settings_mask & SETTING_CUSTOM_MOUNTS) == 0 &&
3174 settings->n_custom_mounts > 0) {
3175
3176 if (!arg_settings_trusted)
3177 log_warning("Ignoring TemporaryFileSystem=, Bind= and BindReadOnly= settings, file %s is not trusted.", p);
3178 else {
3179 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
3180 arg_custom_mounts = settings->custom_mounts;
3181 arg_n_custom_mounts = settings->n_custom_mounts;
3182
3183 settings->custom_mounts = NULL;
3184 settings->n_custom_mounts = 0;
3185 }
3186 }
3187
3188 if ((arg_settings_mask & SETTING_NETWORK) == 0 &&
3189 (settings->private_network >= 0 ||
3190 settings->network_veth >= 0 ||
3191 settings->network_bridge ||
22b28dfd 3192 settings->network_zone ||
f757855e
LP		 3193 settings->network_interfaces ||
3194 settings->network_macvlan ||
f6d6bad1
LP		 3195 settings->network_ipvlan ||
3196 settings->network_veth_extra)) {
f757855e
LP		 3197
3198 if (!arg_settings_trusted)
3199 log_warning("Ignoring network settings, file %s is not trusted.", p);
3200 else {
f6d6bad1 3201 arg_network_veth = settings_network_veth(settings);
0e265674
LP		 3202 arg_private_network = settings_private_network(settings);
3203
f757855e
LP		 3204 strv_free(arg_network_interfaces);
3205 arg_network_interfaces = settings->network_interfaces;
3206 settings->network_interfaces = NULL;
3207
3208 strv_free(arg_network_macvlan);
3209 arg_network_macvlan = settings->network_macvlan;
3210 settings->network_macvlan = NULL;
3211
3212 strv_free(arg_network_ipvlan);
3213 arg_network_ipvlan = settings->network_ipvlan;
3214 settings->network_ipvlan = NULL;
3215
f6d6bad1
LP		 3216 strv_free(arg_network_veth_extra);
3217 arg_network_veth_extra = settings->network_veth_extra;
3218 settings->network_veth_extra = NULL;
3219
f757855e
LP		 3220 free(arg_network_bridge);
3221 arg_network_bridge = settings->network_bridge;
3222 settings->network_bridge = NULL;
22b28dfd
LP		 3223
3224 free(arg_network_zone);
3225 arg_network_zone = settings->network_zone;
3226 settings->network_zone = NULL;
f757855e
LP		 3227 }
3228 }
3229
3230 if ((arg_settings_mask & SETTING_EXPOSE_PORTS) == 0 &&
3231 settings->expose_ports) {
3232
3233 if (!arg_settings_trusted)
3234 log_warning("Ignoring Port= setting, file %s is not trusted.", p);
3235 else {
3236 expose_port_free_all(arg_expose_ports);
3237 arg_expose_ports = settings->expose_ports;
3238 settings->expose_ports = NULL;
3239 }
3240 }
3241
0de7acce
LP		 3242 if ((arg_settings_mask & SETTING_USERNS) == 0 &&
3243 settings->userns_mode != _USER_NAMESPACE_MODE_INVALID) {
3244
3245 if (!arg_settings_trusted)
3246 log_warning("Ignoring PrivateUsers= and PrivateUsersChown= settings, file %s is not trusted.", p);
3247 else {
3248 arg_userns_mode = settings->userns_mode;
3249 arg_uid_shift = settings->uid_shift;
3250 arg_uid_range = settings->uid_range;
3251 arg_userns_chown = settings->userns_chown;
3252 }
3253 }
3254
9c1e04d0
AP		 3255 if ((arg_settings_mask & SETTING_NOTIFY_READY) == 0)
3256 arg_notify_ready = settings->notify_ready;
3257
960e4569
LP		 3258 if ((arg_settings_mask & SETTING_SYSCALL_FILTER) == 0) {
3259
3260 if (!arg_settings_trusted && !strv_isempty(arg_syscall_whitelist))
3261 log_warning("Ignoring SystemCallFilter= settings, file %s is not trusted.", p);
3262 else {
3263 strv_free(arg_syscall_whitelist);
3264 strv_free(arg_syscall_blacklist);
3265
3266 arg_syscall_whitelist = settings->syscall_whitelist;
3267 arg_syscall_blacklist = settings->syscall_blacklist;
3268
3269 settings->syscall_whitelist = settings->syscall_blacklist = NULL;
3270 }
3271 }
3272
f757855e
LP		 3273 return 0;
3274}
3275
b0067625
ZJS		 3276static int run(int master,
3277 const char* console,
2d845785 3278 DissectedImage *dissected_image,
b0067625
ZJS		 3279 bool interactive,
3280 bool secondary,
3281 FDSet *fds,
3282 char veth_name[IFNAMSIZ], bool *veth_created,
3283 union in_addr_union *exposed,
3284 pid_t *pid, int *ret) {
3285
3286 static const struct sigaction sa = {
3287 .sa_handler = nop_signal_handler,
e28c7cd0 3288 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
b0067625
ZJS		 3289 };
3290
3291 _cleanup_release_lock_file_ LockFile uid_shift_lock = LOCK_FILE_INIT;
3292 _cleanup_close_ int etc_passwd_lock = -1;
3293 _cleanup_close_pair_ int
3294 kmsg_socket_pair[2] = { -1, -1 },
3295 rtnl_socket_pair[2] = { -1, -1 },
3296 pid_socket_pair[2] = { -1, -1 },
3297 uuid_socket_pair[2] = { -1, -1 },
3298 notify_socket_pair[2] = { -1, -1 },
8199d554
LP		 3299 uid_shift_socket_pair[2] = { -1, -1 },
3300 unified_cgroup_hierarchy_socket_pair[2] = { -1, -1};
3301
b0067625
ZJS		 3302 _cleanup_close_ int notify_socket= -1;
3303 _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
5773024d 3304 _cleanup_(sd_event_source_unrefp) sd_event_source *notify_event_source = NULL;
b0067625
ZJS		 3305 _cleanup_(sd_event_unrefp) sd_event *event = NULL;
3306 _cleanup_(pty_forward_freep) PTYForward *forward = NULL;
3307 _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL;
abdb9b08 3308 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
b0067625
ZJS		 3309 ContainerStatus container_status = 0;
3310 char last_char = 0;
3311 int ifi = 0, r;
3312 ssize_t l;
3313 sigset_t mask_chld;
3314
3315 assert_se(sigemptyset(&mask_chld) == 0);
3316 assert_se(sigaddset(&mask_chld, SIGCHLD) == 0);
3317
3318 if (arg_userns_mode == USER_NAMESPACE_PICK) {
3319 /* When we shall pick the UID/GID range, let's first lock /etc/passwd, so that we can safely
3320 * check with getpwuid() if the specific user already exists. Note that /etc might be
3321 * read-only, in which case this will fail with EROFS. But that's really OK, as in that case we
3322 * can be reasonably sure that no users are going to be added. Note that getpwuid() checks are
3323 * really just an extra safety net. We kinda assume that the UID range we allocate from is
3324 * really ours. */
3325
3326 etc_passwd_lock = take_etc_passwd_lock(NULL);
3327 if (etc_passwd_lock < 0 && etc_passwd_lock != -EROFS)
3328 return log_error_errno(etc_passwd_lock, "Failed to take /etc/passwd lock: %m");
3329 }
3330
3331 r = barrier_create(&barrier);
3332 if (r < 0)
3333 return log_error_errno(r, "Cannot initialize IPC barrier: %m");
3334
3335 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0)
3336 return log_error_errno(errno, "Failed to create kmsg socket pair: %m");
3337
3338 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0)
3339 return log_error_errno(errno, "Failed to create rtnl socket pair: %m");
3340
3341 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, pid_socket_pair) < 0)
3342 return log_error_errno(errno, "Failed to create pid socket pair: %m");
3343
3344 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uuid_socket_pair) < 0)
3345 return log_error_errno(errno, "Failed to create id socket pair: %m");
3346
3347 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, notify_socket_pair) < 0)
3348 return log_error_errno(errno, "Failed to create notify socket pair: %m");
3349
3350 if (arg_userns_mode != USER_NAMESPACE_NO)
3351 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0)
3352 return log_error_errno(errno, "Failed to create uid shift socket pair: %m");
3353
8199d554
LP		 3354 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN)
3355 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, unified_cgroup_hierarchy_socket_pair) < 0)
3356 return log_error_errno(errno, "Failed to create unified cgroup socket pair: %m");
3357
b0067625
ZJS		 3358 /* Child can be killed before execv(), so handle SIGCHLD in order to interrupt
3359 * parent's blocking calls and give it a chance to call wait() and terminate. */
3360 r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
3361 if (r < 0)
3362 return log_error_errno(errno, "Failed to change the signal mask: %m");
3363
3364 r = sigaction(SIGCHLD, &sa, NULL);
3365 if (r < 0)
3366 return log_error_errno(errno, "Failed to install SIGCHLD handler: %m");
3367
3368 *pid = raw_clone(SIGCHLD|CLONE_NEWNS);
3369 if (*pid < 0)
3370 return log_error_errno(errno, "clone() failed%s: %m",
3371 errno == EINVAL ?
3372 ", do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in)" : "");
3373
3374 if (*pid == 0) {
3375 /* The outer child only has a file system namespace. */
3376 barrier_set_role(&barrier, BARRIER_CHILD);
3377
3378 master = safe_close(master);
3379
3380 kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
3381 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
3382 pid_socket_pair[0] = safe_close(pid_socket_pair[0]);
3383 uuid_socket_pair[0] = safe_close(uuid_socket_pair[0]);
3384 notify_socket_pair[0] = safe_close(notify_socket_pair[0]);
3385 uid_shift_socket_pair[0] = safe_close(uid_shift_socket_pair[0]);
8199d554 3386 unified_cgroup_hierarchy_socket_pair[0] = safe_close(unified_cgroup_hierarchy_socket_pair[0]);
b0067625
ZJS		 3387
3388 (void) reset_all_signal_handlers();
3389 (void) reset_signal_mask();
3390
3391 r = outer_child(&barrier,
3392 arg_directory,
3393 console,
2d845785 3394 dissected_image,
b0067625
ZJS		 3395 interactive,
3396 secondary,
3397 pid_socket_pair[1],
3398 uuid_socket_pair[1],
3399 notify_socket_pair[1],
3400 kmsg_socket_pair[1],
3401 rtnl_socket_pair[1],
3402 uid_shift_socket_pair[1],
8199d554 3403 unified_cgroup_hierarchy_socket_pair[1],
b0067625
ZJS		 3404 fds);
3405 if (r < 0)
3406 _exit(EXIT_FAILURE);
3407
3408 _exit(EXIT_SUCCESS);
3409 }
3410
3411 barrier_set_role(&barrier, BARRIER_PARENT);
3412
3413 fds = fdset_free(fds);
3414
3415 kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
3416 rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
3417 pid_socket_pair[1] = safe_close(pid_socket_pair[1]);
3418 uuid_socket_pair[1] = safe_close(uuid_socket_pair[1]);
3419 notify_socket_pair[1] = safe_close(notify_socket_pair[1]);
3420 uid_shift_socket_pair[1] = safe_close(uid_shift_socket_pair[1]);
8199d554 3421 unified_cgroup_hierarchy_socket_pair[1] = safe_close(unified_cgroup_hierarchy_socket_pair[1]);
b0067625
ZJS		 3422
3423 if (arg_userns_mode != USER_NAMESPACE_NO) {
3424 /* The child just let us know the UID shift it might have read from the image. */
3425 l = recv(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, 0);
3426 if (l < 0)
3427 return log_error_errno(errno, "Failed to read UID shift: %m");
b0067625
ZJS		 3428 if (l != sizeof arg_uid_shift) {
3429 log_error("Short read while reading UID shift.");
3430 return -EIO;
3431 }
3432
3433 if (arg_userns_mode == USER_NAMESPACE_PICK) {
3434 /* If we are supposed to pick the UID shift, let's try to use the shift read from the
3435 * image, but if that's already in use, pick a new one, and report back to the child,
3436 * which one we now picked. */
3437
3438 r = uid_shift_pick(&arg_uid_shift, &uid_shift_lock);
3439 if (r < 0)
3440 return log_error_errno(r, "Failed to pick suitable UID/GID range: %m");
3441
3442 l = send(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, MSG_NOSIGNAL);
3443 if (l < 0)
3444 return log_error_errno(errno, "Failed to send UID shift: %m");
3445 if (l != sizeof arg_uid_shift) {
3446 log_error("Short write while writing UID shift.");
3447 return -EIO;
3448 }
3449 }
3450 }
3451
8199d554
LP		 3452 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
3453 /* The child let us know the support cgroup mode it might have read from the image. */
3454 l = recv(unified_cgroup_hierarchy_socket_pair[0], &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), 0);
3455 if (l < 0)
3456 return log_error_errno(errno, "Failed to read cgroup mode: %m");
3457 if (l != sizeof(arg_unified_cgroup_hierarchy)) {
3458 log_error("Short read while reading cgroup mode.");
3459 return -EIO;
3460 }
3461 }
3462
b0067625
ZJS		 3463 /* Wait for the outer child. */
3464 r = wait_for_terminate_and_warn("namespace helper", *pid, NULL);
3465 if (r != 0)
3466 return r < 0 ? r : -EIO;
3467
3468 /* And now retrieve the PID of the inner child. */
3469 l = recv(pid_socket_pair[0], pid, sizeof *pid, 0);
3470 if (l < 0)
3471 return log_error_errno(errno, "Failed to read inner child PID: %m");
3472 if (l != sizeof *pid) {
3473 log_error("Short read while reading inner child PID.");
3474 return -EIO;
3475 }
3476
3477 /* We also retrieve container UUID in case it was generated by outer child */
3478 l = recv(uuid_socket_pair[0], &arg_uuid, sizeof arg_uuid, 0);
3479 if (l < 0)
3480 return log_error_errno(errno, "Failed to read container machine ID: %m");
3481 if (l != sizeof(arg_uuid)) {
3482 log_error("Short read while reading container machined ID.");
3483 return -EIO;
3484 }
3485
3486 /* We also retrieve the socket used for notifications generated by outer child */
3487 notify_socket = receive_one_fd(notify_socket_pair[0], 0);
3488 if (notify_socket < 0)
3489 return log_error_errno(notify_socket,
3490 "Failed to receive notification socket from the outer child: %m");
3491
3492 log_debug("Init process invoked as PID "PID_FMT, *pid);
3493
3494 if (arg_userns_mode != USER_NAMESPACE_NO) {
3495 if (!barrier_place_and_sync(&barrier)) { /* #1 */
3496 log_error("Child died too early.");
3497 return -ESRCH;
3498 }
3499
3500 r = setup_uid_map(*pid);
3501 if (r < 0)
3502 return r;
3503
3504 (void) barrier_place(&barrier); /* #2 */
3505 }
3506
3507 if (arg_private_network) {
3508
3509 r = move_network_interfaces(*pid, arg_network_interfaces);
3510 if (r < 0)
3511 return r;
3512
3513 if (arg_network_veth) {
3514 r = setup_veth(arg_machine, *pid, veth_name,
3515 arg_network_bridge || arg_network_zone);
3516 if (r < 0)
3517 return r;
3518 else if (r > 0)
3519 ifi = r;
3520
3521 if (arg_network_bridge) {
3522 /* Add the interface to a bridge */
3523 r = setup_bridge(veth_name, arg_network_bridge, false);
3524 if (r < 0)
3525 return r;
3526 if (r > 0)
3527 ifi = r;
3528 } else if (arg_network_zone) {
3529 /* Add the interface to a bridge, possibly creating it */
3530 r = setup_bridge(veth_name, arg_network_zone, true);
3531 if (r < 0)
3532 return r;
3533 if (r > 0)
3534 ifi = r;
3535 }
3536 }
3537
3538 r = setup_veth_extra(arg_machine, *pid, arg_network_veth_extra);
3539 if (r < 0)
3540 return r;
3541
3542 /* We created the primary and extra veth links now; let's remember this, so that we know to
3543 remove them later on. Note that we don't bother with removing veth links that were created
3544 here when their setup failed half-way, because in that case the kernel should be able to
3545 remove them on its own, since they cannot be referenced by anything yet. */
3546 *veth_created = true;
3547
3548 r = setup_macvlan(arg_machine, *pid, arg_network_macvlan);
3549 if (r < 0)
3550 return r;
3551
3552 r = setup_ipvlan(arg_machine, *pid, arg_network_ipvlan);
3553 if (r < 0)
3554 return r;
3555 }
3556
abdb9b08
LP		 3557 if (arg_register || !arg_keep_unit) {
3558 r = sd_bus_default_system(&bus);
3559 if (r < 0)
3560 return log_error_errno(r, "Failed to open system bus: %m");
3561 }
3562
3563 if (!arg_keep_unit) {
3564 /* When a new scope is created for this container, then we'll be registered as its controller, in which
3565 * case PID 1 will send us a friendly RequestStop signal, when it is asked to terminate the
3566 * scope. Let's hook into that, and cleanly shut down the container, and print a friendly message. */
3567
3568 r = sd_bus_add_match(bus, NULL,
3569 "type='signal',"
3570 "sender='org.freedesktop.systemd1',"
3571 "interface='org.freedesktop.systemd1.Scope',"
3572 "member='RequestStop'",
3573 on_request_stop, PID_TO_PTR(*pid));
3574 if (r < 0)
3575 return log_error_errno(r, "Failed to install request stop match: %m");
3576 }
3577
b0067625 3578 if (arg_register) {
abdb9b08 3579
b0067625 3580 r = register_machine(
abdb9b08 3581 bus,
b0067625
ZJS		 3582 arg_machine,
3583 *pid,
3584 arg_directory,
3585 arg_uuid,
3586 ifi,
3587 arg_slice,
3588 arg_custom_mounts, arg_n_custom_mounts,
3589 arg_kill_signal,
3590 arg_property,
3591 arg_keep_unit,
3592 arg_container_service_name);
3593 if (r < 0)
3594 return r;
abdb9b08 3595
cd2dfc6f 3596 } else if (!arg_keep_unit) {
abdb9b08 3597
cd2dfc6f 3598 r = allocate_scope(
abdb9b08 3599 bus,
cd2dfc6f
LP		 3600 arg_machine,
3601 *pid,
3602 arg_slice,
3603 arg_custom_mounts, arg_n_custom_mounts,
3604 arg_kill_signal,
3605 arg_property);
3606 if (r < 0)
3607 return r;
3608
3609 } else if (arg_slice || arg_property)
3610 log_notice("Machine and scope registration turned off, --slice= and --property= settings will have no effect.");
b0067625 3611
f0bef277 3612 r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
b0067625
ZJS		 3613 if (r < 0)
3614 return r;
3615
3616 if (arg_keep_unit) {
3617 r = create_subcgroup(*pid, arg_unified_cgroup_hierarchy);
3618 if (r < 0)
3619 return r;
3620 }
3621
de54e02d 3622 r = chown_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
b0067625
ZJS		 3623 if (r < 0)
3624 return r;
3625
3626 /* Notify the child that the parent is ready with all
3627 * its setup (including cgroup-ification), and that
3628 * the child can now hand over control to the code to
3629 * run inside the container. */
3630 (void) barrier_place(&barrier); /* #3 */
3631
3632 /* Block SIGCHLD here, before notifying child.
3633 * process_pty() will handle it with the other signals. */
3634 assert_se(sigprocmask(SIG_BLOCK, &mask_chld, NULL) >= 0);
3635
3636 /* Reset signal to default */
3637 r = default_signals(SIGCHLD, -1);
3638 if (r < 0)
3639 return log_error_errno(r, "Failed to reset SIGCHLD: %m");
3640
3641 r = sd_event_new(&event);
3642 if (r < 0)
3643 return log_error_errno(r, "Failed to get default event source: %m");
3644
8fd010bb
LP		 3645 (void) sd_event_set_watchdog(event, true);
3646
abdb9b08
LP		 3647 if (bus) {
3648 r = sd_bus_attach_event(bus, event, 0);
3649 if (r < 0)
3650 return log_error_errno(r, "Failed to attach bus to event loop: %m");
3651 }
3652
5773024d 3653 r = setup_sd_notify_parent(event, notify_socket, PID_TO_PTR(*pid), &notify_event_source);
b0067625
ZJS		 3654 if (r < 0)
3655 return r;
3656
3657 /* Let the child know that we are ready and wait that the child is completely ready now. */
3658 if (!barrier_place_and_sync(&barrier)) { /* #4 */
3659 log_error("Child died too early.");
3660 return -ESRCH;
3661 }
3662
3663 /* At this point we have made use of the UID we picked, and thus nss-mymachines
3664 * will make them appear in getpwuid(), thus we can release the /etc/passwd lock. */
3665 etc_passwd_lock = safe_close(etc_passwd_lock);
3666
3667 sd_notifyf(false,
3668 "STATUS=Container running.\n"
3669 "X_NSPAWN_LEADER_PID=" PID_FMT, *pid);
3670 if (!arg_notify_ready)
3671 sd_notify(false, "READY=1\n");
3672
3673 if (arg_kill_signal > 0) {
3674 /* Try to kill the init system on SIGINT or SIGTERM */
3675 sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, PID_TO_PTR(*pid));
3676 sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, PID_TO_PTR(*pid));
3677 } else {
3678 /* Immediately exit */
3679 sd_event_add_signal(event, NULL, SIGINT, NULL, NULL);
3680 sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL);
3681 }
3682
6916b164
AU		 3683 /* Exit when the child exits */
3684 sd_event_add_signal(event, NULL, SIGCHLD, on_sigchld, PID_TO_PTR(*pid));
b0067625
ZJS		 3685
3686 if (arg_expose_ports) {
3687 r = expose_port_watch_rtnl(event, rtnl_socket_pair[0], on_address_change, exposed, &rtnl);
3688 if (r < 0)
3689 return r;
3690
3691 (void) expose_port_execute(rtnl, arg_expose_ports, exposed);
3692 }
3693
3694 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
3695
3696 r = pty_forward_new(event, master,
3697 PTY_FORWARD_IGNORE_VHANGUP | (interactive ? 0 : PTY_FORWARD_READ_ONLY),
3698 &forward);
3699 if (r < 0)
3700 return log_error_errno(r, "Failed to create PTY forwarder: %m");
3701
3702 r = sd_event_loop(event);
3703 if (r < 0)
3704 return log_error_errno(r, "Failed to run event loop: %m");
3705
3706 pty_forward_get_last_char(forward, &last_char);
3707
3708 forward = pty_forward_free(forward);
3709
3710 if (!arg_quiet && last_char != '\n')
3711 putc('\n', stdout);
3712
3713 /* Kill if it is not dead yet anyway */
abdb9b08
LP		 3714 if (arg_register && !arg_keep_unit && bus)
3715 terminate_machine(bus, *pid);
b0067625
ZJS		 3716
3717 /* Normally redundant, but better safe than sorry */
c67b0082 3718 (void) kill(*pid, SIGKILL);
b0067625
ZJS		 3719
3720 r = wait_for_container(*pid, &container_status);
3721 *pid = 0;
3722
3723 if (r < 0)
3724 /* We failed to wait for the container, or the container exited abnormally. */
3725 return r;
3726 if (r > 0 || container_status == CONTAINER_TERMINATED) {
27e29a1e
ZJS		 3727 /* r > 0 → The container exited with a non-zero status.
3728 * As a special case, we need to replace 133 with a different value,
3729 * because 133 is special-cased in the service file to reboot the container.
3730 * otherwise → The container exited with zero status and a reboot was not requested.
3731 */
2a49b612 3732 if (r == EXIT_FORCE_RESTART)
27e29a1e 3733 r = EXIT_FAILURE; /* replace 133 with the general failure code */
b0067625 3734 *ret = r;
b0067625
ZJS		 3735 return 0; /* finito */
3736 }
3737
3738 /* CONTAINER_REBOOTED, loop again */
3739
3740 if (arg_keep_unit) {
3741 /* Special handling if we are running as a service: instead of simply
3742 * restarting the machine we want to restart the entire service, so let's
3743 * inform systemd about this with the special exit code 133. The service
3744 * file uses RestartForceExitStatus=133 so that this results in a full
3745 * nspawn restart. This is necessary since we might have cgroup parameters
3746 * set we want to have flushed out. */
2a49b612
ZJS		 3747 *ret = EXIT_FORCE_RESTART;
3748 return 0; /* finito */
b0067625
ZJS		 3749 }
3750
3751 expose_port_flush(arg_expose_ports, exposed);
3752
3753 (void) remove_veth_links(veth_name, arg_network_veth_extra);
3754 *veth_created = false;
3755 return 1; /* loop again */
3756}
3757
03cfe0d5
LP		 3758int main(int argc, char *argv[]) {
3759
2d845785
LP		 3760 _cleanup_free_ char *console = NULL;
3761 _cleanup_close_ int master = -1;
03cfe0d5 3762 _cleanup_fdset_free_ FDSet *fds = NULL;
2d845785 3763 int r, n_fd_passed, ret = EXIT_SUCCESS;
5aa3eba5 3764 char veth_name[IFNAMSIZ] = "";
17cbb288 3765 bool secondary = false, remove_directory = false, remove_image = false;
03cfe0d5 3766 pid_t pid = 0;
03cfe0d5
LP		 3767 union in_addr_union exposed = {};
3768 _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT;
c67b0082
LP		 3769 bool interactive, veth_created = false, remove_tmprootdir = false;
3770 char tmprootdir[] = "/tmp/nspawn-root-XXXXXX";
2d845785 3771 _cleanup_(loop_device_unrefp) LoopDevice *loop = NULL;
18b5886e
LP		 3772 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
3773 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
03cfe0d5
LP		 3774
3775 log_parse_environment();
3776 log_open();
415fc41c 3777
7732f92b
LP		 3778 /* Make sure rename_process() in the stub init process can work */
3779 saved_argv = argv;
3780 saved_argc = argc;
3781
03cfe0d5
LP		 3782 r = parse_argv(argc, argv);
3783 if (r <= 0)
3784 goto finish;
3785
03cfe0d5
LP		 3786 if (geteuid() != 0) {
3787 log_error("Need to be root.");
3788 r = -EPERM;
3789 goto finish;
3790 }
f757855e
LP		 3791 r = determine_names();
3792 if (r < 0)
3793 goto finish;
3794
3795 r = load_settings();
3796 if (r < 0)
3797 goto finish;
3798
3799 r = verify_arguments();
3800 if (r < 0)
3801 goto finish;
03cfe0d5 3802
8199d554
LP		 3803 r = detect_unified_cgroup_hierarchy_from_environment();
3804 if (r < 0)
3805 goto finish;
3806
03cfe0d5
LP		 3807 n_fd_passed = sd_listen_fds(false);
3808 if (n_fd_passed > 0) {
3809 r = fdset_new_listen_fds(&fds, false);
3810 if (r < 0) {
3811 log_error_errno(r, "Failed to collect file descriptors: %m");
3812 goto finish;
3813 }
3814 }
3815
3816 if (arg_directory) {
3817 assert(!arg_image);
3818
3819 if (path_equal(arg_directory, "/") && !arg_ephemeral) {
3820 log_error("Spawning container on root directory is not supported. Consider using --ephemeral.");
3821 r = -EINVAL;
3822 goto finish;
3823 }
3824
3825 if (arg_ephemeral) {
3826 _cleanup_free_ char *np = NULL;
3827
8d4aa2bb 3828 r = chase_symlinks_and_update(&arg_directory, 0);
3f342ec4
LP		 3829 if (r < 0)
3830 goto finish;
3831
03cfe0d5
LP		 3832 /* If the specified path is a mount point we
3833 * generate the new snapshot immediately
3834 * inside it under a random name. However if
3835 * the specified is not a mount point we
3836 * create the new snapshot in the parent
3837 * directory, just next to it. */
e1873695 3838 r = path_is_mount_point(arg_directory, NULL, 0);
03cfe0d5
LP		 3839 if (r < 0) {
3840 log_error_errno(r, "Failed to determine whether directory %s is mount point: %m", arg_directory);
3841 goto finish;
3842 }
3843 if (r > 0)
770b5ce4 3844 r = tempfn_random_child(arg_directory, "machine.", &np);
03cfe0d5 3845 else
770b5ce4 3846 r = tempfn_random(arg_directory, "machine.", &np);
03cfe0d5 3847 if (r < 0) {
0f3be6ca 3848 log_error_errno(r, "Failed to generate name for directory snapshot: %m");
03cfe0d5
LP		 3849 goto finish;
3850 }
3851
3852 r = image_path_lock(np, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3853 if (r < 0) {
3854 log_error_errno(r, "Failed to lock %s: %m", np);
3855 goto finish;
3856 }
3857
17cbb288
LP		 3858 r = btrfs_subvol_snapshot(arg_directory, np,
3859 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
3860 BTRFS_SNAPSHOT_FALLBACK_COPY |
3861 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
3862 BTRFS_SNAPSHOT_RECURSIVE |
3863 BTRFS_SNAPSHOT_QUOTA);
03cfe0d5
LP		 3864 if (r < 0) {
3865 log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory);
3866 goto finish;
ec16945e
LP		 3867 }
3868
3869 free(arg_directory);
3870 arg_directory = np;
8a16a7b4 3871 np = NULL;
ec16945e 3872
17cbb288 3873 remove_directory = true;
30535c16
LP		 3874
3875 } else {
cb638b5e 3876 r = chase_symlinks_and_update(&arg_directory, arg_template ? CHASE_NONEXISTENT : 0);
8d4aa2bb
LP		 3877 if (r < 0)
3878 goto finish;
3879
30535c16
LP		 3880 r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3881 if (r == -EBUSY) {
3882 log_error_errno(r, "Directory tree %s is currently busy.", arg_directory);
3883 goto finish;
3884 }
3885 if (r < 0) {
3886 log_error_errno(r, "Failed to lock %s: %m", arg_directory);
476b8254 3887 goto finish;
30535c16
LP		 3888 }
3889
3890 if (arg_template) {
8d4aa2bb 3891 r = chase_symlinks_and_update(&arg_template, 0);
3f342ec4
LP		 3892 if (r < 0)
3893 goto finish;
3894
17cbb288
LP		 3895 r = btrfs_subvol_snapshot(arg_template, arg_directory,
3896 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
3897 BTRFS_SNAPSHOT_FALLBACK_COPY |
3898 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
3899 BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE |
3900 BTRFS_SNAPSHOT_RECURSIVE |
3901 BTRFS_SNAPSHOT_QUOTA);
30535c16
LP		 3902 if (r == -EEXIST) {
3903 if (!arg_quiet)
3904 log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template);
3905 } else if (r < 0) {
83521414 3906 log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template);
30535c16
LP		 3907 goto finish;
3908 } else {
3909 if (!arg_quiet)
3910 log_info("Populated %s from template %s.", arg_directory, arg_template);
3911 }
3912 }
ec16945e
LP		 3913 }
3914
7732f92b 3915 if (arg_start_mode == START_BOOT) {
1b9e5b12 3916 if (path_is_os_tree(arg_directory) <= 0) {
5ae4d543 3917 log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory);
ec16945e 3918 r = -EINVAL;
1b9e5b12
LP		 3919 goto finish;
3920 }
3921 } else {
3922 const char *p;
3923
16fb773e
LP		 3924 p = strjoina(arg_directory, "/usr/");
3925 if (laccess(p, F_OK) < 0) {
3926 log_error("Directory %s doesn't look like it has an OS tree. Refusing.", arg_directory);
ec16945e 3927 r = -EINVAL;
1b9e5b12 3928 goto finish;
1b9e5b12
LP		 3929 }
3930 }
ec16945e 3931
6b9132a9 3932 } else {
ec16945e
LP		 3933 assert(arg_image);
3934 assert(!arg_template);
3935
8d4aa2bb 3936 r = chase_symlinks_and_update(&arg_image, 0);
3f342ec4
LP		 3937 if (r < 0)
3938 goto finish;
3939
0f3be6ca
LP		 3940 if (arg_ephemeral) {
3941 _cleanup_free_ char *np = NULL;
3942
3943 r = tempfn_random(arg_image, "machine.", &np);
3944 if (r < 0) {
3945 log_error_errno(r, "Failed to generate name for image snapshot: %m");
3946 goto finish;
3947 }
3948
3949 r = image_path_lock(np, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3950 if (r < 0) {
3951 r = log_error_errno(r, "Failed to create image lock: %m");
3952 goto finish;
3953 }
3954
1c876927 3955 r = copy_file(arg_image, np, O_EXCL, arg_read_only ? 0400 : 0600, FS_NOCOW_FL, COPY_REFLINK);
0f3be6ca
LP		 3956 if (r < 0) {
3957 r = log_error_errno(r, "Failed to copy image file: %m");
3958 goto finish;
3959 }
3960
3961 free(arg_image);
3962 arg_image = np;
3963 np = NULL;
3964
3965 remove_image = true;
3966 } else {
3967 r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
3968 if (r == -EBUSY) {
3969 r = log_error_errno(r, "Disk image %s is currently busy.", arg_image);
3970 goto finish;
3971 }
3972 if (r < 0) {
3973 r = log_error_errno(r, "Failed to create image lock: %m");
3974 goto finish;
3975 }
4623e8e6 3976
78ebe980
LP		 3977 if (!arg_root_hash) {
3978 r = root_hash_load(arg_image, &arg_root_hash, &arg_root_hash_size);
3979 if (r < 0) {
3980 log_error_errno(r, "Failed to load root hash file for %s: %m", arg_image);
3981 goto finish;
3982 }
3983 }
30535c16
LP		 3984 }
3985
c67b0082 3986 if (!mkdtemp(tmprootdir)) {
0f3be6ca 3987 r = log_error_errno(errno, "Failed to create temporary directory: %m");
6b9132a9 3988 goto finish;
1b9e5b12 3989 }
6b9132a9 3990
c67b0082
LP		 3991 remove_tmprootdir = true;
3992
3993 arg_directory = strdup(tmprootdir);
1b9e5b12
LP		 3994 if (!arg_directory) {
3995 r = log_oom();
3996 goto finish;
6b9132a9 3997 }
88213476 3998
2d845785
LP		 3999 r = loop_device_make_by_path(arg_image, arg_read_only ? O_RDONLY : O_RDWR, &loop);
4000 if (r < 0) {
4001 log_error_errno(r, "Failed to set up loopback block device: %m");
842f3b0f
LP		 4002 goto finish;
4003 }
1b9e5b12 4004
e0f9e7bd
LP		 4005 r = dissect_image(
4006 loop->fd,
4007 arg_root_hash, arg_root_hash_size,
4008 DISSECT_IMAGE_REQUIRE_ROOT,
4009 &dissected_image);
2d845785
LP		 4010 if (r == -ENOPKG) {
4011 log_error_errno(r, "Could not find a suitable file system or partition table in image: %s", arg_image);
4012
4013 log_notice("Note that the disk image needs to\n"
4014 " a) either contain only a single MBR partition of type 0x83 that is marked bootable\n"
4015 " b) or contain a single GPT partition of type 0FC63DAF-8483-4772-8E79-3D69D8477DE4\n"
4016 " c) or follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/\n"
4017 " d) or contain a file system without a partition table\n"
4018 "in order to be bootable with systemd-nspawn.");
1b9e5b12 4019 goto finish;
2d845785 4020 }
4623e8e6
LP		 4021 if (r == -EADDRNOTAVAIL) {
4022 log_error_errno(r, "No root partition for specified root hash found.");
4023 goto finish;
4024 }
2d845785
LP		 4025 if (r == -EOPNOTSUPP) {
4026 log_error_errno(r, "--image= is not supported, compiled without blkid support.");
4027 goto finish;
4028 }
759aaedc
LP		 4029 if (r == -EPROTONOSUPPORT) {
4030 log_error_errno(r, "Device is loopback block device with partition scanning turned off, please turn it on.");
4031 goto finish;
4032 }
2d845785
LP		 4033 if (r < 0) {
4034 log_error_errno(r, "Failed to dissect image: %m");
842f3b0f
LP		 4035 goto finish;
4036 }
1b9e5b12 4037
4623e8e6
LP		 4038 if (!arg_root_hash && dissected_image->can_verity)
4039 log_notice("Note: image %s contains verity information, but no root hash specified! Proceeding without integrity checking.", arg_image);
4040
4041 r = dissected_image_decrypt_interactively(dissected_image, NULL, arg_root_hash, arg_root_hash_size, 0, &decrypted_image);
1b9e5b12
LP		 4042 if (r < 0)
4043 goto finish;
0f3be6ca
LP		 4044
4045 /* Now that we mounted the image, let's try to remove it again, if it is ephemeral */
4046 if (remove_image && unlink(arg_image) >= 0)
4047 remove_image = false;
842f3b0f 4048 }
842f3b0f 4049
86c0dd4a 4050 r = custom_mount_prepare_all(arg_directory, arg_custom_mounts, arg_n_custom_mounts);
5a8af538
LP		 4051 if (r < 0)
4052 goto finish;
4053
03cfe0d5
LP		 4054 interactive =
4055 isatty(STDIN_FILENO) > 0 &&
4056 isatty(STDOUT_FILENO) > 0;
9c857b9d 4057
db7feb7e
LP		 4058 master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
4059 if (master < 0) {
ec16945e 4060 r = log_error_errno(errno, "Failed to acquire pseudo tty: %m");
a258bf26
LP		 4061 goto finish;
4062 }
4063
611b312b
LP		 4064 r = ptsname_malloc(master, &console);
4065 if (r < 0) {
4066 r = log_error_errno(r, "Failed to determine tty name: %m");
a258bf26 4067 goto finish;
68b02049
DW		 4068 }
4069
4070 if (arg_selinux_apifs_context) {
4071 r = mac_selinux_apply(console, arg_selinux_apifs_context);
4072 if (r < 0)
4073 goto finish;
a258bf26
LP		 4074 }
4075
a258bf26 4076 if (unlockpt(master) < 0) {
ec16945e 4077 r = log_error_errno(errno, "Failed to unlock tty: %m");
a258bf26
LP		 4078 goto finish;
4079 }
4080
9c857b9d
LP		 4081 if (!arg_quiet)
4082 log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.",
4083 arg_machine, arg_image ?: arg_directory);
4084
72c0a2c2 4085 assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1) >= 0);
a258bf26 4086
03cfe0d5
LP		 4087 if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) {
4088 r = log_error_errno(errno, "Failed to become subreaper: %m");
4089 goto finish;
4090 }
4091
d87be9b0 4092 for (;;) {
b0067625
ZJS		 4093 r = run(master,
4094 console,
2d845785 4095 dissected_image,
b0067625
ZJS		 4096 interactive, secondary,
4097 fds,
4098 veth_name, &veth_created,
4099 &exposed,
4100 &pid, &ret);
4101 if (r <= 0)
d87be9b0 4102 break;
d87be9b0 4103 }
88213476
LP		 4104
4105finish:
af4ec430 4106 sd_notify(false,
2a49b612
ZJS		 4107 r == 0 && ret == EXIT_FORCE_RESTART ? "STOPPING=1\nSTATUS=Restarting..." :
4108 "STOPPING=1\nSTATUS=Terminating...");
af4ec430 4109
9444b1f2 4110 if (pid > 0)
c67b0082 4111 (void) kill(pid, SIGKILL);
88213476 4112
503546da 4113 /* Try to flush whatever is still queued in the pty */
6a0f896b 4114 if (master >= 0) {
1c876927 4115 (void) copy_bytes(master, STDOUT_FILENO, (uint64_t) -1, 0);
6a0f896b
LP		 4116 master = safe_close(master);
4117 }
4118
4119 if (pid > 0)
4120 (void) wait_for_terminate(pid, NULL);
503546da 4121
17cbb288 4122 if (remove_directory && arg_directory) {
ec16945e
LP		 4123 int k;
4124
17cbb288 4125 k = rm_rf(arg_directory, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME);
ec16945e 4126 if (k < 0)
17cbb288 4127 log_warning_errno(k, "Cannot remove '%s', ignoring: %m", arg_directory);
ec16945e
LP		 4128 }
4129
0f3be6ca
LP		 4130 if (remove_image && arg_image) {
4131 if (unlink(arg_image) < 0)
4132 log_warning_errno(errno, "Can't remove image file '%s', ignoring: %m", arg_image);
4133 }
4134
c67b0082
LP		 4135 if (remove_tmprootdir) {
4136 if (rmdir(tmprootdir) < 0)
4137 log_debug_errno(errno, "Can't remove temporary root directory '%s', ignoring: %m", tmprootdir);
4138 }
4139
785890ac
LP		 4140 if (arg_machine) {
4141 const char *p;
4142
63c372cb 4143 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
c6878637 4144 (void) rm_rf(p, REMOVE_ROOT);
785890ac
LP		 4145 }
4146
7a8f6325 4147 expose_port_flush(arg_expose_ports, &exposed);
7513c5b8
LP		 4148
4149 if (veth_created)
4150 (void) remove_veth_links(veth_name, arg_network_veth_extra);
22b28dfd 4151 (void) remove_bridge(arg_network_zone);
f757855e 4152
04d391da 4153 free(arg_directory);
ec16945e
LP		 4154 free(arg_template);
4155 free(arg_image);
7027ff61 4156 free(arg_machine);
c74e630d 4157 free(arg_user);
b53ede69
PW		 4158 free(arg_pivot_root_new);
4159 free(arg_pivot_root_old);
5f932eb9 4160 free(arg_chdir);
c74e630d 4161 strv_free(arg_setenv);
f757855e 4162 free(arg_network_bridge);
c74e630d
LP		 4163 strv_free(arg_network_interfaces);
4164 strv_free(arg_network_macvlan);
4bbfe7ad 4165 strv_free(arg_network_ipvlan);
f6d6bad1 4166 strv_free(arg_network_veth_extra);
f757855e
LP		 4167 strv_free(arg_parameters);
4168 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
4169 expose_port_free_all(arg_expose_ports);
4623e8e6 4170 free(arg_root_hash);
6d0b55c2 4171
ec16945e 4172 return r < 0 ? EXIT_FAILURE : ret;
88213476 4173}
Unnamed repository; edit this file 'description' to name the repository.