]> 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
pid1,nspawn: raise default RLIMIT_MEMLOCK to 8M
[thirdparty/systemd.git] / src / nspawn / nspawn.c
CommitLineData
db9ecf05 1/* SPDX-License-Identifier: LGPL-2.1-or-later */
88213476 2
349cc4a5 3#if HAVE_BLKID
8fe0087e 4#endif
88213476 5#include <errno.h>
88213476 6#include <getopt.h>
503f480f 7#include <linux/fs.h>
1b9e5b12 8#include <linux/loop.h>
349cc4a5 9#if HAVE_SELINUX
8fe0087e 10#include <selinux/selinux.h>
1b9e5b12 11#endif
8fe0087e 12#include <stdlib.h>
8fe0087e 13#include <sys/file.h>
335d2ead 14#include <sys/ioctl.h>
8fe0087e
LP		 15#include <sys/personality.h>
16#include <sys/prctl.h>
17#include <sys/types.h>
6916b164 18#include <sys/wait.h>
335d2ead 19#include <termios.h>
8fe0087e 20#include <unistd.h>
1b9e5b12 21
b053cd5f 22#include "sd-bus.h"
1f0cd86b 23#include "sd-daemon.h"
1f0cd86b 24#include "sd-id128.h"
8fe0087e 25
b5efdb8a 26#include "alloc-util.h"
8fe0087e
LP		 27#include "barrier.h"
28#include "base-filesystem.h"
29#include "blkid-util.h"
30#include "btrfs-util.h"
b8ea7a6e 31#include "bus-error.h"
b053cd5f 32#include "bus-util.h"
8fe0087e 33#include "cap-list.h"
430f0182 34#include "capability-util.h"
04d391da 35#include "cgroup-util.h"
f4351959 36#include "chase-symlinks.h"
8fe0087e 37#include "copy.h"
d107bb7d 38#include "cpu-set-util.h"
786d19fd 39#include "creds-util.h"
4fc9982c 40#include "dev-setup.h"
57f1b61b 41#include "discover-image.h"
2d845785 42#include "dissect-image.h"
8fe0087e 43#include "env-util.h"
3652872a 44#include "escape.h"
3ffd4af2 45#include "fd-util.h"
842f3b0f 46#include "fdset.h"
a5c32cff 47#include "fileio.h"
f97b34a6 48#include "format-util.h"
f4f15635 49#include "fs-util.h"
1b9e5b12 50#include "gpt.h"
4623e8e6 51#include "hexdecoct.h"
e2054217 52#include "hostname-setup.h"
8fe0087e 53#include "hostname-util.h"
910fd145 54#include "id128-util.h"
3652872a 55#include "io-util.h"
8fe0087e 56#include "log.h"
2d845785 57#include "loop-util.h"
8fe0087e 58#include "loopback-setup.h"
8fe0087e 59#include "macro.h"
44dbef90 60#include "main-func.h"
f5947a5e 61#include "missing_sched.h"
8fe0087e 62#include "mkdir.h"
4349cd7c 63#include "mount-util.h"
049af8ad 64#include "mountpoint-util.h"
0cb8e3d1 65#include "namespace-util.h"
8fe0087e 66#include "netlink-util.h"
2f893044 67#include "nspawn-bind-user.h"
07630cea 68#include "nspawn-cgroup.h"
3652872a 69#include "nspawn-creds.h"
3603efde 70#include "nspawn-def.h"
07630cea
LP		 71#include "nspawn-expose-ports.h"
72#include "nspawn-mount.h"
73#include "nspawn-network.h"
de40a303 74#include "nspawn-oci.h"
7336138e 75#include "nspawn-patch-uid.h"
07630cea 76#include "nspawn-register.h"
910fd145 77#include "nspawn-seccomp.h"
07630cea
LP		 78#include "nspawn-settings.h"
79#include "nspawn-setuid.h"
7732f92b 80#include "nspawn-stub-pid1.h"
91181e07 81#include "nspawn.h"
d8b4d14d 82#include "nulstr-util.h"
d58ad743 83#include "os-util.h"
50ebcf6c 84#include "pager.h"
614b022c 85#include "parse-argument.h"
6bedfcbb 86#include "parse-util.h"
8fe0087e 87#include "path-util.h"
294bf0c3 88#include "pretty-print.h"
0b452006 89#include "process-util.h"
8fe0087e
LP		 90#include "ptyfwd.h"
91#include "random-util.h"
8869a0b4 92#include "raw-clone.h"
86775e35 93#include "resolve-util.h"
bf428efb 94#include "rlimit-util.h"
8fe0087e 95#include "rm-rf.h"
de40a303
LP		 96#if HAVE_SECCOMP
97#include "seccomp-util.h"
98#endif
68b02049 99#include "selinux-util.h"
8fe0087e 100#include "signal-util.h"
2583fbea 101#include "socket-util.h"
8fcde012 102#include "stat-util.h"
15a5e950 103#include "stdio-util.h"
5c828e66 104#include "string-table.h"
07630cea 105#include "string-util.h"
8fe0087e 106#include "strv.h"
de40a303 107#include "sysctl-util.h"
8fe0087e 108#include "terminal-util.h"
e4de7287 109#include "tmpfile-util.h"
affb60b1 110#include "umask-util.h"
43c3fb46 111#include "unit-name.h"
b1d4f8e1 112#include "user-util.h"
8fe0087e 113#include "util.h"
e9642be2 114
e96ceaba
LP		 115/* The notify socket inside the container it can use to talk to nspawn using the sd_notify(3) protocol */
116#define NSPAWN_NOTIFY_SOCKET_PATH "/run/host/notify"
0e7ac751 117
2a49b612
ZJS		 118#define EXIT_FORCE_RESTART 133
119
113cea80
DH		 120typedef enum ContainerStatus {
121 CONTAINER_TERMINATED,
6145bb4f 122 CONTAINER_REBOOTED,
113cea80
DH		 123} ContainerStatus;
124
88213476 125static char *arg_directory = NULL;
ec16945e 126static char *arg_template = NULL;
5f932eb9 127static char *arg_chdir = NULL;
b53ede69
PW		 128static char *arg_pivot_root_new = NULL;
129static char *arg_pivot_root_old = NULL;
687d0825 130static char *arg_user = NULL;
de40a303
LP		 131static uid_t arg_uid = UID_INVALID;
132static gid_t arg_gid = GID_INVALID;
133static gid_t* arg_supplementary_gids = NULL;
134static size_t arg_n_supplementary_gids = 0;
9444b1f2 135static sd_id128_t arg_uuid = {};
3a9530e5
LP		 136static char *arg_machine = NULL; /* The name used by the host to refer to this */
137static char *arg_hostname = NULL; /* The name the payload sees by default */
c74e630d
LP		 138static const char *arg_selinux_context = NULL;
139static const char *arg_selinux_apifs_context = NULL;
de40a303 140static char *arg_slice = NULL;
ff01d048 141static bool arg_private_network = false;
bc2f673e 142static bool arg_read_only = false;
7732f92b 143static StartMode arg_start_mode = START_PID1;
ec16945e 144static bool arg_ephemeral = false;
57fb9fb5 145static LinkJournal arg_link_journal = LINK_AUTO;
574edc90 146static bool arg_link_journal_try = false;
520e0d54 147static uint64_t arg_caps_retain =
50b52222
LP		 148 (1ULL << CAP_AUDIT_CONTROL) |
149 (1ULL << CAP_AUDIT_WRITE) |
5076f0cc
LP		 150 (1ULL << CAP_CHOWN) |
151 (1ULL << CAP_DAC_OVERRIDE) |
152 (1ULL << CAP_DAC_READ_SEARCH) |
153 (1ULL << CAP_FOWNER) |
154 (1ULL << CAP_FSETID) |
155 (1ULL << CAP_IPC_OWNER) |
156 (1ULL << CAP_KILL) |
157 (1ULL << CAP_LEASE) |
158 (1ULL << CAP_LINUX_IMMUTABLE) |
50b52222 159 (1ULL << CAP_MKNOD) |
5076f0cc
LP		 160 (1ULL << CAP_NET_BIND_SERVICE) |
161 (1ULL << CAP_NET_BROADCAST) |
162 (1ULL << CAP_NET_RAW) |
5076f0cc 163 (1ULL << CAP_SETFCAP) |
50b52222 164 (1ULL << CAP_SETGID) |
5076f0cc
LP		 165 (1ULL << CAP_SETPCAP) |
166 (1ULL << CAP_SETUID) |
167 (1ULL << CAP_SYS_ADMIN) |
50b52222 168 (1ULL << CAP_SYS_BOOT) |
5076f0cc
LP		 169 (1ULL << CAP_SYS_CHROOT) |
170 (1ULL << CAP_SYS_NICE) |
171 (1ULL << CAP_SYS_PTRACE) |
d87be9b0 172 (1ULL << CAP_SYS_RESOURCE) |
50b52222 173 (1ULL << CAP_SYS_TTY_CONFIG);
88fc9c9b 174static uint64_t arg_caps_ambient = 0;
de40a303 175static CapabilityQuintet arg_full_capabilities = CAPABILITY_QUINTET_NULL;
5a8af538 176static CustomMount *arg_custom_mounts = NULL;
88614c8a 177static size_t arg_n_custom_mounts = 0;
f4889f65 178static char **arg_setenv = NULL;
284c0b91 179static bool arg_quiet = false;
eb91eb18 180static bool arg_register = true;
89f7c846 181static bool arg_keep_unit = false;
aa28aefe 182static char **arg_network_interfaces = NULL;
c74e630d 183static char **arg_network_macvlan = NULL;
4bbfe7ad 184static char **arg_network_ipvlan = NULL;
69c79d3c 185static bool arg_network_veth = false;
f6d6bad1 186static char **arg_network_veth_extra = NULL;
f757855e 187static char *arg_network_bridge = NULL;
22b28dfd 188static char *arg_network_zone = NULL;
d7bea6b6 189static char *arg_network_namespace_path = NULL;
bb068de0 190static PagerFlags arg_pager_flags = 0;
050f7277 191static unsigned long arg_personality = PERSONALITY_INVALID;
ec16945e 192static char *arg_image = NULL;
de40a303 193static char *arg_oci_bundle = NULL;
f757855e 194static VolatileMode arg_volatile_mode = VOLATILE_NO;
6d0b55c2 195static ExposePort *arg_expose_ports = NULL;
f36933fe 196static char **arg_property = NULL;
de40a303 197static sd_bus_message *arg_property_message = NULL;
0de7acce 198static UserNamespaceMode arg_userns_mode = USER_NAMESPACE_NO;
6dac160c 199static uid_t arg_uid_shift = UID_INVALID, arg_uid_range = 0x10000U;
6c045a99 200static UserNamespaceOwnership arg_userns_ownership = _USER_NAMESPACE_OWNERSHIP_INVALID;
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;
1099ceeb 210static MountSettingsMask arg_mount_settings = MOUNT_APPLY_APIVFS_RO|MOUNT_APPLY_TMPFS_TMP;
aee36b4e 211static VeritySettings arg_verity_settings = VERITY_SETTINGS_DEFAULT;
6b000af4
LP		 212static char **arg_syscall_allow_list = NULL;
213static char **arg_syscall_deny_list = NULL;
de40a303
LP		 214#if HAVE_SECCOMP
215static scmp_filter_ctx arg_seccomp = NULL;
216#endif
bf428efb 217static struct rlimit *arg_rlimit[_RLIMIT_MAX] = {};
66edd963 218static bool arg_no_new_privileges = false;
81f345df
LP		 219static int arg_oom_score_adjust = 0;
220static bool arg_oom_score_adjust_set = false;
0985c7c4 221static CPUSet arg_cpu_set = {};
09d423e9 222static ResolvConfMode arg_resolv_conf = RESOLV_CONF_AUTO;
1688841f 223static TimezoneMode arg_timezone = TIMEZONE_AUTO;
f5fbe71d 224static unsigned arg_console_width = UINT_MAX, arg_console_height = UINT_MAX;
de40a303
LP		 225static DeviceNode* arg_extra_nodes = NULL;
226static size_t arg_n_extra_nodes = 0;
227static char **arg_sysctl = NULL;
228static ConsoleMode arg_console_mode = _CONSOLE_MODE_INVALID;
3652872a
LP		 229static Credential *arg_credentials = NULL;
230static size_t arg_n_credentials = 0;
2f893044 231static char **arg_bind_user = NULL;
4a4654e0 232static bool arg_suppress_sync = false;
88213476 233
6145bb4f
LP		 234STATIC_DESTRUCTOR_REGISTER(arg_directory, freep);
235STATIC_DESTRUCTOR_REGISTER(arg_template, freep);
236STATIC_DESTRUCTOR_REGISTER(arg_chdir, freep);
237STATIC_DESTRUCTOR_REGISTER(arg_pivot_root_new, freep);
238STATIC_DESTRUCTOR_REGISTER(arg_pivot_root_old, freep);
239STATIC_DESTRUCTOR_REGISTER(arg_user, freep);
240STATIC_DESTRUCTOR_REGISTER(arg_supplementary_gids, freep);
241STATIC_DESTRUCTOR_REGISTER(arg_machine, freep);
242STATIC_DESTRUCTOR_REGISTER(arg_hostname, freep);
243STATIC_DESTRUCTOR_REGISTER(arg_slice, freep);
244STATIC_DESTRUCTOR_REGISTER(arg_setenv, strv_freep);
245STATIC_DESTRUCTOR_REGISTER(arg_network_interfaces, strv_freep);
246STATIC_DESTRUCTOR_REGISTER(arg_network_macvlan, strv_freep);
247STATIC_DESTRUCTOR_REGISTER(arg_network_ipvlan, strv_freep);
248STATIC_DESTRUCTOR_REGISTER(arg_network_veth_extra, strv_freep);
249STATIC_DESTRUCTOR_REGISTER(arg_network_bridge, freep);
250STATIC_DESTRUCTOR_REGISTER(arg_network_zone, freep);
251STATIC_DESTRUCTOR_REGISTER(arg_network_namespace_path, freep);
252STATIC_DESTRUCTOR_REGISTER(arg_image, freep);
253STATIC_DESTRUCTOR_REGISTER(arg_oci_bundle, freep);
254STATIC_DESTRUCTOR_REGISTER(arg_property, strv_freep);
255STATIC_DESTRUCTOR_REGISTER(arg_property_message, sd_bus_message_unrefp);
256STATIC_DESTRUCTOR_REGISTER(arg_parameters, strv_freep);
89e62e0b 257STATIC_DESTRUCTOR_REGISTER(arg_verity_settings, verity_settings_done);
6b000af4
LP		 258STATIC_DESTRUCTOR_REGISTER(arg_syscall_allow_list, strv_freep);
259STATIC_DESTRUCTOR_REGISTER(arg_syscall_deny_list, strv_freep);
6145bb4f
LP		 260#if HAVE_SECCOMP
261STATIC_DESTRUCTOR_REGISTER(arg_seccomp, seccomp_releasep);
262#endif
0985c7c4 263STATIC_DESTRUCTOR_REGISTER(arg_cpu_set, cpu_set_reset);
6145bb4f 264STATIC_DESTRUCTOR_REGISTER(arg_sysctl, strv_freep);
2f893044 265STATIC_DESTRUCTOR_REGISTER(arg_bind_user, strv_freep);
6145bb4f 266
dce66ffe
ZJS		 267static int handle_arg_console(const char *arg) {
268 if (streq(arg, "help")) {
10e8a60b
LP		 269 puts("autopipe\n"
270 "interactive\n"
dce66ffe 271 "passive\n"
10e8a60b
LP		 272 "pipe\n"
273 "read-only");
dce66ffe
ZJS		 274 return 0;
275 }
276
277 if (streq(arg, "interactive"))
278 arg_console_mode = CONSOLE_INTERACTIVE;
279 else if (streq(arg, "read-only"))
280 arg_console_mode = CONSOLE_READ_ONLY;
281 else if (streq(arg, "passive"))
282 arg_console_mode = CONSOLE_PASSIVE;
554c4beb
LP		 283 else if (streq(arg, "pipe")) {
284 if (isatty(STDIN_FILENO) > 0 && isatty(STDOUT_FILENO) > 0)
285 log_full(arg_quiet ? LOG_DEBUG : LOG_NOTICE,
286 "Console mode 'pipe' selected, but standard input/output are connected to an interactive TTY. "
287 "Most likely you want to use 'interactive' console mode for proper interactivity and shell job control. "
288 "Proceeding anyway.");
289
dce66ffe 290 arg_console_mode = CONSOLE_PIPE;
10e8a60b
LP		 291 } else if (streq(arg, "autopipe")) {
292 if (isatty(STDIN_FILENO) > 0 && isatty(STDOUT_FILENO) > 0)
293 arg_console_mode = CONSOLE_INTERACTIVE;
294 else
295 arg_console_mode = CONSOLE_PIPE;
554c4beb 296 } else
dce66ffe
ZJS		 297 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Unknown console mode: %s", optarg);
298
299 arg_settings_mask |= SETTING_CONSOLE_MODE;
300 return 1;
301}
302
37ec0fdd
LP		 303static int help(void) {
304 _cleanup_free_ char *link = NULL;
305 int r;
306
384c2c32 307 pager_open(arg_pager_flags);
50ebcf6c 308
37ec0fdd
LP		 309 r = terminal_urlify_man("systemd-nspawn", "1", &link);
310 if (r < 0)
311 return log_oom();
312
25148653 313 printf("%1$s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
37a92352 314 "%5$sSpawn a command or OS in a light-weight container.%6$s\n\n"
a8828ed9
DW		 315 " -h --help Show this help\n"
316 " --version Print version string\n"
69c79d3c 317 " -q --quiet Do not show status information\n"
bb068de0 318 " --no-pager Do not pipe output into a pager\n"
25148653
LP		 319 " --settings=BOOLEAN Load additional settings from .nspawn file\n\n"
320 "%3$sImage:%4$s\n"
1b9e5b12 321 " -D --directory=PATH Root directory for the container\n"
ec16945e
LP		 322 " --template=PATH Initialize root directory from template directory,\n"
323 " if missing\n"
324 " -x --ephemeral Run container with snapshot of root directory, and\n"
325 " remove it after exit\n"
25e68fd3
LP		 326 " -i --image=PATH Root file system disk image (or device node) for\n"
327 " the container\n"
de40a303 328 " --oci-bundle=PATH OCI bundle directory\n"
25148653
LP		 329 " --read-only Mount the root directory read-only\n"
330 " --volatile[=MODE] Run the system in volatile mode\n"
25e68fd3 331 " --root-hash=HASH Specify verity root hash for root disk image\n"
c2923fdc
LB		 332 " --root-hash-sig=SIG Specify pkcs7 signature of root hash for verity\n"
333 " as a DER encoded PKCS7, either as a path to a file\n"
334 " or as an ASCII base64 encoded string prefixed by\n"
335 " 'base64:'\n"
e7cbe5cb 336 " --verity-data=PATH Specify hash device for verity\n"
25148653
LP		 337 " --pivot-root=PATH[:PATH]\n"
338 " Pivot root to given directory in the container\n\n"
339 "%3$sExecution:%4$s\n"
7732f92b 340 " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n"
a8828ed9 341 " -b --boot Boot up full system (i.e. invoke init)\n"
5f932eb9 342 " --chdir=PATH Set working directory in the container\n"
0d2a0179 343 " -E --setenv=NAME[=VALUE] Pass an environment variable to PID 1\n"
25148653
LP		 344 " -u --user=USER Run the command under specified user or UID\n"
345 " --kill-signal=SIGNAL Select signal to use for shutting down PID 1\n"
4a4654e0
LP		 346 " --notify-ready=BOOLEAN Receive notifications from the child init process\n"
347 " --suppress-sync=BOOLEAN\n"
348 " Suppress any form of disk data synchronization\n\n"
25148653 349 "%3$sSystem Identity:%4$s\n"
a8828ed9 350 " -M --machine=NAME Set the machine name for the container\n"
3a9530e5 351 " --hostname=NAME Override the hostname for the container\n"
25148653
LP		 352 " --uuid=UUID Set a specific machine UUID for the container\n\n"
353 "%3$sProperties:%4$s\n"
a8828ed9 354 " -S --slice=SLICE Place the container in the specified slice\n"
f36933fe 355 " --property=NAME=VALUE Set scope unit property\n"
25148653
LP		 356 " --register=BOOLEAN Register container as machine\n"
357 " --keep-unit Do not register a scope for the machine, reuse\n"
358 " the service unit nspawn is running in\n\n"
359 "%3$sUser Namespacing:%4$s\n"
90b4a64d 360 " -U --private-users=pick Run within user namespace, autoselect UID/GID range\n"
03cfe0d5 361 " --private-users[=UIDBASE[:NUIDS]]\n"
90b4a64d 362 " Similar, but with user configured UID/GID range\n"
6c045a99
LP		 363 " --private-users-ownership=MODE\n"
364 " Adjust ('chown') or map ('map') OS tree ownership\n"
365 " to private UID/GID range\n\n"
25148653 366 "%3$sNetworking:%4$s\n"
69c79d3c
LP		 367 " --private-network Disable network in container\n"
368 " --network-interface=INTERFACE\n"
369 " Assign an existing network interface to the\n"
370 " container\n"
c74e630d
LP		 371 " --network-macvlan=INTERFACE\n"
372 " Create a macvlan network interface based on an\n"
373 " existing network interface to the container\n"
4bbfe7ad 374 " --network-ipvlan=INTERFACE\n"
387f6955 375 " Create an ipvlan network interface based on an\n"
4bbfe7ad 376 " existing network interface to the container\n"
a8eaaee7 377 " -n --network-veth Add a virtual Ethernet connection between host\n"
69c79d3c 378 " and container\n"
f6d6bad1
LP		 379 " --network-veth-extra=HOSTIF[:CONTAINERIF]\n"
380 " Add an additional virtual Ethernet link between\n"
381 " host and container\n"
ab046dde 382 " --network-bridge=INTERFACE\n"
90b4a64d
ZJS		 383 " Add a virtual Ethernet connection to the container\n"
384 " and attach it to an existing bridge on the host\n"
385 " --network-zone=NAME Similar, but attach the new interface to an\n"
386 " an automatically managed bridge interface\n"
d7bea6b6
DP		 387 " --network-namespace-path=PATH\n"
388 " Set network namespace to the one represented by\n"
389 " the specified kernel namespace file node\n"
6d0b55c2 390 " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n"
25148653
LP		 391 " Expose a container IP port on the host\n\n"
392 "%3$sSecurity:%4$s\n"
a8828ed9
DW		 393 " --capability=CAP In addition to the default, retain specified\n"
394 " capability\n"
395 " --drop-capability=CAP Drop the specified capability from the default set\n"
88fc9c9b
TH		 396 " --ambient-capability=CAP\n"
397 " Sets the specified capability for the started\n"
398 " process. Not useful if booting a machine.\n"
f4e803c8 399 " --no-new-privileges Set PR_SET_NO_NEW_PRIVS flag for container payload\n"
960e4569
LP		 400 " --system-call-filter=LIST|~LIST\n"
401 " Permit/prohibit specific system calls\n"
25148653
LP		 402 " -Z --selinux-context=SECLABEL\n"
403 " Set the SELinux security context to be used by\n"
404 " processes in the container\n"
405 " -L --selinux-apifs-context=SECLABEL\n"
406 " Set the SELinux security context to be used by\n"
407 " API/tmpfs file systems in the container\n\n"
408 "%3$sResources:%4$s\n"
bf428efb 409 " --rlimit=NAME=LIMIT Set a resource limit for the payload\n"
81f345df
LP		 410 " --oom-score-adjust=VALUE\n"
411 " Adjust the OOM score value for the payload\n"
f4e803c8
LP		 412 " --cpu-affinity=CPUS Adjust the CPU affinity of the container\n"
413 " --personality=ARCH Pick personality for this container\n\n"
25148653 414 "%3$sIntegration:%4$s\n"
09d423e9 415 " --resolv-conf=MODE Select mode of /etc/resolv.conf initialization\n"
1688841f 416 " --timezone=MODE Select mode of /etc/localtime initialization\n"
25148653
LP		 417 " --link-journal=MODE Link up guest journal, one of no, auto, guest, \n"
418 " host, try-guest, try-host\n"
419 " -j Equivalent to --link-journal=try-guest\n\n"
420 "%3$sMounts:%4$s\n"
5e5bfa6e
EY		 421 " --bind=PATH[:PATH[:OPTIONS]]\n"
422 " Bind mount a file or directory from the host into\n"
a8828ed9 423 " the container\n"
5e5bfa6e
EY		 424 " --bind-ro=PATH[:PATH[:OPTIONS]\n"
425 " Similar, but creates a read-only bind mount\n"
de40a303
LP		 426 " --inaccessible=PATH Over-mount file node with inaccessible node to mask\n"
427 " it\n"
06c17c39 428 " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
5a8af538
LP		 429 " --overlay=PATH[:PATH...]:PATH\n"
430 " Create an overlay mount from the host to \n"
431 " the container\n"
432 " --overlay-ro=PATH[:PATH...]:PATH\n"
2f893044
LP		 433 " Similar, but creates a read-only overlay mount\n"
434 " --bind-user=NAME Bind user from host to container\n\n"
25148653 435 "%3$sInput/Output:%4$s\n"
de40a303
LP		 436 " --console=MODE Select how stdin/stdout/stderr and /dev/console are\n"
437 " set up for the container.\n"
3652872a
LP		 438 " -P --pipe Equivalent to --console=pipe\n\n"
439 "%3$sCredentials:%4$s\n"
440 " --set-credential=ID:VALUE\n"
441 " Pass a credential with literal value to container.\n"
442 " --load-credential=ID:PATH\n"
443 " Load credential to pass to container from file or\n"
444 " AF_UNIX stream socket.\n"
bc556335
DDM		 445 "\nSee the %2$s for details.\n",
446 program_invocation_short_name,
447 link,
448 ansi_underline(),
449 ansi_normal(),
450 ansi_highlight(),
451 ansi_normal());
37ec0fdd
LP		 452
453 return 0;
88213476
LP		 454}
455
86c0dd4a 456static int custom_mount_check_all(void) {
88614c8a 457 size_t i;
5a8af538 458
5a8af538
LP		 459 for (i = 0; i < arg_n_custom_mounts; i++) {
460 CustomMount *m = &arg_custom_mounts[i];
461
0de7acce 462 if (path_equal(m->destination, "/") && arg_userns_mode != USER_NAMESPACE_NO) {
6c045a99 463 if (arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_OFF)
baaa35ad 464 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
c920b863 465 "--private-users-ownership=own may not be combined with custom root mounts.");
6c045a99 466 if (arg_uid_shift == UID_INVALID)
baaa35ad
ZJS		 467 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
468 "--private-users with automatic UID shift may not be combined with custom root mounts.");
825d5287 469 }
5a8af538
LP		 470 }
471
472 return 0;
473}
474
8199d554 475static int detect_unified_cgroup_hierarchy_from_environment(void) {
c78c095b 476 const char *e, *var = "SYSTEMD_NSPAWN_UNIFIED_HIERARCHY";
415fc41c 477 int r;
5da38d07 478
efdb0237 479 /* Allow the user to control whether the unified hierarchy is used */
c78c095b
ZJS		 480
481 e = getenv(var);
482 if (!e) {
d5fc5b2f 483 /* $UNIFIED_CGROUP_HIERARCHY has been renamed to $SYSTEMD_NSPAWN_UNIFIED_HIERARCHY. */
c78c095b
ZJS		 484 var = "UNIFIED_CGROUP_HIERARCHY";
485 e = getenv(var);
c78c095b
ZJS		 486 }
487
488 if (!isempty(e)) {
efdb0237
LP		 489 r = parse_boolean(e);
490 if (r < 0)
c78c095b 491 return log_error_errno(r, "Failed to parse $%s: %m", var);
5da38d07
TH		 492 if (r > 0)
493 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
494 else
495 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237
LP		 496 }
497
8199d554
LP		 498 return 0;
499}
500
501static int detect_unified_cgroup_hierarchy_from_image(const char *directory) {
502 int r;
503
75b0d8b8
ZJS		 504 /* Let's inherit the mode to use from the host system, but let's take into consideration what systemd
505 * in the image actually supports. */
b4cccbc1
LP		 506 r = cg_all_unified();
507 if (r < 0)
508 return log_error_errno(r, "Failed to determine whether we are in all unified mode.");
509 if (r > 0) {
a8725a06
ZJS		 510 /* Unified cgroup hierarchy support was added in 230. Unfortunately the detection
511 * routine only detects 231, so we'll have a false negative here for 230. */
512 r = systemd_installation_has_version(directory, 230);
513 if (r < 0)
514 return log_error_errno(r, "Failed to determine systemd version in container: %m");
515 if (r > 0)
516 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
517 else
518 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
c22800e4 519 } else if (cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER) > 0) {
2977724b
TH		 520 /* Mixed cgroup hierarchy support was added in 233 */
521 r = systemd_installation_has_version(directory, 233);
0fd9563f
ZJS		 522 if (r < 0)
523 return log_error_errno(r, "Failed to determine systemd version in container: %m");
524 if (r > 0)
525 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD;
526 else
527 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
528 } else
5da38d07 529 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237 530
8199d554
LP		 531 log_debug("Using %s hierarchy for container.",
532 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_NONE ? "legacy" :
533 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_SYSTEMD ? "hybrid" : "unified");
534
efdb0237
LP		 535 return 0;
536}
537
8a99bd0c
ZJS		 538static int parse_capability_spec(const char *spec, uint64_t *ret_mask) {
539 uint64_t mask = 0;
540 int r;
541
542 for (;;) {
543 _cleanup_free_ char *t = NULL;
544
545 r = extract_first_word(&spec, &t, ",", 0);
546 if (r < 0)
547 return log_error_errno(r, "Failed to parse capability %s.", t);
548 if (r == 0)
549 break;
550
551 if (streq(t, "help")) {
552 for (int i = 0; i < capability_list_length(); i++) {
553 const char *name;
554
555 name = capability_to_name(i);
556 if (name)
557 puts(name);
558 }
559
560 return 0; /* quit */
561 }
562
563 if (streq(t, "all"))
f5fbe71d 564 mask = UINT64_MAX;
8a99bd0c
ZJS		 565 else {
566 r = capability_from_name(t);
567 if (r < 0)
568 return log_error_errno(r, "Failed to parse capability %s.", t);
569
570 mask |= 1ULL << r;
571 }
572 }
573
574 *ret_mask = mask;
575 return 1; /* continue */
576}
577
49048684 578static int parse_share_ns_env(const char *name, unsigned long ns_flag) {
0c582db0
LB		 579 int r;
580
581 r = getenv_bool(name);
582 if (r == -ENXIO)
49048684 583 return 0;
0c582db0 584 if (r < 0)
49048684 585 return log_error_errno(r, "Failed to parse $%s: %m", name);
de40a303 586
0c582db0 587 arg_clone_ns_flags = (arg_clone_ns_flags & ~ns_flag) | (r > 0 ? 0 : ns_flag);
de40a303 588 arg_settings_mask |= SETTING_CLONE_NS_FLAGS;
49048684 589 return 0;
0c582db0
LB		 590}
591
49048684 592static int parse_mount_settings_env(void) {
4f086aab 593 const char *e;
1099ceeb
LP		 594 int r;
595
596 r = getenv_bool("SYSTEMD_NSPAWN_TMPFS_TMP");
49048684
ZJS		 597 if (r < 0 && r != -ENXIO)
598 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_TMPFS_TMP: %m");
1099ceeb
LP		 599 if (r >= 0)
600 SET_FLAG(arg_mount_settings, MOUNT_APPLY_TMPFS_TMP, r > 0);
4f086aab
SU		 601
602 e = getenv("SYSTEMD_NSPAWN_API_VFS_WRITABLE");
49048684 603 if (streq_ptr(e, "network"))
4f086aab 604 arg_mount_settings |= MOUNT_APPLY_APIVFS_RO|MOUNT_APPLY_APIVFS_NETNS;
4f086aab 605
49048684
ZJS		 606 else if (e) {
607 r = parse_boolean(e);
608 if (r < 0)
609 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_API_VFS_WRITABLE: %m");
610
611 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_RO, r == 0);
612 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_NETNS, false);
ab8ee0f2 613 }
4f086aab 614
49048684 615 return 0;
4f086aab
SU		 616}
617
49048684 618static int parse_environment(void) {
d5455d2f
LP		 619 const char *e;
620 int r;
621
49048684
ZJS		 622 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_IPC", CLONE_NEWIPC);
623 if (r < 0)
624 return r;
625 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_PID", CLONE_NEWPID);
626 if (r < 0)
627 return r;
628 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_UTS", CLONE_NEWUTS);
629 if (r < 0)
630 return r;
631 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_SYSTEM", CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS);
632 if (r < 0)
633 return r;
d5455d2f 634
49048684
ZJS		 635 r = parse_mount_settings_env();
636 if (r < 0)
637 return r;
d5455d2f 638
489fae52
ZJS		 639 /* SYSTEMD_NSPAWN_USE_CGNS=0 can be used to disable CLONE_NEWCGROUP use,
640 * even if it is supported. If not supported, it has no effect. */
de40a303 641 if (!cg_ns_supported())
489fae52 642 arg_use_cgns = false;
de40a303
LP		 643 else {
644 r = getenv_bool("SYSTEMD_NSPAWN_USE_CGNS");
645 if (r < 0) {
646 if (r != -ENXIO)
49048684 647 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_USE_CGNS: %m");
de40a303
LP		 648
649 arg_use_cgns = true;
650 } else {
651 arg_use_cgns = r > 0;
652 arg_settings_mask |= SETTING_USE_CGNS;
653 }
654 }
d5455d2f
LP		 655
656 e = getenv("SYSTEMD_NSPAWN_CONTAINER_SERVICE");
657 if (e)
658 arg_container_service_name = e;
659
4a4654e0
LP		 660 r = getenv_bool("SYSTEMD_SUPPRESS_SYNC");
661 if (r >= 0)
662 arg_suppress_sync = r;
663 else if (r != -ENXIO)
664 log_debug_errno(r, "Failed to parse $SYSTEMD_SUPPRESS_SYNC, ignoring: %m");
665
49048684 666 return detect_unified_cgroup_hierarchy_from_environment();
d5455d2f
LP		 667}
668
88213476 669static int parse_argv(int argc, char *argv[]) {
a41fe3a2 670 enum {
acbeb427
ZJS		 671 ARG_VERSION = 0x100,
672 ARG_PRIVATE_NETWORK,
bc2f673e 673 ARG_UUID,
5076f0cc 674 ARG_READ_ONLY,
57fb9fb5 675 ARG_CAPABILITY,
88fc9c9b 676 ARG_AMBIENT_CAPABILITY,
420c7379 677 ARG_DROP_CAPABILITY,
17fe0523
LP		 678 ARG_LINK_JOURNAL,
679 ARG_BIND,
f4889f65 680 ARG_BIND_RO,
06c17c39 681 ARG_TMPFS,
5a8af538
LP		 682 ARG_OVERLAY,
683 ARG_OVERLAY_RO,
de40a303 684 ARG_INACCESSIBLE,
eb91eb18 685 ARG_SHARE_SYSTEM,
89f7c846 686 ARG_REGISTER,
aa28aefe 687 ARG_KEEP_UNIT,
69c79d3c 688 ARG_NETWORK_INTERFACE,
c74e630d 689 ARG_NETWORK_MACVLAN,
4bbfe7ad 690 ARG_NETWORK_IPVLAN,
ab046dde 691 ARG_NETWORK_BRIDGE,
22b28dfd 692 ARG_NETWORK_ZONE,
f6d6bad1 693 ARG_NETWORK_VETH_EXTRA,
d7bea6b6 694 ARG_NETWORK_NAMESPACE_PATH,
6afc95b7 695 ARG_PERSONALITY,
4d9f07b4 696 ARG_VOLATILE,
ec16945e 697 ARG_TEMPLATE,
f36933fe 698 ARG_PROPERTY,
6dac160c 699 ARG_PRIVATE_USERS,
c6c8f6e2 700 ARG_KILL_SIGNAL,
f757855e 701 ARG_SETTINGS,
5f932eb9 702 ARG_CHDIR,
b53ede69 703 ARG_PIVOT_ROOT,
7336138e 704 ARG_PRIVATE_USERS_CHOWN,
6c045a99 705 ARG_PRIVATE_USERS_OWNERSHIP,
9c1e04d0 706 ARG_NOTIFY_READY,
4623e8e6 707 ARG_ROOT_HASH,
89e62e0b
LP		 708 ARG_ROOT_HASH_SIG,
709 ARG_VERITY_DATA,
960e4569 710 ARG_SYSTEM_CALL_FILTER,
bf428efb 711 ARG_RLIMIT,
3a9530e5 712 ARG_HOSTNAME,
66edd963 713 ARG_NO_NEW_PRIVILEGES,
81f345df 714 ARG_OOM_SCORE_ADJUST,
d107bb7d 715 ARG_CPU_AFFINITY,
09d423e9 716 ARG_RESOLV_CONF,
1688841f 717 ARG_TIMEZONE,
de40a303
LP		 718 ARG_CONSOLE,
719 ARG_PIPE,
720 ARG_OCI_BUNDLE,
bb068de0 721 ARG_NO_PAGER,
3652872a
LP		 722 ARG_SET_CREDENTIAL,
723 ARG_LOAD_CREDENTIAL,
2f893044 724 ARG_BIND_USER,
4a4654e0 725 ARG_SUPPRESS_SYNC,
a41fe3a2
LP		 726 };
727
88213476 728 static const struct option options[] = {
d7bea6b6
DP		 729 { "help", no_argument, NULL, 'h' },
730 { "version", no_argument, NULL, ARG_VERSION },
731 { "directory", required_argument, NULL, 'D' },
732 { "template", required_argument, NULL, ARG_TEMPLATE },
733 { "ephemeral", no_argument, NULL, 'x' },
734 { "user", required_argument, NULL, 'u' },
735 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
736 { "as-pid2", no_argument, NULL, 'a' },
737 { "boot", no_argument, NULL, 'b' },
738 { "uuid", required_argument, NULL, ARG_UUID },
739 { "read-only", no_argument, NULL, ARG_READ_ONLY },
740 { "capability", required_argument, NULL, ARG_CAPABILITY },
88fc9c9b 741 { "ambient-capability", required_argument, NULL, ARG_AMBIENT_CAPABILITY },
d7bea6b6 742 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
66edd963 743 { "no-new-privileges", required_argument, NULL, ARG_NO_NEW_PRIVILEGES },
d7bea6b6
DP		 744 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
745 { "bind", required_argument, NULL, ARG_BIND },
746 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
747 { "tmpfs", required_argument, NULL, ARG_TMPFS },
748 { "overlay", required_argument, NULL, ARG_OVERLAY },
749 { "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO },
de40a303 750 { "inaccessible", required_argument, NULL, ARG_INACCESSIBLE },
d7bea6b6 751 { "machine", required_argument, NULL, 'M' },
3a9530e5 752 { "hostname", required_argument, NULL, ARG_HOSTNAME },
d7bea6b6
DP		 753 { "slice", required_argument, NULL, 'S' },
754 { "setenv", required_argument, NULL, 'E' },
755 { "selinux-context", required_argument, NULL, 'Z' },
756 { "selinux-apifs-context", required_argument, NULL, 'L' },
757 { "quiet", no_argument, NULL, 'q' },
758 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, /* not documented */
759 { "register", required_argument, NULL, ARG_REGISTER },
760 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
761 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
762 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
763 { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN },
764 { "network-veth", no_argument, NULL, 'n' },
765 { "network-veth-extra", required_argument, NULL, ARG_NETWORK_VETH_EXTRA },
766 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
767 { "network-zone", required_argument, NULL, ARG_NETWORK_ZONE },
768 { "network-namespace-path", required_argument, NULL, ARG_NETWORK_NAMESPACE_PATH },
769 { "personality", required_argument, NULL, ARG_PERSONALITY },
770 { "image", required_argument, NULL, 'i' },
771 { "volatile", optional_argument, NULL, ARG_VOLATILE },
772 { "port", required_argument, NULL, 'p' },
773 { "property", required_argument, NULL, ARG_PROPERTY },
774 { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS },
6c045a99
LP		 775 { "private-users-chown", optional_argument, NULL, ARG_PRIVATE_USERS_CHOWN }, /* obsolete */
776 { "private-users-ownership",required_argument, NULL, ARG_PRIVATE_USERS_OWNERSHIP},
d7bea6b6
DP		 777 { "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL },
778 { "settings", required_argument, NULL, ARG_SETTINGS },
779 { "chdir", required_argument, NULL, ARG_CHDIR },
780 { "pivot-root", required_argument, NULL, ARG_PIVOT_ROOT },
781 { "notify-ready", required_argument, NULL, ARG_NOTIFY_READY },
782 { "root-hash", required_argument, NULL, ARG_ROOT_HASH },
89e62e0b
LP		 783 { "root-hash-sig", required_argument, NULL, ARG_ROOT_HASH_SIG },
784 { "verity-data", required_argument, NULL, ARG_VERITY_DATA },
d7bea6b6 785 { "system-call-filter", required_argument, NULL, ARG_SYSTEM_CALL_FILTER },
bf428efb 786 { "rlimit", required_argument, NULL, ARG_RLIMIT },
81f345df 787 { "oom-score-adjust", required_argument, NULL, ARG_OOM_SCORE_ADJUST },
d107bb7d 788 { "cpu-affinity", required_argument, NULL, ARG_CPU_AFFINITY },
09d423e9 789 { "resolv-conf", required_argument, NULL, ARG_RESOLV_CONF },
1688841f 790 { "timezone", required_argument, NULL, ARG_TIMEZONE },
de40a303
LP		 791 { "console", required_argument, NULL, ARG_CONSOLE },
792 { "pipe", no_argument, NULL, ARG_PIPE },
793 { "oci-bundle", required_argument, NULL, ARG_OCI_BUNDLE },
bb068de0 794 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
3652872a
LP		 795 { "set-credential", required_argument, NULL, ARG_SET_CREDENTIAL },
796 { "load-credential", required_argument, NULL, ARG_LOAD_CREDENTIAL },
2f893044 797 { "bind-user", required_argument, NULL, ARG_BIND_USER },
4a4654e0 798 { "suppress-sync", required_argument, NULL, ARG_SUPPRESS_SYNC },
eb9da376 799 {}
88213476
LP		 800 };
801
9444b1f2 802 int c, r;
a42c8b54 803 uint64_t plus = 0, minus = 0;
f757855e 804 bool mask_all_settings = false, mask_no_settings = false;
88213476
LP		 805
806 assert(argc >= 0);
807 assert(argv);
808
de40a303 809 while ((c = getopt_long(argc, argv, "+hD:u:abL:M:jS:Z:qi:xp:nUE:P", options, NULL)) >= 0)
88213476
LP		 810 switch (c) {
811
812 case 'h':
37ec0fdd 813 return help();
88213476 814
acbeb427 815 case ARG_VERSION:
3f6fd1ba 816 return version();
acbeb427 817
88213476 818 case 'D':
614b022c 819 r = parse_path_argument(optarg, false, &arg_directory);
ec16945e 820 if (r < 0)
0f03c2a4 821 return r;
de40a303
LP		 822
823 arg_settings_mask |= SETTING_DIRECTORY;
ec16945e
LP		 824 break;
825
826 case ARG_TEMPLATE:
614b022c 827 r = parse_path_argument(optarg, false, &arg_template);
ec16945e 828 if (r < 0)
0f03c2a4 829 return r;
de40a303
LP		 830
831 arg_settings_mask |= SETTING_DIRECTORY;
88213476
LP		 832 break;
833
1b9e5b12 834 case 'i':
614b022c 835 r = parse_path_argument(optarg, false, &arg_image);
ec16945e 836 if (r < 0)
0f03c2a4 837 return r;
de40a303
LP		 838
839 arg_settings_mask |= SETTING_DIRECTORY;
840 break;
841
842 case ARG_OCI_BUNDLE:
614b022c 843 r = parse_path_argument(optarg, false, &arg_oci_bundle);
de40a303
LP		 844 if (r < 0)
845 return r;
846
ec16945e
LP		 847 break;
848
849 case 'x':
850 arg_ephemeral = true;
a2f577fc 851 arg_settings_mask |= SETTING_EPHEMERAL;
1b9e5b12
LP		 852 break;
853
687d0825 854 case 'u':
2fc09a9c
DM		 855 r = free_and_strdup(&arg_user, optarg);
856 if (r < 0)
7027ff61 857 return log_oom();
687d0825 858
f757855e 859 arg_settings_mask |= SETTING_USER;
687d0825
MV		 860 break;
861
22b28dfd
LP		 862 case ARG_NETWORK_ZONE: {
863 char *j;
864
b910cc72 865 j = strjoin("vz-", optarg);
22b28dfd
LP		 866 if (!j)
867 return log_oom();
868
869 if (!ifname_valid(j)) {
870 log_error("Network zone name not valid: %s", j);
871 free(j);
872 return -EINVAL;
873 }
874
df1fac6d 875 free_and_replace(arg_network_zone, j);
22b28dfd
LP		 876
877 arg_network_veth = true;
878 arg_private_network = true;
879 arg_settings_mask |= SETTING_NETWORK;
880 break;
881 }
882
ab046dde 883 case ARG_NETWORK_BRIDGE:
ef76dff2 884
baaa35ad
ZJS		 885 if (!ifname_valid(optarg))
886 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
887 "Bridge interface name not valid: %s", optarg);
ef76dff2 888
f757855e
LP		 889 r = free_and_strdup(&arg_network_bridge, optarg);
890 if (r < 0)
891 return log_oom();
ab046dde 892
4831981d 893 _fallthrough_;
0dfaa006 894 case 'n':
69c79d3c
LP		 895 arg_network_veth = true;
896 arg_private_network = true;
f757855e 897 arg_settings_mask |= SETTING_NETWORK;
69c79d3c
LP		 898 break;
899
f6d6bad1
LP		 900 case ARG_NETWORK_VETH_EXTRA:
901 r = veth_extra_parse(&arg_network_veth_extra, optarg);
902 if (r < 0)
903 return log_error_errno(r, "Failed to parse --network-veth-extra= parameter: %s", optarg);
904
905 arg_private_network = true;
906 arg_settings_mask |= SETTING_NETWORK;
907 break;
908
aa28aefe 909 case ARG_NETWORK_INTERFACE:
baaa35ad
ZJS		 910 if (!ifname_valid(optarg))
911 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
912 "Network interface name not valid: %s", optarg);
ef76dff2 913
b390f178
DDM		 914 r = test_network_interface_initialized(optarg);
915 if (r < 0)
916 return r;
917
c74e630d
LP		 918 if (strv_extend(&arg_network_interfaces, optarg) < 0)
919 return log_oom();
920
921 arg_private_network = true;
f757855e 922 arg_settings_mask |= SETTING_NETWORK;
c74e630d
LP		 923 break;
924
925 case ARG_NETWORK_MACVLAN:
ef76dff2 926
baaa35ad
ZJS		 927 if (!ifname_valid(optarg))
928 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
929 "MACVLAN network interface name not valid: %s", optarg);
ef76dff2 930
b390f178
DDM		 931 r = test_network_interface_initialized(optarg);
932 if (r < 0)
933 return r;
934
c74e630d 935 if (strv_extend(&arg_network_macvlan, optarg) < 0)
aa28aefe
LP		 936 return log_oom();
937
4bbfe7ad 938 arg_private_network = true;
f757855e 939 arg_settings_mask |= SETTING_NETWORK;
4bbfe7ad
TG		 940 break;
941
942 case ARG_NETWORK_IPVLAN:
ef76dff2 943
baaa35ad
ZJS		 944 if (!ifname_valid(optarg))
945 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
946 "IPVLAN network interface name not valid: %s", optarg);
ef76dff2 947
b390f178
DDM		 948 r = test_network_interface_initialized(optarg);
949 if (r < 0)
950 return r;
951
4bbfe7ad
TG		 952 if (strv_extend(&arg_network_ipvlan, optarg) < 0)
953 return log_oom();
954
4831981d 955 _fallthrough_;
ff01d048
LP		 956 case ARG_PRIVATE_NETWORK:
957 arg_private_network = true;
f757855e 958 arg_settings_mask |= SETTING_NETWORK;
a41fe3a2
LP		 959 break;
960
d7bea6b6 961 case ARG_NETWORK_NAMESPACE_PATH:
614b022c 962 r = parse_path_argument(optarg, false, &arg_network_namespace_path);
d7bea6b6
DP		 963 if (r < 0)
964 return r;
965
de40a303 966 arg_settings_mask |= SETTING_NETWORK;
d7bea6b6
DP		 967 break;
968
0f0dbc46 969 case 'b':
baaa35ad
ZJS		 970 if (arg_start_mode == START_PID2)
971 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
972 "--boot and --as-pid2 may not be combined.");
7732f92b
LP		 973
974 arg_start_mode = START_BOOT;
975 arg_settings_mask |= SETTING_START_MODE;
976 break;
977
978 case 'a':
baaa35ad
ZJS		 979 if (arg_start_mode == START_BOOT)
980 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
981 "--boot and --as-pid2 may not be combined.");
7732f92b
LP		 982
983 arg_start_mode = START_PID2;
984 arg_settings_mask |= SETTING_START_MODE;
0f0dbc46
LP		 985 break;
986
144f0fc0 987 case ARG_UUID:
9444b1f2 988 r = sd_id128_from_string(optarg, &arg_uuid);
317feb4d
LP		 989 if (r < 0)
990 return log_error_errno(r, "Invalid UUID: %s", optarg);
991
baaa35ad
ZJS		 992 if (sd_id128_is_null(arg_uuid))
993 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
994 "Machine UUID may not be all zeroes.");
f757855e
LP		 995
996 arg_settings_mask |= SETTING_MACHINE_ID;
9444b1f2 997 break;
aa96c6cb 998
43c3fb46
LP		 999 case 'S': {
1000 _cleanup_free_ char *mangled = NULL;
1001
1002 r = unit_name_mangle_with_suffix(optarg, NULL, UNIT_NAME_MANGLE_WARN, ".slice", &mangled);
de40a303
LP		 1003 if (r < 0)
1004 return log_oom();
1005
43c3fb46 1006 free_and_replace(arg_slice, mangled);
de40a303 1007 arg_settings_mask |= SETTING_SLICE;
144f0fc0 1008 break;
43c3fb46 1009 }
144f0fc0 1010
7027ff61 1011 case 'M':
c1521918 1012 if (isempty(optarg))
97b11eed 1013 arg_machine = mfree(arg_machine);
c1521918 1014 else {
52ef5dd7 1015 if (!hostname_is_valid(optarg, 0))
baaa35ad
ZJS		 1016 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1017 "Invalid machine name: %s", optarg);
7027ff61 1018
0c3c4284
LP		 1019 r = free_and_strdup(&arg_machine, optarg);
1020 if (r < 0)
eb91eb18 1021 return log_oom();
eb91eb18 1022 }
9ce6d1b3 1023 break;
7027ff61 1024
3a9530e5
LP		 1025 case ARG_HOSTNAME:
1026 if (isempty(optarg))
1027 arg_hostname = mfree(arg_hostname);
1028 else {
52ef5dd7 1029 if (!hostname_is_valid(optarg, 0))
baaa35ad
ZJS		 1030 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1031 "Invalid hostname: %s", optarg);
3a9530e5
LP		 1032
1033 r = free_and_strdup(&arg_hostname, optarg);
1034 if (r < 0)
1035 return log_oom();
1036 }
1037
1038 arg_settings_mask |= SETTING_HOSTNAME;
1039 break;
1040
82adf6af
LP		 1041 case 'Z':
1042 arg_selinux_context = optarg;
a8828ed9
DW		 1043 break;
1044
82adf6af
LP		 1045 case 'L':
1046 arg_selinux_apifs_context = optarg;
a8828ed9
DW		 1047 break;
1048
bc2f673e
LP		 1049 case ARG_READ_ONLY:
1050 arg_read_only = true;
f757855e 1051 arg_settings_mask |= SETTING_READ_ONLY;
bc2f673e
LP		 1052 break;
1053
88fc9c9b
TH		 1054 case ARG_AMBIENT_CAPABILITY: {
1055 uint64_t m;
1056 r = parse_capability_spec(optarg, &m);
1057 if (r <= 0)
1058 return r;
1059 arg_caps_ambient |= m;
1060 arg_settings_mask |= SETTING_CAPABILITY;
1061 break;
1062 }
420c7379
LP		 1063 case ARG_CAPABILITY:
1064 case ARG_DROP_CAPABILITY: {
8a99bd0c
ZJS		 1065 uint64_t m;
1066 r = parse_capability_spec(optarg, &m);
1067 if (r <= 0)
1068 return r;
5076f0cc 1069
8a99bd0c
ZJS		 1070 if (c == ARG_CAPABILITY)
1071 plus |= m;
1072 else
1073 minus |= m;
f757855e 1074 arg_settings_mask |= SETTING_CAPABILITY;
5076f0cc
LP		 1075 break;
1076 }
66edd963
LP		 1077 case ARG_NO_NEW_PRIVILEGES:
1078 r = parse_boolean(optarg);
1079 if (r < 0)
1080 return log_error_errno(r, "Failed to parse --no-new-privileges= argument: %s", optarg);
1081
1082 arg_no_new_privileges = r;
1083 arg_settings_mask |= SETTING_NO_NEW_PRIVILEGES;
1084 break;
1085
57fb9fb5
LP		 1086 case 'j':
1087 arg_link_journal = LINK_GUEST;
574edc90 1088 arg_link_journal_try = true;
4e1d6aa9 1089 arg_settings_mask |= SETTING_LINK_JOURNAL;
57fb9fb5
LP		 1090 break;
1091
1092 case ARG_LINK_JOURNAL:
4e1d6aa9 1093 r = parse_link_journal(optarg, &arg_link_journal, &arg_link_journal_try);
c6147113
LP		 1094 if (r < 0)
1095 return log_error_errno(r, "Failed to parse link journal mode %s", optarg);
57fb9fb5 1096
4e1d6aa9 1097 arg_settings_mask |= SETTING_LINK_JOURNAL;
57fb9fb5
LP		 1098 break;
1099
17fe0523 1100 case ARG_BIND:
f757855e
LP		 1101 case ARG_BIND_RO:
1102 r = bind_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg, c == ARG_BIND_RO);
1103 if (r < 0)
1104 return log_error_errno(r, "Failed to parse --bind(-ro)= argument %s: %m", optarg);
17fe0523 1105
f757855e 1106 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
17fe0523 1107 break;
06c17c39 1108
f757855e
LP		 1109 case ARG_TMPFS:
1110 r = tmpfs_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg);
1111 if (r < 0)
1112 return log_error_errno(r, "Failed to parse --tmpfs= argument %s: %m", optarg);
5a8af538 1113
f757855e 1114 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
5a8af538 1115 break;
5a8af538
LP		 1116
1117 case ARG_OVERLAY:
ad85779a
LP		 1118 case ARG_OVERLAY_RO:
1119 r = overlay_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg, c == ARG_OVERLAY_RO);
1120 if (r == -EADDRNOTAVAIL)
1121 return log_error_errno(r, "--overlay(-ro)= needs at least two colon-separated directories specified.");
1122 if (r < 0)
1123 return log_error_errno(r, "Failed to parse --overlay(-ro)= argument %s: %m", optarg);
06c17c39 1124
f757855e 1125 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
06c17c39 1126 break;
06c17c39 1127
de40a303
LP		 1128 case ARG_INACCESSIBLE:
1129 r = inaccessible_mount_parse(&arg_custom_mounts, &arg_n_custom_mounts, optarg);
1130 if (r < 0)
1131 return log_error_errno(r, "Failed to parse --inaccessible= argument %s: %m", optarg);
1132
1133 arg_settings_mask |= SETTING_CUSTOM_MOUNTS;
1134 break;
1135
0d2a0179
ZJS		 1136 case 'E':
1137 r = strv_env_replace_strdup_passthrough(&arg_setenv, optarg);
aaf057c4 1138 if (r < 0)
0d2a0179 1139 return log_error_errno(r, "Cannot assign environment variable %s: %m", optarg);
f4889f65 1140
f757855e 1141 arg_settings_mask |= SETTING_ENVIRONMENT;
f4889f65 1142 break;
f4889f65 1143
284c0b91
LP		 1144 case 'q':
1145 arg_quiet = true;
1146 break;
1147
8a96d94e 1148 case ARG_SHARE_SYSTEM:
a6b5216c 1149 /* We don't officially support this anymore, except for compat reasons. People should use the
0c582db0 1150 * $SYSTEMD_NSPAWN_SHARE_* environment variables instead. */
63d1c29f 1151 log_warning("Please do not use --share-system anymore, use $SYSTEMD_NSPAWN_SHARE_* instead.");
0c582db0 1152 arg_clone_ns_flags = 0;
8a96d94e
LP		 1153 break;
1154
eb91eb18
LP		 1155 case ARG_REGISTER:
1156 r = parse_boolean(optarg);
1157 if (r < 0) {
1158 log_error("Failed to parse --register= argument: %s", optarg);
1159 return r;
1160 }
1161
1162 arg_register = r;
1163 break;
1164
89f7c846
LP		 1165 case ARG_KEEP_UNIT:
1166 arg_keep_unit = true;
1167 break;
1168
6afc95b7
LP		 1169 case ARG_PERSONALITY:
1170
ac45f971 1171 arg_personality = personality_from_string(optarg);
baaa35ad
ZJS		 1172 if (arg_personality == PERSONALITY_INVALID)
1173 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1174 "Unknown or unsupported personality '%s'.", optarg);
6afc95b7 1175
f757855e 1176 arg_settings_mask |= SETTING_PERSONALITY;
6afc95b7
LP		 1177 break;
1178
4d9f07b4
LP		 1179 case ARG_VOLATILE:
1180
1181 if (!optarg)
f757855e 1182 arg_volatile_mode = VOLATILE_YES;
5c828e66
LP		 1183 else if (streq(optarg, "help")) {
1184 DUMP_STRING_TABLE(volatile_mode, VolatileMode, _VOLATILE_MODE_MAX);
1185 return 0;
1186 } else {
f757855e 1187 VolatileMode m;
4d9f07b4 1188
f757855e 1189 m = volatile_mode_from_string(optarg);
baaa35ad
ZJS		 1190 if (m < 0)
1191 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1192 "Failed to parse --volatile= argument: %s", optarg);
1193 else
f757855e 1194 arg_volatile_mode = m;
6d0b55c2
LP		 1195 }
1196
f757855e
LP		 1197 arg_settings_mask |= SETTING_VOLATILE_MODE;
1198 break;
6d0b55c2 1199
f757855e
LP		 1200 case 'p':
1201 r = expose_port_parse(&arg_expose_ports, optarg);
1202 if (r == -EEXIST)
1203 return log_error_errno(r, "Duplicate port specification: %s", optarg);
1204 if (r < 0)
1205 return log_error_errno(r, "Failed to parse host port %s: %m", optarg);
6d0b55c2 1206
f757855e 1207 arg_settings_mask |= SETTING_EXPOSE_PORTS;
6d0b55c2 1208 break;
6d0b55c2 1209
f36933fe
LP		 1210 case ARG_PROPERTY:
1211 if (strv_extend(&arg_property, optarg) < 0)
1212 return log_oom();
1213
1214 break;
1215
ae209204 1216 case ARG_PRIVATE_USERS: {
33eac552 1217 int boolean;
0de7acce 1218
ae209204
ZJS		 1219 if (!optarg)
1220 boolean = true;
1221 else if (!in_charset(optarg, DIGITS))
1222 /* do *not* parse numbers as booleans */
1223 boolean = parse_boolean(optarg);
33eac552
LP		 1224 else
1225 boolean = -1;
ae209204 1226
33eac552 1227 if (boolean == 0) {
0de7acce
LP		 1228 /* no: User namespacing off */
1229 arg_userns_mode = USER_NAMESPACE_NO;
1230 arg_uid_shift = UID_INVALID;
1231 arg_uid_range = UINT32_C(0x10000);
33eac552 1232 } else if (boolean > 0) {
0de7acce
LP		 1233 /* yes: User namespacing on, UID range is read from root dir */
1234 arg_userns_mode = USER_NAMESPACE_FIXED;
1235 arg_uid_shift = UID_INVALID;
1236 arg_uid_range = UINT32_C(0x10000);
1237 } else if (streq(optarg, "pick")) {
1238 /* pick: User namespacing on, UID range is picked randomly */
6c045a99
LP		 1239 arg_userns_mode = USER_NAMESPACE_PICK; /* Note that arg_userns_ownership is
1240 * implied by USER_NAMESPACE_PICK
33eac552 1241 * further down. */
0de7acce
LP		 1242 arg_uid_shift = UID_INVALID;
1243 arg_uid_range = UINT32_C(0x10000);
33eac552
LP		 1244
1245 } else if (streq(optarg, "identity")) {
1246 /* identitiy: User namespaces on, UID range is map the 0…0xFFFF range to
1247 * itself, i.e. we don't actually map anything, but do take benefit of
1248 * isolation of capability sets. */
1249 arg_userns_mode = USER_NAMESPACE_FIXED;
1250 arg_uid_shift = 0;
1251 arg_uid_range = UINT32_C(0x10000);
0de7acce 1252 } else {
6c2058b3 1253 _cleanup_free_ char *buffer = NULL;
6dac160c
LP		 1254 const char *range, *shift;
1255
0de7acce
LP		 1256 /* anything else: User namespacing on, UID range is explicitly configured */
1257
6dac160c
LP		 1258 range = strchr(optarg, ':');
1259 if (range) {
6c2058b3
ZJS		 1260 buffer = strndup(optarg, range - optarg);
1261 if (!buffer)
1262 return log_oom();
1263 shift = buffer;
6dac160c
LP		 1264
1265 range++;
bfd292ec
ZJS		 1266 r = safe_atou32(range, &arg_uid_range);
1267 if (r < 0)
be715731 1268 return log_error_errno(r, "Failed to parse UID range \"%s\": %m", range);
6dac160c
LP		 1269 } else
1270 shift = optarg;
1271
be715731
ZJS		 1272 r = parse_uid(shift, &arg_uid_shift);
1273 if (r < 0)
1274 return log_error_errno(r, "Failed to parse UID \"%s\": %m", optarg);
0de7acce
LP		 1275
1276 arg_userns_mode = USER_NAMESPACE_FIXED;
6dac160c 1277
58e13de5
LP		 1278 if (!userns_shift_range_valid(arg_uid_shift, arg_uid_range))
1279 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "UID range cannot be empty or go beyond " UID_FMT ".", UID_INVALID);
1280 }
be715731 1281
0de7acce 1282 arg_settings_mask |= SETTING_USERNS;
6dac160c 1283 break;
ae209204 1284 }
6dac160c 1285
0de7acce 1286 case 'U':
ccabee0d 1287 if (userns_supported()) {
6c045a99
LP		 1288 arg_userns_mode = USER_NAMESPACE_PICK; /* Note that arg_userns_ownership is
1289 * implied by USER_NAMESPACE_PICK
33eac552 1290 * further down. */
ccabee0d
LP		 1291 arg_uid_shift = UID_INVALID;
1292 arg_uid_range = UINT32_C(0x10000);
1293
1294 arg_settings_mask |= SETTING_USERNS;
6dac160c
LP		 1295 }
1296
7336138e
LP		 1297 break;
1298
0de7acce 1299 case ARG_PRIVATE_USERS_CHOWN:
6c045a99
LP		 1300 arg_userns_ownership = USER_NAMESPACE_OWNERSHIP_CHOWN;
1301
1302 arg_settings_mask |= SETTING_USERNS;
1303 break;
1304
1305 case ARG_PRIVATE_USERS_OWNERSHIP:
1306 if (streq(optarg, "help")) {
1307 DUMP_STRING_TABLE(user_namespace_ownership, UserNamespaceOwnership, _USER_NAMESPACE_OWNERSHIP_MAX);
1308 return 0;
1309 }
1310
1311 arg_userns_ownership = user_namespace_ownership_from_string(optarg);
1312 if (arg_userns_ownership < 0)
1313 return log_error_errno(arg_userns_ownership, "Cannot parse --user-namespace-ownership= value: %s", optarg);
0de7acce
LP		 1314
1315 arg_settings_mask |= SETTING_USERNS;
6dac160c
LP		 1316 break;
1317
c6c8f6e2 1318 case ARG_KILL_SIGNAL:
5c828e66
LP		 1319 if (streq(optarg, "help")) {
1320 DUMP_STRING_TABLE(signal, int, _NSIG);
1321 return 0;
1322 }
1323
29a3db75 1324 arg_kill_signal = signal_from_string(optarg);
baaa35ad 1325 if (arg_kill_signal < 0)
7211c853 1326 return log_error_errno(arg_kill_signal, "Cannot parse signal: %s", optarg);
c6c8f6e2 1327
f757855e
LP		 1328 arg_settings_mask |= SETTING_KILL_SIGNAL;
1329 break;
1330
1331 case ARG_SETTINGS:
1332
1333 /* no → do not read files
1334 * yes → read files, do not override cmdline, trust only subset
1335 * override → read files, override cmdline, trust only subset
1336 * trusted → read files, do not override cmdline, trust all
1337 */
1338
1339 r = parse_boolean(optarg);
1340 if (r < 0) {
1341 if (streq(optarg, "trusted")) {
1342 mask_all_settings = false;
1343 mask_no_settings = false;
1344 arg_settings_trusted = true;
1345
1346 } else if (streq(optarg, "override")) {
1347 mask_all_settings = false;
1348 mask_no_settings = true;
1349 arg_settings_trusted = -1;
1350 } else
1351 return log_error_errno(r, "Failed to parse --settings= argument: %s", optarg);
1352 } else if (r > 0) {
1353 /* yes */
1354 mask_all_settings = false;
1355 mask_no_settings = false;
1356 arg_settings_trusted = -1;
1357 } else {
1358 /* no */
1359 mask_all_settings = true;
1360 mask_no_settings = false;
1361 arg_settings_trusted = false;
1362 }
1363
c6c8f6e2
LP		 1364 break;
1365
5f932eb9 1366 case ARG_CHDIR:
baaa35ad
ZJS		 1367 if (!path_is_absolute(optarg))
1368 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1369 "Working directory %s is not an absolute path.", optarg);
5f932eb9
LP		 1370
1371 r = free_and_strdup(&arg_chdir, optarg);
1372 if (r < 0)
1373 return log_oom();
1374
1375 arg_settings_mask |= SETTING_WORKING_DIRECTORY;
1376 break;
1377
b53ede69
PW		 1378 case ARG_PIVOT_ROOT:
1379 r = pivot_root_parse(&arg_pivot_root_new, &arg_pivot_root_old, optarg);
1380 if (r < 0)
1381 return log_error_errno(r, "Failed to parse --pivot-root= argument %s: %m", optarg);
1382
1383 arg_settings_mask |= SETTING_PIVOT_ROOT;
1384 break;
1385
9c1e04d0
AP		 1386 case ARG_NOTIFY_READY:
1387 r = parse_boolean(optarg);
baaa35ad
ZJS		 1388 if (r < 0)
1389 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1390 "%s is not a valid notify mode. Valid modes are: yes, no, and ready.", optarg);
9c1e04d0
AP		 1391 arg_notify_ready = r;
1392 arg_settings_mask |= SETTING_NOTIFY_READY;
1393 break;
1394
4623e8e6 1395 case ARG_ROOT_HASH: {
89e62e0b 1396 _cleanup_free_ void *k = NULL;
4623e8e6
LP		 1397 size_t l;
1398
1399 r = unhexmem(optarg, strlen(optarg), &k, &l);
1400 if (r < 0)
1401 return log_error_errno(r, "Failed to parse root hash: %s", optarg);
89e62e0b 1402 if (l < sizeof(sd_id128_t))
c6147113 1403 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Root hash must be at least 128bit long: %s", optarg);
4623e8e6 1404
89e62e0b
LP		 1405 free_and_replace(arg_verity_settings.root_hash, k);
1406 arg_verity_settings.root_hash_size = l;
4623e8e6
LP		 1407 break;
1408 }
1409
c2923fdc
LB		 1410 case ARG_ROOT_HASH_SIG: {
1411 char *value;
89e62e0b
LP		 1412 size_t l;
1413 void *p;
c2923fdc
LB		 1414
1415 if ((value = startswith(optarg, "base64:"))) {
c2923fdc
LB		 1416 r = unbase64mem(value, strlen(value), &p, &l);
1417 if (r < 0)
1418 return log_error_errno(r, "Failed to parse root hash signature '%s': %m", optarg);
1419
c2923fdc 1420 } else {
89e62e0b 1421 r = read_full_file(optarg, (char**) &p, &l);
c2923fdc 1422 if (r < 0)
89e62e0b 1423 return log_error_errno(r, "Failed parse root hash signature file '%s': %m", optarg);
c2923fdc
LB		 1424 }
1425
89e62e0b
LP		 1426 free_and_replace(arg_verity_settings.root_hash_sig, p);
1427 arg_verity_settings.root_hash_sig_size = l;
c2923fdc
LB		 1428 break;
1429 }
1430
89e62e0b 1431 case ARG_VERITY_DATA:
614b022c 1432 r = parse_path_argument(optarg, false, &arg_verity_settings.data_path);
89e62e0b
LP		 1433 if (r < 0)
1434 return r;
1435 break;
1436
960e4569
LP		 1437 case ARG_SYSTEM_CALL_FILTER: {
1438 bool negative;
1439 const char *items;
1440
1441 negative = optarg[0] == '~';
1442 items = negative ? optarg + 1 : optarg;
1443
1444 for (;;) {
1445 _cleanup_free_ char *word = NULL;
1446
1447 r = extract_first_word(&items, &word, NULL, 0);
1448 if (r == 0)
1449 break;
1450 if (r == -ENOMEM)
1451 return log_oom();
1452 if (r < 0)
1453 return log_error_errno(r, "Failed to parse system call filter: %m");
1454
1455 if (negative)
6b000af4 1456 r = strv_extend(&arg_syscall_deny_list, word);
960e4569 1457 else
6b000af4 1458 r = strv_extend(&arg_syscall_allow_list, word);
960e4569
LP		 1459 if (r < 0)
1460 return log_oom();
1461 }
1462
1463 arg_settings_mask |= SETTING_SYSCALL_FILTER;
1464 break;
1465 }
1466
bf428efb
LP		 1467 case ARG_RLIMIT: {
1468 const char *eq;
622ecfa8 1469 _cleanup_free_ char *name = NULL;
bf428efb
LP		 1470 int rl;
1471
5c828e66
LP		 1472 if (streq(optarg, "help")) {
1473 DUMP_STRING_TABLE(rlimit, int, _RLIMIT_MAX);
1474 return 0;
1475 }
1476
bf428efb 1477 eq = strchr(optarg, '=');
baaa35ad
ZJS		 1478 if (!eq)
1479 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1480 "--rlimit= expects an '=' assignment.");
bf428efb
LP		 1481
1482 name = strndup(optarg, eq - optarg);
1483 if (!name)
1484 return log_oom();
1485
1486 rl = rlimit_from_string_harder(name);
baaa35ad 1487 if (rl < 0)
7211c853 1488 return log_error_errno(rl, "Unknown resource limit: %s", name);
bf428efb
LP		 1489
1490 if (!arg_rlimit[rl]) {
1491 arg_rlimit[rl] = new0(struct rlimit, 1);
1492 if (!arg_rlimit[rl])
1493 return log_oom();
1494 }
1495
1496 r = rlimit_parse(rl, eq + 1, arg_rlimit[rl]);
1497 if (r < 0)
1498 return log_error_errno(r, "Failed to parse resource limit: %s", eq + 1);
1499
1500 arg_settings_mask |= SETTING_RLIMIT_FIRST << rl;
1501 break;
1502 }
1503
81f345df
LP		 1504 case ARG_OOM_SCORE_ADJUST:
1505 r = parse_oom_score_adjust(optarg, &arg_oom_score_adjust);
1506 if (r < 0)
1507 return log_error_errno(r, "Failed to parse --oom-score-adjust= parameter: %s", optarg);
1508
1509 arg_oom_score_adjust_set = true;
1510 arg_settings_mask |= SETTING_OOM_SCORE_ADJUST;
1511 break;
1512
d107bb7d 1513 case ARG_CPU_AFFINITY: {
0985c7c4 1514 CPUSet cpuset;
d107bb7d
LP		 1515
1516 r = parse_cpu_set(optarg, &cpuset);
1517 if (r < 0)
0985c7c4 1518 return log_error_errno(r, "Failed to parse CPU affinity mask %s: %m", optarg);
d107bb7d 1519
0985c7c4
ZJS		 1520 cpu_set_reset(&arg_cpu_set);
1521 arg_cpu_set = cpuset;
d107bb7d
LP		 1522 arg_settings_mask |= SETTING_CPU_AFFINITY;
1523 break;
1524 }
1525
09d423e9
LP		 1526 case ARG_RESOLV_CONF:
1527 if (streq(optarg, "help")) {
1528 DUMP_STRING_TABLE(resolv_conf_mode, ResolvConfMode, _RESOLV_CONF_MODE_MAX);
1529 return 0;
1530 }
1531
1532 arg_resolv_conf = resolv_conf_mode_from_string(optarg);
baaa35ad 1533 if (arg_resolv_conf < 0)
7211c853 1534 return log_error_errno(arg_resolv_conf,
baaa35ad 1535 "Failed to parse /etc/resolv.conf mode: %s", optarg);
09d423e9
LP		 1536
1537 arg_settings_mask |= SETTING_RESOLV_CONF;
1538 break;
1539
1688841f
LP		 1540 case ARG_TIMEZONE:
1541 if (streq(optarg, "help")) {
1542 DUMP_STRING_TABLE(timezone_mode, TimezoneMode, _TIMEZONE_MODE_MAX);
1543 return 0;
1544 }
1545
1546 arg_timezone = timezone_mode_from_string(optarg);
baaa35ad 1547 if (arg_timezone < 0)
7211c853 1548 return log_error_errno(arg_timezone,
baaa35ad 1549 "Failed to parse /etc/localtime mode: %s", optarg);
1688841f
LP		 1550
1551 arg_settings_mask |= SETTING_TIMEZONE;
1552 break;
1553
de40a303 1554 case ARG_CONSOLE:
dce66ffe
ZJS		 1555 r = handle_arg_console(optarg);
1556 if (r <= 0)
1557 return r;
de40a303
LP		 1558 break;
1559
1560 case 'P':
1561 case ARG_PIPE:
dce66ffe
ZJS		 1562 r = handle_arg_console("pipe");
1563 if (r <= 0)
1564 return r;
de40a303
LP		 1565 break;
1566
bb068de0
ZJS		 1567 case ARG_NO_PAGER:
1568 arg_pager_flags |= PAGER_DISABLE;
1569 break;
1570
3652872a
LP		 1571 case ARG_SET_CREDENTIAL: {
1572 _cleanup_free_ char *word = NULL, *data = NULL;
1573 const char *p = optarg;
1574 Credential *a;
e437538f 1575 ssize_t l;
3652872a
LP		 1576
1577 r = extract_first_word(&p, &word, ":", EXTRACT_DONT_COALESCE_SEPARATORS);
1578 if (r == -ENOMEM)
1579 return log_oom();
1580 if (r < 0)
1581 return log_error_errno(r, "Failed to parse --set-credential= parameter: %m");
1582 if (r == 0 || !p)
1583 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Missing value for --set-credential=: %s", optarg);
1584
1585 if (!credential_name_valid(word))
1586 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Credential name is not valid: %s", word);
1587
12d729b2 1588 for (size_t i = 0; i < arg_n_credentials; i++)
3652872a
LP		 1589 if (streq(arg_credentials[i].id, word))
1590 return log_error_errno(SYNTHETIC_ERRNO(EEXIST), "Duplicate credential '%s', refusing.", word);
1591
1592 l = cunescape(p, UNESCAPE_ACCEPT_NUL, &data);
1593 if (l < 0)
1594 return log_error_errno(l, "Failed to unescape credential data: %s", p);
1595
1596 a = reallocarray(arg_credentials, arg_n_credentials + 1, sizeof(Credential));
1597 if (!a)
1598 return log_oom();
1599
1600 a[arg_n_credentials++] = (Credential) {
1601 .id = TAKE_PTR(word),
1602 .data = TAKE_PTR(data),
1603 .size = l,
1604 };
1605
1606 arg_credentials = a;
1607
1608 arg_settings_mask |= SETTING_CREDENTIALS;
1609 break;
1610 }
1611
1612 case ARG_LOAD_CREDENTIAL: {
1613 ReadFullFileFlags flags = READ_FULL_FILE_SECURE;
1614 _cleanup_(erase_and_freep) char *data = NULL;
1615 _cleanup_free_ char *word = NULL, *j = NULL;
1616 const char *p = optarg;
1617 Credential *a;
1618 size_t size, i;
1619
1620 r = extract_first_word(&p, &word, ":", EXTRACT_DONT_COALESCE_SEPARATORS);
1621 if (r == -ENOMEM)
1622 return log_oom();
1623 if (r < 0)
1624 return log_error_errno(r, "Failed to parse --set-credential= parameter: %m");
1625 if (r == 0 || !p)
1626 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Missing value for --set-credential=: %s", optarg);
1627
1628 if (!credential_name_valid(word))
1629 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Credential name is not valid: %s", word);
1630
1631 for (i = 0; i < arg_n_credentials; i++)
1632 if (streq(arg_credentials[i].id, word))
1633 return log_error_errno(SYNTHETIC_ERRNO(EEXIST), "Duplicate credential '%s', refusing.", word);
1634
1635 if (path_is_absolute(p))
1636 flags |= READ_FULL_FILE_CONNECT_SOCKET;
1637 else {
1638 const char *e;
1639
786d19fd
LP		 1640 r = get_credentials_dir(&e);
1641 if (r < 0)
1642 return log_error_errno(r, "Credential not available (no credentials passed at all): %s", word);
3652872a
LP		 1643
1644 j = path_join(e, p);
1645 if (!j)
1646 return log_oom();
1647 }
1648
986311c2
LP		 1649 r = read_full_file_full(AT_FDCWD, j ?: p, UINT64_MAX, SIZE_MAX,
1650 flags,
1651 NULL,
1652 &data, &size);
3652872a
LP		 1653 if (r < 0)
1654 return log_error_errno(r, "Failed to read credential '%s': %m", j ?: p);
1655
1656 a = reallocarray(arg_credentials, arg_n_credentials + 1, sizeof(Credential));
1657 if (!a)
1658 return log_oom();
1659
1660 a[arg_n_credentials++] = (Credential) {
1661 .id = TAKE_PTR(word),
1662 .data = TAKE_PTR(data),
1663 .size = size,
1664 };
1665
1666 arg_credentials = a;
1667
1668 arg_settings_mask |= SETTING_CREDENTIALS;
1669 break;
1670 }
1671
2f893044
LP		 1672 case ARG_BIND_USER:
1673 if (!valid_user_group_name(optarg, 0))
1674 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Invalid user name to bind: %s", optarg);
1675
1676 if (strv_extend(&arg_bind_user, optarg) < 0)
1677 return log_oom();
1678
1679 arg_settings_mask |= SETTING_BIND_USER;
1680 break;
1681
4a4654e0
LP		 1682 case ARG_SUPPRESS_SYNC:
1683 r = parse_boolean_argument("--suppress-sync=", optarg, &arg_suppress_sync);
1684 if (r < 0)
1685 return r;
1686
1687 arg_settings_mask |= SETTING_SUPPRESS_SYNC;
1688 break;
1689
88213476
LP		 1690 case '?':
1691 return -EINVAL;
1692
1693 default:
04499a70 1694 assert_not_reached();
88213476 1695 }
88213476 1696
60f1ec13
LP		 1697 if (argc > optind) {
1698 strv_free(arg_parameters);
1699 arg_parameters = strv_copy(argv + optind);
1700 if (!arg_parameters)
1701 return log_oom();
d7bea6b6 1702
60f1ec13
LP		 1703 arg_settings_mask |= SETTING_START_MODE;
1704 }
1705
1706 if (arg_ephemeral && arg_template && !arg_directory)
1707 /* User asked for ephemeral execution but specified --template= instead of --directory=. Semantically
1708 * such an invocation makes some sense, see https://github.com/systemd/systemd/issues/3667. Let's
1709 * accept this here, and silently make "--ephemeral --template=" equivalent to "--ephemeral
1710 * --directory=". */
1711 arg_directory = TAKE_PTR(arg_template);
1712
bd4b15f2 1713 arg_caps_retain = (arg_caps_retain | plus | (arg_private_network ? UINT64_C(1) << CAP_NET_ADMIN : 0)) & ~minus;
60f1ec13 1714
de40a303 1715 /* Make sure to parse environment before we reset the settings mask below */
49048684
ZJS		 1716 r = parse_environment();
1717 if (r < 0)
1718 return r;
de40a303 1719
60f1ec13
LP		 1720 /* Load all settings from .nspawn files */
1721 if (mask_no_settings)
1722 arg_settings_mask = 0;
1723
1724 /* Don't load any settings from .nspawn files */
1725 if (mask_all_settings)
1726 arg_settings_mask = _SETTINGS_MASK_ALL;
1727
1728 return 1;
1729}
1730
1731static int verify_arguments(void) {
1732 int r;
a6b5216c 1733
75b0d8b8
ZJS		 1734 if (arg_start_mode == START_PID2 && arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
1735 /* If we are running the stub init in the container, we don't need to look at what the init
1736 * in the container supports, because we are not using it. Let's immediately pick the right
1737 * setting based on the host system configuration.
1738 *
1739 * We only do this, if the user didn't use an environment variable to override the detection.
1740 */
1741
1742 r = cg_all_unified();
1743 if (r < 0)
1744 return log_error_errno(r, "Failed to determine whether we are in all unified mode.");
1745 if (r > 0)
1746 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
1747 else if (cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER) > 0)
1748 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD;
1749 else
1750 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
1751 }
1752
4f086aab
SU		 1753 if (arg_userns_mode != USER_NAMESPACE_NO)
1754 arg_mount_settings |= MOUNT_USE_USERNS;
1755
1756 if (arg_private_network)
1757 arg_mount_settings |= MOUNT_APPLY_APIVFS_NETNS;
1758
48a8d337
LB		 1759 if (!(arg_clone_ns_flags & CLONE_NEWPID) ||
1760 !(arg_clone_ns_flags & CLONE_NEWUTS)) {
eb91eb18 1761 arg_register = false;
baaa35ad 1762 if (arg_start_mode != START_PID1)
60f1ec13 1763 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--boot cannot be used without namespacing.");
0c582db0 1764 }
eb91eb18 1765
6c045a99
LP		 1766 if (arg_userns_ownership < 0)
1767 arg_userns_ownership =
f61c7f88 1768 arg_userns_mode == USER_NAMESPACE_PICK ? USER_NAMESPACE_OWNERSHIP_AUTO :
6c045a99 1769 USER_NAMESPACE_OWNERSHIP_OFF;
0e7ac751 1770
60f1ec13
LP		 1771 if (arg_start_mode == START_BOOT && arg_kill_signal <= 0)
1772 arg_kill_signal = SIGRTMIN+3;
1773
e5a4bb0d
LP		 1774 if (arg_volatile_mode != VOLATILE_NO) /* Make sure all file systems contained in the image are mounted read-only if we are in volatile mode */
1775 arg_read_only = true;
1776
2436ea76
DDM		 1777 if (has_custom_root_mount(arg_custom_mounts, arg_n_custom_mounts))
1778 arg_read_only = true;
1779
baaa35ad 1780 if (arg_keep_unit && arg_register && cg_pid_get_owner_uid(0, NULL) >= 0)
8d9c2bca
AJ		 1781 /* Save the user from accidentally registering either user-$SESSION.scope or user@.service.
1782 * The latter is not technically a user session, but we don't need to labour the point. */
60f1ec13 1783 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--keep-unit --register=yes may not be used when invoked from a user session.");
89f7c846 1784
baaa35ad 1785 if (arg_directory && arg_image)
60f1ec13 1786 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--directory= and --image= may not be combined.");
1b9e5b12 1787
baaa35ad 1788 if (arg_template && arg_image)
60f1ec13 1789 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--template= and --image= may not be combined.");
8cd328d8 1790
baaa35ad 1791 if (arg_template && !(arg_directory || arg_machine))
60f1ec13 1792 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--template= needs --directory= or --machine=.");
ec16945e 1793
baaa35ad 1794 if (arg_ephemeral && arg_template)
60f1ec13 1795 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--ephemeral and --template= may not be combined.");
ec16945e 1796
baaa35ad 1797 if (arg_ephemeral && !IN_SET(arg_link_journal, LINK_NO, LINK_AUTO))
60f1ec13 1798 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--ephemeral and --link-journal= may not be combined.");
df9a75e4 1799
baaa35ad 1800 if (arg_userns_mode != USER_NAMESPACE_NO && !userns_supported())
60f1ec13 1801 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), "--private-users= is not supported, kernel compiled without user namespace support.");
7336138e 1802
6c045a99 1803 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_CHOWN && arg_read_only)
de40a303 1804 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
6c045a99 1805 "--read-only and --private-users-ownership=chown may not be combined.");
f757855e 1806
6c045a99
LP		 1807 /* We don't support --private-users-ownership=chown together with any of the volatile modes since we
1808 * couldn't change the read-only part of the tree (i.e. /usr) anyway, or because it would trigger a
1809 * massive copy-up (in case of overlay) making the entire exercise pointless. */
1810 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_CHOWN && arg_volatile_mode != VOLATILE_NO)
1811 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--volatile= and --private-users-ownership=chown may not be combined.");
e5a4bb0d 1812
679ecd36
SZ		 1813 /* If --network-namespace-path is given with any other network-related option (except --private-network),
1814 * we need to error out, to avoid conflicts between different network options. */
60f1ec13
LP		 1815 if (arg_network_namespace_path &&
1816 (arg_network_interfaces || arg_network_macvlan ||
1817 arg_network_ipvlan || arg_network_veth_extra ||
1818 arg_network_bridge || arg_network_zone ||
679ecd36 1819 arg_network_veth))
de40a303 1820 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--network-namespace-path= cannot be combined with other network options.");
86c0dd4a 1821
60f1ec13 1822 if (arg_network_bridge && arg_network_zone)
de40a303
LP		 1823 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1824 "--network-bridge= and --network-zone= may not be combined.");
f757855e 1825
baaa35ad 1826 if (arg_userns_mode != USER_NAMESPACE_NO && (arg_mount_settings & MOUNT_APPLY_APIVFS_NETNS) && !arg_private_network)
60f1ec13 1827 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Invalid namespacing settings. Mounting sysfs with --private-users requires --private-network.");
4f086aab 1828
baaa35ad 1829 if (arg_userns_mode != USER_NAMESPACE_NO && !(arg_mount_settings & MOUNT_APPLY_APIVFS_RO))
60f1ec13 1830 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Cannot combine --private-users with read-write mounts.");
f757855e 1831
baaa35ad 1832 if (arg_expose_ports && !arg_private_network)
60f1ec13 1833 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Cannot use --port= without private networking.");
6d0b55c2 1834
88fc9c9b 1835 if (arg_caps_ambient) {
f5fbe71d 1836 if (arg_caps_ambient == UINT64_MAX)
88fc9c9b
TH		 1837 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= does not support the value all.");
1838
1839 if ((arg_caps_ambient & arg_caps_retain) != arg_caps_ambient)
1840 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= setting is not fully covered by Capability= setting.");
1841
1842 if (arg_start_mode == START_BOOT)
1843 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= setting is not useful for boot mode.");
1844 }
1845
2f893044
LP		 1846 if (arg_userns_mode == USER_NAMESPACE_NO && !strv_isempty(arg_bind_user))
1847 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--bind-user= requires --private-users");
1848
1849 /* Drop duplicate --bind-user= entries */
1850 strv_uniq(arg_bind_user);
1851
60f1ec13
LP		 1852 r = custom_mount_check_all();
1853 if (r < 0)
1854 return r;
c6c8f6e2 1855
f757855e 1856 return 0;
88213476
LP		 1857}
1858
91181e07 1859int userns_lchown(const char *p, uid_t uid, gid_t gid) {
03cfe0d5
LP		 1860 assert(p);
1861
0de7acce 1862 if (arg_userns_mode == USER_NAMESPACE_NO)
03cfe0d5
LP		 1863 return 0;
1864
1865 if (uid == UID_INVALID && gid == GID_INVALID)
1866 return 0;
1867
1868 if (uid != UID_INVALID) {
1869 uid += arg_uid_shift;
1870
1871 if (uid < arg_uid_shift || uid >= arg_uid_shift + arg_uid_range)
1872 return -EOVERFLOW;
1873 }
1874
1875 if (gid != GID_INVALID) {
1876 gid += (gid_t) arg_uid_shift;
1877
1878 if (gid < (gid_t) arg_uid_shift || gid >= (gid_t) (arg_uid_shift + arg_uid_range))
1879 return -EOVERFLOW;
1880 }
1881
7c248223 1882 return RET_NERRNO(lchown(p, uid, gid));
b12afc8c
LP		 1883}
1884
91181e07 1885int userns_mkdir(const char *root, const char *path, mode_t mode, uid_t uid, gid_t gid) {
03cfe0d5 1886 const char *q;
dae8b82e 1887 int r;
03cfe0d5
LP		 1888
1889 q = prefix_roota(root, path);
3f692e2e 1890 r = RET_NERRNO(mkdir(q, mode));
dae8b82e
ZJS		 1891 if (r == -EEXIST)
1892 return 0;
1893 if (r < 0)
1894 return r;
03cfe0d5
LP		 1895
1896 return userns_lchown(q, uid, gid);
1897}
1898
1688841f 1899static const char *timezone_from_path(const char *path) {
da9fc98d
LP		 1900 return PATH_STARTSWITH_SET(
1901 path,
1902 "../usr/share/zoneinfo/",
1903 "/usr/share/zoneinfo/");
1688841f
LP		 1904}
1905
83205269
LP		 1906static bool etc_writable(void) {
1907 return !arg_read_only || IN_SET(arg_volatile_mode, VOLATILE_YES, VOLATILE_OVERLAY);
1908}
1909
e58a1277 1910static int setup_timezone(const char *dest) {
1688841f
LP		 1911 _cleanup_free_ char *p = NULL, *etc = NULL;
1912 const char *where, *check;
1913 TimezoneMode m;
d4036145 1914 int r;
f8440af5 1915
e58a1277
LP		 1916 assert(dest);
1917
1688841f 1918 if (IN_SET(arg_timezone, TIMEZONE_AUTO, TIMEZONE_SYMLINK)) {
1688841f
LP		 1919 r = readlink_malloc("/etc/localtime", &p);
1920 if (r == -ENOENT && arg_timezone == TIMEZONE_AUTO)
83205269 1921 m = etc_writable() ? TIMEZONE_DELETE : TIMEZONE_OFF;
1688841f 1922 else if (r == -EINVAL && arg_timezone == TIMEZONE_AUTO) /* regular file? */
83205269 1923 m = etc_writable() ? TIMEZONE_COPY : TIMEZONE_BIND;
1688841f
LP		 1924 else if (r < 0) {
1925 log_warning_errno(r, "Failed to read host's /etc/localtime symlink, not updating container timezone: %m");
1926 /* To handle warning, delete /etc/localtime and replace it with a symbolic link to a time zone data
1927 * file.
1928 *
1929 * Example:
1930 * ln -s /usr/share/zoneinfo/UTC /etc/localtime
1931 */
1932 return 0;
1933 } else if (arg_timezone == TIMEZONE_AUTO)
83205269 1934 m = etc_writable() ? TIMEZONE_SYMLINK : TIMEZONE_BIND;
1688841f
LP		 1935 else
1936 m = arg_timezone;
1937 } else
1938 m = arg_timezone;
1939
1940 if (m == TIMEZONE_OFF)
1941 return 0;
1942
a5648b80 1943 r = chase_symlinks("/etc", dest, CHASE_PREFIX_ROOT, &etc, NULL);
d4036145 1944 if (r < 0) {
1688841f 1945 log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
d4036145
LP		 1946 return 0;
1947 }
1948
1688841f
LP		 1949 where = strjoina(etc, "/localtime");
1950
1951 switch (m) {
1952
1953 case TIMEZONE_DELETE:
1954 if (unlink(where) < 0)
1955 log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno, "Failed to remove '%s', ignoring: %m", where);
1956
d4036145 1957 return 0;
d4036145 1958
1688841f
LP		 1959 case TIMEZONE_SYMLINK: {
1960 _cleanup_free_ char *q = NULL;
1961 const char *z, *what;
4d1c38b8 1962
1688841f
LP		 1963 z = timezone_from_path(p);
1964 if (!z) {
1965 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
d4036145 1966 return 0;
1688841f 1967 }
d4036145 1968
1688841f
LP		 1969 r = readlink_malloc(where, &q);
1970 if (r >= 0 && streq_ptr(timezone_from_path(q), z))
1971 return 0; /* Already pointing to the right place? Then do nothing .. */
1972
1973 check = strjoina(dest, "/usr/share/zoneinfo/", z);
a5648b80 1974 r = chase_symlinks(check, dest, 0, NULL, NULL);
1688841f
LP		 1975 if (r < 0)
1976 log_debug_errno(r, "Timezone %s does not exist (or is not accessible) in container, not creating symlink: %m", z);
1977 else {
1978 if (unlink(where) < 0 && errno != ENOENT) {
1979 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING, /* Don't complain on read-only images */
1980 errno, "Failed to remove existing timezone info %s in container, ignoring: %m", where);
1981 return 0;
1982 }
1983
1984 what = strjoina("../usr/share/zoneinfo/", z);
1985 if (symlink(what, where) < 0) {
1986 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING,
1987 errno, "Failed to correct timezone of container, ignoring: %m");
1988 return 0;
1989 }
1990
1991 break;
1992 }
1993
1994 _fallthrough_;
d4036145 1995 }
68fb0892 1996
1688841f
LP		 1997 case TIMEZONE_BIND: {
1998 _cleanup_free_ char *resolved = NULL;
1999 int found;
2000
a5648b80 2001 found = chase_symlinks(where, dest, CHASE_NONEXISTENT, &resolved, NULL);
1688841f
LP		 2002 if (found < 0) {
2003 log_warning_errno(found, "Failed to resolve /etc/localtime path in container, ignoring: %m");
2004 return 0;
2005 }
2006
2007 if (found == 0) /* missing? */
2008 (void) touch(resolved);
2009
511a8cfe 2010 r = mount_nofollow_verbose(LOG_WARNING, "/etc/localtime", resolved, NULL, MS_BIND, NULL);
1688841f 2011 if (r >= 0)
511a8cfe 2012 return mount_nofollow_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
1688841f
LP		 2013
2014 _fallthrough_;
79d80fc1 2015 }
4d9f07b4 2016
1688841f
LP		 2017 case TIMEZONE_COPY:
2018 /* If mounting failed, try to copy */
8a016c74 2019 r = copy_file_atomic("/etc/localtime", where, 0644, 0, 0, COPY_REFLINK|COPY_REPLACE);
1688841f
LP		 2020 if (r < 0) {
2021 log_full_errno(IN_SET(r, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
2022 "Failed to copy /etc/localtime to %s, ignoring: %m", where);
2023 return 0;
2024 }
2025
2026 break;
2027
2028 default:
04499a70 2029 assert_not_reached();
d4036145 2030 }
e58a1277 2031
1688841f 2032 /* Fix permissions of the symlink or file copy we just created */
03cfe0d5
LP		 2033 r = userns_lchown(where, 0, 0);
2034 if (r < 0)
1688841f 2035 log_warning_errno(r, "Failed to chown /etc/localtime, ignoring: %m");
03cfe0d5 2036
e58a1277 2037 return 0;
88213476
LP		 2038}
2039
09d423e9
LP		 2040static int have_resolv_conf(const char *path) {
2041 assert(path);
2042
2043 if (access(path, F_OK) < 0) {
2044 if (errno == ENOENT)
2045 return 0;
2046
2047 return log_debug_errno(errno, "Failed to determine whether '%s' is available: %m", path);
2048 }
2049
2050 return 1;
2051}
2052
7357272e 2053static int resolved_listening(void) {
b8ea7a6e 2054 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
b053cd5f 2055 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
7357272e 2056 _cleanup_free_ char *dns_stub_listener_mode = NULL;
b053cd5f
LP		 2057 int r;
2058
7357272e 2059 /* Check if resolved is listening */
b053cd5f
LP		 2060
2061 r = sd_bus_open_system(&bus);
2062 if (r < 0)
b8ea7a6e 2063 return log_debug_errno(r, "Failed to open system bus: %m");
b053cd5f 2064
7357272e 2065 r = bus_name_has_owner(bus, "org.freedesktop.resolve1", NULL);
b8ea7a6e
LP		 2066 if (r < 0)
2067 return log_debug_errno(r, "Failed to check whether the 'org.freedesktop.resolve1' bus name is taken: %m");
2068 if (r == 0)
2069 return 0;
7357272e
DM		 2070
2071 r = sd_bus_get_property_string(bus,
2072 "org.freedesktop.resolve1",
2073 "/org/freedesktop/resolve1",
2074 "org.freedesktop.resolve1.Manager",
2075 "DNSStubListener",
b8ea7a6e 2076 &error,
7357272e
DM		 2077 &dns_stub_listener_mode);
2078 if (r < 0)
b8ea7a6e 2079 return log_debug_errno(r, "Failed to query DNSStubListener property: %s", bus_error_message(&error, r));
7357272e
DM		 2080
2081 return STR_IN_SET(dns_stub_listener_mode, "udp", "yes");
b053cd5f
LP		 2082}
2083
2547bb41 2084static int setup_resolv_conf(const char *dest) {
09d423e9
LP		 2085 _cleanup_free_ char *etc = NULL;
2086 const char *where, *what;
2087 ResolvConfMode m;
2088 int r;
2547bb41
LP		 2089
2090 assert(dest);
2091
09d423e9
LP		 2092 if (arg_resolv_conf == RESOLV_CONF_AUTO) {
2093 if (arg_private_network)
2094 m = RESOLV_CONF_OFF;
86775e35
LP		 2095 else if (have_resolv_conf(PRIVATE_STUB_RESOLV_CONF) > 0 && resolved_listening() > 0)
2096 m = etc_writable() ? RESOLV_CONF_COPY_STUB : RESOLV_CONF_BIND_STUB;
09d423e9 2097 else if (have_resolv_conf("/etc/resolv.conf") > 0)
83205269 2098 m = etc_writable() ? RESOLV_CONF_COPY_HOST : RESOLV_CONF_BIND_HOST;
09d423e9 2099 else
83205269 2100 m = etc_writable() ? RESOLV_CONF_DELETE : RESOLV_CONF_OFF;
86775e35 2101
09d423e9
LP		 2102 } else
2103 m = arg_resolv_conf;
2104
2105 if (m == RESOLV_CONF_OFF)
2547bb41
LP		 2106 return 0;
2107
a5648b80 2108 r = chase_symlinks("/etc", dest, CHASE_PREFIX_ROOT, &etc, NULL);
87447ae4
LP		 2109 if (r < 0) {
2110 log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
2111 return 0;
2112 }
2113
2114 where = strjoina(etc, "/resolv.conf");
09d423e9
LP		 2115
2116 if (m == RESOLV_CONF_DELETE) {
2117 if (unlink(where) < 0)
2118 log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno, "Failed to remove '%s', ignoring: %m", where);
2119
87447ae4
LP		 2120 return 0;
2121 }
79d80fc1 2122
86775e35
LP		 2123 if (IN_SET(m, RESOLV_CONF_BIND_STATIC, RESOLV_CONF_REPLACE_STATIC, RESOLV_CONF_COPY_STATIC))
2124 what = PRIVATE_STATIC_RESOLV_CONF;
2125 else if (IN_SET(m, RESOLV_CONF_BIND_UPLINK, RESOLV_CONF_REPLACE_UPLINK, RESOLV_CONF_COPY_UPLINK))
2126 what = PRIVATE_UPLINK_RESOLV_CONF;
2127 else if (IN_SET(m, RESOLV_CONF_BIND_STUB, RESOLV_CONF_REPLACE_STUB, RESOLV_CONF_COPY_STUB))
2128 what = PRIVATE_STUB_RESOLV_CONF;
09d423e9
LP		 2129 else
2130 what = "/etc/resolv.conf";
87447ae4 2131
86775e35 2132 if (IN_SET(m, RESOLV_CONF_BIND_HOST, RESOLV_CONF_BIND_STATIC, RESOLV_CONF_BIND_UPLINK, RESOLV_CONF_BIND_STUB)) {
09d423e9
LP		 2133 _cleanup_free_ char *resolved = NULL;
2134 int found;
2135
a5648b80 2136 found = chase_symlinks(where, dest, CHASE_NONEXISTENT, &resolved, NULL);
09d423e9
LP		 2137 if (found < 0) {
2138 log_warning_errno(found, "Failed to resolve /etc/resolv.conf path in container, ignoring: %m");
2139 return 0;
2140 }
3539724c 2141
87447ae4
LP		 2142 if (found == 0) /* missing? */
2143 (void) touch(resolved);
5367354d 2144
511a8cfe 2145 r = mount_nofollow_verbose(LOG_WARNING, what, resolved, NULL, MS_BIND, NULL);
60e76d48 2146 if (r >= 0)
511a8cfe 2147 return mount_nofollow_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
86775e35
LP		 2148
2149 /* If that didn't work, let's copy the file */
3539724c
LP		 2150 }
2151
86775e35
LP		 2152 if (IN_SET(m, RESOLV_CONF_REPLACE_HOST, RESOLV_CONF_REPLACE_STATIC, RESOLV_CONF_REPLACE_UPLINK, RESOLV_CONF_REPLACE_STUB))
2153 r = copy_file_atomic(what, where, 0644, 0, 0, COPY_REFLINK|COPY_REPLACE);
2154 else
2155 r = copy_file(what, where, O_TRUNC|O_NOFOLLOW, 0644, 0, 0, COPY_REFLINK);
79d80fc1 2156 if (r < 0) {
3539724c
LP		 2157 /* If the file already exists as symlink, let's suppress the warning, under the assumption that
2158 * resolved or something similar runs inside and the symlink points there.
68a313c5 2159 *
3539724c 2160 * If the disk image is read-only, there's also no point in complaining.
68a313c5 2161 */
86775e35
LP		 2162 log_full_errno(!IN_SET(RESOLV_CONF_COPY_HOST, RESOLV_CONF_COPY_STATIC, RESOLV_CONF_COPY_UPLINK, RESOLV_CONF_COPY_STUB) &&
2163 IN_SET(r, -ELOOP, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
3539724c 2164 "Failed to copy /etc/resolv.conf to %s, ignoring: %m", where);
79d80fc1
TG		 2165 return 0;
2166 }
2547bb41 2167
03cfe0d5
LP		 2168 r = userns_lchown(where, 0, 0);
2169 if (r < 0)
3539724c 2170 log_warning_errno(r, "Failed to chown /etc/resolv.conf, ignoring: %m");
03cfe0d5 2171
2547bb41
LP		 2172 return 0;
2173}
2174
1e4f1671 2175static int setup_boot_id(void) {
cdde6ba6
LP		 2176 _cleanup_(unlink_and_freep) char *from = NULL;
2177 _cleanup_free_ char *path = NULL;
3bbaff3e 2178 sd_id128_t rnd = SD_ID128_NULL;
cdde6ba6 2179 const char *to;
04bc4a3f
LP		 2180 int r;
2181
1eacc470 2182 /* Generate a new randomized boot ID, so that each boot-up of the container gets a new one */
04bc4a3f 2183
1eacc470 2184 r = tempfn_random_child("/run", "proc-sys-kernel-random-boot-id", &path);
cdde6ba6
LP		 2185 if (r < 0)
2186 return log_error_errno(r, "Failed to generate random boot ID path: %m");
04bc4a3f
LP		 2187
2188 r = sd_id128_randomize(&rnd);
f647962d
MS		 2189 if (r < 0)
2190 return log_error_errno(r, "Failed to generate random boot id: %m");
04bc4a3f 2191
cdde6ba6 2192 r = id128_write(path, ID128_UUID, rnd, false);
f647962d
MS		 2193 if (r < 0)
2194 return log_error_errno(r, "Failed to write boot id: %m");
04bc4a3f 2195
cdde6ba6
LP		 2196 from = TAKE_PTR(path);
2197 to = "/proc/sys/kernel/random/boot_id";
2198
511a8cfe 2199 r = mount_nofollow_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL);
cdde6ba6
LP		 2200 if (r < 0)
2201 return r;
04bc4a3f 2202
511a8cfe 2203 return mount_nofollow_verbose(LOG_ERR, NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
04bc4a3f
LP		 2204}
2205
e58a1277 2206static int copy_devnodes(const char *dest) {
88213476
LP		 2207 static const char devnodes[] =
2208 "null\0"
2209 "zero\0"
2210 "full\0"
2211 "random\0"
2212 "urandom\0"
85614d66
TG		 2213 "tty\0"
2214 "net/tun\0";
88213476
LP		 2215
2216 const char *d;
e58a1277 2217 int r = 0;
a258bf26
LP		 2218
2219 assert(dest);
124640f1 2220
52f05ef2 2221 BLOCK_WITH_UMASK(0000);
88213476 2222
03cfe0d5
LP		 2223 /* Create /dev/net, so that we can create /dev/net/tun in it */
2224 if (userns_mkdir(dest, "/dev/net", 0755, 0, 0) < 0)
2225 return log_error_errno(r, "Failed to create /dev/net directory: %m");
2226
88213476 2227 NULSTR_FOREACH(d, devnodes) {
7fd1b19b 2228 _cleanup_free_ char *from = NULL, *to = NULL;
7f112f50 2229 struct stat st;
88213476 2230
c6134d3e 2231 from = path_join("/dev/", d);
8967f291
LP		 2232 if (!from)
2233 return log_oom();
2234
c6134d3e 2235 to = path_join(dest, from);
8967f291
LP		 2236 if (!to)
2237 return log_oom();
88213476
LP		 2238
2239 if (stat(from, &st) < 0) {
2240
4a62c710
MS		 2241 if (errno != ENOENT)
2242 return log_error_errno(errno, "Failed to stat %s: %m", from);
88213476 2243
baaa35ad
ZJS		 2244 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode))
2245 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2246 "%s is not a char or block device, cannot copy.", from);
2247 else {
8dfce114
LP		 2248 _cleanup_free_ char *sl = NULL, *prefixed = NULL, *dn = NULL, *t = NULL;
2249
81f5049b 2250 if (mknod(to, st.st_mode, st.st_rdev) < 0) {
8dbf71ec 2251 /* Explicitly warn the user when /dev is already populated. */
41eb4362 2252 if (errno == EEXIST)
8dbf71ec 2253 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		 2254 if (errno != EPERM)
2255 return log_error_errno(errno, "mknod(%s) failed: %m", to);
2256
8dfce114 2257 /* Some systems abusively restrict mknod but allow bind mounts. */
81f5049b
AC		 2258 r = touch(to);
2259 if (r < 0)
2260 return log_error_errno(r, "touch (%s) failed: %m", to);
511a8cfe 2261 r = mount_nofollow_verbose(LOG_DEBUG, from, to, NULL, MS_BIND, NULL);
60e76d48
ZJS		 2262 if (r < 0)
2263 return log_error_errno(r, "Both mknod and bind mount (%s) failed: %m", to);
81f5049b 2264 }
6278cf60 2265
03cfe0d5
LP		 2266 r = userns_lchown(to, 0, 0);
2267 if (r < 0)
2268 return log_error_errno(r, "chown() of device node %s failed: %m", to);
8dfce114 2269
657ee2d8 2270 dn = path_join("/dev", S_ISCHR(st.st_mode) ? "char" : "block");
8dfce114
LP		 2271 if (!dn)
2272 return log_oom();
2273
2274 r = userns_mkdir(dest, dn, 0755, 0, 0);
2275 if (r < 0)
2276 return log_error_errno(r, "Failed to create '%s': %m", dn);
2277
2278 if (asprintf(&sl, "%s/%u:%u", dn, major(st.st_rdev), minor(st.st_rdev)) < 0)
2279 return log_oom();
2280
c6134d3e 2281 prefixed = path_join(dest, sl);
8dfce114
LP		 2282 if (!prefixed)
2283 return log_oom();
2284
2d9b74ba 2285 t = path_join("..", d);
8dfce114
LP		 2286 if (!t)
2287 return log_oom();
2288
2289 if (symlink(t, prefixed) < 0)
2290 log_debug_errno(errno, "Failed to symlink '%s' to '%s': %m", t, prefixed);
88213476 2291 }
88213476
LP		 2292 }
2293
e58a1277
LP		 2294 return r;
2295}
88213476 2296
de40a303 2297static int make_extra_nodes(const char *dest) {
de40a303
LP		 2298 size_t i;
2299 int r;
2300
52f05ef2 2301 BLOCK_WITH_UMASK(0000);
de40a303
LP		 2302
2303 for (i = 0; i < arg_n_extra_nodes; i++) {
2304 _cleanup_free_ char *path = NULL;
2305 DeviceNode *n = arg_extra_nodes + i;
2306
c6134d3e 2307 path = path_join(dest, n->path);
de40a303
LP		 2308 if (!path)
2309 return log_oom();
2310
2311 if (mknod(path, n->mode, S_ISCHR(n->mode) || S_ISBLK(n->mode) ? makedev(n->major, n->minor) : 0) < 0)
2312 return log_error_errno(errno, "Failed to create device node '%s': %m", path);
2313
2314 r = chmod_and_chown(path, n->mode, n->uid, n->gid);
2315 if (r < 0)
2316 return log_error_errno(r, "Failed to adjust device node ownership of '%s': %m", path);
2317 }
2318
2319 return 0;
2320}
2321
03cfe0d5
LP		 2322static int setup_pts(const char *dest) {
2323 _cleanup_free_ char *options = NULL;
2324 const char *p;
709f6e46 2325 int r;
03cfe0d5 2326
349cc4a5 2327#if HAVE_SELINUX
03cfe0d5
LP		 2328 if (arg_selinux_apifs_context)
2329 (void) asprintf(&options,
3dce8915 2330 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"",
03cfe0d5
LP		 2331 arg_uid_shift + TTY_GID,
2332 arg_selinux_apifs_context);
2333 else
2334#endif
2335 (void) asprintf(&options,
3dce8915 2336 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT,
03cfe0d5 2337 arg_uid_shift + TTY_GID);
f2d88580 2338
03cfe0d5 2339 if (!options)
f2d88580
LP		 2340 return log_oom();
2341
03cfe0d5 2342 /* Mount /dev/pts itself */
cc9fce65 2343 p = prefix_roota(dest, "/dev/pts");
3f692e2e 2344 r = RET_NERRNO(mkdir(p, 0755));
dae8b82e
ZJS		 2345 if (r < 0)
2346 return log_error_errno(r, "Failed to create /dev/pts: %m");
2347
511a8cfe 2348 r = mount_nofollow_verbose(LOG_ERR, "devpts", p, "devpts", MS_NOSUID|MS_NOEXEC, options);
60e76d48
ZJS		 2349 if (r < 0)
2350 return r;
709f6e46
MS		 2351 r = userns_lchown(p, 0, 0);
2352 if (r < 0)
2353 return log_error_errno(r, "Failed to chown /dev/pts: %m");
03cfe0d5
LP		 2354
2355 /* Create /dev/ptmx symlink */
2356 p = prefix_roota(dest, "/dev/ptmx");
4a62c710
MS		 2357 if (symlink("pts/ptmx", p) < 0)
2358 return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
709f6e46
MS		 2359 r = userns_lchown(p, 0, 0);
2360 if (r < 0)
2361 return log_error_errno(r, "Failed to chown /dev/ptmx: %m");
f2d88580 2362
03cfe0d5
LP		 2363 /* And fix /dev/pts/ptmx ownership */
2364 p = prefix_roota(dest, "/dev/pts/ptmx");
709f6e46
MS		 2365 r = userns_lchown(p, 0, 0);
2366 if (r < 0)
2367 return log_error_errno(r, "Failed to chown /dev/pts/ptmx: %m");
6278cf60 2368
f2d88580
LP		 2369 return 0;
2370}
2371
3acc84eb 2372static int setup_stdio_as_dev_console(void) {
2fef50cd 2373 _cleanup_close_ int terminal = -1;
e58a1277 2374 int r;
e58a1277 2375
335d2ead
LP		 2376 /* We open the TTY in O_NOCTTY mode, so that we do not become controller yet. We'll do that later
2377 * explicitly, if we are configured to. */
2378 terminal = open_terminal("/dev/console", O_RDWR|O_NOCTTY);
3acc84eb
FB		 2379 if (terminal < 0)
2380 return log_error_errno(terminal, "Failed to open console: %m");
e58a1277 2381
3acc84eb
FB		 2382 /* Make sure we can continue logging to the original stderr, even if
2383 * stderr points elsewhere now */
2384 r = log_dup_console();
2385 if (r < 0)
2386 return log_error_errno(r, "Failed to duplicate stderr: %m");
de40a303 2387
3acc84eb
FB		 2388 /* invalidates 'terminal' on success and failure */
2389 r = rearrange_stdio(terminal, terminal, terminal);
2fef50cd 2390 TAKE_FD(terminal);
f647962d 2391 if (r < 0)
3acc84eb
FB		 2392 return log_error_errno(r, "Failed to move console to stdin/stdout/stderr: %m");
2393
2394 return 0;
2395}
88213476 2396
3acc84eb
FB		 2397static int setup_dev_console(const char *console) {
2398 _cleanup_free_ char *p = NULL;
2399 int r;
a258bf26 2400
3acc84eb
FB		 2401 /* Create /dev/console symlink */
2402 r = path_make_relative("/dev", console, &p);
81f5049b 2403 if (r < 0)
3acc84eb
FB		 2404 return log_error_errno(r, "Failed to create relative path: %m");
2405
2406 if (symlink(p, "/dev/console") < 0)
2407 return log_error_errno(errno, "Failed to create /dev/console symlink: %m");
a258bf26 2408
3acc84eb 2409 return 0;
e58a1277
LP		 2410}
2411
8e5430c4
LP		 2412static int setup_keyring(void) {
2413 key_serial_t keyring;
2414
6b000af4
LP		 2415 /* Allocate a new session keyring for the container. This makes sure the keyring of the session
2416 * systemd-nspawn was invoked from doesn't leak into the container. Note that by default we block
2417 * keyctl() and request_key() anyway via seccomp so doing this operation isn't strictly necessary,
2418 * but in case people explicitly allow-list these system calls let's make sure we don't leak anything
2419 * into the container. */
8e5430c4
LP		 2420
2421 keyring = keyctl(KEYCTL_JOIN_SESSION_KEYRING, 0, 0, 0, 0);
2422 if (keyring == -1) {
2423 if (errno == ENOSYS)
2424 log_debug_errno(errno, "Kernel keyring not supported, ignoring.");
065b4774 2425 else if (ERRNO_IS_PRIVILEGE(errno))
8e5430c4
LP		 2426 log_debug_errno(errno, "Kernel keyring access prohibited, ignoring.");
2427 else
2428 return log_error_errno(errno, "Setting up kernel keyring failed: %m");
2429 }
2430
2431 return 0;
2432}
2433
3652872a
LP		 2434static int setup_credentials(const char *root) {
2435 const char *q;
2436 int r;
2437
2438 if (arg_n_credentials <= 0)
2439 return 0;
2440
2441 r = userns_mkdir(root, "/run/host", 0755, 0, 0);
2442 if (r < 0)
2443 return log_error_errno(r, "Failed to create /run/host: %m");
2444
2445 r = userns_mkdir(root, "/run/host/credentials", 0700, 0, 0);
2446 if (r < 0)
2447 return log_error_errno(r, "Failed to create /run/host/credentials: %m");
2448
2449 q = prefix_roota(root, "/run/host/credentials");
511a8cfe 2450 r = mount_nofollow_verbose(LOG_ERR, NULL, q, "ramfs", MS_NOSUID|MS_NOEXEC|MS_NODEV, "mode=0700");
3652872a
LP		 2451 if (r < 0)
2452 return r;
2453
2454 for (size_t i = 0; i < arg_n_credentials; i++) {
2455 _cleanup_free_ char *j = NULL;
2456 _cleanup_close_ int fd = -1;
2457
2458 j = path_join(q, arg_credentials[i].id);
2459 if (!j)
2460 return log_oom();
2461
2462 fd = open(j, O_CREAT|O_EXCL|O_WRONLY|O_CLOEXEC|O_NOFOLLOW, 0600);
2463 if (fd < 0)
2464 return log_error_errno(errno, "Failed to create credential file %s: %m", j);
2465
2466 r = loop_write(fd, arg_credentials[i].data, arg_credentials[i].size, /* do_poll= */ false);
2467 if (r < 0)
2468 return log_error_errno(r, "Failed to write credential to file %s: %m", j);
2469
2470 if (fchmod(fd, 0400) < 0)
2471 return log_error_errno(errno, "Failed to adjust access mode of %s: %m", j);
2472
2473 if (arg_userns_mode != USER_NAMESPACE_NO) {
2474 if (fchown(fd, arg_uid_shift, arg_uid_shift) < 0)
2475 return log_error_errno(errno, "Failed to adjust ownership of %s: %m", j);
2476 }
2477 }
2478
2479 if (chmod(q, 0500) < 0)
2480 return log_error_errno(errno, "Failed to adjust access mode of %s: %m", q);
2481
2482 r = userns_lchown(q, 0, 0);
2483 if (r < 0)
2484 return r;
2485
2486 /* Make both mount and superblock read-only now */
511a8cfe 2487 r = mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_REMOUNT|MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
3652872a
LP		 2488 if (r < 0)
2489 return r;
2490
511a8cfe 2491 return mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, "mode=0500");
3652872a
LP		 2492}
2493
1e4f1671 2494static int setup_kmsg(int kmsg_socket) {
9ec5a93c
LP		 2495 _cleanup_(unlink_and_freep) char *from = NULL;
2496 _cleanup_free_ char *fifo = NULL;
2497 _cleanup_close_ int fd = -1;
9ec5a93c 2498 int r;
e58a1277 2499
e58a1277 2500 assert(kmsg_socket >= 0);
a258bf26 2501
52f05ef2 2502 BLOCK_WITH_UMASK(0000);
a258bf26 2503
1eacc470 2504 /* We create the kmsg FIFO as as temporary file in /run, but immediately delete it after bind mounting it to
9ec5a93c
LP		 2505 * /proc/kmsg. While FIFOs on the reading side behave very similar to /proc/kmsg, their writing side behaves
2506 * differently from /dev/kmsg in that writing blocks when nothing is reading. In order to avoid any problems
2507 * with containers deadlocking due to this we simply make /dev/kmsg unavailable to the container. */
2508
1eacc470 2509 r = tempfn_random_child("/run", "proc-kmsg", &fifo);
9ec5a93c
LP		 2510 if (r < 0)
2511 return log_error_errno(r, "Failed to generate kmsg path: %m");
e58a1277 2512
9ec5a93c 2513 if (mkfifo(fifo, 0600) < 0)
03cfe0d5 2514 return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m");
9ec5a93c
LP		 2515
2516 from = TAKE_PTR(fifo);
9ec5a93c 2517
511a8cfe 2518 r = mount_nofollow_verbose(LOG_ERR, from, "/proc/kmsg", NULL, MS_BIND, NULL);
60e76d48
ZJS		 2519 if (r < 0)
2520 return r;
e58a1277 2521
669fc4e5 2522 fd = open(from, O_RDWR|O_NONBLOCK|O_CLOEXEC);
4a62c710
MS		 2523 if (fd < 0)
2524 return log_error_errno(errno, "Failed to open fifo: %m");
e58a1277 2525
9ec5a93c 2526 /* Store away the fd in the socket, so that it stays open as long as we run the child */
3ee897d6 2527 r = send_one_fd(kmsg_socket, fd, 0);
d9603714
DH		 2528 if (r < 0)
2529 return log_error_errno(r, "Failed to send FIFO fd: %m");
a258bf26 2530
25ea79fe 2531 return 0;
88213476
LP		 2532}
2533
761cf19d 2534struct ExposeArgs {
deff68e7
FW		 2535 union in_addr_union address4;
2536 union in_addr_union address6;
761cf19d
FW		 2537 struct FirewallContext *fw_ctx;
2538};
2539
1c4baffc 2540static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) {
761cf19d 2541 struct ExposeArgs *args = userdata;
6d0b55c2
LP		 2542
2543 assert(rtnl);
2544 assert(m);
761cf19d 2545 assert(args);
6d0b55c2 2546
fb9044cb
LP		 2547 (void) expose_port_execute(rtnl, &args->fw_ctx, arg_expose_ports, AF_INET, &args->address4);
2548 (void) expose_port_execute(rtnl, &args->fw_ctx, arg_expose_ports, AF_INET6, &args->address6);
6d0b55c2
LP		 2549 return 0;
2550}
2551
3a74cea5 2552static int setup_hostname(void) {
c818eef1 2553 int r;
3a74cea5 2554
0c582db0 2555 if ((arg_clone_ns_flags & CLONE_NEWUTS) == 0)
eb91eb18
LP		 2556 return 0;
2557
c818eef1
LP		 2558 r = sethostname_idempotent(arg_hostname ?: arg_machine);
2559 if (r < 0)
2560 return log_error_errno(r, "Failed to set hostname: %m");
3a74cea5 2561
7027ff61 2562 return 0;
3a74cea5
LP		 2563}
2564
57fb9fb5 2565static int setup_journal(const char *directory) {
0f5e1382 2566 _cleanup_free_ char *d = NULL;
b2238e38
LP		 2567 const char *dirname, *p, *q;
2568 sd_id128_t this_id;
8054d749 2569 bool try;
57fb9fb5
LP		 2570 int r;
2571
df9a75e4
LP		 2572 /* Don't link journals in ephemeral mode */
2573 if (arg_ephemeral)
2574 return 0;
2575
8054d749
LP		 2576 if (arg_link_journal == LINK_NO)
2577 return 0;
2578
2579 try = arg_link_journal_try || arg_link_journal == LINK_AUTO;
2580
4d680aee 2581 r = sd_id128_get_machine(&this_id);
f647962d
MS		 2582 if (r < 0)
2583 return log_error_errno(r, "Failed to retrieve machine ID: %m");
4d680aee 2584
e01ff70a 2585 if (sd_id128_equal(arg_uuid, this_id)) {
8054d749 2586 log_full(try ? LOG_WARNING : LOG_ERR,
85b55869 2587 "Host and machine ids are equal (%s): refusing to link journals", SD_ID128_TO_STRING(arg_uuid));
8054d749 2588 if (try)
4d680aee 2589 return 0;
df9a75e4 2590 return -EEXIST;
4d680aee
ZJS		 2591 }
2592
369ca6da
ZJS		 2593 FOREACH_STRING(dirname, "/var", "/var/log", "/var/log/journal") {
2594 r = userns_mkdir(directory, dirname, 0755, 0, 0);
2595 if (r < 0) {
2596 bool ignore = r == -EROFS && try;
2597 log_full_errno(ignore ? LOG_DEBUG : LOG_ERR, r,
2598 "Failed to create %s%s: %m", dirname, ignore ? ", ignoring" : "");
2599 return ignore ? 0 : r;
2600 }
2601 }
03cfe0d5 2602
85b55869 2603 p = strjoina("/var/log/journal/", SD_ID128_TO_STRING(arg_uuid));
03cfe0d5 2604 q = prefix_roota(directory, p);
27407a01 2605
e1873695 2606 if (path_is_mount_point(p, NULL, 0) > 0) {
8054d749
LP		 2607 if (try)
2608 return 0;
27407a01 2609
baaa35ad
ZJS		 2610 return log_error_errno(SYNTHETIC_ERRNO(EEXIST),
2611 "%s: already a mount point, refusing to use for journal", p);
57fb9fb5
LP		 2612 }
2613
e1873695 2614 if (path_is_mount_point(q, NULL, 0) > 0) {
8054d749
LP		 2615 if (try)
2616 return 0;
57fb9fb5 2617
baaa35ad
ZJS		 2618 return log_error_errno(SYNTHETIC_ERRNO(EEXIST),
2619 "%s: already a mount point, refusing to use for journal", q);
57fb9fb5
LP		 2620 }
2621
2622 r = readlink_and_make_absolute(p, &d);
2623 if (r >= 0) {
3742095b 2624 if (IN_SET(arg_link_journal, LINK_GUEST, LINK_AUTO) &&
57fb9fb5
LP		 2625 path_equal(d, q)) {
2626
03cfe0d5 2627 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 2628 if (r < 0)
709f6e46 2629 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 2630 return 0;
57fb9fb5
LP		 2631 }
2632
4a62c710
MS		 2633 if (unlink(p) < 0)
2634 return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
57fb9fb5
LP		 2635 } else if (r == -EINVAL) {
2636
2637 if (arg_link_journal == LINK_GUEST &&
2638 rmdir(p) < 0) {
2639
27407a01
ZJS		 2640 if (errno == ENOTDIR) {
2641 log_error("%s already exists and is neither a symlink nor a directory", p);
2642 return r;
4314d33f
MS		 2643 } else
2644 return log_error_errno(errno, "Failed to remove %s: %m", p);
57fb9fb5 2645 }
4314d33f
MS		 2646 } else if (r != -ENOENT)
2647 return log_error_errno(r, "readlink(%s) failed: %m", p);
57fb9fb5
LP		 2648
2649 if (arg_link_journal == LINK_GUEST) {
2650
2651 if (symlink(q, p) < 0) {
8054d749 2652 if (try) {
56f64d95 2653 log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
574edc90 2654 return 0;
4314d33f
MS		 2655 } else
2656 return log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
57fb9fb5
LP		 2657 }
2658
03cfe0d5 2659 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 2660 if (r < 0)
709f6e46 2661 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 2662 return 0;
57fb9fb5
LP		 2663 }
2664
2665 if (arg_link_journal == LINK_HOST) {
ccddd104 2666 /* don't create parents here — if the host doesn't have
574edc90 2667 * permanent journal set up, don't force it here */
ba8e6c4d 2668
3f692e2e 2669 r = RET_NERRNO(mkdir(p, 0755));
dae8b82e 2670 if (r < 0 && r != -EEXIST) {
8054d749 2671 if (try) {
dae8b82e 2672 log_debug_errno(r, "Failed to create %s, skipping journal setup: %m", p);
574edc90 2673 return 0;
4314d33f 2674 } else
dae8b82e 2675 return log_error_errno(r, "Failed to create %s: %m", p);
57fb9fb5
LP		 2676 }
2677
27407a01
ZJS		 2678 } else if (access(p, F_OK) < 0)
2679 return 0;
57fb9fb5 2680
cdb2b9d0
LP		 2681 if (dir_is_empty(q) == 0)
2682 log_warning("%s is not empty, proceeding anyway.", q);
2683
03cfe0d5 2684 r = userns_mkdir(directory, p, 0755, 0, 0);
709f6e46
MS		 2685 if (r < 0)
2686 return log_error_errno(r, "Failed to create %s: %m", q);
57fb9fb5 2687
511a8cfe 2688 r = mount_nofollow_verbose(LOG_DEBUG, p, q, NULL, MS_BIND, NULL);
60e76d48 2689 if (r < 0)
4a62c710 2690 return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
57fb9fb5 2691
27407a01 2692 return 0;
57fb9fb5
LP		 2693}
2694
de40a303
LP		 2695static int drop_capabilities(uid_t uid) {
2696 CapabilityQuintet q;
2697
2698 /* Let's initialize all five capability sets to something valid. If the quintet was configured via
2699 * OCI use that, but fill in missing bits. If it wasn't then derive the quintet in full from
2700 * arg_caps_retain. */
2701
2702 if (capability_quintet_is_set(&arg_full_capabilities)) {
2703 q = arg_full_capabilities;
2704
f5fbe71d 2705 if (q.bounding == UINT64_MAX)
de40a303
LP		 2706 q.bounding = uid == 0 ? arg_caps_retain : 0;
2707
f5fbe71d 2708 if (q.effective == UINT64_MAX)
de40a303
LP		 2709 q.effective = uid == 0 ? q.bounding : 0;
2710
f5fbe71d 2711 if (q.inheritable == UINT64_MAX)
88fc9c9b 2712 q.inheritable = uid == 0 ? q.bounding : arg_caps_ambient;
de40a303 2713
f5fbe71d 2714 if (q.permitted == UINT64_MAX)
88fc9c9b 2715 q.permitted = uid == 0 ? q.bounding : arg_caps_ambient;
de40a303 2716
f5fbe71d 2717 if (q.ambient == UINT64_MAX && ambient_capabilities_supported())
88fc9c9b 2718 q.ambient = arg_caps_ambient;
f66ad460
AZ		 2719
2720 if (capability_quintet_mangle(&q))
2721 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Cannot set capabilities that are not in the current bounding set.");
2722
2723 } else {
de40a303
LP		 2724 q = (CapabilityQuintet) {
2725 .bounding = arg_caps_retain,
2726 .effective = uid == 0 ? arg_caps_retain : 0,
88fc9c9b
TH		 2727 .inheritable = uid == 0 ? arg_caps_retain : arg_caps_ambient,
2728 .permitted = uid == 0 ? arg_caps_retain : arg_caps_ambient,
f5fbe71d 2729 .ambient = ambient_capabilities_supported() ? arg_caps_ambient : UINT64_MAX,
de40a303
LP		 2730 };
2731
f66ad460
AZ		 2732 /* If we're not using OCI, proceed with mangled capabilities (so we don't error out)
2733 * in order to maintain the same behavior as systemd < 242. */
2734 if (capability_quintet_mangle(&q))
0ccdaa79
JT		 2735 log_full(arg_quiet ? LOG_DEBUG : LOG_WARNING,
2736 "Some capabilities will not be set because they are not in the current bounding set.");
f66ad460
AZ		 2737
2738 }
2739
de40a303 2740 return capability_quintet_enforce(&q);
88213476
LP		 2741}
2742
db999e0f
LP		 2743static int reset_audit_loginuid(void) {
2744 _cleanup_free_ char *p = NULL;
2745 int r;
2746
0c582db0 2747 if ((arg_clone_ns_flags & CLONE_NEWPID) == 0)
db999e0f
LP		 2748 return 0;
2749
2750 r = read_one_line_file("/proc/self/loginuid", &p);
13e8ceb8 2751 if (r == -ENOENT)
db999e0f 2752 return 0;
f647962d
MS		 2753 if (r < 0)
2754 return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
db999e0f
LP		 2755
2756 /* Already reset? */
2757 if (streq(p, "4294967295"))
2758 return 0;
2759
57512c89 2760 r = write_string_file("/proc/self/loginuid", "4294967295", WRITE_STRING_FILE_DISABLE_BUFFER);
db999e0f 2761 if (r < 0) {
10a87006
LP		 2762 log_error_errno(r,
2763 "Failed to reset audit login UID. This probably means that your kernel is too\n"
2764 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
2765 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
2766 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
2767 "using systemd-nspawn. Sleeping for 5s... (%m)");
77b6e194 2768
db999e0f 2769 sleep(5);
77b6e194 2770 }
db999e0f
LP		 2771
2772 return 0;
77b6e194
LP		 2773}
2774
785890ac
LP		 2775static int setup_propagate(const char *root) {
2776 const char *p, *q;
709f6e46 2777 int r;
785890ac
LP		 2778
2779 (void) mkdir_p("/run/systemd/nspawn/", 0755);
2780 (void) mkdir_p("/run/systemd/nspawn/propagate", 0600);
63c372cb 2781 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
785890ac
LP		 2782 (void) mkdir_p(p, 0600);
2783
5a27b395 2784 r = userns_mkdir(root, "/run/host", 0755, 0, 0);
709f6e46 2785 if (r < 0)
5a27b395 2786 return log_error_errno(r, "Failed to create /run/host: %m");
03cfe0d5 2787
5a27b395 2788 r = userns_mkdir(root, "/run/host/incoming", 0600, 0, 0);
709f6e46 2789 if (r < 0)
5a27b395 2790 return log_error_errno(r, "Failed to create /run/host/incoming: %m");
03cfe0d5 2791
5a27b395 2792 q = prefix_roota(root, "/run/host/incoming");
511a8cfe 2793 r = mount_nofollow_verbose(LOG_ERR, p, q, NULL, MS_BIND, NULL);
60e76d48
ZJS		 2794 if (r < 0)
2795 return r;
785890ac 2796
511a8cfe 2797 r = mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
60e76d48
ZJS		 2798 if (r < 0)
2799 return r;
785890ac 2800
5a27b395 2801 /* machined will MS_MOVE into that directory, and that's only supported for non-shared mounts. */
511a8cfe 2802 return mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_SLAVE, NULL);
785890ac
LP		 2803}
2804
317feb4d 2805static int setup_machine_id(const char *directory) {
691675ba
LP		 2806 const char *etc_machine_id;
2807 sd_id128_t id;
3bbaff3e 2808 int r;
e01ff70a 2809
317feb4d
LP		 2810 /* If the UUID in the container is already set, then that's what counts, and we use. If it isn't set, and the
2811 * caller passed --uuid=, then we'll pass it in the $container_uuid env var to PID 1 of the container. The
2812 * assumption is that PID 1 will then write it to /etc/machine-id to make it persistent. If --uuid= is not
2813 * passed we generate a random UUID, and pass it via $container_uuid. In effect this means that /etc/machine-id
2814 * in the container and our idea of the container UUID will always be in sync (at least if PID 1 in the
2815 * container behaves nicely). */
2816
e01ff70a
MS		 2817 etc_machine_id = prefix_roota(directory, "/etc/machine-id");
2818
c5fbeedb 2819 r = id128_read(etc_machine_id, ID128_PLAIN_OR_UNINIT, &id);
317feb4d
LP		 2820 if (r < 0) {
2821 if (!IN_SET(r, -ENOENT, -ENOMEDIUM)) /* If the file is missing or empty, we don't mind */
2822 return log_error_errno(r, "Failed to read machine ID from container image: %m");
691675ba 2823
317feb4d
LP		 2824 if (sd_id128_is_null(arg_uuid)) {
2825 r = sd_id128_randomize(&arg_uuid);
2826 if (r < 0)
2827 return log_error_errno(r, "Failed to acquire randomized machine UUID: %m");
2828 }
2829 } else {
baaa35ad
ZJS		 2830 if (sd_id128_is_null(id))
2831 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
2832 "Machine ID in container image is zero, refusing.");
e01ff70a 2833
317feb4d
LP		 2834 arg_uuid = id;
2835 }
691675ba 2836
e01ff70a
MS		 2837 return 0;
2838}
2839
7336138e
LP		 2840static int recursive_chown(const char *directory, uid_t shift, uid_t range) {
2841 int r;
2842
2843 assert(directory);
2844
6c045a99 2845 if (arg_userns_mode == USER_NAMESPACE_NO || arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_CHOWN)
7336138e
LP		 2846 return 0;
2847
2848 r = path_patch_uid(directory, arg_uid_shift, arg_uid_range);
2849 if (r == -EOPNOTSUPP)
2850 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.");
2851 if (r == -EBADE)
2852 return log_error_errno(r, "Upper 16 bits of root directory UID and GID do not match.");
2853 if (r < 0)
2854 return log_error_errno(r, "Failed to adjust UID/GID shift of OS tree: %m");
2855 if (r == 0)
2856 log_debug("Root directory of image is already owned by the right UID/GID range, skipping recursive chown operation.");
2857 else
2858 log_debug("Patched directory tree to match UID/GID range.");
2859
2860 return r;
2861}
2862
113cea80 2863/*
6d416b9c
LS		 2864 * Return values:
2865 * < 0 : wait_for_terminate() failed to get the state of the
2866 * container, the container was terminated by a signal, or
2867 * failed for an unknown reason. No change is made to the
2868 * container argument.
2869 * > 0 : The program executed in the container terminated with an
2870 * error. The exit code of the program executed in the
919699ec
LP		 2871 * container is returned. The container argument has been set
2872 * to CONTAINER_TERMINATED.
6d416b9c
LS		 2873 * 0 : The container is being rebooted, has been shut down or exited
2874 * successfully. The container argument has been set to either
2875 * CONTAINER_TERMINATED or CONTAINER_REBOOTED.
113cea80 2876 *
6d416b9c
LS		 2877 * That is, success is indicated by a return value of zero, and an
2878 * error is indicated by a non-zero value.
113cea80
DH		 2879 */
2880static int wait_for_container(pid_t pid, ContainerStatus *container) {
113cea80 2881 siginfo_t status;
919699ec 2882 int r;
113cea80
DH		 2883
2884 r = wait_for_terminate(pid, &status);
f647962d
MS		 2885 if (r < 0)
2886 return log_warning_errno(r, "Failed to wait for container: %m");
113cea80
DH		 2887
2888 switch (status.si_code) {
fddbb89c 2889
113cea80 2890 case CLD_EXITED:
b5a2179b 2891 if (status.si_status == 0)
919699ec 2892 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s exited successfully.", arg_machine);
b5a2179b 2893 else
919699ec 2894 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s failed with error code %i.", arg_machine, status.si_status);
fddbb89c 2895
919699ec
LP		 2896 *container = CONTAINER_TERMINATED;
2897 return status.si_status;
113cea80
DH		 2898
2899 case CLD_KILLED:
2900 if (status.si_status == SIGINT) {
919699ec 2901 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s has been shut down.", arg_machine);
113cea80 2902 *container = CONTAINER_TERMINATED;
919699ec
LP		 2903 return 0;
2904
113cea80 2905 } else if (status.si_status == SIGHUP) {
919699ec 2906 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s is being rebooted.", arg_machine);
113cea80 2907 *container = CONTAINER_REBOOTED;
919699ec 2908 return 0;
113cea80 2909 }
919699ec 2910
4831981d 2911 _fallthrough_;
113cea80 2912 case CLD_DUMPED:
baaa35ad
ZJS		 2913 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2914 "Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
113cea80
DH		 2915
2916 default:
baaa35ad
ZJS		 2917 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2918 "Container %s failed due to unknown reason.", arg_machine);
113cea80 2919 }
113cea80
DH		 2920}
2921
023fb90b
LP		 2922static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
2923 pid_t pid;
2924
4a0b58c4 2925 pid = PTR_TO_PID(userdata);
023fb90b 2926 if (pid > 0) {
c6c8f6e2 2927 if (kill(pid, arg_kill_signal) >= 0) {
023fb90b
LP		 2928 log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
2929 sd_event_source_set_userdata(s, NULL);
2930 return 0;
2931 }
2932 }
2933
2934 sd_event_exit(sd_event_source_get_event(s), 0);
2935 return 0;
2936}
2937
6916b164 2938static int on_sigchld(sd_event_source *s, const struct signalfd_siginfo *ssi, void *userdata) {
abdb9b08
LP		 2939 pid_t pid;
2940
2941 assert(s);
2942 assert(ssi);
2943
2944 pid = PTR_TO_PID(userdata);
2945
6916b164
AU		 2946 for (;;) {
2947 siginfo_t si = {};
abdb9b08 2948
6916b164
AU		 2949 if (waitid(P_ALL, 0, &si, WNOHANG|WNOWAIT|WEXITED) < 0)
2950 return log_error_errno(errno, "Failed to waitid(): %m");
2951 if (si.si_pid == 0) /* No pending children. */
2952 break;
abdb9b08 2953 if (si.si_pid == pid) {
6916b164
AU		 2954 /* The main process we care for has exited. Return from
2955 * signal handler but leave the zombie. */
2956 sd_event_exit(sd_event_source_get_event(s), 0);
2957 break;
2958 }
abdb9b08 2959
6916b164
AU		 2960 /* Reap all other children. */
2961 (void) waitid(P_PID, si.si_pid, &si, WNOHANG|WEXITED);
2962 }
2963
2964 return 0;
2965}
2966
abdb9b08
LP		 2967static int on_request_stop(sd_bus_message *m, void *userdata, sd_bus_error *error) {
2968 pid_t pid;
2969
2970 assert(m);
2971
2972 pid = PTR_TO_PID(userdata);
2973
2974 if (arg_kill_signal > 0) {
2975 log_info("Container termination requested. Attempting to halt container.");
2976 (void) kill(pid, arg_kill_signal);
2977 } else {
2978 log_info("Container termination requested. Exiting.");
2979 sd_event_exit(sd_bus_get_event(sd_bus_message_get_bus(m)), 0);
2980 }
2981
2982 return 0;
2983}
2984
ec16945e 2985static int determine_names(void) {
1b9cebf6 2986 int r;
ec16945e 2987
c1521918
LP		 2988 if (arg_template && !arg_directory && arg_machine) {
2989
2990 /* If --template= was specified then we should not
2991 * search for a machine, but instead create a new one
2992 * in /var/lib/machine. */
2993
657ee2d8 2994 arg_directory = path_join("/var/lib/machines", arg_machine);
c1521918
LP		 2995 if (!arg_directory)
2996 return log_oom();
2997 }
2998
ec16945e 2999 if (!arg_image && !arg_directory) {
1b9cebf6
LP		 3000 if (arg_machine) {
3001 _cleanup_(image_unrefp) Image *i = NULL;
3002
d577d4a4 3003 r = image_find(IMAGE_MACHINE, arg_machine, NULL, &i);
3a6ce860
LP		 3004 if (r == -ENOENT)
3005 return log_error_errno(r, "No image for machine '%s'.", arg_machine);
1b9cebf6
LP		 3006 if (r < 0)
3007 return log_error_errno(r, "Failed to find image for machine '%s': %m", arg_machine);
1b9cebf6 3008
eb38edce 3009 if (IN_SET(i->type, IMAGE_RAW, IMAGE_BLOCK))
0f03c2a4 3010 r = free_and_strdup(&arg_image, i->path);
1b9cebf6 3011 else
0f03c2a4 3012 r = free_and_strdup(&arg_directory, i->path);
1b9cebf6 3013 if (r < 0)
0f3be6ca 3014 return log_oom();
1b9cebf6 3015
aee327b8
LP		 3016 if (!arg_ephemeral)
3017 arg_read_only = arg_read_only || i->read_only;
d7249575
LP		 3018 } else {
3019 r = safe_getcwd(&arg_directory);
3020 if (r < 0)
3021 return log_error_errno(r, "Failed to determine current directory: %m");
3022 }
ec16945e 3023
c6147113
LP		 3024 if (!arg_directory && !arg_image)
3025 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to determine path, please use -D or -i.");
ec16945e
LP		 3026 }
3027
3028 if (!arg_machine) {
b9ba4dab
LP		 3029 if (arg_directory && path_equal(arg_directory, "/"))
3030 arg_machine = gethostname_malloc();
e9b88a6d
LP		 3031 else if (arg_image) {
3032 char *e;
4827ab48 3033
e9b88a6d 3034 arg_machine = strdup(basename(arg_image));
4827ab48 3035
e9b88a6d
LP		 3036 /* Truncate suffix if there is one */
3037 e = endswith(arg_machine, ".raw");
3038 if (e)
3039 *e = 0;
3040 } else
3041 arg_machine = strdup(basename(arg_directory));
ec16945e
LP		 3042 if (!arg_machine)
3043 return log_oom();
3044
ae691c1d 3045 hostname_cleanup(arg_machine);
52ef5dd7 3046 if (!hostname_is_valid(arg_machine, 0))
c6147113 3047 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to determine machine name automatically, please use -M.");
b9ba4dab 3048
e9b88a6d
LP		 3049 /* Add a random suffix when this is an ephemeral machine, so that we can run many
3050 * instances at once without manually having to specify -M each time. */
3051 if (arg_ephemeral)
3052 if (strextendf(&arg_machine, "-%016" PRIx64, random_u64()) < 0)
b9ba4dab 3053 return log_oom();
ec16945e
LP		 3054 }
3055
3056 return 0;
3057}
3058
8d4aa2bb 3059static int chase_symlinks_and_update(char **p, unsigned flags) {
3f342ec4
LP		 3060 char *chased;
3061 int r;
3062
3063 assert(p);
3064
3065 if (!*p)
3066 return 0;
3067
a5648b80 3068 r = chase_symlinks(*p, NULL, flags, &chased, NULL);
3f342ec4
LP		 3069 if (r < 0)
3070 return log_error_errno(r, "Failed to resolve path %s: %m", *p);
3071
a5648b80 3072 return free_and_replace(*p, chased);
3f342ec4
LP		 3073}
3074
03cfe0d5 3075static int determine_uid_shift(const char *directory) {
6dac160c 3076
0de7acce 3077 if (arg_userns_mode == USER_NAMESPACE_NO) {
03cfe0d5 3078 arg_uid_shift = 0;
6dac160c 3079 return 0;
03cfe0d5 3080 }
6dac160c
LP		 3081
3082 if (arg_uid_shift == UID_INVALID) {
3083 struct stat st;
3084
993da6d4
LP		 3085 /* Read the UID shift off the image. Maybe we can reuse this to avoid chowning. */
3086
3087 if (stat(directory, &st) < 0)
03cfe0d5 3088 return log_error_errno(errno, "Failed to determine UID base of %s: %m", directory);
6dac160c
LP		 3089
3090 arg_uid_shift = st.st_uid & UINT32_C(0xffff0000);
3091
baaa35ad
ZJS		 3092 if (arg_uid_shift != (st.st_gid & UINT32_C(0xffff0000)))
3093 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
3094 "UID and GID base of %s don't match.", directory);
6dac160c
LP		 3095
3096 arg_uid_range = UINT32_C(0x10000);
f61c7f88
LP		 3097
3098 if (arg_uid_shift != 0) {
3099 /* If the image is shifted already, then we'll fall back to classic chowning, for
3100 * compatibility (and simplicity), or refuse if mapping is explicitly requested. */
3101
3102 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_AUTO) {
3103 log_debug("UID base of %s is non-zero, not using UID mapping.", directory);
3104 arg_userns_ownership = USER_NAMESPACE_OWNERSHIP_CHOWN;
3105 } else if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_MAP)
3106 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
3107 "UID base of %s is not zero, UID mapping not supported.", directory);
3108 }
6dac160c
LP		 3109 }
3110
58e13de5
LP		 3111 if (!userns_shift_range_valid(arg_uid_shift, arg_uid_range))
3112 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "UID base too high for UID range.");
6dac160c 3113
6dac160c
LP		 3114 return 0;
3115}
3116
de40a303
LP		 3117static unsigned long effective_clone_ns_flags(void) {
3118 unsigned long flags = arg_clone_ns_flags;
3119
3120 if (arg_private_network)
3121 flags |= CLONE_NEWNET;
3122 if (arg_use_cgns)
3123 flags |= CLONE_NEWCGROUP;
3124 if (arg_userns_mode != USER_NAMESPACE_NO)
3125 flags |= CLONE_NEWUSER;
3126
3127 return flags;
3128}
3129
3130static int patch_sysctl(void) {
3131
3132 /* This table is inspired by runc's sysctl() function */
3133 static const struct {
3134 const char *key;
3135 bool prefix;
3136 unsigned long clone_flags;
3137 } safe_sysctl[] = {
3138 { "kernel.hostname", false, CLONE_NEWUTS },
3139 { "kernel.domainname", false, CLONE_NEWUTS },
3140 { "kernel.msgmax", false, CLONE_NEWIPC },
3141 { "kernel.msgmnb", false, CLONE_NEWIPC },
3142 { "kernel.msgmni", false, CLONE_NEWIPC },
3143 { "kernel.sem", false, CLONE_NEWIPC },
3144 { "kernel.shmall", false, CLONE_NEWIPC },
3145 { "kernel.shmmax", false, CLONE_NEWIPC },
3146 { "kernel.shmmni", false, CLONE_NEWIPC },
3147 { "fs.mqueue.", true, CLONE_NEWIPC },
3148 { "net.", true, CLONE_NEWNET },
3149 };
3150
3151 unsigned long flags;
3152 char **k, **v;
3153 int r;
3154
3155 flags = effective_clone_ns_flags();
3156
3157 STRV_FOREACH_PAIR(k, v, arg_sysctl) {
3158 bool good = false;
3159 size_t i;
3160
3161 for (i = 0; i < ELEMENTSOF(safe_sysctl); i++) {
3162
3163 if (!FLAGS_SET(flags, safe_sysctl[i].clone_flags))
3164 continue;
3165
3166 if (safe_sysctl[i].prefix)
3167 good = startswith(*k, safe_sysctl[i].key);
3168 else
3169 good = streq(*k, safe_sysctl[i].key);
3170
3171 if (good)
3172 break;
3173 }
3174
c6147113
LP		 3175 if (!good)
3176 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Refusing to write to sysctl '%s', as it is not safe in the selected namespaces.", *k);
de40a303
LP		 3177
3178 r = sysctl_write(*k, *v);
3179 if (r < 0)
3180 return log_error_errno(r, "Failed to write sysctl '%s': %m", *k);
3181 }
3182
3183 return 0;
3184}
3185
03cfe0d5
LP		 3186static int inner_child(
3187 Barrier *barrier,
3188 const char *directory,
3189 bool secondary,
3190 int kmsg_socket,
3191 int rtnl_socket,
3acc84eb 3192 int master_pty_socket,
e1bb4b0d
LB		 3193 FDSet *fds,
3194 char **os_release_pairs) {
69c79d3c 3195
03cfe0d5 3196 _cleanup_free_ char *home = NULL;
88614c8a 3197 size_t n_env = 1;
4ab3d29f
ZJS		 3198 char *envp[] = {
3199 (char*) "PATH=" DEFAULT_PATH_COMPAT,
6aadfa4c 3200 NULL, /* container */
03cfe0d5
LP		 3201 NULL, /* TERM */
3202 NULL, /* HOME */
3203 NULL, /* USER */
3204 NULL, /* LOGNAME */
3205 NULL, /* container_uuid */
3206 NULL, /* LISTEN_FDS */
3207 NULL, /* LISTEN_PID */
9c1e04d0 3208 NULL, /* NOTIFY_SOCKET */
3652872a 3209 NULL, /* CREDENTIALS_DIRECTORY */
03cfe0d5
LP		 3210 NULL
3211 };
1a68e1e5 3212 const char *exec_target;
2371271c 3213 _cleanup_strv_free_ char **env_use = NULL;
de40a303 3214 int r, which_failed;
88213476 3215
b37469d7
LP		 3216 /* This is the "inner" child process, i.e. the one forked off by the "outer" child process, which is the one
3217 * the container manager itself forked off. At the time of clone() it gained its own CLONE_NEWNS, CLONE_NEWPID,
3218 * CLONE_NEWUTS, CLONE_NEWIPC, CLONE_NEWUSER namespaces. Note that it has its own CLONE_NEWNS namespace,
3219 * separate from the CLONE_NEWNS created for the "outer" child, and also separate from the host's CLONE_NEWNS
3220 * namespace. The reason for having two levels of CLONE_NEWNS namespaces is that the "inner" one is owned by
3221 * the CLONE_NEWUSER namespace of the container, while the "outer" one is owned by the host's CLONE_NEWUSER
3222 * namespace.
3223 *
3224 * Note at this point we have no CLONE_NEWNET namespace yet. We'll acquire that one later through
3225 * unshare(). See below. */
3226
03cfe0d5
LP		 3227 assert(barrier);
3228 assert(directory);
3229 assert(kmsg_socket >= 0);
88213476 3230
de40a303
LP		 3231 log_debug("Inner child is initializing.");
3232
0de7acce 3233 if (arg_userns_mode != USER_NAMESPACE_NO) {
03cfe0d5
LP		 3234 /* Tell the parent, that it now can write the UID map. */
3235 (void) barrier_place(barrier); /* #1 */
7027ff61 3236
03cfe0d5 3237 /* Wait until the parent wrote the UID map */
baaa35ad 3238 if (!barrier_place_and_sync(barrier)) /* #2 */
2a2e78e9 3239 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Parent died too early");
88213476 3240
2a2e78e9
LP		 3241 /* Become the new root user inside our namespace */
3242 r = reset_uid_gid();
3243 if (r < 0)
3244 return log_error_errno(r, "Couldn't become new root: %m");
3245
3246 /* Creating a new user namespace means all MS_SHARED mounts become MS_SLAVE. Let's put them
3247 * back to MS_SHARED here, since that's what we want as defaults. (This will not reconnect
3248 * propagation, but simply create new peer groups for all our mounts). */
511a8cfe 3249 r = mount_follow_verbose(LOG_ERR, NULL, "/", NULL, MS_SHARED|MS_REC, NULL);
2a2e78e9
LP		 3250 if (r < 0)
3251 return r;
3252 }
6d66bd3b 3253
0de7acce 3254 r = mount_all(NULL,
4f086aab 3255 arg_mount_settings | MOUNT_IN_USERNS,
0de7acce 3256 arg_uid_shift,
0de7acce 3257 arg_selinux_apifs_context);
03cfe0d5
LP		 3258 if (r < 0)
3259 return r;
3260
04413780
ZJS		 3261 if (!arg_network_namespace_path && arg_private_network) {
3262 r = unshare(CLONE_NEWNET);
3263 if (r < 0)
3264 return log_error_errno(errno, "Failed to unshare network namespace: %m");
75116558
PS		 3265
3266 /* Tell the parent that it can setup network interfaces. */
3267 (void) barrier_place(barrier); /* #3 */
04413780
ZJS		 3268 }
3269
4f086aab 3270 r = mount_sysfs(NULL, arg_mount_settings);
d8fc6a00
LP		 3271 if (r < 0)
3272 return r;
3273
03cfe0d5
LP		 3274 /* Wait until we are cgroup-ified, so that we
3275 * can mount the right cgroup path writable */
baaa35ad
ZJS		 3276 if (!barrier_place_and_sync(barrier)) /* #4 */
3277 return log_error_errno(SYNTHETIC_ERRNO(ESRCH),
3278 "Parent died too early");
88213476 3279
489fae52 3280 if (arg_use_cgns) {
0996ef00
CB		 3281 r = unshare(CLONE_NEWCGROUP);
3282 if (r < 0)
04413780 3283 return log_error_errno(errno, "Failed to unshare cgroup namespace: %m");
0996ef00
CB		 3284 r = mount_cgroups(
3285 "",
3286 arg_unified_cgroup_hierarchy,
3287 arg_userns_mode != USER_NAMESPACE_NO,
3288 arg_uid_shift,
3289 arg_uid_range,
5a8ff0e6 3290 arg_selinux_apifs_context,
ada54120 3291 true);
1433e0f2 3292 } else
0996ef00 3293 r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy);
1433e0f2
LP		 3294 if (r < 0)
3295 return r;
ec16945e 3296
1e4f1671 3297 r = setup_boot_id();
03cfe0d5
LP		 3298 if (r < 0)
3299 return r;
ec16945e 3300
1e4f1671 3301 r = setup_kmsg(kmsg_socket);
03cfe0d5
LP		 3302 if (r < 0)
3303 return r;
3304 kmsg_socket = safe_close(kmsg_socket);
ec16945e 3305
de40a303
LP		 3306 r = mount_custom(
3307 "/",
3308 arg_custom_mounts,
3309 arg_n_custom_mounts,
de40a303 3310 0,
c0c8f718 3311 0,
de40a303 3312 arg_selinux_apifs_context,
5f0a6347 3313 MOUNT_NON_ROOT_ONLY | MOUNT_IN_USERNS);
de40a303
LP		 3314 if (r < 0)
3315 return r;
3316
03cfe0d5
LP		 3317 if (setsid() < 0)
3318 return log_error_errno(errno, "setsid() failed: %m");
3319
3320 if (arg_private_network)
df883de9 3321 (void) loopback_setup();
03cfe0d5 3322
7a8f6325
LP		 3323 if (arg_expose_ports) {
3324 r = expose_port_send_rtnl(rtnl_socket);
3325 if (r < 0)
3326 return r;
3327 rtnl_socket = safe_close(rtnl_socket);
3328 }
03cfe0d5 3329
3acc84eb 3330 if (arg_console_mode != CONSOLE_PIPE) {
cd132992 3331 _cleanup_close_ int master = -1;
3acc84eb
FB		 3332 _cleanup_free_ char *console = NULL;
3333
3334 /* Allocate a pty and make it available as /dev/console. */
dc98caea 3335 master = openpt_allocate(O_RDWR|O_NONBLOCK, &console);
3acc84eb 3336 if (master < 0)
dc98caea 3337 return log_error_errno(master, "Failed to allocate a pty: %m");
3acc84eb
FB		 3338
3339 r = setup_dev_console(console);
3340 if (r < 0)
105a1a36 3341 return log_error_errno(r, "Failed to set up /dev/console: %m");
3acc84eb
FB		 3342
3343 r = send_one_fd(master_pty_socket, master, 0);
3344 if (r < 0)
3345 return log_error_errno(r, "Failed to send master fd: %m");
3346 master_pty_socket = safe_close(master_pty_socket);
3347
3348 r = setup_stdio_as_dev_console();
3349 if (r < 0)
3350 return r;
3351 }
3352
de40a303
LP		 3353 r = patch_sysctl();
3354 if (r < 0)
3355 return r;
3356
81f345df
LP		 3357 if (arg_oom_score_adjust_set) {
3358 r = set_oom_score_adjust(arg_oom_score_adjust);
3359 if (r < 0)
3360 return log_error_errno(r, "Failed to adjust OOM score: %m");
3361 }
3362
0985c7c4
ZJS		 3363 if (arg_cpu_set.set)
3364 if (sched_setaffinity(0, arg_cpu_set.allocated, arg_cpu_set.set) < 0)
d107bb7d
LP		 3365 return log_error_errno(errno, "Failed to set CPU affinity: %m");
3366
c818eef1 3367 (void) setup_hostname();
03cfe0d5 3368
050f7277 3369 if (arg_personality != PERSONALITY_INVALID) {
21022b9d
LP		 3370 r = safe_personality(arg_personality);
3371 if (r < 0)
3372 return log_error_errno(r, "personality() failed: %m");
03cfe0d5 3373 } else if (secondary) {
21022b9d
LP		 3374 r = safe_personality(PER_LINUX32);
3375 if (r < 0)
3376 return log_error_errno(r, "personality() failed: %m");
03cfe0d5
LP		 3377 }
3378
de40a303
LP		 3379 r = setrlimit_closest_all((const struct rlimit *const*) arg_rlimit, &which_failed);
3380 if (r < 0)
3381 return log_error_errno(r, "Failed to apply resource limit RLIMIT_%s: %m", rlimit_to_string(which_failed));
3382
3383#if HAVE_SECCOMP
3384 if (arg_seccomp) {
3385
3386 if (is_seccomp_available()) {
3387
3388 r = seccomp_load(arg_seccomp);
7bc5e0b1 3389 if (ERRNO_IS_SECCOMP_FATAL(r))
de40a303
LP		 3390 return log_error_errno(r, "Failed to install seccomp filter: %m");
3391 if (r < 0)
3392 log_debug_errno(r, "Failed to install seccomp filter: %m");
3393 }
3394 } else
3395#endif
3396 {
6b000af4 3397 r = setup_seccomp(arg_caps_retain, arg_syscall_allow_list, arg_syscall_deny_list);
de40a303
LP		 3398 if (r < 0)
3399 return r;
3400 }
3401
4a4654e0 3402 if (arg_suppress_sync) {
20e458ae 3403#if HAVE_SECCOMP
4a4654e0
LP		 3404 r = seccomp_suppress_sync();
3405 if (r < 0)
3406 log_debug_errno(r, "Failed to install sync() suppression seccomp filter, ignoring: %m");
20e458ae 3407#else
2db32618 3408 log_debug("systemd is built without SECCOMP support. Ignoring --suppress-sync= command line option and SuppressSync= setting.");
20e458ae 3409#endif
4a4654e0
LP		 3410 }
3411
349cc4a5 3412#if HAVE_SELINUX
03cfe0d5 3413 if (arg_selinux_context)
2ed96880 3414 if (setexeccon(arg_selinux_context) < 0)
03cfe0d5
LP		 3415 return log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context);
3416#endif
3417
de40a303
LP		 3418 /* Make sure we keep the caps across the uid/gid dropping, so that we can retain some selected caps
3419 * if we need to later on. */
3420 if (prctl(PR_SET_KEEPCAPS, 1) < 0)
3421 return log_error_errno(errno, "Failed to set PR_SET_KEEPCAPS: %m");
3422
3423 if (uid_is_valid(arg_uid) || gid_is_valid(arg_gid))
3462d773 3424 r = change_uid_gid_raw(arg_uid, arg_gid, arg_supplementary_gids, arg_n_supplementary_gids, arg_console_mode != CONSOLE_PIPE);
de40a303 3425 else
3462d773 3426 r = change_uid_gid(arg_user, arg_console_mode != CONSOLE_PIPE, &home);
03cfe0d5
LP		 3427 if (r < 0)
3428 return r;
3429
de40a303
LP		 3430 r = drop_capabilities(getuid());
3431 if (r < 0)
3432 return log_error_errno(r, "Dropping capabilities failed: %m");
3433
66edd963
LP		 3434 if (arg_no_new_privileges)
3435 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0)
3436 return log_error_errno(errno, "Failed to disable new privileges: %m");
3437
6aadfa4c
ILG		 3438 /* LXC sets container=lxc, so follow the scheme here */
3439 envp[n_env++] = strjoina("container=", arg_container_service_name);
3440
03cfe0d5
LP		 3441 envp[n_env] = strv_find_prefix(environ, "TERM=");
3442 if (envp[n_env])
313cefa1 3443 n_env++;
03cfe0d5 3444
de40a303 3445 if (home || !uid_is_valid(arg_uid) || arg_uid == 0)
4ab3d29f 3446 if (asprintf(envp + n_env++, "HOME=%s", home ?: "/root") < 0)
de40a303
LP		 3447 return log_oom();
3448
3449 if (arg_user || !uid_is_valid(arg_uid) || arg_uid == 0)
4ab3d29f
ZJS		 3450 if (asprintf(envp + n_env++, "USER=%s", arg_user ?: "root") < 0 ||
3451 asprintf(envp + n_env++, "LOGNAME=%s", arg_user ? arg_user : "root") < 0)
de40a303 3452 return log_oom();
03cfe0d5 3453
3bbaff3e 3454 assert(!sd_id128_is_null(arg_uuid));
03cfe0d5 3455
b7416360 3456 if (asprintf(envp + n_env++, "container_uuid=%s", SD_ID128_TO_UUID_STRING(arg_uuid)) < 0)
e01ff70a 3457 return log_oom();
03cfe0d5
LP		 3458
3459 if (fdset_size(fds) > 0) {
3460 r = fdset_cloexec(fds, false);
3461 if (r < 0)
3462 return log_error_errno(r, "Failed to unset O_CLOEXEC for file descriptors.");
3463
4ab3d29f
ZJS		 3464 if ((asprintf(envp + n_env++, "LISTEN_FDS=%u", fdset_size(fds)) < 0) ||
3465 (asprintf(envp + n_env++, "LISTEN_PID=1") < 0))
03cfe0d5
LP		 3466 return log_oom();
3467 }
4ab3d29f 3468 if (asprintf(envp + n_env++, "NOTIFY_SOCKET=%s", NSPAWN_NOTIFY_SOCKET_PATH) < 0)
9c1e04d0 3469 return log_oom();
03cfe0d5 3470
3652872a
LP		 3471 if (arg_n_credentials > 0) {
3472 envp[n_env] = strdup("CREDENTIALS_DIRECTORY=/run/host/credentials");
3473 if (!envp[n_env])
3474 return log_oom();
3475 n_env++;
3476 }
3477
4ab3d29f 3478 env_use = strv_env_merge(envp, os_release_pairs, arg_setenv);
2371271c
TG		 3479 if (!env_use)
3480 return log_oom();
03cfe0d5
LP		 3481
3482 /* Let the parent know that we are ready and
3483 * wait until the parent is ready with the
3484 * setup, too... */
baaa35ad 3485 if (!barrier_place_and_sync(barrier)) /* #5 */
335d2ead 3486 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Parent died too early");
03cfe0d5 3487
5f932eb9
LP		 3488 if (arg_chdir)
3489 if (chdir(arg_chdir) < 0)
3490 return log_error_errno(errno, "Failed to change to specified working directory %s: %m", arg_chdir);
3491
7732f92b 3492 if (arg_start_mode == START_PID2) {
75bf701f 3493 r = stub_pid1(arg_uuid);
7732f92b
LP		 3494 if (r < 0)
3495 return r;
3496 }
3497
335d2ead
LP		 3498 if (arg_console_mode != CONSOLE_PIPE) {
3499 /* So far our pty wasn't controlled by any process. Finally, it's time to change that, if we
3500 * are configured for that. Acquire it as controlling tty. */
3501 if (ioctl(STDIN_FILENO, TIOCSCTTY) < 0)
3502 return log_error_errno(errno, "Failed to acquire controlling TTY: %m");
3503 }
3504
de40a303
LP		 3505 log_debug("Inner child completed, invoking payload.");
3506
8ca082b4
LP		 3507 /* Now, explicitly close the log, so that we then can close all remaining fds. Closing the log explicitly first
3508 * has the benefit that the logging subsystem knows about it, and is thus ready to be reopened should we need
3509 * it again. Note that the other fds closed here are at least the locking and barrier fds. */
03cfe0d5 3510 log_close();
8ca082b4
LP		 3511 log_set_open_when_needed(true);
3512
03cfe0d5
LP		 3513 (void) fdset_close_others(fds);
3514
7732f92b 3515 if (arg_start_mode == START_BOOT) {
a5096641 3516 const char *init;
03cfe0d5
LP		 3517 char **a;
3518 size_t m;
3519
3520 /* Automatically search for the init system */
3521
75f32f04
ZJS		 3522 m = strv_length(arg_parameters);
3523 a = newa(char*, m + 2);
3524 memcpy_safe(a + 1, arg_parameters, m * sizeof(char*));
3525 a[1 + m] = NULL;
03cfe0d5 3526
a5096641
LP		 3527 FOREACH_STRING(init,
3528 "/usr/lib/systemd/systemd",
3529 "/lib/systemd/systemd",
3530 "/sbin/init") {
3531 a[0] = (char*) init;
3532 execve(a[0], a, env_use);
3533 }
ced58da7
LP		 3534
3535 exec_target = "/usr/lib/systemd/systemd, /lib/systemd/systemd, /sbin/init";
1a68e1e5 3536 } else if (!strv_isempty(arg_parameters)) {
b6b180b7
LP		 3537 const char *dollar_path;
3538
1a68e1e5 3539 exec_target = arg_parameters[0];
b6b180b7
LP		 3540
3541 /* Use the user supplied search $PATH if there is one, or DEFAULT_PATH_COMPAT if not to search the
3542 * binary. */
3543 dollar_path = strv_env_get(env_use, "PATH");
3544 if (dollar_path) {
6f646e01 3545 if (setenv("PATH", dollar_path, 1) < 0)
b6b180b7
LP		 3546 return log_error_errno(errno, "Failed to update $PATH: %m");
3547 }
3548
f757855e 3549 execvpe(arg_parameters[0], arg_parameters, env_use);
1a68e1e5 3550 } else {
5f932eb9 3551 if (!arg_chdir)
d929b0f9
ZJS		 3552 /* If we cannot change the directory, we'll end up in /, that is expected. */
3553 (void) chdir(home ?: "/root");
5f932eb9 3554
03cfe0d5
LP		 3555 execle("/bin/bash", "-bash", NULL, env_use);
3556 execle("/bin/sh", "-sh", NULL, env_use);
ced58da7
LP		 3557
3558 exec_target = "/bin/bash, /bin/sh";
03cfe0d5
LP		 3559 }
3560
8ca082b4 3561 return log_error_errno(errno, "execv(%s) failed: %m", exec_target);
03cfe0d5
LP		 3562}
3563
e96ceaba 3564static int setup_notify_child(void) {
271f518f 3565 _cleanup_close_ int fd = -1;
9c1e04d0 3566 union sockaddr_union sa = {
44ed5214
LP		 3567 .un.sun_family = AF_UNIX,
3568 .un.sun_path = NSPAWN_NOTIFY_SOCKET_PATH,
9c1e04d0
AP		 3569 };
3570 int r;
3571
3572 fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
3573 if (fd < 0)
3574 return log_error_errno(errno, "Failed to allocate notification socket: %m");
3575
3576 (void) mkdir_parents(NSPAWN_NOTIFY_SOCKET_PATH, 0755);
fbda85b0 3577 (void) sockaddr_un_unlink(&sa.un);
9c1e04d0 3578
9c1e04d0 3579 r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un));
271f518f 3580 if (r < 0)
44ed5214 3581 return log_error_errno(errno, "bind(" NSPAWN_NOTIFY_SOCKET_PATH ") failed: %m");
9c1e04d0 3582
adc7d9f0 3583 r = userns_lchown(NSPAWN_NOTIFY_SOCKET_PATH, 0, 0);
271f518f 3584 if (r < 0)
adc7d9f0 3585 return log_error_errno(r, "Failed to chown " NSPAWN_NOTIFY_SOCKET_PATH ": %m");
adc7d9f0 3586
2ff48e98 3587 r = setsockopt_int(fd, SOL_SOCKET, SO_PASSCRED, true);
271f518f 3588 if (r < 0)
2ff48e98 3589 return log_error_errno(r, "SO_PASSCRED failed: %m");
9c1e04d0 3590
271f518f 3591 return TAKE_FD(fd);
9c1e04d0
AP		 3592}
3593
03cfe0d5
LP		 3594static int outer_child(
3595 Barrier *barrier,
3596 const char *directory,
2d845785 3597 DissectedImage *dissected_image,
03cfe0d5
LP		 3598 bool secondary,
3599 int pid_socket,
e01ff70a 3600 int uuid_socket,
9c1e04d0 3601 int notify_socket,
03cfe0d5
LP		 3602 int kmsg_socket,
3603 int rtnl_socket,
825d5287 3604 int uid_shift_socket,
3acc84eb 3605 int master_pty_socket,
8199d554 3606 int unified_cgroup_hierarchy_socket,
d7bea6b6
DP		 3607 FDSet *fds,
3608 int netns_fd) {
03cfe0d5 3609
2f893044 3610 _cleanup_(bind_user_context_freep) BindUserContext *bind_user_context = NULL;
e1bb4b0d 3611 _cleanup_strv_free_ char **os_release_pairs = NULL;
bf428efb 3612 _cleanup_close_ int fd = -1;
f61c7f88 3613 bool idmap = false;
e5f10caf 3614 const char *p;
03cfe0d5
LP		 3615 pid_t pid;
3616 ssize_t l;
de40a303 3617 int r;
03cfe0d5 3618
b37469d7
LP		 3619 /* This is the "outer" child process, i.e the one forked off by the container manager itself. It already has
3620 * its own CLONE_NEWNS namespace (which was created by the clone()). It still lives in the host's CLONE_NEWPID,
3621 * CLONE_NEWUTS, CLONE_NEWIPC, CLONE_NEWUSER and CLONE_NEWNET namespaces. After it completed a number of
3622 * initializations a second child (the "inner" one) is forked off it, and it exits. */
3623
03cfe0d5
LP		 3624 assert(barrier);
3625 assert(directory);
03cfe0d5 3626 assert(pid_socket >= 0);
e01ff70a 3627 assert(uuid_socket >= 0);
9c1e04d0 3628 assert(notify_socket >= 0);
3acc84eb 3629 assert(master_pty_socket >= 0);
03cfe0d5
LP		 3630 assert(kmsg_socket >= 0);
3631
de40a303
LP		 3632 log_debug("Outer child is initializing.");
3633
e1bb4b0d
LB		 3634 r = load_os_release_pairs_with_prefix("/", "container_host_", &os_release_pairs);
3635 if (r < 0)
3636 log_debug_errno(r, "Failed to read os-release from host for container, ignoring: %m");
3637
03cfe0d5
LP		 3638 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
3639 return log_error_errno(errno, "PR_SET_PDEATHSIG failed: %m");
3640
03cfe0d5
LP		 3641 r = reset_audit_loginuid();
3642 if (r < 0)
3643 return r;
3644
2a2e78e9
LP		 3645 /* Mark everything as slave, so that we still receive mounts from the real root, but don't propagate
3646 * mounts to the real root. */
511a8cfe 3647 r = mount_follow_verbose(LOG_ERR, NULL, "/", NULL, MS_SLAVE|MS_REC, NULL);
60e76d48
ZJS		 3648 if (r < 0)
3649 return r;
03cfe0d5 3650
2d845785 3651 if (dissected_image) {
2d3a5a73
LP		 3652 /* If we are operating on a disk image, then mount its root directory now, but leave out the rest. We
3653 * can read the UID shift from it if we need to. Further down we'll mount the rest, but then with the
3654 * uid shift known. That way we can mount VFAT file systems shifted to the right place right away. This
3655 * makes sure ESP partitions and userns are compatible. */
3656
af187ab2 3657 r = dissected_image_mount_and_warn(
d04faa4e
LP		 3658 dissected_image,
3659 directory,
3660 arg_uid_shift,
21b61b1d 3661 arg_uid_range,
d04faa4e
LP		 3662 DISSECT_IMAGE_MOUNT_ROOT_ONLY|
3663 DISSECT_IMAGE_DISCARD_ON_LOOP|
3664 DISSECT_IMAGE_USR_NO_ROOT|
c65f854a 3665 (arg_read_only ? DISSECT_IMAGE_READ_ONLY : DISSECT_IMAGE_FSCK|DISSECT_IMAGE_GROWFS)|
af187ab2 3666 (arg_start_mode == START_BOOT ? DISSECT_IMAGE_VALIDATE_OS : 0));
2d845785 3667 if (r < 0)
af187ab2 3668 return r;
2d845785 3669 }
03cfe0d5 3670
391567f4
LP		 3671 r = determine_uid_shift(directory);
3672 if (r < 0)
3673 return r;
3674
0de7acce 3675 if (arg_userns_mode != USER_NAMESPACE_NO) {
0e7ac751 3676 /* Let the parent know which UID shift we read from the image */
825d5287
RM		 3677 l = send(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), MSG_NOSIGNAL);
3678 if (l < 0)
3679 return log_error_errno(errno, "Failed to send UID shift: %m");
baaa35ad
ZJS		 3680 if (l != sizeof(arg_uid_shift))
3681 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3682 "Short write while sending UID shift.");
0e7ac751 3683
0de7acce 3684 if (arg_userns_mode == USER_NAMESPACE_PICK) {
0e7ac751
LP		 3685 /* When we are supposed to pick the UID shift, the parent will check now whether the UID shift
3686 * we just read from the image is available. If yes, it will send the UID shift back to us, if
3687 * not it will pick a different one, and send it back to us. */
3688
3689 l = recv(uid_shift_socket, &arg_uid_shift, sizeof(arg_uid_shift), 0);
3690 if (l < 0)
3691 return log_error_errno(errno, "Failed to recv UID shift: %m");
baaa35ad
ZJS		 3692 if (l != sizeof(arg_uid_shift))
3693 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3694 "Short read while receiving UID shift.");
0e7ac751
LP		 3695 }
3696
ff6c6cc1
LP		 3697 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
3698 "Selected user namespace base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range);
825d5287
RM		 3699 }
3700
6f83d3d1
LP		 3701 if (path_equal(directory, "/")) {
3702 /* If the directory we shall boot is the host, let's operate on a bind mount at a different
3703 * place, so that we can make changes to its mount structure (for example, to implement
3704 * --volatile=) without this interfering with our ability to access files such as
3705 * /etc/localtime to copy into the container. Note that we use a fixed place for this
3706 * (instead of a temporary directory, since we are living in our own mount namspace here
7802194a 3707 * already, and thus don't need to be afraid of colliding with anyone else's mounts). */
6f83d3d1
LP		 3708 (void) mkdir_p("/run/systemd/nspawn-root", 0755);
3709
511a8cfe 3710 r = mount_nofollow_verbose(LOG_ERR, "/", "/run/systemd/nspawn-root", NULL, MS_BIND|MS_REC, NULL);
6f83d3d1
LP		 3711 if (r < 0)
3712 return r;
3713
3714 directory = "/run/systemd/nspawn-root";
e50cd82f 3715 }
7d0ecdd6
LP		 3716
3717 r = setup_pivot_root(
3718 directory,
3719 arg_pivot_root_new,
3720 arg_pivot_root_old);
3721 if (r < 0)
3722 return r;
3723
3724 r = setup_volatile_mode(
3725 directory,
3726 arg_volatile_mode,
7d0ecdd6 3727 arg_uid_shift,
8f1ed04a 3728 arg_selinux_apifs_context);
7d0ecdd6
LP		 3729 if (r < 0)
3730 return r;
3731
2f893044
LP		 3732 r = bind_user_prepare(
3733 directory,
3734 arg_bind_user,
3735 arg_uid_shift,
3736 arg_uid_range,
3737 &arg_custom_mounts, &arg_n_custom_mounts,
3738 &bind_user_context);
3739 if (r < 0)
3740 return r;
3741
3742 if (arg_userns_mode != USER_NAMESPACE_NO && bind_user_context) {
3743 /* Send the user maps we determined to the parent, so that it installs it in our user namespace UID map table */
3744
3745 for (size_t i = 0; i < bind_user_context->n_data; i++) {
3746 uid_t map[] = {
3747 bind_user_context->data[i].payload_user->uid,
3748 bind_user_context->data[i].host_user->uid,
3749 (uid_t) bind_user_context->data[i].payload_group->gid,
3750 (uid_t) bind_user_context->data[i].host_group->gid,
3751 };
3752
3753 l = send(uid_shift_socket, map, sizeof(map), MSG_NOSIGNAL);
3754 if (l < 0)
3755 return log_error_errno(errno, "Failed to send user UID map: %m");
3756 if (l != sizeof(map))
3757 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3758 "Short write while sending user UID map.");
3759 }
3760 }
3761
5f0a6347
DDM		 3762 r = mount_custom(
3763 directory,
3764 arg_custom_mounts,
3765 arg_n_custom_mounts,
5f0a6347 3766 arg_uid_shift,
c0c8f718 3767 arg_uid_range,
5f0a6347
DDM		 3768 arg_selinux_apifs_context,
3769 MOUNT_ROOT_ONLY);
3770 if (r < 0)
3771 return r;
3772
5530dc87 3773 /* Make sure we always have a mount that we can move to root later on. */
14a25e1f
LP		 3774 r = make_mount_point(directory);
3775 if (r < 0)
3776 return r;
5530dc87 3777
c0c8f718
AV		 3778 if (arg_userns_mode != USER_NAMESPACE_NO &&
3779 IN_SET(arg_userns_ownership, USER_NAMESPACE_OWNERSHIP_MAP, USER_NAMESPACE_OWNERSHIP_AUTO) &&
3780 arg_uid_shift != 0) {
3781
3782 r = remount_idmap(directory, arg_uid_shift, arg_uid_range);
3783 if (r == -EINVAL || ERRNO_IS_NOT_SUPPORTED(r)) {
3784 /* This might fail because the kernel or file system doesn't support idmapping. We
3785 * can't really distinguish this nicely, nor do we have any guarantees about the
3786 * error codes we see, could be EOPNOTSUPP or EINVAL. */
3787 if (arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_AUTO)
3788 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
3789 "ID mapped mounts are apparently not available, sorry.");
3790
3791 log_debug("ID mapped mounts are apparently not available on this kernel or for the selected file system, reverting to recursive chown()ing.");
3792 arg_userns_ownership = USER_NAMESPACE_OWNERSHIP_CHOWN;
3793 } else if (r < 0)
3794 return log_error_errno(r, "Failed to set up ID mapped mounts: %m");
3795 else {
3796 log_debug("ID mapped mounts available, making use of them.");
3797 idmap = true;
3798 }
3799 }
3800
2d3a5a73
LP		 3801 if (dissected_image) {
3802 /* Now we know the uid shift, let's now mount everything else that might be in the image. */
d04faa4e
LP		 3803 r = dissected_image_mount(
3804 dissected_image,
3805 directory,
3806 arg_uid_shift,
21b61b1d 3807 arg_uid_range,
d04faa4e
LP		 3808 DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY|
3809 DISSECT_IMAGE_DISCARD_ON_LOOP|
3810 DISSECT_IMAGE_USR_NO_ROOT|
f61c7f88
LP		 3811 (arg_read_only ? DISSECT_IMAGE_READ_ONLY : DISSECT_IMAGE_FSCK|DISSECT_IMAGE_GROWFS)|
3812 (idmap ? DISSECT_IMAGE_MOUNT_IDMAPPED : 0));
4fcb96ce
LP		 3813 if (r == -EUCLEAN)
3814 return log_error_errno(r, "File system check for image failed: %m");
2d3a5a73 3815 if (r < 0)
4fcb96ce 3816 return log_error_errno(r, "Failed to mount image file system: %m");
2d3a5a73
LP		 3817 }
3818
8199d554
LP		 3819 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
3820 /* OK, we don't know yet which cgroup mode to use yet. Let's figure it out, and tell the parent. */
3821
3822 r = detect_unified_cgroup_hierarchy_from_image(directory);
3823 if (r < 0)
3824 return r;
3825
3826 l = send(unified_cgroup_hierarchy_socket, &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), MSG_NOSIGNAL);
3827 if (l < 0)
3828 return log_error_errno(errno, "Failed to send cgroup mode: %m");
baaa35ad
ZJS		 3829 if (l != sizeof(arg_unified_cgroup_hierarchy))
3830 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3831 "Short write while sending cgroup mode.");
8199d554
LP		 3832
3833 unified_cgroup_hierarchy_socket = safe_close(unified_cgroup_hierarchy_socket);
3834 }
3835
4ad14eff
LP		 3836 /* Mark everything as shared so our mounts get propagated down. This is
3837 * required to make new bind mounts available in systemd services
5238e957 3838 * inside the container that create a new mount namespace.
4ad14eff
LP		 3839 * See https://github.com/systemd/systemd/issues/3860
3840 * Further submounts (such as /dev) done after this will inherit the
5f0a6347
DDM		 3841 * shared propagation mode.
3842 *
3843 * IMPORTANT: Do not overmount the root directory anymore from now on to
3844 * enable moving the root directory mount to root later on.
3845 * https://github.com/systemd/systemd/issues/3847#issuecomment-562735251
3846 */
511a8cfe 3847 r = mount_nofollow_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL);
4ad14eff
LP		 3848 if (r < 0)
3849 return r;
3850
3851 r = recursive_chown(directory, arg_uid_shift, arg_uid_range);
3852 if (r < 0)
3853 return r;
3854
03cfe0d5
LP		 3855 r = base_filesystem_create(directory, arg_uid_shift, (gid_t) arg_uid_shift);
3856 if (r < 0)
3857 return r;
3858
bbd407ea
DDM		 3859 if (arg_read_only && arg_volatile_mode == VOLATILE_NO &&
3860 !has_custom_root_mount(arg_custom_mounts, arg_n_custom_mounts)) {
64e82c19 3861 r = bind_remount_recursive(directory, MS_RDONLY, MS_RDONLY, NULL);
03cfe0d5
LP		 3862 if (r < 0)
3863 return log_error_errno(r, "Failed to make tree read-only: %m");
3864 }
3865
0de7acce 3866 r = mount_all(directory,
4f086aab 3867 arg_mount_settings,
0de7acce 3868 arg_uid_shift,
0de7acce 3869 arg_selinux_apifs_context);
03cfe0d5
LP		 3870 if (r < 0)
3871 return r;
3872
07fa00f9
LP		 3873 r = copy_devnodes(directory);
3874 if (r < 0)
03cfe0d5
LP		 3875 return r;
3876
de40a303
LP		 3877 r = make_extra_nodes(directory);
3878 if (r < 0)
3879 return r;
3880
3881 (void) dev_setup(directory, arg_uid_shift, arg_uid_shift);
e5f10caf 3882
9fac5029 3883 p = prefix_roota(directory, "/run/host");
e5f10caf 3884 (void) make_inaccessible_nodes(p, arg_uid_shift, arg_uid_shift);
03cfe0d5 3885
07fa00f9
LP		 3886 r = setup_pts(directory);
3887 if (r < 0)
03cfe0d5
LP		 3888 return r;
3889
3890 r = setup_propagate(directory);
3891 if (r < 0)
3892 return r;
3893
8e5430c4
LP		 3894 r = setup_keyring();
3895 if (r < 0)
3896 return r;
3897
3652872a
LP		 3898 r = setup_credentials(directory);
3899 if (r < 0)
3900 return r;
3901
2f893044
LP		 3902 r = bind_user_setup(bind_user_context, directory);
3903 if (r < 0)
3904 return r;
3905
5c4deb9a
MJ		 3906 r = mount_custom(
3907 directory,
3908 arg_custom_mounts,
3909 arg_n_custom_mounts,
3910 arg_uid_shift,
c0c8f718 3911 arg_uid_range,
5c4deb9a
MJ		 3912 arg_selinux_apifs_context,
3913 MOUNT_NON_ROOT_ONLY);
3914 if (r < 0)
3915 return r;
3916
03cfe0d5
LP		 3917 r = setup_timezone(directory);
3918 if (r < 0)
3919 return r;
3920
3921 r = setup_resolv_conf(directory);
3922 if (r < 0)
3923 return r;
3924
e01ff70a
MS		 3925 r = setup_machine_id(directory);
3926 if (r < 0)
3927 return r;
3928
03cfe0d5
LP		 3929 r = setup_journal(directory);
3930 if (r < 0)
3931 return r;
3932
0f48ba7b
LP		 3933 /* The same stuff as the $container env var, but nicely readable for the entire payload */
3934 p = prefix_roota(directory, "/run/host/container-manager");
3935 (void) write_string_file(p, arg_container_service_name, WRITE_STRING_FILE_CREATE);
3936
3937 /* The same stuff as the $container_uuid env var */
3938 p = prefix_roota(directory, "/run/host/container-uuid");
3939 (void) write_string_filef(p, WRITE_STRING_FILE_CREATE, SD_ID128_UUID_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid));
3940
489fae52 3941 if (!arg_use_cgns) {
0996ef00
CB		 3942 r = mount_cgroups(
3943 directory,
3944 arg_unified_cgroup_hierarchy,
3945 arg_userns_mode != USER_NAMESPACE_NO,
3946 arg_uid_shift,
3947 arg_uid_range,
5a8ff0e6 3948 arg_selinux_apifs_context,
ada54120 3949 false);
0996ef00
CB		 3950 if (r < 0)
3951 return r;
3952 }
03cfe0d5
LP		 3953
3954 r = mount_move_root(directory);
3955 if (r < 0)
3956 return log_error_errno(r, "Failed to move root directory: %m");
3957
e96ceaba 3958 fd = setup_notify_child();
9c1e04d0
AP		 3959 if (fd < 0)
3960 return fd;
3961
03cfe0d5 3962 pid = raw_clone(SIGCHLD|CLONE_NEWNS|
0c582db0 3963 arg_clone_ns_flags |
8869a0b4 3964 (arg_userns_mode != USER_NAMESPACE_NO ? CLONE_NEWUSER : 0));
03cfe0d5
LP		 3965 if (pid < 0)
3966 return log_error_errno(errno, "Failed to fork inner child: %m");
03cfe0d5
LP		 3967 if (pid == 0) {
3968 pid_socket = safe_close(pid_socket);
e01ff70a 3969 uuid_socket = safe_close(uuid_socket);
9c1e04d0 3970 notify_socket = safe_close(notify_socket);
825d5287 3971 uid_shift_socket = safe_close(uid_shift_socket);
03cfe0d5 3972
2a2e78e9
LP		 3973 /* The inner child has all namespaces that are requested, so that we all are owned by the
3974 * user if user namespaces are turned on. */
03cfe0d5 3975
d7bea6b6
DP		 3976 if (arg_network_namespace_path) {
3977 r = namespace_enter(-1, -1, netns_fd, -1, -1);
3978 if (r < 0)
e2d39e54 3979 return log_error_errno(r, "Failed to join network namespace: %m");
d7bea6b6
DP		 3980 }
3981
e1bb4b0d 3982 r = inner_child(barrier, directory, secondary, kmsg_socket, rtnl_socket, master_pty_socket, fds, os_release_pairs);
03cfe0d5
LP		 3983 if (r < 0)
3984 _exit(EXIT_FAILURE);
3985
3986 _exit(EXIT_SUCCESS);
3987 }
3988
3989 l = send(pid_socket, &pid, sizeof(pid), MSG_NOSIGNAL);
3990 if (l < 0)
3991 return log_error_errno(errno, "Failed to send PID: %m");
baaa35ad
ZJS		 3992 if (l != sizeof(pid))
3993 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3994 "Short write while sending PID.");
03cfe0d5 3995
e01ff70a
MS		 3996 l = send(uuid_socket, &arg_uuid, sizeof(arg_uuid), MSG_NOSIGNAL);
3997 if (l < 0)
3998 return log_error_errno(errno, "Failed to send machine ID: %m");
baaa35ad
ZJS		 3999 if (l != sizeof(arg_uuid))
4000 return log_error_errno(SYNTHETIC_ERRNO(EIO),
4001 "Short write while sending machine ID.");
e01ff70a 4002
9c1e04d0
AP		 4003 l = send_one_fd(notify_socket, fd, 0);
4004 if (l < 0)
ba72801d 4005 return log_error_errno(l, "Failed to send notify fd: %m");
9c1e04d0 4006
03cfe0d5 4007 pid_socket = safe_close(pid_socket);
e01ff70a 4008 uuid_socket = safe_close(uuid_socket);
9c1e04d0 4009 notify_socket = safe_close(notify_socket);
3acc84eb 4010 master_pty_socket = safe_close(master_pty_socket);
327e26d6
KN		 4011 kmsg_socket = safe_close(kmsg_socket);
4012 rtnl_socket = safe_close(rtnl_socket);
d7bea6b6 4013 netns_fd = safe_close(netns_fd);
03cfe0d5
LP		 4014
4015 return 0;
4016}
4017
0e7ac751 4018static int uid_shift_pick(uid_t *shift, LockFile *ret_lock_file) {
d381c8a6 4019 bool tried_hashed = false;
0e7ac751
LP		 4020 unsigned n_tries = 100;
4021 uid_t candidate;
4022 int r;
4023
4024 assert(shift);
4025 assert(ret_lock_file);
0de7acce 4026 assert(arg_userns_mode == USER_NAMESPACE_PICK);
0e7ac751
LP		 4027 assert(arg_uid_range == 0x10000U);
4028
4029 candidate = *shift;
4030
4031 (void) mkdir("/run/systemd/nspawn-uid", 0755);
4032
4033 for (;;) {
fbd0b64f 4034 char lock_path[STRLEN("/run/systemd/nspawn-uid/") + DECIMAL_STR_MAX(uid_t) + 1];
8e766630 4035 _cleanup_(release_lock_file) LockFile lf = LOCK_FILE_INIT;
0e7ac751
LP		 4036
4037 if (--n_tries <= 0)
4038 return -EBUSY;
4039
87d5e4f2 4040 if (candidate < CONTAINER_UID_BASE_MIN || candidate > CONTAINER_UID_BASE_MAX)
0e7ac751
LP		 4041 goto next;
4042 if ((candidate & UINT32_C(0xFFFF)) != 0)
4043 goto next;
4044
4045 xsprintf(lock_path, "/run/systemd/nspawn-uid/" UID_FMT, candidate);
4046 r = make_lock_file(lock_path, LOCK_EX|LOCK_NB, &lf);
4047 if (r == -EBUSY) /* Range already taken by another nspawn instance */
4048 goto next;
4049 if (r < 0)
4050 return r;
4051
4052 /* Make some superficial checks whether the range is currently known in the user database */
4053 if (getpwuid(candidate))
4054 goto next;
4055 if (getpwuid(candidate + UINT32_C(0xFFFE)))
4056 goto next;
4057 if (getgrgid(candidate))
4058 goto next;
4059 if (getgrgid(candidate + UINT32_C(0xFFFE)))
4060 goto next;
4061
4062 *ret_lock_file = lf;
4063 lf = (struct LockFile) LOCK_FILE_INIT;
4064 *shift = candidate;
4065 return 0;
4066
4067 next:
d381c8a6
LP		 4068 if (arg_machine && !tried_hashed) {
4069 /* Try to hash the base from the container name */
4070
4071 static const uint8_t hash_key[] = {
4072 0xe1, 0x56, 0xe0, 0xf0, 0x4a, 0xf0, 0x41, 0xaf,
4073 0x96, 0x41, 0xcf, 0x41, 0x33, 0x94, 0xff, 0x72
4074 };
4075
4076 candidate = (uid_t) siphash24(arg_machine, strlen(arg_machine), hash_key);
4077
4078 tried_hashed = true;
4079 } else
4080 random_bytes(&candidate, sizeof(candidate));
4081
87d5e4f2 4082 candidate = (candidate % (CONTAINER_UID_BASE_MAX - CONTAINER_UID_BASE_MIN)) + CONTAINER_UID_BASE_MIN;
0e7ac751
LP		 4083 candidate &= (uid_t) UINT32_C(0xFFFF0000);
4084 }
4085}
4086
2f893044
LP		 4087static int add_one_uid_map(
4088 char **p,
4089 uid_t container_uid,
4090 uid_t host_uid,
4091 uid_t range) {
4092
4093 return strextendf(p,
4094 UID_FMT " " UID_FMT " " UID_FMT "\n",
4095 container_uid, host_uid, range);
4096}
4097
4098static int make_uid_map_string(
4099 const uid_t bind_user_uid[],
4100 size_t n_bind_user_uid,
4101 size_t offset,
4102 char **ret) {
4103
4104 _cleanup_free_ char *s = NULL;
4105 uid_t previous_uid = 0;
4106 int r;
4107
4108 assert(n_bind_user_uid == 0 || bind_user_uid);
2f092762 4109 assert(IN_SET(offset, 0, 2)); /* used to switch between UID and GID map */
2f893044
LP		 4110 assert(ret);
4111
4112 /* The bind_user_uid[] array is a series of 4 uid_t values, for each --bind-user= entry one
4113 * quadruplet, consisting of host and container UID + GID. */
4114
4115 for (size_t i = 0; i < n_bind_user_uid; i++) {
4116 uid_t payload_uid = bind_user_uid[i*2+offset],
4117 host_uid = bind_user_uid[i*2+offset+1];
4118
4119 assert(previous_uid <= payload_uid);
4120 assert(payload_uid < arg_uid_range);
4121
4122 /* Add a range to close the gap to previous entry */
4123 if (payload_uid > previous_uid) {
4124 r = add_one_uid_map(&s, previous_uid, arg_uid_shift + previous_uid, payload_uid - previous_uid);
4125 if (r < 0)
4126 return r;
4127 }
4128
4129 /* Map this specific user */
4130 r = add_one_uid_map(&s, payload_uid, host_uid, 1);
4131 if (r < 0)
4132 return r;
4133
4134 previous_uid = payload_uid + 1;
4135 }
4136
4137 /* And add a range to close the gap to finish the range */
4138 if (arg_uid_range > previous_uid) {
4139 r = add_one_uid_map(&s, previous_uid, arg_uid_shift + previous_uid, arg_uid_range - previous_uid);
4140 if (r < 0)
4141 return r;
4142 }
4143
4144 assert(s);
4145
4146 *ret = TAKE_PTR(s);
4147 return 0;
4148}
4149
4150static int setup_uid_map(
4151 pid_t pid,
4152 const uid_t bind_user_uid[],
4153 size_t n_bind_user_uid) {
4154
4155 char uid_map[STRLEN("/proc//uid_map") + DECIMAL_STR_MAX(uid_t) + 1];
4156 _cleanup_free_ char *s = NULL;
03cfe0d5
LP		 4157 int r;
4158
4159 assert(pid > 1);
4160
2f893044
LP		 4161 /* Build the UID map string */
4162 if (make_uid_map_string(bind_user_uid, n_bind_user_uid, 0, &s) < 0) /* offset=0 contains the UID pair */
4163 return log_oom();
4164
03cfe0d5 4165 xsprintf(uid_map, "/proc/" PID_FMT "/uid_map", pid);
2f893044 4166 r = write_string_file(uid_map, s, WRITE_STRING_FILE_DISABLE_BUFFER);
03cfe0d5
LP		 4167 if (r < 0)
4168 return log_error_errno(r, "Failed to write UID map: %m");
4169
2f893044
LP		 4170 /* And now build the GID map string */
4171 s = mfree(s);
4172 if (make_uid_map_string(bind_user_uid, n_bind_user_uid, 2, &s) < 0) /* offset=2 contains the GID pair */
4173 return log_oom();
4174
03cfe0d5 4175 xsprintf(uid_map, "/proc/" PID_FMT "/gid_map", pid);
2f893044 4176 r = write_string_file(uid_map, s, WRITE_STRING_FILE_DISABLE_BUFFER);
03cfe0d5
LP		 4177 if (r < 0)
4178 return log_error_errno(r, "Failed to write GID map: %m");
4179
4180 return 0;
4181}
4182
9c1e04d0 4183static int nspawn_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) {
9c1e04d0
AP		 4184 char buf[NOTIFY_BUFFER_MAX+1];
4185 char *p = NULL;
4186 struct iovec iovec = {
4187 .iov_base = buf,
4188 .iov_len = sizeof(buf)-1,
4189 };
fb29cdbe
LP		 4190 CMSG_BUFFER_TYPE(CMSG_SPACE(sizeof(struct ucred)) +
4191 CMSG_SPACE(sizeof(int) * NOTIFY_FD_MAX)) control;
9c1e04d0
AP		 4192 struct msghdr msghdr = {
4193 .msg_iov = &iovec,
4194 .msg_iovlen = 1,
4195 .msg_control = &control,
4196 .msg_controllen = sizeof(control),
4197 };
371d72e0 4198 struct ucred *ucred;
9c1e04d0
AP		 4199 ssize_t n;
4200 pid_t inner_child_pid;
4201 _cleanup_strv_free_ char **tags = NULL;
4bf4f50f 4202 int r;
9c1e04d0
AP		 4203
4204 assert(userdata);
4205
4206 inner_child_pid = PTR_TO_PID(userdata);
4207
4208 if (revents != EPOLLIN) {
4209 log_warning("Got unexpected poll event for notify fd.");
4210 return 0;
4211 }
4212
3691bcf3 4213 n = recvmsg_safe(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
8add30a0
YW		 4214 if (n < 0) {
4215 if (ERRNO_IS_TRANSIENT(n))
4216 return 0;
4217 if (n == -EXFULL) {
4218 log_warning("Got message with truncated control data (too many fds sent?), ignoring.");
4219 return 0;
4220 }
3691bcf3 4221 return log_warning_errno(n, "Couldn't read notification socket: %m");
8add30a0 4222 }
9c1e04d0 4223
9c1e04d0
AP		 4224 cmsg_close_all(&msghdr);
4225
371d72e0 4226 ucred = CMSG_FIND_DATA(&msghdr, SOL_SOCKET, SCM_CREDENTIALS, struct ucred);
9c1e04d0 4227 if (!ucred || ucred->pid != inner_child_pid) {
8cb57430 4228 log_debug("Received notify message without valid credentials. Ignoring.");
9c1e04d0
AP		 4229 return 0;
4230 }
4231
4232 if ((size_t) n >= sizeof(buf)) {
4233 log_warning("Received notify message exceeded maximum size. Ignoring.");
4234 return 0;
4235 }
4236
4237 buf[n] = 0;
4238 tags = strv_split(buf, "\n\r");
4239 if (!tags)
4240 return log_oom();
4241
4bf4f50f 4242 if (strv_find(tags, "READY=1")) {
d4341b76 4243 r = sd_notify(false, "READY=1\n");
4bf4f50f
ZJS		 4244 if (r < 0)
4245 log_warning_errno(r, "Failed to send readiness notification, ignoring: %m");
4246 }
9c1e04d0
AP		 4247
4248 p = strv_find_startswith(tags, "STATUS=");
4249 if (p)
04f590a4 4250 (void) sd_notifyf(false, "STATUS=Container running: %s", p);
9c1e04d0
AP		 4251
4252 return 0;
4253}
4254
e96ceaba 4255static int setup_notify_parent(sd_event *event, int fd, pid_t *inner_child_pid, sd_event_source **notify_event_source) {
9c1e04d0 4256 int r;
9c1e04d0 4257
5773024d 4258 r = sd_event_add_io(event, notify_event_source, fd, EPOLLIN, nspawn_dispatch_notify_fd, inner_child_pid);
9c1e04d0
AP		 4259 if (r < 0)
4260 return log_error_errno(r, "Failed to allocate notify event source: %m");
4261
5773024d 4262 (void) sd_event_source_set_description(*notify_event_source, "nspawn-notify");
9c1e04d0
AP		 4263
4264 return 0;
4265}
4266
5d961407
LP		 4267static int merge_settings(Settings *settings, const char *path) {
4268 int rl;
f757855e 4269
5d961407
LP		 4270 assert(settings);
4271 assert(path);
f757855e 4272
5d961407
LP		 4273 /* Copy over bits from the settings, unless they have been explicitly masked by command line switches. Note
4274 * that this steals the fields of the Settings* structure, and hence modifies it. */
f757855e 4275
7732f92b
LP		 4276 if ((arg_settings_mask & SETTING_START_MODE) == 0 &&
4277 settings->start_mode >= 0) {
4278 arg_start_mode = settings->start_mode;
130d3d22 4279 strv_free_and_replace(arg_parameters, settings->parameters);
f757855e
LP		 4280 }
4281
d3689b94
LP		 4282 if ((arg_settings_mask & SETTING_EPHEMERAL) == 0 &&
4283 settings->ephemeral >= 0)
a2f577fc
JL		 4284 arg_ephemeral = settings->ephemeral;
4285
de40a303
LP		 4286 if ((arg_settings_mask & SETTING_DIRECTORY) == 0 &&
4287 settings->root) {
4288
4289 if (!arg_settings_trusted)
4290 log_warning("Ignoring root directory setting, file %s is not trusted.", path);
4291 else
4292 free_and_replace(arg_directory, settings->root);
4293 }
4294
b53ede69
PW		 4295 if ((arg_settings_mask & SETTING_PIVOT_ROOT) == 0 &&
4296 settings->pivot_root_new) {
4297 free_and_replace(arg_pivot_root_new, settings->pivot_root_new);
4298 free_and_replace(arg_pivot_root_old, settings->pivot_root_old);
4299 }
4300
5f932eb9 4301 if ((arg_settings_mask & SETTING_WORKING_DIRECTORY) == 0 &&
1cc6c93a
YW		 4302 settings->working_directory)
4303 free_and_replace(arg_chdir, settings->working_directory);
5f932eb9 4304
f757855e 4305 if ((arg_settings_mask & SETTING_ENVIRONMENT) == 0 &&
130d3d22
YW		 4306 settings->environment)
4307 strv_free_and_replace(arg_setenv, settings->environment);
f757855e 4308
de40a303
LP		 4309 if ((arg_settings_mask & SETTING_USER) == 0) {
4310
4311 if (settings->user)
4312 free_and_replace(arg_user, settings->user);
4313
4314 if (uid_is_valid(settings->uid))
4315 arg_uid = settings->uid;
4316 if (gid_is_valid(settings->gid))
4317 arg_gid = settings->gid;
4318 if (settings->n_supplementary_gids > 0) {
4319 free_and_replace(arg_supplementary_gids, settings->supplementary_gids);
4320 arg_n_supplementary_gids = settings->n_supplementary_gids;
4321 }
4322 }
f757855e
LP		 4323
4324 if ((arg_settings_mask & SETTING_CAPABILITY) == 0) {
a3fc6b55 4325 uint64_t plus, minus;
7be830c6 4326 uint64_t network_minus = 0;
88fc9c9b 4327 uint64_t ambient;
f757855e 4328
de40a303
LP		 4329 /* Note that we copy both the simple plus/minus caps here, and the full quintet from the
4330 * Settings structure */
4331
0e265674 4332 plus = settings->capability;
a3fc6b55
LP		 4333 minus = settings->drop_capability;
4334
9baa294c
LP		 4335 if ((arg_settings_mask & SETTING_NETWORK) == 0 &&
4336 settings_network_configured(settings)) {
a3fc6b55
LP		 4337 if (settings_private_network(settings))
4338 plus |= UINT64_C(1) << CAP_NET_ADMIN;
4339 else
7be830c6 4340 network_minus |= UINT64_C(1) << CAP_NET_ADMIN;
a3fc6b55 4341 }
0e265674
LP		 4342
4343 if (!arg_settings_trusted && plus != 0) {
4344 if (settings->capability != 0)
5d961407 4345 log_warning("Ignoring Capability= setting, file %s is not trusted.", path);
7be830c6
TH		 4346 } else {
4347 arg_caps_retain &= ~network_minus;
520e0d54 4348 arg_caps_retain |= plus;
7be830c6 4349 }
f757855e 4350
a3fc6b55 4351 arg_caps_retain &= ~minus;
de40a303
LP		 4352
4353 /* Copy the full capabilities over too */
4354 if (capability_quintet_is_set(&settings->full_capabilities)) {
4355 if (!arg_settings_trusted)
5238e957 4356 log_warning("Ignoring capability settings, file %s is not trusted.", path);
de40a303
LP		 4357 else
4358 arg_full_capabilities = settings->full_capabilities;
4359 }
88fc9c9b
TH		 4360
4361 ambient = settings->ambient_capability;
4362 if (!arg_settings_trusted && ambient != 0)
4363 log_warning("Ignoring AmbientCapability= setting, file %s is not trusted.", path);
4364 else
4365 arg_caps_ambient |= ambient;
f757855e
LP		 4366 }
4367
4368 if ((arg_settings_mask & SETTING_KILL_SIGNAL) == 0 &&
4369 settings->kill_signal > 0)
4370 arg_kill_signal = settings->kill_signal;
4371
4372 if ((arg_settings_mask & SETTING_PERSONALITY) == 0 &&
4373 settings->personality != PERSONALITY_INVALID)
4374 arg_personality = settings->personality;
4375
4376 if ((arg_settings_mask & SETTING_MACHINE_ID) == 0 &&
4377 !sd_id128_is_null(settings->machine_id)) {
4378
4379 if (!arg_settings_trusted)
5d961407 4380 log_warning("Ignoring MachineID= setting, file %s is not trusted.", path);
f757855e
LP		 4381 else
4382 arg_uuid = settings->machine_id;
4383 }
4384
4385 if ((arg_settings_mask & SETTING_READ_ONLY) == 0 &&
4386 settings->read_only >= 0)
4387 arg_read_only = settings->read_only;
4388
4389 if ((arg_settings_mask & SETTING_VOLATILE_MODE) == 0 &&
4390 settings->volatile_mode != _VOLATILE_MODE_INVALID)
4391 arg_volatile_mode = settings->volatile_mode;
4392
4393 if ((arg_settings_mask & SETTING_CUSTOM_MOUNTS) == 0 &&
4394 settings->n_custom_mounts > 0) {
4395
4396 if (!arg_settings_trusted)
5d961407 4397 log_warning("Ignoring TemporaryFileSystem=, Bind= and BindReadOnly= settings, file %s is not trusted.", path);
f757855e
LP		 4398 else {
4399 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
1cc6c93a 4400 arg_custom_mounts = TAKE_PTR(settings->custom_mounts);
f757855e 4401 arg_n_custom_mounts = settings->n_custom_mounts;
f757855e
LP		 4402 settings->n_custom_mounts = 0;
4403 }
4404 }
4405
4406 if ((arg_settings_mask & SETTING_NETWORK) == 0 &&
a1dfd585 4407 settings_network_configured(settings)) {
f757855e
LP		 4408
4409 if (!arg_settings_trusted)
5d961407 4410 log_warning("Ignoring network settings, file %s is not trusted.", path);
f757855e 4411 else {
f6d6bad1 4412 arg_network_veth = settings_network_veth(settings);
0e265674
LP		 4413 arg_private_network = settings_private_network(settings);
4414
130d3d22
YW		 4415 strv_free_and_replace(arg_network_interfaces, settings->network_interfaces);
4416 strv_free_and_replace(arg_network_macvlan, settings->network_macvlan);
4417 strv_free_and_replace(arg_network_ipvlan, settings->network_ipvlan);
4418 strv_free_and_replace(arg_network_veth_extra, settings->network_veth_extra);
f6d6bad1 4419
1cc6c93a
YW		 4420 free_and_replace(arg_network_bridge, settings->network_bridge);
4421 free_and_replace(arg_network_zone, settings->network_zone);
de40a303
LP		 4422
4423 free_and_replace(arg_network_namespace_path, settings->network_namespace_path);
f757855e
LP		 4424 }
4425 }
4426
4427 if ((arg_settings_mask & SETTING_EXPOSE_PORTS) == 0 &&
4428 settings->expose_ports) {
4429
4430 if (!arg_settings_trusted)
5d961407 4431 log_warning("Ignoring Port= setting, file %s is not trusted.", path);
f757855e
LP		 4432 else {
4433 expose_port_free_all(arg_expose_ports);
1cc6c93a 4434 arg_expose_ports = TAKE_PTR(settings->expose_ports);
f757855e
LP		 4435 }
4436 }
4437
0de7acce
LP		 4438 if ((arg_settings_mask & SETTING_USERNS) == 0 &&
4439 settings->userns_mode != _USER_NAMESPACE_MODE_INVALID) {
4440
4441 if (!arg_settings_trusted)
5d961407 4442 log_warning("Ignoring PrivateUsers= and PrivateUsersChown= settings, file %s is not trusted.", path);
0de7acce
LP		 4443 else {
4444 arg_userns_mode = settings->userns_mode;
4445 arg_uid_shift = settings->uid_shift;
4446 arg_uid_range = settings->uid_range;
6c045a99 4447 arg_userns_ownership = settings->userns_ownership;
0de7acce
LP		 4448 }
4449 }
4450
0cc3c9f9
LP		 4451 if ((arg_settings_mask & SETTING_BIND_USER) == 0 &&
4452 !strv_isempty(settings->bind_user))
2f893044
LP		 4453 strv_free_and_replace(arg_bind_user, settings->bind_user);
4454
d3689b94
LP		 4455 if ((arg_settings_mask & SETTING_NOTIFY_READY) == 0 &&
4456 settings->notify_ready >= 0)
9c1e04d0
AP		 4457 arg_notify_ready = settings->notify_ready;
4458
960e4569
LP		 4459 if ((arg_settings_mask & SETTING_SYSCALL_FILTER) == 0) {
4460
2d09ea44
LP		 4461 if (!strv_isempty(settings->syscall_allow_list) || !strv_isempty(settings->syscall_deny_list)) {
4462 if (!arg_settings_trusted && !strv_isempty(settings->syscall_allow_list))
4463 log_warning("Ignoring SystemCallFilter= settings, file %s is not trusted.", path);
4464 else {
4465 strv_free_and_replace(arg_syscall_allow_list, settings->syscall_allow_list);
4466 strv_free_and_replace(arg_syscall_deny_list, settings->syscall_deny_list);
4467 }
960e4569 4468 }
de40a303
LP		 4469
4470#if HAVE_SECCOMP
2d09ea44
LP		 4471 if (settings->seccomp) {
4472 if (!arg_settings_trusted)
4473 log_warning("Ignoring SECCOMP filter, file %s is not trusted.", path);
4474 else {
4475 seccomp_release(arg_seccomp);
4476 arg_seccomp = TAKE_PTR(settings->seccomp);
4477 }
de40a303
LP		 4478 }
4479#endif
960e4569
LP		 4480 }
4481
bf428efb
LP		 4482 for (rl = 0; rl < _RLIMIT_MAX; rl ++) {
4483 if ((arg_settings_mask & (SETTING_RLIMIT_FIRST << rl)))
4484 continue;
4485
4486 if (!settings->rlimit[rl])
4487 continue;
4488
4489 if (!arg_settings_trusted) {
5d961407 4490 log_warning("Ignoring Limit%s= setting, file '%s' is not trusted.", rlimit_to_string(rl), path);
bf428efb
LP		 4491 continue;
4492 }
4493
4494 free_and_replace(arg_rlimit[rl], settings->rlimit[rl]);
4495 }
4496
3a9530e5
LP		 4497 if ((arg_settings_mask & SETTING_HOSTNAME) == 0 &&
4498 settings->hostname)
4499 free_and_replace(arg_hostname, settings->hostname);
4500
66edd963
LP		 4501 if ((arg_settings_mask & SETTING_NO_NEW_PRIVILEGES) == 0 &&
4502 settings->no_new_privileges >= 0)
4503 arg_no_new_privileges = settings->no_new_privileges;
4504
81f345df
LP		 4505 if ((arg_settings_mask & SETTING_OOM_SCORE_ADJUST) == 0 &&
4506 settings->oom_score_adjust_set) {
4507
4508 if (!arg_settings_trusted)
5d961407 4509 log_warning("Ignoring OOMScoreAdjust= setting, file '%s' is not trusted.", path);
81f345df
LP		 4510 else {
4511 arg_oom_score_adjust = settings->oom_score_adjust;
4512 arg_oom_score_adjust_set = true;
4513 }
4514 }
4515
d107bb7d 4516 if ((arg_settings_mask & SETTING_CPU_AFFINITY) == 0 &&
0985c7c4 4517 settings->cpu_set.set) {
d107bb7d
LP		 4518
4519 if (!arg_settings_trusted)
5d961407 4520 log_warning("Ignoring CPUAffinity= setting, file '%s' is not trusted.", path);
d107bb7d 4521 else {
0985c7c4
ZJS		 4522 cpu_set_reset(&arg_cpu_set);
4523 arg_cpu_set = settings->cpu_set;
4524 settings->cpu_set = (CPUSet) {};
d107bb7d
LP		 4525 }
4526 }
4527
09d423e9
LP		 4528 if ((arg_settings_mask & SETTING_RESOLV_CONF) == 0 &&
4529 settings->resolv_conf != _RESOLV_CONF_MODE_INVALID)
4530 arg_resolv_conf = settings->resolv_conf;
4531
4e1d6aa9
LP		 4532 if ((arg_settings_mask & SETTING_LINK_JOURNAL) == 0 &&
4533 settings->link_journal != _LINK_JOURNAL_INVALID) {
4534
4535 if (!arg_settings_trusted)
4536 log_warning("Ignoring journal link setting, file '%s' is not trusted.", path);
4537 else {
4538 arg_link_journal = settings->link_journal;
4539 arg_link_journal_try = settings->link_journal_try;
4540 }
4541 }
4542
1688841f
LP		 4543 if ((arg_settings_mask & SETTING_TIMEZONE) == 0 &&
4544 settings->timezone != _TIMEZONE_MODE_INVALID)
4545 arg_timezone = settings->timezone;
4546
de40a303
LP		 4547 if ((arg_settings_mask & SETTING_SLICE) == 0 &&
4548 settings->slice) {
4549
4550 if (!arg_settings_trusted)
4551 log_warning("Ignoring slice setting, file '%s' is not trusted.", path);
4552 else
4553 free_and_replace(arg_slice, settings->slice);
4554 }
4555
4556 if ((arg_settings_mask & SETTING_USE_CGNS) == 0 &&
4557 settings->use_cgns >= 0) {
4558
4559 if (!arg_settings_trusted)
4560 log_warning("Ignoring cgroup namespace setting, file '%s' is not trusted.", path);
4561 else
4562 arg_use_cgns = settings->use_cgns;
4563 }
4564
4565 if ((arg_settings_mask & SETTING_CLONE_NS_FLAGS) == 0 &&
f5fbe71d 4566 settings->clone_ns_flags != ULONG_MAX) {
de40a303
LP		 4567
4568 if (!arg_settings_trusted)
4569 log_warning("Ignoring namespace setting, file '%s' is not trusted.", path);
4570 else
4571 arg_clone_ns_flags = settings->clone_ns_flags;
4572 }
4573
4574 if ((arg_settings_mask & SETTING_CONSOLE_MODE) == 0 &&
4575 settings->console_mode >= 0) {
4576
4577 if (!arg_settings_trusted)
4578 log_warning("Ignoring console mode setting, file '%s' is not trusted.", path);
4579 else
4580 arg_console_mode = settings->console_mode;
4581 }
4582
d3689b94
LP		 4583 if ((arg_settings_mask & SETTING_SUPPRESS_SYNC) == 0 &&
4584 settings->suppress_sync >= 0)
4a4654e0
LP		 4585 arg_suppress_sync = settings->suppress_sync;
4586
de40a303
LP		 4587 /* The following properties can only be set through the OCI settings logic, not from the command line, hence we
4588 * don't consult arg_settings_mask for them. */
4589
4590 sd_bus_message_unref(arg_property_message);
4591 arg_property_message = TAKE_PTR(settings->properties);
4592
4593 arg_console_width = settings->console_width;
4594 arg_console_height = settings->console_height;
4595
b2645747 4596 device_node_array_free(arg_extra_nodes, arg_n_extra_nodes);
de40a303
LP		 4597 arg_extra_nodes = TAKE_PTR(settings->extra_nodes);
4598 arg_n_extra_nodes = settings->n_extra_nodes;
4599
f757855e
LP		 4600 return 0;
4601}
4602
5d961407
LP		 4603static int load_settings(void) {
4604 _cleanup_(settings_freep) Settings *settings = NULL;
4605 _cleanup_fclose_ FILE *f = NULL;
4606 _cleanup_free_ char *p = NULL;
4607 const char *fn, *i;
4608 int r;
4609
de40a303
LP		 4610 if (arg_oci_bundle)
4611 return 0;
4612
5d961407
LP		 4613 /* If all settings are masked, there's no point in looking for
4614 * the settings file */
d7a0f1f4 4615 if (FLAGS_SET(arg_settings_mask, _SETTINGS_MASK_ALL))
5d961407
LP		 4616 return 0;
4617
4618 fn = strjoina(arg_machine, ".nspawn");
4619
4620 /* We first look in the admin's directories in /etc and /run */
4621 FOREACH_STRING(i, "/etc/systemd/nspawn", "/run/systemd/nspawn") {
4622 _cleanup_free_ char *j = NULL;
4623
657ee2d8 4624 j = path_join(i, fn);
5d961407
LP		 4625 if (!j)
4626 return log_oom();
4627
4628 f = fopen(j, "re");
4629 if (f) {
4630 p = TAKE_PTR(j);
4631
4632 /* By default, we trust configuration from /etc and /run */
4633 if (arg_settings_trusted < 0)
4634 arg_settings_trusted = true;
4635
4636 break;
4637 }
4638
4639 if (errno != ENOENT)
4640 return log_error_errno(errno, "Failed to open %s: %m", j);
4641 }
4642
4643 if (!f) {
4644 /* After that, let's look for a file next to the
4645 * actual image we shall boot. */
4646
4647 if (arg_image) {
4648 p = file_in_same_dir(arg_image, fn);
4649 if (!p)
4650 return log_oom();
cd6e3914 4651 } else if (arg_directory && !path_equal(arg_directory, "/")) {
5d961407
LP		 4652 p = file_in_same_dir(arg_directory, fn);
4653 if (!p)
4654 return log_oom();
4655 }
4656
4657 if (p) {
4658 f = fopen(p, "re");
4659 if (!f && errno != ENOENT)
4660 return log_error_errno(errno, "Failed to open %s: %m", p);
4661
4662 /* By default, we do not trust configuration from /var/lib/machines */
4663 if (arg_settings_trusted < 0)
4664 arg_settings_trusted = false;
4665 }
4666 }
4667
4668 if (!f)
4669 return 0;
4670
4671 log_debug("Settings are trusted: %s", yes_no(arg_settings_trusted));
4672
4673 r = settings_load(f, p, &settings);
4674 if (r < 0)
4675 return r;
4676
4677 return merge_settings(settings, p);
4678}
4679
de40a303
LP		 4680static int load_oci_bundle(void) {
4681 _cleanup_(settings_freep) Settings *settings = NULL;
4682 int r;
4683
4684 if (!arg_oci_bundle)
4685 return 0;
4686
4687 /* By default let's trust OCI bundles */
4688 if (arg_settings_trusted < 0)
4689 arg_settings_trusted = true;
4690
4691 r = oci_load(NULL, arg_oci_bundle, &settings);
4692 if (r < 0)
4693 return r;
4694
4695 return merge_settings(settings, arg_oci_bundle);
4696}
4697
3acc84eb 4698static int run_container(
2d845785 4699 DissectedImage *dissected_image,
b0067625
ZJS		 4700 bool secondary,
4701 FDSet *fds,
4702 char veth_name[IFNAMSIZ], bool *veth_created,
761cf19d 4703 struct ExposeArgs *expose_args,
3acc84eb 4704 int *master, pid_t *pid, int *ret) {
b0067625
ZJS		 4705
4706 static const struct sigaction sa = {
4707 .sa_handler = nop_signal_handler,
e28c7cd0 4708 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
b0067625
ZJS		 4709 };
4710
8e766630 4711 _cleanup_(release_lock_file) LockFile uid_shift_lock = LOCK_FILE_INIT;
b0067625
ZJS		 4712 _cleanup_close_ int etc_passwd_lock = -1;
4713 _cleanup_close_pair_ int
4714 kmsg_socket_pair[2] = { -1, -1 },
4715 rtnl_socket_pair[2] = { -1, -1 },
4716 pid_socket_pair[2] = { -1, -1 },
4717 uuid_socket_pair[2] = { -1, -1 },
4718 notify_socket_pair[2] = { -1, -1 },
8199d554 4719 uid_shift_socket_pair[2] = { -1, -1 },
3acc84eb 4720 master_pty_socket_pair[2] = { -1, -1 },
8199d554
LP		 4721 unified_cgroup_hierarchy_socket_pair[2] = { -1, -1};
4722
3acc84eb 4723 _cleanup_close_ int notify_socket = -1;
b0067625 4724 _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
5773024d 4725 _cleanup_(sd_event_source_unrefp) sd_event_source *notify_event_source = NULL;
b0067625
ZJS		 4726 _cleanup_(sd_event_unrefp) sd_event *event = NULL;
4727 _cleanup_(pty_forward_freep) PTYForward *forward = NULL;
4728 _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL;
abdb9b08 4729 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
2f893044
LP		 4730 _cleanup_free_ uid_t *bind_user_uid = NULL;
4731 size_t n_bind_user_uid = 0;
b0067625 4732 ContainerStatus container_status = 0;
b0067625
ZJS		 4733 int ifi = 0, r;
4734 ssize_t l;
4735 sigset_t mask_chld;
5b4855ab 4736 _cleanup_close_ int child_netns_fd = -1;
b0067625
ZJS		 4737
4738 assert_se(sigemptyset(&mask_chld) == 0);
4739 assert_se(sigaddset(&mask_chld, SIGCHLD) == 0);
4740
4741 if (arg_userns_mode == USER_NAMESPACE_PICK) {
4742 /* When we shall pick the UID/GID range, let's first lock /etc/passwd, so that we can safely
4743 * check with getpwuid() if the specific user already exists. Note that /etc might be
4744 * read-only, in which case this will fail with EROFS. But that's really OK, as in that case we
4745 * can be reasonably sure that no users are going to be added. Note that getpwuid() checks are
4746 * really just an extra safety net. We kinda assume that the UID range we allocate from is
4747 * really ours. */
4748
4749 etc_passwd_lock = take_etc_passwd_lock(NULL);
4750 if (etc_passwd_lock < 0 && etc_passwd_lock != -EROFS)
4751 return log_error_errno(etc_passwd_lock, "Failed to take /etc/passwd lock: %m");
4752 }
4753
4754 r = barrier_create(&barrier);
4755 if (r < 0)
4756 return log_error_errno(r, "Cannot initialize IPC barrier: %m");
4757
4758 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0)
4759 return log_error_errno(errno, "Failed to create kmsg socket pair: %m");
4760
4761 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0)
4762 return log_error_errno(errno, "Failed to create rtnl socket pair: %m");
4763
4764 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, pid_socket_pair) < 0)
4765 return log_error_errno(errno, "Failed to create pid socket pair: %m");
4766
4767 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uuid_socket_pair) < 0)
4768 return log_error_errno(errno, "Failed to create id socket pair: %m");
4769
4770 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, notify_socket_pair) < 0)
4771 return log_error_errno(errno, "Failed to create notify socket pair: %m");
4772
3acc84eb
FB		 4773 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, master_pty_socket_pair) < 0)
4774 return log_error_errno(errno, "Failed to create console socket pair: %m");
4775
b0067625
ZJS		 4776 if (arg_userns_mode != USER_NAMESPACE_NO)
4777 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0)
4778 return log_error_errno(errno, "Failed to create uid shift socket pair: %m");
4779
8199d554
LP		 4780 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN)
4781 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, unified_cgroup_hierarchy_socket_pair) < 0)
4782 return log_error_errno(errno, "Failed to create unified cgroup socket pair: %m");
4783
b0067625
ZJS		 4784 /* Child can be killed before execv(), so handle SIGCHLD in order to interrupt
4785 * parent's blocking calls and give it a chance to call wait() and terminate. */
4786 r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
4787 if (r < 0)
4788 return log_error_errno(errno, "Failed to change the signal mask: %m");
4789
4790 r = sigaction(SIGCHLD, &sa, NULL);
4791 if (r < 0)
4792 return log_error_errno(errno, "Failed to install SIGCHLD handler: %m");
4793
d7bea6b6 4794 if (arg_network_namespace_path) {
5b4855ab
DDM		 4795 child_netns_fd = open(arg_network_namespace_path, O_RDONLY|O_NOCTTY|O_CLOEXEC);
4796 if (child_netns_fd < 0)
d7bea6b6
DP		 4797 return log_error_errno(errno, "Cannot open file %s: %m", arg_network_namespace_path);
4798
54c2459d 4799 r = fd_is_ns(child_netns_fd, CLONE_NEWNET);
6619ad88
LP		 4800 if (r == -EUCLEAN)
4801 log_debug_errno(r, "Cannot determine if passed network namespace path '%s' really refers to a network namespace, assuming it does.", arg_network_namespace_path);
4802 else if (r < 0)
d7bea6b6 4803 return log_error_errno(r, "Failed to check %s fs type: %m", arg_network_namespace_path);
c6147113
LP		 4804 else if (r == 0)
4805 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4806 "Path %s doesn't refer to a network namespace, refusing.", arg_network_namespace_path);
d7bea6b6
DP		 4807 }
4808
b0067625
ZJS		 4809 *pid = raw_clone(SIGCHLD|CLONE_NEWNS);
4810 if (*pid < 0)
4811 return log_error_errno(errno, "clone() failed%s: %m",
4812 errno == EINVAL ?
4813 ", do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in)" : "");
4814
4815 if (*pid == 0) {
4816 /* The outer child only has a file system namespace. */
4817 barrier_set_role(&barrier, BARRIER_CHILD);
4818
b0067625
ZJS		 4819 kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
4820 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
4821 pid_socket_pair[0] = safe_close(pid_socket_pair[0]);
4822 uuid_socket_pair[0] = safe_close(uuid_socket_pair[0]);
4823 notify_socket_pair[0] = safe_close(notify_socket_pair[0]);
3acc84eb 4824 master_pty_socket_pair[0] = safe_close(master_pty_socket_pair[0]);
b0067625 4825 uid_shift_socket_pair[0] = safe_close(uid_shift_socket_pair[0]);
8199d554 4826 unified_cgroup_hierarchy_socket_pair[0] = safe_close(unified_cgroup_hierarchy_socket_pair[0]);
b0067625
ZJS		 4827
4828 (void) reset_all_signal_handlers();
4829 (void) reset_signal_mask();
4830
4831 r = outer_child(&barrier,
4832 arg_directory,
2d845785 4833 dissected_image,
b0067625
ZJS		 4834 secondary,
4835 pid_socket_pair[1],
4836 uuid_socket_pair[1],
4837 notify_socket_pair[1],
4838 kmsg_socket_pair[1],
4839 rtnl_socket_pair[1],
4840 uid_shift_socket_pair[1],
3acc84eb 4841 master_pty_socket_pair[1],
8199d554 4842 unified_cgroup_hierarchy_socket_pair[1],
d7bea6b6 4843 fds,
5b4855ab 4844 child_netns_fd);
b0067625
ZJS		 4845 if (r < 0)
4846 _exit(EXIT_FAILURE);
4847
4848 _exit(EXIT_SUCCESS);
4849 }
4850
4851 barrier_set_role(&barrier, BARRIER_PARENT);
4852
e4077ff6 4853 fdset_close(fds);
b0067625
ZJS		 4854
4855 kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
4856 rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
4857 pid_socket_pair[1] = safe_close(pid_socket_pair[1]);
4858 uuid_socket_pair[1] = safe_close(uuid_socket_pair[1]);
4859 notify_socket_pair[1] = safe_close(notify_socket_pair[1]);
3acc84eb 4860 master_pty_socket_pair[1] = safe_close(master_pty_socket_pair[1]);
b0067625 4861 uid_shift_socket_pair[1] = safe_close(uid_shift_socket_pair[1]);
8199d554 4862 unified_cgroup_hierarchy_socket_pair[1] = safe_close(unified_cgroup_hierarchy_socket_pair[1]);
b0067625
ZJS		 4863
4864 if (arg_userns_mode != USER_NAMESPACE_NO) {
4865 /* The child just let us know the UID shift it might have read from the image. */
4866 l = recv(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, 0);
4867 if (l < 0)
4868 return log_error_errno(errno, "Failed to read UID shift: %m");
c6147113
LP		 4869 if (l != sizeof arg_uid_shift)
4870 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading UID shift.");
b0067625
ZJS		 4871
4872 if (arg_userns_mode == USER_NAMESPACE_PICK) {
4873 /* If we are supposed to pick the UID shift, let's try to use the shift read from the
4874 * image, but if that's already in use, pick a new one, and report back to the child,
4875 * which one we now picked. */
4876
4877 r = uid_shift_pick(&arg_uid_shift, &uid_shift_lock);
4878 if (r < 0)
4879 return log_error_errno(r, "Failed to pick suitable UID/GID range: %m");
4880
4881 l = send(uid_shift_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, MSG_NOSIGNAL);
4882 if (l < 0)
4883 return log_error_errno(errno, "Failed to send UID shift: %m");
c6147113
LP		 4884 if (l != sizeof arg_uid_shift)
4885 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short write while writing UID shift.");
b0067625 4886 }
2f893044
LP		 4887
4888 n_bind_user_uid = strv_length(arg_bind_user);
4889 if (n_bind_user_uid > 0) {
4890 /* Right after the UID shift, we'll receive the list of UID mappings for the
4891 * --bind-user= logic. Always a quadruplet of payload and host UID + GID. */
4892
4893 bind_user_uid = new(uid_t, n_bind_user_uid*4);
4894 if (!bind_user_uid)
4895 return log_oom();
4896
4897 for (size_t i = 0; i < n_bind_user_uid; i++) {
4898 l = recv(uid_shift_socket_pair[0], bind_user_uid + i*4, sizeof(uid_t)*4, 0);
4899 if (l < 0)
4900 return log_error_errno(errno, "Failed to read user UID map pair: %m");
4901 if (l != sizeof(uid_t)*4)
4902 return log_full_errno(l == 0 ? LOG_DEBUG : LOG_WARNING,
4903 SYNTHETIC_ERRNO(EIO),
4904 "Short read while reading bind user UID pairs.");
4905 }
4906 }
b0067625
ZJS		 4907 }
4908
8199d554
LP		 4909 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
4910 /* The child let us know the support cgroup mode it might have read from the image. */
4911 l = recv(unified_cgroup_hierarchy_socket_pair[0], &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), 0);
4912 if (l < 0)
4913 return log_error_errno(errno, "Failed to read cgroup mode: %m");
c6147113
LP		 4914 if (l != sizeof(arg_unified_cgroup_hierarchy))
4915 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading cgroup mode (%zu bytes).%s",
4916 l, l == 0 ? " The child is most likely dead." : "");
8199d554
LP		 4917 }
4918
b0067625 4919 /* Wait for the outer child. */
d2e0ac3d
LP		 4920 r = wait_for_terminate_and_check("(sd-namespace)", *pid, WAIT_LOG_ABNORMAL);
4921 if (r < 0)
4922 return r;
4923 if (r != EXIT_SUCCESS)
4924 return -EIO;
b0067625
ZJS		 4925
4926 /* And now retrieve the PID of the inner child. */
4927 l = recv(pid_socket_pair[0], pid, sizeof *pid, 0);
4928 if (l < 0)
4929 return log_error_errno(errno, "Failed to read inner child PID: %m");
c6147113
LP		 4930 if (l != sizeof *pid)
4931 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading inner child PID.");
b0067625
ZJS		 4932
4933 /* We also retrieve container UUID in case it was generated by outer child */
4934 l = recv(uuid_socket_pair[0], &arg_uuid, sizeof arg_uuid, 0);
4935 if (l < 0)
4936 return log_error_errno(errno, "Failed to read container machine ID: %m");
c6147113
LP		 4937 if (l != sizeof(arg_uuid))
4938 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading container machined ID.");
b0067625
ZJS		 4939
4940 /* We also retrieve the socket used for notifications generated by outer child */
4941 notify_socket = receive_one_fd(notify_socket_pair[0], 0);
4942 if (notify_socket < 0)
4943 return log_error_errno(notify_socket,
4944 "Failed to receive notification socket from the outer child: %m");
4945
4946 log_debug("Init process invoked as PID "PID_FMT, *pid);
4947
4948 if (arg_userns_mode != USER_NAMESPACE_NO) {
c6147113
LP		 4949 if (!barrier_place_and_sync(&barrier)) /* #1 */
4950 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early.");
b0067625 4951
2f893044 4952 r = setup_uid_map(*pid, bind_user_uid, n_bind_user_uid);
b0067625
ZJS		 4953 if (r < 0)
4954 return r;
4955
4956 (void) barrier_place(&barrier); /* #2 */
4957 }
4958
4959 if (arg_private_network) {
75116558
PS		 4960 if (!arg_network_namespace_path) {
4961 /* Wait until the child has unshared its network namespace. */
c6147113
LP		 4962 if (!barrier_place_and_sync(&barrier)) /* #3 */
4963 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early");
75116558
PS		 4964 }
4965
5b4855ab
DDM		 4966 if (child_netns_fd < 0) {
4967 /* Make sure we have an open file descriptor to the child's network
4968 * namespace so it stays alive even if the child exits. */
4969 r = namespace_open(*pid, NULL, NULL, &child_netns_fd, NULL, NULL);
4970 if (r < 0)
4971 return log_error_errno(r, "Failed to open child network namespace: %m");
4972 }
4973
4974 r = move_network_interfaces(child_netns_fd, arg_network_interfaces);
b0067625
ZJS		 4975 if (r < 0)
4976 return r;
4977
4978 if (arg_network_veth) {
4979 r = setup_veth(arg_machine, *pid, veth_name,
4980 arg_network_bridge || arg_network_zone);
4981 if (r < 0)
4982 return r;
4983 else if (r > 0)
4984 ifi = r;
4985
4986 if (arg_network_bridge) {
4987 /* Add the interface to a bridge */
4988 r = setup_bridge(veth_name, arg_network_bridge, false);
4989 if (r < 0)
4990 return r;
4991 if (r > 0)
4992 ifi = r;
4993 } else if (arg_network_zone) {
4994 /* Add the interface to a bridge, possibly creating it */
4995 r = setup_bridge(veth_name, arg_network_zone, true);
4996 if (r < 0)
4997 return r;
4998 if (r > 0)
4999 ifi = r;
5000 }
5001 }
5002
5003 r = setup_veth_extra(arg_machine, *pid, arg_network_veth_extra);
5004 if (r < 0)
5005 return r;
5006
5007 /* We created the primary and extra veth links now; let's remember this, so that we know to
5008 remove them later on. Note that we don't bother with removing veth links that were created
5009 here when their setup failed half-way, because in that case the kernel should be able to
5010 remove them on its own, since they cannot be referenced by anything yet. */
5011 *veth_created = true;
5012
5013 r = setup_macvlan(arg_machine, *pid, arg_network_macvlan);
5014 if (r < 0)
5015 return r;
5016
5017 r = setup_ipvlan(arg_machine, *pid, arg_network_ipvlan);
5018 if (r < 0)
5019 return r;
5020 }
5021
abdb9b08
LP		 5022 if (arg_register || !arg_keep_unit) {
5023 r = sd_bus_default_system(&bus);
5024 if (r < 0)
5025 return log_error_errno(r, "Failed to open system bus: %m");
e5a2d8b5
LP		 5026
5027 r = sd_bus_set_close_on_exit(bus, false);
5028 if (r < 0)
5029 return log_error_errno(r, "Failed to disable close-on-exit behaviour: %m");
abdb9b08
LP		 5030 }
5031
5032 if (!arg_keep_unit) {
5033 /* When a new scope is created for this container, then we'll be registered as its controller, in which
5034 * case PID 1 will send us a friendly RequestStop signal, when it is asked to terminate the
5035 * scope. Let's hook into that, and cleanly shut down the container, and print a friendly message. */
5036
75152a4d
LP		 5037 r = sd_bus_match_signal_async(
5038 bus,
5039 NULL,
5040 "org.freedesktop.systemd1",
5041 NULL,
5042 "org.freedesktop.systemd1.Scope",
5043 "RequestStop",
5044 on_request_stop, NULL, PID_TO_PTR(*pid));
abdb9b08 5045 if (r < 0)
75152a4d 5046 return log_error_errno(r, "Failed to request RequestStop match: %m");
abdb9b08
LP		 5047 }
5048
b0067625
ZJS		 5049 if (arg_register) {
5050 r = register_machine(
abdb9b08 5051 bus,
b0067625
ZJS		 5052 arg_machine,
5053 *pid,
5054 arg_directory,
5055 arg_uuid,
5056 ifi,
5057 arg_slice,
5058 arg_custom_mounts, arg_n_custom_mounts,
5059 arg_kill_signal,
5060 arg_property,
de40a303 5061 arg_property_message,
b0067625
ZJS		 5062 arg_keep_unit,
5063 arg_container_service_name);
5064 if (r < 0)
5065 return r;
abdb9b08 5066
cd2dfc6f
LP		 5067 } else if (!arg_keep_unit) {
5068 r = allocate_scope(
abdb9b08 5069 bus,
cd2dfc6f
LP		 5070 arg_machine,
5071 *pid,
5072 arg_slice,
5073 arg_custom_mounts, arg_n_custom_mounts,
5074 arg_kill_signal,
de40a303
LP		 5075 arg_property,
5076 arg_property_message);
cd2dfc6f
LP		 5077 if (r < 0)
5078 return r;
5079
5080 } else if (arg_slice || arg_property)
5081 log_notice("Machine and scope registration turned off, --slice= and --property= settings will have no effect.");
b0067625 5082
27da7ef0 5083 r = create_subcgroup(*pid, arg_keep_unit, arg_unified_cgroup_hierarchy);
b0067625
ZJS		 5084 if (r < 0)
5085 return r;
5086
27da7ef0 5087 r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
720f0a2f
LP		 5088 if (r < 0)
5089 return r;
b0067625 5090
de54e02d 5091 r = chown_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
b0067625
ZJS		 5092 if (r < 0)
5093 return r;
5094
5095 /* Notify the child that the parent is ready with all
5096 * its setup (including cgroup-ification), and that
5097 * the child can now hand over control to the code to
5098 * run inside the container. */
75116558 5099 (void) barrier_place(&barrier); /* #4 */
b0067625
ZJS		 5100
5101 /* Block SIGCHLD here, before notifying child.
5102 * process_pty() will handle it with the other signals. */
5103 assert_se(sigprocmask(SIG_BLOCK, &mask_chld, NULL) >= 0);
5104
5105 /* Reset signal to default */
9c274488 5106 r = default_signals(SIGCHLD);
b0067625
ZJS		 5107 if (r < 0)
5108 return log_error_errno(r, "Failed to reset SIGCHLD: %m");
5109
5110 r = sd_event_new(&event);
5111 if (r < 0)
5112 return log_error_errno(r, "Failed to get default event source: %m");
5113
8fd010bb
LP		 5114 (void) sd_event_set_watchdog(event, true);
5115
abdb9b08
LP		 5116 if (bus) {
5117 r = sd_bus_attach_event(bus, event, 0);
5118 if (r < 0)
5119 return log_error_errno(r, "Failed to attach bus to event loop: %m");
5120 }
5121
e96ceaba 5122 r = setup_notify_parent(event, notify_socket, PID_TO_PTR(*pid), &notify_event_source);
b0067625
ZJS		 5123 if (r < 0)
5124 return r;
5125
5126 /* Let the child know that we are ready and wait that the child is completely ready now. */
c6147113
LP		 5127 if (!barrier_place_and_sync(&barrier)) /* #5 */
5128 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early.");
b0067625 5129
38ccb557 5130 /* At this point we have made use of the UID we picked, and thus nss-systemd/systemd-machined.service
b0067625
ZJS		 5131 * will make them appear in getpwuid(), thus we can release the /etc/passwd lock. */
5132 etc_passwd_lock = safe_close(etc_passwd_lock);
5133
04f590a4
LP		 5134 (void) sd_notifyf(false,
5135 "STATUS=Container running.\n"
5136 "X_NSPAWN_LEADER_PID=" PID_FMT, *pid);
4bf4f50f
ZJS		 5137 if (!arg_notify_ready) {
5138 r = sd_notify(false, "READY=1\n");
5139 if (r < 0)
5140 log_warning_errno(r, "Failed to send readiness notification, ignoring: %m");
5141 }
b0067625
ZJS		 5142
5143 if (arg_kill_signal > 0) {
5144 /* Try to kill the init system on SIGINT or SIGTERM */
919f5ae0
LP		 5145 (void) sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, PID_TO_PTR(*pid));
5146 (void) sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, PID_TO_PTR(*pid));
b0067625
ZJS		 5147 } else {
5148 /* Immediately exit */
919f5ae0
LP		 5149 (void) sd_event_add_signal(event, NULL, SIGINT, NULL, NULL);
5150 (void) sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL);
b0067625
ZJS		 5151 }
5152
6916b164 5153 /* Exit when the child exits */
919f5ae0 5154 (void) sd_event_add_signal(event, NULL, SIGCHLD, on_sigchld, PID_TO_PTR(*pid));
b0067625
ZJS		 5155
5156 if (arg_expose_ports) {
761cf19d 5157 r = expose_port_watch_rtnl(event, rtnl_socket_pair[0], on_address_change, expose_args, &rtnl);
b0067625
ZJS		 5158 if (r < 0)
5159 return r;
5160
deff68e7
FW		 5161 (void) expose_port_execute(rtnl, &expose_args->fw_ctx, arg_expose_ports, AF_INET, &expose_args->address4);
5162 (void) expose_port_execute(rtnl, &expose_args->fw_ctx, arg_expose_ports, AF_INET6, &expose_args->address6);
b0067625
ZJS		 5163 }
5164
5165 rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
5166
3acc84eb
FB		 5167 if (arg_console_mode != CONSOLE_PIPE) {
5168 _cleanup_close_ int fd = -1;
5169 PTYForwardFlags flags = 0;
de40a303 5170
3acc84eb
FB		 5171 /* Retrieve the master pty allocated by inner child */
5172 fd = receive_one_fd(master_pty_socket_pair[0], 0);
5173 if (fd < 0)
5174 return log_error_errno(fd, "Failed to receive master pty from the inner child: %m");
5175
5176 switch (arg_console_mode) {
de40a303 5177
3acc84eb
FB		 5178 case CONSOLE_READ_ONLY:
5179 flags |= PTY_FORWARD_READ_ONLY;
5180
5181 _fallthrough_;
5182
5183 case CONSOLE_INTERACTIVE:
5184 flags |= PTY_FORWARD_IGNORE_VHANGUP;
5185
5186 r = pty_forward_new(event, fd, flags, &forward);
5187 if (r < 0)
5188 return log_error_errno(r, "Failed to create PTY forwarder: %m");
5189
f5fbe71d 5190 if (arg_console_width != UINT_MAX || arg_console_height != UINT_MAX)
3acc84eb
FB		 5191 (void) pty_forward_set_width_height(forward,
5192 arg_console_width,
5193 arg_console_height);
5194 break;
5195
5196 default:
5197 assert(arg_console_mode == CONSOLE_PASSIVE);
5198 }
5199
5200 *master = TAKE_FD(fd);
de40a303 5201 }
b0067625
ZJS		 5202
5203 r = sd_event_loop(event);
5204 if (r < 0)
5205 return log_error_errno(r, "Failed to run event loop: %m");
5206
de40a303
LP		 5207 if (forward) {
5208 char last_char = 0;
b0067625 5209
de40a303
LP		 5210 (void) pty_forward_get_last_char(forward, &last_char);
5211 forward = pty_forward_free(forward);
b0067625 5212
de40a303
LP		 5213 if (!arg_quiet && last_char != '\n')
5214 putc('\n', stdout);
5215 }
b0067625
ZJS		 5216
5217 /* Kill if it is not dead yet anyway */
0bb0a9fa
ZJS		 5218 if (!arg_register && !arg_keep_unit && bus)
5219 terminate_scope(bus, arg_machine);
b0067625
ZJS		 5220
5221 /* Normally redundant, but better safe than sorry */
c67b0082 5222 (void) kill(*pid, SIGKILL);
b0067625 5223
5b4855ab
DDM		 5224 if (arg_private_network) {
5225 /* Move network interfaces back to the parent network namespace. We use `safe_fork`
5226 * to avoid having to move the parent to the child network namespace. */
5227 r = safe_fork(NULL, FORK_RESET_SIGNALS|FORK_DEATHSIG|FORK_WAIT|FORK_LOG, NULL);
5228 if (r < 0)
5229 return r;
5230
5231 if (r == 0) {
5232 _cleanup_close_ int parent_netns_fd = -1;
5233
5234 r = namespace_open(getpid(), NULL, NULL, &parent_netns_fd, NULL, NULL);
5235 if (r < 0) {
5236 log_error_errno(r, "Failed to open parent network namespace: %m");
5237 _exit(EXIT_FAILURE);
5238 }
5239
5240 r = namespace_enter(-1, -1, child_netns_fd, -1, -1);
5241 if (r < 0) {
5242 log_error_errno(r, "Failed to enter child network namespace: %m");
5243 _exit(EXIT_FAILURE);
5244 }
5245
5246 r = move_network_interfaces(parent_netns_fd, arg_network_interfaces);
5247 if (r < 0)
5248 log_error_errno(r, "Failed to move network interfaces back to parent network namespace: %m");
5249
5250 _exit(r < 0 ? EXIT_FAILURE : EXIT_SUCCESS);
5251 }
5252 }
5253
8f03de53 5254 r = wait_for_container(TAKE_PID(*pid), &container_status);
b0067625 5255
0bb0a9fa
ZJS		 5256 /* Tell machined that we are gone. */
5257 if (bus)
5258 (void) unregister_machine(bus, arg_machine);
5259
b0067625
ZJS		 5260 if (r < 0)
5261 /* We failed to wait for the container, or the container exited abnormally. */
5262 return r;
5263 if (r > 0 || container_status == CONTAINER_TERMINATED) {
27e29a1e
ZJS		 5264 /* r > 0 → The container exited with a non-zero status.
5265 * As a special case, we need to replace 133 with a different value,
5266 * because 133 is special-cased in the service file to reboot the container.
5267 * otherwise → The container exited with zero status and a reboot was not requested.
5268 */
2a49b612 5269 if (r == EXIT_FORCE_RESTART)
27e29a1e 5270 r = EXIT_FAILURE; /* replace 133 with the general failure code */
b0067625 5271 *ret = r;
b0067625
ZJS		 5272 return 0; /* finito */
5273 }
5274
5275 /* CONTAINER_REBOOTED, loop again */
5276
5277 if (arg_keep_unit) {
5278 /* Special handling if we are running as a service: instead of simply
5279 * restarting the machine we want to restart the entire service, so let's
5280 * inform systemd about this with the special exit code 133. The service
5281 * file uses RestartForceExitStatus=133 so that this results in a full
5282 * nspawn restart. This is necessary since we might have cgroup parameters
5283 * set we want to have flushed out. */
2a49b612
ZJS		 5284 *ret = EXIT_FORCE_RESTART;
5285 return 0; /* finito */
b0067625
ZJS		 5286 }
5287
deff68e7
FW		 5288 expose_port_flush(&expose_args->fw_ctx, arg_expose_ports, AF_INET, &expose_args->address4);
5289 expose_port_flush(&expose_args->fw_ctx, arg_expose_ports, AF_INET6, &expose_args->address6);
b0067625
ZJS		 5290
5291 (void) remove_veth_links(veth_name, arg_network_veth_extra);
5292 *veth_created = false;
5293 return 1; /* loop again */
5294}
5295
bf428efb 5296static int initialize_rlimits(void) {
852b6250 5297 /* The default resource limits the kernel passes to PID 1, as per kernel 5.16. Let's pass our container payload
bf428efb
LP		 5298 * the same values as the kernel originally passed to PID 1, in order to minimize differences between host and
5299 * container execution environments. */
5300
5301 static const struct rlimit kernel_defaults[_RLIMIT_MAX] = {
852b6250
LP		 5302 [RLIMIT_AS] = { RLIM_INFINITY, RLIM_INFINITY },
5303 [RLIMIT_CORE] = { 0, RLIM_INFINITY },
5304 [RLIMIT_CPU] = { RLIM_INFINITY, RLIM_INFINITY },
5305 [RLIMIT_DATA] = { RLIM_INFINITY, RLIM_INFINITY },
5306 [RLIMIT_FSIZE] = { RLIM_INFINITY, RLIM_INFINITY },
5307 [RLIMIT_LOCKS] = { RLIM_INFINITY, RLIM_INFINITY },
5308 [RLIMIT_MEMLOCK] = { DEFAULT_RLIMIT_MEMLOCK, DEFAULT_RLIMIT_MEMLOCK },
5309 [RLIMIT_MSGQUEUE] = { 819200, 819200 },
5310 [RLIMIT_NICE] = { 0, 0 },
5311 [RLIMIT_NOFILE] = { 1024, 4096 },
5312 [RLIMIT_RSS] = { RLIM_INFINITY, RLIM_INFINITY },
5313 [RLIMIT_RTPRIO] = { 0, 0 },
5314 [RLIMIT_RTTIME] = { RLIM_INFINITY, RLIM_INFINITY },
5315 [RLIMIT_STACK] = { 8388608, RLIM_INFINITY },
bf428efb
LP		 5316
5317 /* The kernel scales the default for RLIMIT_NPROC and RLIMIT_SIGPENDING based on the system's amount of
5318 * RAM. To provide best compatibility we'll read these limits off PID 1 instead of hardcoding them
5319 * here. This is safe as we know that PID 1 doesn't change these two limits and thus the original
5320 * kernel's initialization should still be valid during runtime — at least if PID 1 is systemd. Note
5321 * that PID 1 changes a number of other resource limits during early initialization which is why we
5322 * don't read the other limits from PID 1 but prefer the static table above. */
5323 };
5324
5325 int rl;
5326
5327 for (rl = 0; rl < _RLIMIT_MAX; rl++) {
bf428efb
LP		 5328 /* Let's only fill in what the user hasn't explicitly configured anyway */
5329 if ((arg_settings_mask & (SETTING_RLIMIT_FIRST << rl)) == 0) {
5330 const struct rlimit *v;
5331 struct rlimit buffer;
5332
5333 if (IN_SET(rl, RLIMIT_NPROC, RLIMIT_SIGPENDING)) {
5334 /* For these two let's read the limits off PID 1. See above for an explanation. */
5335
5336 if (prlimit(1, rl, NULL, &buffer) < 0)
5337 return log_error_errno(errno, "Failed to read resource limit RLIMIT_%s of PID 1: %m", rlimit_to_string(rl));
5338
dbf1aca6
LP		 5339 v = &buffer;
5340 } else if (rl == RLIMIT_NOFILE) {
5341 /* We nowadays bump RLIMIT_NOFILE's hard limit early in PID 1 for all
5342 * userspace. Given that nspawn containers are often run without our PID 1,
5343 * let's grant the containers a raised RLIMIT_NOFILE hard limit by default,
5344 * so that container userspace gets similar resources as host userspace
5345 * gets. */
5346 buffer = kernel_defaults[rl];
5347 buffer.rlim_max = MIN((rlim_t) read_nr_open(), (rlim_t) HIGH_RLIMIT_NOFILE);
bf428efb
LP		 5348 v = &buffer;
5349 } else
5350 v = kernel_defaults + rl;
5351
5352 arg_rlimit[rl] = newdup(struct rlimit, v, 1);
5353 if (!arg_rlimit[rl])
5354 return log_oom();
5355 }
5356
5357 if (DEBUG_LOGGING) {
5358 _cleanup_free_ char *k = NULL;
5359
5360 (void) rlimit_format(arg_rlimit[rl], &k);
5361 log_debug("Setting RLIMIT_%s to %s.", rlimit_to_string(rl), k);
5362 }
5363 }
5364
5365 return 0;
5366}
5367
287b7376
LP		 5368static int cant_be_in_netns(void) {
5369 union sockaddr_union sa = {
5370 .un = {
5371 .sun_family = AF_UNIX,
5372 .sun_path = "/run/udev/control",
5373 },
5374 };
5375 char udev_path[STRLEN("/proc//ns/net") + DECIMAL_STR_MAX(pid_t)];
5376 _cleanup_free_ char *udev_ns = NULL, *our_ns = NULL;
5377 _cleanup_close_ int fd = -1;
5378 struct ucred ucred;
5379 int r;
5380
5381 /* Check if we are in the same netns as udev. If we aren't, then device monitoring (and thus waiting
5382 * for loopback block devices) won't work, and we will hang. Detect this case and exit early with a
5383 * nice message. */
5384
5385 if (!arg_image) /* only matters if --image= us used, i.e. we actually need to use loopback devices */
5386 return 0;
5387
5388 fd = socket(AF_UNIX, SOCK_SEQPACKET|SOCK_NONBLOCK|SOCK_CLOEXEC, 0);
5389 if (fd < 0)
5390 return log_error_errno(errno, "Failed to allocate udev control socket: %m");
5391
32b9736a 5392 if (connect(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un)) < 0) {
287b7376
LP		 5393
5394 if (errno == ENOENT || ERRNO_IS_DISCONNECT(errno))
5395 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
5396 "Sorry, but --image= requires access to the host's /run/ hierarchy, since we need access to udev.");
5397
5398 return log_error_errno(errno, "Failed to connect socket to udev control socket: %m");
5399 }
5400
5401 r = getpeercred(fd, &ucred);
5402 if (r < 0)
5403 return log_error_errno(r, "Failed to determine peer of udev control socket: %m");
5404
5405 xsprintf(udev_path, "/proc/" PID_FMT "/ns/net", ucred.pid);
5406 r = readlink_malloc(udev_path, &udev_ns);
5407 if (r < 0)
5408 return log_error_errno(r, "Failed to read network namespace of udev: %m");
5409
5410 r = readlink_malloc("/proc/self/ns/net", &our_ns);
5411 if (r < 0)
5412 return log_error_errno(r, "Failed to read our own network namespace: %m");
5413
5414 if (!streq(our_ns, udev_ns))
5415 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
5416 "Sorry, but --image= is only supported in the main network namespace, since we need access to udev/AF_NETLINK.");
5417 return 0;
5418}
5419
44dbef90 5420static int run(int argc, char *argv[]) {
7bf011e3
LP		 5421 bool secondary = false, remove_directory = false, remove_image = false,
5422 veth_created = false, remove_tmprootdir = false;
2d845785 5423 _cleanup_close_ int master = -1;
03cfe0d5 5424 _cleanup_fdset_free_ FDSet *fds = NULL;
2d845785 5425 int r, n_fd_passed, ret = EXIT_SUCCESS;
5aa3eba5 5426 char veth_name[IFNAMSIZ] = "";
761cf19d 5427 struct ExposeArgs expose_args = {};
8e766630 5428 _cleanup_(release_lock_file) LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT;
c67b0082 5429 char tmprootdir[] = "/tmp/nspawn-root-XXXXXX";
2d845785 5430 _cleanup_(loop_device_unrefp) LoopDevice *loop = NULL;
18b5886e
LP		 5431 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
5432 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
761cf19d 5433 _cleanup_(fw_ctx_freep) FirewallContext *fw_ctx = NULL;
7bf011e3 5434 pid_t pid = 0;
03cfe0d5
LP		 5435
5436 log_parse_environment();
5437 log_open();
415fc41c 5438
03cfe0d5
LP		 5439 r = parse_argv(argc, argv);
5440 if (r <= 0)
5441 goto finish;
5442
38ee19c0
ZJS		 5443 if (geteuid() != 0) {
5444 r = log_warning_errno(SYNTHETIC_ERRNO(EPERM),
5445 argc >= 2 ? "Need to be root." :
5446 "Need to be root (and some arguments are usually required).\nHint: try --help");
03cfe0d5 5447 goto finish;
38ee19c0 5448 }
fba868fa 5449
287b7376
LP		 5450 r = cant_be_in_netns();
5451 if (r < 0)
5452 goto finish;
5453
bf428efb
LP		 5454 r = initialize_rlimits();
5455 if (r < 0)
5456 goto finish;
5457
de40a303
LP		 5458 r = load_oci_bundle();
5459 if (r < 0)
5460 goto finish;
5461
f757855e
LP		 5462 r = determine_names();
5463 if (r < 0)
5464 goto finish;
5465
5466 r = load_settings();
5467 if (r < 0)
5468 goto finish;
5469
d4d99bc6 5470 r = cg_unified();
5eee8290
LP		 5471 if (r < 0) {
5472 log_error_errno(r, "Failed to determine whether the unified cgroups hierarchy is used: %m");
5473 goto finish;
5474 }
5475
f757855e
LP		 5476 r = verify_arguments();
5477 if (r < 0)
5478 goto finish;
03cfe0d5 5479
49048684
ZJS		 5480 /* Reapply environment settings. */
5481 (void) detect_unified_cgroup_hierarchy_from_environment();
8199d554 5482
2949ff26
LP		 5483 /* Ignore SIGPIPE here, because we use splice() on the ptyfwd stuff and that will generate SIGPIPE if
5484 * the result is closed. Note that the container payload child will reset signal mask+handler anyway,
5485 * so just turning this off here means we only turn it off in nspawn itself, not any children. */
9c274488 5486 (void) ignore_signals(SIGPIPE);
2949ff26 5487
03cfe0d5
LP		 5488 n_fd_passed = sd_listen_fds(false);
5489 if (n_fd_passed > 0) {
5490 r = fdset_new_listen_fds(&fds, false);
5491 if (r < 0) {
5492 log_error_errno(r, "Failed to collect file descriptors: %m");
5493 goto finish;
5494 }
5495 }
5496
83e803a9
ZJS		 5497 /* The "default" umask. This is appropriate for most file and directory
5498 * operations performed by nspawn, and is the umask that will be used for
5499 * the child. Functions like copy_devnodes() change the umask temporarily. */
5500 umask(0022);
5501
03cfe0d5
LP		 5502 if (arg_directory) {
5503 assert(!arg_image);
5504
b35ca61a
LP		 5505 /* Safety precaution: let's not allow running images from the live host OS image, as long as
5506 * /var from the host will propagate into container dynamically (because bad things happen if
5507 * two systems write to the same /var). Let's allow it for the special cases where /var is
5508 * either copied (i.e. --ephemeral) or replaced (i.e. --volatile=yes|state). */
5509 if (path_equal(arg_directory, "/") && !(arg_ephemeral || IN_SET(arg_volatile_mode, VOLATILE_YES, VOLATILE_STATE))) {
5510 log_error("Spawning container on root directory is not supported. Consider using --ephemeral, --volatile=yes or --volatile=state.");
03cfe0d5
LP		 5511 r = -EINVAL;
5512 goto finish;
5513 }
5514
5515 if (arg_ephemeral) {
5516 _cleanup_free_ char *np = NULL;
5517
8d4aa2bb 5518 r = chase_symlinks_and_update(&arg_directory, 0);
3f342ec4
LP		 5519 if (r < 0)
5520 goto finish;
5521
7bf011e3
LP		 5522 /* If the specified path is a mount point we generate the new snapshot immediately
5523 * inside it under a random name. However if the specified is not a mount point we
5524 * create the new snapshot in the parent directory, just next to it. */
e1873695 5525 r = path_is_mount_point(arg_directory, NULL, 0);
03cfe0d5
LP		 5526 if (r < 0) {
5527 log_error_errno(r, "Failed to determine whether directory %s is mount point: %m", arg_directory);
5528 goto finish;
5529 }
5530 if (r > 0)
770b5ce4 5531 r = tempfn_random_child(arg_directory, "machine.", &np);
03cfe0d5 5532 else
770b5ce4 5533 r = tempfn_random(arg_directory, "machine.", &np);
03cfe0d5 5534 if (r < 0) {
0f3be6ca 5535 log_error_errno(r, "Failed to generate name for directory snapshot: %m");
03cfe0d5
LP		 5536 goto finish;
5537 }
5538
6992459c 5539 /* We take an exclusive lock on this image, since it's our private, ephemeral copy
162392b7 5540 * only owned by us and no one else. */
6992459c 5541 r = image_path_lock(np, LOCK_EX|LOCK_NB, &tree_global_lock, &tree_local_lock);
03cfe0d5
LP		 5542 if (r < 0) {
5543 log_error_errno(r, "Failed to lock %s: %m", np);
5544 goto finish;
5545 }
5546
7bf011e3
LP		 5547 {
5548 BLOCK_SIGNALS(SIGINT);
5549 r = btrfs_subvol_snapshot(arg_directory, np,
5550 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
5551 BTRFS_SNAPSHOT_FALLBACK_COPY |
5552 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
5553 BTRFS_SNAPSHOT_RECURSIVE |
5554 BTRFS_SNAPSHOT_QUOTA |
5555 BTRFS_SNAPSHOT_SIGINT);
5556 }
5557 if (r == -EINTR) {
5558 log_error_errno(r, "Interrupted while copying file system tree to %s, removed again.", np);
5559 goto finish;
5560 }
03cfe0d5
LP		 5561 if (r < 0) {
5562 log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory);
5563 goto finish;
ec16945e
LP		 5564 }
5565
1cc6c93a 5566 free_and_replace(arg_directory, np);
17cbb288 5567 remove_directory = true;
30535c16 5568 } else {
cb638b5e 5569 r = chase_symlinks_and_update(&arg_directory, arg_template ? CHASE_NONEXISTENT : 0);
8d4aa2bb
LP		 5570 if (r < 0)
5571 goto finish;
5572
30535c16
LP		 5573 r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
5574 if (r == -EBUSY) {
5575 log_error_errno(r, "Directory tree %s is currently busy.", arg_directory);
5576 goto finish;
5577 }
5578 if (r < 0) {
5579 log_error_errno(r, "Failed to lock %s: %m", arg_directory);
476b8254 5580 goto finish;
30535c16
LP		 5581 }
5582
5583 if (arg_template) {
8d4aa2bb 5584 r = chase_symlinks_and_update(&arg_template, 0);
3f342ec4
LP		 5585 if (r < 0)
5586 goto finish;
5587
7bf011e3
LP		 5588 {
5589 BLOCK_SIGNALS(SIGINT);
5590 r = btrfs_subvol_snapshot(arg_template, arg_directory,
5591 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
5592 BTRFS_SNAPSHOT_FALLBACK_COPY |
5593 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
5594 BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE |
5595 BTRFS_SNAPSHOT_RECURSIVE |
5596 BTRFS_SNAPSHOT_QUOTA |
5597 BTRFS_SNAPSHOT_SIGINT);
5598 }
ff6c6cc1
LP		 5599 if (r == -EEXIST)
5600 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
5601 "Directory %s already exists, not populating from template %s.", arg_directory, arg_template);
7bf011e3
LP		 5602 else if (r == -EINTR) {
5603 log_error_errno(r, "Interrupted while copying file system tree to %s, removed again.", arg_directory);
5604 goto finish;
5605 } else if (r < 0) {
83521414 5606 log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template);
30535c16 5607 goto finish;
ff6c6cc1
LP		 5608 } else
5609 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
5610 "Populated %s from template %s.", arg_directory, arg_template);
30535c16 5611 }
ec16945e
LP		 5612 }
5613
7732f92b 5614 if (arg_start_mode == START_BOOT) {
a5201ed6 5615 const char *p;
c9fe05e0 5616
a5201ed6
LP		 5617 if (arg_pivot_root_new)
5618 p = prefix_roota(arg_directory, arg_pivot_root_new);
5619 else
5620 p = arg_directory;
c9fe05e0
AR		 5621
5622 if (path_is_os_tree(p) <= 0) {
5623 log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", p);
ec16945e 5624 r = -EINVAL;
1b9e5b12
LP		 5625 goto finish;
5626 }
5627 } else {
c9fe05e0
AR		 5628 const char *p, *q;
5629
a5201ed6
LP		 5630 if (arg_pivot_root_new)
5631 p = prefix_roota(arg_directory, arg_pivot_root_new);
5632 else
5633 p = arg_directory;
c9fe05e0
AR		 5634
5635 q = strjoina(p, "/usr/");
1b9e5b12 5636
c9fe05e0
AR		 5637 if (laccess(q, F_OK) < 0) {
5638 log_error("Directory %s doesn't look like it has an OS tree. Refusing.", p);
ec16945e 5639 r = -EINVAL;
1b9e5b12 5640 goto finish;
1b9e5b12
LP		 5641 }
5642 }
ec16945e 5643
6b9132a9 5644 } else {
d04faa4e 5645 DissectImageFlags dissect_image_flags =
4b5de5dd 5646 DISSECT_IMAGE_GENERIC_ROOT |
d04faa4e
LP		 5647 DISSECT_IMAGE_REQUIRE_ROOT |
5648 DISSECT_IMAGE_RELAX_VAR_CHECK |
5649 DISSECT_IMAGE_USR_NO_ROOT;
ec16945e
LP		 5650 assert(arg_image);
5651 assert(!arg_template);
5652
8d4aa2bb 5653 r = chase_symlinks_and_update(&arg_image, 0);
3f342ec4
LP		 5654 if (r < 0)
5655 goto finish;
5656
0f3be6ca
LP		 5657 if (arg_ephemeral) {
5658 _cleanup_free_ char *np = NULL;
5659
5660 r = tempfn_random(arg_image, "machine.", &np);
5661 if (r < 0) {
5662 log_error_errno(r, "Failed to generate name for image snapshot: %m");
5663 goto finish;
5664 }
5665
6992459c
LP		 5666 /* Always take an exclusive lock on our own ephemeral copy. */
5667 r = image_path_lock(np, LOCK_EX|LOCK_NB, &tree_global_lock, &tree_local_lock);
0f3be6ca
LP		 5668 if (r < 0) {
5669 r = log_error_errno(r, "Failed to create image lock: %m");
5670 goto finish;
5671 }
5672
7bf011e3
LP		 5673 {
5674 BLOCK_SIGNALS(SIGINT);
5675 r = copy_file(arg_image, np, O_EXCL, arg_read_only ? 0400 : 0600, FS_NOCOW_FL, FS_NOCOW_FL, COPY_REFLINK|COPY_CRTIME|COPY_SIGINT);
5676 }
5677 if (r == -EINTR) {
5678 log_error_errno(r, "Interrupted while copying image file to %s, removed again.", np);
5679 goto finish;
5680 }
0f3be6ca
LP		 5681 if (r < 0) {
5682 r = log_error_errno(r, "Failed to copy image file: %m");
5683 goto finish;
5684 }
5685
1cc6c93a 5686 free_and_replace(arg_image, np);
0f3be6ca
LP		 5687 remove_image = true;
5688 } else {
5689 r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
5690 if (r == -EBUSY) {
5691 r = log_error_errno(r, "Disk image %s is currently busy.", arg_image);
5692 goto finish;
5693 }
5694 if (r < 0) {
5695 r = log_error_errno(r, "Failed to create image lock: %m");
5696 goto finish;
5697 }
4623e8e6 5698
89e62e0b
LP		 5699 r = verity_settings_load(
5700 &arg_verity_settings,
5701 arg_image, NULL, NULL);
e7cbe5cb
LB		 5702 if (r < 0) {
5703 log_error_errno(r, "Failed to read verity artefacts for %s: %m", arg_image);
5704 goto finish;
78ebe980 5705 }
89e62e0b
LP		 5706
5707 if (arg_verity_settings.data_path)
5708 dissect_image_flags |= DISSECT_IMAGE_NO_PARTITION_TABLE;
30535c16
LP		 5709 }
5710
c67b0082 5711 if (!mkdtemp(tmprootdir)) {
0f3be6ca 5712 r = log_error_errno(errno, "Failed to create temporary directory: %m");
6b9132a9 5713 goto finish;
1b9e5b12 5714 }
6b9132a9 5715
c67b0082
LP		 5716 remove_tmprootdir = true;
5717
5718 arg_directory = strdup(tmprootdir);
1b9e5b12
LP		 5719 if (!arg_directory) {
5720 r = log_oom();
5721 goto finish;
6b9132a9 5722 }
88213476 5723
89e62e0b
LP		 5724 r = loop_device_make_by_path(
5725 arg_image,
5726 arg_read_only ? O_RDONLY : O_RDWR,
5727 FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN,
5728 &loop);
2d845785
LP		 5729 if (r < 0) {
5730 log_error_errno(r, "Failed to set up loopback block device: %m");
842f3b0f
LP		 5731 goto finish;
5732 }
1b9e5b12 5733
4526113f 5734 r = dissect_image_and_warn(
e0f9e7bd 5735 loop->fd,
4526113f 5736 arg_image,
89e62e0b 5737 &arg_verity_settings,
18d73705 5738 NULL,
a3642997 5739 loop->diskseq,
75dc190d 5740 loop->uevent_seqnum_not_before,
4a62257d 5741 loop->timestamp_not_before,
e7cbe5cb 5742 dissect_image_flags,
e0f9e7bd 5743 &dissected_image);
2d845785 5744 if (r == -ENOPKG) {
4526113f 5745 /* dissected_image_and_warn() already printed a brief error message. Extend on that with more details */
2d845785
LP		 5746 log_notice("Note that the disk image needs to\n"
5747 " a) either contain only a single MBR partition of type 0x83 that is marked bootable\n"
5748 " b) or contain a single GPT partition of type 0FC63DAF-8483-4772-8E79-3D69D8477DE4\n"
19ac32cd 5749 " c) or follow https://systemd.io/DISCOVERABLE_PARTITIONS\n"
2d845785
LP		 5750 " d) or contain a file system without a partition table\n"
5751 "in order to be bootable with systemd-nspawn.");
1b9e5b12 5752 goto finish;
2d845785 5753 }
4526113f 5754 if (r < 0)
842f3b0f 5755 goto finish;
1b9e5b12 5756
88b3300f
LP		 5757 r = dissected_image_load_verity_sig_partition(
5758 dissected_image,
5759 loop->fd,
5760 &arg_verity_settings);
5761 if (r < 0)
5762 goto finish;
5763
8ee9615e
LP		 5764 if (dissected_image->has_verity && !arg_verity_settings.root_hash && !dissected_image->has_verity_sig)
5765 log_notice("Note: image %s contains verity information, but no root hash specified and no embedded "
5766 "root hash signature found! Proceeding without integrity checking.", arg_image);
4623e8e6 5767
89e62e0b
LP		 5768 r = dissected_image_decrypt_interactively(
5769 dissected_image,
5770 NULL,
5771 &arg_verity_settings,
5772 0,
5773 &decrypted_image);
1b9e5b12
LP		 5774 if (r < 0)
5775 goto finish;
0f3be6ca
LP		 5776
5777 /* Now that we mounted the image, let's try to remove it again, if it is ephemeral */
5778 if (remove_image && unlink(arg_image) >= 0)
5779 remove_image = false;
842f3b0f 5780 }
842f3b0f 5781
86c0dd4a 5782 r = custom_mount_prepare_all(arg_directory, arg_custom_mounts, arg_n_custom_mounts);
5a8af538
LP		 5783 if (r < 0)
5784 goto finish;
5785
de40a303
LP		 5786 if (arg_console_mode < 0)
5787 arg_console_mode =
5788 isatty(STDIN_FILENO) > 0 &&
5789 isatty(STDOUT_FILENO) > 0 ? CONSOLE_INTERACTIVE : CONSOLE_READ_ONLY;
9c857b9d 5790
de40a303
LP		 5791 if (arg_console_mode == CONSOLE_PIPE) /* if we pass STDERR on to the container, don't add our own logs into it too */
5792 arg_quiet = true;
a258bf26 5793
9c857b9d
LP		 5794 if (!arg_quiet)
5795 log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.",
5796 arg_machine, arg_image ?: arg_directory);
5797
72c0a2c2 5798 assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1) >= 0);
a258bf26 5799
66edd963 5800 if (prctl(PR_SET_CHILD_SUBREAPER, 1, 0, 0, 0) < 0) {
03cfe0d5
LP		 5801 r = log_error_errno(errno, "Failed to become subreaper: %m");
5802 goto finish;
5803 }
5804
761cf19d
FW		 5805 if (arg_expose_ports) {
5806 r = fw_ctx_new(&fw_ctx);
5807 if (r < 0) {
5808 log_error_errno(r, "Cannot expose configured ports, firewall initialization failed: %m");
5809 goto finish;
5810 }
5811 expose_args.fw_ctx = fw_ctx;
5812 }
d87be9b0 5813 for (;;) {
3acc84eb 5814 r = run_container(dissected_image,
44dbef90
LP		 5815 secondary,
5816 fds,
5817 veth_name, &veth_created,
761cf19d 5818 &expose_args, &master,
44dbef90 5819 &pid, &ret);
b0067625 5820 if (r <= 0)
d87be9b0 5821 break;
d87be9b0 5822 }
88213476
LP		 5823
5824finish:
04f590a4
LP		 5825 (void) sd_notify(false,
5826 r == 0 && ret == EXIT_FORCE_RESTART ? "STOPPING=1\nSTATUS=Restarting..." :
5827 "STOPPING=1\nSTATUS=Terminating...");
af4ec430 5828
9444b1f2 5829 if (pid > 0)
c67b0082 5830 (void) kill(pid, SIGKILL);
88213476 5831
503546da 5832 /* Try to flush whatever is still queued in the pty */
6a0f896b 5833 if (master >= 0) {
f5fbe71d 5834 (void) copy_bytes(master, STDOUT_FILENO, UINT64_MAX, 0);
6a0f896b
LP		 5835 master = safe_close(master);
5836 }
5837
5838 if (pid > 0)
5839 (void) wait_for_terminate(pid, NULL);
503546da 5840
50ebcf6c
LP		 5841 pager_close();
5842
17cbb288 5843 if (remove_directory && arg_directory) {
ec16945e
LP		 5844 int k;
5845
17cbb288 5846 k = rm_rf(arg_directory, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME);
ec16945e 5847 if (k < 0)
17cbb288 5848 log_warning_errno(k, "Cannot remove '%s', ignoring: %m", arg_directory);
ec16945e
LP		 5849 }
5850
0f3be6ca
LP		 5851 if (remove_image && arg_image) {
5852 if (unlink(arg_image) < 0)
5853 log_warning_errno(errno, "Can't remove image file '%s', ignoring: %m", arg_image);
5854 }
5855
c67b0082
LP		 5856 if (remove_tmprootdir) {
5857 if (rmdir(tmprootdir) < 0)
5858 log_debug_errno(errno, "Can't remove temporary root directory '%s', ignoring: %m", tmprootdir);
5859 }
5860
785890ac
LP		 5861 if (arg_machine) {
5862 const char *p;
5863
63c372cb 5864 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
c6878637 5865 (void) rm_rf(p, REMOVE_ROOT);
785890ac
LP		 5866 }
5867
deff68e7
FW		 5868 expose_port_flush(&fw_ctx, arg_expose_ports, AF_INET, &expose_args.address4);
5869 expose_port_flush(&fw_ctx, arg_expose_ports, AF_INET6, &expose_args.address6);
7513c5b8
LP		 5870
5871 if (veth_created)
5872 (void) remove_veth_links(veth_name, arg_network_veth_extra);
22b28dfd 5873 (void) remove_bridge(arg_network_zone);
f757855e 5874
f757855e
LP		 5875 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
5876 expose_port_free_all(arg_expose_ports);
bf428efb 5877 rlimit_free_all(arg_rlimit);
b2645747 5878 device_node_array_free(arg_extra_nodes, arg_n_extra_nodes);
3652872a 5879 credential_free_all(arg_credentials, arg_n_credentials);
6d0b55c2 5880
44dbef90
LP		 5881 if (r < 0)
5882 return r;
5883
5884 return ret;
88213476 5885}
44dbef90
LP		 5886
5887DEFINE_MAIN_FUNCTION_WITH_POSITIVE_FAILURE(run);
Unnamed repository; edit this file 'description' to name the repository.