]> 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
core: escape spaces in paths during serialization
[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"
813dbff4 27#include "ether-addr-util.h"
8fe0087e
LP		 28#include "barrier.h"
29#include "base-filesystem.h"
30#include "blkid-util.h"
31#include "btrfs-util.h"
d6b4d1c7 32#include "build.h"
b8ea7a6e 33#include "bus-error.h"
7f8a85e6 34#include "bus-locator.h"
b053cd5f 35#include "bus-util.h"
8fe0087e 36#include "cap-list.h"
430f0182 37#include "capability-util.h"
04d391da 38#include "cgroup-util.h"
f461a28d 39#include "chase.h"
988851b6 40#include "common-signal.h"
8fe0087e 41#include "copy.h"
d107bb7d 42#include "cpu-set-util.h"
786d19fd 43#include "creds-util.h"
4fc9982c 44#include "dev-setup.h"
57f1b61b 45#include "discover-image.h"
2d845785 46#include "dissect-image.h"
8fe0087e 47#include "env-util.h"
3652872a 48#include "escape.h"
3ffd4af2 49#include "fd-util.h"
842f3b0f 50#include "fdset.h"
a5c32cff 51#include "fileio.h"
f97b34a6 52#include "format-util.h"
f4f15635 53#include "fs-util.h"
1b9e5b12 54#include "gpt.h"
4623e8e6 55#include "hexdecoct.h"
e2054217 56#include "hostname-setup.h"
8fe0087e 57#include "hostname-util.h"
910fd145 58#include "id128-util.h"
3652872a 59#include "io-util.h"
8fe0087e 60#include "log.h"
2d845785 61#include "loop-util.h"
8fe0087e 62#include "loopback-setup.h"
e8ac916e 63#include "machine-credential.h"
8fe0087e 64#include "macro.h"
44dbef90 65#include "main-func.h"
f5947a5e 66#include "missing_sched.h"
8fe0087e 67#include "mkdir.h"
4349cd7c 68#include "mount-util.h"
049af8ad 69#include "mountpoint-util.h"
0cb8e3d1 70#include "namespace-util.h"
8fe0087e 71#include "netlink-util.h"
2f893044 72#include "nspawn-bind-user.h"
07630cea 73#include "nspawn-cgroup.h"
3603efde 74#include "nspawn-def.h"
07630cea
LP		 75#include "nspawn-expose-ports.h"
76#include "nspawn-mount.h"
77#include "nspawn-network.h"
de40a303 78#include "nspawn-oci.h"
7336138e 79#include "nspawn-patch-uid.h"
07630cea 80#include "nspawn-register.h"
910fd145 81#include "nspawn-seccomp.h"
07630cea
LP		 82#include "nspawn-settings.h"
83#include "nspawn-setuid.h"
7732f92b 84#include "nspawn-stub-pid1.h"
c9394f4f 85#include "nspawn-util.h"
91181e07 86#include "nspawn.h"
d8b4d14d 87#include "nulstr-util.h"
d58ad743 88#include "os-util.h"
50ebcf6c 89#include "pager.h"
614b022c 90#include "parse-argument.h"
6bedfcbb 91#include "parse-util.h"
294bf0c3 92#include "pretty-print.h"
0b452006 93#include "process-util.h"
8fe0087e
LP		 94#include "ptyfwd.h"
95#include "random-util.h"
8869a0b4 96#include "raw-clone.h"
86775e35 97#include "resolve-util.h"
bf428efb 98#include "rlimit-util.h"
8fe0087e 99#include "rm-rf.h"
de40a303 100#include "seccomp-util.h"
68b02049 101#include "selinux-util.h"
8fe0087e 102#include "signal-util.h"
2583fbea 103#include "socket-util.h"
8fcde012 104#include "stat-util.h"
15a5e950 105#include "stdio-util.h"
5c828e66 106#include "string-table.h"
07630cea 107#include "string-util.h"
8fe0087e 108#include "strv.h"
de40a303 109#include "sysctl-util.h"
8fe0087e 110#include "terminal-util.h"
e4de7287 111#include "tmpfile-util.h"
affb60b1 112#include "umask-util.h"
43c3fb46 113#include "unit-name.h"
b1d4f8e1 114#include "user-util.h"
300a03be 115#include "vpick.h"
e9642be2 116
e96ceaba
LP		 117/* The notify socket inside the container it can use to talk to nspawn using the sd_notify(3) protocol */
118#define NSPAWN_NOTIFY_SOCKET_PATH "/run/host/notify"
e79581dd 119#define NSPAWN_MOUNT_TUNNEL "/run/host/incoming"
0e7ac751 120
2a49b612
ZJS		 121#define EXIT_FORCE_RESTART 133
122
113cea80
DH		 123typedef enum ContainerStatus {
124 CONTAINER_TERMINATED,
6145bb4f 125 CONTAINER_REBOOTED,
113cea80
DH		 126} ContainerStatus;
127
88213476 128static char *arg_directory = NULL;
ec16945e 129static char *arg_template = NULL;
5f932eb9 130static char *arg_chdir = NULL;
b53ede69
PW		 131static char *arg_pivot_root_new = NULL;
132static char *arg_pivot_root_old = NULL;
687d0825 133static char *arg_user = NULL;
de40a303
LP		 134static uid_t arg_uid = UID_INVALID;
135static gid_t arg_gid = GID_INVALID;
136static gid_t* arg_supplementary_gids = NULL;
137static size_t arg_n_supplementary_gids = 0;
9444b1f2 138static sd_id128_t arg_uuid = {};
3a9530e5
LP		 139static char *arg_machine = NULL; /* The name used by the host to refer to this */
140static char *arg_hostname = NULL; /* The name the payload sees by default */
c74e630d
LP		 141static const char *arg_selinux_context = NULL;
142static const char *arg_selinux_apifs_context = NULL;
de40a303 143static char *arg_slice = NULL;
ff01d048 144static bool arg_private_network = false;
bc2f673e 145static bool arg_read_only = false;
7732f92b 146static StartMode arg_start_mode = START_PID1;
ec16945e 147static bool arg_ephemeral = false;
57fb9fb5 148static LinkJournal arg_link_journal = LINK_AUTO;
574edc90 149static bool arg_link_journal_try = false;
520e0d54 150static uint64_t arg_caps_retain =
50b52222
LP		 151 (1ULL << CAP_AUDIT_CONTROL) |
152 (1ULL << CAP_AUDIT_WRITE) |
5076f0cc
LP		 153 (1ULL << CAP_CHOWN) |
154 (1ULL << CAP_DAC_OVERRIDE) |
155 (1ULL << CAP_DAC_READ_SEARCH) |
156 (1ULL << CAP_FOWNER) |
157 (1ULL << CAP_FSETID) |
158 (1ULL << CAP_IPC_OWNER) |
159 (1ULL << CAP_KILL) |
160 (1ULL << CAP_LEASE) |
161 (1ULL << CAP_LINUX_IMMUTABLE) |
50b52222 162 (1ULL << CAP_MKNOD) |
5076f0cc
LP		 163 (1ULL << CAP_NET_BIND_SERVICE) |
164 (1ULL << CAP_NET_BROADCAST) |
165 (1ULL << CAP_NET_RAW) |
5076f0cc 166 (1ULL << CAP_SETFCAP) |
50b52222 167 (1ULL << CAP_SETGID) |
5076f0cc
LP		 168 (1ULL << CAP_SETPCAP) |
169 (1ULL << CAP_SETUID) |
170 (1ULL << CAP_SYS_ADMIN) |
50b52222 171 (1ULL << CAP_SYS_BOOT) |
5076f0cc
LP		 172 (1ULL << CAP_SYS_CHROOT) |
173 (1ULL << CAP_SYS_NICE) |
174 (1ULL << CAP_SYS_PTRACE) |
d87be9b0 175 (1ULL << CAP_SYS_RESOURCE) |
50b52222 176 (1ULL << CAP_SYS_TTY_CONFIG);
88fc9c9b 177static uint64_t arg_caps_ambient = 0;
de40a303 178static CapabilityQuintet arg_full_capabilities = CAPABILITY_QUINTET_NULL;
5a8af538 179static CustomMount *arg_custom_mounts = NULL;
88614c8a 180static size_t arg_n_custom_mounts = 0;
f4889f65 181static char **arg_setenv = NULL;
284c0b91 182static bool arg_quiet = false;
eb91eb18 183static bool arg_register = true;
89f7c846 184static bool arg_keep_unit = false;
aa28aefe 185static char **arg_network_interfaces = NULL;
c74e630d 186static char **arg_network_macvlan = NULL;
4bbfe7ad 187static char **arg_network_ipvlan = NULL;
69c79d3c 188static bool arg_network_veth = false;
f6d6bad1 189static char **arg_network_veth_extra = NULL;
f757855e 190static char *arg_network_bridge = NULL;
22b28dfd 191static char *arg_network_zone = NULL;
d7bea6b6 192static char *arg_network_namespace_path = NULL;
813dbff4 193struct ether_addr arg_network_provided_mac = {};
bb068de0 194static PagerFlags arg_pager_flags = 0;
050f7277 195static unsigned long arg_personality = PERSONALITY_INVALID;
ec16945e 196static char *arg_image = NULL;
de40a303 197static char *arg_oci_bundle = NULL;
f757855e 198static VolatileMode arg_volatile_mode = VOLATILE_NO;
6d0b55c2 199static ExposePort *arg_expose_ports = NULL;
f36933fe 200static char **arg_property = NULL;
de40a303 201static sd_bus_message *arg_property_message = NULL;
0de7acce 202static UserNamespaceMode arg_userns_mode = USER_NAMESPACE_NO;
6dac160c 203static uid_t arg_uid_shift = UID_INVALID, arg_uid_range = 0x10000U;
6c045a99 204static UserNamespaceOwnership arg_userns_ownership = _USER_NAMESPACE_OWNERSHIP_INVALID;
c6c8f6e2 205static int arg_kill_signal = 0;
5da38d07 206static CGroupUnified arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_UNKNOWN;
f757855e
LP		 207static SettingsMask arg_settings_mask = 0;
208static int arg_settings_trusted = -1;
209static char **arg_parameters = NULL;
6aadfa4c 210static const char *arg_container_service_name = "systemd-nspawn";
9c1e04d0 211static bool arg_notify_ready = false;
5a8ff0e6 212static bool arg_use_cgns = true;
0c582db0 213static unsigned long arg_clone_ns_flags = CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS;
1099ceeb 214static MountSettingsMask arg_mount_settings = MOUNT_APPLY_APIVFS_RO|MOUNT_APPLY_TMPFS_TMP;
aee36b4e 215static VeritySettings arg_verity_settings = VERITY_SETTINGS_DEFAULT;
6b000af4
LP		 216static char **arg_syscall_allow_list = NULL;
217static char **arg_syscall_deny_list = NULL;
de40a303
LP		 218#if HAVE_SECCOMP
219static scmp_filter_ctx arg_seccomp = NULL;
220#endif
bf428efb 221static struct rlimit *arg_rlimit[_RLIMIT_MAX] = {};
66edd963 222static bool arg_no_new_privileges = false;
81f345df
LP		 223static int arg_oom_score_adjust = 0;
224static bool arg_oom_score_adjust_set = false;
0985c7c4 225static CPUSet arg_cpu_set = {};
09d423e9 226static ResolvConfMode arg_resolv_conf = RESOLV_CONF_AUTO;
1688841f 227static TimezoneMode arg_timezone = TIMEZONE_AUTO;
f5fbe71d 228static unsigned arg_console_width = UINT_MAX, arg_console_height = UINT_MAX;
de40a303
LP		 229static DeviceNode* arg_extra_nodes = NULL;
230static size_t arg_n_extra_nodes = 0;
231static char **arg_sysctl = NULL;
232static ConsoleMode arg_console_mode = _CONSOLE_MODE_INVALID;
bd546b9b 233static MachineCredentialContext arg_credentials = {};
2f893044 234static char **arg_bind_user = NULL;
4a4654e0 235static bool arg_suppress_sync = false;
3603f151 236static char *arg_settings_filename = NULL;
4c27749b 237static Architecture arg_architecture = _ARCHITECTURE_INVALID;
84be0c71 238static ImagePolicy *arg_image_policy = NULL;
88213476 239
6145bb4f
LP		 240STATIC_DESTRUCTOR_REGISTER(arg_directory, freep);
241STATIC_DESTRUCTOR_REGISTER(arg_template, freep);
242STATIC_DESTRUCTOR_REGISTER(arg_chdir, freep);
243STATIC_DESTRUCTOR_REGISTER(arg_pivot_root_new, freep);
244STATIC_DESTRUCTOR_REGISTER(arg_pivot_root_old, freep);
245STATIC_DESTRUCTOR_REGISTER(arg_user, freep);
246STATIC_DESTRUCTOR_REGISTER(arg_supplementary_gids, freep);
247STATIC_DESTRUCTOR_REGISTER(arg_machine, freep);
248STATIC_DESTRUCTOR_REGISTER(arg_hostname, freep);
249STATIC_DESTRUCTOR_REGISTER(arg_slice, freep);
250STATIC_DESTRUCTOR_REGISTER(arg_setenv, strv_freep);
251STATIC_DESTRUCTOR_REGISTER(arg_network_interfaces, strv_freep);
252STATIC_DESTRUCTOR_REGISTER(arg_network_macvlan, strv_freep);
253STATIC_DESTRUCTOR_REGISTER(arg_network_ipvlan, strv_freep);
254STATIC_DESTRUCTOR_REGISTER(arg_network_veth_extra, strv_freep);
255STATIC_DESTRUCTOR_REGISTER(arg_network_bridge, freep);
256STATIC_DESTRUCTOR_REGISTER(arg_network_zone, freep);
257STATIC_DESTRUCTOR_REGISTER(arg_network_namespace_path, freep);
258STATIC_DESTRUCTOR_REGISTER(arg_image, freep);
259STATIC_DESTRUCTOR_REGISTER(arg_oci_bundle, freep);
260STATIC_DESTRUCTOR_REGISTER(arg_property, strv_freep);
261STATIC_DESTRUCTOR_REGISTER(arg_property_message, sd_bus_message_unrefp);
262STATIC_DESTRUCTOR_REGISTER(arg_parameters, strv_freep);
89e62e0b 263STATIC_DESTRUCTOR_REGISTER(arg_verity_settings, verity_settings_done);
6b000af4
LP		 264STATIC_DESTRUCTOR_REGISTER(arg_syscall_allow_list, strv_freep);
265STATIC_DESTRUCTOR_REGISTER(arg_syscall_deny_list, strv_freep);
6145bb4f
LP		 266#if HAVE_SECCOMP
267STATIC_DESTRUCTOR_REGISTER(arg_seccomp, seccomp_releasep);
268#endif
bd546b9b 269STATIC_DESTRUCTOR_REGISTER(arg_credentials, machine_credential_context_done);
0985c7c4 270STATIC_DESTRUCTOR_REGISTER(arg_cpu_set, cpu_set_reset);
6145bb4f 271STATIC_DESTRUCTOR_REGISTER(arg_sysctl, strv_freep);
2f893044 272STATIC_DESTRUCTOR_REGISTER(arg_bind_user, strv_freep);
3603f151 273STATIC_DESTRUCTOR_REGISTER(arg_settings_filename, freep);
84be0c71 274STATIC_DESTRUCTOR_REGISTER(arg_image_policy, image_policy_freep);
6145bb4f 275
dce66ffe
ZJS		 276static int handle_arg_console(const char *arg) {
277 if (streq(arg, "help")) {
10e8a60b
LP		 278 puts("autopipe\n"
279 "interactive\n"
dce66ffe 280 "passive\n"
10e8a60b
LP		 281 "pipe\n"
282 "read-only");
dce66ffe
ZJS		 283 return 0;
284 }
285
286 if (streq(arg, "interactive"))
287 arg_console_mode = CONSOLE_INTERACTIVE;
288 else if (streq(arg, "read-only"))
289 arg_console_mode = CONSOLE_READ_ONLY;
290 else if (streq(arg, "passive"))
291 arg_console_mode = CONSOLE_PASSIVE;
554c4beb 292 else if (streq(arg, "pipe")) {
dd9c8da8 293 if (isatty(STDIN_FILENO) && isatty(STDOUT_FILENO))
554c4beb
LP		 294 log_full(arg_quiet ? LOG_DEBUG : LOG_NOTICE,
295 "Console mode 'pipe' selected, but standard input/output are connected to an interactive TTY. "
296 "Most likely you want to use 'interactive' console mode for proper interactivity and shell job control. "
297 "Proceeding anyway.");
298
dce66ffe 299 arg_console_mode = CONSOLE_PIPE;
10e8a60b 300 } else if (streq(arg, "autopipe")) {
dd9c8da8 301 if (isatty(STDIN_FILENO) && isatty(STDOUT_FILENO))
10e8a60b
LP		 302 arg_console_mode = CONSOLE_INTERACTIVE;
303 else
304 arg_console_mode = CONSOLE_PIPE;
554c4beb 305 } else
dce66ffe
ZJS		 306 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Unknown console mode: %s", optarg);
307
308 arg_settings_mask |= SETTING_CONSOLE_MODE;
309 return 1;
310}
311
37ec0fdd
LP		 312static int help(void) {
313 _cleanup_free_ char *link = NULL;
314 int r;
315
384c2c32 316 pager_open(arg_pager_flags);
50ebcf6c 317
37ec0fdd
LP		 318 r = terminal_urlify_man("systemd-nspawn", "1", &link);
319 if (r < 0)
320 return log_oom();
321
25148653 322 printf("%1$s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
37a92352 323 "%5$sSpawn a command or OS in a light-weight container.%6$s\n\n"
a8828ed9
DW		 324 " -h --help Show this help\n"
325 " --version Print version string\n"
69c79d3c 326 " -q --quiet Do not show status information\n"
bb068de0 327 " --no-pager Do not pipe output into a pager\n"
25148653
LP		 328 " --settings=BOOLEAN Load additional settings from .nspawn file\n\n"
329 "%3$sImage:%4$s\n"
1b9e5b12 330 " -D --directory=PATH Root directory for the container\n"
ec16945e
LP		 331 " --template=PATH Initialize root directory from template directory,\n"
332 " if missing\n"
333 " -x --ephemeral Run container with snapshot of root directory, and\n"
334 " remove it after exit\n"
25e68fd3
LP		 335 " -i --image=PATH Root file system disk image (or device node) for\n"
336 " the container\n"
84be0c71 337 " --image-policy=POLICY Specify disk image dissection policy\n"
de40a303 338 " --oci-bundle=PATH OCI bundle directory\n"
25148653
LP		 339 " --read-only Mount the root directory read-only\n"
340 " --volatile[=MODE] Run the system in volatile mode\n"
25e68fd3 341 " --root-hash=HASH Specify verity root hash for root disk image\n"
c2923fdc
LB		 342 " --root-hash-sig=SIG Specify pkcs7 signature of root hash for verity\n"
343 " as a DER encoded PKCS7, either as a path to a file\n"
344 " or as an ASCII base64 encoded string prefixed by\n"
345 " 'base64:'\n"
e7cbe5cb 346 " --verity-data=PATH Specify hash device for verity\n"
25148653
LP		 347 " --pivot-root=PATH[:PATH]\n"
348 " Pivot root to given directory in the container\n\n"
349 "%3$sExecution:%4$s\n"
7732f92b 350 " -a --as-pid2 Maintain a stub init as PID1, invoke binary as PID2\n"
a8828ed9 351 " -b --boot Boot up full system (i.e. invoke init)\n"
5f932eb9 352 " --chdir=PATH Set working directory in the container\n"
0d2a0179 353 " -E --setenv=NAME[=VALUE] Pass an environment variable to PID 1\n"
25148653
LP		 354 " -u --user=USER Run the command under specified user or UID\n"
355 " --kill-signal=SIGNAL Select signal to use for shutting down PID 1\n"
4a4654e0
LP		 356 " --notify-ready=BOOLEAN Receive notifications from the child init process\n"
357 " --suppress-sync=BOOLEAN\n"
358 " Suppress any form of disk data synchronization\n\n"
25148653 359 "%3$sSystem Identity:%4$s\n"
a8828ed9 360 " -M --machine=NAME Set the machine name for the container\n"
3a9530e5 361 " --hostname=NAME Override the hostname for the container\n"
25148653
LP		 362 " --uuid=UUID Set a specific machine UUID for the container\n\n"
363 "%3$sProperties:%4$s\n"
a8828ed9 364 " -S --slice=SLICE Place the container in the specified slice\n"
f36933fe 365 " --property=NAME=VALUE Set scope unit property\n"
25148653
LP		 366 " --register=BOOLEAN Register container as machine\n"
367 " --keep-unit Do not register a scope for the machine, reuse\n"
368 " the service unit nspawn is running in\n\n"
369 "%3$sUser Namespacing:%4$s\n"
b917743d
YW		 370 " --private-users=no Run without user namespacing\n"
371 " --private-users=yes|pick|identity\n"
372 " Run within user namespace, autoselect UID/GID range\n"
373 " --private-users=UIDBASE[:NUIDS]\n"
90b4a64d 374 " Similar, but with user configured UID/GID range\n"
6c045a99
LP		 375 " --private-users-ownership=MODE\n"
376 " Adjust ('chown') or map ('map') OS tree ownership\n"
b917743d
YW		 377 " to private UID/GID range\n"
378 " -U Equivalent to --private-users=pick and\n"
379 " --private-users-ownership=auto\n\n"
25148653 380 "%3$sNetworking:%4$s\n"
69c79d3c 381 " --private-network Disable network in container\n"
2f091b1b 382 " --network-interface=HOSTIF[:CONTAINERIF]\n"
69c79d3c
LP		 383 " Assign an existing network interface to the\n"
384 " container\n"
2f091b1b 385 " --network-macvlan=HOSTIF[:CONTAINERIF]\n"
c74e630d
LP		 386 " Create a macvlan network interface based on an\n"
387 " existing network interface to the container\n"
2f091b1b 388 " --network-ipvlan=HOSTIF[:CONTAINERIF]\n"
387f6955 389 " Create an ipvlan network interface based on an\n"
4bbfe7ad 390 " existing network interface to the container\n"
a8eaaee7 391 " -n --network-veth Add a virtual Ethernet connection between host\n"
69c79d3c 392 " and container\n"
f6d6bad1
LP		 393 " --network-veth-extra=HOSTIF[:CONTAINERIF]\n"
394 " Add an additional virtual Ethernet link between\n"
395 " host and container\n"
ab046dde 396 " --network-bridge=INTERFACE\n"
90b4a64d
ZJS		 397 " Add a virtual Ethernet connection to the container\n"
398 " and attach it to an existing bridge on the host\n"
399 " --network-zone=NAME Similar, but attach the new interface to an\n"
400 " an automatically managed bridge interface\n"
d7bea6b6
DP		 401 " --network-namespace-path=PATH\n"
402 " Set network namespace to the one represented by\n"
403 " the specified kernel namespace file node\n"
6d0b55c2 404 " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n"
25148653
LP		 405 " Expose a container IP port on the host\n\n"
406 "%3$sSecurity:%4$s\n"
a8828ed9
DW		 407 " --capability=CAP In addition to the default, retain specified\n"
408 " capability\n"
409 " --drop-capability=CAP Drop the specified capability from the default set\n"
88fc9c9b
TH		 410 " --ambient-capability=CAP\n"
411 " Sets the specified capability for the started\n"
412 " process. Not useful if booting a machine.\n"
f4e803c8 413 " --no-new-privileges Set PR_SET_NO_NEW_PRIVS flag for container payload\n"
960e4569
LP		 414 " --system-call-filter=LIST|~LIST\n"
415 " Permit/prohibit specific system calls\n"
25148653
LP		 416 " -Z --selinux-context=SECLABEL\n"
417 " Set the SELinux security context to be used by\n"
418 " processes in the container\n"
419 " -L --selinux-apifs-context=SECLABEL\n"
420 " Set the SELinux security context to be used by\n"
421 " API/tmpfs file systems in the container\n\n"
422 "%3$sResources:%4$s\n"
bf428efb 423 " --rlimit=NAME=LIMIT Set a resource limit for the payload\n"
81f345df
LP		 424 " --oom-score-adjust=VALUE\n"
425 " Adjust the OOM score value for the payload\n"
f4e803c8
LP		 426 " --cpu-affinity=CPUS Adjust the CPU affinity of the container\n"
427 " --personality=ARCH Pick personality for this container\n\n"
25148653 428 "%3$sIntegration:%4$s\n"
09d423e9 429 " --resolv-conf=MODE Select mode of /etc/resolv.conf initialization\n"
1688841f 430 " --timezone=MODE Select mode of /etc/localtime initialization\n"
25148653
LP		 431 " --link-journal=MODE Link up guest journal, one of no, auto, guest, \n"
432 " host, try-guest, try-host\n"
433 " -j Equivalent to --link-journal=try-guest\n\n"
434 "%3$sMounts:%4$s\n"
5e5bfa6e
EY		 435 " --bind=PATH[:PATH[:OPTIONS]]\n"
436 " Bind mount a file or directory from the host into\n"
a8828ed9 437 " the container\n"
5e5bfa6e
EY		 438 " --bind-ro=PATH[:PATH[:OPTIONS]\n"
439 " Similar, but creates a read-only bind mount\n"
de40a303
LP		 440 " --inaccessible=PATH Over-mount file node with inaccessible node to mask\n"
441 " it\n"
06c17c39 442 " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n"
5a8af538
LP		 443 " --overlay=PATH[:PATH...]:PATH\n"
444 " Create an overlay mount from the host to \n"
445 " the container\n"
446 " --overlay-ro=PATH[:PATH...]:PATH\n"
2f893044
LP		 447 " Similar, but creates a read-only overlay mount\n"
448 " --bind-user=NAME Bind user from host to container\n\n"
25148653 449 "%3$sInput/Output:%4$s\n"
de40a303
LP		 450 " --console=MODE Select how stdin/stdout/stderr and /dev/console are\n"
451 " set up for the container.\n"
3652872a
LP		 452 " -P --pipe Equivalent to --console=pipe\n\n"
453 "%3$sCredentials:%4$s\n"
454 " --set-credential=ID:VALUE\n"
455 " Pass a credential with literal value to container.\n"
456 " --load-credential=ID:PATH\n"
457 " Load credential to pass to container from file or\n"
458 " AF_UNIX stream socket.\n"
bc556335
DDM		 459 "\nSee the %2$s for details.\n",
460 program_invocation_short_name,
461 link,
462 ansi_underline(),
463 ansi_normal(),
464 ansi_highlight(),
465 ansi_normal());
37ec0fdd
LP		 466
467 return 0;
88213476
LP		 468}
469
86c0dd4a 470static int custom_mount_check_all(void) {
88614c8a 471 size_t i;
5a8af538 472
5a8af538
LP		 473 for (i = 0; i < arg_n_custom_mounts; i++) {
474 CustomMount *m = &arg_custom_mounts[i];
475
0de7acce 476 if (path_equal(m->destination, "/") && arg_userns_mode != USER_NAMESPACE_NO) {
6c045a99 477 if (arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_OFF)
baaa35ad 478 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
c920b863 479 "--private-users-ownership=own may not be combined with custom root mounts.");
6c045a99 480 if (arg_uid_shift == UID_INVALID)
baaa35ad
ZJS		 481 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
482 "--private-users with automatic UID shift may not be combined with custom root mounts.");
825d5287 483 }
5a8af538
LP		 484 }
485
486 return 0;
487}
488
8199d554 489static int detect_unified_cgroup_hierarchy_from_environment(void) {
c78c095b 490 const char *e, *var = "SYSTEMD_NSPAWN_UNIFIED_HIERARCHY";
415fc41c 491 int r;
5da38d07 492
efdb0237 493 /* Allow the user to control whether the unified hierarchy is used */
c78c095b
ZJS		 494
495 e = getenv(var);
496 if (!e) {
d5fc5b2f 497 /* $UNIFIED_CGROUP_HIERARCHY has been renamed to $SYSTEMD_NSPAWN_UNIFIED_HIERARCHY. */
c78c095b
ZJS		 498 var = "UNIFIED_CGROUP_HIERARCHY";
499 e = getenv(var);
c78c095b
ZJS		 500 }
501
502 if (!isempty(e)) {
efdb0237
LP		 503 r = parse_boolean(e);
504 if (r < 0)
c78c095b 505 return log_error_errno(r, "Failed to parse $%s: %m", var);
5da38d07
TH		 506 if (r > 0)
507 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
508 else
509 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237
LP		 510 }
511
8199d554
LP		 512 return 0;
513}
514
515static int detect_unified_cgroup_hierarchy_from_image(const char *directory) {
516 int r;
517
75b0d8b8
ZJS		 518 /* Let's inherit the mode to use from the host system, but let's take into consideration what systemd
519 * in the image actually supports. */
b4cccbc1
LP		 520 r = cg_all_unified();
521 if (r < 0)
522 return log_error_errno(r, "Failed to determine whether we are in all unified mode.");
523 if (r > 0) {
a8725a06
ZJS		 524 /* Unified cgroup hierarchy support was added in 230. Unfortunately the detection
525 * routine only detects 231, so we'll have a false negative here for 230. */
7e6821ed 526 r = systemd_installation_has_version(directory, "230");
a8725a06
ZJS		 527 if (r < 0)
528 return log_error_errno(r, "Failed to determine systemd version in container: %m");
529 if (r > 0)
530 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
531 else
532 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
c22800e4 533 } else if (cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER) > 0) {
2977724b 534 /* Mixed cgroup hierarchy support was added in 233 */
7e6821ed 535 r = systemd_installation_has_version(directory, "233");
0fd9563f
ZJS		 536 if (r < 0)
537 return log_error_errno(r, "Failed to determine systemd version in container: %m");
538 if (r > 0)
539 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD;
540 else
541 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
542 } else
5da38d07 543 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
efdb0237 544
8199d554
LP		 545 log_debug("Using %s hierarchy for container.",
546 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_NONE ? "legacy" :
547 arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_SYSTEMD ? "hybrid" : "unified");
548
efdb0237
LP		 549 return 0;
550}
551
8a99bd0c
ZJS		 552static int parse_capability_spec(const char *spec, uint64_t *ret_mask) {
553 uint64_t mask = 0;
554 int r;
555
556 for (;;) {
557 _cleanup_free_ char *t = NULL;
558
559 r = extract_first_word(&spec, &t, ",", 0);
560 if (r < 0)
561 return log_error_errno(r, "Failed to parse capability %s.", t);
562 if (r == 0)
563 break;
564
565 if (streq(t, "help")) {
566 for (int i = 0; i < capability_list_length(); i++) {
567 const char *name;
568
569 name = capability_to_name(i);
570 if (name)
571 puts(name);
572 }
573
574 return 0; /* quit */
575 }
576
577 if (streq(t, "all"))
f5fbe71d 578 mask = UINT64_MAX;
8a99bd0c
ZJS		 579 else {
580 r = capability_from_name(t);
581 if (r < 0)
582 return log_error_errno(r, "Failed to parse capability %s.", t);
583
584 mask |= 1ULL << r;
585 }
586 }
587
588 *ret_mask = mask;
589 return 1; /* continue */
590}
591
49048684 592static int parse_share_ns_env(const char *name, unsigned long ns_flag) {
0c582db0
LB		 593 int r;
594
595 r = getenv_bool(name);
596 if (r == -ENXIO)
49048684 597 return 0;
0c582db0 598 if (r < 0)
49048684 599 return log_error_errno(r, "Failed to parse $%s: %m", name);
de40a303 600
0c582db0 601 arg_clone_ns_flags = (arg_clone_ns_flags & ~ns_flag) | (r > 0 ? 0 : ns_flag);
de40a303 602 arg_settings_mask |= SETTING_CLONE_NS_FLAGS;
49048684 603 return 0;
0c582db0
LB		 604}
605
49048684 606static int parse_mount_settings_env(void) {
4f086aab 607 const char *e;
1099ceeb
LP		 608 int r;
609
610 r = getenv_bool("SYSTEMD_NSPAWN_TMPFS_TMP");
49048684
ZJS		 611 if (r < 0 && r != -ENXIO)
612 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_TMPFS_TMP: %m");
1099ceeb
LP		 613 if (r >= 0)
614 SET_FLAG(arg_mount_settings, MOUNT_APPLY_TMPFS_TMP, r > 0);
4f086aab
SU		 615
616 e = getenv("SYSTEMD_NSPAWN_API_VFS_WRITABLE");
49048684 617 if (streq_ptr(e, "network"))
4f086aab 618 arg_mount_settings |= MOUNT_APPLY_APIVFS_RO|MOUNT_APPLY_APIVFS_NETNS;
4f086aab 619
49048684
ZJS		 620 else if (e) {
621 r = parse_boolean(e);
622 if (r < 0)
623 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_API_VFS_WRITABLE: %m");
624
625 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_RO, r == 0);
626 SET_FLAG(arg_mount_settings, MOUNT_APPLY_APIVFS_NETNS, false);
ab8ee0f2 627 }
4f086aab 628
49048684 629 return 0;
4f086aab
SU		 630}
631
49048684 632static int parse_environment(void) {
d5455d2f
LP		 633 const char *e;
634 int r;
635
49048684
ZJS		 636 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_IPC", CLONE_NEWIPC);
637 if (r < 0)
638 return r;
639 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_PID", CLONE_NEWPID);
640 if (r < 0)
641 return r;
642 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_NS_UTS", CLONE_NEWUTS);
643 if (r < 0)
644 return r;
645 r = parse_share_ns_env("SYSTEMD_NSPAWN_SHARE_SYSTEM", CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS);
646 if (r < 0)
647 return r;
d5455d2f 648
49048684
ZJS		 649 r = parse_mount_settings_env();
650 if (r < 0)
651 return r;
d5455d2f 652
489fae52
ZJS		 653 /* SYSTEMD_NSPAWN_USE_CGNS=0 can be used to disable CLONE_NEWCGROUP use,
654 * even if it is supported. If not supported, it has no effect. */
de40a303 655 if (!cg_ns_supported())
489fae52 656 arg_use_cgns = false;
de40a303
LP		 657 else {
658 r = getenv_bool("SYSTEMD_NSPAWN_USE_CGNS");
659 if (r < 0) {
660 if (r != -ENXIO)
49048684 661 return log_error_errno(r, "Failed to parse $SYSTEMD_NSPAWN_USE_CGNS: %m");
de40a303
LP		 662
663 arg_use_cgns = true;
664 } else {
665 arg_use_cgns = r > 0;
666 arg_settings_mask |= SETTING_USE_CGNS;
667 }
668 }
d5455d2f
LP		 669
670 e = getenv("SYSTEMD_NSPAWN_CONTAINER_SERVICE");
671 if (e)
672 arg_container_service_name = e;
673
813dbff4
RC		 674 e = getenv("SYSTEMD_NSPAWN_NETWORK_MAC");
675 if (e) {
676 r = parse_ether_addr(e, &arg_network_provided_mac);
677 if (r < 0)
678 return log_error_errno(r, "Failed to parse provided MAC address via environment variable");
679 }
680
4a4654e0
LP		 681 r = getenv_bool("SYSTEMD_SUPPRESS_SYNC");
682 if (r >= 0)
683 arg_suppress_sync = r;
684 else if (r != -ENXIO)
685 log_debug_errno(r, "Failed to parse $SYSTEMD_SUPPRESS_SYNC, ignoring: %m");
686
49048684 687 return detect_unified_cgroup_hierarchy_from_environment();
d5455d2f
LP		 688}
689
88213476 690static int parse_argv(int argc, char *argv[]) {
a41fe3a2 691 enum {
acbeb427
ZJS		 692 ARG_VERSION = 0x100,
693 ARG_PRIVATE_NETWORK,
bc2f673e 694 ARG_UUID,
5076f0cc 695 ARG_READ_ONLY,
57fb9fb5 696 ARG_CAPABILITY,
88fc9c9b 697 ARG_AMBIENT_CAPABILITY,
420c7379 698 ARG_DROP_CAPABILITY,
17fe0523
LP		 699 ARG_LINK_JOURNAL,
700 ARG_BIND,
f4889f65 701 ARG_BIND_RO,
06c17c39 702 ARG_TMPFS,
5a8af538
LP		 703 ARG_OVERLAY,
704 ARG_OVERLAY_RO,
de40a303 705 ARG_INACCESSIBLE,
eb91eb18 706 ARG_SHARE_SYSTEM,
89f7c846 707 ARG_REGISTER,
aa28aefe 708 ARG_KEEP_UNIT,
69c79d3c 709 ARG_NETWORK_INTERFACE,
c74e630d 710 ARG_NETWORK_MACVLAN,
4bbfe7ad 711 ARG_NETWORK_IPVLAN,
ab046dde 712 ARG_NETWORK_BRIDGE,
22b28dfd 713 ARG_NETWORK_ZONE,
f6d6bad1 714 ARG_NETWORK_VETH_EXTRA,
d7bea6b6 715 ARG_NETWORK_NAMESPACE_PATH,
6afc95b7 716 ARG_PERSONALITY,
4d9f07b4 717 ARG_VOLATILE,
ec16945e 718 ARG_TEMPLATE,
f36933fe 719 ARG_PROPERTY,
6dac160c 720 ARG_PRIVATE_USERS,
c6c8f6e2 721 ARG_KILL_SIGNAL,
f757855e 722 ARG_SETTINGS,
5f932eb9 723 ARG_CHDIR,
b53ede69 724 ARG_PIVOT_ROOT,
7336138e 725 ARG_PRIVATE_USERS_CHOWN,
6c045a99 726 ARG_PRIVATE_USERS_OWNERSHIP,
9c1e04d0 727 ARG_NOTIFY_READY,
4623e8e6 728 ARG_ROOT_HASH,
89e62e0b
LP		 729 ARG_ROOT_HASH_SIG,
730 ARG_VERITY_DATA,
960e4569 731 ARG_SYSTEM_CALL_FILTER,
bf428efb 732 ARG_RLIMIT,
3a9530e5 733 ARG_HOSTNAME,
66edd963 734 ARG_NO_NEW_PRIVILEGES,
81f345df 735 ARG_OOM_SCORE_ADJUST,
d107bb7d 736 ARG_CPU_AFFINITY,
09d423e9 737 ARG_RESOLV_CONF,
1688841f 738 ARG_TIMEZONE,
de40a303
LP		 739 ARG_CONSOLE,
740 ARG_PIPE,
741 ARG_OCI_BUNDLE,
bb068de0 742 ARG_NO_PAGER,
3652872a
LP		 743 ARG_SET_CREDENTIAL,
744 ARG_LOAD_CREDENTIAL,
2f893044 745 ARG_BIND_USER,
4a4654e0 746 ARG_SUPPRESS_SYNC,
84be0c71 747 ARG_IMAGE_POLICY,
a41fe3a2
LP		 748 };
749
88213476 750 static const struct option options[] = {
d7bea6b6
DP		 751 { "help", no_argument, NULL, 'h' },
752 { "version", no_argument, NULL, ARG_VERSION },
753 { "directory", required_argument, NULL, 'D' },
754 { "template", required_argument, NULL, ARG_TEMPLATE },
755 { "ephemeral", no_argument, NULL, 'x' },
756 { "user", required_argument, NULL, 'u' },
757 { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
758 { "as-pid2", no_argument, NULL, 'a' },
759 { "boot", no_argument, NULL, 'b' },
760 { "uuid", required_argument, NULL, ARG_UUID },
761 { "read-only", no_argument, NULL, ARG_READ_ONLY },
762 { "capability", required_argument, NULL, ARG_CAPABILITY },
88fc9c9b 763 { "ambient-capability", required_argument, NULL, ARG_AMBIENT_CAPABILITY },
d7bea6b6 764 { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY },
66edd963 765 { "no-new-privileges", required_argument, NULL, ARG_NO_NEW_PRIVILEGES },
d7bea6b6
DP		 766 { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL },
767 { "bind", required_argument, NULL, ARG_BIND },
768 { "bind-ro", required_argument, NULL, ARG_BIND_RO },
769 { "tmpfs", required_argument, NULL, ARG_TMPFS },
770 { "overlay", required_argument, NULL, ARG_OVERLAY },
771 { "overlay-ro", required_argument, NULL, ARG_OVERLAY_RO },
de40a303 772 { "inaccessible", required_argument, NULL, ARG_INACCESSIBLE },
d7bea6b6 773 { "machine", required_argument, NULL, 'M' },
3a9530e5 774 { "hostname", required_argument, NULL, ARG_HOSTNAME },
d7bea6b6
DP		 775 { "slice", required_argument, NULL, 'S' },
776 { "setenv", required_argument, NULL, 'E' },
777 { "selinux-context", required_argument, NULL, 'Z' },
778 { "selinux-apifs-context", required_argument, NULL, 'L' },
779 { "quiet", no_argument, NULL, 'q' },
780 { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, /* not documented */
781 { "register", required_argument, NULL, ARG_REGISTER },
782 { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT },
783 { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE },
784 { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN },
785 { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN },
786 { "network-veth", no_argument, NULL, 'n' },
787 { "network-veth-extra", required_argument, NULL, ARG_NETWORK_VETH_EXTRA },
788 { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE },
789 { "network-zone", required_argument, NULL, ARG_NETWORK_ZONE },
790 { "network-namespace-path", required_argument, NULL, ARG_NETWORK_NAMESPACE_PATH },
791 { "personality", required_argument, NULL, ARG_PERSONALITY },
792 { "image", required_argument, NULL, 'i' },
793 { "volatile", optional_argument, NULL, ARG_VOLATILE },
794 { "port", required_argument, NULL, 'p' },
795 { "property", required_argument, NULL, ARG_PROPERTY },
796 { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS },
6c045a99
LP		 797 { "private-users-chown", optional_argument, NULL, ARG_PRIVATE_USERS_CHOWN }, /* obsolete */
798 { "private-users-ownership",required_argument, NULL, ARG_PRIVATE_USERS_OWNERSHIP},
d7bea6b6
DP		 799 { "kill-signal", required_argument, NULL, ARG_KILL_SIGNAL },
800 { "settings", required_argument, NULL, ARG_SETTINGS },
801 { "chdir", required_argument, NULL, ARG_CHDIR },
802 { "pivot-root", required_argument, NULL, ARG_PIVOT_ROOT },
803 { "notify-ready", required_argument, NULL, ARG_NOTIFY_READY },
804 { "root-hash", required_argument, NULL, ARG_ROOT_HASH },
89e62e0b
LP		 805 { "root-hash-sig", required_argument, NULL, ARG_ROOT_HASH_SIG },
806 { "verity-data", required_argument, NULL, ARG_VERITY_DATA },
d7bea6b6 807 { "system-call-filter", required_argument, NULL, ARG_SYSTEM_CALL_FILTER },
bf428efb 808 { "rlimit", required_argument, NULL, ARG_RLIMIT },
81f345df 809 { "oom-score-adjust", required_argument, NULL, ARG_OOM_SCORE_ADJUST },
d107bb7d 810 { "cpu-affinity", required_argument, NULL, ARG_CPU_AFFINITY },
09d423e9 811 { "resolv-conf", required_argument, NULL, ARG_RESOLV_CONF },
1688841f 812 { "timezone", required_argument, NULL, ARG_TIMEZONE },
de40a303
LP		 813 { "console", required_argument, NULL, ARG_CONSOLE },
814 { "pipe", no_argument, NULL, ARG_PIPE },
815 { "oci-bundle", required_argument, NULL, ARG_OCI_BUNDLE },
bb068de0 816 { "no-pager", no_argument, NULL, ARG_NO_PAGER },
3652872a
LP		 817 { "set-credential", required_argument, NULL, ARG_SET_CREDENTIAL },
818 { "load-credential", required_argument, NULL, ARG_LOAD_CREDENTIAL },
2f893044 819 { "bind-user", required_argument, NULL, ARG_BIND_USER },
4a4654e0 820 { "suppress-sync", required_argument, NULL, ARG_SUPPRESS_SYNC },
84be0c71 821 { "image-policy", required_argument, NULL, ARG_IMAGE_POLICY },
eb9da376 822 {}
88213476
LP		 823 };
824
9444b1f2 825 int c, r;
a42c8b54 826 uint64_t plus = 0, minus = 0;
f757855e 827 bool mask_all_settings = false, mask_no_settings = false;
88213476
LP		 828
829 assert(argc >= 0);
830 assert(argv);
831
ef9c12b1
YW		 832 /* Resetting to 0 forces the invocation of an internal initialization routine of getopt_long()
833 * that checks for GNU extensions in optstring ('-' or '+' at the beginning). */
834 optind = 0;
de40a303 835 while ((c = getopt_long(argc, argv, "+hD:u:abL:M:jS:Z:qi:xp:nUE:P", options, NULL)) >= 0)
88213476
LP		 836 switch (c) {
837
838 case 'h':
37ec0fdd 839 return help();
88213476 840
acbeb427 841 case ARG_VERSION:
3f6fd1ba 842 return version();
acbeb427 843
88213476 844 case 'D':
614b022c 845 r = parse_path_argument(optarg, false, &arg_directory);
ec16945e 846 if (r < 0)
0f03c2a4 847 return r;
de40a303
LP		 848
849 arg_settings_mask |= SETTING_DIRECTORY;
ec16945e
LP		 850 break;
851
852 case ARG_TEMPLATE:
614b022c 853 r = parse_path_argument(optarg, false, &arg_template);
ec16945e 854 if (r < 0)
0f03c2a4 855 return r;
de40a303
LP		 856
857 arg_settings_mask |= SETTING_DIRECTORY;
88213476
LP		 858 break;
859
1b9e5b12 860 case 'i':
614b022c 861 r = parse_path_argument(optarg, false, &arg_image);
ec16945e 862 if (r < 0)
0f03c2a4 863 return r;
de40a303
LP		 864
865 arg_settings_mask |= SETTING_DIRECTORY;
866 break;
867
868 case ARG_OCI_BUNDLE:
614b022c 869 r = parse_path_argument(optarg, false, &arg_oci_bundle);
de40a303
LP		 870 if (r < 0)
871 return r;
872
ec16945e
LP		 873 break;
874
875 case 'x':
876 arg_ephemeral = true;
a2f577fc 877 arg_settings_mask |= SETTING_EPHEMERAL;
1b9e5b12
LP		 878 break;
879
687d0825 880 case 'u':
2fc09a9c
DM		 881 r = free_and_strdup(&arg_user, optarg);
882 if (r < 0)
7027ff61 883 return log_oom();
687d0825 884
f757855e 885 arg_settings_mask |= SETTING_USER;
687d0825
MV		 886 break;
887
22b28dfd 888 case ARG_NETWORK_ZONE: {
fee9f7b5 889 _cleanup_free_ char *j = NULL;
22b28dfd 890
b910cc72 891 j = strjoin("vz-", optarg);
22b28dfd
LP		 892 if (!j)
893 return log_oom();
894
fee9f7b5
FS		 895 if (!ifname_valid(j))
896 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
897 "Network zone name not valid: %s", j);
22b28dfd 898
df1fac6d 899 free_and_replace(arg_network_zone, j);
22b28dfd
LP		 900
901 arg_network_veth = true;
902 arg_private_network = true;
903 arg_settings_mask |= SETTING_NETWORK;
904 break;
905 }
906
ab046dde 907 case ARG_NETWORK_BRIDGE:
ef76dff2 908
baaa35ad
ZJS		 909 if (!ifname_valid(optarg))
910 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
911 "Bridge interface name not valid: %s", optarg);
ef76dff2 912
f757855e
LP		 913 r = free_and_strdup(&arg_network_bridge, optarg);
914 if (r < 0)
915 return log_oom();
ab046dde 916
4831981d 917 _fallthrough_;
0dfaa006 918 case 'n':
69c79d3c
LP		 919 arg_network_veth = true;
920 arg_private_network = true;
f757855e 921 arg_settings_mask |= SETTING_NETWORK;
69c79d3c
LP		 922 break;
923
f6d6bad1
LP		 924 case ARG_NETWORK_VETH_EXTRA:
925 r = veth_extra_parse(&arg_network_veth_extra, optarg);
926 if (r < 0)
927 return log_error_errno(r, "Failed to parse --network-veth-extra= parameter: %s", optarg);
928
929 arg_private_network = true;
930 arg_settings_mask |= SETTING_NETWORK;
931 break;
932
aa28aefe 933 case ARG_NETWORK_INTERFACE:
2f091b1b 934 r = interface_pair_parse(&arg_network_interfaces, optarg);
b390f178
DDM		 935 if (r < 0)
936 return r;
937
c74e630d 938 arg_private_network = true;
f757855e 939 arg_settings_mask |= SETTING_NETWORK;
c74e630d
LP		 940 break;
941
942 case ARG_NETWORK_MACVLAN:
2f091b1b 943 r = macvlan_pair_parse(&arg_network_macvlan, optarg);
b390f178
DDM		 944 if (r < 0)
945 return r;
946
4bbfe7ad 947 arg_private_network = true;
f757855e 948 arg_settings_mask |= SETTING_NETWORK;
4bbfe7ad
TG		 949 break;
950
951 case ARG_NETWORK_IPVLAN:
2f091b1b 952 r = ipvlan_pair_parse(&arg_network_ipvlan, optarg);
b390f178
DDM		 953 if (r < 0)
954 return r;
955
4831981d 956 _fallthrough_;
ff01d048
LP		 957 case ARG_PRIVATE_NETWORK:
958 arg_private_network = true;
f757855e 959 arg_settings_mask |= SETTING_NETWORK;
a41fe3a2
LP		 960 break;
961
d7bea6b6 962 case ARG_NETWORK_NAMESPACE_PATH:
614b022c 963 r = parse_path_argument(optarg, false, &arg_network_namespace_path);
d7bea6b6
DP		 964 if (r < 0)
965 return r;
966
de40a303 967 arg_settings_mask |= SETTING_NETWORK;
d7bea6b6
DP		 968 break;
969
0f0dbc46 970 case 'b':
baaa35ad
ZJS		 971 if (arg_start_mode == START_PID2)
972 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
973 "--boot and --as-pid2 may not be combined.");
7732f92b
LP		 974
975 arg_start_mode = START_BOOT;
976 arg_settings_mask |= SETTING_START_MODE;
977 break;
978
979 case 'a':
baaa35ad
ZJS		 980 if (arg_start_mode == START_BOOT)
981 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
982 "--boot and --as-pid2 may not be combined.");
7732f92b
LP		 983
984 arg_start_mode = START_PID2;
985 arg_settings_mask |= SETTING_START_MODE;
0f0dbc46
LP		 986 break;
987
144f0fc0 988 case ARG_UUID:
aea3f594
ZJS		 989 r = id128_from_string_nonzero(optarg, &arg_uuid);
990 if (r == -ENXIO)
baaa35ad
ZJS		 991 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
992 "Machine UUID may not be all zeroes.");
aea3f594
ZJS		 993 if (r < 0)
994 return log_error_errno(r, "Invalid UUID: %s", optarg);
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")) {
6c2d70ce 1246 /* identity: User namespaces on, UID range is map the 0…0xFFFF range to
33eac552
LP		 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))
da890466 1403 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Root hash must be at least 128-bit 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
e8ac916e 1571 case ARG_SET_CREDENTIAL:
bd546b9b 1572 r = machine_credential_set(&arg_credentials, optarg);
3652872a 1573 if (r < 0)
6045958b
LP		 1574 return r;
1575
3652872a
LP		 1576 arg_settings_mask |= SETTING_CREDENTIALS;
1577 break;
3652872a 1578
e8ac916e 1579 case ARG_LOAD_CREDENTIAL:
bd546b9b 1580 r = machine_credential_load(&arg_credentials, optarg);
3652872a 1581 if (r < 0)
6045958b 1582 return r;
3652872a
LP		 1583
1584 arg_settings_mask |= SETTING_CREDENTIALS;
1585 break;
3652872a 1586
2f893044
LP		 1587 case ARG_BIND_USER:
1588 if (!valid_user_group_name(optarg, 0))
1589 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Invalid user name to bind: %s", optarg);
1590
1591 if (strv_extend(&arg_bind_user, optarg) < 0)
1592 return log_oom();
1593
1594 arg_settings_mask |= SETTING_BIND_USER;
1595 break;
1596
4a4654e0
LP		 1597 case ARG_SUPPRESS_SYNC:
1598 r = parse_boolean_argument("--suppress-sync=", optarg, &arg_suppress_sync);
1599 if (r < 0)
1600 return r;
1601
1602 arg_settings_mask |= SETTING_SUPPRESS_SYNC;
1603 break;
1604
06e78680
YW		 1605 case ARG_IMAGE_POLICY:
1606 r = parse_image_policy_argument(optarg, &arg_image_policy);
84be0c71 1607 if (r < 0)
06e78680 1608 return r;
84be0c71 1609 break;
84be0c71 1610
88213476
LP		 1611 case '?':
1612 return -EINVAL;
1613
1614 default:
04499a70 1615 assert_not_reached();
88213476 1616 }
88213476 1617
60f1ec13
LP		 1618 if (argc > optind) {
1619 strv_free(arg_parameters);
1620 arg_parameters = strv_copy(argv + optind);
1621 if (!arg_parameters)
1622 return log_oom();
d7bea6b6 1623
60f1ec13
LP		 1624 arg_settings_mask |= SETTING_START_MODE;
1625 }
1626
1627 if (arg_ephemeral && arg_template && !arg_directory)
1628 /* User asked for ephemeral execution but specified --template= instead of --directory=. Semantically
1629 * such an invocation makes some sense, see https://github.com/systemd/systemd/issues/3667. Let's
1630 * accept this here, and silently make "--ephemeral --template=" equivalent to "--ephemeral
1631 * --directory=". */
1632 arg_directory = TAKE_PTR(arg_template);
1633
2642d22a
DDM		 1634 arg_caps_retain |= plus;
1635 arg_caps_retain |= arg_private_network ? UINT64_C(1) << CAP_NET_ADMIN : 0;
2642d22a 1636 arg_caps_retain &= ~minus;
60f1ec13 1637
de40a303 1638 /* Make sure to parse environment before we reset the settings mask below */
49048684
ZJS		 1639 r = parse_environment();
1640 if (r < 0)
1641 return r;
de40a303 1642
60f1ec13
LP		 1643 /* Load all settings from .nspawn files */
1644 if (mask_no_settings)
1645 arg_settings_mask = 0;
1646
1647 /* Don't load any settings from .nspawn files */
1648 if (mask_all_settings)
1649 arg_settings_mask = _SETTINGS_MASK_ALL;
1650
1651 return 1;
1652}
1653
1654static int verify_arguments(void) {
1655 int r;
a6b5216c 1656
75b0d8b8
ZJS		 1657 if (arg_start_mode == START_PID2 && arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
1658 /* If we are running the stub init in the container, we don't need to look at what the init
1659 * in the container supports, because we are not using it. Let's immediately pick the right
1660 * setting based on the host system configuration.
1661 *
1662 * We only do this, if the user didn't use an environment variable to override the detection.
1663 */
1664
1665 r = cg_all_unified();
1666 if (r < 0)
1667 return log_error_errno(r, "Failed to determine whether we are in all unified mode.");
1668 if (r > 0)
1669 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_ALL;
1670 else if (cg_unified_controller(SYSTEMD_CGROUP_CONTROLLER) > 0)
1671 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_SYSTEMD;
1672 else
1673 arg_unified_cgroup_hierarchy = CGROUP_UNIFIED_NONE;
1674 }
1675
4f086aab
SU		 1676 if (arg_userns_mode != USER_NAMESPACE_NO)
1677 arg_mount_settings |= MOUNT_USE_USERNS;
1678
1679 if (arg_private_network)
1680 arg_mount_settings |= MOUNT_APPLY_APIVFS_NETNS;
1681
48a8d337
LB		 1682 if (!(arg_clone_ns_flags & CLONE_NEWPID) ||
1683 !(arg_clone_ns_flags & CLONE_NEWUTS)) {
eb91eb18 1684 arg_register = false;
baaa35ad 1685 if (arg_start_mode != START_PID1)
60f1ec13 1686 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--boot cannot be used without namespacing.");
0c582db0 1687 }
eb91eb18 1688
6c045a99
LP		 1689 if (arg_userns_ownership < 0)
1690 arg_userns_ownership =
f61c7f88 1691 arg_userns_mode == USER_NAMESPACE_PICK ? USER_NAMESPACE_OWNERSHIP_AUTO :
6c045a99 1692 USER_NAMESPACE_OWNERSHIP_OFF;
0e7ac751 1693
60f1ec13
LP		 1694 if (arg_start_mode == START_BOOT && arg_kill_signal <= 0)
1695 arg_kill_signal = SIGRTMIN+3;
1696
e5a4bb0d
LP		 1697 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 */
1698 arg_read_only = true;
1699
2436ea76
DDM		 1700 if (has_custom_root_mount(arg_custom_mounts, arg_n_custom_mounts))
1701 arg_read_only = true;
1702
baaa35ad 1703 if (arg_keep_unit && arg_register && cg_pid_get_owner_uid(0, NULL) >= 0)
8d9c2bca
AJ		 1704 /* Save the user from accidentally registering either user-$SESSION.scope or user@.service.
1705 * The latter is not technically a user session, but we don't need to labour the point. */
60f1ec13 1706 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--keep-unit --register=yes may not be used when invoked from a user session.");
89f7c846 1707
baaa35ad 1708 if (arg_directory && arg_image)
60f1ec13 1709 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--directory= and --image= may not be combined.");
1b9e5b12 1710
baaa35ad 1711 if (arg_template && arg_image)
60f1ec13 1712 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--template= and --image= may not be combined.");
8cd328d8 1713
baaa35ad 1714 if (arg_template && !(arg_directory || arg_machine))
60f1ec13 1715 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--template= needs --directory= or --machine=.");
ec16945e 1716
baaa35ad 1717 if (arg_ephemeral && arg_template)
60f1ec13 1718 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--ephemeral and --template= may not be combined.");
ec16945e 1719
baaa35ad 1720 if (arg_ephemeral && !IN_SET(arg_link_journal, LINK_NO, LINK_AUTO))
60f1ec13 1721 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--ephemeral and --link-journal= may not be combined.");
df9a75e4 1722
baaa35ad 1723 if (arg_userns_mode != USER_NAMESPACE_NO && !userns_supported())
60f1ec13 1724 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), "--private-users= is not supported, kernel compiled without user namespace support.");
7336138e 1725
6c045a99 1726 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_CHOWN && arg_read_only)
de40a303 1727 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
6c045a99 1728 "--read-only and --private-users-ownership=chown may not be combined.");
f757855e 1729
6c045a99
LP		 1730 /* We don't support --private-users-ownership=chown together with any of the volatile modes since we
1731 * couldn't change the read-only part of the tree (i.e. /usr) anyway, or because it would trigger a
1732 * massive copy-up (in case of overlay) making the entire exercise pointless. */
1733 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_CHOWN && arg_volatile_mode != VOLATILE_NO)
1734 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--volatile= and --private-users-ownership=chown may not be combined.");
e5a4bb0d 1735
679ecd36
SZ		 1736 /* If --network-namespace-path is given with any other network-related option (except --private-network),
1737 * we need to error out, to avoid conflicts between different network options. */
60f1ec13
LP		 1738 if (arg_network_namespace_path &&
1739 (arg_network_interfaces || arg_network_macvlan ||
1740 arg_network_ipvlan || arg_network_veth_extra ||
1741 arg_network_bridge || arg_network_zone ||
679ecd36 1742 arg_network_veth))
de40a303 1743 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--network-namespace-path= cannot be combined with other network options.");
86c0dd4a 1744
60f1ec13 1745 if (arg_network_bridge && arg_network_zone)
de40a303
LP		 1746 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
1747 "--network-bridge= and --network-zone= may not be combined.");
f757855e 1748
baaa35ad 1749 if (arg_userns_mode != USER_NAMESPACE_NO && (arg_mount_settings & MOUNT_APPLY_APIVFS_NETNS) && !arg_private_network)
60f1ec13 1750 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Invalid namespacing settings. Mounting sysfs with --private-users requires --private-network.");
4f086aab 1751
baaa35ad 1752 if (arg_userns_mode != USER_NAMESPACE_NO && !(arg_mount_settings & MOUNT_APPLY_APIVFS_RO))
60f1ec13 1753 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Cannot combine --private-users with read-write mounts.");
f757855e 1754
baaa35ad 1755 if (arg_expose_ports && !arg_private_network)
60f1ec13 1756 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Cannot use --port= without private networking.");
6d0b55c2 1757
88fc9c9b 1758 if (arg_caps_ambient) {
f5fbe71d 1759 if (arg_caps_ambient == UINT64_MAX)
88fc9c9b
TH		 1760 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= does not support the value all.");
1761
1762 if ((arg_caps_ambient & arg_caps_retain) != arg_caps_ambient)
1763 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= setting is not fully covered by Capability= setting.");
1764
1765 if (arg_start_mode == START_BOOT)
1766 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "AmbientCapability= setting is not useful for boot mode.");
1767 }
1768
2f893044
LP		 1769 if (arg_userns_mode == USER_NAMESPACE_NO && !strv_isempty(arg_bind_user))
1770 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "--bind-user= requires --private-users");
1771
1772 /* Drop duplicate --bind-user= entries */
1773 strv_uniq(arg_bind_user);
1774
60f1ec13
LP		 1775 r = custom_mount_check_all();
1776 if (r < 0)
1777 return r;
c6c8f6e2 1778
f757855e 1779 return 0;
88213476
LP		 1780}
1781
2f091b1b
TM		 1782static int verify_network_interfaces_initialized(void) {
1783 int r;
1784 r = test_network_interfaces_initialized(arg_network_interfaces);
1785 if (r < 0)
1786 return r;
1787
1788 r = test_network_interfaces_initialized(arg_network_macvlan);
1789 if (r < 0)
1790 return r;
1791
1792 r = test_network_interfaces_initialized(arg_network_ipvlan);
1793 if (r < 0)
1794 return r;
1795
1796 return 0;
1797}
1798
91181e07 1799int userns_lchown(const char *p, uid_t uid, gid_t gid) {
03cfe0d5
LP		 1800 assert(p);
1801
0de7acce 1802 if (arg_userns_mode == USER_NAMESPACE_NO)
03cfe0d5
LP		 1803 return 0;
1804
1805 if (uid == UID_INVALID && gid == GID_INVALID)
1806 return 0;
1807
1808 if (uid != UID_INVALID) {
1809 uid += arg_uid_shift;
1810
1811 if (uid < arg_uid_shift || uid >= arg_uid_shift + arg_uid_range)
1812 return -EOVERFLOW;
1813 }
1814
1815 if (gid != GID_INVALID) {
1816 gid += (gid_t) arg_uid_shift;
1817
1818 if (gid < (gid_t) arg_uid_shift || gid >= (gid_t) (arg_uid_shift + arg_uid_range))
1819 return -EOVERFLOW;
1820 }
1821
7c248223 1822 return RET_NERRNO(lchown(p, uid, gid));
b12afc8c
LP		 1823}
1824
91181e07 1825int userns_mkdir(const char *root, const char *path, mode_t mode, uid_t uid, gid_t gid) {
03cfe0d5 1826 const char *q;
dae8b82e 1827 int r;
03cfe0d5
LP		 1828
1829 q = prefix_roota(root, path);
3f692e2e 1830 r = RET_NERRNO(mkdir(q, mode));
dae8b82e
ZJS		 1831 if (r == -EEXIST)
1832 return 0;
1833 if (r < 0)
1834 return r;
03cfe0d5
LP		 1835
1836 return userns_lchown(q, uid, gid);
1837}
1838
1688841f 1839static const char *timezone_from_path(const char *path) {
da9fc98d
LP		 1840 return PATH_STARTSWITH_SET(
1841 path,
1842 "../usr/share/zoneinfo/",
1843 "/usr/share/zoneinfo/");
1688841f
LP		 1844}
1845
83205269
LP		 1846static bool etc_writable(void) {
1847 return !arg_read_only || IN_SET(arg_volatile_mode, VOLATILE_YES, VOLATILE_OVERLAY);
1848}
1849
e58a1277 1850static int setup_timezone(const char *dest) {
1688841f
LP		 1851 _cleanup_free_ char *p = NULL, *etc = NULL;
1852 const char *where, *check;
1853 TimezoneMode m;
d4036145 1854 int r;
f8440af5 1855
e58a1277
LP		 1856 assert(dest);
1857
1688841f 1858 if (IN_SET(arg_timezone, TIMEZONE_AUTO, TIMEZONE_SYMLINK)) {
1688841f
LP		 1859 r = readlink_malloc("/etc/localtime", &p);
1860 if (r == -ENOENT && arg_timezone == TIMEZONE_AUTO)
83205269 1861 m = etc_writable() ? TIMEZONE_DELETE : TIMEZONE_OFF;
1688841f 1862 else if (r == -EINVAL && arg_timezone == TIMEZONE_AUTO) /* regular file? */
83205269 1863 m = etc_writable() ? TIMEZONE_COPY : TIMEZONE_BIND;
1688841f
LP		 1864 else if (r < 0) {
1865 log_warning_errno(r, "Failed to read host's /etc/localtime symlink, not updating container timezone: %m");
1866 /* To handle warning, delete /etc/localtime and replace it with a symbolic link to a time zone data
1867 * file.
1868 *
1869 * Example:
1870 * ln -s /usr/share/zoneinfo/UTC /etc/localtime
1871 */
1872 return 0;
1873 } else if (arg_timezone == TIMEZONE_AUTO)
83205269 1874 m = etc_writable() ? TIMEZONE_SYMLINK : TIMEZONE_BIND;
1688841f
LP		 1875 else
1876 m = arg_timezone;
1877 } else
1878 m = arg_timezone;
1879
1880 if (m == TIMEZONE_OFF)
1881 return 0;
1882
f461a28d 1883 r = chase("/etc", dest, CHASE_PREFIX_ROOT, &etc, NULL);
d4036145 1884 if (r < 0) {
1688841f 1885 log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
d4036145
LP		 1886 return 0;
1887 }
1888
1688841f
LP		 1889 where = strjoina(etc, "/localtime");
1890
1891 switch (m) {
1892
1893 case TIMEZONE_DELETE:
1894 if (unlink(where) < 0)
1895 log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno, "Failed to remove '%s', ignoring: %m", where);
1896
d4036145 1897 return 0;
d4036145 1898
1688841f
LP		 1899 case TIMEZONE_SYMLINK: {
1900 _cleanup_free_ char *q = NULL;
1901 const char *z, *what;
4d1c38b8 1902
1688841f
LP		 1903 z = timezone_from_path(p);
1904 if (!z) {
1905 log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
d4036145 1906 return 0;
1688841f 1907 }
d4036145 1908
1688841f
LP		 1909 r = readlink_malloc(where, &q);
1910 if (r >= 0 && streq_ptr(timezone_from_path(q), z))
1911 return 0; /* Already pointing to the right place? Then do nothing .. */
1912
1913 check = strjoina(dest, "/usr/share/zoneinfo/", z);
f461a28d 1914 r = chase(check, dest, 0, NULL, NULL);
1688841f
LP		 1915 if (r < 0)
1916 log_debug_errno(r, "Timezone %s does not exist (or is not accessible) in container, not creating symlink: %m", z);
1917 else {
1918 if (unlink(where) < 0 && errno != ENOENT) {
1919 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING, /* Don't complain on read-only images */
1920 errno, "Failed to remove existing timezone info %s in container, ignoring: %m", where);
1921 return 0;
1922 }
1923
1924 what = strjoina("../usr/share/zoneinfo/", z);
1925 if (symlink(what, where) < 0) {
1926 log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING,
1927 errno, "Failed to correct timezone of container, ignoring: %m");
1928 return 0;
1929 }
1930
1931 break;
1932 }
1933
1934 _fallthrough_;
d4036145 1935 }
68fb0892 1936
1688841f
LP		 1937 case TIMEZONE_BIND: {
1938 _cleanup_free_ char *resolved = NULL;
1939 int found;
1940
f461a28d 1941 found = chase(where, dest, CHASE_NONEXISTENT, &resolved, NULL);
1688841f
LP		 1942 if (found < 0) {
1943 log_warning_errno(found, "Failed to resolve /etc/localtime path in container, ignoring: %m");
1944 return 0;
1945 }
1946
1947 if (found == 0) /* missing? */
1948 (void) touch(resolved);
1949
511a8cfe 1950 r = mount_nofollow_verbose(LOG_WARNING, "/etc/localtime", resolved, NULL, MS_BIND, NULL);
1688841f 1951 if (r >= 0)
511a8cfe 1952 return mount_nofollow_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
1688841f
LP		 1953
1954 _fallthrough_;
79d80fc1 1955 }
4d9f07b4 1956
1688841f
LP		 1957 case TIMEZONE_COPY:
1958 /* If mounting failed, try to copy */
7c2f5495 1959 r = copy_file_atomic("/etc/localtime", where, 0644, COPY_REFLINK|COPY_REPLACE);
1688841f
LP		 1960 if (r < 0) {
1961 log_full_errno(IN_SET(r, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
1962 "Failed to copy /etc/localtime to %s, ignoring: %m", where);
1963 return 0;
1964 }
1965
1966 break;
1967
1968 default:
04499a70 1969 assert_not_reached();
d4036145 1970 }
e58a1277 1971
1688841f 1972 /* Fix permissions of the symlink or file copy we just created */
03cfe0d5
LP		 1973 r = userns_lchown(where, 0, 0);
1974 if (r < 0)
1688841f 1975 log_warning_errno(r, "Failed to chown /etc/localtime, ignoring: %m");
03cfe0d5 1976
e58a1277 1977 return 0;
88213476
LP		 1978}
1979
09d423e9
LP		 1980static int have_resolv_conf(const char *path) {
1981 assert(path);
1982
1983 if (access(path, F_OK) < 0) {
1984 if (errno == ENOENT)
1985 return 0;
1986
1987 return log_debug_errno(errno, "Failed to determine whether '%s' is available: %m", path);
1988 }
1989
1990 return 1;
1991}
1992
7357272e 1993static int resolved_listening(void) {
b8ea7a6e 1994 _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL;
b053cd5f 1995 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
7357272e 1996 _cleanup_free_ char *dns_stub_listener_mode = NULL;
b053cd5f
LP		 1997 int r;
1998
7357272e 1999 /* Check if resolved is listening */
b053cd5f
LP		 2000
2001 r = sd_bus_open_system(&bus);
2002 if (r < 0)
b8ea7a6e 2003 return log_debug_errno(r, "Failed to open system bus: %m");
b053cd5f 2004
7357272e 2005 r = bus_name_has_owner(bus, "org.freedesktop.resolve1", NULL);
b8ea7a6e
LP		 2006 if (r < 0)
2007 return log_debug_errno(r, "Failed to check whether the 'org.freedesktop.resolve1' bus name is taken: %m");
2008 if (r == 0)
2009 return 0;
7357272e 2010
7f8a85e6 2011 r = bus_get_property_string(bus, bus_resolve_mgr, "DNSStubListener", &error, &dns_stub_listener_mode);
7357272e 2012 if (r < 0)
b8ea7a6e 2013 return log_debug_errno(r, "Failed to query DNSStubListener property: %s", bus_error_message(&error, r));
7357272e
DM		 2014
2015 return STR_IN_SET(dns_stub_listener_mode, "udp", "yes");
b053cd5f
LP		 2016}
2017
2547bb41 2018static int setup_resolv_conf(const char *dest) {
09d423e9
LP		 2019 _cleanup_free_ char *etc = NULL;
2020 const char *where, *what;
2021 ResolvConfMode m;
2022 int r;
2547bb41
LP		 2023
2024 assert(dest);
2025
09d423e9
LP		 2026 if (arg_resolv_conf == RESOLV_CONF_AUTO) {
2027 if (arg_private_network)
2028 m = RESOLV_CONF_OFF;
86775e35
LP		 2029 else if (have_resolv_conf(PRIVATE_STUB_RESOLV_CONF) > 0 && resolved_listening() > 0)
2030 m = etc_writable() ? RESOLV_CONF_COPY_STUB : RESOLV_CONF_BIND_STUB;
09d423e9 2031 else if (have_resolv_conf("/etc/resolv.conf") > 0)
83205269 2032 m = etc_writable() ? RESOLV_CONF_COPY_HOST : RESOLV_CONF_BIND_HOST;
09d423e9 2033 else
83205269 2034 m = etc_writable() ? RESOLV_CONF_DELETE : RESOLV_CONF_OFF;
86775e35 2035
09d423e9
LP		 2036 } else
2037 m = arg_resolv_conf;
2038
2039 if (m == RESOLV_CONF_OFF)
2547bb41
LP		 2040 return 0;
2041
f461a28d 2042 r = chase("/etc", dest, CHASE_PREFIX_ROOT, &etc, NULL);
87447ae4
LP		 2043 if (r < 0) {
2044 log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
2045 return 0;
2046 }
2047
2048 where = strjoina(etc, "/resolv.conf");
09d423e9
LP		 2049
2050 if (m == RESOLV_CONF_DELETE) {
2051 if (unlink(where) < 0)
2052 log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno, "Failed to remove '%s', ignoring: %m", where);
2053
87447ae4
LP		 2054 return 0;
2055 }
79d80fc1 2056
86775e35
LP		 2057 if (IN_SET(m, RESOLV_CONF_BIND_STATIC, RESOLV_CONF_REPLACE_STATIC, RESOLV_CONF_COPY_STATIC))
2058 what = PRIVATE_STATIC_RESOLV_CONF;
2059 else if (IN_SET(m, RESOLV_CONF_BIND_UPLINK, RESOLV_CONF_REPLACE_UPLINK, RESOLV_CONF_COPY_UPLINK))
2060 what = PRIVATE_UPLINK_RESOLV_CONF;
2061 else if (IN_SET(m, RESOLV_CONF_BIND_STUB, RESOLV_CONF_REPLACE_STUB, RESOLV_CONF_COPY_STUB))
2062 what = PRIVATE_STUB_RESOLV_CONF;
09d423e9
LP		 2063 else
2064 what = "/etc/resolv.conf";
87447ae4 2065
86775e35 2066 if (IN_SET(m, RESOLV_CONF_BIND_HOST, RESOLV_CONF_BIND_STATIC, RESOLV_CONF_BIND_UPLINK, RESOLV_CONF_BIND_STUB)) {
09d423e9
LP		 2067 _cleanup_free_ char *resolved = NULL;
2068 int found;
2069
d404c8d8 2070 found = chase(where, dest, CHASE_NONEXISTENT|CHASE_NOFOLLOW, &resolved, NULL);
09d423e9
LP		 2071 if (found < 0) {
2072 log_warning_errno(found, "Failed to resolve /etc/resolv.conf path in container, ignoring: %m");
2073 return 0;
2074 }
3539724c 2075
87447ae4
LP		 2076 if (found == 0) /* missing? */
2077 (void) touch(resolved);
5367354d 2078
511a8cfe 2079 r = mount_nofollow_verbose(LOG_WARNING, what, resolved, NULL, MS_BIND, NULL);
60e76d48 2080 if (r >= 0)
511a8cfe 2081 return mount_nofollow_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
86775e35
LP		 2082
2083 /* If that didn't work, let's copy the file */
3539724c
LP		 2084 }
2085
86775e35 2086 if (IN_SET(m, RESOLV_CONF_REPLACE_HOST, RESOLV_CONF_REPLACE_STATIC, RESOLV_CONF_REPLACE_UPLINK, RESOLV_CONF_REPLACE_STUB))
7c2f5495 2087 r = copy_file_atomic(what, where, 0644, COPY_REFLINK|COPY_REPLACE);
86775e35 2088 else
7c2f5495 2089 r = copy_file(what, where, O_TRUNC|O_NOFOLLOW, 0644, COPY_REFLINK);
79d80fc1 2090 if (r < 0) {
3539724c
LP		 2091 /* If the file already exists as symlink, let's suppress the warning, under the assumption that
2092 * resolved or something similar runs inside and the symlink points there.
68a313c5 2093 *
3539724c 2094 * If the disk image is read-only, there's also no point in complaining.
68a313c5 2095 */
86775e35
LP		 2096 log_full_errno(!IN_SET(RESOLV_CONF_COPY_HOST, RESOLV_CONF_COPY_STATIC, RESOLV_CONF_COPY_UPLINK, RESOLV_CONF_COPY_STUB) &&
2097 IN_SET(r, -ELOOP, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
3539724c 2098 "Failed to copy /etc/resolv.conf to %s, ignoring: %m", where);
79d80fc1
TG		 2099 return 0;
2100 }
2547bb41 2101
03cfe0d5
LP		 2102 r = userns_lchown(where, 0, 0);
2103 if (r < 0)
3539724c 2104 log_warning_errno(r, "Failed to chown /etc/resolv.conf, ignoring: %m");
03cfe0d5 2105
2547bb41
LP		 2106 return 0;
2107}
2108
1e4f1671 2109static int setup_boot_id(void) {
cdde6ba6
LP		 2110 _cleanup_(unlink_and_freep) char *from = NULL;
2111 _cleanup_free_ char *path = NULL;
3bbaff3e 2112 sd_id128_t rnd = SD_ID128_NULL;
cdde6ba6 2113 const char *to;
04bc4a3f
LP		 2114 int r;
2115
1eacc470 2116 /* Generate a new randomized boot ID, so that each boot-up of the container gets a new one */
04bc4a3f 2117
1eacc470 2118 r = tempfn_random_child("/run", "proc-sys-kernel-random-boot-id", &path);
cdde6ba6
LP		 2119 if (r < 0)
2120 return log_error_errno(r, "Failed to generate random boot ID path: %m");
04bc4a3f
LP		 2121
2122 r = sd_id128_randomize(&rnd);
f647962d
MS		 2123 if (r < 0)
2124 return log_error_errno(r, "Failed to generate random boot id: %m");
04bc4a3f 2125
b40c8ebd 2126 r = id128_write(path, ID128_FORMAT_UUID, rnd);
f647962d
MS		 2127 if (r < 0)
2128 return log_error_errno(r, "Failed to write boot id: %m");
04bc4a3f 2129
cdde6ba6
LP		 2130 from = TAKE_PTR(path);
2131 to = "/proc/sys/kernel/random/boot_id";
2132
511a8cfe 2133 r = mount_nofollow_verbose(LOG_ERR, from, to, NULL, MS_BIND, NULL);
cdde6ba6
LP		 2134 if (r < 0)
2135 return r;
04bc4a3f 2136
511a8cfe 2137 return mount_nofollow_verbose(LOG_ERR, NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
04bc4a3f
LP		 2138}
2139
e58a1277 2140static int copy_devnodes(const char *dest) {
88213476
LP		 2141 static const char devnodes[] =
2142 "null\0"
2143 "zero\0"
2144 "full\0"
2145 "random\0"
2146 "urandom\0"
85614d66
TG		 2147 "tty\0"
2148 "net/tun\0";
88213476 2149
e58a1277 2150 int r = 0;
a258bf26
LP		 2151
2152 assert(dest);
124640f1 2153
52f05ef2 2154 BLOCK_WITH_UMASK(0000);
88213476 2155
03cfe0d5
LP		 2156 /* Create /dev/net, so that we can create /dev/net/tun in it */
2157 if (userns_mkdir(dest, "/dev/net", 0755, 0, 0) < 0)
2158 return log_error_errno(r, "Failed to create /dev/net directory: %m");
2159
88213476 2160 NULSTR_FOREACH(d, devnodes) {
7fd1b19b 2161 _cleanup_free_ char *from = NULL, *to = NULL;
7f112f50 2162 struct stat st;
88213476 2163
c6134d3e 2164 from = path_join("/dev/", d);
8967f291
LP		 2165 if (!from)
2166 return log_oom();
2167
c6134d3e 2168 to = path_join(dest, from);
8967f291
LP		 2169 if (!to)
2170 return log_oom();
88213476
LP		 2171
2172 if (stat(from, &st) < 0) {
2173
4a62c710
MS		 2174 if (errno != ENOENT)
2175 return log_error_errno(errno, "Failed to stat %s: %m", from);
88213476 2176
baaa35ad
ZJS		 2177 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode))
2178 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2179 "%s is not a char or block device, cannot copy.", from);
2180 else {
8dfce114
LP		 2181 _cleanup_free_ char *sl = NULL, *prefixed = NULL, *dn = NULL, *t = NULL;
2182
81f5049b 2183 if (mknod(to, st.st_mode, st.st_rdev) < 0) {
8dbf71ec 2184 /* Explicitly warn the user when /dev is already populated. */
41eb4362 2185 if (errno == EEXIST)
1f87cc8c 2186 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		 2187 if (errno != EPERM)
2188 return log_error_errno(errno, "mknod(%s) failed: %m", to);
2189
8dfce114 2190 /* Some systems abusively restrict mknod but allow bind mounts. */
81f5049b
AC		 2191 r = touch(to);
2192 if (r < 0)
2193 return log_error_errno(r, "touch (%s) failed: %m", to);
511a8cfe 2194 r = mount_nofollow_verbose(LOG_DEBUG, from, to, NULL, MS_BIND, NULL);
60e76d48
ZJS		 2195 if (r < 0)
2196 return log_error_errno(r, "Both mknod and bind mount (%s) failed: %m", to);
81f5049b 2197 }
6278cf60 2198
03cfe0d5
LP		 2199 r = userns_lchown(to, 0, 0);
2200 if (r < 0)
2201 return log_error_errno(r, "chown() of device node %s failed: %m", to);
8dfce114 2202
657ee2d8 2203 dn = path_join("/dev", S_ISCHR(st.st_mode) ? "char" : "block");
8dfce114
LP		 2204 if (!dn)
2205 return log_oom();
2206
2207 r = userns_mkdir(dest, dn, 0755, 0, 0);
2208 if (r < 0)
2209 return log_error_errno(r, "Failed to create '%s': %m", dn);
2210
2211 if (asprintf(&sl, "%s/%u:%u", dn, major(st.st_rdev), minor(st.st_rdev)) < 0)
2212 return log_oom();
2213
c6134d3e 2214 prefixed = path_join(dest, sl);
8dfce114
LP		 2215 if (!prefixed)
2216 return log_oom();
2217
2d9b74ba 2218 t = path_join("..", d);
8dfce114
LP		 2219 if (!t)
2220 return log_oom();
2221
2222 if (symlink(t, prefixed) < 0)
2223 log_debug_errno(errno, "Failed to symlink '%s' to '%s': %m", t, prefixed);
88213476 2224 }
88213476
LP		 2225 }
2226
e58a1277
LP		 2227 return r;
2228}
88213476 2229
de40a303 2230static int make_extra_nodes(const char *dest) {
de40a303
LP		 2231 size_t i;
2232 int r;
2233
52f05ef2 2234 BLOCK_WITH_UMASK(0000);
de40a303
LP		 2235
2236 for (i = 0; i < arg_n_extra_nodes; i++) {
2237 _cleanup_free_ char *path = NULL;
2238 DeviceNode *n = arg_extra_nodes + i;
2239
c6134d3e 2240 path = path_join(dest, n->path);
de40a303
LP		 2241 if (!path)
2242 return log_oom();
2243
2244 if (mknod(path, n->mode, S_ISCHR(n->mode) || S_ISBLK(n->mode) ? makedev(n->major, n->minor) : 0) < 0)
2245 return log_error_errno(errno, "Failed to create device node '%s': %m", path);
2246
2247 r = chmod_and_chown(path, n->mode, n->uid, n->gid);
2248 if (r < 0)
2249 return log_error_errno(r, "Failed to adjust device node ownership of '%s': %m", path);
2250 }
2251
2252 return 0;
2253}
2254
03cfe0d5
LP		 2255static int setup_pts(const char *dest) {
2256 _cleanup_free_ char *options = NULL;
2257 const char *p;
709f6e46 2258 int r;
03cfe0d5 2259
349cc4a5 2260#if HAVE_SELINUX
03cfe0d5
LP		 2261 if (arg_selinux_apifs_context)
2262 (void) asprintf(&options,
3dce8915 2263 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"",
03cfe0d5
LP		 2264 arg_uid_shift + TTY_GID,
2265 arg_selinux_apifs_context);
2266 else
2267#endif
2268 (void) asprintf(&options,
3dce8915 2269 "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT,
03cfe0d5 2270 arg_uid_shift + TTY_GID);
f2d88580 2271
03cfe0d5 2272 if (!options)
f2d88580
LP		 2273 return log_oom();
2274
03cfe0d5 2275 /* Mount /dev/pts itself */
cc9fce65 2276 p = prefix_roota(dest, "/dev/pts");
3f692e2e 2277 r = RET_NERRNO(mkdir(p, 0755));
dae8b82e
ZJS		 2278 if (r < 0)
2279 return log_error_errno(r, "Failed to create /dev/pts: %m");
2280
511a8cfe 2281 r = mount_nofollow_verbose(LOG_ERR, "devpts", p, "devpts", MS_NOSUID|MS_NOEXEC, options);
60e76d48
ZJS		 2282 if (r < 0)
2283 return r;
709f6e46
MS		 2284 r = userns_lchown(p, 0, 0);
2285 if (r < 0)
2286 return log_error_errno(r, "Failed to chown /dev/pts: %m");
03cfe0d5
LP		 2287
2288 /* Create /dev/ptmx symlink */
2289 p = prefix_roota(dest, "/dev/ptmx");
4a62c710
MS		 2290 if (symlink("pts/ptmx", p) < 0)
2291 return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
709f6e46
MS		 2292 r = userns_lchown(p, 0, 0);
2293 if (r < 0)
2294 return log_error_errno(r, "Failed to chown /dev/ptmx: %m");
f2d88580 2295
03cfe0d5
LP		 2296 /* And fix /dev/pts/ptmx ownership */
2297 p = prefix_roota(dest, "/dev/pts/ptmx");
709f6e46
MS		 2298 r = userns_lchown(p, 0, 0);
2299 if (r < 0)
2300 return log_error_errno(r, "Failed to chown /dev/pts/ptmx: %m");
6278cf60 2301
f2d88580
LP		 2302 return 0;
2303}
2304
3acc84eb 2305static int setup_stdio_as_dev_console(void) {
5bb1d7fb 2306 _cleanup_close_ int terminal = -EBADF;
e58a1277 2307 int r;
e58a1277 2308
335d2ead
LP		 2309 /* We open the TTY in O_NOCTTY mode, so that we do not become controller yet. We'll do that later
2310 * explicitly, if we are configured to. */
2311 terminal = open_terminal("/dev/console", O_RDWR|O_NOCTTY);
3acc84eb
FB		 2312 if (terminal < 0)
2313 return log_error_errno(terminal, "Failed to open console: %m");
e58a1277 2314
3acc84eb
FB		 2315 /* Make sure we can continue logging to the original stderr, even if
2316 * stderr points elsewhere now */
2317 r = log_dup_console();
2318 if (r < 0)
2319 return log_error_errno(r, "Failed to duplicate stderr: %m");
de40a303 2320
3acc84eb
FB		 2321 /* invalidates 'terminal' on success and failure */
2322 r = rearrange_stdio(terminal, terminal, terminal);
2fef50cd 2323 TAKE_FD(terminal);
f647962d 2324 if (r < 0)
3acc84eb
FB		 2325 return log_error_errno(r, "Failed to move console to stdin/stdout/stderr: %m");
2326
2327 return 0;
2328}
88213476 2329
3acc84eb
FB		 2330static int setup_dev_console(const char *console) {
2331 _cleanup_free_ char *p = NULL;
2332 int r;
a258bf26 2333
3acc84eb
FB		 2334 /* Create /dev/console symlink */
2335 r = path_make_relative("/dev", console, &p);
81f5049b 2336 if (r < 0)
3acc84eb
FB		 2337 return log_error_errno(r, "Failed to create relative path: %m");
2338
2339 if (symlink(p, "/dev/console") < 0)
2340 return log_error_errno(errno, "Failed to create /dev/console symlink: %m");
a258bf26 2341
3acc84eb 2342 return 0;
e58a1277
LP		 2343}
2344
8e5430c4
LP		 2345static int setup_keyring(void) {
2346 key_serial_t keyring;
2347
6b000af4
LP		 2348 /* Allocate a new session keyring for the container. This makes sure the keyring of the session
2349 * systemd-nspawn was invoked from doesn't leak into the container. Note that by default we block
2350 * keyctl() and request_key() anyway via seccomp so doing this operation isn't strictly necessary,
2351 * but in case people explicitly allow-list these system calls let's make sure we don't leak anything
2352 * into the container. */
8e5430c4
LP		 2353
2354 keyring = keyctl(KEYCTL_JOIN_SESSION_KEYRING, 0, 0, 0, 0);
2355 if (keyring == -1) {
2356 if (errno == ENOSYS)
2357 log_debug_errno(errno, "Kernel keyring not supported, ignoring.");
065b4774 2358 else if (ERRNO_IS_PRIVILEGE(errno))
8e5430c4
LP		 2359 log_debug_errno(errno, "Kernel keyring access prohibited, ignoring.");
2360 else
2361 return log_error_errno(errno, "Setting up kernel keyring failed: %m");
2362 }
2363
2364 return 0;
2365}
2366
32fa2458
LP		 2367int make_run_host(const char *root) {
2368 int r;
2369
2370 assert(root);
2371
2372 r = userns_mkdir(root, "/run/host", 0755, 0, 0);
2373 if (r < 0)
2374 return log_error_errno(r, "Failed to create /run/host/: %m");
2375
2376 return 0;
2377}
2378
3652872a
LP		 2379static int setup_credentials(const char *root) {
2380 const char *q;
2381 int r;
2382
bd546b9b 2383 if (arg_credentials.n_credentials == 0)
3652872a
LP		 2384 return 0;
2385
32fa2458 2386 r = make_run_host(root);
3652872a 2387 if (r < 0)
32fa2458 2388 return r;
3652872a
LP		 2389
2390 r = userns_mkdir(root, "/run/host/credentials", 0700, 0, 0);
2391 if (r < 0)
2392 return log_error_errno(r, "Failed to create /run/host/credentials: %m");
2393
2394 q = prefix_roota(root, "/run/host/credentials");
511a8cfe 2395 r = mount_nofollow_verbose(LOG_ERR, NULL, q, "ramfs", MS_NOSUID|MS_NOEXEC|MS_NODEV, "mode=0700");
3652872a
LP		 2396 if (r < 0)
2397 return r;
2398
bd546b9b 2399 FOREACH_ARRAY(cred, arg_credentials.credentials, arg_credentials.n_credentials) {
3652872a 2400 _cleanup_free_ char *j = NULL;
254d1313 2401 _cleanup_close_ int fd = -EBADF;
3652872a 2402
bd546b9b 2403 j = path_join(q, cred->id);
3652872a
LP		 2404 if (!j)
2405 return log_oom();
2406
2407 fd = open(j, O_CREAT|O_EXCL|O_WRONLY|O_CLOEXEC|O_NOFOLLOW, 0600);
2408 if (fd < 0)
2409 return log_error_errno(errno, "Failed to create credential file %s: %m", j);
2410
bd546b9b 2411 r = loop_write(fd, cred->data, cred->size);
3652872a
LP		 2412 if (r < 0)
2413 return log_error_errno(r, "Failed to write credential to file %s: %m", j);
2414
2415 if (fchmod(fd, 0400) < 0)
2416 return log_error_errno(errno, "Failed to adjust access mode of %s: %m", j);
2417
2418 if (arg_userns_mode != USER_NAMESPACE_NO) {
2419 if (fchown(fd, arg_uid_shift, arg_uid_shift) < 0)
2420 return log_error_errno(errno, "Failed to adjust ownership of %s: %m", j);
2421 }
2422 }
2423
2424 if (chmod(q, 0500) < 0)
2425 return log_error_errno(errno, "Failed to adjust access mode of %s: %m", q);
2426
2427 r = userns_lchown(q, 0, 0);
2428 if (r < 0)
2429 return r;
2430
2431 /* Make both mount and superblock read-only now */
511a8cfe 2432 r = mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_REMOUNT|MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
3652872a
LP		 2433 if (r < 0)
2434 return r;
2435
511a8cfe 2436 return mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, "mode=0500");
3652872a
LP		 2437}
2438
5d9d3fcb 2439static int setup_kmsg(int fd_inner_socket) {
9ec5a93c
LP		 2440 _cleanup_(unlink_and_freep) char *from = NULL;
2441 _cleanup_free_ char *fifo = NULL;
254d1313 2442 _cleanup_close_ int fd = -EBADF;
9ec5a93c 2443 int r;
e58a1277 2444
5d9d3fcb 2445 assert(fd_inner_socket >= 0);
a258bf26 2446
52f05ef2 2447 BLOCK_WITH_UMASK(0000);
a258bf26 2448
30fd9a2d 2449 /* We create the kmsg FIFO as a temporary file in /run, but immediately delete it after bind mounting it to
9ec5a93c
LP		 2450 * /proc/kmsg. While FIFOs on the reading side behave very similar to /proc/kmsg, their writing side behaves
2451 * differently from /dev/kmsg in that writing blocks when nothing is reading. In order to avoid any problems
2452 * with containers deadlocking due to this we simply make /dev/kmsg unavailable to the container. */
2453
1eacc470 2454 r = tempfn_random_child("/run", "proc-kmsg", &fifo);
9ec5a93c
LP		 2455 if (r < 0)
2456 return log_error_errno(r, "Failed to generate kmsg path: %m");
e58a1277 2457
9ec5a93c 2458 if (mkfifo(fifo, 0600) < 0)
03cfe0d5 2459 return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m");
9ec5a93c
LP		 2460
2461 from = TAKE_PTR(fifo);
9ec5a93c 2462
511a8cfe 2463 r = mount_nofollow_verbose(LOG_ERR, from, "/proc/kmsg", NULL, MS_BIND, NULL);
60e76d48
ZJS		 2464 if (r < 0)
2465 return r;
e58a1277 2466
669fc4e5 2467 fd = open(from, O_RDWR|O_NONBLOCK|O_CLOEXEC);
4a62c710
MS		 2468 if (fd < 0)
2469 return log_error_errno(errno, "Failed to open fifo: %m");
e58a1277 2470
9ec5a93c 2471 /* Store away the fd in the socket, so that it stays open as long as we run the child */
5d9d3fcb 2472 r = send_one_fd(fd_inner_socket, fd, 0);
d9603714
DH		 2473 if (r < 0)
2474 return log_error_errno(r, "Failed to send FIFO fd: %m");
a258bf26 2475
25ea79fe 2476 return 0;
88213476
LP		 2477}
2478
761cf19d 2479struct ExposeArgs {
deff68e7
FW		 2480 union in_addr_union address4;
2481 union in_addr_union address6;
761cf19d
FW		 2482 struct FirewallContext *fw_ctx;
2483};
2484
1c4baffc 2485static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) {
99534007 2486 struct ExposeArgs *args = ASSERT_PTR(userdata);
6d0b55c2
LP		 2487
2488 assert(rtnl);
2489 assert(m);
6d0b55c2 2490
fb9044cb
LP		 2491 (void) expose_port_execute(rtnl, &args->fw_ctx, arg_expose_ports, AF_INET, &args->address4);
2492 (void) expose_port_execute(rtnl, &args->fw_ctx, arg_expose_ports, AF_INET6, &args->address6);
6d0b55c2
LP		 2493 return 0;
2494}
2495
3a74cea5 2496static int setup_hostname(void) {
c818eef1 2497 int r;
3a74cea5 2498
0c582db0 2499 if ((arg_clone_ns_flags & CLONE_NEWUTS) == 0)
eb91eb18
LP		 2500 return 0;
2501
c818eef1
LP		 2502 r = sethostname_idempotent(arg_hostname ?: arg_machine);
2503 if (r < 0)
2504 return log_error_errno(r, "Failed to set hostname: %m");
3a74cea5 2505
7027ff61 2506 return 0;
3a74cea5
LP		 2507}
2508
57fb9fb5 2509static int setup_journal(const char *directory) {
0f5e1382 2510 _cleanup_free_ char *d = NULL;
5980d463 2511 const char *p, *q;
b2238e38 2512 sd_id128_t this_id;
8054d749 2513 bool try;
57fb9fb5
LP		 2514 int r;
2515
df9a75e4
LP		 2516 /* Don't link journals in ephemeral mode */
2517 if (arg_ephemeral)
2518 return 0;
2519
8054d749
LP		 2520 if (arg_link_journal == LINK_NO)
2521 return 0;
2522
2523 try = arg_link_journal_try || arg_link_journal == LINK_AUTO;
2524
4d680aee 2525 r = sd_id128_get_machine(&this_id);
f647962d
MS		 2526 if (r < 0)
2527 return log_error_errno(r, "Failed to retrieve machine ID: %m");
4d680aee 2528
e01ff70a 2529 if (sd_id128_equal(arg_uuid, this_id)) {
8054d749 2530 log_full(try ? LOG_WARNING : LOG_ERR,
85b55869 2531 "Host and machine ids are equal (%s): refusing to link journals", SD_ID128_TO_STRING(arg_uuid));
8054d749 2532 if (try)
4d680aee 2533 return 0;
df9a75e4 2534 return -EEXIST;
4d680aee
ZJS		 2535 }
2536
369ca6da
ZJS		 2537 FOREACH_STRING(dirname, "/var", "/var/log", "/var/log/journal") {
2538 r = userns_mkdir(directory, dirname, 0755, 0, 0);
2539 if (r < 0) {
2540 bool ignore = r == -EROFS && try;
2541 log_full_errno(ignore ? LOG_DEBUG : LOG_ERR, r,
2542 "Failed to create %s%s: %m", dirname, ignore ? ", ignoring" : "");
2543 return ignore ? 0 : r;
2544 }
2545 }
03cfe0d5 2546
85b55869 2547 p = strjoina("/var/log/journal/", SD_ID128_TO_STRING(arg_uuid));
03cfe0d5 2548 q = prefix_roota(directory, p);
27407a01 2549
e1873695 2550 if (path_is_mount_point(p, NULL, 0) > 0) {
8054d749
LP		 2551 if (try)
2552 return 0;
27407a01 2553
baaa35ad
ZJS		 2554 return log_error_errno(SYNTHETIC_ERRNO(EEXIST),
2555 "%s: already a mount point, refusing to use for journal", p);
57fb9fb5
LP		 2556 }
2557
e1873695 2558 if (path_is_mount_point(q, NULL, 0) > 0) {
8054d749
LP		 2559 if (try)
2560 return 0;
57fb9fb5 2561
baaa35ad
ZJS		 2562 return log_error_errno(SYNTHETIC_ERRNO(EEXIST),
2563 "%s: already a mount point, refusing to use for journal", q);
57fb9fb5
LP		 2564 }
2565
2566 r = readlink_and_make_absolute(p, &d);
2567 if (r >= 0) {
3742095b 2568 if (IN_SET(arg_link_journal, LINK_GUEST, LINK_AUTO) &&
57fb9fb5
LP		 2569 path_equal(d, q)) {
2570
03cfe0d5 2571 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 2572 if (r < 0)
709f6e46 2573 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 2574 return 0;
57fb9fb5
LP		 2575 }
2576
4a62c710
MS		 2577 if (unlink(p) < 0)
2578 return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
57fb9fb5
LP		 2579 } else if (r == -EINVAL) {
2580
2581 if (arg_link_journal == LINK_GUEST &&
2582 rmdir(p) < 0) {
2583
27407a01
ZJS		 2584 if (errno == ENOTDIR) {
2585 log_error("%s already exists and is neither a symlink nor a directory", p);
2586 return r;
4314d33f
MS		 2587 } else
2588 return log_error_errno(errno, "Failed to remove %s: %m", p);
57fb9fb5 2589 }
4314d33f
MS		 2590 } else if (r != -ENOENT)
2591 return log_error_errno(r, "readlink(%s) failed: %m", p);
57fb9fb5
LP		 2592
2593 if (arg_link_journal == LINK_GUEST) {
2594
2595 if (symlink(q, p) < 0) {
8054d749 2596 if (try) {
56f64d95 2597 log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
574edc90 2598 return 0;
4314d33f
MS		 2599 } else
2600 return log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
57fb9fb5
LP		 2601 }
2602
03cfe0d5 2603 r = userns_mkdir(directory, p, 0755, 0, 0);
27407a01 2604 if (r < 0)
709f6e46 2605 log_warning_errno(r, "Failed to create directory %s: %m", q);
27407a01 2606 return 0;
57fb9fb5
LP		 2607 }
2608
2609 if (arg_link_journal == LINK_HOST) {
ccddd104 2610 /* don't create parents here — if the host doesn't have
574edc90 2611 * permanent journal set up, don't force it here */
ba8e6c4d 2612
3f692e2e 2613 r = RET_NERRNO(mkdir(p, 0755));
dae8b82e 2614 if (r < 0 && r != -EEXIST) {
8054d749 2615 if (try) {
dae8b82e 2616 log_debug_errno(r, "Failed to create %s, skipping journal setup: %m", p);
574edc90 2617 return 0;
4314d33f 2618 } else
dae8b82e 2619 return log_error_errno(r, "Failed to create %s: %m", p);
57fb9fb5
LP		 2620 }
2621
27407a01
ZJS		 2622 } else if (access(p, F_OK) < 0)
2623 return 0;
57fb9fb5 2624
db55bbf2 2625 if (dir_is_empty(q, /* ignore_hidden_or_backup= */ false) == 0)
cdb2b9d0
LP		 2626 log_warning("%s is not empty, proceeding anyway.", q);
2627
03cfe0d5 2628 r = userns_mkdir(directory, p, 0755, 0, 0);
709f6e46
MS		 2629 if (r < 0)
2630 return log_error_errno(r, "Failed to create %s: %m", q);
57fb9fb5 2631
511a8cfe 2632 r = mount_nofollow_verbose(LOG_DEBUG, p, q, NULL, MS_BIND, NULL);
60e76d48 2633 if (r < 0)
4a62c710 2634 return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
57fb9fb5 2635
27407a01 2636 return 0;
57fb9fb5
LP		 2637}
2638
de40a303
LP		 2639static int drop_capabilities(uid_t uid) {
2640 CapabilityQuintet q;
2641
2642 /* Let's initialize all five capability sets to something valid. If the quintet was configured via
2643 * OCI use that, but fill in missing bits. If it wasn't then derive the quintet in full from
2644 * arg_caps_retain. */
2645
2646 if (capability_quintet_is_set(&arg_full_capabilities)) {
2647 q = arg_full_capabilities;
2648
f5fbe71d 2649 if (q.bounding == UINT64_MAX)
de40a303
LP		 2650 q.bounding = uid == 0 ? arg_caps_retain : 0;
2651
f5fbe71d 2652 if (q.effective == UINT64_MAX)
de40a303
LP		 2653 q.effective = uid == 0 ? q.bounding : 0;
2654
f5fbe71d 2655 if (q.inheritable == UINT64_MAX)
88fc9c9b 2656 q.inheritable = uid == 0 ? q.bounding : arg_caps_ambient;
de40a303 2657
f5fbe71d 2658 if (q.permitted == UINT64_MAX)
88fc9c9b 2659 q.permitted = uid == 0 ? q.bounding : arg_caps_ambient;
de40a303 2660
f5fbe71d 2661 if (q.ambient == UINT64_MAX && ambient_capabilities_supported())
88fc9c9b 2662 q.ambient = arg_caps_ambient;
f66ad460
AZ		 2663
2664 if (capability_quintet_mangle(&q))
2665 return log_error_errno(SYNTHETIC_ERRNO(EPERM), "Cannot set capabilities that are not in the current bounding set.");
2666
2667 } else {
de40a303
LP		 2668 q = (CapabilityQuintet) {
2669 .bounding = arg_caps_retain,
2670 .effective = uid == 0 ? arg_caps_retain : 0,
88fc9c9b
TH		 2671 .inheritable = uid == 0 ? arg_caps_retain : arg_caps_ambient,
2672 .permitted = uid == 0 ? arg_caps_retain : arg_caps_ambient,
f5fbe71d 2673 .ambient = ambient_capabilities_supported() ? arg_caps_ambient : UINT64_MAX,
de40a303
LP		 2674 };
2675
f66ad460
AZ		 2676 /* If we're not using OCI, proceed with mangled capabilities (so we don't error out)
2677 * in order to maintain the same behavior as systemd < 242. */
2678 if (capability_quintet_mangle(&q))
0ccdaa79
JT		 2679 log_full(arg_quiet ? LOG_DEBUG : LOG_WARNING,
2680 "Some capabilities will not be set because they are not in the current bounding set.");
f66ad460
AZ		 2681
2682 }
2683
de40a303 2684 return capability_quintet_enforce(&q);
88213476
LP		 2685}
2686
db999e0f
LP		 2687static int reset_audit_loginuid(void) {
2688 _cleanup_free_ char *p = NULL;
2689 int r;
2690
0c582db0 2691 if ((arg_clone_ns_flags & CLONE_NEWPID) == 0)
db999e0f
LP		 2692 return 0;
2693
2694 r = read_one_line_file("/proc/self/loginuid", &p);
13e8ceb8 2695 if (r == -ENOENT)
db999e0f 2696 return 0;
f647962d
MS		 2697 if (r < 0)
2698 return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
db999e0f
LP		 2699
2700 /* Already reset? */
2701 if (streq(p, "4294967295"))
2702 return 0;
2703
57512c89 2704 r = write_string_file("/proc/self/loginuid", "4294967295", WRITE_STRING_FILE_DISABLE_BUFFER);
db999e0f 2705 if (r < 0) {
10a87006
LP		 2706 log_error_errno(r,
2707 "Failed to reset audit login UID. This probably means that your kernel is too\n"
2708 "old and you have audit enabled. Note that the auditing subsystem is known to\n"
2709 "be incompatible with containers on old kernels. Please make sure to upgrade\n"
2710 "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
2711 "using systemd-nspawn. Sleeping for 5s... (%m)");
77b6e194 2712
db999e0f 2713 sleep(5);
77b6e194 2714 }
db999e0f
LP		 2715
2716 return 0;
77b6e194
LP		 2717}
2718
e79581dd 2719static int mount_tunnel_dig(const char *root) {
785890ac 2720 const char *p, *q;
709f6e46 2721 int r;
785890ac
LP		 2722
2723 (void) mkdir_p("/run/systemd/nspawn/", 0755);
2724 (void) mkdir_p("/run/systemd/nspawn/propagate", 0600);
63c372cb 2725 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
785890ac
LP		 2726 (void) mkdir_p(p, 0600);
2727
32fa2458 2728 r = make_run_host(root);
709f6e46 2729 if (r < 0)
32fa2458 2730 return r;
03cfe0d5 2731
e79581dd 2732 r = userns_mkdir(root, NSPAWN_MOUNT_TUNNEL, 0600, 0, 0);
709f6e46 2733 if (r < 0)
e79581dd 2734 return log_error_errno(r, "Failed to create "NSPAWN_MOUNT_TUNNEL": %m");
03cfe0d5 2735
e79581dd 2736 q = prefix_roota(root, NSPAWN_MOUNT_TUNNEL);
511a8cfe 2737 r = mount_nofollow_verbose(LOG_ERR, p, q, NULL, MS_BIND, NULL);
60e76d48
ZJS		 2738 if (r < 0)
2739 return r;
785890ac 2740
511a8cfe 2741 r = mount_nofollow_verbose(LOG_ERR, NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
60e76d48
ZJS		 2742 if (r < 0)
2743 return r;
785890ac 2744
e79581dd
CB		 2745 return 0;
2746}
2747
2748static int mount_tunnel_open(void) {
2749 int r;
2750
2751 r = mount_follow_verbose(LOG_ERR, NULL, NSPAWN_MOUNT_TUNNEL, NULL, MS_SLAVE, NULL);
2752 if (r < 0)
2753 return r;
2754
2755 return 0;
785890ac
LP		 2756}
2757
317feb4d 2758static int setup_machine_id(const char *directory) {
3bbaff3e 2759 int r;
e01ff70a 2760
317feb4d
LP		 2761 /* If the UUID in the container is already set, then that's what counts, and we use. If it isn't set, and the
2762 * caller passed --uuid=, then we'll pass it in the $container_uuid env var to PID 1 of the container. The
2763 * assumption is that PID 1 will then write it to /etc/machine-id to make it persistent. If --uuid= is not
2764 * passed we generate a random UUID, and pass it via $container_uuid. In effect this means that /etc/machine-id
2765 * in the container and our idea of the container UUID will always be in sync (at least if PID 1 in the
2766 * container behaves nicely). */
2767
319477f1 2768 r = id128_get_machine(directory, &arg_uuid);
bb44fd07
ZJS		 2769 if (ERRNO_IS_NEG_MACHINE_ID_UNSET(r)) {
2770 /* If the file is missing, empty, or uninitialized, we don't mind */
317feb4d
LP		 2771 if (sd_id128_is_null(arg_uuid)) {
2772 r = sd_id128_randomize(&arg_uuid);
2773 if (r < 0)
2774 return log_error_errno(r, "Failed to acquire randomized machine UUID: %m");
2775 }
bb44fd07
ZJS		 2776 } else if (r < 0)
2777 return log_error_errno(r, "Failed to read machine ID from container image: %m");
691675ba 2778
e01ff70a
MS		 2779 return 0;
2780}
2781
7336138e
LP		 2782static int recursive_chown(const char *directory, uid_t shift, uid_t range) {
2783 int r;
2784
2785 assert(directory);
2786
6c045a99 2787 if (arg_userns_mode == USER_NAMESPACE_NO || arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_CHOWN)
7336138e
LP		 2788 return 0;
2789
2790 r = path_patch_uid(directory, arg_uid_shift, arg_uid_range);
2791 if (r == -EOPNOTSUPP)
2792 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.");
2793 if (r == -EBADE)
2794 return log_error_errno(r, "Upper 16 bits of root directory UID and GID do not match.");
2795 if (r < 0)
2796 return log_error_errno(r, "Failed to adjust UID/GID shift of OS tree: %m");
2797 if (r == 0)
2798 log_debug("Root directory of image is already owned by the right UID/GID range, skipping recursive chown operation.");
2799 else
2800 log_debug("Patched directory tree to match UID/GID range.");
2801
2802 return r;
2803}
2804
113cea80 2805/*
6d416b9c
LS		 2806 * Return values:
2807 * < 0 : wait_for_terminate() failed to get the state of the
2808 * container, the container was terminated by a signal, or
2809 * failed for an unknown reason. No change is made to the
2810 * container argument.
2811 * > 0 : The program executed in the container terminated with an
2812 * error. The exit code of the program executed in the
919699ec
LP		 2813 * container is returned. The container argument has been set
2814 * to CONTAINER_TERMINATED.
6d416b9c
LS		 2815 * 0 : The container is being rebooted, has been shut down or exited
2816 * successfully. The container argument has been set to either
2817 * CONTAINER_TERMINATED or CONTAINER_REBOOTED.
113cea80 2818 *
6d416b9c
LS		 2819 * That is, success is indicated by a return value of zero, and an
2820 * error is indicated by a non-zero value.
113cea80
DH		 2821 */
2822static int wait_for_container(pid_t pid, ContainerStatus *container) {
113cea80 2823 siginfo_t status;
919699ec 2824 int r;
113cea80
DH		 2825
2826 r = wait_for_terminate(pid, &status);
f647962d
MS		 2827 if (r < 0)
2828 return log_warning_errno(r, "Failed to wait for container: %m");
113cea80
DH		 2829
2830 switch (status.si_code) {
fddbb89c 2831
113cea80 2832 case CLD_EXITED:
b5a2179b 2833 if (status.si_status == 0)
919699ec 2834 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s exited successfully.", arg_machine);
b5a2179b 2835 else
919699ec 2836 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s failed with error code %i.", arg_machine, status.si_status);
fddbb89c 2837
919699ec
LP		 2838 *container = CONTAINER_TERMINATED;
2839 return status.si_status;
113cea80
DH		 2840
2841 case CLD_KILLED:
2842 if (status.si_status == SIGINT) {
919699ec 2843 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s has been shut down.", arg_machine);
113cea80 2844 *container = CONTAINER_TERMINATED;
919699ec
LP		 2845 return 0;
2846
113cea80 2847 } else if (status.si_status == SIGHUP) {
919699ec 2848 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO, "Container %s is being rebooted.", arg_machine);
113cea80 2849 *container = CONTAINER_REBOOTED;
919699ec 2850 return 0;
113cea80 2851 }
919699ec 2852
4831981d 2853 _fallthrough_;
113cea80 2854 case CLD_DUMPED:
baaa35ad
ZJS		 2855 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2856 "Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status));
113cea80
DH		 2857
2858 default:
baaa35ad
ZJS		 2859 return log_error_errno(SYNTHETIC_ERRNO(EIO),
2860 "Container %s failed due to unknown reason.", arg_machine);
113cea80 2861 }
113cea80
DH		 2862}
2863
023fb90b
LP		 2864static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
2865 pid_t pid;
2866
4a0b58c4 2867 pid = PTR_TO_PID(userdata);
023fb90b 2868 if (pid > 0) {
c6c8f6e2 2869 if (kill(pid, arg_kill_signal) >= 0) {
023fb90b
LP		 2870 log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
2871 sd_event_source_set_userdata(s, NULL);
2872 return 0;
2873 }
2874 }
2875
2876 sd_event_exit(sd_event_source_get_event(s), 0);
2877 return 0;
2878}
2879
6916b164 2880static int on_sigchld(sd_event_source *s, const struct signalfd_siginfo *ssi, void *userdata) {
abdb9b08
LP		 2881 pid_t pid;
2882
2883 assert(s);
2884 assert(ssi);
2885
2886 pid = PTR_TO_PID(userdata);
2887
6916b164
AU		 2888 for (;;) {
2889 siginfo_t si = {};
abdb9b08 2890
6916b164
AU		 2891 if (waitid(P_ALL, 0, &si, WNOHANG|WNOWAIT|WEXITED) < 0)
2892 return log_error_errno(errno, "Failed to waitid(): %m");
2893 if (si.si_pid == 0) /* No pending children. */
2894 break;
abdb9b08 2895 if (si.si_pid == pid) {
6916b164
AU		 2896 /* The main process we care for has exited. Return from
2897 * signal handler but leave the zombie. */
2898 sd_event_exit(sd_event_source_get_event(s), 0);
2899 break;
2900 }
abdb9b08 2901
6916b164
AU		 2902 /* Reap all other children. */
2903 (void) waitid(P_PID, si.si_pid, &si, WNOHANG|WEXITED);
2904 }
2905
2906 return 0;
2907}
2908
abdb9b08
LP		 2909static int on_request_stop(sd_bus_message *m, void *userdata, sd_bus_error *error) {
2910 pid_t pid;
2911
2912 assert(m);
2913
2914 pid = PTR_TO_PID(userdata);
2915
2916 if (arg_kill_signal > 0) {
2917 log_info("Container termination requested. Attempting to halt container.");
2918 (void) kill(pid, arg_kill_signal);
2919 } else {
2920 log_info("Container termination requested. Exiting.");
2921 sd_event_exit(sd_bus_get_event(sd_bus_message_get_bus(m)), 0);
2922 }
2923
2924 return 0;
2925}
2926
300a03be
LP		 2927static int pick_paths(void) {
2928 int r;
2929
2930 if (arg_directory) {
2931 _cleanup_(pick_result_done) PickResult result = PICK_RESULT_NULL;
2932 PickFilter filter = pick_filter_image_dir;
2933
2934 filter.architecture = arg_architecture;
2935
2936 r = path_pick_update_warn(
2937 &arg_directory,
2938 &filter,
2939 PICK_ARCHITECTURE|PICK_TRIES,
2940 &result);
2941 if (r < 0) {
2942 /* Accept ENOENT here so that the --template= logic can work */
2943 if (r != -ENOENT)
2944 return r;
2945 } else
2946 arg_architecture = result.architecture;
2947 }
2948
2949 if (arg_image) {
2950 _cleanup_(pick_result_done) PickResult result = PICK_RESULT_NULL;
2951 PickFilter filter = pick_filter_image_raw;
2952
2953 filter.architecture = arg_architecture;
2954
2955 r = path_pick_update_warn(
2956 &arg_image,
2957 &filter,
2958 PICK_ARCHITECTURE|PICK_TRIES,
2959 &result);
2960 if (r < 0)
2961 return r;
2962
2963 arg_architecture = result.architecture;
2964 }
2965
2966 if (arg_template) {
2967 _cleanup_(pick_result_done) PickResult result = PICK_RESULT_NULL;
2968 PickFilter filter = pick_filter_image_dir;
2969
2970 filter.architecture = arg_architecture;
2971
2972 r = path_pick_update_warn(
2973 &arg_template,
2974 &filter,
2975 PICK_ARCHITECTURE,
2976 &result);
2977 if (r < 0)
2978 return r;
2979
2980 arg_architecture = result.architecture;
2981 }
2982
2983 return 0;
2984}
2985
ec16945e 2986static int determine_names(void) {
1b9cebf6 2987 int r;
ec16945e 2988
c1521918
LP		 2989 if (arg_template && !arg_directory && arg_machine) {
2990
300a03be
LP		 2991 /* If --template= was specified then we should not search for a machine, but instead create a
2992 * new one in /var/lib/machine. */
c1521918 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
b36e39d2
LP		 3034 r = path_extract_filename(arg_image, &arg_machine);
3035 if (r < 0)
3036 return log_error_errno(r, "Failed to extract file name from '%s': %m", arg_image);
4827ab48 3037
e9b88a6d
LP		 3038 /* Truncate suffix if there is one */
3039 e = endswith(arg_machine, ".raw");
3040 if (e)
3041 *e = 0;
b36e39d2
LP		 3042 } else {
3043 r = path_extract_filename(arg_directory, &arg_machine);
3044 if (r < 0)
3045 return log_error_errno(r, "Failed to extract file name from '%s': %m", arg_directory);
3046 }
ec16945e 3047
ae691c1d 3048 hostname_cleanup(arg_machine);
52ef5dd7 3049 if (!hostname_is_valid(arg_machine, 0))
c6147113 3050 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to determine machine name automatically, please use -M.");
b9ba4dab 3051
3603f151
LB		 3052 /* Copy the machine name before the random suffix is added below, otherwise we won't be able
3053 * to match fixed config file names. */
3054 arg_settings_filename = strjoin(arg_machine, ".nspawn");
3055 if (!arg_settings_filename)
3056 return log_oom();
3057
e9b88a6d
LP		 3058 /* Add a random suffix when this is an ephemeral machine, so that we can run many
3059 * instances at once without manually having to specify -M each time. */
3060 if (arg_ephemeral)
3061 if (strextendf(&arg_machine, "-%016" PRIx64, random_u64()) < 0)
b9ba4dab 3062 return log_oom();
3603f151
LB		 3063 } else {
3064 arg_settings_filename = strjoin(arg_machine, ".nspawn");
3065 if (!arg_settings_filename)
3066 return log_oom();
ec16945e
LP		 3067 }
3068
3069 return 0;
3070}
3071
f461a28d 3072static int chase_and_update(char **p, unsigned flags) {
3f342ec4
LP		 3073 char *chased;
3074 int r;
3075
3076 assert(p);
3077
3078 if (!*p)
3079 return 0;
3080
f461a28d 3081 r = chase(*p, NULL, flags, &chased, NULL);
3f342ec4
LP		 3082 if (r < 0)
3083 return log_error_errno(r, "Failed to resolve path %s: %m", *p);
3084
a5648b80 3085 return free_and_replace(*p, chased);
3f342ec4
LP		 3086}
3087
03cfe0d5 3088static int determine_uid_shift(const char *directory) {
6dac160c 3089
0de7acce 3090 if (arg_userns_mode == USER_NAMESPACE_NO) {
03cfe0d5 3091 arg_uid_shift = 0;
6dac160c 3092 return 0;
03cfe0d5 3093 }
6dac160c
LP		 3094
3095 if (arg_uid_shift == UID_INVALID) {
3096 struct stat st;
3097
993da6d4
LP		 3098 /* Read the UID shift off the image. Maybe we can reuse this to avoid chowning. */
3099
3100 if (stat(directory, &st) < 0)
03cfe0d5 3101 return log_error_errno(errno, "Failed to determine UID base of %s: %m", directory);
6dac160c
LP		 3102
3103 arg_uid_shift = st.st_uid & UINT32_C(0xffff0000);
3104
baaa35ad
ZJS		 3105 if (arg_uid_shift != (st.st_gid & UINT32_C(0xffff0000)))
3106 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
3107 "UID and GID base of %s don't match.", directory);
6dac160c
LP		 3108
3109 arg_uid_range = UINT32_C(0x10000);
f61c7f88
LP		 3110
3111 if (arg_uid_shift != 0) {
3112 /* If the image is shifted already, then we'll fall back to classic chowning, for
3113 * compatibility (and simplicity), or refuse if mapping is explicitly requested. */
3114
3115 if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_AUTO) {
3116 log_debug("UID base of %s is non-zero, not using UID mapping.", directory);
3117 arg_userns_ownership = USER_NAMESPACE_OWNERSHIP_CHOWN;
3118 } else if (arg_userns_ownership == USER_NAMESPACE_OWNERSHIP_MAP)
3119 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
3120 "UID base of %s is not zero, UID mapping not supported.", directory);
3121 }
6dac160c
LP		 3122 }
3123
58e13de5
LP		 3124 if (!userns_shift_range_valid(arg_uid_shift, arg_uid_range))
3125 return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "UID base too high for UID range.");
6dac160c 3126
6dac160c
LP		 3127 return 0;
3128}
3129
de40a303
LP		 3130static unsigned long effective_clone_ns_flags(void) {
3131 unsigned long flags = arg_clone_ns_flags;
3132
3133 if (arg_private_network)
3134 flags |= CLONE_NEWNET;
3135 if (arg_use_cgns)
3136 flags |= CLONE_NEWCGROUP;
3137 if (arg_userns_mode != USER_NAMESPACE_NO)
3138 flags |= CLONE_NEWUSER;
3139
3140 return flags;
3141}
3142
3143static int patch_sysctl(void) {
3144
3145 /* This table is inspired by runc's sysctl() function */
3146 static const struct {
3147 const char *key;
3148 bool prefix;
3149 unsigned long clone_flags;
3150 } safe_sysctl[] = {
3151 { "kernel.hostname", false, CLONE_NEWUTS },
3152 { "kernel.domainname", false, CLONE_NEWUTS },
3153 { "kernel.msgmax", false, CLONE_NEWIPC },
3154 { "kernel.msgmnb", false, CLONE_NEWIPC },
3155 { "kernel.msgmni", false, CLONE_NEWIPC },
3156 { "kernel.sem", false, CLONE_NEWIPC },
3157 { "kernel.shmall", false, CLONE_NEWIPC },
3158 { "kernel.shmmax", false, CLONE_NEWIPC },
3159 { "kernel.shmmni", false, CLONE_NEWIPC },
3160 { "fs.mqueue.", true, CLONE_NEWIPC },
3161 { "net.", true, CLONE_NEWNET },
3162 };
3163
3164 unsigned long flags;
de40a303
LP		 3165 int r;
3166
3167 flags = effective_clone_ns_flags();
3168
3169 STRV_FOREACH_PAIR(k, v, arg_sysctl) {
3170 bool good = false;
3171 size_t i;
3172
3173 for (i = 0; i < ELEMENTSOF(safe_sysctl); i++) {
3174
3175 if (!FLAGS_SET(flags, safe_sysctl[i].clone_flags))
3176 continue;
3177
3178 if (safe_sysctl[i].prefix)
3179 good = startswith(*k, safe_sysctl[i].key);
3180 else
3181 good = streq(*k, safe_sysctl[i].key);
3182
3183 if (good)
3184 break;
3185 }
3186
c6147113
LP		 3187 if (!good)
3188 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		 3189
3190 r = sysctl_write(*k, *v);
3191 if (r < 0)
3192 return log_error_errno(r, "Failed to write sysctl '%s': %m", *k);
3193 }
3194
3195 return 0;
3196}
3197
03cfe0d5
LP		 3198static int inner_child(
3199 Barrier *barrier,
5d9d3fcb 3200 int fd_inner_socket,
e1bb4b0d
LB		 3201 FDSet *fds,
3202 char **os_release_pairs) {
69c79d3c 3203
03cfe0d5 3204 _cleanup_free_ char *home = NULL;
88614c8a 3205 size_t n_env = 1;
4ab3d29f
ZJS		 3206 char *envp[] = {
3207 (char*) "PATH=" DEFAULT_PATH_COMPAT,
6aadfa4c 3208 NULL, /* container */
03cfe0d5
LP		 3209 NULL, /* TERM */
3210 NULL, /* HOME */
3211 NULL, /* USER */
3212 NULL, /* LOGNAME */
3213 NULL, /* container_uuid */
3214 NULL, /* LISTEN_FDS */
3215 NULL, /* LISTEN_PID */
9c1e04d0 3216 NULL, /* NOTIFY_SOCKET */
3652872a 3217 NULL, /* CREDENTIALS_DIRECTORY */
b626f695 3218 NULL, /* LANG */
03cfe0d5
LP		 3219 NULL
3220 };
1a68e1e5 3221 const char *exec_target;
2371271c 3222 _cleanup_strv_free_ char **env_use = NULL;
de40a303 3223 int r, which_failed;
88213476 3224
b37469d7
LP		 3225 /* This is the "inner" child process, i.e. the one forked off by the "outer" child process, which is the one
3226 * the container manager itself forked off. At the time of clone() it gained its own CLONE_NEWNS, CLONE_NEWPID,
3227 * CLONE_NEWUTS, CLONE_NEWIPC, CLONE_NEWUSER namespaces. Note that it has its own CLONE_NEWNS namespace,
3228 * separate from the CLONE_NEWNS created for the "outer" child, and also separate from the host's CLONE_NEWNS
3229 * namespace. The reason for having two levels of CLONE_NEWNS namespaces is that the "inner" one is owned by
3230 * the CLONE_NEWUSER namespace of the container, while the "outer" one is owned by the host's CLONE_NEWUSER
3231 * namespace.
3232 *
3233 * Note at this point we have no CLONE_NEWNET namespace yet. We'll acquire that one later through
3234 * unshare(). See below. */
3235
03cfe0d5 3236 assert(barrier);
5d9d3fcb 3237 assert(fd_inner_socket >= 0);
88213476 3238
de40a303
LP		 3239 log_debug("Inner child is initializing.");
3240
0de7acce 3241 if (arg_userns_mode != USER_NAMESPACE_NO) {
03cfe0d5
LP		 3242 /* Tell the parent, that it now can write the UID map. */
3243 (void) barrier_place(barrier); /* #1 */
7027ff61 3244
03cfe0d5 3245 /* Wait until the parent wrote the UID map */
baaa35ad 3246 if (!barrier_place_and_sync(barrier)) /* #2 */
2a2e78e9 3247 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Parent died too early");
88213476 3248
2a2e78e9
LP		 3249 /* Become the new root user inside our namespace */
3250 r = reset_uid_gid();
3251 if (r < 0)
3252 return log_error_errno(r, "Couldn't become new root: %m");
3253
3254 /* Creating a new user namespace means all MS_SHARED mounts become MS_SLAVE. Let's put them
3255 * back to MS_SHARED here, since that's what we want as defaults. (This will not reconnect
3256 * propagation, but simply create new peer groups for all our mounts). */
511a8cfe 3257 r = mount_follow_verbose(LOG_ERR, NULL, "/", NULL, MS_SHARED|MS_REC, NULL);
2a2e78e9
LP		 3258 if (r < 0)
3259 return r;
3260 }
6d66bd3b 3261
0de7acce 3262 r = mount_all(NULL,
4f086aab 3263 arg_mount_settings | MOUNT_IN_USERNS,
0de7acce 3264 arg_uid_shift,
0de7acce 3265 arg_selinux_apifs_context);
03cfe0d5
LP		 3266 if (r < 0)
3267 return r;
3268
04413780
ZJS		 3269 if (!arg_network_namespace_path && arg_private_network) {
3270 r = unshare(CLONE_NEWNET);
3271 if (r < 0)
3272 return log_error_errno(errno, "Failed to unshare network namespace: %m");
75116558
PS		 3273
3274 /* Tell the parent that it can setup network interfaces. */
3275 (void) barrier_place(barrier); /* #3 */
04413780
ZJS		 3276 }
3277
4f086aab 3278 r = mount_sysfs(NULL, arg_mount_settings);
d8fc6a00
LP		 3279 if (r < 0)
3280 return r;
3281
03cfe0d5
LP		 3282 /* Wait until we are cgroup-ified, so that we
3283 * can mount the right cgroup path writable */
baaa35ad
ZJS		 3284 if (!barrier_place_and_sync(barrier)) /* #4 */
3285 return log_error_errno(SYNTHETIC_ERRNO(ESRCH),
3286 "Parent died too early");
88213476 3287
489fae52 3288 if (arg_use_cgns) {
0996ef00
CB		 3289 r = unshare(CLONE_NEWCGROUP);
3290 if (r < 0)
04413780 3291 return log_error_errno(errno, "Failed to unshare cgroup namespace: %m");
0996ef00
CB		 3292 r = mount_cgroups(
3293 "",
3294 arg_unified_cgroup_hierarchy,
3295 arg_userns_mode != USER_NAMESPACE_NO,
3296 arg_uid_shift,
3297 arg_uid_range,
5a8ff0e6 3298 arg_selinux_apifs_context,
ada54120 3299 true);
1433e0f2 3300 } else
0996ef00 3301 r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy);
1433e0f2
LP		 3302 if (r < 0)
3303 return r;
ec16945e 3304
1e4f1671 3305 r = setup_boot_id();
03cfe0d5
LP		 3306 if (r < 0)
3307 return r;
ec16945e 3308
5d9d3fcb 3309 r = setup_kmsg(fd_inner_socket);
03cfe0d5
LP		 3310 if (r < 0)
3311 return r;
ec16945e 3312
de40a303
LP		 3313 r = mount_custom(
3314 "/",
3315 arg_custom_mounts,
3316 arg_n_custom_mounts,
de40a303 3317 0,
c0c8f718 3318 0,
de40a303 3319 arg_selinux_apifs_context,
5f0a6347 3320 MOUNT_NON_ROOT_ONLY | MOUNT_IN_USERNS);
de40a303
LP		 3321 if (r < 0)
3322 return r;
3323
03cfe0d5
LP		 3324 if (setsid() < 0)
3325 return log_error_errno(errno, "setsid() failed: %m");
3326
3327 if (arg_private_network)
df883de9 3328 (void) loopback_setup();
03cfe0d5 3329
7a8f6325 3330 if (arg_expose_ports) {
b07ee903 3331 r = expose_port_send_rtnl(fd_inner_socket);
7a8f6325
LP		 3332 if (r < 0)
3333 return r;
7a8f6325 3334 }
03cfe0d5 3335
3acc84eb 3336 if (arg_console_mode != CONSOLE_PIPE) {
5bb1d7fb 3337 _cleanup_close_ int master = -EBADF;
3acc84eb
FB		 3338 _cleanup_free_ char *console = NULL;
3339
3340 /* Allocate a pty and make it available as /dev/console. */
dc98caea 3341 master = openpt_allocate(O_RDWR|O_NONBLOCK, &console);
3acc84eb 3342 if (master < 0)
dc98caea 3343 return log_error_errno(master, "Failed to allocate a pty: %m");
3acc84eb
FB		 3344
3345 r = setup_dev_console(console);
3346 if (r < 0)
105a1a36 3347 return log_error_errno(r, "Failed to set up /dev/console: %m");
3acc84eb 3348
bb1aa185 3349 r = send_one_fd(fd_inner_socket, master, 0);
3acc84eb
FB		 3350 if (r < 0)
3351 return log_error_errno(r, "Failed to send master fd: %m");
3acc84eb
FB		 3352
3353 r = setup_stdio_as_dev_console();
3354 if (r < 0)
3355 return r;
3356 }
3357
de40a303
LP		 3358 r = patch_sysctl();
3359 if (r < 0)
3360 return r;
3361
81f345df
LP		 3362 if (arg_oom_score_adjust_set) {
3363 r = set_oom_score_adjust(arg_oom_score_adjust);
3364 if (r < 0)
3365 return log_error_errno(r, "Failed to adjust OOM score: %m");
3366 }
3367
0985c7c4
ZJS		 3368 if (arg_cpu_set.set)
3369 if (sched_setaffinity(0, arg_cpu_set.allocated, arg_cpu_set.set) < 0)
d107bb7d
LP		 3370 return log_error_errno(errno, "Failed to set CPU affinity: %m");
3371
c818eef1 3372 (void) setup_hostname();
03cfe0d5 3373
050f7277 3374 if (arg_personality != PERSONALITY_INVALID) {
21022b9d
LP		 3375 r = safe_personality(arg_personality);
3376 if (r < 0)
3377 return log_error_errno(r, "personality() failed: %m");
4c27749b
LP		 3378#ifdef ARCHITECTURE_SECONDARY
3379 } else if (arg_architecture == ARCHITECTURE_SECONDARY) {
21022b9d
LP		 3380 r = safe_personality(PER_LINUX32);
3381 if (r < 0)
3382 return log_error_errno(r, "personality() failed: %m");
4c27749b 3383#endif
af262e5f
LB		 3384 } else if (!arg_quiet && arg_architecture >= 0 && arg_architecture != native_architecture())
3385 log_notice("Selected architecture '%s' not supported natively on the local CPU, assuming "
3386 "invocation with qemu userspace emulator (or equivalent) in effect.",
3387 architecture_to_string(arg_architecture));
03cfe0d5 3388
de40a303
LP		 3389 r = setrlimit_closest_all((const struct rlimit *const*) arg_rlimit, &which_failed);
3390 if (r < 0)
3391 return log_error_errno(r, "Failed to apply resource limit RLIMIT_%s: %m", rlimit_to_string(which_failed));
3392
3393#if HAVE_SECCOMP
3394 if (arg_seccomp) {
3395
3396 if (is_seccomp_available()) {
de40a303 3397 r = seccomp_load(arg_seccomp);
3c098014
ZJS		 3398 if (ERRNO_IS_NEG_SECCOMP_FATAL(r))
3399 return log_error_errno(r, "Failed to install seccomp filter: %m");
3400 if (r < 0)
de40a303
LP		 3401 log_debug_errno(r, "Failed to install seccomp filter: %m");
3402 }
3403 } else
3404#endif
3405 {
6b000af4 3406 r = setup_seccomp(arg_caps_retain, arg_syscall_allow_list, arg_syscall_deny_list);
de40a303
LP		 3407 if (r < 0)
3408 return r;
3409 }
3410
4a4654e0 3411 if (arg_suppress_sync) {
20e458ae 3412#if HAVE_SECCOMP
4a4654e0
LP		 3413 r = seccomp_suppress_sync();
3414 if (r < 0)
3415 log_debug_errno(r, "Failed to install sync() suppression seccomp filter, ignoring: %m");
20e458ae 3416#else
2db32618 3417 log_debug("systemd is built without SECCOMP support. Ignoring --suppress-sync= command line option and SuppressSync= setting.");
20e458ae 3418#endif
4a4654e0
LP		 3419 }
3420
349cc4a5 3421#if HAVE_SELINUX
03cfe0d5 3422 if (arg_selinux_context)
2ed96880 3423 if (setexeccon(arg_selinux_context) < 0)
03cfe0d5
LP		 3424 return log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context);
3425#endif
3426
de40a303
LP		 3427 /* Make sure we keep the caps across the uid/gid dropping, so that we can retain some selected caps
3428 * if we need to later on. */
3429 if (prctl(PR_SET_KEEPCAPS, 1) < 0)
3430 return log_error_errno(errno, "Failed to set PR_SET_KEEPCAPS: %m");
3431
3432 if (uid_is_valid(arg_uid) || gid_is_valid(arg_gid))
3462d773 3433 r = change_uid_gid_raw(arg_uid, arg_gid, arg_supplementary_gids, arg_n_supplementary_gids, arg_console_mode != CONSOLE_PIPE);
de40a303 3434 else
3462d773 3435 r = change_uid_gid(arg_user, arg_console_mode != CONSOLE_PIPE, &home);
03cfe0d5
LP		 3436 if (r < 0)
3437 return r;
3438
de40a303
LP		 3439 r = drop_capabilities(getuid());
3440 if (r < 0)
3441 return log_error_errno(r, "Dropping capabilities failed: %m");
3442
66edd963
LP		 3443 if (arg_no_new_privileges)
3444 if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0)
3445 return log_error_errno(errno, "Failed to disable new privileges: %m");
3446
6aadfa4c
ILG		 3447 /* LXC sets container=lxc, so follow the scheme here */
3448 envp[n_env++] = strjoina("container=", arg_container_service_name);
3449
03cfe0d5
LP		 3450 envp[n_env] = strv_find_prefix(environ, "TERM=");
3451 if (envp[n_env])
313cefa1 3452 n_env++;
03cfe0d5 3453
de40a303 3454 if (home || !uid_is_valid(arg_uid) || arg_uid == 0)
4ab3d29f 3455 if (asprintf(envp + n_env++, "HOME=%s", home ?: "/root") < 0)
de40a303
LP		 3456 return log_oom();
3457
3458 if (arg_user || !uid_is_valid(arg_uid) || arg_uid == 0)
4ab3d29f 3459 if (asprintf(envp + n_env++, "USER=%s", arg_user ?: "root") < 0 ||
1da3cb81 3460 asprintf(envp + n_env++, "LOGNAME=%s", arg_user ?: "root") < 0)
de40a303 3461 return log_oom();
03cfe0d5 3462
3bbaff3e 3463 assert(!sd_id128_is_null(arg_uuid));
03cfe0d5 3464
b7416360 3465 if (asprintf(envp + n_env++, "container_uuid=%s", SD_ID128_TO_UUID_STRING(arg_uuid)) < 0)
e01ff70a 3466 return log_oom();
03cfe0d5 3467
43127aeb 3468 if (!fdset_isempty(fds)) {
03cfe0d5
LP		 3469 r = fdset_cloexec(fds, false);
3470 if (r < 0)
3471 return log_error_errno(r, "Failed to unset O_CLOEXEC for file descriptors.");
3472
4ab3d29f
ZJS		 3473 if ((asprintf(envp + n_env++, "LISTEN_FDS=%u", fdset_size(fds)) < 0) ||
3474 (asprintf(envp + n_env++, "LISTEN_PID=1") < 0))
03cfe0d5
LP		 3475 return log_oom();
3476 }
4ab3d29f 3477 if (asprintf(envp + n_env++, "NOTIFY_SOCKET=%s", NSPAWN_NOTIFY_SOCKET_PATH) < 0)
9c1e04d0 3478 return log_oom();
03cfe0d5 3479
bd546b9b 3480 if (arg_credentials.n_credentials > 0) {
3652872a
LP		 3481 envp[n_env] = strdup("CREDENTIALS_DIRECTORY=/run/host/credentials");
3482 if (!envp[n_env])
3483 return log_oom();
3484 n_env++;
3485 }
3486
b626f695 3487 if (arg_start_mode != START_BOOT) {
a22f5186 3488 envp[n_env] = strdup("LANG=" SYSTEMD_NSPAWN_LOCALE);
b626f695
DDM		 3489 if (!envp[n_env])
3490 return log_oom();
3491 n_env++;
3492 }
3493
4ab3d29f 3494 env_use = strv_env_merge(envp, os_release_pairs, arg_setenv);
2371271c
TG		 3495 if (!env_use)
3496 return log_oom();
03cfe0d5 3497
1a8d7814 3498 /* Let the parent know that we are ready and wait until the parent is ready with the setup, too... */
baaa35ad 3499 if (!barrier_place_and_sync(barrier)) /* #5 */
335d2ead 3500 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Parent died too early");
03cfe0d5 3501
5f932eb9
LP		 3502 if (arg_chdir)
3503 if (chdir(arg_chdir) < 0)
3504 return log_error_errno(errno, "Failed to change to specified working directory %s: %m", arg_chdir);
3505
7732f92b 3506 if (arg_start_mode == START_PID2) {
75bf701f 3507 r = stub_pid1(arg_uuid);
7732f92b
LP		 3508 if (r < 0)
3509 return r;
3510 }
3511
335d2ead
LP		 3512 if (arg_console_mode != CONSOLE_PIPE) {
3513 /* So far our pty wasn't controlled by any process. Finally, it's time to change that, if we
3514 * are configured for that. Acquire it as controlling tty. */
3515 if (ioctl(STDIN_FILENO, TIOCSCTTY) < 0)
3516 return log_error_errno(errno, "Failed to acquire controlling TTY: %m");
3517 }
3518
de40a303
LP		 3519 log_debug("Inner child completed, invoking payload.");
3520
8ca082b4
LP		 3521 /* Now, explicitly close the log, so that we then can close all remaining fds. Closing the log explicitly first
3522 * has the benefit that the logging subsystem knows about it, and is thus ready to be reopened should we need
3523 * it again. Note that the other fds closed here are at least the locking and barrier fds. */
03cfe0d5 3524 log_close();
8ca082b4 3525 log_set_open_when_needed(true);
a3b00f91 3526 log_settle_target();
8ca082b4 3527
03cfe0d5
LP		 3528 (void) fdset_close_others(fds);
3529
7732f92b 3530 if (arg_start_mode == START_BOOT) {
03cfe0d5
LP		 3531 char **a;
3532 size_t m;
3533
3534 /* Automatically search for the init system */
3535
75f32f04
ZJS		 3536 m = strv_length(arg_parameters);
3537 a = newa(char*, m + 2);
3538 memcpy_safe(a + 1, arg_parameters, m * sizeof(char*));
3539 a[1 + m] = NULL;
03cfe0d5 3540
a5096641
LP		 3541 FOREACH_STRING(init,
3542 "/usr/lib/systemd/systemd",
3543 "/lib/systemd/systemd",
3544 "/sbin/init") {
3545 a[0] = (char*) init;
3546 execve(a[0], a, env_use);
3547 }
ced58da7
LP		 3548
3549 exec_target = "/usr/lib/systemd/systemd, /lib/systemd/systemd, /sbin/init";
1a68e1e5 3550 } else if (!strv_isempty(arg_parameters)) {
b6b180b7
LP		 3551 const char *dollar_path;
3552
1a68e1e5 3553 exec_target = arg_parameters[0];
b6b180b7
LP		 3554
3555 /* Use the user supplied search $PATH if there is one, or DEFAULT_PATH_COMPAT if not to search the
3556 * binary. */
3557 dollar_path = strv_env_get(env_use, "PATH");
3558 if (dollar_path) {
6f646e01 3559 if (setenv("PATH", dollar_path, 1) < 0)
b6b180b7
LP		 3560 return log_error_errno(errno, "Failed to update $PATH: %m");
3561 }
3562
f757855e 3563 execvpe(arg_parameters[0], arg_parameters, env_use);
1a68e1e5 3564 } else {
5f932eb9 3565 if (!arg_chdir)
d929b0f9
ZJS		 3566 /* If we cannot change the directory, we'll end up in /, that is expected. */
3567 (void) chdir(home ?: "/root");
5f932eb9 3568
53350c7b 3569 execle(DEFAULT_USER_SHELL, "-" DEFAULT_USER_SHELL_NAME, NULL, env_use);
3570 if (!streq(DEFAULT_USER_SHELL, "/bin/bash"))
3571 execle("/bin/bash", "-bash", NULL, env_use);
3572 if (!streq(DEFAULT_USER_SHELL, "/bin/sh"))
3573 execle("/bin/sh", "-sh", NULL, env_use);
ced58da7 3574
53350c7b 3575 exec_target = DEFAULT_USER_SHELL ", /bin/bash, /bin/sh";
03cfe0d5
LP		 3576 }
3577
8ca082b4 3578 return log_error_errno(errno, "execv(%s) failed: %m", exec_target);
03cfe0d5
LP		 3579}
3580
e96ceaba 3581static int setup_notify_child(void) {
254d1313 3582 _cleanup_close_ int fd = -EBADF;
1eb874b9 3583 static const union sockaddr_union sa = {
44ed5214
LP		 3584 .un.sun_family = AF_UNIX,
3585 .un.sun_path = NSPAWN_NOTIFY_SOCKET_PATH,
9c1e04d0
AP		 3586 };
3587 int r;
3588
3589 fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
3590 if (fd < 0)
3591 return log_error_errno(errno, "Failed to allocate notification socket: %m");
3592
3593 (void) mkdir_parents(NSPAWN_NOTIFY_SOCKET_PATH, 0755);
fbda85b0 3594 (void) sockaddr_un_unlink(&sa.un);
9c1e04d0 3595
9c1e04d0 3596 r = bind(fd, &sa.sa, SOCKADDR_UN_LEN(sa.un));
271f518f 3597 if (r < 0)
44ed5214 3598 return log_error_errno(errno, "bind(" NSPAWN_NOTIFY_SOCKET_PATH ") failed: %m");
9c1e04d0 3599
adc7d9f0 3600 r = userns_lchown(NSPAWN_NOTIFY_SOCKET_PATH, 0, 0);
271f518f 3601 if (r < 0)
adc7d9f0 3602 return log_error_errno(r, "Failed to chown " NSPAWN_NOTIFY_SOCKET_PATH ": %m");
adc7d9f0 3603
2ff48e98 3604 r = setsockopt_int(fd, SOL_SOCKET, SO_PASSCRED, true);
271f518f 3605 if (r < 0)
2ff48e98 3606 return log_error_errno(r, "SO_PASSCRED failed: %m");
9c1e04d0 3607
271f518f 3608 return TAKE_FD(fd);
9c1e04d0
AP		 3609}
3610
03cfe0d5
LP		 3611static int outer_child(
3612 Barrier *barrier,
3613 const char *directory,
2d845785 3614 DissectedImage *dissected_image,
af06cd30 3615 int fd_outer_socket,
5d9d3fcb 3616 int fd_inner_socket,
d7bea6b6
DP		 3617 FDSet *fds,
3618 int netns_fd) {
03cfe0d5 3619
2f893044 3620 _cleanup_(bind_user_context_freep) BindUserContext *bind_user_context = NULL;
e1bb4b0d 3621 _cleanup_strv_free_ char **os_release_pairs = NULL;
254d1313 3622 _cleanup_close_ int fd = -EBADF, mntns_fd = -EBADF;
f61c7f88 3623 bool idmap = false;
e5f10caf 3624 const char *p;
03cfe0d5
LP		 3625 pid_t pid;
3626 ssize_t l;
de40a303 3627 int r;
03cfe0d5 3628
d1d0b895
LP		 3629 /* This is the "outer" child process, i.e the one forked off by the container manager itself. It
3630 * already has its own CLONE_NEWNS namespace (which was created by the clone()). It still lives in
3631 * the host's CLONE_NEWPID, CLONE_NEWUTS, CLONE_NEWIPC, CLONE_NEWUSER and CLONE_NEWNET
3632 * namespaces. After it completed a number of initializations a second child (the "inner" one) is
3633 * forked off it, and it exits. */
b37469d7 3634
03cfe0d5
LP		 3635 assert(barrier);
3636 assert(directory);
af06cd30 3637 assert(fd_outer_socket >= 0);
5d9d3fcb 3638 assert(fd_inner_socket >= 0);
03cfe0d5 3639
de40a303
LP		 3640 log_debug("Outer child is initializing.");
3641
e1bb4b0d
LB		 3642 r = load_os_release_pairs_with_prefix("/", "container_host_", &os_release_pairs);
3643 if (r < 0)
3644 log_debug_errno(r, "Failed to read os-release from host for container, ignoring: %m");
3645
03cfe0d5
LP		 3646 if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
3647 return log_error_errno(errno, "PR_SET_PDEATHSIG failed: %m");
3648
03cfe0d5
LP		 3649 r = reset_audit_loginuid();
3650 if (r < 0)
3651 return r;
3652
2a2e78e9
LP		 3653 /* Mark everything as slave, so that we still receive mounts from the real root, but don't propagate
3654 * mounts to the real root. */
511a8cfe 3655 r = mount_follow_verbose(LOG_ERR, NULL, "/", NULL, MS_SLAVE|MS_REC, NULL);
60e76d48
ZJS		 3656 if (r < 0)
3657 return r;
03cfe0d5 3658
2d845785 3659 if (dissected_image) {
d1d0b895
LP		 3660 /* If we are operating on a disk image, then mount its root directory now, but leave out the
3661 * rest. We can read the UID shift from it if we need to. Further down we'll mount the rest,
3662 * but then with the uid shift known. That way we can mount VFAT file systems shifted to the
3663 * right place right away. This makes sure ESP partitions and userns are compatible. */
2d3a5a73 3664
af187ab2 3665 r = dissected_image_mount_and_warn(
d04faa4e
LP		 3666 dissected_image,
3667 directory,
3668 arg_uid_shift,
21b61b1d 3669 arg_uid_range,
8d9a1d59 3670 /* userns_fd= */ -EBADF,
d04faa4e
LP		 3671 DISSECT_IMAGE_MOUNT_ROOT_ONLY|
3672 DISSECT_IMAGE_DISCARD_ON_LOOP|
3673 DISSECT_IMAGE_USR_NO_ROOT|
c65f854a 3674 (arg_read_only ? DISSECT_IMAGE_READ_ONLY : DISSECT_IMAGE_FSCK|DISSECT_IMAGE_GROWFS)|
af187ab2 3675 (arg_start_mode == START_BOOT ? DISSECT_IMAGE_VALIDATE_OS : 0));
2d845785 3676 if (r < 0)
af187ab2 3677 return r;
2d845785 3678 }
03cfe0d5 3679
391567f4
LP		 3680 r = determine_uid_shift(directory);
3681 if (r < 0)
3682 return r;
3683
0de7acce 3684 if (arg_userns_mode != USER_NAMESPACE_NO) {
b71a0192
CB		 3685 r = namespace_open(0, NULL, &mntns_fd, NULL, NULL, NULL);
3686 if (r < 0)
3687 return log_error_errno(r, "Failed to pin outer mount namespace: %m");
3688
af06cd30 3689 l = send_one_fd(fd_outer_socket, mntns_fd, 0);
b71a0192
CB		 3690 if (l < 0)
3691 return log_error_errno(l, "Failed to send outer mount namespace fd: %m");
3692 mntns_fd = safe_close(mntns_fd);
3693
0e7ac751 3694 /* Let the parent know which UID shift we read from the image */
af06cd30 3695 l = send(fd_outer_socket, &arg_uid_shift, sizeof(arg_uid_shift), MSG_NOSIGNAL);
825d5287
RM		 3696 if (l < 0)
3697 return log_error_errno(errno, "Failed to send UID shift: %m");
baaa35ad
ZJS		 3698 if (l != sizeof(arg_uid_shift))
3699 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3700 "Short write while sending UID shift.");
0e7ac751 3701
0de7acce 3702 if (arg_userns_mode == USER_NAMESPACE_PICK) {
d1d0b895
LP		 3703 /* When we are supposed to pick the UID shift, the parent will check now whether the
3704 * UID shift we just read from the image is available. If yes, it will send the UID
3705 * shift back to us, if not it will pick a different one, and send it back to us. */
0e7ac751 3706
af06cd30 3707 l = recv(fd_outer_socket, &arg_uid_shift, sizeof(arg_uid_shift), 0);
0e7ac751
LP		 3708 if (l < 0)
3709 return log_error_errno(errno, "Failed to recv UID shift: %m");
baaa35ad
ZJS		 3710 if (l != sizeof(arg_uid_shift))
3711 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3712 "Short read while receiving UID shift.");
0e7ac751
LP		 3713 }
3714
ff6c6cc1
LP		 3715 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
3716 "Selected user namespace base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range);
825d5287
RM		 3717 }
3718
6f83d3d1
LP		 3719 if (path_equal(directory, "/")) {
3720 /* If the directory we shall boot is the host, let's operate on a bind mount at a different
3721 * place, so that we can make changes to its mount structure (for example, to implement
3722 * --volatile=) without this interfering with our ability to access files such as
3723 * /etc/localtime to copy into the container. Note that we use a fixed place for this
6c2d70ce 3724 * (instead of a temporary directory, since we are living in our own mount namespace here
7802194a 3725 * already, and thus don't need to be afraid of colliding with anyone else's mounts). */
6f83d3d1
LP		 3726 (void) mkdir_p("/run/systemd/nspawn-root", 0755);
3727
511a8cfe 3728 r = mount_nofollow_verbose(LOG_ERR, "/", "/run/systemd/nspawn-root", NULL, MS_BIND|MS_REC, NULL);
6f83d3d1
LP		 3729 if (r < 0)
3730 return r;
3731
3732 directory = "/run/systemd/nspawn-root";
e50cd82f 3733 }
7d0ecdd6 3734
75f81732
LP		 3735 /* Make sure we always have a mount that we can move to root later on. */
3736 r = make_mount_point(directory);
3737 if (r < 0)
3738 return r;
3739
3740 /* So the whole tree is now MS_SLAVE, i.e. we'll still receive mount/umount events from the host
3741 * mount namespace. For the directory we are going to run our container let's turn this off, so that
3742 * we'll live in our own little world from now on, and propagation from the host may only happen via
3743 * the mount tunnel dir, or not at all. */
3744 r = mount_follow_verbose(LOG_ERR, NULL, directory, NULL, MS_PRIVATE|MS_REC, NULL);
3745 if (r < 0)
3746 return r;
3747
7d0ecdd6
LP		 3748 r = setup_pivot_root(
3749 directory,
3750 arg_pivot_root_new,
3751 arg_pivot_root_old);
3752 if (r < 0)
3753 return r;
3754
3755 r = setup_volatile_mode(
3756 directory,
3757 arg_volatile_mode,
7d0ecdd6 3758 arg_uid_shift,
8f1ed04a 3759 arg_selinux_apifs_context);
7d0ecdd6
LP		 3760 if (r < 0)
3761 return r;
3762
2f893044
LP		 3763 r = bind_user_prepare(
3764 directory,
3765 arg_bind_user,
3766 arg_uid_shift,
3767 arg_uid_range,
3768 &arg_custom_mounts, &arg_n_custom_mounts,
3769 &bind_user_context);
3770 if (r < 0)
3771 return r;
3772
3773 if (arg_userns_mode != USER_NAMESPACE_NO && bind_user_context) {
d1d0b895
LP		 3774 /* Send the user maps we determined to the parent, so that it installs it in our user
3775 * namespace UID map table */
2f893044
LP		 3776
3777 for (size_t i = 0; i < bind_user_context->n_data; i++) {
3778 uid_t map[] = {
3779 bind_user_context->data[i].payload_user->uid,
3780 bind_user_context->data[i].host_user->uid,
3781 (uid_t) bind_user_context->data[i].payload_group->gid,
3782 (uid_t) bind_user_context->data[i].host_group->gid,
3783 };
3784
af06cd30 3785 l = send(fd_outer_socket, map, sizeof(map), MSG_NOSIGNAL);
2f893044
LP		 3786 if (l < 0)
3787 return log_error_errno(errno, "Failed to send user UID map: %m");
3788 if (l != sizeof(map))
3789 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3790 "Short write while sending user UID map.");
3791 }
3792 }
3793
5f0a6347
DDM		 3794 r = mount_custom(
3795 directory,
3796 arg_custom_mounts,
3797 arg_n_custom_mounts,
5f0a6347 3798 arg_uid_shift,
c0c8f718 3799 arg_uid_range,
5f0a6347
DDM		 3800 arg_selinux_apifs_context,
3801 MOUNT_ROOT_ONLY);
3802 if (r < 0)
3803 return r;
3804
c0c8f718
AV		 3805 if (arg_userns_mode != USER_NAMESPACE_NO &&
3806 IN_SET(arg_userns_ownership, USER_NAMESPACE_OWNERSHIP_MAP, USER_NAMESPACE_OWNERSHIP_AUTO) &&
3807 arg_uid_shift != 0) {
dba4fa89
LP		 3808 _cleanup_free_ char *usr_subtree = NULL;
3809 char *dirs[3];
3810 size_t i = 0;
c0c8f718 3811
dba4fa89
LP		 3812 dirs[i++] = (char*) directory;
3813
3814 if (dissected_image && dissected_image->partitions[PARTITION_USR].found) {
3815 usr_subtree = path_join(directory, "/usr");
3816 if (!usr_subtree)
3817 return log_oom();
3818
3819 dirs[i++] = usr_subtree;
3820 }
3821
3822 dirs[i] = NULL;
3823
3824 r = remount_idmap(dirs, arg_uid_shift, arg_uid_range, UID_INVALID, REMOUNT_IDMAPPING_HOST_ROOT);
bb44fd07
ZJS		 3825 if (r == -EINVAL || ERRNO_IS_NEG_NOT_SUPPORTED(r)) {
3826 /* This might fail because the kernel or file system doesn't support idmapping. We
3827 * can't really distinguish this nicely, nor do we have any guarantees about the
3828 * error codes we see, could be EOPNOTSUPP or EINVAL. */
3829 if (arg_userns_ownership != USER_NAMESPACE_OWNERSHIP_AUTO)
3830 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
3831 "ID mapped mounts are apparently not available, sorry.");
3832
3833 log_debug("ID mapped mounts are apparently not available on this kernel or for the selected file system, reverting to recursive chown()ing.");
3834 arg_userns_ownership = USER_NAMESPACE_OWNERSHIP_CHOWN;
3835 } else if (r < 0)
3836 return log_error_errno(r, "Failed to set up ID mapped mounts: %m");
3837 else {
c0c8f718
AV		 3838 log_debug("ID mapped mounts available, making use of them.");
3839 idmap = true;
3840 }
3841 }
3842
2d3a5a73
LP		 3843 if (dissected_image) {
3844 /* Now we know the uid shift, let's now mount everything else that might be in the image. */
d04faa4e
LP		 3845 r = dissected_image_mount(
3846 dissected_image,
3847 directory,
3848 arg_uid_shift,
21b61b1d 3849 arg_uid_range,
8d9a1d59 3850 /* userns_fd= */ -EBADF,
d04faa4e
LP		 3851 DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY|
3852 DISSECT_IMAGE_DISCARD_ON_LOOP|
3853 DISSECT_IMAGE_USR_NO_ROOT|
f61c7f88
LP		 3854 (arg_read_only ? DISSECT_IMAGE_READ_ONLY : DISSECT_IMAGE_FSCK|DISSECT_IMAGE_GROWFS)|
3855 (idmap ? DISSECT_IMAGE_MOUNT_IDMAPPED : 0));
4fcb96ce
LP		 3856 if (r == -EUCLEAN)
3857 return log_error_errno(r, "File system check for image failed: %m");
2d3a5a73 3858 if (r < 0)
4fcb96ce 3859 return log_error_errno(r, "Failed to mount image file system: %m");
2d3a5a73
LP		 3860 }
3861
8199d554
LP		 3862 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
3863 /* OK, we don't know yet which cgroup mode to use yet. Let's figure it out, and tell the parent. */
3864
3865 r = detect_unified_cgroup_hierarchy_from_image(directory);
3866 if (r < 0)
3867 return r;
3868
fefb7a6d 3869 l = send(fd_outer_socket, &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), MSG_NOSIGNAL);
8199d554
LP		 3870 if (l < 0)
3871 return log_error_errno(errno, "Failed to send cgroup mode: %m");
baaa35ad
ZJS		 3872 if (l != sizeof(arg_unified_cgroup_hierarchy))
3873 return log_error_errno(SYNTHETIC_ERRNO(EIO),
3874 "Short write while sending cgroup mode.");
8199d554
LP		 3875 }
3876
4ad14eff
LP		 3877 r = recursive_chown(directory, arg_uid_shift, arg_uid_range);
3878 if (r < 0)
3879 return r;
3880
03cfe0d5
LP		 3881 r = base_filesystem_create(directory, arg_uid_shift, (gid_t) arg_uid_shift);
3882 if (r < 0)
3883 return r;
3884
bbd407ea
DDM		 3885 if (arg_read_only && arg_volatile_mode == VOLATILE_NO &&
3886 !has_custom_root_mount(arg_custom_mounts, arg_n_custom_mounts)) {
64e82c19 3887 r = bind_remount_recursive(directory, MS_RDONLY, MS_RDONLY, NULL);
03cfe0d5
LP		 3888 if (r < 0)
3889 return log_error_errno(r, "Failed to make tree read-only: %m");
3890 }
3891
0de7acce 3892 r = mount_all(directory,
4f086aab 3893 arg_mount_settings,
0de7acce 3894 arg_uid_shift,
0de7acce 3895 arg_selinux_apifs_context);
03cfe0d5
LP		 3896 if (r < 0)
3897 return r;
3898
07fa00f9
LP		 3899 r = copy_devnodes(directory);
3900 if (r < 0)
03cfe0d5
LP		 3901 return r;
3902
de40a303
LP		 3903 r = make_extra_nodes(directory);
3904 if (r < 0)
3905 return r;
3906
3907 (void) dev_setup(directory, arg_uid_shift, arg_uid_shift);
e5f10caf 3908
9fac5029 3909 p = prefix_roota(directory, "/run/host");
e5f10caf 3910 (void) make_inaccessible_nodes(p, arg_uid_shift, arg_uid_shift);
03cfe0d5 3911
07fa00f9
LP		 3912 r = setup_pts(directory);
3913 if (r < 0)
03cfe0d5
LP		 3914 return r;
3915
e79581dd 3916 r = mount_tunnel_dig(directory);
03cfe0d5
LP		 3917 if (r < 0)
3918 return r;
3919
8e5430c4
LP		 3920 r = setup_keyring();
3921 if (r < 0)
3922 return r;
3923
3652872a
LP		 3924 r = setup_credentials(directory);
3925 if (r < 0)
3926 return r;
3927
2f893044
LP		 3928 r = bind_user_setup(bind_user_context, directory);
3929 if (r < 0)
3930 return r;
3931
5c4deb9a
MJ		 3932 r = mount_custom(
3933 directory,
3934 arg_custom_mounts,
3935 arg_n_custom_mounts,
3936 arg_uid_shift,
c0c8f718 3937 arg_uid_range,
5c4deb9a
MJ		 3938 arg_selinux_apifs_context,
3939 MOUNT_NON_ROOT_ONLY);
3940 if (r < 0)
3941 return r;
3942
03cfe0d5
LP		 3943 r = setup_timezone(directory);
3944 if (r < 0)
3945 return r;
3946
3947 r = setup_resolv_conf(directory);
3948 if (r < 0)
3949 return r;
3950
e01ff70a
MS		 3951 r = setup_machine_id(directory);
3952 if (r < 0)
3953 return r;
3954
03cfe0d5
LP		 3955 r = setup_journal(directory);
3956 if (r < 0)
3957 return r;
3958
0f48ba7b
LP		 3959 /* The same stuff as the $container env var, but nicely readable for the entire payload */
3960 p = prefix_roota(directory, "/run/host/container-manager");
05794f5c 3961 (void) write_string_file(p, arg_container_service_name, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_MODE_0444);
0f48ba7b
LP		 3962
3963 /* The same stuff as the $container_uuid env var */
3964 p = prefix_roota(directory, "/run/host/container-uuid");
05794f5c 3965 (void) write_string_filef(p, WRITE_STRING_FILE_CREATE|WRITE_STRING_FILE_MODE_0444, SD_ID128_UUID_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid));
0f48ba7b 3966
489fae52 3967 if (!arg_use_cgns) {
0996ef00
CB		 3968 r = mount_cgroups(
3969 directory,
3970 arg_unified_cgroup_hierarchy,
3971 arg_userns_mode != USER_NAMESPACE_NO,
3972 arg_uid_shift,
3973 arg_uid_range,
5a8ff0e6 3974 arg_selinux_apifs_context,
ada54120 3975 false);
0996ef00
CB		 3976 if (r < 0)
3977 return r;
3978 }
03cfe0d5 3979
57c10a56
CB		 3980 /* Mark everything as shared so our mounts get propagated down. This is required to make new bind
3981 * mounts available in systemd services inside the container that create a new mount namespace. See
3982 * https://github.com/systemd/systemd/issues/3860 Further submounts (such as /dev) done after this
3983 * will inherit the shared propagation mode.
3984 *
3985 * IMPORTANT: Do not overmount the root directory anymore from now on to enable moving the root
3986 * directory mount to root later on.
3987 * https://github.com/systemd/systemd/issues/3847#issuecomment-562735251
3988 */
9d50f850 3989 r = mount_switch_root(directory, MS_SHARED);
03cfe0d5
LP		 3990 if (r < 0)
3991 return log_error_errno(r, "Failed to move root directory: %m");
3992
e79581dd
CB		 3993 /* We finished setting up the rootfs which is a shared mount. The mount tunnel needs to be a
3994 * dependent mount otherwise we can't MS_MOVE mounts that were propagated from the host into
3995 * the container. */
3996 r = mount_tunnel_open();
3997 if (r < 0)
3998 return r;
3999
b71a0192
CB		 4000 if (arg_userns_mode != USER_NAMESPACE_NO) {
4001 /* In order to mount procfs and sysfs in an unprivileged container the kernel
4002 * requires that a fully visible instance is already present in the target mount
4003 * namespace. Mount one here so the inner child can mount its own instances. Later
4004 * we umount the temporary instances created here before we actually exec the
4005 * payload. Since the rootfs is shared the umount will propagate into the container.
4006 * Note, the inner child wouldn't be able to unmount the instances on its own since
4007 * it doesn't own the originating mount namespace. IOW, the outer child needs to do
4008 * this. */
4009 r = pin_fully_visible_fs();
4010 if (r < 0)
4011 return r;
4012 }
4013
e96ceaba 4014 fd = setup_notify_child();
9c1e04d0
AP		 4015 if (fd < 0)
4016 return fd;
4017
03cfe0d5 4018 pid = raw_clone(SIGCHLD|CLONE_NEWNS|
0c582db0 4019 arg_clone_ns_flags |
8869a0b4 4020 (arg_userns_mode != USER_NAMESPACE_NO ? CLONE_NEWUSER : 0));
03cfe0d5
LP		 4021 if (pid < 0)
4022 return log_error_errno(errno, "Failed to fork inner child: %m");
03cfe0d5 4023 if (pid == 0) {
af06cd30 4024 fd_outer_socket = safe_close(fd_outer_socket);
03cfe0d5 4025
2a2e78e9
LP		 4026 /* The inner child has all namespaces that are requested, so that we all are owned by the
4027 * user if user namespaces are turned on. */
03cfe0d5 4028
d7bea6b6
DP		 4029 if (arg_network_namespace_path) {
4030 r = namespace_enter(-1, -1, netns_fd, -1, -1);
4031 if (r < 0)
e2d39e54 4032 return log_error_errno(r, "Failed to join network namespace: %m");
d7bea6b6
DP		 4033 }
4034
11875a98 4035 r = inner_child(barrier, fd_inner_socket, fds, os_release_pairs);
03cfe0d5
LP		 4036 if (r < 0)
4037 _exit(EXIT_FAILURE);
4038
4039 _exit(EXIT_SUCCESS);
4040 }
4041
af06cd30 4042 l = send(fd_outer_socket, &pid, sizeof(pid), MSG_NOSIGNAL);
03cfe0d5
LP		 4043 if (l < 0)
4044 return log_error_errno(errno, "Failed to send PID: %m");
baaa35ad
ZJS		 4045 if (l != sizeof(pid))
4046 return log_error_errno(SYNTHETIC_ERRNO(EIO),
4047 "Short write while sending PID.");
03cfe0d5 4048
af06cd30 4049 l = send(fd_outer_socket, &arg_uuid, sizeof(arg_uuid), MSG_NOSIGNAL);
e01ff70a
MS		 4050 if (l < 0)
4051 return log_error_errno(errno, "Failed to send machine ID: %m");
baaa35ad
ZJS		 4052 if (l != sizeof(arg_uuid))
4053 return log_error_errno(SYNTHETIC_ERRNO(EIO),
4054 "Short write while sending machine ID.");
e01ff70a 4055
af06cd30 4056 l = send_one_fd(fd_outer_socket, fd, 0);
9c1e04d0 4057 if (l < 0)
ba72801d 4058 return log_error_errno(l, "Failed to send notify fd: %m");
9c1e04d0 4059
af06cd30 4060 fd_outer_socket = safe_close(fd_outer_socket);
5d9d3fcb 4061 fd_inner_socket = safe_close(fd_inner_socket);
d7bea6b6 4062 netns_fd = safe_close(netns_fd);
03cfe0d5
LP		 4063
4064 return 0;
4065}
4066
0e7ac751 4067static int uid_shift_pick(uid_t *shift, LockFile *ret_lock_file) {
d381c8a6 4068 bool tried_hashed = false;
0e7ac751
LP		 4069 unsigned n_tries = 100;
4070 uid_t candidate;
4071 int r;
4072
4073 assert(shift);
4074 assert(ret_lock_file);
0de7acce 4075 assert(arg_userns_mode == USER_NAMESPACE_PICK);
0e7ac751
LP		 4076 assert(arg_uid_range == 0x10000U);
4077
4078 candidate = *shift;
4079
4080 (void) mkdir("/run/systemd/nspawn-uid", 0755);
4081
4082 for (;;) {
fbd0b64f 4083 char lock_path[STRLEN("/run/systemd/nspawn-uid/") + DECIMAL_STR_MAX(uid_t) + 1];
8e766630 4084 _cleanup_(release_lock_file) LockFile lf = LOCK_FILE_INIT;
0e7ac751
LP		 4085
4086 if (--n_tries <= 0)
4087 return -EBUSY;
4088
87d5e4f2 4089 if (candidate < CONTAINER_UID_BASE_MIN || candidate > CONTAINER_UID_BASE_MAX)
0e7ac751
LP		 4090 goto next;
4091 if ((candidate & UINT32_C(0xFFFF)) != 0)
4092 goto next;
4093
4094 xsprintf(lock_path, "/run/systemd/nspawn-uid/" UID_FMT, candidate);
4095 r = make_lock_file(lock_path, LOCK_EX|LOCK_NB, &lf);
4096 if (r == -EBUSY) /* Range already taken by another nspawn instance */
4097 goto next;
4098 if (r < 0)
4099 return r;
4100
4101 /* Make some superficial checks whether the range is currently known in the user database */
4102 if (getpwuid(candidate))
4103 goto next;
4104 if (getpwuid(candidate + UINT32_C(0xFFFE)))
4105 goto next;
4106 if (getgrgid(candidate))
4107 goto next;
4108 if (getgrgid(candidate + UINT32_C(0xFFFE)))
4109 goto next;
4110
4111 *ret_lock_file = lf;
4112 lf = (struct LockFile) LOCK_FILE_INIT;
4113 *shift = candidate;
4114 return 0;
4115
4116 next:
d381c8a6
LP		 4117 if (arg_machine && !tried_hashed) {
4118 /* Try to hash the base from the container name */
4119
4120 static const uint8_t hash_key[] = {
4121 0xe1, 0x56, 0xe0, 0xf0, 0x4a, 0xf0, 0x41, 0xaf,
4122 0x96, 0x41, 0xcf, 0x41, 0x33, 0x94, 0xff, 0x72
4123 };
4124
4125 candidate = (uid_t) siphash24(arg_machine, strlen(arg_machine), hash_key);
4126
4127 tried_hashed = true;
4128 } else
4129 random_bytes(&candidate, sizeof(candidate));
4130
87d5e4f2 4131 candidate = (candidate % (CONTAINER_UID_BASE_MAX - CONTAINER_UID_BASE_MIN)) + CONTAINER_UID_BASE_MIN;
0e7ac751
LP		 4132 candidate &= (uid_t) UINT32_C(0xFFFF0000);
4133 }
4134}
4135
2f893044
LP		 4136static int add_one_uid_map(
4137 char **p,
4138 uid_t container_uid,
4139 uid_t host_uid,
4140 uid_t range) {
4141
4142 return strextendf(p,
4143 UID_FMT " " UID_FMT " " UID_FMT "\n",
4144 container_uid, host_uid, range);
4145}
4146
4147static int make_uid_map_string(
4148 const uid_t bind_user_uid[],
4149 size_t n_bind_user_uid,
4150 size_t offset,
4151 char **ret) {
4152
4153 _cleanup_free_ char *s = NULL;
4154 uid_t previous_uid = 0;
4155 int r;
4156
4157 assert(n_bind_user_uid == 0 || bind_user_uid);
2f092762 4158 assert(IN_SET(offset, 0, 2)); /* used to switch between UID and GID map */
2f893044
LP		 4159 assert(ret);
4160
4161 /* The bind_user_uid[] array is a series of 4 uid_t values, for each --bind-user= entry one
4162 * quadruplet, consisting of host and container UID + GID. */
4163
4164 for (size_t i = 0; i < n_bind_user_uid; i++) {
05ab439a
YW		 4165 uid_t payload_uid = bind_user_uid[i*4+offset],
4166 host_uid = bind_user_uid[i*4+offset+1];
2f893044
LP		 4167
4168 assert(previous_uid <= payload_uid);
4169 assert(payload_uid < arg_uid_range);
4170
4171 /* Add a range to close the gap to previous entry */
4172 if (payload_uid > previous_uid) {
4173 r = add_one_uid_map(&s, previous_uid, arg_uid_shift + previous_uid, payload_uid - previous_uid);
4174 if (r < 0)
4175 return r;
4176 }
4177
4178 /* Map this specific user */
4179 r = add_one_uid_map(&s, payload_uid, host_uid, 1);
4180 if (r < 0)
4181 return r;
4182
4183 previous_uid = payload_uid + 1;
4184 }
4185
4186 /* And add a range to close the gap to finish the range */
4187 if (arg_uid_range > previous_uid) {
4188 r = add_one_uid_map(&s, previous_uid, arg_uid_shift + previous_uid, arg_uid_range - previous_uid);
4189 if (r < 0)
4190 return r;
4191 }
4192
4193 assert(s);
4194
4195 *ret = TAKE_PTR(s);
4196 return 0;
4197}
4198
4199static int setup_uid_map(
4200 pid_t pid,
4201 const uid_t bind_user_uid[],
4202 size_t n_bind_user_uid) {
4203
4204 char uid_map[STRLEN("/proc//uid_map") + DECIMAL_STR_MAX(uid_t) + 1];
4205 _cleanup_free_ char *s = NULL;
03cfe0d5
LP		 4206 int r;
4207
4208 assert(pid > 1);
4209
2f893044
LP		 4210 /* Build the UID map string */
4211 if (make_uid_map_string(bind_user_uid, n_bind_user_uid, 0, &s) < 0) /* offset=0 contains the UID pair */
4212 return log_oom();
4213
03cfe0d5 4214 xsprintf(uid_map, "/proc/" PID_FMT "/uid_map", pid);
2f893044 4215 r = write_string_file(uid_map, s, WRITE_STRING_FILE_DISABLE_BUFFER);
03cfe0d5
LP		 4216 if (r < 0)
4217 return log_error_errno(r, "Failed to write UID map: %m");
4218
2f893044
LP		 4219 /* And now build the GID map string */
4220 s = mfree(s);
4221 if (make_uid_map_string(bind_user_uid, n_bind_user_uid, 2, &s) < 0) /* offset=2 contains the GID pair */
4222 return log_oom();
4223
03cfe0d5 4224 xsprintf(uid_map, "/proc/" PID_FMT "/gid_map", pid);
2f893044 4225 r = write_string_file(uid_map, s, WRITE_STRING_FILE_DISABLE_BUFFER);
03cfe0d5
LP		 4226 if (r < 0)
4227 return log_error_errno(r, "Failed to write GID map: %m");
4228
4229 return 0;
4230}
4231
9c1e04d0 4232static int nspawn_dispatch_notify_fd(sd_event_source *source, int fd, uint32_t revents, void *userdata) {
9c1e04d0
AP		 4233 char buf[NOTIFY_BUFFER_MAX+1];
4234 char *p = NULL;
4235 struct iovec iovec = {
4236 .iov_base = buf,
4237 .iov_len = sizeof(buf)-1,
4238 };
fb29cdbe
LP		 4239 CMSG_BUFFER_TYPE(CMSG_SPACE(sizeof(struct ucred)) +
4240 CMSG_SPACE(sizeof(int) * NOTIFY_FD_MAX)) control;
9c1e04d0
AP		 4241 struct msghdr msghdr = {
4242 .msg_iov = &iovec,
4243 .msg_iovlen = 1,
4244 .msg_control = &control,
4245 .msg_controllen = sizeof(control),
4246 };
371d72e0 4247 struct ucred *ucred;
9c1e04d0
AP		 4248 ssize_t n;
4249 pid_t inner_child_pid;
4250 _cleanup_strv_free_ char **tags = NULL;
4bf4f50f 4251 int r;
9c1e04d0
AP		 4252
4253 assert(userdata);
4254
4255 inner_child_pid = PTR_TO_PID(userdata);
4256
4257 if (revents != EPOLLIN) {
4258 log_warning("Got unexpected poll event for notify fd.");
4259 return 0;
4260 }
4261
3691bcf3 4262 n = recvmsg_safe(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
bb44fd07
ZJS		 4263 if (ERRNO_IS_NEG_TRANSIENT(n))
4264 return 0;
4265 else if (n == -EXFULL) {
4266 log_warning("Got message with truncated control data (too many fds sent?), ignoring.");
4267 return 0;
4268 } else if (n < 0)
3691bcf3 4269 return log_warning_errno(n, "Couldn't read notification socket: %m");
9c1e04d0 4270
9c1e04d0
AP		 4271 cmsg_close_all(&msghdr);
4272
371d72e0 4273 ucred = CMSG_FIND_DATA(&msghdr, SOL_SOCKET, SCM_CREDENTIALS, struct ucred);
9c1e04d0 4274 if (!ucred || ucred->pid != inner_child_pid) {
8cb57430 4275 log_debug("Received notify message without valid credentials. Ignoring.");
9c1e04d0
AP		 4276 return 0;
4277 }
4278
4279 if ((size_t) n >= sizeof(buf)) {
4280 log_warning("Received notify message exceeded maximum size. Ignoring.");
4281 return 0;
4282 }
4283
4284 buf[n] = 0;
4285 tags = strv_split(buf, "\n\r");
4286 if (!tags)
4287 return log_oom();
4288
d29cc4d6 4289 if (strv_contains(tags, "READY=1")) {
d4341b76 4290 r = sd_notify(false, "READY=1\n");
4bf4f50f
ZJS		 4291 if (r < 0)
4292 log_warning_errno(r, "Failed to send readiness notification, ignoring: %m");
4293 }
9c1e04d0
AP		 4294
4295 p = strv_find_startswith(tags, "STATUS=");
4296 if (p)
04f590a4 4297 (void) sd_notifyf(false, "STATUS=Container running: %s", p);
9c1e04d0
AP		 4298
4299 return 0;
4300}
4301
e96ceaba 4302static int setup_notify_parent(sd_event *event, int fd, pid_t *inner_child_pid, sd_event_source **notify_event_source) {
9c1e04d0 4303 int r;
9c1e04d0 4304
5773024d 4305 r = sd_event_add_io(event, notify_event_source, fd, EPOLLIN, nspawn_dispatch_notify_fd, inner_child_pid);
9c1e04d0
AP		 4306 if (r < 0)
4307 return log_error_errno(r, "Failed to allocate notify event source: %m");
4308
5773024d 4309 (void) sd_event_source_set_description(*notify_event_source, "nspawn-notify");
9c1e04d0
AP		 4310
4311 return 0;
4312}
4313
5d961407
LP		 4314static int merge_settings(Settings *settings, const char *path) {
4315 int rl;
f757855e 4316
5d961407
LP		 4317 assert(settings);
4318 assert(path);
f757855e 4319
5d961407
LP		 4320 /* Copy over bits from the settings, unless they have been explicitly masked by command line switches. Note
4321 * that this steals the fields of the Settings* structure, and hence modifies it. */
f757855e 4322
7732f92b
LP		 4323 if ((arg_settings_mask & SETTING_START_MODE) == 0 &&
4324 settings->start_mode >= 0) {
4325 arg_start_mode = settings->start_mode;
130d3d22 4326 strv_free_and_replace(arg_parameters, settings->parameters);
f757855e
LP		 4327 }
4328
d3689b94
LP		 4329 if ((arg_settings_mask & SETTING_EPHEMERAL) == 0 &&
4330 settings->ephemeral >= 0)
a2f577fc
JL		 4331 arg_ephemeral = settings->ephemeral;
4332
de40a303
LP		 4333 if ((arg_settings_mask & SETTING_DIRECTORY) == 0 &&
4334 settings->root) {
4335
4336 if (!arg_settings_trusted)
4337 log_warning("Ignoring root directory setting, file %s is not trusted.", path);
4338 else
4339 free_and_replace(arg_directory, settings->root);
4340 }
4341
b53ede69
PW		 4342 if ((arg_settings_mask & SETTING_PIVOT_ROOT) == 0 &&
4343 settings->pivot_root_new) {
4344 free_and_replace(arg_pivot_root_new, settings->pivot_root_new);
4345 free_and_replace(arg_pivot_root_old, settings->pivot_root_old);
4346 }
4347
5f932eb9 4348 if ((arg_settings_mask & SETTING_WORKING_DIRECTORY) == 0 &&
1cc6c93a
YW		 4349 settings->working_directory)
4350 free_and_replace(arg_chdir, settings->working_directory);
5f932eb9 4351
f757855e 4352 if ((arg_settings_mask & SETTING_ENVIRONMENT) == 0 &&
130d3d22
YW		 4353 settings->environment)
4354 strv_free_and_replace(arg_setenv, settings->environment);
f757855e 4355
de40a303
LP		 4356 if ((arg_settings_mask & SETTING_USER) == 0) {
4357
4358 if (settings->user)
4359 free_and_replace(arg_user, settings->user);
4360
4361 if (uid_is_valid(settings->uid))
4362 arg_uid = settings->uid;
4363 if (gid_is_valid(settings->gid))
4364 arg_gid = settings->gid;
4365 if (settings->n_supplementary_gids > 0) {
4366 free_and_replace(arg_supplementary_gids, settings->supplementary_gids);
4367 arg_n_supplementary_gids = settings->n_supplementary_gids;
4368 }
4369 }
f757855e
LP		 4370
4371 if ((arg_settings_mask & SETTING_CAPABILITY) == 0) {
a3fc6b55 4372 uint64_t plus, minus;
7be830c6 4373 uint64_t network_minus = 0;
88fc9c9b 4374 uint64_t ambient;
f757855e 4375
de40a303
LP		 4376 /* Note that we copy both the simple plus/minus caps here, and the full quintet from the
4377 * Settings structure */
4378
0e265674 4379 plus = settings->capability;
a3fc6b55
LP		 4380 minus = settings->drop_capability;
4381
9baa294c
LP		 4382 if ((arg_settings_mask & SETTING_NETWORK) == 0 &&
4383 settings_network_configured(settings)) {
a3fc6b55
LP		 4384 if (settings_private_network(settings))
4385 plus |= UINT64_C(1) << CAP_NET_ADMIN;
4386 else
7be830c6 4387 network_minus |= UINT64_C(1) << CAP_NET_ADMIN;
a3fc6b55 4388 }
0e265674
LP		 4389
4390 if (!arg_settings_trusted && plus != 0) {
4391 if (settings->capability != 0)
5d961407 4392 log_warning("Ignoring Capability= setting, file %s is not trusted.", path);
7be830c6
TH		 4393 } else {
4394 arg_caps_retain &= ~network_minus;
520e0d54 4395 arg_caps_retain |= plus;
7be830c6 4396 }
f757855e 4397
a3fc6b55 4398 arg_caps_retain &= ~minus;
de40a303
LP		 4399
4400 /* Copy the full capabilities over too */
4401 if (capability_quintet_is_set(&settings->full_capabilities)) {
4402 if (!arg_settings_trusted)
5238e957 4403 log_warning("Ignoring capability settings, file %s is not trusted.", path);
de40a303
LP		 4404 else
4405 arg_full_capabilities = settings->full_capabilities;
4406 }
88fc9c9b
TH		 4407
4408 ambient = settings->ambient_capability;
4409 if (!arg_settings_trusted && ambient != 0)
4410 log_warning("Ignoring AmbientCapability= setting, file %s is not trusted.", path);
4411 else
4412 arg_caps_ambient |= ambient;
f757855e
LP		 4413 }
4414
4415 if ((arg_settings_mask & SETTING_KILL_SIGNAL) == 0 &&
4416 settings->kill_signal > 0)
4417 arg_kill_signal = settings->kill_signal;
4418
4419 if ((arg_settings_mask & SETTING_PERSONALITY) == 0 &&
4420 settings->personality != PERSONALITY_INVALID)
4421 arg_personality = settings->personality;
4422
4423 if ((arg_settings_mask & SETTING_MACHINE_ID) == 0 &&
4424 !sd_id128_is_null(settings->machine_id)) {
4425
4426 if (!arg_settings_trusted)
5d961407 4427 log_warning("Ignoring MachineID= setting, file %s is not trusted.", path);
f757855e
LP		 4428 else
4429 arg_uuid = settings->machine_id;
4430 }
4431
4432 if ((arg_settings_mask & SETTING_READ_ONLY) == 0 &&
4433 settings->read_only >= 0)
4434 arg_read_only = settings->read_only;
4435
4436 if ((arg_settings_mask & SETTING_VOLATILE_MODE) == 0 &&
4437 settings->volatile_mode != _VOLATILE_MODE_INVALID)
4438 arg_volatile_mode = settings->volatile_mode;
4439
4440 if ((arg_settings_mask & SETTING_CUSTOM_MOUNTS) == 0 &&
4441 settings->n_custom_mounts > 0) {
4442
4443 if (!arg_settings_trusted)
5d961407 4444 log_warning("Ignoring TemporaryFileSystem=, Bind= and BindReadOnly= settings, file %s is not trusted.", path);
f757855e
LP		 4445 else {
4446 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
1cc6c93a 4447 arg_custom_mounts = TAKE_PTR(settings->custom_mounts);
f757855e 4448 arg_n_custom_mounts = settings->n_custom_mounts;
f757855e
LP		 4449 settings->n_custom_mounts = 0;
4450 }
4451 }
4452
4453 if ((arg_settings_mask & SETTING_NETWORK) == 0 &&
a1dfd585 4454 settings_network_configured(settings)) {
f757855e
LP		 4455
4456 if (!arg_settings_trusted)
5d961407 4457 log_warning("Ignoring network settings, file %s is not trusted.", path);
f757855e 4458 else {
f6d6bad1 4459 arg_network_veth = settings_network_veth(settings);
0e265674
LP		 4460 arg_private_network = settings_private_network(settings);
4461
130d3d22
YW		 4462 strv_free_and_replace(arg_network_interfaces, settings->network_interfaces);
4463 strv_free_and_replace(arg_network_macvlan, settings->network_macvlan);
4464 strv_free_and_replace(arg_network_ipvlan, settings->network_ipvlan);
4465 strv_free_and_replace(arg_network_veth_extra, settings->network_veth_extra);
f6d6bad1 4466
1cc6c93a
YW		 4467 free_and_replace(arg_network_bridge, settings->network_bridge);
4468 free_and_replace(arg_network_zone, settings->network_zone);
de40a303
LP		 4469
4470 free_and_replace(arg_network_namespace_path, settings->network_namespace_path);
f757855e
LP		 4471 }
4472 }
4473
4474 if ((arg_settings_mask & SETTING_EXPOSE_PORTS) == 0 &&
4475 settings->expose_ports) {
4476
4477 if (!arg_settings_trusted)
5d961407 4478 log_warning("Ignoring Port= setting, file %s is not trusted.", path);
f757855e
LP		 4479 else {
4480 expose_port_free_all(arg_expose_ports);
1cc6c93a 4481 arg_expose_ports = TAKE_PTR(settings->expose_ports);
f757855e
LP		 4482 }
4483 }
4484
0de7acce
LP		 4485 if ((arg_settings_mask & SETTING_USERNS) == 0 &&
4486 settings->userns_mode != _USER_NAMESPACE_MODE_INVALID) {
4487
4488 if (!arg_settings_trusted)
5d961407 4489 log_warning("Ignoring PrivateUsers= and PrivateUsersChown= settings, file %s is not trusted.", path);
0de7acce
LP		 4490 else {
4491 arg_userns_mode = settings->userns_mode;
4492 arg_uid_shift = settings->uid_shift;
4493 arg_uid_range = settings->uid_range;
6c045a99 4494 arg_userns_ownership = settings->userns_ownership;
0de7acce
LP		 4495 }
4496 }
4497
0cc3c9f9
LP		 4498 if ((arg_settings_mask & SETTING_BIND_USER) == 0 &&
4499 !strv_isempty(settings->bind_user))
2f893044
LP		 4500 strv_free_and_replace(arg_bind_user, settings->bind_user);
4501
d3689b94
LP		 4502 if ((arg_settings_mask & SETTING_NOTIFY_READY) == 0 &&
4503 settings->notify_ready >= 0)
9c1e04d0
AP		 4504 arg_notify_ready = settings->notify_ready;
4505
960e4569
LP		 4506 if ((arg_settings_mask & SETTING_SYSCALL_FILTER) == 0) {
4507
2d09ea44
LP		 4508 if (!strv_isempty(settings->syscall_allow_list) || !strv_isempty(settings->syscall_deny_list)) {
4509 if (!arg_settings_trusted && !strv_isempty(settings->syscall_allow_list))
4510 log_warning("Ignoring SystemCallFilter= settings, file %s is not trusted.", path);
4511 else {
4512 strv_free_and_replace(arg_syscall_allow_list, settings->syscall_allow_list);
4513 strv_free_and_replace(arg_syscall_deny_list, settings->syscall_deny_list);
4514 }
960e4569 4515 }
de40a303
LP		 4516
4517#if HAVE_SECCOMP
2d09ea44
LP		 4518 if (settings->seccomp) {
4519 if (!arg_settings_trusted)
4520 log_warning("Ignoring SECCOMP filter, file %s is not trusted.", path);
4521 else {
4522 seccomp_release(arg_seccomp);
4523 arg_seccomp = TAKE_PTR(settings->seccomp);
4524 }
de40a303
LP		 4525 }
4526#endif
960e4569
LP		 4527 }
4528
b3a9d980 4529 for (rl = 0; rl < _RLIMIT_MAX; rl++) {
bf428efb
LP		 4530 if ((arg_settings_mask & (SETTING_RLIMIT_FIRST << rl)))
4531 continue;
4532
4533 if (!settings->rlimit[rl])
4534 continue;
4535
4536 if (!arg_settings_trusted) {
5d961407 4537 log_warning("Ignoring Limit%s= setting, file '%s' is not trusted.", rlimit_to_string(rl), path);
bf428efb
LP		 4538 continue;
4539 }
4540
4541 free_and_replace(arg_rlimit[rl], settings->rlimit[rl]);
4542 }
4543
3a9530e5
LP		 4544 if ((arg_settings_mask & SETTING_HOSTNAME) == 0 &&
4545 settings->hostname)
4546 free_and_replace(arg_hostname, settings->hostname);
4547
66edd963
LP		 4548 if ((arg_settings_mask & SETTING_NO_NEW_PRIVILEGES) == 0 &&
4549 settings->no_new_privileges >= 0)
4550 arg_no_new_privileges = settings->no_new_privileges;
4551
81f345df
LP		 4552 if ((arg_settings_mask & SETTING_OOM_SCORE_ADJUST) == 0 &&
4553 settings->oom_score_adjust_set) {
4554
4555 if (!arg_settings_trusted)
5d961407 4556 log_warning("Ignoring OOMScoreAdjust= setting, file '%s' is not trusted.", path);
81f345df
LP		 4557 else {
4558 arg_oom_score_adjust = settings->oom_score_adjust;
4559 arg_oom_score_adjust_set = true;
4560 }
4561 }
4562
d107bb7d 4563 if ((arg_settings_mask & SETTING_CPU_AFFINITY) == 0 &&
0985c7c4 4564 settings->cpu_set.set) {
d107bb7d
LP		 4565
4566 if (!arg_settings_trusted)
5d961407 4567 log_warning("Ignoring CPUAffinity= setting, file '%s' is not trusted.", path);
d107bb7d 4568 else {
0985c7c4 4569 cpu_set_reset(&arg_cpu_set);
088d71f8 4570 arg_cpu_set = TAKE_STRUCT(settings->cpu_set);
d107bb7d
LP		 4571 }
4572 }
4573
09d423e9
LP		 4574 if ((arg_settings_mask & SETTING_RESOLV_CONF) == 0 &&
4575 settings->resolv_conf != _RESOLV_CONF_MODE_INVALID)
4576 arg_resolv_conf = settings->resolv_conf;
4577
4e1d6aa9
LP		 4578 if ((arg_settings_mask & SETTING_LINK_JOURNAL) == 0 &&
4579 settings->link_journal != _LINK_JOURNAL_INVALID) {
4580
4581 if (!arg_settings_trusted)
4582 log_warning("Ignoring journal link setting, file '%s' is not trusted.", path);
4583 else {
4584 arg_link_journal = settings->link_journal;
4585 arg_link_journal_try = settings->link_journal_try;
4586 }
4587 }
4588
1688841f
LP		 4589 if ((arg_settings_mask & SETTING_TIMEZONE) == 0 &&
4590 settings->timezone != _TIMEZONE_MODE_INVALID)
4591 arg_timezone = settings->timezone;
4592
de40a303
LP		 4593 if ((arg_settings_mask & SETTING_SLICE) == 0 &&
4594 settings->slice) {
4595
4596 if (!arg_settings_trusted)
4597 log_warning("Ignoring slice setting, file '%s' is not trusted.", path);
4598 else
4599 free_and_replace(arg_slice, settings->slice);
4600 }
4601
4602 if ((arg_settings_mask & SETTING_USE_CGNS) == 0 &&
4603 settings->use_cgns >= 0) {
4604
4605 if (!arg_settings_trusted)
4606 log_warning("Ignoring cgroup namespace setting, file '%s' is not trusted.", path);
4607 else
4608 arg_use_cgns = settings->use_cgns;
4609 }
4610
4611 if ((arg_settings_mask & SETTING_CLONE_NS_FLAGS) == 0 &&
f5fbe71d 4612 settings->clone_ns_flags != ULONG_MAX) {
de40a303
LP		 4613
4614 if (!arg_settings_trusted)
4615 log_warning("Ignoring namespace setting, file '%s' is not trusted.", path);
4616 else
4617 arg_clone_ns_flags = settings->clone_ns_flags;
4618 }
4619
4620 if ((arg_settings_mask & SETTING_CONSOLE_MODE) == 0 &&
4621 settings->console_mode >= 0) {
4622
4623 if (!arg_settings_trusted)
4624 log_warning("Ignoring console mode setting, file '%s' is not trusted.", path);
4625 else
4626 arg_console_mode = settings->console_mode;
4627 }
4628
d3689b94
LP		 4629 if ((arg_settings_mask & SETTING_SUPPRESS_SYNC) == 0 &&
4630 settings->suppress_sync >= 0)
4a4654e0
LP		 4631 arg_suppress_sync = settings->suppress_sync;
4632
de40a303
LP		 4633 /* The following properties can only be set through the OCI settings logic, not from the command line, hence we
4634 * don't consult arg_settings_mask for them. */
4635
4636 sd_bus_message_unref(arg_property_message);
4637 arg_property_message = TAKE_PTR(settings->properties);
4638
4639 arg_console_width = settings->console_width;
4640 arg_console_height = settings->console_height;
4641
b2645747 4642 device_node_array_free(arg_extra_nodes, arg_n_extra_nodes);
de40a303
LP		 4643 arg_extra_nodes = TAKE_PTR(settings->extra_nodes);
4644 arg_n_extra_nodes = settings->n_extra_nodes;
825210d4 4645 settings->n_extra_nodes = 0;
de40a303 4646
f757855e
LP		 4647 return 0;
4648}
4649
5d961407
LP		 4650static int load_settings(void) {
4651 _cleanup_(settings_freep) Settings *settings = NULL;
4652 _cleanup_fclose_ FILE *f = NULL;
3603f151 4653 _cleanup_free_ char *p = NULL;
5d961407
LP		 4654 int r;
4655
de40a303
LP		 4656 if (arg_oci_bundle)
4657 return 0;
4658
5d961407
LP		 4659 /* If all settings are masked, there's no point in looking for
4660 * the settings file */
d7a0f1f4 4661 if (FLAGS_SET(arg_settings_mask, _SETTINGS_MASK_ALL))
5d961407
LP		 4662 return 0;
4663
5d961407
LP		 4664 /* We first look in the admin's directories in /etc and /run */
4665 FOREACH_STRING(i, "/etc/systemd/nspawn", "/run/systemd/nspawn") {
4666 _cleanup_free_ char *j = NULL;
4667
3603f151 4668 j = path_join(i, arg_settings_filename);
5d961407
LP		 4669 if (!j)
4670 return log_oom();
4671
4672 f = fopen(j, "re");
4673 if (f) {
4674 p = TAKE_PTR(j);
4675
4676 /* By default, we trust configuration from /etc and /run */
4677 if (arg_settings_trusted < 0)
4678 arg_settings_trusted = true;
4679
4680 break;
4681 }
4682
4683 if (errno != ENOENT)
4684 return log_error_errno(errno, "Failed to open %s: %m", j);
4685 }
4686
4687 if (!f) {
4688 /* After that, let's look for a file next to the
4689 * actual image we shall boot. */
4690
4691 if (arg_image) {
162f6477
LP		 4692 r = file_in_same_dir(arg_image, arg_settings_filename, &p);
4693 if (r < 0)
4694 return log_error_errno(r, "Failed to generate settings path from image path: %m");
4695 } else if (arg_directory) {
4696 r = file_in_same_dir(arg_directory, arg_settings_filename, &p);
4697 if (r < 0 && r != -EADDRNOTAVAIL) /* if directory is root fs, don't complain */
4698 return log_error_errno(r, "Failed to generate settings path from directory path: %m");
5d961407
LP		 4699 }
4700
4701 if (p) {
4702 f = fopen(p, "re");
4703 if (!f && errno != ENOENT)
4704 return log_error_errno(errno, "Failed to open %s: %m", p);
4705
4706 /* By default, we do not trust configuration from /var/lib/machines */
4707 if (arg_settings_trusted < 0)
4708 arg_settings_trusted = false;
4709 }
4710 }
4711
4712 if (!f)
4713 return 0;
4714
4715 log_debug("Settings are trusted: %s", yes_no(arg_settings_trusted));
4716
4717 r = settings_load(f, p, &settings);
4718 if (r < 0)
4719 return r;
4720
4721 return merge_settings(settings, p);
4722}
4723
de40a303
LP		 4724static int load_oci_bundle(void) {
4725 _cleanup_(settings_freep) Settings *settings = NULL;
4726 int r;
4727
4728 if (!arg_oci_bundle)
4729 return 0;
4730
4731 /* By default let's trust OCI bundles */
4732 if (arg_settings_trusted < 0)
4733 arg_settings_trusted = true;
4734
4735 r = oci_load(NULL, arg_oci_bundle, &settings);
4736 if (r < 0)
4737 return r;
4738
4739 return merge_settings(settings, arg_oci_bundle);
4740}
4741
3acc84eb 4742static int run_container(
2d845785 4743 DissectedImage *dissected_image,
b0067625
ZJS		 4744 FDSet *fds,
4745 char veth_name[IFNAMSIZ], bool *veth_created,
761cf19d 4746 struct ExposeArgs *expose_args,
3acc84eb 4747 int *master, pid_t *pid, int *ret) {
b0067625
ZJS		 4748
4749 static const struct sigaction sa = {
4750 .sa_handler = nop_signal_handler,
e28c7cd0 4751 .sa_flags = SA_NOCLDSTOP|SA_RESTART,
b0067625
ZJS		 4752 };
4753
8e766630 4754 _cleanup_(release_lock_file) LockFile uid_shift_lock = LOCK_FILE_INIT;
5bb1d7fb 4755 _cleanup_close_ int etc_passwd_lock = -EBADF;
b0067625 4756 _cleanup_close_pair_ int
71136404
LP		 4757 fd_inner_socket_pair[2] = EBADF_PAIR,
4758 fd_outer_socket_pair[2] = EBADF_PAIR;
8199d554 4759
5bb1d7fb 4760 _cleanup_close_ int notify_socket = -EBADF, mntns_fd = -EBADF, fd_kmsg_fifo = -EBADF;
b0067625 4761 _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
5773024d 4762 _cleanup_(sd_event_source_unrefp) sd_event_source *notify_event_source = NULL;
b0067625
ZJS		 4763 _cleanup_(sd_event_unrefp) sd_event *event = NULL;
4764 _cleanup_(pty_forward_freep) PTYForward *forward = NULL;
4765 _cleanup_(sd_netlink_unrefp) sd_netlink *rtnl = NULL;
abdb9b08 4766 _cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
2f893044
LP		 4767 _cleanup_free_ uid_t *bind_user_uid = NULL;
4768 size_t n_bind_user_uid = 0;
b0067625 4769 ContainerStatus container_status = 0;
b0067625
ZJS		 4770 int ifi = 0, r;
4771 ssize_t l;
4772 sigset_t mask_chld;
254d1313 4773 _cleanup_close_ int child_netns_fd = -EBADF;
b0067625
ZJS		 4774
4775 assert_se(sigemptyset(&mask_chld) == 0);
4776 assert_se(sigaddset(&mask_chld, SIGCHLD) == 0);
4777
4778 if (arg_userns_mode == USER_NAMESPACE_PICK) {
4779 /* When we shall pick the UID/GID range, let's first lock /etc/passwd, so that we can safely
4780 * check with getpwuid() if the specific user already exists. Note that /etc might be
4781 * read-only, in which case this will fail with EROFS. But that's really OK, as in that case we
4782 * can be reasonably sure that no users are going to be added. Note that getpwuid() checks are
4783 * really just an extra safety net. We kinda assume that the UID range we allocate from is
4784 * really ours. */
4785
4786 etc_passwd_lock = take_etc_passwd_lock(NULL);
4787 if (etc_passwd_lock < 0 && etc_passwd_lock != -EROFS)
4788 return log_error_errno(etc_passwd_lock, "Failed to take /etc/passwd lock: %m");
4789 }
4790
4791 r = barrier_create(&barrier);
4792 if (r < 0)
4793 return log_error_errno(r, "Cannot initialize IPC barrier: %m");
4794
5d9d3fcb
CB		 4795 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, fd_inner_socket_pair) < 0)
4796 return log_error_errno(errno, "Failed to create inner socket pair: %m");
4797
af06cd30
CB		 4798 if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, fd_outer_socket_pair) < 0)
4799 return log_error_errno(errno, "Failed to create outer socket pair: %m");
b0067625 4800
b0067625
ZJS		 4801 /* Child can be killed before execv(), so handle SIGCHLD in order to interrupt
4802 * parent's blocking calls and give it a chance to call wait() and terminate. */
4803 r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL);
4804 if (r < 0)
4805 return log_error_errno(errno, "Failed to change the signal mask: %m");
4806
4807 r = sigaction(SIGCHLD, &sa, NULL);
4808 if (r < 0)
4809 return log_error_errno(errno, "Failed to install SIGCHLD handler: %m");
4810
d7bea6b6 4811 if (arg_network_namespace_path) {
5b4855ab
DDM		 4812 child_netns_fd = open(arg_network_namespace_path, O_RDONLY|O_NOCTTY|O_CLOEXEC);
4813 if (child_netns_fd < 0)
d7bea6b6
DP		 4814 return log_error_errno(errno, "Cannot open file %s: %m", arg_network_namespace_path);
4815
54c2459d 4816 r = fd_is_ns(child_netns_fd, CLONE_NEWNET);
6619ad88
LP		 4817 if (r == -EUCLEAN)
4818 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);
4819 else if (r < 0)
d7bea6b6 4820 return log_error_errno(r, "Failed to check %s fs type: %m", arg_network_namespace_path);
c6147113
LP		 4821 else if (r == 0)
4822 return log_error_errno(SYNTHETIC_ERRNO(EINVAL),
4823 "Path %s doesn't refer to a network namespace, refusing.", arg_network_namespace_path);
d7bea6b6
DP		 4824 }
4825
b0067625
ZJS		 4826 *pid = raw_clone(SIGCHLD|CLONE_NEWNS);
4827 if (*pid < 0)
4828 return log_error_errno(errno, "clone() failed%s: %m",
4829 errno == EINVAL ?
4830 ", do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in)" : "");
4831
4832 if (*pid == 0) {
4833 /* The outer child only has a file system namespace. */
4834 barrier_set_role(&barrier, BARRIER_CHILD);
4835
5d9d3fcb 4836 fd_inner_socket_pair[0] = safe_close(fd_inner_socket_pair[0]);
af06cd30 4837 fd_outer_socket_pair[0] = safe_close(fd_outer_socket_pair[0]);
b0067625
ZJS		 4838
4839 (void) reset_all_signal_handlers();
4840 (void) reset_signal_mask();
4841
4842 r = outer_child(&barrier,
4843 arg_directory,
2d845785 4844 dissected_image,
af06cd30 4845 fd_outer_socket_pair[1],
5d9d3fcb 4846 fd_inner_socket_pair[1],
d7bea6b6 4847 fds,
5b4855ab 4848 child_netns_fd);
b0067625
ZJS		 4849 if (r < 0)
4850 _exit(EXIT_FAILURE);
4851
4852 _exit(EXIT_SUCCESS);
4853 }
4854
4855 barrier_set_role(&barrier, BARRIER_PARENT);
4856
e4077ff6 4857 fdset_close(fds);
b0067625 4858
5d9d3fcb 4859 fd_inner_socket_pair[1] = safe_close(fd_inner_socket_pair[1]);
af06cd30 4860 fd_outer_socket_pair[1] = safe_close(fd_outer_socket_pair[1]);
b0067625
ZJS		 4861
4862 if (arg_userns_mode != USER_NAMESPACE_NO) {
af06cd30 4863 mntns_fd = receive_one_fd(fd_outer_socket_pair[0], 0);
b71a0192
CB		 4864 if (mntns_fd < 0)
4865 return log_error_errno(mntns_fd, "Failed to receive mount namespace fd from outer child: %m");
4866
b0067625 4867 /* The child just let us know the UID shift it might have read from the image. */
af06cd30 4868 l = recv(fd_outer_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, 0);
b0067625
ZJS		 4869 if (l < 0)
4870 return log_error_errno(errno, "Failed to read UID shift: %m");
c6147113
LP		 4871 if (l != sizeof arg_uid_shift)
4872 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading UID shift.");
b0067625
ZJS		 4873
4874 if (arg_userns_mode == USER_NAMESPACE_PICK) {
4875 /* If we are supposed to pick the UID shift, let's try to use the shift read from the
4876 * image, but if that's already in use, pick a new one, and report back to the child,
4877 * which one we now picked. */
4878
4879 r = uid_shift_pick(&arg_uid_shift, &uid_shift_lock);
4880 if (r < 0)
4881 return log_error_errno(r, "Failed to pick suitable UID/GID range: %m");
4882
af06cd30 4883 l = send(fd_outer_socket_pair[0], &arg_uid_shift, sizeof arg_uid_shift, MSG_NOSIGNAL);
b0067625
ZJS		 4884 if (l < 0)
4885 return log_error_errno(errno, "Failed to send UID shift: %m");
c6147113
LP		 4886 if (l != sizeof arg_uid_shift)
4887 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short write while writing UID shift.");
b0067625 4888 }
2f893044
LP		 4889
4890 n_bind_user_uid = strv_length(arg_bind_user);
4891 if (n_bind_user_uid > 0) {
4892 /* Right after the UID shift, we'll receive the list of UID mappings for the
4893 * --bind-user= logic. Always a quadruplet of payload and host UID + GID. */
4894
4895 bind_user_uid = new(uid_t, n_bind_user_uid*4);
4896 if (!bind_user_uid)
4897 return log_oom();
4898
4899 for (size_t i = 0; i < n_bind_user_uid; i++) {
af06cd30 4900 l = recv(fd_outer_socket_pair[0], bind_user_uid + i*4, sizeof(uid_t)*4, 0);
2f893044
LP		 4901 if (l < 0)
4902 return log_error_errno(errno, "Failed to read user UID map pair: %m");
4903 if (l != sizeof(uid_t)*4)
4904 return log_full_errno(l == 0 ? LOG_DEBUG : LOG_WARNING,
4905 SYNTHETIC_ERRNO(EIO),
4906 "Short read while reading bind user UID pairs.");
4907 }
4908 }
b0067625
ZJS		 4909 }
4910
8199d554
LP		 4911 if (arg_unified_cgroup_hierarchy == CGROUP_UNIFIED_UNKNOWN) {
4912 /* The child let us know the support cgroup mode it might have read from the image. */
fefb7a6d 4913 l = recv(fd_outer_socket_pair[0], &arg_unified_cgroup_hierarchy, sizeof(arg_unified_cgroup_hierarchy), 0);
8199d554
LP		 4914 if (l < 0)
4915 return log_error_errno(errno, "Failed to read cgroup mode: %m");
c6147113 4916 if (l != sizeof(arg_unified_cgroup_hierarchy))
c0f86d66 4917 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading cgroup mode (%zi bytes).%s",
c6147113 4918 l, l == 0 ? " The child is most likely dead." : "");
8199d554
LP		 4919 }
4920
b0067625 4921 /* Wait for the outer child. */
d2e0ac3d
LP		 4922 r = wait_for_terminate_and_check("(sd-namespace)", *pid, WAIT_LOG_ABNORMAL);
4923 if (r < 0)
4924 return r;
4925 if (r != EXIT_SUCCESS)
4926 return -EIO;
b0067625
ZJS		 4927
4928 /* And now retrieve the PID of the inner child. */
af06cd30 4929 l = recv(fd_outer_socket_pair[0], pid, sizeof *pid, 0);
b0067625
ZJS		 4930 if (l < 0)
4931 return log_error_errno(errno, "Failed to read inner child PID: %m");
c6147113
LP		 4932 if (l != sizeof *pid)
4933 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading inner child PID.");
b0067625
ZJS		 4934
4935 /* We also retrieve container UUID in case it was generated by outer child */
af06cd30 4936 l = recv(fd_outer_socket_pair[0], &arg_uuid, sizeof arg_uuid, 0);
b0067625
ZJS		 4937 if (l < 0)
4938 return log_error_errno(errno, "Failed to read container machine ID: %m");
c6147113
LP		 4939 if (l != sizeof(arg_uuid))
4940 return log_error_errno(SYNTHETIC_ERRNO(EIO), "Short read while reading container machined ID.");
b0067625
ZJS		 4941
4942 /* We also retrieve the socket used for notifications generated by outer child */
af06cd30 4943 notify_socket = receive_one_fd(fd_outer_socket_pair[0], 0);
b0067625
ZJS		 4944 if (notify_socket < 0)
4945 return log_error_errno(notify_socket,
4946 "Failed to receive notification socket from the outer child: %m");
4947
4948 log_debug("Init process invoked as PID "PID_FMT, *pid);
4949
4950 if (arg_userns_mode != USER_NAMESPACE_NO) {
c6147113
LP		 4951 if (!barrier_place_and_sync(&barrier)) /* #1 */
4952 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early.");
b0067625 4953
2f893044 4954 r = setup_uid_map(*pid, bind_user_uid, n_bind_user_uid);
b0067625
ZJS		 4955 if (r < 0)
4956 return r;
4957
4958 (void) barrier_place(&barrier); /* #2 */
4959 }
4960
4961 if (arg_private_network) {
75116558
PS		 4962 if (!arg_network_namespace_path) {
4963 /* Wait until the child has unshared its network namespace. */
c6147113
LP		 4964 if (!barrier_place_and_sync(&barrier)) /* #3 */
4965 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early");
75116558
PS		 4966 }
4967
5b4855ab
DDM		 4968 if (child_netns_fd < 0) {
4969 /* Make sure we have an open file descriptor to the child's network
4970 * namespace so it stays alive even if the child exits. */
4971 r = namespace_open(*pid, NULL, NULL, &child_netns_fd, NULL, NULL);
4972 if (r < 0)
4973 return log_error_errno(r, "Failed to open child network namespace: %m");
4974 }
4975
4976 r = move_network_interfaces(child_netns_fd, arg_network_interfaces);
b0067625
ZJS		 4977 if (r < 0)
4978 return r;
4979
4980 if (arg_network_veth) {
4981 r = setup_veth(arg_machine, *pid, veth_name,
813dbff4 4982 arg_network_bridge || arg_network_zone, &arg_network_provided_mac);
b0067625
ZJS		 4983 if (r < 0)
4984 return r;
4985 else if (r > 0)
4986 ifi = r;
4987
4988 if (arg_network_bridge) {
4989 /* Add the interface to a bridge */
4990 r = setup_bridge(veth_name, arg_network_bridge, false);
4991 if (r < 0)
4992 return r;
4993 if (r > 0)
4994 ifi = r;
4995 } else if (arg_network_zone) {
4996 /* Add the interface to a bridge, possibly creating it */
4997 r = setup_bridge(veth_name, arg_network_zone, true);
4998 if (r < 0)
4999 return r;
5000 if (r > 0)
5001 ifi = r;
5002 }
5003 }
5004
5005 r = setup_veth_extra(arg_machine, *pid, arg_network_veth_extra);
5006 if (r < 0)
5007 return r;
5008
5009 /* We created the primary and extra veth links now; let's remember this, so that we know to
5010 remove them later on. Note that we don't bother with removing veth links that were created
5011 here when their setup failed half-way, because in that case the kernel should be able to
5012 remove them on its own, since they cannot be referenced by anything yet. */
5013 *veth_created = true;
5014
5015 r = setup_macvlan(arg_machine, *pid, arg_network_macvlan);
5016 if (r < 0)
5017 return r;
5018
5019 r = setup_ipvlan(arg_machine, *pid, arg_network_ipvlan);
5020 if (r < 0)
5021 return r;
5022 }
5023
abdb9b08
LP		 5024 if (arg_register || !arg_keep_unit) {
5025 r = sd_bus_default_system(&bus);
5026 if (r < 0)
5027 return log_error_errno(r, "Failed to open system bus: %m");
e5a2d8b5
LP		 5028
5029 r = sd_bus_set_close_on_exit(bus, false);
5030 if (r < 0)
5031 return log_error_errno(r, "Failed to disable close-on-exit behaviour: %m");
abdb9b08
LP		 5032 }
5033
5034 if (!arg_keep_unit) {
5035 /* When a new scope is created for this container, then we'll be registered as its controller, in which
5036 * case PID 1 will send us a friendly RequestStop signal, when it is asked to terminate the
5037 * scope. Let's hook into that, and cleanly shut down the container, and print a friendly message. */
5038
75152a4d
LP		 5039 r = sd_bus_match_signal_async(
5040 bus,
5041 NULL,
5042 "org.freedesktop.systemd1",
5043 NULL,
5044 "org.freedesktop.systemd1.Scope",
5045 "RequestStop",
5046 on_request_stop, NULL, PID_TO_PTR(*pid));
abdb9b08 5047 if (r < 0)
75152a4d 5048 return log_error_errno(r, "Failed to request RequestStop match: %m");
abdb9b08
LP		 5049 }
5050
b0067625
ZJS		 5051 if (arg_register) {
5052 r = register_machine(
abdb9b08 5053 bus,
b0067625
ZJS		 5054 arg_machine,
5055 *pid,
5056 arg_directory,
5057 arg_uuid,
5058 ifi,
5059 arg_slice,
5060 arg_custom_mounts, arg_n_custom_mounts,
5061 arg_kill_signal,
5062 arg_property,
de40a303 5063 arg_property_message,
b0067625 5064 arg_keep_unit,
411d8c72
NR		 5065 arg_container_service_name,
5066 arg_start_mode);
b0067625
ZJS		 5067 if (r < 0)
5068 return r;
abdb9b08 5069
cd2dfc6f
LP		 5070 } else if (!arg_keep_unit) {
5071 r = allocate_scope(
abdb9b08 5072 bus,
cd2dfc6f
LP		 5073 arg_machine,
5074 *pid,
5075 arg_slice,
5076 arg_custom_mounts, arg_n_custom_mounts,
5077 arg_kill_signal,
de40a303 5078 arg_property,
7eda208f 5079 arg_property_message,
411d8c72
NR		 5080 /* allow_pidfds= */ true,
5081 arg_start_mode);
cd2dfc6f
LP		 5082 if (r < 0)
5083 return r;
5084
5085 } else if (arg_slice || arg_property)
5086 log_notice("Machine and scope registration turned off, --slice= and --property= settings will have no effect.");
b0067625 5087
27da7ef0 5088 r = create_subcgroup(*pid, arg_keep_unit, arg_unified_cgroup_hierarchy);
b0067625
ZJS		 5089 if (r < 0)
5090 return r;
5091
27da7ef0 5092 r = sync_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
720f0a2f
LP		 5093 if (r < 0)
5094 return r;
b0067625 5095
de54e02d 5096 r = chown_cgroup(*pid, arg_unified_cgroup_hierarchy, arg_uid_shift);
b0067625
ZJS		 5097 if (r < 0)
5098 return r;
5099
5100 /* Notify the child that the parent is ready with all
5101 * its setup (including cgroup-ification), and that
5102 * the child can now hand over control to the code to
5103 * run inside the container. */
75116558 5104 (void) barrier_place(&barrier); /* #4 */
b0067625
ZJS		 5105
5106 /* Block SIGCHLD here, before notifying child.
5107 * process_pty() will handle it with the other signals. */
5108 assert_se(sigprocmask(SIG_BLOCK, &mask_chld, NULL) >= 0);
5109
5110 /* Reset signal to default */
9c274488 5111 r = default_signals(SIGCHLD);
b0067625
ZJS		 5112 if (r < 0)
5113 return log_error_errno(r, "Failed to reset SIGCHLD: %m");
5114
5115 r = sd_event_new(&event);
5116 if (r < 0)
5117 return log_error_errno(r, "Failed to get default event source: %m");
5118
8fd010bb
LP		 5119 (void) sd_event_set_watchdog(event, true);
5120
abdb9b08
LP		 5121 if (bus) {
5122 r = sd_bus_attach_event(bus, event, 0);
5123 if (r < 0)
5124 return log_error_errno(r, "Failed to attach bus to event loop: %m");
5125 }
5126
e96ceaba 5127 r = setup_notify_parent(event, notify_socket, PID_TO_PTR(*pid), &notify_event_source);
b0067625
ZJS		 5128 if (r < 0)
5129 return r;
5130
1a8d7814
LP		 5131 /* Wait that the child is completely ready now, and has mounted their own copies of procfs and so on,
5132 * before we take the fully visible instances away. */
5133 if (!barrier_sync(&barrier)) /* #5.1 */
5134 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early.");
5135
b71a0192
CB		 5136 if (arg_userns_mode != USER_NAMESPACE_NO) {
5137 r = wipe_fully_visible_fs(mntns_fd);
5138 if (r < 0)
5139 return r;
5140 mntns_fd = safe_close(mntns_fd);
5141 }
5142
1a8d7814
LP		 5143 /* And now let the child know that we completed removing the procfs instances, and it can start the
5144 * payload. */
5145 if (!barrier_place(&barrier)) /* #5.2 */
c6147113 5146 return log_error_errno(SYNTHETIC_ERRNO(ESRCH), "Child died too early.");
b0067625 5147
38ccb557 5148 /* At this point we have made use of the UID we picked, and thus nss-systemd/systemd-machined.service
b0067625
ZJS		 5149 * will make them appear in getpwuid(), thus we can release the /etc/passwd lock. */
5150 etc_passwd_lock = safe_close(etc_passwd_lock);
5151
04f590a4
LP		 5152 (void) sd_notifyf(false,
5153 "STATUS=Container running.\n"
5154 "X_NSPAWN_LEADER_PID=" PID_FMT, *pid);
4bf4f50f
ZJS		 5155 if (!arg_notify_ready) {
5156 r = sd_notify(false, "READY=1\n");
5157 if (r < 0)
5158 log_warning_errno(r, "Failed to send readiness notification, ignoring: %m");
5159 }
b0067625
ZJS		 5160
5161 if (arg_kill_signal > 0) {
5162 /* Try to kill the init system on SIGINT or SIGTERM */
919f5ae0
LP		 5163 (void) sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, PID_TO_PTR(*pid));
5164 (void) sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, PID_TO_PTR(*pid));
b0067625
ZJS		 5165 } else {
5166 /* Immediately exit */
919f5ae0
LP		 5167 (void) sd_event_add_signal(event, NULL, SIGINT, NULL, NULL);
5168 (void) sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL);
b0067625
ZJS		 5169 }
5170
988851b6
LP		 5171 (void) sd_event_add_signal(event, NULL, SIGRTMIN+18, sigrtmin18_handler, NULL);
5172
5173 r = sd_event_add_memory_pressure(event, NULL, NULL, NULL);
5174 if (r < 0)
5175 log_debug_errno(r, "Failed allocate memory pressure event source, ignoring: %m");
5176
6916b164 5177 /* Exit when the child exits */
919f5ae0 5178 (void) sd_event_add_signal(event, NULL, SIGCHLD, on_sigchld, PID_TO_PTR(*pid));
b0067625 5179
b07ee903
CB		 5180 /* Retrieve the kmsg fifo allocated by inner child */
5181 fd_kmsg_fifo = receive_one_fd(fd_inner_socket_pair[0], 0);
5182 if (fd_kmsg_fifo < 0)
5183 return log_error_errno(fd_kmsg_fifo, "Failed to receive kmsg fifo from inner child: %m");
5184
b0067625 5185 if (arg_expose_ports) {
b07ee903 5186 r = expose_port_watch_rtnl(event, fd_inner_socket_pair[0], on_address_change, expose_args, &rtnl);
b0067625
ZJS		 5187 if (r < 0)
5188 return r;
5189
deff68e7
FW		 5190 (void) expose_port_execute(rtnl, &expose_args->fw_ctx, arg_expose_ports, AF_INET, &expose_args->address4);
5191 (void) expose_port_execute(rtnl, &expose_args->fw_ctx, arg_expose_ports, AF_INET6, &expose_args->address6);
b0067625
ZJS		 5192 }
5193
3acc84eb 5194 if (arg_console_mode != CONSOLE_PIPE) {
254d1313 5195 _cleanup_close_ int fd = -EBADF;
3acc84eb 5196 PTYForwardFlags flags = 0;
de40a303 5197
3acc84eb 5198 /* Retrieve the master pty allocated by inner child */
bb1aa185 5199 fd = receive_one_fd(fd_inner_socket_pair[0], 0);
3acc84eb
FB		 5200 if (fd < 0)
5201 return log_error_errno(fd, "Failed to receive master pty from the inner child: %m");
5202
5203 switch (arg_console_mode) {
de40a303 5204
3acc84eb
FB		 5205 case CONSOLE_READ_ONLY:
5206 flags |= PTY_FORWARD_READ_ONLY;
5207
5208 _fallthrough_;
5209
5210 case CONSOLE_INTERACTIVE:
5211 flags |= PTY_FORWARD_IGNORE_VHANGUP;
5212
5213 r = pty_forward_new(event, fd, flags, &forward);
5214 if (r < 0)
5215 return log_error_errno(r, "Failed to create PTY forwarder: %m");
5216
f5fbe71d 5217 if (arg_console_width != UINT_MAX || arg_console_height != UINT_MAX)
3acc84eb
FB		 5218 (void) pty_forward_set_width_height(forward,
5219 arg_console_width,
5220 arg_console_height);
5221 break;
5222
5223 default:
5224 assert(arg_console_mode == CONSOLE_PASSIVE);
5225 }
5226
5227 *master = TAKE_FD(fd);
de40a303 5228 }
b0067625 5229
5d9d3fcb
CB		 5230 fd_inner_socket_pair[0] = safe_close(fd_inner_socket_pair[0]);
5231
b0067625
ZJS		 5232 r = sd_event_loop(event);
5233 if (r < 0)
5234 return log_error_errno(r, "Failed to run event loop: %m");
5235
de40a303
LP		 5236 if (forward) {
5237 char last_char = 0;
b0067625 5238
de40a303
LP		 5239 (void) pty_forward_get_last_char(forward, &last_char);
5240 forward = pty_forward_free(forward);
b0067625 5241
de40a303
LP		 5242 if (!arg_quiet && last_char != '\n')
5243 putc('\n', stdout);
5244 }
b0067625
ZJS		 5245
5246 /* Kill if it is not dead yet anyway */
0bb0a9fa
ZJS		 5247 if (!arg_register && !arg_keep_unit && bus)
5248 terminate_scope(bus, arg_machine);
b0067625
ZJS		 5249
5250 /* Normally redundant, but better safe than sorry */
c67b0082 5251 (void) kill(*pid, SIGKILL);
b0067625 5252
5d9d3fcb
CB		 5253 fd_kmsg_fifo = safe_close(fd_kmsg_fifo);
5254
5b4855ab
DDM		 5255 if (arg_private_network) {
5256 /* Move network interfaces back to the parent network namespace. We use `safe_fork`
5257 * to avoid having to move the parent to the child network namespace. */
e9ccae31 5258 r = safe_fork(NULL, FORK_RESET_SIGNALS|FORK_DEATHSIG_SIGTERM|FORK_WAIT|FORK_LOG, NULL);
5b4855ab
DDM		 5259 if (r < 0)
5260 return r;
5261
5262 if (r == 0) {
254d1313 5263 _cleanup_close_ int parent_netns_fd = -EBADF;
5b4855ab 5264
19b761a0 5265 r = namespace_open(getpid_cached(), NULL, NULL, &parent_netns_fd, NULL, NULL);
5b4855ab
DDM		 5266 if (r < 0) {
5267 log_error_errno(r, "Failed to open parent network namespace: %m");
5268 _exit(EXIT_FAILURE);
5269 }
5270
5271 r = namespace_enter(-1, -1, child_netns_fd, -1, -1);
5272 if (r < 0) {
5273 log_error_errno(r, "Failed to enter child network namespace: %m");
5274 _exit(EXIT_FAILURE);
5275 }
5276
2f091b1b
TM		 5277 /* Reverse network interfaces pair list so that interfaces get their initial name back.
5278 * This is about ensuring interfaces get their old name back when being moved back. */
5279 arg_network_interfaces = strv_reverse(arg_network_interfaces);
5280
5b4855ab
DDM		 5281 r = move_network_interfaces(parent_netns_fd, arg_network_interfaces);
5282 if (r < 0)
5283 log_error_errno(r, "Failed to move network interfaces back to parent network namespace: %m");
5284
5285 _exit(r < 0 ? EXIT_FAILURE : EXIT_SUCCESS);
5286 }
5287 }
5288
8f03de53 5289 r = wait_for_container(TAKE_PID(*pid), &container_status);
b0067625 5290
0bb0a9fa
ZJS		 5291 /* Tell machined that we are gone. */
5292 if (bus)
5293 (void) unregister_machine(bus, arg_machine);
5294
b0067625
ZJS		 5295 if (r < 0)
5296 /* We failed to wait for the container, or the container exited abnormally. */
5297 return r;
5298 if (r > 0 || container_status == CONTAINER_TERMINATED) {
27e29a1e
ZJS		 5299 /* r > 0 → The container exited with a non-zero status.
5300 * As a special case, we need to replace 133 with a different value,
5301 * because 133 is special-cased in the service file to reboot the container.
5302 * otherwise → The container exited with zero status and a reboot was not requested.
5303 */
2a49b612 5304 if (r == EXIT_FORCE_RESTART)
27e29a1e 5305 r = EXIT_FAILURE; /* replace 133 with the general failure code */
b0067625 5306 *ret = r;
b0067625
ZJS		 5307 return 0; /* finito */
5308 }
5309
5310 /* CONTAINER_REBOOTED, loop again */
5311
5312 if (arg_keep_unit) {
5313 /* Special handling if we are running as a service: instead of simply
5314 * restarting the machine we want to restart the entire service, so let's
5315 * inform systemd about this with the special exit code 133. The service
5316 * file uses RestartForceExitStatus=133 so that this results in a full
5317 * nspawn restart. This is necessary since we might have cgroup parameters
5318 * set we want to have flushed out. */
2a49b612
ZJS		 5319 *ret = EXIT_FORCE_RESTART;
5320 return 0; /* finito */
b0067625
ZJS		 5321 }
5322
deff68e7
FW		 5323 expose_port_flush(&expose_args->fw_ctx, arg_expose_ports, AF_INET, &expose_args->address4);
5324 expose_port_flush(&expose_args->fw_ctx, arg_expose_ports, AF_INET6, &expose_args->address6);
b0067625
ZJS		 5325
5326 (void) remove_veth_links(veth_name, arg_network_veth_extra);
5327 *veth_created = false;
5328 return 1; /* loop again */
5329}
5330
bf428efb 5331static int initialize_rlimits(void) {
852b6250 5332 /* The default resource limits the kernel passes to PID 1, as per kernel 5.16. Let's pass our container payload
bf428efb
LP		 5333 * the same values as the kernel originally passed to PID 1, in order to minimize differences between host and
5334 * container execution environments. */
5335
5336 static const struct rlimit kernel_defaults[_RLIMIT_MAX] = {
852b6250
LP		 5337 [RLIMIT_AS] = { RLIM_INFINITY, RLIM_INFINITY },
5338 [RLIMIT_CORE] = { 0, RLIM_INFINITY },
5339 [RLIMIT_CPU] = { RLIM_INFINITY, RLIM_INFINITY },
5340 [RLIMIT_DATA] = { RLIM_INFINITY, RLIM_INFINITY },
5341 [RLIMIT_FSIZE] = { RLIM_INFINITY, RLIM_INFINITY },
5342 [RLIMIT_LOCKS] = { RLIM_INFINITY, RLIM_INFINITY },
5343 [RLIMIT_MEMLOCK] = { DEFAULT_RLIMIT_MEMLOCK, DEFAULT_RLIMIT_MEMLOCK },
5344 [RLIMIT_MSGQUEUE] = { 819200, 819200 },
5345 [RLIMIT_NICE] = { 0, 0 },
5346 [RLIMIT_NOFILE] = { 1024, 4096 },
5347 [RLIMIT_RSS] = { RLIM_INFINITY, RLIM_INFINITY },
5348 [RLIMIT_RTPRIO] = { 0, 0 },
5349 [RLIMIT_RTTIME] = { RLIM_INFINITY, RLIM_INFINITY },
5350 [RLIMIT_STACK] = { 8388608, RLIM_INFINITY },
bf428efb
LP		 5351
5352 /* The kernel scales the default for RLIMIT_NPROC and RLIMIT_SIGPENDING based on the system's amount of
5353 * RAM. To provide best compatibility we'll read these limits off PID 1 instead of hardcoding them
5354 * here. This is safe as we know that PID 1 doesn't change these two limits and thus the original
5355 * kernel's initialization should still be valid during runtime — at least if PID 1 is systemd. Note
5356 * that PID 1 changes a number of other resource limits during early initialization which is why we
5357 * don't read the other limits from PID 1 but prefer the static table above. */
5358 };
5359
21c43631 5360 int rl, r;
bf428efb
LP		 5361
5362 for (rl = 0; rl < _RLIMIT_MAX; rl++) {
bf428efb
LP		 5363 /* Let's only fill in what the user hasn't explicitly configured anyway */
5364 if ((arg_settings_mask & (SETTING_RLIMIT_FIRST << rl)) == 0) {
5365 const struct rlimit *v;
5366 struct rlimit buffer;
5367
5368 if (IN_SET(rl, RLIMIT_NPROC, RLIMIT_SIGPENDING)) {
5369 /* For these two let's read the limits off PID 1. See above for an explanation. */
5370
21c43631
LP		 5371 r = pid_getrlimit(1, rl, &buffer);
5372 if (r < 0)
5373 return log_error_errno(r, "Failed to read resource limit RLIMIT_%s of PID 1: %m", rlimit_to_string(rl));
bf428efb 5374
dbf1aca6
LP		 5375 v = &buffer;
5376 } else if (rl == RLIMIT_NOFILE) {
5377 /* We nowadays bump RLIMIT_NOFILE's hard limit early in PID 1 for all
5378 * userspace. Given that nspawn containers are often run without our PID 1,
5379 * let's grant the containers a raised RLIMIT_NOFILE hard limit by default,
5380 * so that container userspace gets similar resources as host userspace
5381 * gets. */
5382 buffer = kernel_defaults[rl];
5383 buffer.rlim_max = MIN((rlim_t) read_nr_open(), (rlim_t) HIGH_RLIMIT_NOFILE);
bf428efb
LP		 5384 v = &buffer;
5385 } else
5386 v = kernel_defaults + rl;
5387
5388 arg_rlimit[rl] = newdup(struct rlimit, v, 1);
5389 if (!arg_rlimit[rl])
5390 return log_oom();
5391 }
5392
5393 if (DEBUG_LOGGING) {
5394 _cleanup_free_ char *k = NULL;
5395
5396 (void) rlimit_format(arg_rlimit[rl], &k);
5397 log_debug("Setting RLIMIT_%s to %s.", rlimit_to_string(rl), k);
5398 }
5399 }
5400
5401 return 0;
5402}
5403
287b7376 5404static int cant_be_in_netns(void) {
254d1313 5405 _cleanup_close_ int fd = -EBADF;
287b7376
LP		 5406 struct ucred ucred;
5407 int r;
5408
5409 /* Check if we are in the same netns as udev. If we aren't, then device monitoring (and thus waiting
5410 * for loopback block devices) won't work, and we will hang. Detect this case and exit early with a
5411 * nice message. */
5412
5413 if (!arg_image) /* only matters if --image= us used, i.e. we actually need to use loopback devices */
5414 return 0;
5415
5416 fd = socket(AF_UNIX, SOCK_SEQPACKET|SOCK_NONBLOCK|SOCK_CLOEXEC, 0);
5417 if (fd < 0)
5418 return log_error_errno(errno, "Failed to allocate udev control socket: %m");
5419
1861986a 5420 r = connect_unix_path(fd, AT_FDCWD, "/run/udev/control");
bb44fd07
ZJS		 5421 if (r == -ENOENT || ERRNO_IS_NEG_DISCONNECT(r))
5422 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
5423 "Sorry, but --image= requires access to the host's /run/ hierarchy, since we need access to udev.");
5424 if (r < 0)
1861986a 5425 return log_error_errno(r, "Failed to connect socket to udev control socket: %m");
287b7376
LP		 5426
5427 r = getpeercred(fd, &ucred);
5428 if (r < 0)
5429 return log_error_errno(r, "Failed to determine peer of udev control socket: %m");
5430
f7a2dc3d 5431 r = in_same_namespace(ucred.pid, 0, NAMESPACE_NET);
287b7376 5432 if (r < 0)
f7a2dc3d
CB		 5433 return log_error_errno(r, "Failed to determine network namespace of udev: %m");
5434 if (r == 0)
287b7376
LP		 5435 return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP),
5436 "Sorry, but --image= is only supported in the main network namespace, since we need access to udev/AF_NETLINK.");
5437 return 0;
5438}
5439
44dbef90 5440static int run(int argc, char *argv[]) {
4c27749b 5441 bool remove_directory = false, remove_image = false, veth_created = false, remove_tmprootdir = false;
5bb1d7fb 5442 _cleanup_close_ int master = -EBADF;
03cfe0d5 5443 _cleanup_fdset_free_ FDSet *fds = NULL;
2d845785 5444 int r, n_fd_passed, ret = EXIT_SUCCESS;
5aa3eba5 5445 char veth_name[IFNAMSIZ] = "";
761cf19d 5446 struct ExposeArgs expose_args = {};
8e766630 5447 _cleanup_(release_lock_file) LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT;
c67b0082 5448 char tmprootdir[] = "/tmp/nspawn-root-XXXXXX";
2d845785 5449 _cleanup_(loop_device_unrefp) LoopDevice *loop = NULL;
18b5886e 5450 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
761cf19d 5451 _cleanup_(fw_ctx_freep) FirewallContext *fw_ctx = NULL;
7bf011e3 5452 pid_t pid = 0;
03cfe0d5
LP		 5453
5454 log_parse_environment();
5455 log_open();
415fc41c 5456
03cfe0d5
LP		 5457 r = parse_argv(argc, argv);
5458 if (r <= 0)
5459 goto finish;
5460
38ee19c0
ZJS		 5461 if (geteuid() != 0) {
5462 r = log_warning_errno(SYNTHETIC_ERRNO(EPERM),
5463 argc >= 2 ? "Need to be root." :
5464 "Need to be root (and some arguments are usually required).\nHint: try --help");
03cfe0d5 5465 goto finish;
38ee19c0 5466 }
fba868fa 5467
287b7376
LP		 5468 r = cant_be_in_netns();
5469 if (r < 0)
5470 goto finish;
5471
bf428efb
LP		 5472 r = initialize_rlimits();
5473 if (r < 0)
5474 goto finish;
5475
de40a303
LP		 5476 r = load_oci_bundle();
5477 if (r < 0)
5478 goto finish;
5479
300a03be
LP		 5480 r = pick_paths();
5481 if (r < 0)
5482 goto finish;
5483
f757855e
LP		 5484 r = determine_names();
5485 if (r < 0)
5486 goto finish;
5487
5488 r = load_settings();
5489 if (r < 0)
5490 goto finish;
5491
dd78141c
DDM		 5492 /* If we're not unsharing the network namespace and are unsharing the user namespace, we won't have
5493 * permissions to bind ports in the container, so let's drop the CAP_NET_BIND_SERVICE capability to
5494 * indicate that. */
5495 if (!arg_private_network && arg_userns_mode != USER_NAMESPACE_NO && arg_uid_shift > 0)
5496 arg_caps_retain &= ~(UINT64_C(1) << CAP_NET_BIND_SERVICE);
5497
d4d99bc6 5498 r = cg_unified();
5eee8290
LP		 5499 if (r < 0) {
5500 log_error_errno(r, "Failed to determine whether the unified cgroups hierarchy is used: %m");
5501 goto finish;
5502 }
5503
f757855e
LP		 5504 r = verify_arguments();
5505 if (r < 0)
5506 goto finish;
03cfe0d5 5507
2f091b1b
TM		 5508 r = verify_network_interfaces_initialized();
5509 if (r < 0)
5510 goto finish;
5511
49048684
ZJS		 5512 /* Reapply environment settings. */
5513 (void) detect_unified_cgroup_hierarchy_from_environment();
8199d554 5514
2949ff26
LP		 5515 /* Ignore SIGPIPE here, because we use splice() on the ptyfwd stuff and that will generate SIGPIPE if
5516 * the result is closed. Note that the container payload child will reset signal mask+handler anyway,
5517 * so just turning this off here means we only turn it off in nspawn itself, not any children. */
9c274488 5518 (void) ignore_signals(SIGPIPE);
2949ff26 5519
03cfe0d5
LP		 5520 n_fd_passed = sd_listen_fds(false);
5521 if (n_fd_passed > 0) {
5522 r = fdset_new_listen_fds(&fds, false);
5523 if (r < 0) {
5524 log_error_errno(r, "Failed to collect file descriptors: %m");
5525 goto finish;
5526 }
5527 }
5528
83e803a9
ZJS		 5529 /* The "default" umask. This is appropriate for most file and directory
5530 * operations performed by nspawn, and is the umask that will be used for
5531 * the child. Functions like copy_devnodes() change the umask temporarily. */
5532 umask(0022);
5533
03cfe0d5
LP		 5534 if (arg_directory) {
5535 assert(!arg_image);
5536
b35ca61a
LP		 5537 /* Safety precaution: let's not allow running images from the live host OS image, as long as
5538 * /var from the host will propagate into container dynamically (because bad things happen if
5539 * two systems write to the same /var). Let's allow it for the special cases where /var is
5540 * either copied (i.e. --ephemeral) or replaced (i.e. --volatile=yes|state). */
5541 if (path_equal(arg_directory, "/") && !(arg_ephemeral || IN_SET(arg_volatile_mode, VOLATILE_YES, VOLATILE_STATE))) {
1406bd66
LP		 5542 r = log_error_errno(SYNTHETIC_ERRNO(EINVAL),
5543 "Spawning container on root directory is not supported. Consider using --ephemeral, --volatile=yes or --volatile=state.");
03cfe0d5
LP		 5544 goto finish;
5545 }
5546
5547 if (arg_ephemeral) {
5548 _cleanup_free_ char *np = NULL;
5549
f461a28d 5550 r = chase_and_update(&arg_directory, 0);
3f342ec4
LP		 5551 if (r < 0)
5552 goto finish;
5553
7bf011e3
LP		 5554 /* If the specified path is a mount point we generate the new snapshot immediately
5555 * inside it under a random name. However if the specified is not a mount point we
5556 * create the new snapshot in the parent directory, just next to it. */
e1873695 5557 r = path_is_mount_point(arg_directory, NULL, 0);
03cfe0d5
LP		 5558 if (r < 0) {
5559 log_error_errno(r, "Failed to determine whether directory %s is mount point: %m", arg_directory);
5560 goto finish;
5561 }
5562 if (r > 0)
770b5ce4 5563 r = tempfn_random_child(arg_directory, "machine.", &np);
03cfe0d5 5564 else
770b5ce4 5565 r = tempfn_random(arg_directory, "machine.", &np);
03cfe0d5 5566 if (r < 0) {
0f3be6ca 5567 log_error_errno(r, "Failed to generate name for directory snapshot: %m");
03cfe0d5
LP		 5568 goto finish;
5569 }
5570
6992459c 5571 /* We take an exclusive lock on this image, since it's our private, ephemeral copy
162392b7 5572 * only owned by us and no one else. */
6992459c 5573 r = image_path_lock(np, LOCK_EX|LOCK_NB, &tree_global_lock, &tree_local_lock);
03cfe0d5
LP		 5574 if (r < 0) {
5575 log_error_errno(r, "Failed to lock %s: %m", np);
5576 goto finish;
5577 }
5578
7bf011e3
LP		 5579 {
5580 BLOCK_SIGNALS(SIGINT);
fab4ef72
DDM		 5581 r = btrfs_subvol_snapshot_at(AT_FDCWD, arg_directory, AT_FDCWD, np,
5582 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
5583 BTRFS_SNAPSHOT_FALLBACK_COPY |
5584 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
5585 BTRFS_SNAPSHOT_RECURSIVE |
5586 BTRFS_SNAPSHOT_QUOTA |
5587 BTRFS_SNAPSHOT_SIGINT);
7bf011e3
LP		 5588 }
5589 if (r == -EINTR) {
5590 log_error_errno(r, "Interrupted while copying file system tree to %s, removed again.", np);
5591 goto finish;
5592 }
03cfe0d5
LP		 5593 if (r < 0) {
5594 log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory);
5595 goto finish;
ec16945e
LP		 5596 }
5597
1cc6c93a 5598 free_and_replace(arg_directory, np);
17cbb288 5599 remove_directory = true;
30535c16 5600 } else {
f461a28d 5601 r = chase_and_update(&arg_directory, arg_template ? CHASE_NONEXISTENT : 0);
8d4aa2bb
LP		 5602 if (r < 0)
5603 goto finish;
5604
30535c16
LP		 5605 r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
5606 if (r == -EBUSY) {
5607 log_error_errno(r, "Directory tree %s is currently busy.", arg_directory);
5608 goto finish;
5609 }
5610 if (r < 0) {
5611 log_error_errno(r, "Failed to lock %s: %m", arg_directory);
476b8254 5612 goto finish;
30535c16
LP		 5613 }
5614
5615 if (arg_template) {
f461a28d 5616 r = chase_and_update(&arg_template, 0);
3f342ec4
LP		 5617 if (r < 0)
5618 goto finish;
5619
7bf011e3
LP		 5620 {
5621 BLOCK_SIGNALS(SIGINT);
fab4ef72
DDM		 5622 r = btrfs_subvol_snapshot_at(AT_FDCWD, arg_template, AT_FDCWD, arg_directory,
5623 (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) |
5624 BTRFS_SNAPSHOT_FALLBACK_COPY |
5625 BTRFS_SNAPSHOT_FALLBACK_DIRECTORY |
5626 BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE |
5627 BTRFS_SNAPSHOT_RECURSIVE |
5628 BTRFS_SNAPSHOT_QUOTA |
5629 BTRFS_SNAPSHOT_SIGINT);
7bf011e3 5630 }
ff6c6cc1
LP		 5631 if (r == -EEXIST)
5632 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
5633 "Directory %s already exists, not populating from template %s.", arg_directory, arg_template);
7bf011e3
LP		 5634 else if (r == -EINTR) {
5635 log_error_errno(r, "Interrupted while copying file system tree to %s, removed again.", arg_directory);
5636 goto finish;
5637 } else if (r < 0) {
83521414 5638 log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template);
30535c16 5639 goto finish;
ff6c6cc1
LP		 5640 } else
5641 log_full(arg_quiet ? LOG_DEBUG : LOG_INFO,
5642 "Populated %s from template %s.", arg_directory, arg_template);
30535c16 5643 }
ec16945e
LP		 5644 }
5645
7732f92b 5646 if (arg_start_mode == START_BOOT) {
aff7ae0d 5647 _cleanup_free_ char *b = NULL;
a5201ed6 5648 const char *p;
d4317fe1 5649 int check_os_release, is_os_tree;
c9fe05e0 5650
aff7ae0d
LP		 5651 if (arg_pivot_root_new) {
5652 b = path_join(arg_directory, arg_pivot_root_new);
41de458a
LP		 5653 if (!b) {
5654 r = log_oom();
5655 goto finish;
5656 }
aff7ae0d
LP		 5657
5658 p = b;
5659 } else
a5201ed6 5660 p = arg_directory;
c9fe05e0 5661
d4317fe1
FS		 5662 check_os_release = getenv_bool("SYSTEMD_NSPAWN_CHECK_OS_RELEASE");
5663 if (check_os_release < 0 && check_os_release != -ENXIO) {
5664 r = log_error_errno(check_os_release, "Failed to parse $SYSTEMD_NSPAWN_CHECK_OS_RELEASE: %m");
5665 goto finish;
5666 }
5667
5668 is_os_tree = path_is_os_tree(p);
5669 if (is_os_tree == 0 && check_os_release == 0)
5670 log_debug("Directory %s is missing an os-release file, continuing anyway.", p);
5671 else if (is_os_tree <= 0) {
aff7ae0d
LP		 5672 r = log_error_errno(SYNTHETIC_ERRNO(EINVAL),
5673 "Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", p);
1b9e5b12
LP		 5674 goto finish;
5675 }
5676 } else {
aff7ae0d 5677 _cleanup_free_ char *p = NULL;
c9fe05e0 5678
a5201ed6 5679 if (arg_pivot_root_new)
aff7ae0d 5680 p = path_join(arg_directory, arg_pivot_root_new, "/usr/");
a5201ed6 5681 else
aff7ae0d 5682 p = path_join(arg_directory, "/usr/");
41de458a
LP		 5683 if (!p) {
5684 r = log_oom();
5685 goto finish;
5686 }
1b9e5b12 5687
aff7ae0d
LP		 5688 if (laccess(p, F_OK) < 0) {
5689 r = log_error_errno(SYNTHETIC_ERRNO(EINVAL),
5690 "Directory %s doesn't look like it has an OS tree (/usr/ directory is missing). Refusing.", arg_directory);
1b9e5b12 5691 goto finish;
1b9e5b12
LP		 5692 }
5693 }
ec16945e 5694
6b9132a9 5695 } else {
d04faa4e 5696 DissectImageFlags dissect_image_flags =
4b5de5dd 5697 DISSECT_IMAGE_GENERIC_ROOT |
d04faa4e
LP		 5698 DISSECT_IMAGE_REQUIRE_ROOT |
5699 DISSECT_IMAGE_RELAX_VAR_CHECK |
73d88b80
LP		 5700 DISSECT_IMAGE_USR_NO_ROOT |
5701 DISSECT_IMAGE_ADD_PARTITION_DEVICES |
5702 DISSECT_IMAGE_PIN_PARTITION_DEVICES;
ec16945e
LP		 5703 assert(arg_image);
5704 assert(!arg_template);
5705
f461a28d 5706 r = chase_and_update(&arg_image, 0);
3f342ec4
LP		 5707 if (r < 0)
5708 goto finish;
5709
0f3be6ca
LP		 5710 if (arg_ephemeral) {
5711 _cleanup_free_ char *np = NULL;
5712
5713 r = tempfn_random(arg_image, "machine.", &np);
5714 if (r < 0) {
5715 log_error_errno(r, "Failed to generate name for image snapshot: %m");
5716 goto finish;
5717 }
5718
6992459c
LP		 5719 /* Always take an exclusive lock on our own ephemeral copy. */
5720 r = image_path_lock(np, LOCK_EX|LOCK_NB, &tree_global_lock, &tree_local_lock);
0f3be6ca 5721 if (r < 0) {
af255804 5722 log_error_errno(r, "Failed to create image lock: %m");
0f3be6ca
LP		 5723 goto finish;
5724 }
5725
7bf011e3
LP		 5726 {
5727 BLOCK_SIGNALS(SIGINT);
7c2f5495
DDM		 5728 r = copy_file_full(arg_image, np, O_EXCL, arg_read_only ? 0400 : 0600,
5729 FS_NOCOW_FL, FS_NOCOW_FL,
5730 COPY_REFLINK|COPY_CRTIME|COPY_SIGINT,
5731 NULL, NULL);
7bf011e3
LP		 5732 }
5733 if (r == -EINTR) {
5734 log_error_errno(r, "Interrupted while copying image file to %s, removed again.", np);
5735 goto finish;
5736 }
0f3be6ca
LP		 5737 if (r < 0) {
5738 r = log_error_errno(r, "Failed to copy image file: %m");
5739 goto finish;
5740 }
5741
1cc6c93a 5742 free_and_replace(arg_image, np);
0f3be6ca
LP		 5743 remove_image = true;
5744 } else {
5745 r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock);
5746 if (r == -EBUSY) {
af255804 5747 log_error_errno(r, "Disk image %s is currently busy.", arg_image);
0f3be6ca
LP		 5748 goto finish;
5749 }
5750 if (r < 0) {
af255804 5751 log_error_errno(r, "Failed to create image lock: %m");
0f3be6ca
LP		 5752 goto finish;
5753 }
4623e8e6 5754
89e62e0b
LP		 5755 r = verity_settings_load(
5756 &arg_verity_settings,
5757 arg_image, NULL, NULL);
e7cbe5cb
LB		 5758 if (r < 0) {
5759 log_error_errno(r, "Failed to read verity artefacts for %s: %m", arg_image);
5760 goto finish;
78ebe980 5761 }
89e62e0b
LP		 5762
5763 if (arg_verity_settings.data_path)
5764 dissect_image_flags |= DISSECT_IMAGE_NO_PARTITION_TABLE;
30535c16
LP		 5765 }
5766
c67b0082 5767 if (!mkdtemp(tmprootdir)) {
0f3be6ca 5768 r = log_error_errno(errno, "Failed to create temporary directory: %m");
6b9132a9 5769 goto finish;
1b9e5b12 5770 }
6b9132a9 5771
c67b0082
LP		 5772 remove_tmprootdir = true;
5773
5774 arg_directory = strdup(tmprootdir);
1b9e5b12
LP		 5775 if (!arg_directory) {
5776 r = log_oom();
5777 goto finish;
6b9132a9 5778 }
88213476 5779
89e62e0b
LP		 5780 r = loop_device_make_by_path(
5781 arg_image,
5782 arg_read_only ? O_RDONLY : O_RDWR,
22ee78a8 5783 /* sector_size= */ UINT32_MAX,
89e62e0b 5784 FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN,
7f52206a 5785 LOCK_SH,
89e62e0b 5786 &loop);
2d845785
LP		 5787 if (r < 0) {
5788 log_error_errno(r, "Failed to set up loopback block device: %m");
842f3b0f
LP		 5789 goto finish;
5790 }
1b9e5b12 5791
bad31660 5792 r = dissect_loop_device_and_warn(
bad31660 5793 loop,
89e62e0b 5794 &arg_verity_settings,
84be0c71
LP		 5795 /* mount_options=*/ NULL,
5796 arg_image_policy ?: &image_policy_container,
e7cbe5cb 5797 dissect_image_flags,
e0f9e7bd 5798 &dissected_image);
2d845785 5799 if (r == -ENOPKG) {
4526113f 5800 /* dissected_image_and_warn() already printed a brief error message. Extend on that with more details */
2d845785
LP		 5801 log_notice("Note that the disk image needs to\n"
5802 " a) either contain only a single MBR partition of type 0x83 that is marked bootable\n"
5803 " b) or contain a single GPT partition of type 0FC63DAF-8483-4772-8E79-3D69D8477DE4\n"
db811444 5804 " c) or follow https://uapi-group.org/specifications/specs/discoverable_partitions_specification\n"
2d845785
LP		 5805 " d) or contain a file system without a partition table\n"
5806 "in order to be bootable with systemd-nspawn.");
1b9e5b12 5807 goto finish;
2d845785 5808 }
4526113f 5809 if (r < 0)
842f3b0f 5810 goto finish;
1b9e5b12 5811
88b3300f
LP		 5812 r = dissected_image_load_verity_sig_partition(
5813 dissected_image,
5814 loop->fd,
5815 &arg_verity_settings);
5816 if (r < 0)
5817 goto finish;
5818
8ee9615e
LP		 5819 if (dissected_image->has_verity && !arg_verity_settings.root_hash && !dissected_image->has_verity_sig)
5820 log_notice("Note: image %s contains verity information, but no root hash specified and no embedded "
5821 "root hash signature found! Proceeding without integrity checking.", arg_image);
4623e8e6 5822
89e62e0b
LP		 5823 r = dissected_image_decrypt_interactively(
5824 dissected_image,
5825 NULL,
5826 &arg_verity_settings,
e330f97a 5827 0);
1b9e5b12
LP		 5828 if (r < 0)
5829 goto finish;
0f3be6ca
LP		 5830
5831 /* Now that we mounted the image, let's try to remove it again, if it is ephemeral */
5832 if (remove_image && unlink(arg_image) >= 0)
5833 remove_image = false;
4c27749b
LP		 5834
5835 if (arg_architecture < 0)
5836 arg_architecture = dissected_image_architecture(dissected_image);
842f3b0f 5837 }
842f3b0f 5838
86c0dd4a 5839 r = custom_mount_prepare_all(arg_directory, arg_custom_mounts, arg_n_custom_mounts);
5a8af538
LP		 5840 if (r < 0)
5841 goto finish;
5842
de40a303 5843 if (arg_console_mode < 0)
dd9c8da8
MY		 5844 arg_console_mode = isatty(STDIN_FILENO) && isatty(STDOUT_FILENO) ?
5845 CONSOLE_INTERACTIVE : CONSOLE_READ_ONLY;
9c857b9d 5846
de40a303
LP		 5847 if (arg_console_mode == CONSOLE_PIPE) /* if we pass STDERR on to the container, don't add our own logs into it too */
5848 arg_quiet = true;
a258bf26 5849
9c857b9d 5850 if (!arg_quiet)
c85c2f79 5851 log_info("Spawning container %s on %s.\nPress Ctrl-] three times within 1s to kill container.",
9c857b9d
LP		 5852 arg_machine, arg_image ?: arg_directory);
5853
988851b6 5854 assert_se(sigprocmask_many(SIG_BLOCK, NULL, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, SIGRTMIN+18, -1) >= 0);
a258bf26 5855
8c3fe1b5
LP		 5856 r = make_reaper_process(true);
5857 if (r < 0) {
5858 log_error_errno(r, "Failed to become subreaper: %m");
03cfe0d5
LP		 5859 goto finish;
5860 }
5861
761cf19d
FW		 5862 if (arg_expose_ports) {
5863 r = fw_ctx_new(&fw_ctx);
5864 if (r < 0) {
5865 log_error_errno(r, "Cannot expose configured ports, firewall initialization failed: %m");
5866 goto finish;
5867 }
5868 expose_args.fw_ctx = fw_ctx;
5869 }
d87be9b0 5870 for (;;) {
3acc84eb 5871 r = run_container(dissected_image,
44dbef90
LP		 5872 fds,
5873 veth_name, &veth_created,
761cf19d 5874 &expose_args, &master,
44dbef90 5875 &pid, &ret);
b0067625 5876 if (r <= 0)
d87be9b0 5877 break;
d87be9b0 5878 }
88213476
LP		 5879
5880finish:
04f590a4
LP		 5881 (void) sd_notify(false,
5882 r == 0 && ret == EXIT_FORCE_RESTART ? "STOPPING=1\nSTATUS=Restarting..." :
5883 "STOPPING=1\nSTATUS=Terminating...");
af4ec430 5884
9444b1f2 5885 if (pid > 0)
c67b0082 5886 (void) kill(pid, SIGKILL);
88213476 5887
503546da 5888 /* Try to flush whatever is still queued in the pty */
6a0f896b 5889 if (master >= 0) {
f5fbe71d 5890 (void) copy_bytes(master, STDOUT_FILENO, UINT64_MAX, 0);
6a0f896b
LP		 5891 master = safe_close(master);
5892 }
5893
5894 if (pid > 0)
5895 (void) wait_for_terminate(pid, NULL);
503546da 5896
50ebcf6c
LP		 5897 pager_close();
5898
17cbb288 5899 if (remove_directory && arg_directory) {
ec16945e
LP		 5900 int k;
5901
17cbb288 5902 k = rm_rf(arg_directory, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME);
ec16945e 5903 if (k < 0)
17cbb288 5904 log_warning_errno(k, "Cannot remove '%s', ignoring: %m", arg_directory);
ec16945e
LP		 5905 }
5906
0f3be6ca
LP		 5907 if (remove_image && arg_image) {
5908 if (unlink(arg_image) < 0)
5909 log_warning_errno(errno, "Can't remove image file '%s', ignoring: %m", arg_image);
5910 }
5911
c67b0082
LP		 5912 if (remove_tmprootdir) {
5913 if (rmdir(tmprootdir) < 0)
5914 log_debug_errno(errno, "Can't remove temporary root directory '%s', ignoring: %m", tmprootdir);
5915 }
5916
785890ac
LP		 5917 if (arg_machine) {
5918 const char *p;
5919
63c372cb 5920 p = strjoina("/run/systemd/nspawn/propagate/", arg_machine);
c6878637 5921 (void) rm_rf(p, REMOVE_ROOT);
785890ac
LP		 5922 }
5923
deff68e7
FW		 5924 expose_port_flush(&fw_ctx, arg_expose_ports, AF_INET, &expose_args.address4);
5925 expose_port_flush(&fw_ctx, arg_expose_ports, AF_INET6, &expose_args.address6);
7513c5b8
LP		 5926
5927 if (veth_created)
5928 (void) remove_veth_links(veth_name, arg_network_veth_extra);
22b28dfd 5929 (void) remove_bridge(arg_network_zone);
f757855e 5930
f757855e
LP		 5931 custom_mount_free_all(arg_custom_mounts, arg_n_custom_mounts);
5932 expose_port_free_all(arg_expose_ports);
bf428efb 5933 rlimit_free_all(arg_rlimit);
b2645747 5934 device_node_array_free(arg_extra_nodes, arg_n_extra_nodes);
6d0b55c2 5935
44dbef90
LP		 5936 if (r < 0)
5937 return r;
5938
5939 return ret;
88213476 5940}
44dbef90
LP		 5941
5942DEFINE_MAIN_FUNCTION_WITH_POSITIVE_FAILURE(run);
Unnamed repository; edit this file 'description' to name the repository.