]>
Commit | Line | Data |
---|---|---|
88213476 LP |
1 | /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ |
2 | ||
3 | /*** | |
4 | This file is part of systemd. | |
5 | ||
6 | Copyright 2010 Lennart Poettering | |
7 | ||
8 | systemd is free software; you can redistribute it and/or modify it | |
5430f7f2 LP |
9 | under the terms of the GNU Lesser General Public License as published by |
10 | the Free Software Foundation; either version 2.1 of the License, or | |
88213476 LP |
11 | (at your option) any later version. |
12 | ||
13 | systemd is distributed in the hope that it will be useful, but | |
14 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
5430f7f2 | 16 | Lesser General Public License for more details. |
88213476 | 17 | |
5430f7f2 | 18 | You should have received a copy of the GNU Lesser General Public License |
88213476 LP |
19 | along with systemd; If not, see <http://www.gnu.org/licenses/>. |
20 | ***/ | |
21 | ||
22 | #include <signal.h> | |
23 | #include <sched.h> | |
24 | #include <unistd.h> | |
25 | #include <sys/types.h> | |
26 | #include <sys/syscall.h> | |
27 | #include <sys/mount.h> | |
28 | #include <sys/wait.h> | |
29 | #include <stdlib.h> | |
30 | #include <string.h> | |
31 | #include <stdio.h> | |
32 | #include <errno.h> | |
33 | #include <sys/prctl.h> | |
34 | #include <sys/capability.h> | |
35 | #include <getopt.h> | |
a258bf26 LP |
36 | #include <termios.h> |
37 | #include <sys/signalfd.h> | |
687d0825 | 38 | #include <grp.h> |
5ed27dbd | 39 | #include <linux/fs.h> |
9537eab0 LP |
40 | #include <sys/un.h> |
41 | #include <sys/socket.h> | |
aea38d80 | 42 | #include <linux/netlink.h> |
aa28aefe | 43 | #include <linux/rtnetlink.h> |
354bfd2b | 44 | #include <sys/eventfd.h> |
aa28aefe | 45 | #include <net/if.h> |
69c79d3c | 46 | #include <linux/veth.h> |
aa28aefe | 47 | |
5d63309c | 48 | #ifdef HAVE_SELINUX |
a8828ed9 DW |
49 | #include <selinux/selinux.h> |
50 | #endif | |
88213476 | 51 | |
24fb1112 LP |
52 | #ifdef HAVE_SECCOMP |
53 | #include <seccomp.h> | |
54 | #endif | |
55 | ||
1f0cd86b LP |
56 | #include "sd-daemon.h" |
57 | #include "sd-bus.h" | |
58 | #include "sd-id128.h" | |
aa28aefe | 59 | #include "sd-rtnl.h" |
88213476 LP |
60 | #include "log.h" |
61 | #include "util.h" | |
49e942b2 | 62 | #include "mkdir.h" |
6b2d0e85 | 63 | #include "macro.h" |
d7832d2c | 64 | #include "audit.h" |
94d82985 | 65 | #include "missing.h" |
04d391da | 66 | #include "cgroup-util.h" |
a258bf26 | 67 | #include "strv.h" |
9eb977db | 68 | #include "path-util.h" |
a41fe3a2 | 69 | #include "loopback-setup.h" |
4fc9982c | 70 | #include "dev-setup.h" |
842f3b0f | 71 | #include "fdset.h" |
acbeb427 | 72 | #include "build.h" |
a5c32cff | 73 | #include "fileio.h" |
40ca29a1 | 74 | #include "bus-util.h" |
1f0cd86b | 75 | #include "bus-error.h" |
4ba93280 | 76 | #include "ptyfwd.h" |
9bd37b40 | 77 | #include "bus-kernel.h" |
f4889f65 | 78 | #include "env-util.h" |
7f112f50 | 79 | #include "def.h" |
aa28aefe | 80 | #include "rtnl-util.h" |
7e227024 | 81 | #include "udev-util.h" |
f2d88580 | 82 | |
57fb9fb5 LP |
83 | typedef enum LinkJournal { |
84 | LINK_NO, | |
85 | LINK_AUTO, | |
86 | LINK_HOST, | |
87 | LINK_GUEST | |
88 | } LinkJournal; | |
88213476 LP |
89 | |
90 | static char *arg_directory = NULL; | |
687d0825 | 91 | static char *arg_user = NULL; |
9444b1f2 | 92 | static sd_id128_t arg_uuid = {}; |
7027ff61 | 93 | static char *arg_machine = NULL; |
82adf6af LP |
94 | static char *arg_selinux_context = NULL; |
95 | static char *arg_selinux_apifs_context = NULL; | |
9444b1f2 | 96 | static const char *arg_slice = NULL; |
ff01d048 | 97 | static bool arg_private_network = false; |
bc2f673e | 98 | static bool arg_read_only = false; |
0f0dbc46 | 99 | static bool arg_boot = false; |
57fb9fb5 | 100 | static LinkJournal arg_link_journal = LINK_AUTO; |
5076f0cc LP |
101 | static uint64_t arg_retain = |
102 | (1ULL << CAP_CHOWN) | | |
103 | (1ULL << CAP_DAC_OVERRIDE) | | |
104 | (1ULL << CAP_DAC_READ_SEARCH) | | |
105 | (1ULL << CAP_FOWNER) | | |
106 | (1ULL << CAP_FSETID) | | |
107 | (1ULL << CAP_IPC_OWNER) | | |
108 | (1ULL << CAP_KILL) | | |
109 | (1ULL << CAP_LEASE) | | |
110 | (1ULL << CAP_LINUX_IMMUTABLE) | | |
111 | (1ULL << CAP_NET_BIND_SERVICE) | | |
112 | (1ULL << CAP_NET_BROADCAST) | | |
113 | (1ULL << CAP_NET_RAW) | | |
114 | (1ULL << CAP_SETGID) | | |
115 | (1ULL << CAP_SETFCAP) | | |
116 | (1ULL << CAP_SETPCAP) | | |
117 | (1ULL << CAP_SETUID) | | |
118 | (1ULL << CAP_SYS_ADMIN) | | |
119 | (1ULL << CAP_SYS_CHROOT) | | |
120 | (1ULL << CAP_SYS_NICE) | | |
121 | (1ULL << CAP_SYS_PTRACE) | | |
122 | (1ULL << CAP_SYS_TTY_CONFIG) | | |
d87be9b0 | 123 | (1ULL << CAP_SYS_RESOURCE) | |
88d04e31 LP |
124 | (1ULL << CAP_SYS_BOOT) | |
125 | (1ULL << CAP_AUDIT_WRITE) | | |
7f112f50 LP |
126 | (1ULL << CAP_AUDIT_CONTROL) | |
127 | (1ULL << CAP_MKNOD); | |
17fe0523 LP |
128 | static char **arg_bind = NULL; |
129 | static char **arg_bind_ro = NULL; | |
f4889f65 | 130 | static char **arg_setenv = NULL; |
284c0b91 | 131 | static bool arg_quiet = false; |
8a96d94e | 132 | static bool arg_share_system = false; |
eb91eb18 | 133 | static bool arg_register = true; |
89f7c846 | 134 | static bool arg_keep_unit = false; |
aa28aefe | 135 | static char **arg_network_interfaces = NULL; |
69c79d3c | 136 | static bool arg_network_veth = false; |
88213476 LP |
137 | |
138 | static int help(void) { | |
139 | ||
140 | printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" | |
141 | "Spawn a minimal namespace container for debugging, testing and building.\n\n" | |
a8828ed9 DW |
142 | " -h --help Show this help\n" |
143 | " --version Print version string\n" | |
69c79d3c | 144 | " -q --quiet Do not show status information\n" |
a8828ed9 DW |
145 | " -D --directory=NAME Root directory for the container\n" |
146 | " -b --boot Boot up full system (i.e. invoke init)\n" | |
147 | " -u --user=USER Run the command under specified user or uid\n" | |
a8828ed9 | 148 | " -M --machine=NAME Set the machine name for the container\n" |
69c79d3c | 149 | " --uuid=UUID Set a specific machine UUID for the container\n" |
a8828ed9 | 150 | " -S --slice=SLICE Place the container in the specified slice\n" |
69c79d3c LP |
151 | " --private-network Disable network in container\n" |
152 | " --network-interface=INTERFACE\n" | |
153 | " Assign an existing network interface to the\n" | |
154 | " container\n" | |
155 | " --network-veth Add a a virtual ethernet connection between host\n" | |
156 | " and container\n" | |
82adf6af LP |
157 | " -Z --selinux-context=SECLABEL\n" |
158 | " Set the SELinux security context to be used by\n" | |
159 | " processes in the container\n" | |
160 | " -L --selinux-apifs-context=SECLABEL\n" | |
161 | " Set the SELinux security context to be used by\n" | |
162 | " API/tmpfs file systems in the container\n" | |
a8828ed9 DW |
163 | " --capability=CAP In addition to the default, retain specified\n" |
164 | " capability\n" | |
165 | " --drop-capability=CAP Drop the specified capability from the default set\n" | |
166 | " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" | |
167 | " -j Equivalent to --link-journal=host\n" | |
69c79d3c | 168 | " --read-only Mount the root directory read-only\n" |
a8828ed9 DW |
169 | " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" |
170 | " the container\n" | |
171 | " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n" | |
284c0b91 | 172 | " --setenv=NAME=VALUE Pass an environment variable to PID 1\n" |
69c79d3c | 173 | " --share-system Share system namespaces with host\n" |
eb91eb18 | 174 | " --register=BOOLEAN Register container as machine\n" |
89f7c846 | 175 | " --keep-unit Do not register a scope for the machine, reuse\n" |
69c79d3c | 176 | " the service unit nspawn is running in\n", |
88213476 LP |
177 | program_invocation_short_name); |
178 | ||
179 | return 0; | |
180 | } | |
181 | ||
182 | static int parse_argv(int argc, char *argv[]) { | |
183 | ||
a41fe3a2 | 184 | enum { |
acbeb427 ZJS |
185 | ARG_VERSION = 0x100, |
186 | ARG_PRIVATE_NETWORK, | |
bc2f673e | 187 | ARG_UUID, |
5076f0cc | 188 | ARG_READ_ONLY, |
57fb9fb5 | 189 | ARG_CAPABILITY, |
420c7379 | 190 | ARG_DROP_CAPABILITY, |
17fe0523 LP |
191 | ARG_LINK_JOURNAL, |
192 | ARG_BIND, | |
f4889f65 LP |
193 | ARG_BIND_RO, |
194 | ARG_SETENV, | |
eb91eb18 | 195 | ARG_SHARE_SYSTEM, |
89f7c846 | 196 | ARG_REGISTER, |
aa28aefe | 197 | ARG_KEEP_UNIT, |
69c79d3c LP |
198 | ARG_NETWORK_INTERFACE, |
199 | ARG_NETWORK_VETH, | |
a41fe3a2 LP |
200 | }; |
201 | ||
88213476 | 202 | static const struct option options[] = { |
aa28aefe LP |
203 | { "help", no_argument, NULL, 'h' }, |
204 | { "version", no_argument, NULL, ARG_VERSION }, | |
205 | { "directory", required_argument, NULL, 'D' }, | |
206 | { "user", required_argument, NULL, 'u' }, | |
207 | { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, | |
208 | { "boot", no_argument, NULL, 'b' }, | |
209 | { "uuid", required_argument, NULL, ARG_UUID }, | |
210 | { "read-only", no_argument, NULL, ARG_READ_ONLY }, | |
211 | { "capability", required_argument, NULL, ARG_CAPABILITY }, | |
212 | { "drop-capability", required_argument, NULL, ARG_DROP_CAPABILITY }, | |
213 | { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, | |
214 | { "bind", required_argument, NULL, ARG_BIND }, | |
215 | { "bind-ro", required_argument, NULL, ARG_BIND_RO }, | |
216 | { "machine", required_argument, NULL, 'M' }, | |
217 | { "slice", required_argument, NULL, 'S' }, | |
218 | { "setenv", required_argument, NULL, ARG_SETENV }, | |
219 | { "selinux-context", required_argument, NULL, 'Z' }, | |
220 | { "selinux-apifs-context", required_argument, NULL, 'L' }, | |
221 | { "quiet", no_argument, NULL, 'q' }, | |
222 | { "share-system", no_argument, NULL, ARG_SHARE_SYSTEM }, | |
223 | { "register", required_argument, NULL, ARG_REGISTER }, | |
224 | { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, | |
225 | { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, | |
69c79d3c | 226 | { "network-veth", no_argument, NULL, ARG_NETWORK_VETH }, |
eb9da376 | 227 | {} |
88213476 LP |
228 | }; |
229 | ||
9444b1f2 | 230 | int c, r; |
a42c8b54 | 231 | uint64_t plus = 0, minus = 0; |
88213476 LP |
232 | |
233 | assert(argc >= 0); | |
234 | assert(argv); | |
235 | ||
284c0b91 | 236 | while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:q", options, NULL)) >= 0) { |
88213476 LP |
237 | |
238 | switch (c) { | |
239 | ||
240 | case 'h': | |
eb9da376 | 241 | return help(); |
88213476 | 242 | |
acbeb427 ZJS |
243 | case ARG_VERSION: |
244 | puts(PACKAGE_STRING); | |
245 | puts(SYSTEMD_FEATURES); | |
246 | return 0; | |
247 | ||
88213476 LP |
248 | case 'D': |
249 | free(arg_directory); | |
3a74cea5 LP |
250 | arg_directory = canonicalize_file_name(optarg); |
251 | if (!arg_directory) { | |
898d5c91 | 252 | log_error("Invalid root directory: %m"); |
88213476 LP |
253 | return -ENOMEM; |
254 | } | |
255 | ||
256 | break; | |
257 | ||
687d0825 MV |
258 | case 'u': |
259 | free(arg_user); | |
7027ff61 LP |
260 | arg_user = strdup(optarg); |
261 | if (!arg_user) | |
262 | return log_oom(); | |
687d0825 MV |
263 | |
264 | break; | |
265 | ||
69c79d3c LP |
266 | case ARG_NETWORK_VETH: |
267 | arg_network_veth = true; | |
268 | arg_private_network = true; | |
269 | break; | |
270 | ||
aa28aefe LP |
271 | case ARG_NETWORK_INTERFACE: |
272 | if (strv_push(&arg_network_interfaces, optarg) < 0) | |
273 | return log_oom(); | |
274 | ||
275 | /* fall through */ | |
276 | ||
ff01d048 LP |
277 | case ARG_PRIVATE_NETWORK: |
278 | arg_private_network = true; | |
a41fe3a2 LP |
279 | break; |
280 | ||
0f0dbc46 LP |
281 | case 'b': |
282 | arg_boot = true; | |
283 | break; | |
284 | ||
144f0fc0 | 285 | case ARG_UUID: |
9444b1f2 LP |
286 | r = sd_id128_from_string(optarg, &arg_uuid); |
287 | if (r < 0) { | |
aa96c6cb | 288 | log_error("Invalid UUID: %s", optarg); |
9444b1f2 | 289 | return r; |
aa96c6cb | 290 | } |
9444b1f2 | 291 | break; |
aa96c6cb | 292 | |
9444b1f2 LP |
293 | case 'S': |
294 | arg_slice = strdup(optarg); | |
b3451bed DH |
295 | if (!arg_slice) |
296 | return log_oom(); | |
297 | ||
144f0fc0 LP |
298 | break; |
299 | ||
7027ff61 | 300 | case 'M': |
eb91eb18 LP |
301 | if (isempty(optarg)) { |
302 | free(arg_machine); | |
303 | arg_machine = NULL; | |
304 | } else { | |
7027ff61 | 305 | |
eb91eb18 LP |
306 | if (!hostname_is_valid(optarg)) { |
307 | log_error("Invalid machine name: %s", optarg); | |
308 | return -EINVAL; | |
309 | } | |
7027ff61 | 310 | |
eb91eb18 LP |
311 | free(arg_machine); |
312 | arg_machine = strdup(optarg); | |
313 | if (!arg_machine) | |
314 | return log_oom(); | |
315 | ||
316 | break; | |
317 | } | |
7027ff61 | 318 | |
82adf6af LP |
319 | case 'Z': |
320 | arg_selinux_context = optarg; | |
a8828ed9 DW |
321 | break; |
322 | ||
82adf6af LP |
323 | case 'L': |
324 | arg_selinux_apifs_context = optarg; | |
a8828ed9 DW |
325 | break; |
326 | ||
bc2f673e LP |
327 | case ARG_READ_ONLY: |
328 | arg_read_only = true; | |
329 | break; | |
330 | ||
420c7379 LP |
331 | case ARG_CAPABILITY: |
332 | case ARG_DROP_CAPABILITY: { | |
5076f0cc LP |
333 | char *state, *word; |
334 | size_t length; | |
335 | ||
336 | FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) { | |
39ed67d1 | 337 | _cleanup_free_ char *t; |
5076f0cc | 338 | cap_value_t cap; |
5076f0cc LP |
339 | |
340 | t = strndup(word, length); | |
0d0f0c50 SL |
341 | if (!t) |
342 | return log_oom(); | |
5076f0cc | 343 | |
39ed67d1 LP |
344 | if (streq(t, "all")) { |
345 | if (c == ARG_CAPABILITY) | |
a42c8b54 | 346 | plus = (uint64_t) -1; |
39ed67d1 | 347 | else |
a42c8b54 | 348 | minus = (uint64_t) -1; |
39ed67d1 LP |
349 | } else { |
350 | if (cap_from_name(t, &cap) < 0) { | |
351 | log_error("Failed to parse capability %s.", t); | |
352 | return -EINVAL; | |
353 | } | |
354 | ||
355 | if (c == ARG_CAPABILITY) | |
a42c8b54 | 356 | plus |= 1ULL << (uint64_t) cap; |
39ed67d1 | 357 | else |
a42c8b54 | 358 | minus |= 1ULL << (uint64_t) cap; |
5076f0cc | 359 | } |
5076f0cc LP |
360 | } |
361 | ||
362 | break; | |
363 | } | |
364 | ||
57fb9fb5 LP |
365 | case 'j': |
366 | arg_link_journal = LINK_GUEST; | |
367 | break; | |
368 | ||
369 | case ARG_LINK_JOURNAL: | |
370 | if (streq(optarg, "auto")) | |
371 | arg_link_journal = LINK_AUTO; | |
372 | else if (streq(optarg, "no")) | |
373 | arg_link_journal = LINK_NO; | |
374 | else if (streq(optarg, "guest")) | |
375 | arg_link_journal = LINK_GUEST; | |
376 | else if (streq(optarg, "host")) | |
377 | arg_link_journal = LINK_HOST; | |
378 | else { | |
379 | log_error("Failed to parse link journal mode %s", optarg); | |
380 | return -EINVAL; | |
381 | } | |
382 | ||
383 | break; | |
384 | ||
17fe0523 LP |
385 | case ARG_BIND: |
386 | case ARG_BIND_RO: { | |
387 | _cleanup_free_ char *a = NULL, *b = NULL; | |
388 | char *e; | |
389 | char ***x; | |
17fe0523 LP |
390 | |
391 | x = c == ARG_BIND ? &arg_bind : &arg_bind_ro; | |
392 | ||
393 | e = strchr(optarg, ':'); | |
394 | if (e) { | |
395 | a = strndup(optarg, e - optarg); | |
396 | b = strdup(e + 1); | |
397 | } else { | |
398 | a = strdup(optarg); | |
399 | b = strdup(optarg); | |
400 | } | |
401 | ||
402 | if (!a || !b) | |
403 | return log_oom(); | |
404 | ||
405 | if (!path_is_absolute(a) || !path_is_absolute(b)) { | |
406 | log_error("Invalid bind mount specification: %s", optarg); | |
407 | return -EINVAL; | |
408 | } | |
409 | ||
410 | r = strv_extend(x, a); | |
411 | if (r < 0) | |
b3451bed | 412 | return log_oom(); |
17fe0523 LP |
413 | |
414 | r = strv_extend(x, b); | |
415 | if (r < 0) | |
b3451bed | 416 | return log_oom(); |
17fe0523 LP |
417 | |
418 | break; | |
419 | } | |
420 | ||
f4889f65 LP |
421 | case ARG_SETENV: { |
422 | char **n; | |
423 | ||
424 | if (!env_assignment_is_valid(optarg)) { | |
425 | log_error("Environment variable assignment '%s' is not valid.", optarg); | |
426 | return -EINVAL; | |
427 | } | |
428 | ||
429 | n = strv_env_set(arg_setenv, optarg); | |
430 | if (!n) | |
431 | return log_oom(); | |
432 | ||
433 | strv_free(arg_setenv); | |
434 | arg_setenv = n; | |
435 | break; | |
436 | } | |
437 | ||
284c0b91 LP |
438 | case 'q': |
439 | arg_quiet = true; | |
440 | break; | |
441 | ||
8a96d94e LP |
442 | case ARG_SHARE_SYSTEM: |
443 | arg_share_system = true; | |
444 | break; | |
445 | ||
eb91eb18 LP |
446 | case ARG_REGISTER: |
447 | r = parse_boolean(optarg); | |
448 | if (r < 0) { | |
449 | log_error("Failed to parse --register= argument: %s", optarg); | |
450 | return r; | |
451 | } | |
452 | ||
453 | arg_register = r; | |
454 | break; | |
455 | ||
89f7c846 LP |
456 | case ARG_KEEP_UNIT: |
457 | arg_keep_unit = true; | |
458 | break; | |
459 | ||
88213476 LP |
460 | case '?': |
461 | return -EINVAL; | |
462 | ||
463 | default: | |
eb9da376 | 464 | assert_not_reached("Unhandled option"); |
88213476 LP |
465 | } |
466 | } | |
467 | ||
eb91eb18 LP |
468 | if (arg_share_system) |
469 | arg_register = false; | |
470 | ||
471 | if (arg_boot && arg_share_system) { | |
472 | log_error("--boot and --share-system may not be combined."); | |
473 | return -EINVAL; | |
474 | } | |
475 | ||
89f7c846 LP |
476 | if (arg_keep_unit && cg_pid_get_owner_uid(0, NULL) >= 0) { |
477 | log_error("--keep-unit may not be used when invoked from a user session."); | |
478 | return -EINVAL; | |
479 | } | |
480 | ||
a42c8b54 LP |
481 | arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; |
482 | ||
88213476 LP |
483 | return 1; |
484 | } | |
485 | ||
486 | static int mount_all(const char *dest) { | |
487 | ||
488 | typedef struct MountPoint { | |
489 | const char *what; | |
490 | const char *where; | |
491 | const char *type; | |
492 | const char *options; | |
493 | unsigned long flags; | |
3bd66c05 | 494 | bool fatal; |
88213476 LP |
495 | } MountPoint; |
496 | ||
497 | static const MountPoint mount_table[] = { | |
4b7a6af4 | 498 | { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, |
b4c59701 LP |
499 | { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */ |
500 | { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ | |
e65aec12 | 501 | { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, |
635f7d8c | 502 | { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true }, |
f2d88580 | 503 | { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true }, |
ede89845 | 504 | { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, |
635f7d8c | 505 | { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, |
9b634ea5 | 506 | #ifdef HAVE_SELINUX |
b4c59701 LP |
507 | { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */ |
508 | { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ | |
9b634ea5 | 509 | #endif |
88213476 LP |
510 | }; |
511 | ||
512 | unsigned k; | |
513 | int r = 0; | |
514 | ||
515 | for (k = 0; k < ELEMENTSOF(mount_table); k++) { | |
7fd1b19b | 516 | _cleanup_free_ char *where = NULL; |
d002827b | 517 | #ifdef HAVE_SELINUX |
a8828ed9 | 518 | _cleanup_free_ char *options = NULL; |
d002827b LP |
519 | #endif |
520 | const char *o; | |
88213476 LP |
521 | int t; |
522 | ||
17fe0523 LP |
523 | where = strjoin(dest, "/", mount_table[k].where, NULL); |
524 | if (!where) | |
525 | return log_oom(); | |
88213476 | 526 | |
e65aec12 | 527 | t = path_is_mount_point(where, true); |
68fb0892 | 528 | if (t < 0) { |
88213476 | 529 | log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t)); |
88213476 LP |
530 | |
531 | if (r == 0) | |
532 | r = t; | |
533 | ||
534 | continue; | |
535 | } | |
536 | ||
9c1c7f71 LP |
537 | /* Skip this entry if it is not a remount. */ |
538 | if (mount_table[k].what && t > 0) | |
014a9c77 LP |
539 | continue; |
540 | ||
17fe0523 | 541 | mkdir_p(where, 0755); |
88213476 | 542 | |
a8828ed9 | 543 | #ifdef HAVE_SELINUX |
82adf6af LP |
544 | if (arg_selinux_apifs_context && |
545 | (streq_ptr(mount_table[k].what, "tmpfs") || streq_ptr(mount_table[k].what, "devpts"))) { | |
546 | options = strjoin(mount_table[k].options, ",context=\"", arg_selinux_apifs_context, "\"", NULL); | |
d002827b LP |
547 | if (!options) |
548 | return log_oom(); | |
549 | ||
550 | o = options; | |
551 | } else | |
a8828ed9 | 552 | #endif |
d002827b | 553 | o = mount_table[k].options; |
a8828ed9 | 554 | |
a8828ed9 | 555 | |
88213476 LP |
556 | if (mount(mount_table[k].what, |
557 | where, | |
558 | mount_table[k].type, | |
559 | mount_table[k].flags, | |
d002827b | 560 | o) < 0 && |
3bd66c05 | 561 | mount_table[k].fatal) { |
88213476 LP |
562 | |
563 | log_error("mount(%s) failed: %m", where); | |
564 | ||
565 | if (r == 0) | |
566 | r = -errno; | |
567 | } | |
88213476 LP |
568 | } |
569 | ||
e58a1277 LP |
570 | return r; |
571 | } | |
f8440af5 | 572 | |
17fe0523 LP |
573 | static int mount_binds(const char *dest, char **l, unsigned long flags) { |
574 | char **x, **y; | |
575 | ||
576 | STRV_FOREACH_PAIR(x, y, l) { | |
2ed4e5e0 | 577 | char *where; |
d2421337 | 578 | struct stat source_st, dest_st; |
2ed4e5e0 | 579 | int r; |
d2421337 DR |
580 | |
581 | if (stat(*x, &source_st) < 0) { | |
582 | log_error("failed to stat %s: %m", *x); | |
583 | return -errno; | |
584 | } | |
17fe0523 | 585 | |
2ed4e5e0 SL |
586 | where = strappenda(dest, *y); |
587 | r = stat(where, &dest_st); | |
588 | if (r == 0) { | |
d2421337 | 589 | if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) { |
cecf24e7 | 590 | log_error("The file types of %s and %s do not match. Refusing bind mount", |
d2421337 DR |
591 | *x, where); |
592 | return -EINVAL; | |
593 | } | |
2ed4e5e0 SL |
594 | } else if (errno == ENOENT) { |
595 | r = mkdir_parents_label(where, 0755); | |
596 | if (r < 0) { | |
597 | log_error("Failed to bind mount %s: %s", *x, strerror(-r)); | |
598 | return r; | |
d2421337 | 599 | } |
2ed4e5e0 SL |
600 | } else { |
601 | log_error("Failed to bind mount %s: %s", *x, strerror(errno)); | |
602 | return -errno; | |
603 | } | |
604 | /* Create the mount point, but be conservative -- refuse to create block | |
605 | * and char devices. */ | |
606 | if (S_ISDIR(source_st.st_mode)) | |
607 | mkdir_label(where, 0755); | |
608 | else if (S_ISFIFO(source_st.st_mode)) | |
609 | mkfifo(where, 0644); | |
610 | else if (S_ISSOCK(source_st.st_mode)) | |
611 | mknod(where, 0644 | S_IFSOCK, 0); | |
612 | else if (S_ISREG(source_st.st_mode)) | |
613 | touch(where); | |
614 | else { | |
615 | log_error("Refusing to create mountpoint for file: %s", *x); | |
616 | return -ENOTSUP; | |
d2421337 | 617 | } |
17fe0523 LP |
618 | |
619 | if (mount(*x, where, "bind", MS_BIND, NULL) < 0) { | |
620 | log_error("mount(%s) failed: %m", where); | |
621 | return -errno; | |
622 | } | |
623 | ||
624 | if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) { | |
625 | log_error("mount(%s) failed: %m", where); | |
626 | return -errno; | |
627 | } | |
628 | } | |
629 | ||
630 | return 0; | |
631 | } | |
632 | ||
e58a1277 | 633 | static int setup_timezone(const char *dest) { |
d4036145 LP |
634 | _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL; |
635 | char *z, *y; | |
636 | int r; | |
f8440af5 | 637 | |
e58a1277 LP |
638 | assert(dest); |
639 | ||
640 | /* Fix the timezone, if possible */ | |
d4036145 LP |
641 | r = readlink_malloc("/etc/localtime", &p); |
642 | if (r < 0) { | |
643 | log_warning("/etc/localtime is not a symlink, not updating container timezone."); | |
644 | return 0; | |
645 | } | |
646 | ||
647 | z = path_startswith(p, "../usr/share/zoneinfo/"); | |
648 | if (!z) | |
649 | z = path_startswith(p, "/usr/share/zoneinfo/"); | |
650 | if (!z) { | |
651 | log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone."); | |
652 | return 0; | |
653 | } | |
654 | ||
04bc4a3f LP |
655 | where = strappend(dest, "/etc/localtime"); |
656 | if (!where) | |
0d0f0c50 | 657 | return log_oom(); |
715ac17a | 658 | |
d4036145 LP |
659 | r = readlink_malloc(where, &q); |
660 | if (r >= 0) { | |
661 | y = path_startswith(q, "../usr/share/zoneinfo/"); | |
662 | if (!y) | |
663 | y = path_startswith(q, "/usr/share/zoneinfo/"); | |
4d1c38b8 | 664 | |
4d1c38b8 | 665 | |
d4036145 LP |
666 | /* Already pointing to the right place? Then do nothing .. */ |
667 | if (y && streq(y, z)) | |
668 | return 0; | |
669 | } | |
670 | ||
671 | check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL); | |
672 | if (!check) | |
0d0f0c50 | 673 | return log_oom(); |
4d1c38b8 | 674 | |
d4036145 LP |
675 | if (access(check, F_OK) < 0) { |
676 | log_warning("Timezone %s does not exist in container, not updating container timezone.", z); | |
677 | return 0; | |
678 | } | |
68fb0892 | 679 | |
d4036145 LP |
680 | what = strappend("../usr/share/zoneinfo/", z); |
681 | if (!what) | |
682 | return log_oom(); | |
683 | ||
684 | unlink(where); | |
685 | if (symlink(what, where) < 0) { | |
686 | log_error("Failed to correct timezone of container: %m"); | |
687 | return 0; | |
688 | } | |
e58a1277 LP |
689 | |
690 | return 0; | |
88213476 LP |
691 | } |
692 | ||
2547bb41 | 693 | static int setup_resolv_conf(const char *dest) { |
f333fbb1 | 694 | char _cleanup_free_ *where = NULL; |
2547bb41 LP |
695 | |
696 | assert(dest); | |
697 | ||
698 | if (arg_private_network) | |
699 | return 0; | |
700 | ||
701 | /* Fix resolv.conf, if possible */ | |
04bc4a3f LP |
702 | where = strappend(dest, "/etc/resolv.conf"); |
703 | if (!where) | |
0d0f0c50 | 704 | return log_oom(); |
2547bb41 | 705 | |
77e63faf LP |
706 | /* We don't really care for the results of this really. If it |
707 | * fails, it fails, but meh... */ | |
51045322 | 708 | copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW); |
2547bb41 LP |
709 | |
710 | return 0; | |
711 | } | |
712 | ||
04bc4a3f | 713 | static int setup_boot_id(const char *dest) { |
7fd1b19b | 714 | _cleanup_free_ char *from = NULL, *to = NULL; |
04bc4a3f LP |
715 | sd_id128_t rnd; |
716 | char as_uuid[37]; | |
717 | int r; | |
718 | ||
719 | assert(dest); | |
720 | ||
eb91eb18 LP |
721 | if (arg_share_system) |
722 | return 0; | |
723 | ||
04bc4a3f LP |
724 | /* Generate a new randomized boot ID, so that each boot-up of |
725 | * the container gets a new one */ | |
726 | ||
727 | from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id"); | |
04bc4a3f | 728 | to = strappend(dest, "/proc/sys/kernel/random/boot_id"); |
ed8b7a3e ZJS |
729 | if (!from || !to) |
730 | return log_oom(); | |
04bc4a3f LP |
731 | |
732 | r = sd_id128_randomize(&rnd); | |
733 | if (r < 0) { | |
734 | log_error("Failed to generate random boot id: %s", strerror(-r)); | |
ed8b7a3e | 735 | return r; |
04bc4a3f LP |
736 | } |
737 | ||
738 | snprintf(as_uuid, sizeof(as_uuid), | |
739 | "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", | |
740 | SD_ID128_FORMAT_VAL(rnd)); | |
741 | char_array_0(as_uuid); | |
742 | ||
574d5f2d | 743 | r = write_string_file(from, as_uuid); |
04bc4a3f LP |
744 | if (r < 0) { |
745 | log_error("Failed to write boot id: %s", strerror(-r)); | |
ed8b7a3e | 746 | return r; |
04bc4a3f LP |
747 | } |
748 | ||
749 | if (mount(from, to, "bind", MS_BIND, NULL) < 0) { | |
750 | log_error("Failed to bind mount boot id: %m"); | |
751 | r = -errno; | |
10d18763 ZJS |
752 | } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL)) |
753 | log_warning("Failed to make boot id read-only: %m"); | |
04bc4a3f LP |
754 | |
755 | unlink(from); | |
04bc4a3f LP |
756 | return r; |
757 | } | |
758 | ||
e58a1277 | 759 | static int copy_devnodes(const char *dest) { |
88213476 LP |
760 | |
761 | static const char devnodes[] = | |
762 | "null\0" | |
763 | "zero\0" | |
764 | "full\0" | |
765 | "random\0" | |
766 | "urandom\0" | |
f2d88580 | 767 | "tty\0"; |
88213476 LP |
768 | |
769 | const char *d; | |
e58a1277 | 770 | int r = 0; |
7fd1b19b | 771 | _cleanup_umask_ mode_t u; |
a258bf26 LP |
772 | |
773 | assert(dest); | |
124640f1 LP |
774 | |
775 | u = umask(0000); | |
88213476 LP |
776 | |
777 | NULSTR_FOREACH(d, devnodes) { | |
7fd1b19b | 778 | _cleanup_free_ char *from = NULL, *to = NULL; |
7f112f50 | 779 | struct stat st; |
88213476 | 780 | |
7f112f50 LP |
781 | from = strappend("/dev/", d); |
782 | to = strjoin(dest, "/dev/", d, NULL); | |
783 | if (!from || !to) | |
784 | return log_oom(); | |
88213476 LP |
785 | |
786 | if (stat(from, &st) < 0) { | |
787 | ||
788 | if (errno != ENOENT) { | |
789 | log_error("Failed to stat %s: %m", from); | |
7f112f50 | 790 | return -errno; |
88213476 LP |
791 | } |
792 | ||
a258bf26 | 793 | } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) { |
88213476 | 794 | |
ed8b7a3e | 795 | log_error("%s is not a char or block device, cannot copy", from); |
7f112f50 | 796 | return -EIO; |
a258bf26 LP |
797 | |
798 | } else if (mknod(to, st.st_mode, st.st_rdev) < 0) { | |
799 | ||
800 | log_error("mknod(%s) failed: %m", dest); | |
7f112f50 | 801 | return -errno; |
88213476 | 802 | } |
88213476 LP |
803 | } |
804 | ||
e58a1277 LP |
805 | return r; |
806 | } | |
88213476 | 807 | |
f2d88580 LP |
808 | static int setup_ptmx(const char *dest) { |
809 | _cleanup_free_ char *p = NULL; | |
810 | ||
811 | p = strappend(dest, "/dev/ptmx"); | |
812 | if (!p) | |
813 | return log_oom(); | |
814 | ||
815 | if (symlink("pts/ptmx", p) < 0) { | |
816 | log_error("Failed to create /dev/ptmx symlink: %m"); | |
817 | return -errno; | |
818 | } | |
819 | ||
820 | return 0; | |
821 | } | |
822 | ||
e58a1277 LP |
823 | static int setup_dev_console(const char *dest, const char *console) { |
824 | struct stat st; | |
7fd1b19b | 825 | _cleanup_free_ char *to = NULL; |
e58a1277 | 826 | int r; |
7fd1b19b | 827 | _cleanup_umask_ mode_t u; |
e58a1277 LP |
828 | |
829 | assert(dest); | |
830 | assert(console); | |
831 | ||
832 | u = umask(0000); | |
833 | ||
834 | if (stat(console, &st) < 0) { | |
835 | log_error("Failed to stat %s: %m", console); | |
25ea79fe | 836 | return -errno; |
88213476 | 837 | |
a258bf26 | 838 | } else if (!S_ISCHR(st.st_mode)) { |
25ea79fe ZJS |
839 | log_error("/dev/console is not a char device"); |
840 | return -EIO; | |
e58a1277 | 841 | } |
88213476 | 842 | |
e58a1277 LP |
843 | r = chmod_and_chown(console, 0600, 0, 0); |
844 | if (r < 0) { | |
845 | log_error("Failed to correct access mode for TTY: %s", strerror(-r)); | |
25ea79fe | 846 | return r; |
a258bf26 | 847 | } |
88213476 | 848 | |
25ea79fe ZJS |
849 | if (asprintf(&to, "%s/dev/console", dest) < 0) |
850 | return log_oom(); | |
88213476 | 851 | |
a258bf26 LP |
852 | /* We need to bind mount the right tty to /dev/console since |
853 | * ptys can only exist on pts file systems. To have something | |
854 | * to bind mount things on we create a device node first, that | |
855 | * has the right major/minor (note that the major minor | |
856 | * doesn't actually matter here, since we mount it over | |
857 | * anyway). */ | |
858 | ||
e58a1277 LP |
859 | if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { |
860 | log_error("mknod() for /dev/console failed: %m"); | |
25ea79fe | 861 | return -errno; |
e58a1277 | 862 | } |
a258bf26 LP |
863 | |
864 | if (mount(console, to, "bind", MS_BIND, NULL) < 0) { | |
e58a1277 | 865 | log_error("Bind mount for /dev/console failed: %m"); |
25ea79fe | 866 | return -errno; |
a258bf26 LP |
867 | } |
868 | ||
25ea79fe | 869 | return 0; |
e58a1277 LP |
870 | } |
871 | ||
872 | static int setup_kmsg(const char *dest, int kmsg_socket) { | |
7fd1b19b | 873 | _cleanup_free_ char *from = NULL, *to = NULL; |
e58a1277 | 874 | int r, fd, k; |
7fd1b19b | 875 | _cleanup_umask_ mode_t u; |
e58a1277 LP |
876 | union { |
877 | struct cmsghdr cmsghdr; | |
878 | uint8_t buf[CMSG_SPACE(sizeof(int))]; | |
b92bea5d ZJS |
879 | } control = {}; |
880 | struct msghdr mh = { | |
881 | .msg_control = &control, | |
882 | .msg_controllen = sizeof(control), | |
883 | }; | |
e58a1277 LP |
884 | struct cmsghdr *cmsg; |
885 | ||
886 | assert(dest); | |
887 | assert(kmsg_socket >= 0); | |
a258bf26 | 888 | |
e58a1277 | 889 | u = umask(0000); |
a258bf26 | 890 | |
f1e5dfe2 LP |
891 | /* We create the kmsg FIFO as /dev/kmsg, but immediately |
892 | * delete it after bind mounting it to /proc/kmsg. While FIFOs | |
893 | * on the reading side behave very similar to /proc/kmsg, | |
894 | * their writing side behaves differently from /dev/kmsg in | |
895 | * that writing blocks when nothing is reading. In order to | |
896 | * avoid any problems with containers deadlocking due to this | |
897 | * we simply make /dev/kmsg unavailable to the container. */ | |
25ea79fe ZJS |
898 | if (asprintf(&from, "%s/dev/kmsg", dest) < 0 || |
899 | asprintf(&to, "%s/proc/kmsg", dest) < 0) | |
900 | return log_oom(); | |
e58a1277 LP |
901 | |
902 | if (mkfifo(from, 0600) < 0) { | |
903 | log_error("mkfifo() for /dev/kmsg failed: %m"); | |
25ea79fe | 904 | return -errno; |
e58a1277 LP |
905 | } |
906 | ||
907 | r = chmod_and_chown(from, 0600, 0, 0); | |
908 | if (r < 0) { | |
909 | log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r)); | |
25ea79fe | 910 | return r; |
e58a1277 LP |
911 | } |
912 | ||
913 | if (mount(from, to, "bind", MS_BIND, NULL) < 0) { | |
914 | log_error("Bind mount for /proc/kmsg failed: %m"); | |
25ea79fe | 915 | return -errno; |
e58a1277 LP |
916 | } |
917 | ||
918 | fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC); | |
919 | if (fd < 0) { | |
920 | log_error("Failed to open fifo: %m"); | |
25ea79fe | 921 | return -errno; |
e58a1277 LP |
922 | } |
923 | ||
e58a1277 LP |
924 | cmsg = CMSG_FIRSTHDR(&mh); |
925 | cmsg->cmsg_level = SOL_SOCKET; | |
926 | cmsg->cmsg_type = SCM_RIGHTS; | |
927 | cmsg->cmsg_len = CMSG_LEN(sizeof(int)); | |
928 | memcpy(CMSG_DATA(cmsg), &fd, sizeof(int)); | |
929 | ||
930 | mh.msg_controllen = cmsg->cmsg_len; | |
931 | ||
932 | /* Store away the fd in the socket, so that it stays open as | |
933 | * long as we run the child */ | |
934 | k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL); | |
935 | close_nointr_nofail(fd); | |
936 | ||
937 | if (k < 0) { | |
938 | log_error("Failed to send FIFO fd: %m"); | |
25ea79fe | 939 | return -errno; |
a258bf26 LP |
940 | } |
941 | ||
f1e5dfe2 LP |
942 | /* And now make the FIFO unavailable as /dev/kmsg... */ |
943 | unlink(from); | |
25ea79fe | 944 | return 0; |
88213476 LP |
945 | } |
946 | ||
3a74cea5 | 947 | static int setup_hostname(void) { |
3a74cea5 | 948 | |
eb91eb18 LP |
949 | if (arg_share_system) |
950 | return 0; | |
951 | ||
7027ff61 LP |
952 | if (sethostname(arg_machine, strlen(arg_machine)) < 0) |
953 | return -errno; | |
3a74cea5 | 954 | |
7027ff61 | 955 | return 0; |
3a74cea5 LP |
956 | } |
957 | ||
57fb9fb5 | 958 | static int setup_journal(const char *directory) { |
4d680aee | 959 | sd_id128_t machine_id, this_id; |
7fd1b19b | 960 | _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL; |
27407a01 | 961 | char *id; |
57fb9fb5 LP |
962 | int r; |
963 | ||
57fb9fb5 | 964 | p = strappend(directory, "/etc/machine-id"); |
27407a01 ZJS |
965 | if (!p) |
966 | return log_oom(); | |
57fb9fb5 LP |
967 | |
968 | r = read_one_line_file(p, &b); | |
27407a01 ZJS |
969 | if (r == -ENOENT && arg_link_journal == LINK_AUTO) |
970 | return 0; | |
971 | else if (r < 0) { | |
972 | log_error("Failed to read machine ID from %s: %s", p, strerror(-r)); | |
57fb9fb5 LP |
973 | return r; |
974 | } | |
975 | ||
27407a01 ZJS |
976 | id = strstrip(b); |
977 | if (isempty(id) && arg_link_journal == LINK_AUTO) | |
978 | return 0; | |
57fb9fb5 | 979 | |
27407a01 ZJS |
980 | /* Verify validity */ |
981 | r = sd_id128_from_string(id, &machine_id); | |
57fb9fb5 | 982 | if (r < 0) { |
27407a01 ZJS |
983 | log_error("Failed to parse machine ID from %s: %s", p, strerror(-r)); |
984 | return r; | |
57fb9fb5 LP |
985 | } |
986 | ||
4d680aee ZJS |
987 | r = sd_id128_get_machine(&this_id); |
988 | if (r < 0) { | |
989 | log_error("Failed to retrieve machine ID: %s", strerror(-r)); | |
990 | return r; | |
991 | } | |
992 | ||
993 | if (sd_id128_equal(machine_id, this_id)) { | |
994 | log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR, | |
995 | "Host and machine ids are equal (%s): refusing to link journals", id); | |
996 | if (arg_link_journal == LINK_AUTO) | |
997 | return 0; | |
998 | return | |
999 | -EEXIST; | |
1000 | } | |
1001 | ||
1002 | if (arg_link_journal == LINK_NO) | |
1003 | return 0; | |
1004 | ||
57fb9fb5 | 1005 | free(p); |
27407a01 ZJS |
1006 | p = strappend("/var/log/journal/", id); |
1007 | q = strjoin(directory, "/var/log/journal/", id, NULL); | |
1008 | if (!p || !q) | |
1009 | return log_oom(); | |
1010 | ||
1011 | if (path_is_mount_point(p, false) > 0) { | |
1012 | if (arg_link_journal != LINK_AUTO) { | |
1013 | log_error("%s: already a mount point, refusing to use for journal", p); | |
1014 | return -EEXIST; | |
1015 | } | |
1016 | ||
1017 | return 0; | |
57fb9fb5 LP |
1018 | } |
1019 | ||
27407a01 | 1020 | if (path_is_mount_point(q, false) > 0) { |
57fb9fb5 | 1021 | if (arg_link_journal != LINK_AUTO) { |
27407a01 ZJS |
1022 | log_error("%s: already a mount point, refusing to use for journal", q); |
1023 | return -EEXIST; | |
57fb9fb5 LP |
1024 | } |
1025 | ||
27407a01 | 1026 | return 0; |
57fb9fb5 LP |
1027 | } |
1028 | ||
1029 | r = readlink_and_make_absolute(p, &d); | |
1030 | if (r >= 0) { | |
1031 | if ((arg_link_journal == LINK_GUEST || | |
1032 | arg_link_journal == LINK_AUTO) && | |
1033 | path_equal(d, q)) { | |
1034 | ||
27407a01 ZJS |
1035 | r = mkdir_p(q, 0755); |
1036 | if (r < 0) | |
1037 | log_warning("failed to create directory %s: %m", q); | |
1038 | return 0; | |
57fb9fb5 LP |
1039 | } |
1040 | ||
1041 | if (unlink(p) < 0) { | |
1042 | log_error("Failed to remove symlink %s: %m", p); | |
27407a01 | 1043 | return -errno; |
57fb9fb5 LP |
1044 | } |
1045 | } else if (r == -EINVAL) { | |
1046 | ||
1047 | if (arg_link_journal == LINK_GUEST && | |
1048 | rmdir(p) < 0) { | |
1049 | ||
27407a01 ZJS |
1050 | if (errno == ENOTDIR) { |
1051 | log_error("%s already exists and is neither a symlink nor a directory", p); | |
1052 | return r; | |
1053 | } else { | |
57fb9fb5 | 1054 | log_error("Failed to remove %s: %m", p); |
27407a01 | 1055 | return -errno; |
57fb9fb5 | 1056 | } |
57fb9fb5 LP |
1057 | } |
1058 | } else if (r != -ENOENT) { | |
1059 | log_error("readlink(%s) failed: %m", p); | |
27407a01 | 1060 | return r; |
57fb9fb5 LP |
1061 | } |
1062 | ||
1063 | if (arg_link_journal == LINK_GUEST) { | |
1064 | ||
1065 | if (symlink(q, p) < 0) { | |
1066 | log_error("Failed to symlink %s to %s: %m", q, p); | |
27407a01 | 1067 | return -errno; |
57fb9fb5 LP |
1068 | } |
1069 | ||
27407a01 ZJS |
1070 | r = mkdir_p(q, 0755); |
1071 | if (r < 0) | |
1072 | log_warning("failed to create directory %s: %m", q); | |
1073 | return 0; | |
57fb9fb5 LP |
1074 | } |
1075 | ||
1076 | if (arg_link_journal == LINK_HOST) { | |
1077 | r = mkdir_p(p, 0755); | |
1078 | if (r < 0) { | |
1079 | log_error("Failed to create %s: %m", p); | |
27407a01 | 1080 | return r; |
57fb9fb5 LP |
1081 | } |
1082 | ||
27407a01 ZJS |
1083 | } else if (access(p, F_OK) < 0) |
1084 | return 0; | |
57fb9fb5 LP |
1085 | |
1086 | if (dir_is_empty(q) == 0) { | |
1087 | log_error("%s not empty.", q); | |
27407a01 | 1088 | return -ENOTEMPTY; |
57fb9fb5 LP |
1089 | } |
1090 | ||
1091 | r = mkdir_p(q, 0755); | |
1092 | if (r < 0) { | |
1093 | log_error("Failed to create %s: %m", q); | |
27407a01 | 1094 | return r; |
57fb9fb5 LP |
1095 | } |
1096 | ||
1097 | if (mount(p, q, "bind", MS_BIND, NULL) < 0) { | |
1098 | log_error("Failed to bind mount journal from host into guest: %m"); | |
27407a01 | 1099 | return -errno; |
57fb9fb5 LP |
1100 | } |
1101 | ||
27407a01 | 1102 | return 0; |
57fb9fb5 LP |
1103 | } |
1104 | ||
9bd37b40 LP |
1105 | static int setup_kdbus(const char *dest, const char *path) { |
1106 | const char *p; | |
1107 | ||
1108 | if (!path) | |
1109 | return 0; | |
1110 | ||
1111 | p = strappenda(dest, "/dev/kdbus"); | |
1112 | if (mkdir(p, 0755) < 0) { | |
1113 | log_error("Failed to create kdbus path: %m"); | |
1114 | return -errno; | |
1115 | } | |
1116 | ||
1117 | if (mount(path, p, "bind", MS_BIND, NULL) < 0) { | |
486e99a3 | 1118 | log_error("Failed to mount kdbus domain path: %m"); |
9bd37b40 LP |
1119 | return -errno; |
1120 | } | |
1121 | ||
1122 | return 0; | |
1123 | } | |
1124 | ||
88213476 | 1125 | static int drop_capabilities(void) { |
5076f0cc | 1126 | return capability_bounding_set_drop(~arg_retain, false); |
88213476 LP |
1127 | } |
1128 | ||
354bfd2b | 1129 | static int register_machine(pid_t pid) { |
9444b1f2 LP |
1130 | _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; |
1131 | _cleanup_bus_unref_ sd_bus *bus = NULL; | |
1132 | int r; | |
1133 | ||
eb91eb18 LP |
1134 | if (!arg_register) |
1135 | return 0; | |
1136 | ||
1c03020c | 1137 | r = sd_bus_default_system(&bus); |
9444b1f2 LP |
1138 | if (r < 0) { |
1139 | log_error("Failed to open system bus: %s", strerror(-r)); | |
1140 | return r; | |
1141 | } | |
1142 | ||
89f7c846 LP |
1143 | if (arg_keep_unit) { |
1144 | r = sd_bus_call_method( | |
1145 | bus, | |
1146 | "org.freedesktop.machine1", | |
1147 | "/org/freedesktop/machine1", | |
1148 | "org.freedesktop.machine1.Manager", | |
1149 | "RegisterMachine", | |
1150 | &error, | |
1151 | NULL, | |
1152 | "sayssus", | |
1153 | arg_machine, | |
1154 | SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), | |
1155 | "nspawn", | |
1156 | "container", | |
1157 | (uint32_t) pid, | |
1158 | strempty(arg_directory)); | |
1159 | } else { | |
1160 | r = sd_bus_call_method( | |
1161 | bus, | |
1162 | "org.freedesktop.machine1", | |
1163 | "/org/freedesktop/machine1", | |
1164 | "org.freedesktop.machine1.Manager", | |
1165 | "CreateMachine", | |
1166 | &error, | |
1167 | NULL, | |
1168 | "sayssusa(sv)", | |
1169 | arg_machine, | |
1170 | SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), | |
1171 | "nspawn", | |
1172 | "container", | |
1173 | (uint32_t) pid, | |
1174 | strempty(arg_directory), | |
1175 | !isempty(arg_slice), "Slice", "s", arg_slice); | |
1176 | } | |
1177 | ||
9444b1f2 | 1178 | if (r < 0) { |
1f0cd86b LP |
1179 | log_error("Failed to register machine: %s", bus_error_message(&error, r)); |
1180 | return r; | |
1181 | } | |
1182 | ||
1183 | return 0; | |
1184 | } | |
1185 | ||
1186 | static int terminate_machine(pid_t pid) { | |
1187 | _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; | |
1188 | _cleanup_bus_message_unref_ sd_bus_message *reply = NULL; | |
1189 | _cleanup_bus_unref_ sd_bus *bus = NULL; | |
1190 | const char *path; | |
1191 | int r; | |
1192 | ||
eb91eb18 LP |
1193 | if (!arg_register) |
1194 | return 0; | |
1195 | ||
76b54375 | 1196 | r = sd_bus_default_system(&bus); |
1f0cd86b LP |
1197 | if (r < 0) { |
1198 | log_error("Failed to open system bus: %s", strerror(-r)); | |
1199 | return r; | |
1200 | } | |
1201 | ||
1202 | r = sd_bus_call_method( | |
1203 | bus, | |
1204 | "org.freedesktop.machine1", | |
1205 | "/org/freedesktop/machine1", | |
1206 | "org.freedesktop.machine1.Manager", | |
1207 | "GetMachineByPID", | |
1208 | &error, | |
1209 | &reply, | |
1210 | "u", | |
1211 | (uint32_t) pid); | |
1212 | if (r < 0) { | |
1213 | /* Note that the machine might already have been | |
1214 | * cleaned up automatically, hence don't consider it a | |
1215 | * failure if we cannot get the machine object. */ | |
1216 | log_debug("Failed to get machine: %s", bus_error_message(&error, r)); | |
1217 | return 0; | |
1218 | } | |
1219 | ||
1220 | r = sd_bus_message_read(reply, "o", &path); | |
5b30bef8 LP |
1221 | if (r < 0) |
1222 | return bus_log_parse_error(r); | |
9444b1f2 | 1223 | |
1f0cd86b LP |
1224 | r = sd_bus_call_method( |
1225 | bus, | |
1226 | "org.freedesktop.machine1", | |
1227 | path, | |
1228 | "org.freedesktop.machine1.Machine", | |
1229 | "Terminate", | |
1230 | &error, | |
1231 | NULL, | |
1232 | NULL); | |
1233 | if (r < 0) { | |
1234 | log_debug("Failed to terminate machine: %s", bus_error_message(&error, r)); | |
1235 | return 0; | |
1236 | } | |
1237 | ||
9444b1f2 LP |
1238 | return 0; |
1239 | } | |
1240 | ||
db999e0f LP |
1241 | static int reset_audit_loginuid(void) { |
1242 | _cleanup_free_ char *p = NULL; | |
1243 | int r; | |
1244 | ||
1245 | if (arg_share_system) | |
1246 | return 0; | |
1247 | ||
1248 | r = read_one_line_file("/proc/self/loginuid", &p); | |
1249 | if (r == -EEXIST) | |
1250 | return 0; | |
1251 | if (r < 0) { | |
1252 | log_error("Failed to read /proc/self/loginuid: %s", strerror(-r)); | |
1253 | return r; | |
1254 | } | |
1255 | ||
1256 | /* Already reset? */ | |
1257 | if (streq(p, "4294967295")) | |
1258 | return 0; | |
1259 | ||
1260 | r = write_string_file("/proc/self/loginuid", "4294967295"); | |
1261 | if (r < 0) { | |
1262 | log_error("Failed to reset audit login UID. This probably means that your kernel is too\n" | |
1263 | "old and you have audit enabled. Note that the auditing subsystem is known to\n" | |
1264 | "be incompatible with containers on old kernels. Please make sure to upgrade\n" | |
1265 | "your kernel or to off auditing with 'audit=0' on the kernel command line before\n" | |
1266 | "using systemd-nspawn. Sleeping for 5s... (%s)\n", strerror(-r)); | |
77b6e194 | 1267 | |
db999e0f | 1268 | sleep(5); |
77b6e194 | 1269 | } |
db999e0f LP |
1270 | |
1271 | return 0; | |
77b6e194 LP |
1272 | } |
1273 | ||
69c79d3c LP |
1274 | static int setup_veth(int netns_fd) { |
1275 | _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; | |
cf6a8911 | 1276 | _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; |
69c79d3c LP |
1277 | char iface_name[IFNAMSIZ] = "ve-"; |
1278 | int r; | |
1279 | ||
1280 | if (!arg_private_network) | |
1281 | return 0; | |
1282 | ||
1283 | if (!arg_network_veth) | |
1284 | return 0; | |
1285 | ||
1286 | strncpy(iface_name+3, arg_machine, sizeof(iface_name) - 3); | |
1287 | ||
1288 | r = sd_rtnl_open(0, &rtnl); | |
1289 | if (r < 0) { | |
1290 | log_error("Failed to connect to netlink: %s", strerror(-r)); | |
1291 | return r; | |
1292 | } | |
1293 | ||
1294 | r = sd_rtnl_message_new_link(RTM_NEWLINK, 0, &m); | |
1295 | if (r < 0) { | |
1296 | log_error("Failed to allocate netlink message: %s", strerror(-r)); | |
1297 | return r; | |
1298 | } | |
1299 | ||
1300 | r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0"); | |
1301 | if (r < 0) { | |
1302 | log_error("Failed to append netlink kind: %s", strerror(-r)); | |
1303 | return r; | |
1304 | } | |
1305 | ||
1306 | r = sd_rtnl_message_open_container(m, IFLA_LINKINFO, 0); | |
1307 | if (r < 0) { | |
1308 | log_error("Failed to open netlink container: %s", strerror(-r)); | |
1309 | return r; | |
1310 | } | |
1311 | ||
1312 | r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth"); | |
1313 | if (r < 0) { | |
1314 | log_error("Failed to append netlink kind: %s", strerror(-r)); | |
1315 | return r; | |
1316 | } | |
1317 | ||
1318 | r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA, 0); | |
1319 | if (r < 0) { | |
1320 | log_error("Failed to open netlink container: %s", strerror(-r)); | |
1321 | return r; | |
1322 | } | |
1323 | ||
1324 | r = sd_rtnl_message_open_container(m, VETH_INFO_PEER, sizeof(struct ifinfomsg)); | |
1325 | if (r < 0) { | |
1326 | log_error("z Failed to open netlink container: %s", strerror(-r)); | |
1327 | return r; | |
1328 | } | |
1329 | ||
1330 | r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name); | |
1331 | if (r < 0) { | |
1332 | log_error("Failed to append netlink kind: %s", strerror(-r)); | |
1333 | return r; | |
1334 | } | |
1335 | ||
1336 | r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_FD, netns_fd); | |
1337 | if (r < 0) { | |
1338 | log_error("Failed to add netlink namespace field: %s", strerror(-r)); | |
1339 | return r; | |
1340 | } | |
1341 | ||
1342 | r = sd_rtnl_message_close_container(m); | |
1343 | if (r < 0) { | |
1344 | log_error("Failed to close netlink container: %s", strerror(-r)); | |
1345 | return r; | |
1346 | } | |
1347 | ||
1348 | r = sd_rtnl_message_close_container(m); | |
1349 | if (r < 0) { | |
1350 | log_error("Failed to close netlink container: %s", strerror(-r)); | |
1351 | return r; | |
1352 | } | |
1353 | ||
1354 | r = sd_rtnl_message_close_container(m); | |
1355 | if (r < 0) { | |
1356 | log_error("Failed to close netlink container: %s", strerror(-r)); | |
1357 | return r; | |
1358 | } | |
1359 | ||
1360 | r = sd_rtnl_call(rtnl, m, 0, NULL); | |
1361 | if (r < 0) { | |
1362 | log_error("Failed to add new veth interfaces: %s", strerror(-r)); | |
1363 | return r; | |
1364 | } | |
1365 | ||
1366 | return 0; | |
1367 | } | |
1368 | ||
1369 | static int move_network_interfaces(pid_t pid) { | |
7e227024 | 1370 | _cleanup_udev_unref_ struct udev *udev = NULL; |
69c79d3c | 1371 | _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; |
aa28aefe LP |
1372 | char **i; |
1373 | int r; | |
1374 | ||
1375 | if (!arg_private_network) | |
1376 | return 0; | |
1377 | ||
1378 | if (strv_isempty(arg_network_interfaces)) | |
1379 | return 0; | |
1380 | ||
b88eb17a | 1381 | r = sd_rtnl_open(0, &rtnl); |
aa28aefe LP |
1382 | if (r < 0) { |
1383 | log_error("Failed to connect to netlink: %s", strerror(-r)); | |
1384 | return r; | |
1385 | } | |
1386 | ||
7e227024 LP |
1387 | udev = udev_new(); |
1388 | if (!udev) { | |
1389 | log_error("Failed to connect to udev."); | |
1390 | return -ENOMEM; | |
1391 | } | |
1392 | ||
aa28aefe | 1393 | STRV_FOREACH(i, arg_network_interfaces) { |
cf6a8911 | 1394 | _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; |
7e227024 LP |
1395 | _cleanup_udev_device_unref_ struct udev_device *d = NULL; |
1396 | char ifi_str[2 + DECIMAL_STR_MAX(int)]; | |
b88eb17a | 1397 | int ifi; |
aa28aefe | 1398 | |
b88eb17a LP |
1399 | ifi = (int) if_nametoindex(*i); |
1400 | if (ifi <= 0) { | |
aa28aefe LP |
1401 | log_error("Failed to resolve interface %s: %m", *i); |
1402 | return -errno; | |
1403 | } | |
1404 | ||
7e227024 LP |
1405 | sprintf(ifi_str, "n%i", ifi); |
1406 | d = udev_device_new_from_device_id(udev, ifi_str); | |
1407 | if (!d) { | |
1408 | log_error("Failed to get udev device for interface %s: %m", *i); | |
1409 | return -errno; | |
1410 | } | |
1411 | ||
1412 | if (udev_device_get_is_initialized(d) <= 0) { | |
1413 | log_error("Network interface %s is not initialized yet.", *i); | |
1414 | return -EBUSY; | |
1415 | } | |
1416 | ||
d595c5cc | 1417 | r = sd_rtnl_message_new_link(RTM_NEWLINK, ifi, &m); |
aa28aefe LP |
1418 | if (r < 0) { |
1419 | log_error("Failed to allocate netlink message: %s", strerror(-r)); | |
1420 | return r; | |
1421 | } | |
1422 | ||
1423 | r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); | |
1424 | if (r < 0) { | |
1425 | log_error("Failed to append namespace PID to netlink message: %s", strerror(-r)); | |
1426 | return r; | |
1427 | } | |
1428 | ||
1429 | r = sd_rtnl_call(rtnl, m, 0, NULL); | |
1430 | if (r < 0) { | |
b88eb17a | 1431 | log_error("Failed to move interface %s to namespace: %s", *i, strerror(-r)); |
aa28aefe LP |
1432 | return r; |
1433 | } | |
1434 | } | |
1435 | ||
1436 | return 0; | |
1437 | } | |
1438 | ||
24fb1112 LP |
1439 | static int audit_still_doesnt_work_in_containers(void) { |
1440 | ||
1441 | #ifdef HAVE_SECCOMP | |
1442 | scmp_filter_ctx seccomp; | |
1443 | int r; | |
1444 | ||
1445 | /* | |
1446 | Audit is broken in containers, much of the userspace audit | |
1447 | hookup will fail if running inside a container. We don't | |
1448 | care and just turn off creation of audit sockets. | |
1449 | ||
1450 | This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail | |
1451 | with EAFNOSUPPORT which audit userspace uses as indication | |
1452 | that audit is disabled in the kernel. | |
1453 | */ | |
1454 | ||
1455 | seccomp = seccomp_init(SCMP_ACT_ALLOW); | |
1456 | if (!seccomp) | |
1457 | return log_oom(); | |
1458 | ||
1459 | r = seccomp_rule_add_exact( | |
1460 | seccomp, | |
1461 | SCMP_ACT_ERRNO(EAFNOSUPPORT), | |
1462 | SCMP_SYS(socket), | |
1463 | 2, | |
1464 | SCMP_A0(SCMP_CMP_EQ, AF_NETLINK), | |
1465 | SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT)); | |
1466 | if (r < 0) { | |
1467 | log_error("Failed to add audit seccomp rule: %s", strerror(-r)); | |
1468 | goto finish; | |
1469 | } | |
1470 | ||
1471 | r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); | |
1472 | if (r < 0) { | |
1473 | log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r)); | |
1474 | goto finish; | |
1475 | } | |
1476 | ||
1477 | r = seccomp_load(seccomp); | |
1478 | if (r < 0) | |
1479 | log_error("Failed to install seccomp audit filter: %s", strerror(-r)); | |
1480 | ||
1481 | finish: | |
1482 | seccomp_release(seccomp); | |
1483 | return r; | |
1484 | #else | |
1485 | return 0; | |
1486 | #endif | |
1487 | ||
1488 | } | |
1489 | ||
88213476 | 1490 | int main(int argc, char *argv[]) { |
69c79d3c LP |
1491 | |
1492 | _cleanup_close_ int master = -1, kdbus_fd = -1, sync_fd = -1, netns_fd = -1; | |
1493 | _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 }; | |
1494 | _cleanup_free_ char *kdbus_domain = NULL; | |
1495 | _cleanup_fdset_free_ FDSet *fds = NULL; | |
1496 | const char *console = NULL; | |
04d391da | 1497 | int r = EXIT_FAILURE, k; |
7027ff61 | 1498 | int n_fd_passed; |
69c79d3c | 1499 | pid_t pid = 0; |
a258bf26 | 1500 | sigset_t mask; |
88213476 LP |
1501 | |
1502 | log_parse_environment(); | |
1503 | log_open(); | |
1504 | ||
05947bef LP |
1505 | k = parse_argv(argc, argv); |
1506 | if (k < 0) | |
88213476 | 1507 | goto finish; |
05947bef LP |
1508 | else if (k == 0) { |
1509 | r = EXIT_SUCCESS; | |
1510 | goto finish; | |
1511 | } | |
88213476 LP |
1512 | |
1513 | if (arg_directory) { | |
1514 | char *p; | |
1515 | ||
1516 | p = path_make_absolute_cwd(arg_directory); | |
1517 | free(arg_directory); | |
1518 | arg_directory = p; | |
1519 | } else | |
1520 | arg_directory = get_current_dir_name(); | |
1521 | ||
1522 | if (!arg_directory) { | |
a383724e | 1523 | log_error("Failed to determine path, please use -D."); |
88213476 LP |
1524 | goto finish; |
1525 | } | |
1526 | ||
1527 | path_kill_slashes(arg_directory); | |
1528 | ||
7027ff61 | 1529 | if (!arg_machine) { |
2b6bf07d | 1530 | arg_machine = strdup(basename(arg_directory)); |
7027ff61 LP |
1531 | if (!arg_machine) { |
1532 | log_oom(); | |
1533 | goto finish; | |
1534 | } | |
1535 | ||
e724b063 | 1536 | hostname_cleanup(arg_machine, false); |
7027ff61 LP |
1537 | if (isempty(arg_machine)) { |
1538 | log_error("Failed to determine machine name automatically, please use -M."); | |
1539 | goto finish; | |
1540 | } | |
1541 | } | |
1542 | ||
88213476 LP |
1543 | if (geteuid() != 0) { |
1544 | log_error("Need to be root."); | |
1545 | goto finish; | |
1546 | } | |
1547 | ||
04d391da LP |
1548 | if (sd_booted() <= 0) { |
1549 | log_error("Not running on a systemd system."); | |
1550 | goto finish; | |
1551 | } | |
1552 | ||
88213476 | 1553 | if (path_equal(arg_directory, "/")) { |
6df6b939 | 1554 | log_error("Spawning container on root directory not supported."); |
88213476 LP |
1555 | goto finish; |
1556 | } | |
1557 | ||
6b9132a9 LP |
1558 | if (arg_boot) { |
1559 | if (path_is_os_tree(arg_directory) <= 0) { | |
1560 | log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); | |
1561 | goto finish; | |
1562 | } | |
1563 | } else { | |
1564 | const char *p; | |
1565 | ||
1566 | p = strappenda(arg_directory, | |
1567 | argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/"); | |
1568 | if (access(p, F_OK) < 0) { | |
1569 | log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory); | |
1570 | goto finish; | |
1571 | ||
1572 | } | |
88213476 LP |
1573 | } |
1574 | ||
842f3b0f LP |
1575 | log_close(); |
1576 | n_fd_passed = sd_listen_fds(false); | |
1577 | if (n_fd_passed > 0) { | |
1578 | k = fdset_new_listen_fds(&fds, false); | |
1579 | if (k < 0) { | |
1580 | log_error("Failed to collect file descriptors: %s", strerror(-k)); | |
1581 | goto finish; | |
1582 | } | |
1583 | } | |
1584 | fdset_close_others(fds); | |
1585 | log_open(); | |
1586 | ||
db7feb7e LP |
1587 | master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); |
1588 | if (master < 0) { | |
a258bf26 LP |
1589 | log_error("Failed to acquire pseudo tty: %m"); |
1590 | goto finish; | |
1591 | } | |
1592 | ||
db7feb7e LP |
1593 | console = ptsname(master); |
1594 | if (!console) { | |
a258bf26 LP |
1595 | log_error("Failed to determine tty name: %m"); |
1596 | goto finish; | |
1597 | } | |
1598 | ||
284c0b91 LP |
1599 | if (!arg_quiet) |
1600 | log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_directory); | |
a258bf26 LP |
1601 | |
1602 | if (unlockpt(master) < 0) { | |
1603 | log_error("Failed to unlock tty: %m"); | |
1604 | goto finish; | |
1605 | } | |
1606 | ||
69c79d3c LP |
1607 | if (arg_network_veth) { |
1608 | netns_fd = open("/proc/self/ns/net", O_RDWR|O_CLOEXEC); | |
1609 | if (netns_fd < 0) { | |
1610 | log_error("Failed to open network namespace fd: %m"); | |
1611 | goto finish; | |
1612 | } | |
1613 | } | |
eb91eb18 LP |
1614 | |
1615 | if (access("/dev/kdbus/control", F_OK) >= 0) { | |
1616 | ||
1617 | if (arg_share_system) { | |
1618 | kdbus_domain = strdup("/dev/kdbus"); | |
1619 | if (!kdbus_domain) { | |
1620 | log_oom(); | |
1621 | goto finish; | |
1622 | } | |
1623 | } else { | |
1624 | const char *ns; | |
1625 | ||
1626 | ns = strappenda("machine-", arg_machine); | |
1627 | kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain); | |
1628 | if (r < 0) | |
1629 | log_debug("Failed to create kdbus domain: %s", strerror(-r)); | |
1630 | else | |
1631 | log_debug("Successfully created kdbus domain as %s", kdbus_domain); | |
1632 | } | |
1633 | } | |
9bd37b40 | 1634 | |
e58a1277 | 1635 | if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { |
354bfd2b LP |
1636 | log_error("Failed to create kmsg socket pair: %m"); |
1637 | goto finish; | |
1638 | } | |
1639 | ||
05947bef LP |
1640 | sd_notify(0, "READY=1"); |
1641 | ||
a258bf26 LP |
1642 | assert_se(sigemptyset(&mask) == 0); |
1643 | sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); | |
1644 | assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); | |
1645 | ||
d87be9b0 LP |
1646 | for (;;) { |
1647 | siginfo_t status; | |
a383724e | 1648 | |
40ddbdf8 LP |
1649 | sync_fd = eventfd(0, EFD_CLOEXEC); |
1650 | if (sync_fd < 0) { | |
1651 | log_error("Failed to create event fd: %m"); | |
1652 | goto finish; | |
1653 | } | |
1654 | ||
8a96d94e LP |
1655 | pid = syscall(__NR_clone, |
1656 | SIGCHLD|CLONE_NEWNS| | |
1657 | (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| | |
1658 | (arg_private_network ? CLONE_NEWNET : 0), NULL); | |
d87be9b0 LP |
1659 | if (pid < 0) { |
1660 | if (errno == EINVAL) | |
1661 | log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); | |
1662 | else | |
1663 | log_error("clone() failed: %m"); | |
a258bf26 | 1664 | |
d87be9b0 LP |
1665 | goto finish; |
1666 | } | |
a258bf26 | 1667 | |
d87be9b0 LP |
1668 | if (pid == 0) { |
1669 | /* child */ | |
d87be9b0 LP |
1670 | const char *home = NULL; |
1671 | uid_t uid = (uid_t) -1; | |
1672 | gid_t gid = (gid_t) -1; | |
5674767e | 1673 | unsigned n_env = 2; |
d87be9b0 | 1674 | const char *envp[] = { |
e10a55fd | 1675 | "PATH=" DEFAULT_PATH_SPLIT_USR, |
d87be9b0 LP |
1676 | "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ |
1677 | NULL, /* TERM */ | |
1678 | NULL, /* HOME */ | |
1679 | NULL, /* USER */ | |
1680 | NULL, /* LOGNAME */ | |
1681 | NULL, /* container_uuid */ | |
842f3b0f LP |
1682 | NULL, /* LISTEN_FDS */ |
1683 | NULL, /* LISTEN_PID */ | |
d87be9b0 LP |
1684 | NULL |
1685 | }; | |
f4889f65 | 1686 | char **env_use; |
354bfd2b | 1687 | eventfd_t x; |
a258bf26 | 1688 | |
5674767e ZJS |
1689 | envp[n_env] = strv_find_prefix(environ, "TERM="); |
1690 | if (envp[n_env]) | |
1691 | n_env ++; | |
a258bf26 | 1692 | |
d87be9b0 | 1693 | close_nointr_nofail(master); |
842f3b0f | 1694 | master = -1; |
a258bf26 | 1695 | |
d87be9b0 LP |
1696 | close_nointr(STDIN_FILENO); |
1697 | close_nointr(STDOUT_FILENO); | |
1698 | close_nointr(STDERR_FILENO); | |
db7feb7e | 1699 | |
842f3b0f LP |
1700 | close_nointr_nofail(kmsg_socket_pair[0]); |
1701 | kmsg_socket_pair[0] = -1; | |
a258bf26 | 1702 | |
d87be9b0 | 1703 | reset_all_signal_handlers(); |
88213476 | 1704 | |
d87be9b0 LP |
1705 | assert_se(sigemptyset(&mask) == 0); |
1706 | assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); | |
f5c1b9ee | 1707 | |
842f3b0f LP |
1708 | k = open_terminal(console, O_RDWR); |
1709 | if (k != STDIN_FILENO) { | |
1710 | if (k >= 0) { | |
1711 | close_nointr_nofail(k); | |
1712 | k = -EINVAL; | |
1713 | } | |
1714 | ||
1715 | log_error("Failed to open console: %s", strerror(-k)); | |
1716 | goto child_fail; | |
1717 | } | |
1718 | ||
1719 | if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || | |
1720 | dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { | |
1721 | log_error("Failed to duplicate console: %m"); | |
d87be9b0 | 1722 | goto child_fail; |
842f3b0f | 1723 | } |
bc2f673e | 1724 | |
d87be9b0 LP |
1725 | if (setsid() < 0) { |
1726 | log_error("setsid() failed: %m"); | |
bc2f673e LP |
1727 | goto child_fail; |
1728 | } | |
1729 | ||
db999e0f LP |
1730 | if (reset_audit_loginuid() < 0) |
1731 | goto child_fail; | |
1732 | ||
d87be9b0 LP |
1733 | if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { |
1734 | log_error("PR_SET_PDEATHSIG failed: %m"); | |
1735 | goto child_fail; | |
1736 | } | |
e58a1277 | 1737 | |
d87be9b0 LP |
1738 | /* Mark everything as slave, so that we still |
1739 | * receive mounts from the real root, but don't | |
1740 | * propagate mounts to the real root. */ | |
1741 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
1742 | log_error("MS_SLAVE|MS_REC failed: %m"); | |
1743 | goto child_fail; | |
1744 | } | |
04bc4a3f | 1745 | |
d87be9b0 LP |
1746 | /* Turn directory into bind mount */ |
1747 | if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) { | |
1748 | log_error("Failed to make bind mount."); | |
1749 | goto child_fail; | |
1750 | } | |
88213476 | 1751 | |
d87be9b0 LP |
1752 | if (arg_read_only) |
1753 | if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) { | |
1754 | log_error("Failed to make read-only."); | |
1755 | goto child_fail; | |
1756 | } | |
2547bb41 | 1757 | |
d87be9b0 LP |
1758 | if (mount_all(arg_directory) < 0) |
1759 | goto child_fail; | |
57fb9fb5 | 1760 | |
d87be9b0 LP |
1761 | if (copy_devnodes(arg_directory) < 0) |
1762 | goto child_fail; | |
a258bf26 | 1763 | |
f2d88580 LP |
1764 | if (setup_ptmx(arg_directory) < 0) |
1765 | goto child_fail; | |
1766 | ||
d87be9b0 | 1767 | dev_setup(arg_directory); |
88213476 | 1768 | |
69c79d3c LP |
1769 | if (setup_veth(netns_fd) < 0) |
1770 | goto child_fail; | |
1771 | ||
1772 | if (netns_fd >= 0) { | |
1773 | close_nointr_nofail(netns_fd); | |
1774 | netns_fd = -1; | |
1775 | } | |
1776 | ||
24fb1112 LP |
1777 | if (audit_still_doesnt_work_in_containers() < 0) |
1778 | goto child_fail; | |
1779 | ||
d87be9b0 LP |
1780 | if (setup_dev_console(arg_directory, console) < 0) |
1781 | goto child_fail; | |
88213476 | 1782 | |
d87be9b0 LP |
1783 | if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) |
1784 | goto child_fail; | |
88213476 | 1785 | |
d87be9b0 | 1786 | close_nointr_nofail(kmsg_socket_pair[1]); |
842f3b0f | 1787 | kmsg_socket_pair[1] = -1; |
a258bf26 | 1788 | |
d87be9b0 LP |
1789 | if (setup_boot_id(arg_directory) < 0) |
1790 | goto child_fail; | |
a41fe3a2 | 1791 | |
d87be9b0 LP |
1792 | if (setup_timezone(arg_directory) < 0) |
1793 | goto child_fail; | |
88213476 | 1794 | |
d87be9b0 LP |
1795 | if (setup_resolv_conf(arg_directory) < 0) |
1796 | goto child_fail; | |
687d0825 | 1797 | |
d87be9b0 | 1798 | if (setup_journal(arg_directory) < 0) |
687d0825 | 1799 | goto child_fail; |
687d0825 | 1800 | |
17fe0523 LP |
1801 | if (mount_binds(arg_directory, arg_bind, 0) < 0) |
1802 | goto child_fail; | |
1803 | ||
1804 | if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) | |
1805 | goto child_fail; | |
1806 | ||
486e99a3 | 1807 | if (setup_kdbus(arg_directory, kdbus_domain) < 0) |
9bd37b40 LP |
1808 | goto child_fail; |
1809 | ||
d87be9b0 LP |
1810 | if (chdir(arg_directory) < 0) { |
1811 | log_error("chdir(%s) failed: %m", arg_directory); | |
687d0825 MV |
1812 | goto child_fail; |
1813 | } | |
1814 | ||
d87be9b0 LP |
1815 | if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) { |
1816 | log_error("mount(MS_MOVE) failed: %m"); | |
687d0825 MV |
1817 | goto child_fail; |
1818 | } | |
1819 | ||
d87be9b0 LP |
1820 | if (chroot(".") < 0) { |
1821 | log_error("chroot() failed: %m"); | |
687d0825 MV |
1822 | goto child_fail; |
1823 | } | |
1824 | ||
d87be9b0 LP |
1825 | if (chdir("/") < 0) { |
1826 | log_error("chdir() failed: %m"); | |
687d0825 MV |
1827 | goto child_fail; |
1828 | } | |
1829 | ||
d87be9b0 LP |
1830 | umask(0022); |
1831 | ||
eb91eb18 LP |
1832 | if (arg_private_network) |
1833 | loopback_setup(); | |
d87be9b0 LP |
1834 | |
1835 | if (drop_capabilities() < 0) { | |
1836 | log_error("drop_capabilities() failed: %m"); | |
687d0825 MV |
1837 | goto child_fail; |
1838 | } | |
687d0825 | 1839 | |
d87be9b0 LP |
1840 | if (arg_user) { |
1841 | ||
963ddb91 LP |
1842 | /* Note that this resolves user names |
1843 | * inside the container, and hence | |
1844 | * accesses the NSS modules from the | |
1845 | * container and not the host. This is | |
1846 | * a bit weird... */ | |
1847 | ||
d87be9b0 LP |
1848 | if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { |
1849 | log_error("get_user_creds() failed: %m"); | |
1850 | goto child_fail; | |
1851 | } | |
1852 | ||
1853 | if (mkdir_parents_label(home, 0775) < 0) { | |
1854 | log_error("mkdir_parents_label() failed: %m"); | |
1855 | goto child_fail; | |
1856 | } | |
1857 | ||
1858 | if (mkdir_safe_label(home, 0775, uid, gid) < 0) { | |
1859 | log_error("mkdir_safe_label() failed: %m"); | |
1860 | goto child_fail; | |
1861 | } | |
1862 | ||
1863 | if (initgroups((const char*)arg_user, gid) < 0) { | |
1864 | log_error("initgroups() failed: %m"); | |
1865 | goto child_fail; | |
1866 | } | |
144f0fc0 | 1867 | |
d87be9b0 LP |
1868 | if (setresgid(gid, gid, gid) < 0) { |
1869 | log_error("setregid() failed: %m"); | |
1870 | goto child_fail; | |
1871 | } | |
1872 | ||
1873 | if (setresuid(uid, uid, uid) < 0) { | |
1874 | log_error("setreuid() failed: %m"); | |
1875 | goto child_fail; | |
1876 | } | |
3c957acf LP |
1877 | } else { |
1878 | /* Reset everything fully to 0, just in case */ | |
1879 | ||
1880 | if (setgroups(0, NULL) < 0) { | |
1881 | log_error("setgroups() failed: %m"); | |
1882 | goto child_fail; | |
1883 | } | |
1884 | ||
1885 | if (setresgid(0, 0, 0) < 0) { | |
1886 | log_error("setregid() failed: %m"); | |
1887 | goto child_fail; | |
1888 | } | |
1889 | ||
1890 | if (setresuid(0, 0, 0) < 0) { | |
1891 | log_error("setreuid() failed: %m"); | |
1892 | goto child_fail; | |
1893 | } | |
d87be9b0 LP |
1894 | } |
1895 | ||
842f3b0f LP |
1896 | if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || |
1897 | (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || | |
1898 | (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { | |
0d0f0c50 | 1899 | log_oom(); |
144f0fc0 LP |
1900 | goto child_fail; |
1901 | } | |
687d0825 | 1902 | |
9444b1f2 LP |
1903 | if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) { |
1904 | if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) { | |
842f3b0f LP |
1905 | log_oom(); |
1906 | goto child_fail; | |
1907 | } | |
1908 | } | |
1909 | ||
1910 | if (fdset_size(fds) > 0) { | |
1911 | k = fdset_cloexec(fds, false); | |
1912 | if (k < 0) { | |
1913 | log_error("Failed to unset O_CLOEXEC for file descriptors."); | |
1914 | goto child_fail; | |
1915 | } | |
1916 | ||
1917 | if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) || | |
d1826146 | 1918 | (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) { |
d87be9b0 LP |
1919 | log_oom(); |
1920 | goto child_fail; | |
1921 | } | |
1922 | } | |
1923 | ||
1924 | setup_hostname(); | |
1925 | ||
354bfd2b LP |
1926 | eventfd_read(sync_fd, &x); |
1927 | close_nointr_nofail(sync_fd); | |
1928 | sync_fd = -1; | |
1929 | ||
f4889f65 LP |
1930 | if (!strv_isempty(arg_setenv)) { |
1931 | char **n; | |
1932 | ||
1933 | n = strv_env_merge(2, envp, arg_setenv); | |
1934 | if (!n) { | |
1935 | log_oom(); | |
1936 | goto child_fail; | |
1937 | } | |
1938 | ||
1939 | env_use = n; | |
1940 | } else | |
1941 | env_use = (char**) envp; | |
1942 | ||
5d63309c | 1943 | #ifdef HAVE_SELINUX |
82adf6af LP |
1944 | if (arg_selinux_context) |
1945 | if (setexeccon(arg_selinux_context) < 0) | |
1946 | log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); | |
a8828ed9 | 1947 | #endif |
d87be9b0 LP |
1948 | if (arg_boot) { |
1949 | char **a; | |
1950 | size_t l; | |
88213476 | 1951 | |
d87be9b0 | 1952 | /* Automatically search for the init system */ |
0f0dbc46 | 1953 | |
d87be9b0 LP |
1954 | l = 1 + argc - optind; |
1955 | a = newa(char*, l + 1); | |
1956 | memcpy(a + 1, argv + optind, l * sizeof(char*)); | |
0f0dbc46 | 1957 | |
d87be9b0 | 1958 | a[0] = (char*) "/usr/lib/systemd/systemd"; |
f4889f65 | 1959 | execve(a[0], a, env_use); |
0f0dbc46 | 1960 | |
d87be9b0 | 1961 | a[0] = (char*) "/lib/systemd/systemd"; |
f4889f65 | 1962 | execve(a[0], a, env_use); |
0f0dbc46 | 1963 | |
d87be9b0 | 1964 | a[0] = (char*) "/sbin/init"; |
f4889f65 | 1965 | execve(a[0], a, env_use); |
d87be9b0 | 1966 | } else if (argc > optind) |
f4889f65 | 1967 | execvpe(argv[optind], argv + optind, env_use); |
d87be9b0 LP |
1968 | else { |
1969 | chdir(home ? home : "/root"); | |
f4889f65 | 1970 | execle("/bin/bash", "-bash", NULL, env_use); |
262d10e6 | 1971 | execle("/bin/sh", "-sh", NULL, env_use); |
d87be9b0 LP |
1972 | } |
1973 | ||
1974 | log_error("execv() failed: %m"); | |
0f0dbc46 | 1975 | |
d87be9b0 LP |
1976 | child_fail: |
1977 | _exit(EXIT_FAILURE); | |
da5b3bad | 1978 | } |
88213476 | 1979 | |
842f3b0f LP |
1980 | fdset_free(fds); |
1981 | fds = NULL; | |
1982 | ||
354bfd2b LP |
1983 | r = register_machine(pid); |
1984 | if (r < 0) | |
1985 | goto finish; | |
1986 | ||
aa28aefe LP |
1987 | r = move_network_interfaces(pid); |
1988 | if (r < 0) | |
1989 | goto finish; | |
1990 | ||
354bfd2b LP |
1991 | eventfd_write(sync_fd, 1); |
1992 | close_nointr_nofail(sync_fd); | |
1993 | sync_fd = -1; | |
1994 | ||
04d39279 LP |
1995 | k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); |
1996 | if (k < 0) { | |
1997 | r = EXIT_FAILURE; | |
1998 | break; | |
1999 | } | |
88213476 | 2000 | |
284c0b91 LP |
2001 | if (!arg_quiet) |
2002 | putc('\n', stdout); | |
04d39279 LP |
2003 | |
2004 | /* Kill if it is not dead yet anyway */ | |
1f0cd86b LP |
2005 | terminate_machine(pid); |
2006 | ||
2007 | /* Redundant, but better safe than sorry */ | |
04d39279 | 2008 | kill(pid, SIGKILL); |
a258bf26 | 2009 | |
05947bef | 2010 | k = wait_for_terminate(pid, &status); |
04d39279 LP |
2011 | pid = 0; |
2012 | ||
05947bef | 2013 | if (k < 0) { |
d87be9b0 LP |
2014 | r = EXIT_FAILURE; |
2015 | break; | |
2016 | } | |
a258bf26 | 2017 | |
d87be9b0 | 2018 | if (status.si_code == CLD_EXITED) { |
a5f5f8a0 | 2019 | r = status.si_status; |
d87be9b0 | 2020 | if (status.si_status != 0) { |
04d39279 | 2021 | log_error("Container %s failed with error code %i.", arg_machine, status.si_status); |
d87be9b0 LP |
2022 | break; |
2023 | } | |
2024 | ||
284c0b91 LP |
2025 | if (!arg_quiet) |
2026 | log_debug("Container %s exited successfully.", arg_machine); | |
d87be9b0 LP |
2027 | break; |
2028 | } else if (status.si_code == CLD_KILLED && | |
2029 | status.si_status == SIGINT) { | |
284c0b91 LP |
2030 | |
2031 | if (!arg_quiet) | |
2032 | log_info("Container %s has been shut down.", arg_machine); | |
d87be9b0 LP |
2033 | r = 0; |
2034 | break; | |
2035 | } else if (status.si_code == CLD_KILLED && | |
2036 | status.si_status == SIGHUP) { | |
284c0b91 LP |
2037 | |
2038 | if (!arg_quiet) | |
2039 | log_info("Container %s is being rebooted.", arg_machine); | |
d87be9b0 LP |
2040 | continue; |
2041 | } else if (status.si_code == CLD_KILLED || | |
2042 | status.si_code == CLD_DUMPED) { | |
88213476 | 2043 | |
eb91eb18 | 2044 | log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); |
d87be9b0 LP |
2045 | r = EXIT_FAILURE; |
2046 | break; | |
2047 | } else { | |
04d39279 | 2048 | log_error("Container %s failed due to unknown reason.", arg_machine); |
d87be9b0 LP |
2049 | r = EXIT_FAILURE; |
2050 | break; | |
2051 | } | |
2052 | } | |
88213476 LP |
2053 | |
2054 | finish: | |
9444b1f2 LP |
2055 | if (pid > 0) |
2056 | kill(pid, SIGKILL); | |
88213476 | 2057 | |
04d391da | 2058 | free(arg_directory); |
7027ff61 | 2059 | free(arg_machine); |
f4889f65 | 2060 | free(arg_setenv); |
aa28aefe | 2061 | free(arg_network_interfaces); |
88213476 LP |
2062 | |
2063 | return r; | |
2064 | } |