]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/core/namespace.c
man: fix typo in systemd.unit man page
[thirdparty/systemd.git] / src / core / namespace.c
CommitLineData
53e1b683 1/* SPDX-License-Identifier: LGPL-2.1+ */
15ae422b
LP
2
3#include <errno.h>
e08f94ac 4#include <linux/loop.h>
07630cea 5#include <sched.h>
15ae422b 6#include <stdio.h>
07630cea 7#include <sys/mount.h>
07630cea 8#include <unistd.h>
25e870b5 9#include <linux/fs.h>
15ae422b 10
b5efdb8a 11#include "alloc-util.h"
10404d52 12#include "base-filesystem.h"
7f112f50 13#include "dev-setup.h"
3ffd4af2 14#include "fd-util.h"
e5f10caf 15#include "format-util.h"
d944dc95 16#include "fs-util.h"
e908468b 17#include "label.h"
915e6d16 18#include "loop-util.h"
07630cea 19#include "loopback-setup.h"
07630cea 20#include "mkdir.h"
4349cd7c 21#include "mount-util.h"
049af8ad 22#include "mountpoint-util.h"
0cb8e3d1 23#include "namespace-util.h"
3ffd4af2 24#include "namespace.h"
d8b4d14d 25#include "nulstr-util.h"
07630cea 26#include "path-util.h"
d7b8eec7 27#include "selinux-util.h"
2583fbea 28#include "socket-util.h"
760877e9 29#include "sort-util.h"
36ce7110 30#include "stat-util.h"
8b43440b 31#include "string-table.h"
07630cea
LP
32#include "string-util.h"
33#include "strv.h"
a652f050 34#include "tmpfile-util.h"
affb60b1 35#include "umask-util.h"
ee104e11 36#include "user-util.h"
15ae422b 37
737ba3c8 38#define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
39
c17ec25e 40typedef enum MountMode {
15ae422b
LP
41 /* This is ordered by priority! */
42 INACCESSIBLE,
d2d6c096
LP
43 BIND_MOUNT,
44 BIND_MOUNT_RECURSIVE,
ac0930c8 45 PRIVATE_TMP,
7f112f50 46 PRIVATE_DEV,
5d997827 47 BIND_DEV,
6c47cd7d 48 EMPTY_DIR,
5d997827
LP
49 SYSFS,
50 PROCFS,
51 READONLY,
59eeb84b 52 READWRITE,
2abd4e38 53 TMPFS,
1e05071d 54 READWRITE_IMPLICIT, /* Should have the lowest priority. */
5beb8688 55 _MOUNT_MODE_MAX,
c17ec25e 56} MountMode;
15ae422b 57
34de407a 58typedef struct MountEntry {
5327c910 59 const char *path_const; /* Memory allocated on stack or static */
cfbeb4ef 60 MountMode mode:5;
5327c910
LP
61 bool ignore:1; /* Ignore if path does not exist? */
62 bool has_prefix:1; /* Already is prefixed by the root dir? */
cfbeb4ef 63 bool read_only:1; /* Shall this mount point be read-only? */
9ce4e4b0 64 bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */
088696fe 65 bool applied:1; /* Already applied */
55fe7432 66 char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */
d2d6c096
LP
67 const char *source_const; /* The source path, for bind mounts */
68 char *source_malloc;
2abd4e38
YW
69 const char *options_const;/* Mount options for tmpfs */
70 char *options_malloc;
71 unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
088696fe 72 unsigned n_followed;
34de407a 73} MountEntry;
15ae422b 74
5d997827 75/* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
3fe91079 76 * something there already. These mounts are hence overridden by any other explicitly configured mounts. */
5d997827
LP
77static const MountEntry apivfs_table[] = {
78 { "/proc", PROCFS, false },
79 { "/dev", BIND_DEV, false },
80 { "/sys", SYSFS, false },
81};
f471b2af 82
11a30cec 83/* ProtectKernelTunables= option and the related filesystem APIs */
34de407a 84static const MountEntry protect_kernel_tunables_table[] = {
1e05071d
YW
85 { "/proc/acpi", READONLY, true },
86 { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */
87 { "/proc/asound", READONLY, true },
88 { "/proc/bus", READONLY, true },
89 { "/proc/fs", READONLY, true },
90 { "/proc/irq", READONLY, true },
91 { "/proc/kallsyms", INACCESSIBLE, true },
92 { "/proc/kcore", INACCESSIBLE, true },
93 { "/proc/latency_stats", READONLY, true },
94 { "/proc/mtrr", READONLY, true },
95 { "/proc/scsi", READONLY, true },
96 { "/proc/sys", READONLY, false },
97 { "/proc/sysrq-trigger", READONLY, true },
98 { "/proc/timer_stats", READONLY, true },
99 { "/sys", READONLY, false },
100 { "/sys/fs/bpf", READONLY, true },
101 { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */
102 { "/sys/fs/selinux", READWRITE_IMPLICIT, true },
103 { "/sys/kernel/debug", READONLY, true },
104 { "/sys/kernel/tracing", READONLY, true },
11a30cec
DH
105};
106
c575770b 107/* ProtectKernelModules= option */
34de407a 108static const MountEntry protect_kernel_modules_table[] = {
349cc4a5 109#if HAVE_SPLIT_USR
c6232fb0 110 { "/lib/modules", INACCESSIBLE, true },
c575770b 111#endif
c6232fb0 112 { "/usr/lib/modules", INACCESSIBLE, true },
c575770b
DH
113};
114
94a7b275
KK
115/* ProtectKernelLogs= option */
116static const MountEntry protect_kernel_logs_table[] = {
117 { "/proc/kmsg", INACCESSIBLE, true },
118 { "/dev/kmsg", INACCESSIBLE, true },
119};
120
b6c432ca
DH
121/*
122 * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
123 * system should be protected by ProtectSystem=
124 */
34de407a 125static const MountEntry protect_home_read_only_table[] = {
c6232fb0
LP
126 { "/home", READONLY, true },
127 { "/run/user", READONLY, true },
128 { "/root", READONLY, true },
b6c432ca
DH
129};
130
e4da7d8c
YW
131/* ProtectHome=tmpfs table */
132static const MountEntry protect_home_tmpfs_table[] = {
133 { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME },
134 { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME },
135 { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME },
136};
137
b6c432ca 138/* ProtectHome=yes table */
34de407a 139static const MountEntry protect_home_yes_table[] = {
c6232fb0
LP
140 { "/home", INACCESSIBLE, true },
141 { "/run/user", INACCESSIBLE, true },
142 { "/root", INACCESSIBLE, true },
b6c432ca
DH
143};
144
f471b2af 145/* ProtectSystem=yes table */
34de407a 146static const MountEntry protect_system_yes_table[] = {
c6232fb0
LP
147 { "/usr", READONLY, false },
148 { "/boot", READONLY, true },
149 { "/efi", READONLY, true },
7486f305
AB
150#if HAVE_SPLIT_USR
151 { "/lib", READONLY, true },
152 { "/lib64", READONLY, true },
153 { "/bin", READONLY, true },
671f0f8d 154# if HAVE_SPLIT_BIN
7486f305 155 { "/sbin", READONLY, true },
671f0f8d 156# endif
7486f305 157#endif
f471b2af
DH
158};
159
160/* ProtectSystem=full includes ProtectSystem=yes */
34de407a 161static const MountEntry protect_system_full_table[] = {
c6232fb0
LP
162 { "/usr", READONLY, false },
163 { "/boot", READONLY, true },
164 { "/efi", READONLY, true },
165 { "/etc", READONLY, false },
7486f305
AB
166#if HAVE_SPLIT_USR
167 { "/lib", READONLY, true },
168 { "/lib64", READONLY, true },
169 { "/bin", READONLY, true },
671f0f8d 170# if HAVE_SPLIT_BIN
7486f305 171 { "/sbin", READONLY, true },
671f0f8d 172# endif
7486f305 173#endif
f471b2af
DH
174};
175
176/*
177 * ProtectSystem=strict table. In this strict mode, we mount everything
178 * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
179 * which are left writable, but PrivateDevices= + ProtectKernelTunables=
180 * protect those, and these options should be fully orthogonal.
181 * (And of course /home and friends are also left writable, as ProtectHome=
182 * shall manage those, orthogonally).
183 */
34de407a 184static const MountEntry protect_system_strict_table[] = {
1e05071d
YW
185 { "/", READONLY, false },
186 { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */
187 { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */
188 { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */
189 { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */
190 { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */
191 { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */
f471b2af
DH
192};
193
5beb8688
YW
194static const char * const mount_mode_table[_MOUNT_MODE_MAX] = {
195 [INACCESSIBLE] = "inaccessible",
196 [BIND_MOUNT] = "bind",
197 [BIND_MOUNT_RECURSIVE] = "rbind",
198 [PRIVATE_TMP] = "private-tmp",
199 [PRIVATE_DEV] = "private-dev",
200 [BIND_DEV] = "bind-dev",
201 [EMPTY_DIR] = "empty",
202 [SYSFS] = "sysfs",
203 [PROCFS] = "procfs",
204 [READONLY] = "read-only",
205 [READWRITE] = "read-write",
206 [TMPFS] = "tmpfs",
207 [READWRITE_IMPLICIT] = "rw-implicit",
208};
209
210DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode);
211
34de407a 212static const char *mount_entry_path(const MountEntry *p) {
f0a4feb0
DH
213 assert(p);
214
5327c910
LP
215 /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
216 * otherwise the stack/static ->path field is returned. */
f0a4feb0 217
5327c910 218 return p->path_malloc ?: p->path_const;
f0a4feb0
DH
219}
220
34de407a 221static bool mount_entry_read_only(const MountEntry *p) {
cfbeb4ef
LP
222 assert(p);
223
224 return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE);
225}
226
d2d6c096
LP
227static const char *mount_entry_source(const MountEntry *p) {
228 assert(p);
229
230 return p->source_malloc ?: p->source_const;
231}
232
2abd4e38
YW
233static const char *mount_entry_options(const MountEntry *p) {
234 assert(p);
235
236 return p->options_malloc ?: p->options_const;
237}
238
1eb7e08e
LP
239static void mount_entry_done(MountEntry *p) {
240 assert(p);
241
242 p->path_malloc = mfree(p->path_malloc);
243 p->source_malloc = mfree(p->source_malloc);
2abd4e38 244 p->options_malloc = mfree(p->options_malloc);
1eb7e08e
LP
245}
246
d18aff04 247static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) {
15ae422b
LP
248 char **i;
249
613b411c
LP
250 assert(p);
251
1e05071d 252 /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */
5327c910 253
15ae422b 254 STRV_FOREACH(i, strv) {
5327c910
LP
255 bool ignore = false, needs_prefix = false;
256 const char *e = *i;
15ae422b 257
5327c910
LP
258 /* Look for any prefixes */
259 if (startswith(e, "-")) {
260 e++;
9c94d52e 261 ignore = true;
ea92ae33 262 }
5327c910
LP
263 if (startswith(e, "+")) {
264 e++;
265 needs_prefix = true;
266 }
ea92ae33 267
baaa35ad
ZJS
268 if (!path_is_absolute(e))
269 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
270 "Path is not absolute: %s", e);
15ae422b 271
34de407a 272 *((*p)++) = (MountEntry) {
5327c910
LP
273 .path_const = e,
274 .mode = mode,
275 .ignore = ignore,
d18aff04 276 .has_prefix = !needs_prefix && !forcibly_require_prefix,
5327c910 277 };
15ae422b
LP
278 }
279
280 return 0;
281}
282
6c47cd7d
LP
283static int append_empty_dir_mounts(MountEntry **p, char **strv) {
284 char **i;
285
286 assert(p);
287
288 /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the
289 * "/private/" boundary directories for DynamicUser=1. */
290
291 STRV_FOREACH(i, strv) {
292
293 *((*p)++) = (MountEntry) {
294 .path_const = *i,
295 .mode = EMPTY_DIR,
296 .ignore = false,
6c47cd7d 297 .read_only = true,
2abd4e38
YW
298 .options_const = "mode=755",
299 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
6c47cd7d
LP
300 };
301 }
302
303 return 0;
304}
305
da6053d0
LP
306static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) {
307 size_t i;
d2d6c096
LP
308
309 assert(p);
310
311 for (i = 0; i < n; i++) {
312 const BindMount *b = binds + i;
313
314 *((*p)++) = (MountEntry) {
315 .path_const = b->destination,
316 .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT,
317 .read_only = b->read_only,
9ce4e4b0 318 .nosuid = b->nosuid,
d2d6c096 319 .source_const = b->source,
4ca763a9 320 .ignore = b->ignore_enoent,
d2d6c096
LP
321 };
322 }
323
324 return 0;
325}
326
da6053d0
LP
327static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) {
328 size_t i;
2abd4e38
YW
329 int r;
330
331 assert(p);
332
333 for (i = 0; i < n; i++) {
334 const TemporaryFileSystem *t = tmpfs + i;
335 _cleanup_free_ char *o = NULL, *str = NULL;
ad8e66dc 336 unsigned long flags;
2abd4e38
YW
337 bool ro = false;
338
baaa35ad
ZJS
339 if (!path_is_absolute(t->path))
340 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
341 "Path is not absolute: %s",
342 t->path);
2abd4e38 343
ad8e66dc
AJ
344 str = strjoin("mode=0755,", t->options);
345 if (!str)
346 return -ENOMEM;
2abd4e38 347
ad8e66dc
AJ
348 r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o);
349 if (r < 0)
350 return log_debug_errno(r, "Failed to parse mount option '%s': %m", str);
2abd4e38 351
ad8e66dc
AJ
352 ro = flags & MS_RDONLY;
353 if (ro)
354 flags ^= MS_RDONLY;
2abd4e38
YW
355
356 *((*p)++) = (MountEntry) {
357 .path_const = t->path,
358 .mode = TMPFS,
359 .read_only = ro,
ad8e66dc 360 .options_malloc = TAKE_PTR(o),
2abd4e38
YW
361 .flags = flags,
362 };
2abd4e38
YW
363 }
364
365 return 0;
366}
367
da6053d0
LP
368static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) {
369 size_t i;
11a30cec
DH
370
371 assert(p);
f471b2af 372 assert(mounts);
11a30cec 373
5327c910 374 /* Adds a list of static pre-defined entries */
f471b2af 375
5327c910 376 for (i = 0; i < n; i++)
34de407a
LP
377 *((*p)++) = (MountEntry) {
378 .path_const = mount_entry_path(mounts+i),
5327c910
LP
379 .mode = mounts[i].mode,
380 .ignore = mounts[i].ignore || ignore_protect,
381 };
f471b2af
DH
382
383 return 0;
384}
385
34de407a 386static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) {
c575770b
DH
387 assert(p);
388
5327c910 389 switch (protect_home) {
b6c432ca 390
5327c910 391 case PROTECT_HOME_NO:
b6c432ca
DH
392 return 0;
393
b6c432ca 394 case PROTECT_HOME_READ_ONLY:
5327c910
LP
395 return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect);
396
e4da7d8c
YW
397 case PROTECT_HOME_TMPFS:
398 return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect);
399
b6c432ca 400 case PROTECT_HOME_YES:
5327c910
LP
401 return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect);
402
b6c432ca 403 default:
5327c910 404 assert_not_reached("Unexpected ProtectHome= value");
b6c432ca 405 }
b6c432ca
DH
406}
407
34de407a 408static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) {
f471b2af
DH
409 assert(p);
410
5327c910
LP
411 switch (protect_system) {
412
413 case PROTECT_SYSTEM_NO:
f471b2af
DH
414 return 0;
415
f471b2af 416 case PROTECT_SYSTEM_STRICT:
5327c910
LP
417 return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect);
418
f471b2af 419 case PROTECT_SYSTEM_YES:
5327c910
LP
420 return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect);
421
f471b2af 422 case PROTECT_SYSTEM_FULL:
5327c910
LP
423 return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect);
424
f471b2af 425 default:
5327c910 426 assert_not_reached("Unexpected ProtectSystem= value");
f471b2af 427 }
11a30cec
DH
428}
429
93bab288 430static int mount_path_compare(const MountEntry *a, const MountEntry *b) {
a0827e2b 431 int d;
15ae422b 432
6ee1a919 433 /* If the paths are not equal, then order prefixes first */
93bab288 434 d = path_compare(mount_entry_path(a), mount_entry_path(b));
6ee1a919
LP
435 if (d != 0)
436 return d;
15ae422b 437
6ee1a919 438 /* If the paths are equal, check the mode */
93bab288 439 return CMP((int) a->mode, (int) b->mode);
15ae422b
LP
440}
441
da6053d0
LP
442static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) {
443 size_t i;
5327c910 444
4a756839 445 /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */
5327c910
LP
446
447 for (i = 0; i < n; i++) {
448 char *s;
449
450 if (m[i].has_prefix)
451 continue;
452
c6134d3e 453 s = path_join(root_directory, mount_entry_path(m+i));
5327c910
LP
454 if (!s)
455 return -ENOMEM;
456
e282f51f 457 free_and_replace(m[i].path_malloc, s);
5327c910
LP
458 m[i].has_prefix = true;
459 }
460
461 return 0;
462}
463
da6053d0 464static void drop_duplicates(MountEntry *m, size_t *n) {
34de407a 465 MountEntry *f, *t, *previous;
15ae422b 466
c17ec25e 467 assert(m);
15ae422b 468 assert(n);
15ae422b 469
fe3c2583
LP
470 /* Drops duplicate entries. Expects that the array is properly ordered already. */
471
1d54cd5d 472 for (f = m, t = m, previous = NULL; f < m + *n; f++) {
15ae422b 473
fe3c2583 474 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
088696fe
LP
475 * above. Note that we only drop duplicates that haven't been mounted yet. */
476 if (previous &&
477 path_equal(mount_entry_path(f), mount_entry_path(previous)) &&
478 !f->applied && !previous->applied) {
5beb8688 479 log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode));
34de407a 480 previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */
1eb7e08e 481 mount_entry_done(f);
15ae422b 482 continue;
fe3c2583 483 }
15ae422b 484
e2d7c1a0 485 *t = *f;
15ae422b 486 previous = t;
fe3c2583
LP
487 t++;
488 }
489
490 *n = t - m;
491}
492
da6053d0 493static void drop_inaccessible(MountEntry *m, size_t *n) {
34de407a 494 MountEntry *f, *t;
fe3c2583
LP
495 const char *clear = NULL;
496
497 assert(m);
498 assert(n);
499
500 /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
501 * ordered already. */
502
1d54cd5d 503 for (f = m, t = m; f < m + *n; f++) {
fe3c2583
LP
504
505 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
506 * it, as inaccessible paths really should drop the entire subtree. */
34de407a
LP
507 if (clear && path_startswith(mount_entry_path(f), clear)) {
508 log_debug("%s is masked by %s.", mount_entry_path(f), clear);
1eb7e08e 509 mount_entry_done(f);
fe3c2583
LP
510 continue;
511 }
15ae422b 512
34de407a 513 clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL;
fe3c2583
LP
514
515 *t = *f;
15ae422b
LP
516 t++;
517 }
518
c17ec25e 519 *n = t - m;
15ae422b
LP
520}
521
da6053d0 522static void drop_nop(MountEntry *m, size_t *n) {
34de407a 523 MountEntry *f, *t;
7648a565
LP
524
525 assert(m);
526 assert(n);
527
528 /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
529 * list is ordered by prefixes. */
530
1d54cd5d 531 for (f = m, t = m; f < m + *n; f++) {
7648a565 532
1e05071d
YW
533 /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */
534 if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) {
34de407a 535 MountEntry *p;
7648a565
LP
536 bool found = false;
537
538 /* Now let's find the first parent of the entry we are looking at. */
539 for (p = t-1; p >= m; p--) {
34de407a 540 if (path_startswith(mount_entry_path(f), mount_entry_path(p))) {
7648a565
LP
541 found = true;
542 break;
543 }
544 }
545
546 /* We found it, let's see if it's the same mode, if so, we can drop this entry */
547 if (found && p->mode == f->mode) {
5beb8688
YW
548 log_debug("%s (%s) is made redundant by %s (%s)",
549 mount_entry_path(f), mount_mode_to_string(f->mode),
550 mount_entry_path(p), mount_mode_to_string(p->mode));
1eb7e08e 551 mount_entry_done(f);
7648a565
LP
552 continue;
553 }
554 }
555
556 *t = *f;
557 t++;
558 }
559
560 *n = t - m;
561}
562
da6053d0 563static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) {
34de407a 564 MountEntry *f, *t;
cd2902c9
LP
565
566 assert(m);
567 assert(n);
568
1d54cd5d 569 /* Nothing to do */
cd2902c9
LP
570 if (!root_directory)
571 return;
572
573 /* Drops all mounts that are outside of the root directory. */
574
1d54cd5d 575 for (f = m, t = m; f < m + *n; f++) {
cd2902c9 576
34de407a
LP
577 if (!path_startswith(mount_entry_path(f), root_directory)) {
578 log_debug("%s is outside of root directory.", mount_entry_path(f));
1eb7e08e 579 mount_entry_done(f);
cd2902c9
LP
580 continue;
581 }
582
583 *t = *f;
584 t++;
585 }
586
587 *n = t - m;
588}
589
b2a60844
LP
590static int clone_device_node(
591 const char *d,
592 const char *temporary_mount,
593 bool *make_devnode) {
594
595 _cleanup_free_ char *sl = NULL;
596 const char *dn, *bn, *t;
b5e99f23
ДГ
597 struct stat st;
598 int r;
599
414b304b 600 if (stat(d, &st) < 0) {
b2a60844
LP
601 if (errno == ENOENT) {
602 log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d);
af984e13 603 return -ENXIO;
b2a60844
LP
604 }
605
606 return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d);
b5e99f23
ДГ
607 }
608
609 if (!S_ISBLK(st.st_mode) &&
baaa35ad
ZJS
610 !S_ISCHR(st.st_mode))
611 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
612 "Device node '%s' to clone is not a device node, ignoring.",
613 d);
b5e99f23 614
6f7f3a33 615 dn = strjoina(temporary_mount, d);
b5e99f23 616
b2a60844 617 /* First, try to create device node properly */
16498617
CB
618 if (*make_devnode) {
619 mac_selinux_create_file_prepare(d, st.st_mode);
620 r = mknod(dn, st.st_mode, st.st_rdev);
621 mac_selinux_create_file_clear();
b2a60844
LP
622 if (r >= 0)
623 goto add_symlink;
16498617
CB
624 if (errno != EPERM)
625 return log_debug_errno(errno, "mknod failed for %s: %m", d);
626
b2a60844 627 /* This didn't work, let's not try this again for the next iterations. */
16498617
CB
628 *make_devnode = false;
629 }
630
631 /* We're about to fallback to bind-mounting the device
632 * node. So create a dummy bind-mount target. */
633 mac_selinux_create_file_prepare(d, 0);
634 r = mknod(dn, S_IFREG, 0);
b5e99f23 635 mac_selinux_create_file_clear();
16498617 636 if (r < 0 && errno != EEXIST)
b2a60844 637 return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d);
16498617
CB
638
639 /* Fallback to bind-mounting:
640 * The assumption here is that all used device nodes carry standard
641 * properties. Specifically, the devices nodes we bind-mount should
642 * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx)
643 * and should not carry ACLs. */
644 if (mount(d, dn, NULL, MS_BIND, NULL) < 0)
b2a60844
LP
645 return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d);
646
647add_symlink:
648 bn = path_startswith(d, "/dev/");
649 if (!bn)
650 return 0;
651
652 /* Create symlinks like /dev/char/1:9 → ../urandom */
653 if (asprintf(&sl, "%s/dev/%s/%u:%u", temporary_mount, S_ISCHR(st.st_mode) ? "char" : "block", major(st.st_rdev), minor(st.st_rdev)) < 0)
654 return log_oom();
655
656 (void) mkdir_parents(sl, 0755);
657
658 t = strjoina("../", bn);
659
660 if (symlink(t, sl) < 0)
2e4a4fae 661 log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl);
b5e99f23 662
af984e13 663 return 0;
b5e99f23
ДГ
664}
665
5d997827 666static int mount_private_dev(MountEntry *m) {
7f112f50
LP
667 static const char devnodes[] =
668 "/dev/null\0"
669 "/dev/zero\0"
670 "/dev/full\0"
671 "/dev/random\0"
672 "/dev/urandom\0"
673 "/dev/tty\0";
674
2b85f4e1 675 char temporary_mount[] = "/tmp/namespace-dev-XXXXXX";
63cc4c31 676 const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL;
16498617 677 bool can_mknod = true;
7f112f50
LP
678 _cleanup_umask_ mode_t u;
679 int r;
680
681 assert(m);
682
683 u = umask(0000);
684
2b85f4e1 685 if (!mkdtemp(temporary_mount))
2e4a4fae 686 return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount);
2b85f4e1 687
63c372cb 688 dev = strjoina(temporary_mount, "/dev");
dc751688 689 (void) mkdir(dev, 0755);
737ba3c8 690 if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) {
2e4a4fae 691 r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev);
2b85f4e1
LP
692 goto fail;
693 }
694
63c372cb 695 devpts = strjoina(temporary_mount, "/dev/pts");
dc751688 696 (void) mkdir(devpts, 0755);
2b85f4e1 697 if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) {
2e4a4fae 698 r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts);
2b85f4e1
LP
699 goto fail;
700 }
701
2e4a4fae
YW
702 /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx.
703 * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible.
704 * Thus, in that case make a clone.
705 * In nspawn and other containers it will be a symlink, in that case make it a symlink. */
36ce7110 706 r = is_symlink("/dev/ptmx");
2e4a4fae
YW
707 if (r < 0) {
708 log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m");
3164e3cb 709 goto fail;
2e4a4fae 710 } else if (r > 0) {
414b304b
ДГ
711 devptmx = strjoina(temporary_mount, "/dev/ptmx");
712 if (symlink("pts/ptmx", devptmx) < 0) {
2e4a4fae 713 r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx);
414b304b
ДГ
714 goto fail;
715 }
716 } else {
16498617 717 r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod);
152c475f
LP
718 if (r < 0)
719 goto fail;
414b304b 720 }
e06b6479 721
63c372cb 722 devshm = strjoina(temporary_mount, "/dev/shm");
8d953682 723 (void) mkdir(devshm, 0755);
2b85f4e1
LP
724 r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL);
725 if (r < 0) {
2e4a4fae 726 r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm);
2b85f4e1
LP
727 goto fail;
728 }
729
63c372cb 730 devmqueue = strjoina(temporary_mount, "/dev/mqueue");
dc751688 731 (void) mkdir(devmqueue, 0755);
2e4a4fae
YW
732 if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0)
733 log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue);
2b85f4e1 734
63c372cb 735 devhugepages = strjoina(temporary_mount, "/dev/hugepages");
dc751688 736 (void) mkdir(devhugepages, 0755);
2e4a4fae
YW
737 if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0)
738 log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages);
2b85f4e1 739
63c372cb 740 devlog = strjoina(temporary_mount, "/dev/log");
2e4a4fae
YW
741 if (symlink("/run/systemd/journal/dev-log", devlog) < 0)
742 log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog);
82d25240 743
7f112f50 744 NULSTR_FOREACH(d, devnodes) {
16498617 745 r = clone_device_node(d, temporary_mount, &can_mknod);
af984e13
ZJS
746 /* ENXIO means the the *source* is not a device file, skip creation in that case */
747 if (r < 0 && r != -ENXIO)
2b85f4e1 748 goto fail;
7f112f50
LP
749 }
750
2e4a4fae
YW
751 r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID);
752 if (r < 0)
753 log_debug_errno(r, "Failed to setup basic device tree at '%s', ignoring: %m", temporary_mount);
7f112f50 754
ee818b89
AC
755 /* Create the /dev directory if missing. It is more likely to be
756 * missing when the service is started with RootDirectory. This is
757 * consistent with mount units creating the mount points when missing.
758 */
34de407a 759 (void) mkdir_p_label(mount_entry_path(m), 0755);
ee818b89 760
9e5f8252 761 /* Unmount everything in old /dev */
2e4a4fae
YW
762 r = umount_recursive(mount_entry_path(m), 0);
763 if (r < 0)
764 log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m));
765
34de407a 766 if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) {
2e4a4fae 767 r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m));
2b85f4e1
LP
768 goto fail;
769 }
7f112f50 770
1019a48f
LP
771 (void) rmdir(dev);
772 (void) rmdir(temporary_mount);
7f112f50 773
2b85f4e1 774 return 0;
7f112f50 775
2b85f4e1
LP
776fail:
777 if (devpts)
1019a48f 778 (void) umount(devpts);
7f112f50 779
2b85f4e1 780 if (devshm)
1019a48f 781 (void) umount(devshm);
7f112f50 782
2b85f4e1 783 if (devhugepages)
1019a48f 784 (void) umount(devhugepages);
7f112f50 785
2b85f4e1 786 if (devmqueue)
1019a48f 787 (void) umount(devmqueue);
7f112f50 788
1019a48f
LP
789 (void) umount(dev);
790 (void) rmdir(dev);
791 (void) rmdir(temporary_mount);
7f112f50 792
2b85f4e1 793 return r;
7f112f50
LP
794}
795
2a2969fd 796static int mount_bind_dev(const MountEntry *m) {
5d997827
LP
797 int r;
798
799 assert(m);
800
801 /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
802 * /dev. This is only used when RootDirectory= is set. */
803
645767d6
LP
804 (void) mkdir_p_label(mount_entry_path(m), 0755);
805
5d997827
LP
806 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
807 if (r < 0)
808 return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m");
809 if (r > 0) /* make this a NOP if /dev is already a mount point */
810 return 0;
811
812 if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
813 return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m));
814
815 return 1;
816}
817
2a2969fd 818static int mount_sysfs(const MountEntry *m) {
5d997827
LP
819 int r;
820
821 assert(m);
822
645767d6
LP
823 (void) mkdir_p_label(mount_entry_path(m), 0755);
824
5d997827
LP
825 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
826 if (r < 0)
827 return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m");
828 if (r > 0) /* make this a NOP if /sys is already a mount point */
829 return 0;
830
831 /* Bind mount the host's version so that we get all child mounts of it, too. */
832 if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
833 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
834
835 return 1;
836}
837
2a2969fd 838static int mount_procfs(const MountEntry *m) {
5d997827
LP
839 int r;
840
841 assert(m);
842
645767d6
LP
843 (void) mkdir_p_label(mount_entry_path(m), 0755);
844
5d997827
LP
845 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
846 if (r < 0)
847 return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m");
848 if (r > 0) /* make this a NOP if /proc is already a mount point */
849 return 0;
850
851 /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
852 if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0)
853 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
854
855 return 1;
856}
857
2abd4e38 858static int mount_tmpfs(const MountEntry *m) {
6c47cd7d
LP
859 assert(m);
860
2abd4e38 861 /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
6c47cd7d
LP
862
863 (void) mkdir_p_label(mount_entry_path(m), 0755);
864 (void) umount_recursive(mount_entry_path(m), 0);
865
2abd4e38 866 if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0)
6c47cd7d
LP
867 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
868
869 return 1;
870}
871
088696fe 872static int follow_symlink(
d2d6c096 873 const char *root_directory,
088696fe 874 MountEntry *m) {
d2d6c096 875
088696fe 876 _cleanup_free_ char *target = NULL;
8fceda93
LP
877 int r;
878
088696fe
LP
879 /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we
880 * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at
881 * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the
882 * end and already have a fully normalized name. */
8fceda93 883
a5648b80 884 r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL);
088696fe
LP
885 if (r < 0)
886 return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m));
887 if (r > 0) /* Reached the end, nothing more to resolve */
888 return 1;
8fceda93 889
baaa35ad
ZJS
890 if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */
891 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
892 "Symlink loop on '%s'.",
893 mount_entry_path(m));
8fceda93 894
088696fe 895 log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target);
8fceda93 896
088696fe
LP
897 free_and_replace(m->path_malloc, target);
898 m->has_prefix = true;
8fceda93 899
088696fe
LP
900 m->n_followed ++;
901
902 return 0;
8fceda93
LP
903}
904
ac0930c8 905static int apply_mount(
8fceda93 906 const char *root_directory,
89bd586c 907 MountEntry *m) {
ac0930c8 908
e5f10caf 909 _cleanup_free_ char *inaccessible = NULL;
a227a4be 910 bool rbind = true, make = false;
15ae422b 911 const char *what;
15ae422b 912 int r;
15ae422b 913
c17ec25e 914 assert(m);
15ae422b 915
34de407a 916 log_debug("Applying namespace mount on %s", mount_entry_path(m));
fe3c2583 917
c17ec25e 918 switch (m->mode) {
15ae422b 919
160cfdbe 920 case INACCESSIBLE: {
e5f10caf
AZ
921 _cleanup_free_ char *tmp = NULL;
922 const char *runtime_dir;
160cfdbe 923 struct stat target;
6d313367
LP
924
925 /* First, get rid of everything that is below if there
926 * is anything... Then, overmount it with an
c4b41707 927 * inaccessible path. */
34de407a 928 (void) umount_recursive(mount_entry_path(m), 0);
6d313367 929
088696fe
LP
930 if (lstat(mount_entry_path(m), &target) < 0) {
931 if (errno == ENOENT && m->ignore)
932 return 0;
933
34de407a 934 return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m));
088696fe 935 }
15ae422b 936
e5f10caf
AZ
937 if (geteuid() == 0)
938 runtime_dir = "/run/systemd";
939 else {
940 if (asprintf(&tmp, "/run/user/"UID_FMT, geteuid()) < 0)
941 log_oom();
942
943 runtime_dir = tmp;
944 }
945
946 r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible);
947 if (r < 0)
baaa35ad
ZJS
948 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
949 "File type not supported for inaccessible mounts. Note that symlinks are not allowed");
e5f10caf 950 what = inaccessible;
c4b41707 951 break;
160cfdbe 952 }
fe3c2583 953
15ae422b 954 case READONLY:
15ae422b 955 case READWRITE:
1e05071d 956 case READWRITE_IMPLICIT:
8fceda93 957 r = path_is_mount_point(mount_entry_path(m), root_directory, 0);
088696fe
LP
958 if (r == -ENOENT && m->ignore)
959 return 0;
d944dc95 960 if (r < 0)
34de407a 961 return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m));
6b7c9f8b
LP
962 if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY bit for the mount point if needed. */
963 return 0;
6b7c9f8b 964 /* This isn't a mount point yet, let's make it one. */
34de407a 965 what = mount_entry_path(m);
6b7c9f8b 966 break;
15ae422b 967
d2d6c096
LP
968 case BIND_MOUNT:
969 rbind = false;
d2d6c096 970
4831981d 971 _fallthrough_;
088696fe
LP
972 case BIND_MOUNT_RECURSIVE: {
973 _cleanup_free_ char *chased = NULL;
5d997827 974
088696fe
LP
975 /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind
976 * mount source paths are always relative to the host root, hence we pass NULL as root directory to
977 * chase_symlinks() here. */
978
a5648b80 979 r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL);
088696fe
LP
980 if (r == -ENOENT && m->ignore) {
981 log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m));
982 return 0;
983 }
984 if (r < 0)
985 return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m));
986
987 log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased);
988
989 free_and_replace(m->source_malloc, chased);
d2d6c096
LP
990
991 what = mount_entry_source(m);
a227a4be 992 make = true;
d2d6c096 993 break;
088696fe 994 }
d2d6c096 995
6c47cd7d 996 case EMPTY_DIR:
2abd4e38
YW
997 case TMPFS:
998 return mount_tmpfs(m);
6c47cd7d 999
ac0930c8 1000 case PRIVATE_TMP:
89bd586c 1001 what = mount_entry_source(m);
a227a4be 1002 make = true;
15ae422b 1003 break;
e364ad06 1004
d6797c92 1005 case PRIVATE_DEV:
5d997827
LP
1006 return mount_private_dev(m);
1007
1008 case BIND_DEV:
1009 return mount_bind_dev(m);
1010
1011 case SYSFS:
1012 return mount_sysfs(m);
1013
1014 case PROCFS:
1015 return mount_procfs(m);
d6797c92 1016
e364ad06
LP
1017 default:
1018 assert_not_reached("Unknown mode");
15ae422b
LP
1019 }
1020
ac0930c8 1021 assert(what);
15ae422b 1022
a227a4be
LP
1023 if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) {
1024 bool try_again = false;
1025 r = -errno;
1026
1027 if (r == -ENOENT && make) {
1028 struct stat st;
1029
1030 /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */
1031
e8717862
LP
1032 if (stat(what, &st) < 0)
1033 log_debug_errno(errno, "Mount point source '%s' is not accessible: %m", what);
1034 else {
1035 int q;
a227a4be
LP
1036
1037 (void) mkdir_parents(mount_entry_path(m), 0755);
1038
1039 if (S_ISDIR(st.st_mode))
e8717862 1040 q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0;
a227a4be 1041 else
e8717862
LP
1042 q = touch(mount_entry_path(m));
1043
1044 if (q < 0)
1045 log_debug_errno(q, "Failed to create destination mount point node '%s': %m", mount_entry_path(m));
1046 else
1047 try_again = true;
a227a4be
LP
1048 }
1049 }
1050
1051 if (try_again) {
1052 if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0)
1053 r = -errno;
1054 else
1055 r = 0;
1056 }
1057
1058 if (r < 0)
1059 return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m));
1060 }
6b7c9f8b 1061
34de407a 1062 log_debug("Successfully mounted %s to %s", what, mount_entry_path(m));
6b7c9f8b 1063 return 0;
ac0930c8 1064}
15ae422b 1065
2a2969fd 1066static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) {
9ce4e4b0 1067 unsigned long new_flags = 0, flags_mask = 0;
763a260a 1068 bool submounts = false;
6b7c9f8b 1069 int r = 0;
15ae422b 1070
c17ec25e 1071 assert(m);
ac9de0b3 1072 assert(proc_self_mountinfo);
ac0930c8 1073
9ce4e4b0
LP
1074 if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) {
1075 new_flags |= MS_RDONLY;
1076 flags_mask |= MS_RDONLY;
1077 }
1078
1079 if (m->nosuid) {
1080 new_flags |= MS_NOSUID;
1081 flags_mask |= MS_NOSUID;
1082 }
1083
1084 if (flags_mask == 0) /* No Change? */
6b7c9f8b
LP
1085 return 0;
1086
9ce4e4b0
LP
1087 /* We generally apply these changes recursively, except for /dev, and the cases we know there's
1088 * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the
1089 * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace
1090 * and running Linux <= 4.17. */
1091 submounts =
1092 mount_entry_read_only(m) &&
1093 !IN_SET(m->mode, EMPTY_DIR, TMPFS);
1094 if (submounts)
1095 r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, blacklist, proc_self_mountinfo);
1096 else
7cce68e1 1097 r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo);
9ce4e4b0 1098
867189b5
LP
1099 /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked
1100 * read-only already stays this way. This improves compatibility with container managers, where we
1101 * won't attempt to undo read-only mounts already applied. */
ac0930c8 1102
8fceda93 1103 if (r == -ENOENT && m->ignore)
867189b5 1104 return 0;
763a260a 1105 if (r < 0)
9ce4e4b0 1106 return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m),
763a260a 1107 submounts ? " and its submounts" : "");
763a260a 1108 return 0;
d944dc95
LP
1109}
1110
9b68367b 1111static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) {
5d997827
LP
1112 assert(ns_info);
1113
9c988f93
DH
1114 /*
1115 * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1116 * since to protect the API VFS mounts, they need to be around in the
9b68367b 1117 * first place...
9c988f93 1118 */
5d997827 1119
9b68367b
YW
1120 return ns_info->mount_apivfs ||
1121 ns_info->protect_control_groups ||
1122 ns_info->protect_kernel_tunables;
5d997827
LP
1123}
1124
da6053d0 1125static size_t namespace_calculate_mounts(
bb0ff3fb 1126 const NamespaceInfo *ns_info,
2652c6c1
DH
1127 char** read_write_paths,
1128 char** read_only_paths,
1129 char** inaccessible_paths,
6c47cd7d 1130 char** empty_directories,
da6053d0
LP
1131 size_t n_bind_mounts,
1132 size_t n_temporary_filesystems,
2652c6c1
DH
1133 const char* tmp_dir,
1134 const char* var_tmp_dir,
91dd5f7c 1135 const char* log_namespace,
2652c6c1
DH
1136 ProtectHome protect_home,
1137 ProtectSystem protect_system) {
1138
da6053d0
LP
1139 size_t protect_home_cnt;
1140 size_t protect_system_cnt =
f471b2af
DH
1141 (protect_system == PROTECT_SYSTEM_STRICT ?
1142 ELEMENTSOF(protect_system_strict_table) :
1143 ((protect_system == PROTECT_SYSTEM_FULL) ?
1144 ELEMENTSOF(protect_system_full_table) :
1145 ((protect_system == PROTECT_SYSTEM_YES) ?
1146 ELEMENTSOF(protect_system_yes_table) : 0)));
1147
b6c432ca
DH
1148 protect_home_cnt =
1149 (protect_home == PROTECT_HOME_YES ?
1150 ELEMENTSOF(protect_home_yes_table) :
1151 ((protect_home == PROTECT_HOME_READ_ONLY) ?
e4da7d8c
YW
1152 ELEMENTSOF(protect_home_read_only_table) :
1153 ((protect_home == PROTECT_HOME_TMPFS) ?
1154 ELEMENTSOF(protect_home_tmpfs_table) : 0)));
b6c432ca 1155
2652c6c1
DH
1156 return !!tmp_dir + !!var_tmp_dir +
1157 strv_length(read_write_paths) +
1158 strv_length(read_only_paths) +
1159 strv_length(inaccessible_paths) +
6c47cd7d 1160 strv_length(empty_directories) +
d2d6c096 1161 n_bind_mounts +
2abd4e38 1162 n_temporary_filesystems +
c575770b
DH
1163 ns_info->private_dev +
1164 (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) +
c575770b 1165 (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) +
94a7b275
KK
1166 (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) +
1167 (ns_info->protect_control_groups ? 1 : 0) +
5d997827 1168 protect_home_cnt + protect_system_cnt +
aecd5ac6 1169 (ns_info->protect_hostname ? 2 : 0) +
91dd5f7c
LP
1170 (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) +
1171 !!log_namespace;
2652c6c1
DH
1172}
1173
da6053d0 1174static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) {
9b68367b 1175 assert(root_directory);
f8b64b57
LP
1176 assert(n_mounts);
1177 assert(mounts || *n_mounts == 0);
1178
93bab288 1179 typesafe_qsort(mounts, *n_mounts, mount_path_compare);
f8b64b57
LP
1180
1181 drop_duplicates(mounts, n_mounts);
1182 drop_outside_root(root_directory, mounts, n_mounts);
1183 drop_inaccessible(mounts, n_mounts);
1184 drop_nop(mounts, n_mounts);
1185}
1186
c8c535d5
LP
1187static bool root_read_only(
1188 char **read_only_paths,
1189 ProtectSystem protect_system) {
1190
1191 /* Determine whether the root directory is going to be read-only given the configured settings. */
1192
1193 if (protect_system == PROTECT_SYSTEM_STRICT)
1194 return true;
1195
1196 if (path_strv_contains(read_only_paths, "/"))
1197 return true;
1198
1199 return false;
1200}
1201
1202static bool home_read_only(
1203 char** read_only_paths,
1204 char** inaccessible_paths,
1205 char** empty_directories,
1206 const BindMount *bind_mounts,
1207 size_t n_bind_mounts,
1208 const TemporaryFileSystem *temporary_filesystems,
1209 size_t n_temporary_filesystems,
1210 ProtectHome protect_home) {
1211
1212 size_t i;
1213
1214 /* Determine whether the /home directory is going to be read-only given the configured settings. Yes,
1215 * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple
1216 * settings. */
1217
1218 if (protect_home != PROTECT_HOME_NO)
1219 return true;
1220
1221 if (path_strv_contains(read_only_paths, "/home") ||
1222 path_strv_contains(inaccessible_paths, "/home") ||
1223 path_strv_contains(empty_directories, "/home"))
1224 return true;
1225
1226 for (i = 0; i < n_temporary_filesystems; i++)
1227 if (path_equal(temporary_filesystems[i].path, "/home"))
1228 return true;
1229
1230 /* If /home is overmounted with some dir from the host it's not writable. */
1231 for (i = 0; i < n_bind_mounts; i++)
1232 if (path_equal(bind_mounts[i].destination, "/home"))
1233 return true;
1234
1235 return false;
1236}
1237
613b411c 1238int setup_namespace(
ee818b89 1239 const char* root_directory,
915e6d16 1240 const char* root_image,
bb0ff3fb 1241 const NamespaceInfo *ns_info,
2a624c36
AP
1242 char** read_write_paths,
1243 char** read_only_paths,
1244 char** inaccessible_paths,
6c47cd7d 1245 char** empty_directories,
d2d6c096 1246 const BindMount *bind_mounts,
da6053d0 1247 size_t n_bind_mounts,
2abd4e38 1248 const TemporaryFileSystem *temporary_filesystems,
da6053d0 1249 size_t n_temporary_filesystems,
a004cb4c
LP
1250 const char* tmp_dir,
1251 const char* var_tmp_dir,
91dd5f7c 1252 const char *log_namespace,
1b8689f9
LP
1253 ProtectHome protect_home,
1254 ProtectSystem protect_system,
915e6d16 1255 unsigned long mount_flags,
7cc5ef5f
ZJS
1256 DissectImageFlags dissect_image_flags,
1257 char **error_path) {
15ae422b 1258
915e6d16 1259 _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
78ebe980 1260 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
915e6d16 1261 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
78ebe980 1262 _cleanup_free_ void *root_hash = NULL;
5f7a690a 1263 MountEntry *m = NULL, *mounts = NULL;
9b68367b 1264 size_t n_mounts, root_hash_size = 0;
d18aff04 1265 bool require_prefix = false;
9b68367b 1266 const char *root;
c17ec25e 1267 int r = 0;
15ae422b 1268
915e6d16
LP
1269 assert(ns_info);
1270
613b411c 1271 if (mount_flags == 0)
c17ec25e 1272 mount_flags = MS_SHARED;
ac0930c8 1273
915e6d16
LP
1274 if (root_image) {
1275 dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT;
1276
c8c535d5
LP
1277 /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */
1278 if (root_read_only(read_only_paths,
1279 protect_system) &&
1280 home_read_only(read_only_paths, inaccessible_paths, empty_directories,
1281 bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems,
1282 protect_home) &&
c9ef8573 1283 strv_isempty(read_write_paths))
915e6d16
LP
1284 dissect_image_flags |= DISSECT_IMAGE_READ_ONLY;
1285
1286 r = loop_device_make_by_path(root_image,
b0a94268 1287 FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */,
e08f94ac 1288 LO_FLAGS_PARTSCAN,
915e6d16
LP
1289 &loop_device);
1290 if (r < 0)
763a260a 1291 return log_debug_errno(r, "Failed to create loop device for root image: %m");
915e6d16 1292
78ebe980
LP
1293 r = root_hash_load(root_image, &root_hash, &root_hash_size);
1294 if (r < 0)
763a260a 1295 return log_debug_errno(r, "Failed to load root hash: %m");
78ebe980
LP
1296
1297 r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image);
1298 if (r < 0)
763a260a 1299 return log_debug_errno(r, "Failed to dissect image: %m");
78ebe980
LP
1300
1301 r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image);
915e6d16 1302 if (r < 0)
763a260a 1303 return log_debug_errno(r, "Failed to decrypt dissected image: %m");
915e6d16
LP
1304 }
1305
e908468b
LP
1306 if (root_directory)
1307 root = root_directory;
0722b359
JS
1308 else {
1309 /* Always create the mount namespace in a temporary directory, instead of operating
1310 * directly in the root. The temporary directory prevents any mounts from being
1311 * potentially obscured my other mounts we already applied.
1312 * We use the same mount point for all images, which is safe, since they all live
1313 * in their own namespaces after all, and hence won't see each other. */
e908468b
LP
1314
1315 root = "/run/systemd/unit-root";
1316 (void) mkdir_label(root, 0700);
d18aff04 1317 require_prefix = true;
0722b359 1318 }
e908468b 1319
cfbeb4ef
LP
1320 n_mounts = namespace_calculate_mounts(
1321 ns_info,
1322 read_write_paths,
1323 read_only_paths,
1324 inaccessible_paths,
6c47cd7d 1325 empty_directories,
f5c52a77 1326 n_bind_mounts,
2abd4e38 1327 n_temporary_filesystems,
cfbeb4ef 1328 tmp_dir, var_tmp_dir,
91dd5f7c 1329 log_namespace,
cfbeb4ef 1330 protect_home, protect_system);
613b411c 1331
f0a4feb0 1332 if (n_mounts > 0) {
5f7a690a
LP
1333 m = mounts = new0(MountEntry, n_mounts);
1334 if (!mounts)
1335 return -ENOMEM;
1336
d18aff04 1337 r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix);
613b411c 1338 if (r < 0)
f0a4feb0 1339 goto finish;
613b411c 1340
d18aff04 1341 r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix);
613b411c 1342 if (r < 0)
f0a4feb0 1343 goto finish;
613b411c 1344
d18aff04 1345 r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix);
613b411c 1346 if (r < 0)
f0a4feb0 1347 goto finish;
7ff7394d 1348
6c47cd7d
LP
1349 r = append_empty_dir_mounts(&m, empty_directories);
1350 if (r < 0)
1351 goto finish;
1352
d2d6c096
LP
1353 r = append_bind_mounts(&m, bind_mounts, n_bind_mounts);
1354 if (r < 0)
1355 goto finish;
1356
2abd4e38
YW
1357 r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems);
1358 if (r < 0)
1359 goto finish;
1360
613b411c 1361 if (tmp_dir) {
34de407a 1362 *(m++) = (MountEntry) {
5327c910
LP
1363 .path_const = "/tmp",
1364 .mode = PRIVATE_TMP,
89bd586c 1365 .source_const = tmp_dir,
5327c910 1366 };
613b411c 1367 }
7ff7394d 1368
613b411c 1369 if (var_tmp_dir) {
34de407a 1370 *(m++) = (MountEntry) {
5327c910 1371 .path_const = "/var/tmp",
89bd586c
YW
1372 .mode = PRIVATE_TMP,
1373 .source_const = var_tmp_dir,
5327c910 1374 };
7ff7394d 1375 }
ac0930c8 1376
c575770b 1377 if (ns_info->private_dev) {
34de407a 1378 *(m++) = (MountEntry) {
5327c910
LP
1379 .path_const = "/dev",
1380 .mode = PRIVATE_DEV,
9ce4e4b0 1381 .flags = DEV_MOUNT_OPTIONS,
5327c910 1382 };
7f112f50
LP
1383 }
1384
c575770b 1385 if (ns_info->protect_kernel_tunables) {
5327c910 1386 r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths);
c575770b 1387 if (r < 0)
f0a4feb0 1388 goto finish;
c575770b
DH
1389 }
1390
1391 if (ns_info->protect_kernel_modules) {
5327c910 1392 r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths);
c575770b 1393 if (r < 0)
f0a4feb0 1394 goto finish;
c575770b 1395 }
59eeb84b 1396
94a7b275
KK
1397 if (ns_info->protect_kernel_logs) {
1398 r = append_static_mounts(&m, protect_kernel_logs_table, ELEMENTSOF(protect_kernel_logs_table), ns_info->ignore_protect_paths);
1399 if (r < 0)
1400 goto finish;
1401 }
1402
c575770b 1403 if (ns_info->protect_control_groups) {
34de407a 1404 *(m++) = (MountEntry) {
5327c910
LP
1405 .path_const = "/sys/fs/cgroup",
1406 .mode = READONLY,
1407 };
59eeb84b
LP
1408 }
1409
5327c910 1410 r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths);
b6c432ca 1411 if (r < 0)
f0a4feb0 1412 goto finish;
417116f2 1413
5327c910 1414 r = append_protect_system(&m, protect_system, false);
f471b2af 1415 if (r < 0)
f0a4feb0 1416 goto finish;
417116f2 1417
9b68367b 1418 if (namespace_info_mount_apivfs(ns_info)) {
5d997827
LP
1419 r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths);
1420 if (r < 0)
1421 goto finish;
1422 }
1423
aecd5ac6
TM
1424 if (ns_info->protect_hostname) {
1425 *(m++) = (MountEntry) {
1426 .path_const = "/proc/sys/kernel/hostname",
1427 .mode = READONLY,
1428 };
1429 *(m++) = (MountEntry) {
1430 .path_const = "/proc/sys/kernel/domainname",
1431 .mode = READONLY,
1432 };
1433 }
1434
91dd5f7c
LP
1435 if (log_namespace) {
1436 _cleanup_free_ char *q;
1437
1438 q = strjoin("/run/systemd/journal.", log_namespace);
1439 if (!q) {
1440 r = -ENOMEM;
1441 goto finish;
1442 }
1443
1444 *(m++) = (MountEntry) {
1445 .path_const = "/run/systemd/journal",
1446 .mode = BIND_MOUNT_RECURSIVE,
1447 .read_only = true,
1448 .source_malloc = TAKE_PTR(q),
1449 };
1450 }
1451
f0a4feb0 1452 assert(mounts + n_mounts == m);
ac0930c8 1453
5327c910 1454 /* Prepend the root directory where that's necessary */
e908468b 1455 r = prefix_where_needed(mounts, n_mounts, root);
5327c910
LP
1456 if (r < 0)
1457 goto finish;
1458
839f1877 1459 normalize_mounts(root, mounts, &n_mounts);
15ae422b
LP
1460 }
1461
1beab8b0
LP
1462 /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */
1463
d944dc95 1464 if (unshare(CLONE_NEWNS) < 0) {
763a260a 1465 r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m");
1beab8b0
LP
1466 if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS))
1467 /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place
1468 * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable
1469 * error back, which the caller can use to detect this case (and only this) and optionally
1470 * continue without namespacing applied. */
1471 r = -ENOANO;
1472
d944dc95
LP
1473 goto finish;
1474 }
1e4e94c8 1475
9b68367b
YW
1476 /* Remount / as SLAVE so that nothing now mounted in the namespace
1477 * shows up in the parent */
1478 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
763a260a 1479 r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m");
9b68367b 1480 goto finish;
ee818b89
AC
1481 }
1482
915e6d16 1483 if (root_image) {
e908468b 1484 /* A root image is specified, mount it to the right place */
2d3a5a73 1485 r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags);
763a260a
YW
1486 if (r < 0) {
1487 log_debug_errno(r, "Failed to mount root image: %m");
915e6d16 1488 goto finish;
763a260a 1489 }
915e6d16 1490
07ce7407
TM
1491 if (decrypted_image) {
1492 r = decrypted_image_relinquish(decrypted_image);
763a260a
YW
1493 if (r < 0) {
1494 log_debug_errno(r, "Failed to relinquish decrypted image: %m");
07ce7407 1495 goto finish;
763a260a 1496 }
07ce7407 1497 }
78ebe980 1498
915e6d16
LP
1499 loop_device_relinquish(loop_device);
1500
1501 } else if (root_directory) {
1502
e908468b
LP
1503 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1504 r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW);
763a260a
YW
1505 if (r < 0) {
1506 log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root);
d944dc95 1507 goto finish;
763a260a 1508 }
8f1ad200 1509 if (r == 0) {
e908468b 1510 if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) {
763a260a 1511 r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root);
8f1ad200
LP
1512 goto finish;
1513 }
d944dc95 1514 }
e908468b 1515
9b68367b 1516 } else {
e908468b
LP
1517
1518 /* Let's mount the main root directory to the root directory to use */
1519 if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) {
763a260a 1520 r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root);
e908468b
LP
1521 goto finish;
1522 }
ee818b89 1523 }
c2c13f2d 1524
4e0c20de
LP
1525 /* Try to set up the new root directory before mounting anything else there. */
1526 if (root_image || root_directory)
1527 (void) base_filesystem_create(root, UID_INVALID, GID_INVALID);
1528
f0a4feb0 1529 if (n_mounts > 0) {
ac9de0b3 1530 _cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
5f7a690a 1531 _cleanup_free_ char **blacklist = NULL;
da6053d0 1532 size_t j;
6b7c9f8b 1533
ac9de0b3
TR
1534 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc.
1535 * For example, this is the case with the option: 'InaccessiblePaths=/proc' */
1536 proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
1537 if (!proc_self_mountinfo) {
763a260a 1538 r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m");
7cc5ef5f
ZJS
1539 if (error_path)
1540 *error_path = strdup("/proc/self/mountinfo");
ac9de0b3
TR
1541 goto finish;
1542 }
1543
088696fe
LP
1544 /* First round, establish all mounts we need */
1545 for (;;) {
1546 bool again = false;
1547
1548 for (m = mounts; m < mounts + n_mounts; ++m) {
1549
1550 if (m->applied)
1551 continue;
1552
1553 r = follow_symlink(root, m);
7cc5ef5f
ZJS
1554 if (r < 0) {
1555 if (error_path && mount_entry_path(m))
1556 *error_path = strdup(mount_entry_path(m));
088696fe 1557 goto finish;
7cc5ef5f 1558 }
088696fe
LP
1559 if (r == 0) {
1560 /* We hit a symlinked mount point. The entry got rewritten and might point to a
1561 * very different place now. Let's normalize the changed list, and start from
1562 * the beginning. After all to mount the entry at the new location we might
1563 * need some other mounts first */
1564 again = true;
1565 break;
1566 }
1567
1568 r = apply_mount(root, m);
7cc5ef5f
ZJS
1569 if (r < 0) {
1570 if (error_path && mount_entry_path(m))
1571 *error_path = strdup(mount_entry_path(m));
088696fe 1572 goto finish;
7cc5ef5f 1573 }
088696fe
LP
1574
1575 m->applied = true;
1576 }
1577
1578 if (!again)
1579 break;
1580
839f1877 1581 normalize_mounts(root, mounts, &n_mounts);
c2c13f2d 1582 }
15ae422b 1583
6b7c9f8b 1584 /* Create a blacklist we can pass to bind_mount_recursive() */
5f7a690a
LP
1585 blacklist = new(char*, n_mounts+1);
1586 if (!blacklist) {
1587 r = -ENOMEM;
1588 goto finish;
1589 }
f0a4feb0 1590 for (j = 0; j < n_mounts; j++)
34de407a 1591 blacklist[j] = (char*) mount_entry_path(mounts+j);
6b7c9f8b
LP
1592 blacklist[j] = NULL;
1593
1594 /* Second round, flip the ro bits if necessary. */
f0a4feb0 1595 for (m = mounts; m < mounts + n_mounts; ++m) {
ac9de0b3 1596 r = make_read_only(m, blacklist, proc_self_mountinfo);
7cc5ef5f
ZJS
1597 if (r < 0) {
1598 if (error_path && mount_entry_path(m))
1599 *error_path = strdup(mount_entry_path(m));
d944dc95 1600 goto finish;
7cc5ef5f 1601 }
c2c13f2d 1602 }
15ae422b
LP
1603 }
1604
9b68367b
YW
1605 /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1606 r = mount_move_root(root);
763a260a
YW
1607 if (r < 0) {
1608 log_debug_errno(r, "Failed to mount root with MS_MOVE: %m");
9b68367b 1609 goto finish;
763a260a 1610 }
ee818b89 1611
55fe7432 1612 /* Remount / as the desired mode. Note that this will not
c2c13f2d
LP
1613 * reestablish propagation from our side to the host, since
1614 * what's disconnected is disconnected. */
d944dc95 1615 if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
763a260a 1616 r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m");
d944dc95
LP
1617 goto finish;
1618 }
15ae422b 1619
d944dc95 1620 r = 0;
15ae422b 1621
d944dc95 1622finish:
f0a4feb0 1623 for (m = mounts; m < mounts + n_mounts; m++)
1eb7e08e 1624 mount_entry_done(m);
613b411c 1625
5f7a690a
LP
1626 free(mounts);
1627
613b411c
LP
1628 return r;
1629}
1630
da6053d0
LP
1631void bind_mount_free_many(BindMount *b, size_t n) {
1632 size_t i;
d2d6c096
LP
1633
1634 assert(b || n == 0);
1635
1636 for (i = 0; i < n; i++) {
1637 free(b[i].source);
1638 free(b[i].destination);
1639 }
1640
1641 free(b);
1642}
1643
da6053d0 1644int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) {
d2d6c096
LP
1645 _cleanup_free_ char *s = NULL, *d = NULL;
1646 BindMount *c;
1647
1648 assert(b);
1649 assert(n);
1650 assert(item);
1651
1652 s = strdup(item->source);
1653 if (!s)
1654 return -ENOMEM;
1655
1656 d = strdup(item->destination);
1657 if (!d)
1658 return -ENOMEM;
1659
aa484f35 1660 c = reallocarray(*b, *n + 1, sizeof(BindMount));
d2d6c096
LP
1661 if (!c)
1662 return -ENOMEM;
1663
1664 *b = c;
1665
1666 c[(*n) ++] = (BindMount) {
1cc6c93a
YW
1667 .source = TAKE_PTR(s),
1668 .destination = TAKE_PTR(d),
d2d6c096 1669 .read_only = item->read_only,
9ce4e4b0 1670 .nosuid = item->nosuid,
d2d6c096
LP
1671 .recursive = item->recursive,
1672 .ignore_enoent = item->ignore_enoent,
1673 };
1674
d2d6c096
LP
1675 return 0;
1676}
1677
da6053d0
LP
1678void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) {
1679 size_t i;
2abd4e38
YW
1680
1681 assert(t || n == 0);
1682
1683 for (i = 0; i < n; i++) {
1684 free(t[i].path);
1685 free(t[i].options);
1686 }
1687
1688 free(t);
1689}
1690
1691int temporary_filesystem_add(
1692 TemporaryFileSystem **t,
da6053d0 1693 size_t *n,
2abd4e38
YW
1694 const char *path,
1695 const char *options) {
1696
1697 _cleanup_free_ char *p = NULL, *o = NULL;
1698 TemporaryFileSystem *c;
1699
1700 assert(t);
1701 assert(n);
1702 assert(path);
1703
1704 p = strdup(path);
1705 if (!p)
1706 return -ENOMEM;
1707
1708 if (!isempty(options)) {
1709 o = strdup(options);
1710 if (!o)
1711 return -ENOMEM;
1712 }
1713
aa484f35 1714 c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem));
2abd4e38
YW
1715 if (!c)
1716 return -ENOMEM;
1717
1718 *t = c;
1719
1720 c[(*n) ++] = (TemporaryFileSystem) {
1cc6c93a
YW
1721 .path = TAKE_PTR(p),
1722 .options = TAKE_PTR(o),
2abd4e38
YW
1723 };
1724
2abd4e38
YW
1725 return 0;
1726}
1727
a652f050
JR
1728static int make_tmp_prefix(const char *prefix) {
1729 _cleanup_free_ char *t = NULL;
1730 int r;
1731
1732 /* Don't do anything unless we know the dir is actually missing */
1733 r = access(prefix, F_OK);
1734 if (r >= 0)
1735 return 0;
1736 if (errno != ENOENT)
1737 return -errno;
1738
1739 r = mkdir_parents(prefix, 0755);
1740 if (r < 0)
1741 return r;
1742
1743 r = tempfn_random(prefix, NULL, &t);
1744 if (r < 0)
1745 return r;
1746
1747 if (mkdir(t, 0777) < 0)
1748 return -errno;
1749
1750 if (chmod(t, 01777) < 0) {
1751 r = -errno;
1752 (void) rmdir(t);
1753 return r;
1754 }
1755
1756 if (rename(t, prefix) < 0) {
1757 r = -errno;
1758 (void) rmdir(t);
1759 return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */
1760 }
1761
1762 return 0;
1763
1764}
1765
613b411c
LP
1766static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) {
1767 _cleanup_free_ char *x = NULL;
6b46ea73
LP
1768 char bid[SD_ID128_STRING_MAX];
1769 sd_id128_t boot_id;
1770 int r;
613b411c
LP
1771
1772 assert(id);
1773 assert(prefix);
1774 assert(path);
1775
6b46ea73
LP
1776 /* We include the boot id in the directory so that after a
1777 * reboot we can easily identify obsolete directories. */
1778
1779 r = sd_id128_get_boot(&boot_id);
1780 if (r < 0)
1781 return r;
1782
605405c6 1783 x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX");
613b411c
LP
1784 if (!x)
1785 return -ENOMEM;
1786
a652f050
JR
1787 r = make_tmp_prefix(prefix);
1788 if (r < 0)
1789 return r;
1790
613b411c
LP
1791 RUN_WITH_UMASK(0077)
1792 if (!mkdtemp(x))
1793 return -errno;
1794
1795 RUN_WITH_UMASK(0000) {
1796 char *y;
1797
63c372cb 1798 y = strjoina(x, "/tmp");
613b411c
LP
1799
1800 if (mkdir(y, 0777 | S_ISVTX) < 0)
1801 return -errno;
c17ec25e 1802 }
15ae422b 1803
1cc6c93a 1804 *path = TAKE_PTR(x);
613b411c
LP
1805
1806 return 0;
1807}
1808
1809int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) {
1810 char *a, *b;
1811 int r;
1812
1813 assert(id);
1814 assert(tmp_dir);
1815 assert(var_tmp_dir);
1816
1817 r = setup_one_tmp_dir(id, "/tmp", &a);
1818 if (r < 0)
1819 return r;
1820
1821 r = setup_one_tmp_dir(id, "/var/tmp", &b);
1822 if (r < 0) {
1823 char *t;
1824
63c372cb 1825 t = strjoina(a, "/tmp");
6990fb6b
LP
1826 (void) rmdir(t);
1827 (void) rmdir(a);
613b411c
LP
1828
1829 free(a);
1830 return r;
1831 }
1832
1833 *tmp_dir = a;
1834 *var_tmp_dir = b;
1835
1836 return 0;
1837}
1838
2caa38e9 1839int setup_netns(const int netns_storage_socket[static 2]) {
613b411c 1840 _cleanup_close_ int netns = -1;
3ee897d6 1841 int r, q;
613b411c
LP
1842
1843 assert(netns_storage_socket);
1844 assert(netns_storage_socket[0] >= 0);
1845 assert(netns_storage_socket[1] >= 0);
1846
1847 /* We use the passed socketpair as a storage buffer for our
76cd584b
LP
1848 * namespace reference fd. Whatever process runs this first
1849 * shall create a new namespace, all others should just join
1850 * it. To serialize that we use a file lock on the socket
1851 * pair.
613b411c
LP
1852 *
1853 * It's a bit crazy, but hey, works great! */
1854
1855 if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
1856 return -errno;
1857
3ee897d6
LP
1858 netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
1859 if (netns == -EAGAIN) {
44ffcbae 1860 /* Nothing stored yet, so let's create a new namespace. */
613b411c
LP
1861
1862 if (unshare(CLONE_NEWNET) < 0) {
1863 r = -errno;
1864 goto fail;
1865 }
1866
44ffcbae 1867 (void) loopback_setup();
613b411c
LP
1868
1869 netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY);
1870 if (netns < 0) {
1871 r = -errno;
1872 goto fail;
1873 }
1874
1875 r = 1;
613b411c 1876
3ee897d6
LP
1877 } else if (netns < 0) {
1878 r = netns;
1879 goto fail;
613b411c 1880
3ee897d6
LP
1881 } else {
1882 /* Yay, found something, so let's join the namespace */
613b411c
LP
1883 if (setns(netns, CLONE_NEWNET) < 0) {
1884 r = -errno;
1885 goto fail;
1886 }
1887
1888 r = 0;
1889 }
1890
3ee897d6
LP
1891 q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
1892 if (q < 0) {
1893 r = q;
613b411c
LP
1894 goto fail;
1895 }
1896
1897fail:
fe048ce5 1898 (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
15ae422b
LP
1899 return r;
1900}
417116f2 1901
2caa38e9 1902int open_netns_path(const int netns_storage_socket[static 2], const char *path) {
51af7fb2
LP
1903 _cleanup_close_ int netns = -1;
1904 int q, r;
1905
1906 assert(netns_storage_socket);
1907 assert(netns_storage_socket[0] >= 0);
1908 assert(netns_storage_socket[1] >= 0);
1909 assert(path);
1910
1911 /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in
1912 * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a
1913 * new anonymous netns if needed. */
1914
1915 if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
1916 return -errno;
1917
1918 netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
1919 if (netns == -EAGAIN) {
1920 /* Nothing stored yet. Open the file from the file system. */
1921
1922 netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC);
1923 if (netns < 0) {
1924 r = -errno;
1925 goto fail;
1926 }
1927
1928 r = fd_is_network_ns(netns);
1929 if (r == 0) { /* Not a netns? Refuse early. */
1930 r = -EINVAL;
1931 goto fail;
1932 }
1933 if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */
1934 goto fail;
1935
1936 r = 1;
1937
1938 } else if (netns < 0) {
1939 r = netns;
1940 goto fail;
1941 } else
1942 r = 0; /* Already allocated */
1943
1944 q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
1945 if (q < 0) {
1946 r = q;
1947 goto fail;
1948 }
1949
1950fail:
1951 (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
1952 return r;
1953}
1954
6e2d7c4f
MS
1955bool ns_type_supported(NamespaceType type) {
1956 const char *t, *ns_proc;
1957
0fa5b831
LP
1958 t = namespace_type_to_string(type);
1959 if (!t) /* Don't know how to translate this? Then it's not supported */
6e2d7c4f
MS
1960 return false;
1961
6e2d7c4f 1962 ns_proc = strjoina("/proc/self/ns/", t);
6e2d7c4f
MS
1963 return access(ns_proc, F_OK) == 0;
1964}
1965
1b8689f9
LP
1966static const char *const protect_home_table[_PROTECT_HOME_MAX] = {
1967 [PROTECT_HOME_NO] = "no",
1968 [PROTECT_HOME_YES] = "yes",
1969 [PROTECT_HOME_READ_ONLY] = "read-only",
e4da7d8c 1970 [PROTECT_HOME_TMPFS] = "tmpfs",
417116f2
LP
1971};
1972
1e8c7bd5 1973DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES);
5e1c6154 1974
1b8689f9
LP
1975static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = {
1976 [PROTECT_SYSTEM_NO] = "no",
1977 [PROTECT_SYSTEM_YES] = "yes",
1978 [PROTECT_SYSTEM_FULL] = "full",
3f815163 1979 [PROTECT_SYSTEM_STRICT] = "strict",
1b8689f9
LP
1980};
1981
1e8c7bd5 1982DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES);
03c791aa 1983
6e2d7c4f
MS
1984static const char* const namespace_type_table[] = {
1985 [NAMESPACE_MOUNT] = "mnt",
1986 [NAMESPACE_CGROUP] = "cgroup",
1987 [NAMESPACE_UTS] = "uts",
1988 [NAMESPACE_IPC] = "ipc",
1989 [NAMESPACE_USER] = "user",
1990 [NAMESPACE_PID] = "pid",
1991 [NAMESPACE_NET] = "net",
1992};
1993
1994DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType);