]> git.ipfire.org Git - thirdparty/systemd.git/blame - src/core/namespace.c
projects / thirdparty / systemd.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
namespace: move protect_{home|system} into NamespaceInfo
[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"
b3d13314 18#include "list.h"
915e6d16 19#include "loop-util.h"
07630cea 20#include "loopback-setup.h"
07630cea 21#include "mkdir.h"
4349cd7c 22#include "mount-util.h"
049af8ad 23#include "mountpoint-util.h"
0cb8e3d1 24#include "namespace-util.h"
3ffd4af2 25#include "namespace.h"
d8b4d14d 26#include "nulstr-util.h"
07630cea 27#include "path-util.h"
d7b8eec7 28#include "selinux-util.h"
2583fbea 29#include "socket-util.h"
760877e9 30#include "sort-util.h"
36ce7110 31#include "stat-util.h"
8b43440b 32#include "string-table.h"
07630cea
LP		 33#include "string-util.h"
34#include "strv.h"
a652f050 35#include "tmpfile-util.h"
affb60b1 36#include "umask-util.h"
ee104e11 37#include "user-util.h"
15ae422b 38
737ba3c8 39#define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC)
40
c17ec25e 41typedef enum MountMode {
15ae422b
LP		 42 /* This is ordered by priority! */
43 INACCESSIBLE,
b3d13314 44 MOUNT_IMAGES,
d2d6c096
LP		 45 BIND_MOUNT,
46 BIND_MOUNT_RECURSIVE,
ac0930c8 47 PRIVATE_TMP,
56a13a49 48 PRIVATE_TMP_READONLY,
7f112f50 49 PRIVATE_DEV,
5d997827 50 BIND_DEV,
6c47cd7d 51 EMPTY_DIR,
5d997827
LP		 52 SYSFS,
53 PROCFS,
54 READONLY,
59eeb84b 55 READWRITE,
2abd4e38 56 TMPFS,
1e05071d 57 READWRITE_IMPLICIT, /* Should have the lowest priority. */
5beb8688 58 _MOUNT_MODE_MAX,
c17ec25e 59} MountMode;
15ae422b 60
34de407a 61typedef struct MountEntry {
5327c910 62 const char *path_const; /* Memory allocated on stack or static */
cfbeb4ef 63 MountMode mode:5;
5327c910
LP		 64 bool ignore:1; /* Ignore if path does not exist? */
65 bool has_prefix:1; /* Already is prefixed by the root dir? */
cfbeb4ef 66 bool read_only:1; /* Shall this mount point be read-only? */
9ce4e4b0 67 bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */
088696fe 68 bool applied:1; /* Already applied */
55fe7432 69 char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */
b3d13314 70 const char *source_const; /* The source path, for bind mounts or images */
d2d6c096 71 char *source_malloc;
2abd4e38
YW		 72 const char *options_const;/* Mount options for tmpfs */
73 char *options_malloc;
74 unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */
088696fe 75 unsigned n_followed;
427353f6 76 LIST_HEAD(MountOptions, image_options);
34de407a 77} MountEntry;
15ae422b 78
5d997827 79/* 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 80 * something there already. These mounts are hence overridden by any other explicitly configured mounts. */
5d997827
LP		 81static const MountEntry apivfs_table[] = {
82 { "/proc", PROCFS, false },
83 { "/dev", BIND_DEV, false },
84 { "/sys", SYSFS, false },
85};
f471b2af 86
11a30cec 87/* ProtectKernelTunables= option and the related filesystem APIs */
34de407a 88static const MountEntry protect_kernel_tunables_table[] = {
1e05071d
YW		 89 { "/proc/acpi", READONLY, true },
90 { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */
91 { "/proc/asound", READONLY, true },
92 { "/proc/bus", READONLY, true },
93 { "/proc/fs", READONLY, true },
94 { "/proc/irq", READONLY, true },
95 { "/proc/kallsyms", INACCESSIBLE, true },
96 { "/proc/kcore", INACCESSIBLE, true },
97 { "/proc/latency_stats", READONLY, true },
98 { "/proc/mtrr", READONLY, true },
99 { "/proc/scsi", READONLY, true },
100 { "/proc/sys", READONLY, false },
101 { "/proc/sysrq-trigger", READONLY, true },
102 { "/proc/timer_stats", READONLY, true },
103 { "/sys", READONLY, false },
104 { "/sys/fs/bpf", READONLY, true },
105 { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */
106 { "/sys/fs/selinux", READWRITE_IMPLICIT, true },
107 { "/sys/kernel/debug", READONLY, true },
108 { "/sys/kernel/tracing", READONLY, true },
11a30cec
DH		 109};
110
c575770b 111/* ProtectKernelModules= option */
34de407a 112static const MountEntry protect_kernel_modules_table[] = {
349cc4a5 113#if HAVE_SPLIT_USR
c6232fb0 114 { "/lib/modules", INACCESSIBLE, true },
c575770b 115#endif
c6232fb0 116 { "/usr/lib/modules", INACCESSIBLE, true },
c575770b
DH		 117};
118
94a7b275
KK		 119/* ProtectKernelLogs= option */
120static const MountEntry protect_kernel_logs_table[] = {
121 { "/proc/kmsg", INACCESSIBLE, true },
122 { "/dev/kmsg", INACCESSIBLE, true },
123};
124
b6c432ca
DH		 125/*
126 * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of
127 * system should be protected by ProtectSystem=
128 */
34de407a 129static const MountEntry protect_home_read_only_table[] = {
c6232fb0
LP		 130 { "/home", READONLY, true },
131 { "/run/user", READONLY, true },
132 { "/root", READONLY, true },
b6c432ca
DH		 133};
134
e4da7d8c
YW		 135/* ProtectHome=tmpfs table */
136static const MountEntry protect_home_tmpfs_table[] = {
7d85383e
TM		 137 { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME },
138 { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME },
139 { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME },
e4da7d8c
YW		 140};
141
b6c432ca 142/* ProtectHome=yes table */
34de407a 143static const MountEntry protect_home_yes_table[] = {
c6232fb0
LP		 144 { "/home", INACCESSIBLE, true },
145 { "/run/user", INACCESSIBLE, true },
146 { "/root", INACCESSIBLE, true },
b6c432ca
DH		 147};
148
f471b2af 149/* ProtectSystem=yes table */
34de407a 150static const MountEntry protect_system_yes_table[] = {
c6232fb0
LP		 151 { "/usr", READONLY, false },
152 { "/boot", READONLY, true },
153 { "/efi", READONLY, true },
7486f305
AB		 154#if HAVE_SPLIT_USR
155 { "/lib", READONLY, true },
156 { "/lib64", READONLY, true },
157 { "/bin", READONLY, true },
671f0f8d 158# if HAVE_SPLIT_BIN
7486f305 159 { "/sbin", READONLY, true },
671f0f8d 160# endif
7486f305 161#endif
f471b2af
DH		 162};
163
164/* ProtectSystem=full includes ProtectSystem=yes */
34de407a 165static const MountEntry protect_system_full_table[] = {
c6232fb0
LP		 166 { "/usr", READONLY, false },
167 { "/boot", READONLY, true },
168 { "/efi", READONLY, true },
169 { "/etc", READONLY, false },
7486f305
AB		 170#if HAVE_SPLIT_USR
171 { "/lib", READONLY, true },
172 { "/lib64", READONLY, true },
173 { "/bin", READONLY, true },
671f0f8d 174# if HAVE_SPLIT_BIN
7486f305 175 { "/sbin", READONLY, true },
671f0f8d 176# endif
7486f305 177#endif
f471b2af
DH		 178};
179
180/*
181 * ProtectSystem=strict table. In this strict mode, we mount everything
182 * read-only, except for /proc, /dev, /sys which are the kernel API VFS,
183 * which are left writable, but PrivateDevices= + ProtectKernelTunables=
184 * protect those, and these options should be fully orthogonal.
185 * (And of course /home and friends are also left writable, as ProtectHome=
186 * shall manage those, orthogonally).
187 */
34de407a 188static const MountEntry protect_system_strict_table[] = {
1e05071d
YW		 189 { "/", READONLY, false },
190 { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */
191 { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */
192 { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */
193 { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */
194 { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */
195 { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */
f471b2af
DH		 196};
197
5beb8688
YW		 198static const char * const mount_mode_table[_MOUNT_MODE_MAX] = {
199 [INACCESSIBLE] = "inaccessible",
200 [BIND_MOUNT] = "bind",
201 [BIND_MOUNT_RECURSIVE] = "rbind",
202 [PRIVATE_TMP] = "private-tmp",
203 [PRIVATE_DEV] = "private-dev",
204 [BIND_DEV] = "bind-dev",
205 [EMPTY_DIR] = "empty",
206 [SYSFS] = "sysfs",
207 [PROCFS] = "procfs",
208 [READONLY] = "read-only",
209 [READWRITE] = "read-write",
210 [TMPFS] = "tmpfs",
b3d13314 211 [MOUNT_IMAGES] = "mount-images",
5beb8688
YW		 212 [READWRITE_IMPLICIT] = "rw-implicit",
213};
214
215DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode);
216
34de407a 217static const char *mount_entry_path(const MountEntry *p) {
f0a4feb0
DH		 218 assert(p);
219
5327c910
LP		 220 /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that,
221 * otherwise the stack/static ->path field is returned. */
f0a4feb0 222
5327c910 223 return p->path_malloc ?: p->path_const;
f0a4feb0
DH		 224}
225
34de407a 226static bool mount_entry_read_only(const MountEntry *p) {
cfbeb4ef
LP		 227 assert(p);
228
56a13a49 229 return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE, PRIVATE_TMP_READONLY);
cfbeb4ef
LP		 230}
231
d2d6c096
LP		 232static const char *mount_entry_source(const MountEntry *p) {
233 assert(p);
234
235 return p->source_malloc ?: p->source_const;
236}
237
2abd4e38
YW		 238static const char *mount_entry_options(const MountEntry *p) {
239 assert(p);
240
241 return p->options_malloc ?: p->options_const;
242}
243
1eb7e08e
LP		 244static void mount_entry_done(MountEntry *p) {
245 assert(p);
246
247 p->path_malloc = mfree(p->path_malloc);
248 p->source_malloc = mfree(p->source_malloc);
2abd4e38 249 p->options_malloc = mfree(p->options_malloc);
427353f6 250 p->image_options = mount_options_free_all(p->image_options);
1eb7e08e
LP		 251}
252
d18aff04 253static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) {
15ae422b
LP		 254 char **i;
255
613b411c
LP		 256 assert(p);
257
1e05071d 258 /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */
5327c910 259
15ae422b 260 STRV_FOREACH(i, strv) {
5327c910
LP		 261 bool ignore = false, needs_prefix = false;
262 const char *e = *i;
15ae422b 263
5327c910
LP		 264 /* Look for any prefixes */
265 if (startswith(e, "-")) {
266 e++;
9c94d52e 267 ignore = true;
ea92ae33 268 }
5327c910
LP		 269 if (startswith(e, "+")) {
270 e++;
271 needs_prefix = true;
272 }
ea92ae33 273
baaa35ad
ZJS		 274 if (!path_is_absolute(e))
275 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
276 "Path is not absolute: %s", e);
15ae422b 277
34de407a 278 *((*p)++) = (MountEntry) {
5327c910
LP		 279 .path_const = e,
280 .mode = mode,
281 .ignore = ignore,
d18aff04 282 .has_prefix = !needs_prefix && !forcibly_require_prefix,
5327c910 283 };
15ae422b
LP		 284 }
285
286 return 0;
287}
288
6c47cd7d
LP		 289static int append_empty_dir_mounts(MountEntry **p, char **strv) {
290 char **i;
291
292 assert(p);
293
294 /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the
295 * "/private/" boundary directories for DynamicUser=1. */
296
297 STRV_FOREACH(i, strv) {
298
299 *((*p)++) = (MountEntry) {
300 .path_const = *i,
301 .mode = EMPTY_DIR,
302 .ignore = false,
6c47cd7d 303 .read_only = true,
7d85383e 304 .options_const = "mode=755" TMPFS_LIMITS_EMPTY_OR_ALMOST,
2abd4e38 305 .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
6c47cd7d
LP		 306 };
307 }
308
309 return 0;
310}
311
da6053d0
LP		 312static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) {
313 size_t i;
d2d6c096
LP		 314
315 assert(p);
316
317 for (i = 0; i < n; i++) {
318 const BindMount *b = binds + i;
319
320 *((*p)++) = (MountEntry) {
321 .path_const = b->destination,
322 .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT,
323 .read_only = b->read_only,
9ce4e4b0 324 .nosuid = b->nosuid,
d2d6c096 325 .source_const = b->source,
4ca763a9 326 .ignore = b->ignore_enoent,
d2d6c096
LP		 327 };
328 }
329
330 return 0;
331}
332
b3d13314
LB		 333static int append_mount_images(MountEntry **p, const MountImage *mount_images, size_t n) {
334 assert(p);
335
336 for (size_t i = 0; i < n; i++) {
337 const MountImage *m = mount_images + i;
338
339 *((*p)++) = (MountEntry) {
340 .path_const = m->destination,
341 .mode = MOUNT_IMAGES,
342 .source_const = m->source,
427353f6 343 .image_options = m->mount_options,
b3d13314
LB		 344 .ignore = m->ignore_enoent,
345 };
346 }
347
348 return 0;
349}
350
da6053d0 351static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) {
2abd4e38
YW		 352 assert(p);
353
b67ec8e5 354 for (size_t i = 0; i < n; i++) {
2abd4e38
YW		 355 const TemporaryFileSystem *t = tmpfs + i;
356 _cleanup_free_ char *o = NULL, *str = NULL;
ad8e66dc 357 unsigned long flags;
2abd4e38 358 bool ro = false;
b67ec8e5 359 int r;
2abd4e38 360
baaa35ad
ZJS		 361 if (!path_is_absolute(t->path))
362 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
363 "Path is not absolute: %s",
364 t->path);
2abd4e38 365
b67ec8e5 366 str = strjoin("mode=0755" NESTED_TMPFS_LIMITS ",", t->options);
ad8e66dc
AJ		 367 if (!str)
368 return -ENOMEM;
2abd4e38 369
ad8e66dc
AJ		 370 r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o);
371 if (r < 0)
372 return log_debug_errno(r, "Failed to parse mount option '%s': %m", str);
2abd4e38 373
ad8e66dc
AJ		 374 ro = flags & MS_RDONLY;
375 if (ro)
376 flags ^= MS_RDONLY;
2abd4e38
YW		 377
378 *((*p)++) = (MountEntry) {
379 .path_const = t->path,
380 .mode = TMPFS,
381 .read_only = ro,
ad8e66dc 382 .options_malloc = TAKE_PTR(o),
2abd4e38
YW		 383 .flags = flags,
384 };
2abd4e38
YW		 385 }
386
387 return 0;
388}
389
da6053d0
LP		 390static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) {
391 size_t i;
11a30cec
DH		 392
393 assert(p);
f471b2af 394 assert(mounts);
11a30cec 395
5327c910 396 /* Adds a list of static pre-defined entries */
f471b2af 397
5327c910 398 for (i = 0; i < n; i++)
34de407a
LP		 399 *((*p)++) = (MountEntry) {
400 .path_const = mount_entry_path(mounts+i),
5327c910
LP		 401 .mode = mounts[i].mode,
402 .ignore = mounts[i].ignore || ignore_protect,
403 };
f471b2af
DH		 404
405 return 0;
406}
407
34de407a 408static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) {
c575770b
DH		 409 assert(p);
410
5327c910 411 switch (protect_home) {
b6c432ca 412
5327c910 413 case PROTECT_HOME_NO:
b6c432ca
DH		 414 return 0;
415
b6c432ca 416 case PROTECT_HOME_READ_ONLY:
5327c910
LP		 417 return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect);
418
e4da7d8c
YW		 419 case PROTECT_HOME_TMPFS:
420 return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect);
421
b6c432ca 422 case PROTECT_HOME_YES:
5327c910
LP		 423 return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect);
424
b6c432ca 425 default:
5327c910 426 assert_not_reached("Unexpected ProtectHome= value");
b6c432ca 427 }
b6c432ca
DH		 428}
429
34de407a 430static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) {
f471b2af
DH		 431 assert(p);
432
5327c910
LP		 433 switch (protect_system) {
434
435 case PROTECT_SYSTEM_NO:
f471b2af
DH		 436 return 0;
437
f471b2af 438 case PROTECT_SYSTEM_STRICT:
5327c910
LP		 439 return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect);
440
f471b2af 441 case PROTECT_SYSTEM_YES:
5327c910
LP		 442 return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect);
443
f471b2af 444 case PROTECT_SYSTEM_FULL:
5327c910
LP		 445 return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect);
446
f471b2af 447 default:
5327c910 448 assert_not_reached("Unexpected ProtectSystem= value");
f471b2af 449 }
11a30cec
DH		 450}
451
93bab288 452static int mount_path_compare(const MountEntry *a, const MountEntry *b) {
a0827e2b 453 int d;
15ae422b 454
6ee1a919 455 /* If the paths are not equal, then order prefixes first */
93bab288 456 d = path_compare(mount_entry_path(a), mount_entry_path(b));
6ee1a919
LP		 457 if (d != 0)
458 return d;
15ae422b 459
6ee1a919 460 /* If the paths are equal, check the mode */
93bab288 461 return CMP((int) a->mode, (int) b->mode);
15ae422b
LP		 462}
463
da6053d0
LP		 464static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) {
465 size_t i;
5327c910 466
4a756839 467 /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */
5327c910
LP		 468
469 for (i = 0; i < n; i++) {
470 char *s;
471
472 if (m[i].has_prefix)
473 continue;
474
c6134d3e 475 s = path_join(root_directory, mount_entry_path(m+i));
5327c910
LP		 476 if (!s)
477 return -ENOMEM;
478
e282f51f 479 free_and_replace(m[i].path_malloc, s);
5327c910
LP		 480 m[i].has_prefix = true;
481 }
482
483 return 0;
484}
485
da6053d0 486static void drop_duplicates(MountEntry *m, size_t *n) {
34de407a 487 MountEntry *f, *t, *previous;
15ae422b 488
c17ec25e 489 assert(m);
15ae422b 490 assert(n);
15ae422b 491
fe3c2583
LP		 492 /* Drops duplicate entries. Expects that the array is properly ordered already. */
493
1d54cd5d 494 for (f = m, t = m, previous = NULL; f < m + *n; f++) {
15ae422b 495
fe3c2583 496 /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare()
088696fe
LP		 497 * above. Note that we only drop duplicates that haven't been mounted yet. */
498 if (previous &&
499 path_equal(mount_entry_path(f), mount_entry_path(previous)) &&
500 !f->applied && !previous->applied) {
5beb8688 501 log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode));
34de407a 502 previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */
1eb7e08e 503 mount_entry_done(f);
15ae422b 504 continue;
fe3c2583 505 }
15ae422b 506
e2d7c1a0 507 *t = *f;
15ae422b 508 previous = t;
fe3c2583
LP		 509 t++;
510 }
511
512 *n = t - m;
513}
514
da6053d0 515static void drop_inaccessible(MountEntry *m, size_t *n) {
34de407a 516 MountEntry *f, *t;
fe3c2583
LP		 517 const char *clear = NULL;
518
519 assert(m);
520 assert(n);
521
522 /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly
523 * ordered already. */
524
1d54cd5d 525 for (f = m, t = m; f < m + *n; f++) {
fe3c2583
LP		 526
527 /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop
528 * it, as inaccessible paths really should drop the entire subtree. */
34de407a
LP		 529 if (clear && path_startswith(mount_entry_path(f), clear)) {
530 log_debug("%s is masked by %s.", mount_entry_path(f), clear);
1eb7e08e 531 mount_entry_done(f);
fe3c2583
LP		 532 continue;
533 }
15ae422b 534
34de407a 535 clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL;
fe3c2583
LP		 536
537 *t = *f;
15ae422b
LP		 538 t++;
539 }
540
c17ec25e 541 *n = t - m;
15ae422b
LP		 542}
543
da6053d0 544static void drop_nop(MountEntry *m, size_t *n) {
34de407a 545 MountEntry *f, *t;
7648a565
LP		 546
547 assert(m);
548 assert(n);
549
550 /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the
551 * list is ordered by prefixes. */
552
1d54cd5d 553 for (f = m, t = m; f < m + *n; f++) {
7648a565 554
1e05071d
YW		 555 /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */
556 if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) {
34de407a 557 MountEntry *p;
7648a565
LP		 558 bool found = false;
559
560 /* Now let's find the first parent of the entry we are looking at. */
561 for (p = t-1; p >= m; p--) {
34de407a 562 if (path_startswith(mount_entry_path(f), mount_entry_path(p))) {
7648a565
LP		 563 found = true;
564 break;
565 }
566 }
567
568 /* We found it, let's see if it's the same mode, if so, we can drop this entry */
569 if (found && p->mode == f->mode) {
5beb8688
YW		 570 log_debug("%s (%s) is made redundant by %s (%s)",
571 mount_entry_path(f), mount_mode_to_string(f->mode),
572 mount_entry_path(p), mount_mode_to_string(p->mode));
1eb7e08e 573 mount_entry_done(f);
7648a565
LP		 574 continue;
575 }
576 }
577
578 *t = *f;
579 t++;
580 }
581
582 *n = t - m;
583}
584
da6053d0 585static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) {
34de407a 586 MountEntry *f, *t;
cd2902c9
LP		 587
588 assert(m);
589 assert(n);
590
1d54cd5d 591 /* Nothing to do */
cd2902c9
LP		 592 if (!root_directory)
593 return;
594
595 /* Drops all mounts that are outside of the root directory. */
596
1d54cd5d 597 for (f = m, t = m; f < m + *n; f++) {
cd2902c9 598
34de407a
LP		 599 if (!path_startswith(mount_entry_path(f), root_directory)) {
600 log_debug("%s is outside of root directory.", mount_entry_path(f));
1eb7e08e 601 mount_entry_done(f);
cd2902c9
LP		 602 continue;
603 }
604
605 *t = *f;
606 t++;
607 }
608
609 *n = t - m;
610}
611
b2a60844
LP		 612static int clone_device_node(
613 const char *d,
614 const char *temporary_mount,
615 bool *make_devnode) {
616
617 _cleanup_free_ char *sl = NULL;
618 const char *dn, *bn, *t;
b5e99f23
ДГ		 619 struct stat st;
620 int r;
621
414b304b 622 if (stat(d, &st) < 0) {
b2a60844
LP		 623 if (errno == ENOENT) {
624 log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d);
af984e13 625 return -ENXIO;
b2a60844
LP		 626 }
627
628 return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d);
b5e99f23
ДГ		 629 }
630
631 if (!S_ISBLK(st.st_mode) &&
baaa35ad
ZJS		 632 !S_ISCHR(st.st_mode))
633 return log_debug_errno(SYNTHETIC_ERRNO(EINVAL),
634 "Device node '%s' to clone is not a device node, ignoring.",
635 d);
b5e99f23 636
6f7f3a33 637 dn = strjoina(temporary_mount, d);
b5e99f23 638
b2a60844 639 /* First, try to create device node properly */
16498617
CB		 640 if (*make_devnode) {
641 mac_selinux_create_file_prepare(d, st.st_mode);
642 r = mknod(dn, st.st_mode, st.st_rdev);
643 mac_selinux_create_file_clear();
b2a60844
LP		 644 if (r >= 0)
645 goto add_symlink;
16498617
CB		 646 if (errno != EPERM)
647 return log_debug_errno(errno, "mknod failed for %s: %m", d);
648
b2a60844 649 /* This didn't work, let's not try this again for the next iterations. */
16498617
CB		 650 *make_devnode = false;
651 }
652
2aed63f4 653 /* We're about to fall back to bind-mounting the device
1acf344d
CG		 654 * node. So create a dummy bind-mount target.
655 * Do not prepare device-node SELinux label (see issue 13762) */
16498617 656 r = mknod(dn, S_IFREG, 0);
16498617 657 if (r < 0 && errno != EEXIST)
b2a60844 658 return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d);
16498617
CB		 659
660 /* Fallback to bind-mounting:
661 * The assumption here is that all used device nodes carry standard
662 * properties. Specifically, the devices nodes we bind-mount should
663 * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx)
664 * and should not carry ACLs. */
665 if (mount(d, dn, NULL, MS_BIND, NULL) < 0)
b2a60844
LP		 666 return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d);
667
668add_symlink:
669 bn = path_startswith(d, "/dev/");
670 if (!bn)
671 return 0;
672
673 /* Create symlinks like /dev/char/1:9 → ../urandom */
cbc056c8
ZJS		 674 if (asprintf(&sl, "%s/dev/%s/%u:%u",
675 temporary_mount,
676 S_ISCHR(st.st_mode) ? "char" : "block",
677 major(st.st_rdev), minor(st.st_rdev)) < 0)
b2a60844
LP		 678 return log_oom();
679
680 (void) mkdir_parents(sl, 0755);
681
682 t = strjoina("../", bn);
b2a60844 683 if (symlink(t, sl) < 0)
2e4a4fae 684 log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl);
b5e99f23 685
af984e13 686 return 0;
b5e99f23
ДГ		 687}
688
5d997827 689static int mount_private_dev(MountEntry *m) {
7f112f50
LP		 690 static const char devnodes[] =
691 "/dev/null\0"
692 "/dev/zero\0"
693 "/dev/full\0"
694 "/dev/random\0"
695 "/dev/urandom\0"
696 "/dev/tty\0";
697
2b85f4e1 698 char temporary_mount[] = "/tmp/namespace-dev-XXXXXX";
63cc4c31 699 const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL;
16498617 700 bool can_mknod = true;
7f112f50
LP		 701 _cleanup_umask_ mode_t u;
702 int r;
703
704 assert(m);
705
706 u = umask(0000);
707
2b85f4e1 708 if (!mkdtemp(temporary_mount))
2e4a4fae 709 return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount);
2b85f4e1 710
63c372cb 711 dev = strjoina(temporary_mount, "/dev");
dc751688 712 (void) mkdir(dev, 0755);
7d85383e 713 if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755" TMPFS_LIMITS_DEV) < 0) {
2e4a4fae 714 r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev);
2b85f4e1
LP		 715 goto fail;
716 }
c3151977
TM		 717 r = label_fix_container(dev, "/dev", 0);
718 if (r < 0) {
719 log_debug_errno(errno, "Failed to fix label of '%s' as /dev: %m", dev);
720 goto fail;
721 }
2b85f4e1 722
63c372cb 723 devpts = strjoina(temporary_mount, "/dev/pts");
dc751688 724 (void) mkdir(devpts, 0755);
2b85f4e1 725 if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) {
2e4a4fae 726 r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts);
2b85f4e1
LP		 727 goto fail;
728 }
729
2e4a4fae
YW		 730 /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx.
731 * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible.
732 * Thus, in that case make a clone.
733 * In nspawn and other containers it will be a symlink, in that case make it a symlink. */
36ce7110 734 r = is_symlink("/dev/ptmx");
2e4a4fae
YW		 735 if (r < 0) {
736 log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m");
3164e3cb 737 goto fail;
2e4a4fae 738 } else if (r > 0) {
414b304b
ДГ		 739 devptmx = strjoina(temporary_mount, "/dev/ptmx");
740 if (symlink("pts/ptmx", devptmx) < 0) {
2e4a4fae 741 r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx);
414b304b
ДГ		 742 goto fail;
743 }
744 } else {
16498617 745 r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod);
152c475f
LP		 746 if (r < 0)
747 goto fail;
414b304b 748 }
e06b6479 749
63c372cb 750 devshm = strjoina(temporary_mount, "/dev/shm");
8d953682 751 (void) mkdir(devshm, 0755);
2b85f4e1
LP		 752 r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL);
753 if (r < 0) {
2e4a4fae 754 r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm);
2b85f4e1
LP		 755 goto fail;
756 }
757
63c372cb 758 devmqueue = strjoina(temporary_mount, "/dev/mqueue");
dc751688 759 (void) mkdir(devmqueue, 0755);
2e4a4fae
YW		 760 if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0)
761 log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue);
2b85f4e1 762
63c372cb 763 devhugepages = strjoina(temporary_mount, "/dev/hugepages");
dc751688 764 (void) mkdir(devhugepages, 0755);
2e4a4fae
YW		 765 if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0)
766 log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages);
2b85f4e1 767
63c372cb 768 devlog = strjoina(temporary_mount, "/dev/log");
2e4a4fae
YW		 769 if (symlink("/run/systemd/journal/dev-log", devlog) < 0)
770 log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog);
82d25240 771
7f112f50 772 NULSTR_FOREACH(d, devnodes) {
16498617 773 r = clone_device_node(d, temporary_mount, &can_mknod);
37b22b3b 774 /* ENXIO means the *source* is not a device file, skip creation in that case */
af984e13 775 if (r < 0 && r != -ENXIO)
2b85f4e1 776 goto fail;
7f112f50
LP		 777 }
778
2e4a4fae
YW		 779 r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID);
780 if (r < 0)
105a1a36 781 log_debug_errno(r, "Failed to set up basic device tree at '%s', ignoring: %m", temporary_mount);
7f112f50 782
ee818b89
AC		 783 /* Create the /dev directory if missing. It is more likely to be
784 * missing when the service is started with RootDirectory. This is
785 * consistent with mount units creating the mount points when missing.
786 */
34de407a 787 (void) mkdir_p_label(mount_entry_path(m), 0755);
ee818b89 788
9e5f8252 789 /* Unmount everything in old /dev */
2e4a4fae
YW		 790 r = umount_recursive(mount_entry_path(m), 0);
791 if (r < 0)
792 log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m));
793
34de407a 794 if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) {
2e4a4fae 795 r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m));
2b85f4e1
LP		 796 goto fail;
797 }
7f112f50 798
1019a48f
LP		 799 (void) rmdir(dev);
800 (void) rmdir(temporary_mount);
7f112f50 801
2b85f4e1 802 return 0;
7f112f50 803
2b85f4e1
LP		 804fail:
805 if (devpts)
1019a48f 806 (void) umount(devpts);
7f112f50 807
2b85f4e1 808 if (devshm)
1019a48f 809 (void) umount(devshm);
7f112f50 810
2b85f4e1 811 if (devhugepages)
1019a48f 812 (void) umount(devhugepages);
7f112f50 813
2b85f4e1 814 if (devmqueue)
1019a48f 815 (void) umount(devmqueue);
7f112f50 816
1019a48f
LP		 817 (void) umount(dev);
818 (void) rmdir(dev);
819 (void) rmdir(temporary_mount);
7f112f50 820
2b85f4e1 821 return r;
7f112f50
LP		 822}
823
2a2969fd 824static int mount_bind_dev(const MountEntry *m) {
5d997827
LP		 825 int r;
826
827 assert(m);
828
829 /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
830 * /dev. This is only used when RootDirectory= is set. */
831
645767d6
LP		 832 (void) mkdir_p_label(mount_entry_path(m), 0755);
833
5d997827
LP		 834 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
835 if (r < 0)
836 return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m");
837 if (r > 0) /* make this a NOP if /dev is already a mount point */
838 return 0;
839
840 if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
841 return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m));
842
843 return 1;
844}
845
2a2969fd 846static int mount_sysfs(const MountEntry *m) {
5d997827
LP		 847 int r;
848
849 assert(m);
850
645767d6
LP		 851 (void) mkdir_p_label(mount_entry_path(m), 0755);
852
5d997827
LP		 853 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
854 if (r < 0)
855 return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m");
856 if (r > 0) /* make this a NOP if /sys is already a mount point */
857 return 0;
858
859 /* Bind mount the host's version so that we get all child mounts of it, too. */
860 if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
861 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
862
863 return 1;
864}
865
2a2969fd 866static int mount_procfs(const MountEntry *m) {
5d997827
LP		 867 int r;
868
869 assert(m);
870
645767d6
LP		 871 (void) mkdir_p_label(mount_entry_path(m), 0755);
872
5d997827
LP		 873 r = path_is_mount_point(mount_entry_path(m), NULL, 0);
874 if (r < 0)
875 return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m");
876 if (r > 0) /* make this a NOP if /proc is already a mount point */
877 return 0;
878
879 /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
880 if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0)
881 return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
882
883 return 1;
884}
885
2abd4e38 886static int mount_tmpfs(const MountEntry *m) {
abad72be
CG		 887 int r;
888 const char *entry_path = mount_entry_path(m);
889 const char *source_path = m->path_const;
890
6c47cd7d
LP		 891 assert(m);
892
2abd4e38 893 /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */
6c47cd7d 894
abad72be
CG		 895 (void) mkdir_p_label(entry_path, 0755);
896 (void) umount_recursive(entry_path, 0);
6c47cd7d 897
abad72be
CG		 898 if (mount("tmpfs", entry_path, "tmpfs", m->flags, mount_entry_options(m)) < 0)
899 return log_debug_errno(errno, "Failed to mount %s: %m", entry_path);
900
901 r = label_fix_container(entry_path, source_path, 0);
902 if (r < 0)
f2df56bf 903 return log_debug_errno(r, "Failed to fix label of '%s' as '%s': %m", entry_path, source_path);
6c47cd7d
LP		 904
905 return 1;
906}
907
b3d13314
LB		 908static int mount_images(const MountEntry *m) {
909 _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
910 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
911 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
912 _cleanup_free_ void *root_hash_decoded = NULL;
913 _cleanup_free_ char *verity_data = NULL, *hash_sig = NULL;
914 DissectImageFlags dissect_image_flags = m->read_only ? DISSECT_IMAGE_READ_ONLY : 0;
915 size_t root_hash_size = 0;
916 int r;
917
918 r = verity_metadata_load(mount_entry_source(m), NULL, &root_hash_decoded, &root_hash_size, &verity_data, &hash_sig);
919 if (r < 0)
920 return log_debug_errno(r, "Failed to load root hash: %m");
921 dissect_image_flags |= verity_data ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0;
922
923 r = loop_device_make_by_path(mount_entry_source(m),
924 m->read_only ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */,
925 verity_data ? 0 : LO_FLAGS_PARTSCAN,
926 &loop_device);
927 if (r < 0)
928 return log_debug_errno(r, "Failed to create loop device for image: %m");
929
427353f6 930 r = dissect_image(loop_device->fd, root_hash_decoded, root_hash_size, verity_data, m->image_options, dissect_image_flags, &dissected_image);
b3d13314
LB		 931 /* No partition table? Might be a single-filesystem image, try again */
932 if (!verity_data && r < 0 && r == -ENOPKG)
427353f6 933 r = dissect_image(loop_device->fd, root_hash_decoded, root_hash_size, verity_data, m->image_options, dissect_image_flags|DISSECT_IMAGE_NO_PARTITION_TABLE, &dissected_image);
b3d13314
LB		 934 if (r < 0)
935 return log_debug_errno(r, "Failed to dissect image: %m");
936
937 r = dissected_image_decrypt(dissected_image, NULL, root_hash_decoded, root_hash_size, verity_data, hash_sig, NULL, 0, dissect_image_flags, &decrypted_image);
938 if (r < 0)
939 return log_debug_errno(r, "Failed to decrypt dissected image: %m");
940
941 r = mkdir_p_label(mount_entry_path(m), 0755);
942 if (r < 0)
943 return log_debug_errno(r, "Failed to create destination directory %s: %m", mount_entry_path(m));
944 r = umount_recursive(mount_entry_path(m), 0);
945 if (r < 0)
946 return log_debug_errno(r, "Failed to umount under destination directory %s: %m", mount_entry_path(m));
947
948 r = dissected_image_mount(dissected_image, mount_entry_path(m), UID_INVALID, dissect_image_flags);
949 if (r < 0)
950 return log_debug_errno(r, "Failed to mount image: %m");
951
952 if (decrypted_image) {
953 r = decrypted_image_relinquish(decrypted_image);
954 if (r < 0)
955 return log_debug_errno(r, "Failed to relinquish decrypted image: %m");
956 }
957
958 loop_device_relinquish(loop_device);
959
960 return 1;
961}
962
088696fe 963static int follow_symlink(
d2d6c096 964 const char *root_directory,
088696fe 965 MountEntry *m) {
d2d6c096 966
088696fe 967 _cleanup_free_ char *target = NULL;
8fceda93
LP		 968 int r;
969
088696fe
LP		 970 /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we
971 * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at
972 * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the
973 * end and already have a fully normalized name. */
8fceda93 974
a5648b80 975 r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL);
088696fe
LP		 976 if (r < 0)
977 return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m));
978 if (r > 0) /* Reached the end, nothing more to resolve */
979 return 1;
8fceda93 980
baaa35ad
ZJS		 981 if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */
982 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
983 "Symlink loop on '%s'.",
984 mount_entry_path(m));
8fceda93 985
088696fe 986 log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target);
8fceda93 987
088696fe
LP		 988 free_and_replace(m->path_malloc, target);
989 m->has_prefix = true;
8fceda93 990
088696fe
LP		 991 m->n_followed ++;
992
993 return 0;
8fceda93
LP		 994}
995
ac0930c8 996static int apply_mount(
8fceda93 997 const char *root_directory,
89bd586c 998 MountEntry *m) {
ac0930c8 999
e5f10caf 1000 _cleanup_free_ char *inaccessible = NULL;
a227a4be 1001 bool rbind = true, make = false;
15ae422b 1002 const char *what;
15ae422b 1003 int r;
15ae422b 1004
c17ec25e 1005 assert(m);
15ae422b 1006
34de407a 1007 log_debug("Applying namespace mount on %s", mount_entry_path(m));
fe3c2583 1008
c17ec25e 1009 switch (m->mode) {
15ae422b 1010
160cfdbe 1011 case INACCESSIBLE: {
e5f10caf
AZ		 1012 _cleanup_free_ char *tmp = NULL;
1013 const char *runtime_dir;
160cfdbe 1014 struct stat target;
6d313367
LP		 1015
1016 /* First, get rid of everything that is below if there
1017 * is anything... Then, overmount it with an
c4b41707 1018 * inaccessible path. */
34de407a 1019 (void) umount_recursive(mount_entry_path(m), 0);
6d313367 1020
088696fe
LP		 1021 if (lstat(mount_entry_path(m), &target) < 0) {
1022 if (errno == ENOENT && m->ignore)
1023 return 0;
1024
cbc056c8
ZJS		 1025 return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m",
1026 mount_entry_path(m));
088696fe 1027 }
15ae422b 1028
e5f10caf 1029 if (geteuid() == 0)
48b747fa 1030 runtime_dir = "/run";
e5f10caf 1031 else {
48b747fa
LP		 1032 if (asprintf(&tmp, "/run/user/" UID_FMT, geteuid()) < 0)
1033 return -ENOMEM;
e5f10caf
AZ		 1034
1035 runtime_dir = tmp;
1036 }
1037
1038 r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible);
1039 if (r < 0)
baaa35ad
ZJS		 1040 return log_debug_errno(SYNTHETIC_ERRNO(ELOOP),
1041 "File type not supported for inaccessible mounts. Note that symlinks are not allowed");
e5f10caf 1042 what = inaccessible;
c4b41707 1043 break;
160cfdbe 1044 }
fe3c2583 1045
15ae422b 1046 case READONLY:
15ae422b 1047 case READWRITE:
1e05071d 1048 case READWRITE_IMPLICIT:
8fceda93 1049 r = path_is_mount_point(mount_entry_path(m), root_directory, 0);
088696fe
LP		 1050 if (r == -ENOENT && m->ignore)
1051 return 0;
d944dc95 1052 if (r < 0)
cbc056c8
ZJS		 1053 return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m",
1054 mount_entry_path(m));
1055 if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY
1056 * bit for the mount point if needed. */
6b7c9f8b 1057 return 0;
6b7c9f8b 1058 /* This isn't a mount point yet, let's make it one. */
34de407a 1059 what = mount_entry_path(m);
6b7c9f8b 1060 break;
15ae422b 1061
d2d6c096
LP		 1062 case BIND_MOUNT:
1063 rbind = false;
d2d6c096 1064
4831981d 1065 _fallthrough_;
088696fe
LP		 1066 case BIND_MOUNT_RECURSIVE: {
1067 _cleanup_free_ char *chased = NULL;
5d997827 1068
cbc056c8
ZJS		 1069 /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note
1070 * that bind mount source paths are always relative to the host root, hence we pass NULL as
1071 * root directory to chase_symlinks() here. */
088696fe 1072
a5648b80 1073 r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL);
088696fe
LP		 1074 if (r == -ENOENT && m->ignore) {
1075 log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m));
1076 return 0;
1077 }
1078 if (r < 0)
1079 return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m));
1080
1081 log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased);
1082
1083 free_and_replace(m->source_malloc, chased);
d2d6c096
LP		 1084
1085 what = mount_entry_source(m);
a227a4be 1086 make = true;
d2d6c096 1087 break;
088696fe 1088 }
d2d6c096 1089
6c47cd7d 1090 case EMPTY_DIR:
2abd4e38
YW		 1091 case TMPFS:
1092 return mount_tmpfs(m);
6c47cd7d 1093
ac0930c8 1094 case PRIVATE_TMP:
56a13a49 1095 case PRIVATE_TMP_READONLY:
89bd586c 1096 what = mount_entry_source(m);
a227a4be 1097 make = true;
15ae422b 1098 break;
e364ad06 1099
d6797c92 1100 case PRIVATE_DEV:
5d997827
LP		 1101 return mount_private_dev(m);
1102
1103 case BIND_DEV:
1104 return mount_bind_dev(m);
1105
1106 case SYSFS:
1107 return mount_sysfs(m);
1108
1109 case PROCFS:
1110 return mount_procfs(m);
d6797c92 1111
b3d13314
LB		 1112 case MOUNT_IMAGES:
1113 return mount_images(m);
1114
e364ad06
LP		 1115 default:
1116 assert_not_reached("Unknown mode");
15ae422b
LP		 1117 }
1118
ac0930c8 1119 assert(what);
15ae422b 1120
a227a4be
LP		 1121 if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) {
1122 bool try_again = false;
1123 r = -errno;
1124
1125 if (r == -ENOENT && make) {
1126 struct stat st;
1127
cbc056c8
ZJS		 1128 /* Hmm, either the source or the destination are missing. Let's see if we can create
1129 the destination, then try again. */
a227a4be 1130
e8717862 1131 if (stat(what, &st) < 0)
5dc60faa 1132 log_error_errno(errno, "Mount point source '%s' is not accessible: %m", what);
e8717862
LP		 1133 else {
1134 int q;
a227a4be
LP		 1135
1136 (void) mkdir_parents(mount_entry_path(m), 0755);
1137
1138 if (S_ISDIR(st.st_mode))
e8717862 1139 q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0;
a227a4be 1140 else
e8717862
LP		 1141 q = touch(mount_entry_path(m));
1142
1143 if (q < 0)
cbc056c8
ZJS		 1144 log_error_errno(q, "Failed to create destination mount point node '%s': %m",
1145 mount_entry_path(m));
e8717862
LP		 1146 else
1147 try_again = true;
a227a4be
LP		 1148 }
1149 }
1150
1151 if (try_again) {
1152 if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0)
1153 r = -errno;
1154 else
1155 r = 0;
1156 }
1157
1158 if (r < 0)
5dc60faa 1159 return log_error_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m));
a227a4be 1160 }
6b7c9f8b 1161
34de407a 1162 log_debug("Successfully mounted %s to %s", what, mount_entry_path(m));
6b7c9f8b 1163 return 0;
ac0930c8 1164}
15ae422b 1165
6b000af4 1166static int make_read_only(const MountEntry *m, char **deny_list, FILE *proc_self_mountinfo) {
9ce4e4b0 1167 unsigned long new_flags = 0, flags_mask = 0;
763a260a 1168 bool submounts = false;
6b7c9f8b 1169 int r = 0;
15ae422b 1170
c17ec25e 1171 assert(m);
ac9de0b3 1172 assert(proc_self_mountinfo);
ac0930c8 1173
9ce4e4b0
LP		 1174 if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) {
1175 new_flags |= MS_RDONLY;
1176 flags_mask |= MS_RDONLY;
1177 }
1178
1179 if (m->nosuid) {
1180 new_flags |= MS_NOSUID;
1181 flags_mask |= MS_NOSUID;
1182 }
1183
1184 if (flags_mask == 0) /* No Change? */
6b7c9f8b
LP		 1185 return 0;
1186
9ce4e4b0
LP		 1187 /* We generally apply these changes recursively, except for /dev, and the cases we know there's
1188 * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the
1189 * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace
1190 * and running Linux <= 4.17. */
1191 submounts =
1192 mount_entry_read_only(m) &&
1193 !IN_SET(m->mode, EMPTY_DIR, TMPFS);
1194 if (submounts)
6b000af4 1195 r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, deny_list, proc_self_mountinfo);
9ce4e4b0 1196 else
7cce68e1 1197 r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo);
9ce4e4b0 1198
867189b5
LP		 1199 /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked
1200 * read-only already stays this way. This improves compatibility with container managers, where we
1201 * won't attempt to undo read-only mounts already applied. */
ac0930c8 1202
8fceda93 1203 if (r == -ENOENT && m->ignore)
867189b5 1204 return 0;
763a260a 1205 if (r < 0)
9ce4e4b0 1206 return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m),
763a260a 1207 submounts ? " and its submounts" : "");
763a260a 1208 return 0;
d944dc95
LP		 1209}
1210
9b68367b 1211static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) {
5d997827
LP		 1212 assert(ns_info);
1213
9c988f93
DH		 1214 /*
1215 * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=,
1216 * since to protect the API VFS mounts, they need to be around in the
9b68367b 1217 * first place...
9c988f93 1218 */
5d997827 1219
9b68367b
YW		 1220 return ns_info->mount_apivfs ||
1221 ns_info->protect_control_groups ||
1222 ns_info->protect_kernel_tunables;
5d997827
LP		 1223}
1224
da6053d0 1225static size_t namespace_calculate_mounts(
bb0ff3fb 1226 const NamespaceInfo *ns_info,
2652c6c1
DH		 1227 char** read_write_paths,
1228 char** read_only_paths,
1229 char** inaccessible_paths,
6c47cd7d 1230 char** empty_directories,
da6053d0
LP		 1231 size_t n_bind_mounts,
1232 size_t n_temporary_filesystems,
b3d13314 1233 size_t n_mount_images,
2652c6c1
DH		 1234 const char* tmp_dir,
1235 const char* var_tmp_dir,
52b3d652 1236 const char* log_namespace) {
2652c6c1 1237
da6053d0
LP		 1238 size_t protect_home_cnt;
1239 size_t protect_system_cnt =
52b3d652 1240 (ns_info->protect_system == PROTECT_SYSTEM_STRICT ?
f471b2af 1241 ELEMENTSOF(protect_system_strict_table) :
52b3d652 1242 ((ns_info->protect_system == PROTECT_SYSTEM_FULL) ?
f471b2af 1243 ELEMENTSOF(protect_system_full_table) :
52b3d652 1244 ((ns_info->protect_system == PROTECT_SYSTEM_YES) ?
f471b2af
DH		 1245 ELEMENTSOF(protect_system_yes_table) : 0)));
1246
b6c432ca 1247 protect_home_cnt =
52b3d652 1248 (ns_info->protect_home == PROTECT_HOME_YES ?
b6c432ca 1249 ELEMENTSOF(protect_home_yes_table) :
52b3d652 1250 ((ns_info->protect_home == PROTECT_HOME_READ_ONLY) ?
e4da7d8c 1251 ELEMENTSOF(protect_home_read_only_table) :
52b3d652 1252 ((ns_info->protect_home == PROTECT_HOME_TMPFS) ?
e4da7d8c 1253 ELEMENTSOF(protect_home_tmpfs_table) : 0)));
b6c432ca 1254
2652c6c1
DH		 1255 return !!tmp_dir + !!var_tmp_dir +
1256 strv_length(read_write_paths) +
1257 strv_length(read_only_paths) +
1258 strv_length(inaccessible_paths) +
6c47cd7d 1259 strv_length(empty_directories) +
d2d6c096 1260 n_bind_mounts +
b3d13314 1261 n_mount_images +
2abd4e38 1262 n_temporary_filesystems +
c575770b
DH		 1263 ns_info->private_dev +
1264 (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) +
c575770b 1265 (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) +
94a7b275
KK		 1266 (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) +
1267 (ns_info->protect_control_groups ? 1 : 0) +
5d997827 1268 protect_home_cnt + protect_system_cnt +
aecd5ac6 1269 (ns_info->protect_hostname ? 2 : 0) +
91dd5f7c
LP		 1270 (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) +
1271 !!log_namespace;
2652c6c1
DH		 1272}
1273
da6053d0 1274static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) {
9b68367b 1275 assert(root_directory);
f8b64b57
LP		 1276 assert(n_mounts);
1277 assert(mounts || *n_mounts == 0);
1278
93bab288 1279 typesafe_qsort(mounts, *n_mounts, mount_path_compare);
f8b64b57
LP		 1280
1281 drop_duplicates(mounts, n_mounts);
1282 drop_outside_root(root_directory, mounts, n_mounts);
1283 drop_inaccessible(mounts, n_mounts);
1284 drop_nop(mounts, n_mounts);
1285}
1286
c8c535d5
LP		 1287static bool root_read_only(
1288 char **read_only_paths,
1289 ProtectSystem protect_system) {
1290
1291 /* Determine whether the root directory is going to be read-only given the configured settings. */
1292
1293 if (protect_system == PROTECT_SYSTEM_STRICT)
1294 return true;
1295
de46b2be 1296 if (prefixed_path_strv_contains(read_only_paths, "/"))
c8c535d5
LP		 1297 return true;
1298
1299 return false;
1300}
1301
1302static bool home_read_only(
1303 char** read_only_paths,
1304 char** inaccessible_paths,
1305 char** empty_directories,
1306 const BindMount *bind_mounts,
1307 size_t n_bind_mounts,
1308 const TemporaryFileSystem *temporary_filesystems,
1309 size_t n_temporary_filesystems,
1310 ProtectHome protect_home) {
1311
1312 size_t i;
1313
1314 /* Determine whether the /home directory is going to be read-only given the configured settings. Yes,
1315 * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple
1316 * settings. */
1317
1318 if (protect_home != PROTECT_HOME_NO)
1319 return true;
1320
de46b2be
TM		 1321 if (prefixed_path_strv_contains(read_only_paths, "/home") ||
1322 prefixed_path_strv_contains(inaccessible_paths, "/home") ||
1323 prefixed_path_strv_contains(empty_directories, "/home"))
c8c535d5
LP		 1324 return true;
1325
1326 for (i = 0; i < n_temporary_filesystems; i++)
1327 if (path_equal(temporary_filesystems[i].path, "/home"))
1328 return true;
1329
1330 /* If /home is overmounted with some dir from the host it's not writable. */
1331 for (i = 0; i < n_bind_mounts; i++)
1332 if (path_equal(bind_mounts[i].destination, "/home"))
1333 return true;
1334
1335 return false;
1336}
1337
613b411c 1338int setup_namespace(
ee818b89 1339 const char* root_directory,
915e6d16 1340 const char* root_image,
18d73705 1341 const MountOptions *root_image_options,
bb0ff3fb 1342 const NamespaceInfo *ns_info,
2a624c36
AP		 1343 char** read_write_paths,
1344 char** read_only_paths,
1345 char** inaccessible_paths,
6c47cd7d 1346 char** empty_directories,
d2d6c096 1347 const BindMount *bind_mounts,
da6053d0 1348 size_t n_bind_mounts,
2abd4e38 1349 const TemporaryFileSystem *temporary_filesystems,
da6053d0 1350 size_t n_temporary_filesystems,
b3d13314
LB		 1351 const MountImage *mount_images,
1352 size_t n_mount_images,
a004cb4c
LP		 1353 const char* tmp_dir,
1354 const char* var_tmp_dir,
91dd5f7c 1355 const char *log_namespace,
915e6d16 1356 unsigned long mount_flags,
0389f4fa
LB		 1357 const void *root_hash,
1358 size_t root_hash_size,
1359 const char *root_hash_path,
d4d55b0d
LB		 1360 const void *root_hash_sig,
1361 size_t root_hash_sig_size,
1362 const char *root_hash_sig_path,
0389f4fa 1363 const char *root_verity,
7cc5ef5f
ZJS		 1364 DissectImageFlags dissect_image_flags,
1365 char **error_path) {
15ae422b 1366
915e6d16 1367 _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
78ebe980 1368 _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
915e6d16 1369 _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
0389f4fa 1370 _cleanup_free_ void *root_hash_decoded = NULL;
c2923fdc 1371 _cleanup_free_ char *verity_data = NULL, *hash_sig_path = NULL;
5f7a690a 1372 MountEntry *m = NULL, *mounts = NULL;
0389f4fa 1373 size_t n_mounts;
d18aff04 1374 bool require_prefix = false;
9b68367b 1375 const char *root;
c17ec25e 1376 int r = 0;
15ae422b 1377
915e6d16
LP		 1378 assert(ns_info);
1379
613b411c 1380 if (mount_flags == 0)
c17ec25e 1381 mount_flags = MS_SHARED;
ac0930c8 1382
915e6d16
LP		 1383 if (root_image) {
1384 dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT;
1385
c8c535d5
LP		 1386 /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */
1387 if (root_read_only(read_only_paths,
52b3d652 1388 ns_info->protect_system) &&
c8c535d5
LP		 1389 home_read_only(read_only_paths, inaccessible_paths, empty_directories,
1390 bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems,
52b3d652 1391 ns_info->protect_home) &&
c9ef8573 1392 strv_isempty(read_write_paths))
915e6d16
LP		 1393 dissect_image_flags |= DISSECT_IMAGE_READ_ONLY;
1394
1395 r = loop_device_make_by_path(root_image,
b0a94268 1396 FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */,
e08f94ac 1397 LO_FLAGS_PARTSCAN,
915e6d16
LP		 1398 &loop_device);
1399 if (r < 0)
763a260a 1400 return log_debug_errno(r, "Failed to create loop device for root image: %m");
915e6d16 1401
cbc056c8
ZJS		 1402 r = verity_metadata_load(root_image,
1403 root_hash_path,
1404 root_hash ? NULL : &root_hash_decoded,
1405 root_hash ? NULL : &root_hash_size,
1406 root_verity ? NULL : &verity_data,
1407 root_hash_sig || root_hash_sig_path ? NULL : &hash_sig_path);
78ebe980 1408 if (r < 0)
763a260a 1409 return log_debug_errno(r, "Failed to load root hash: %m");
0389f4fa 1410 dissect_image_flags |= root_verity || verity_data ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0;
78ebe980 1411
cbc056c8
ZJS		 1412 r = dissect_image(loop_device->fd,
1413 root_hash ?: root_hash_decoded,
1414 root_hash_size,
1415 root_verity ?: verity_data,
18d73705 1416 root_image_options,
cbc056c8
ZJS		 1417 dissect_image_flags,
1418 &dissected_image);
78ebe980 1419 if (r < 0)
763a260a 1420 return log_debug_errno(r, "Failed to dissect image: %m");
78ebe980 1421
cbc056c8
ZJS		 1422 r = dissected_image_decrypt(dissected_image,
1423 NULL,
1424 root_hash ?: root_hash_decoded,
1425 root_hash_size,
1426 root_verity ?: verity_data,
1427 root_hash_sig_path ?: hash_sig_path,
1428 root_hash_sig,
1429 root_hash_sig_size,
1430 dissect_image_flags,
1431 &decrypted_image);
915e6d16 1432 if (r < 0)
763a260a 1433 return log_debug_errno(r, "Failed to decrypt dissected image: %m");
915e6d16
LP		 1434 }
1435
e908468b
LP		 1436 if (root_directory)
1437 root = root_directory;
0722b359
JS		 1438 else {
1439 /* Always create the mount namespace in a temporary directory, instead of operating
1440 * directly in the root. The temporary directory prevents any mounts from being
1441 * potentially obscured my other mounts we already applied.
1442 * We use the same mount point for all images, which is safe, since they all live
1443 * in their own namespaces after all, and hence won't see each other. */
e908468b
LP		 1444
1445 root = "/run/systemd/unit-root";
1446 (void) mkdir_label(root, 0700);
d18aff04 1447 require_prefix = true;
0722b359 1448 }
e908468b 1449
cfbeb4ef
LP		 1450 n_mounts = namespace_calculate_mounts(
1451 ns_info,
1452 read_write_paths,
1453 read_only_paths,
1454 inaccessible_paths,
6c47cd7d 1455 empty_directories,
f5c52a77 1456 n_bind_mounts,
2abd4e38 1457 n_temporary_filesystems,
b3d13314 1458 n_mount_images,
cfbeb4ef 1459 tmp_dir, var_tmp_dir,
52b3d652 1460 log_namespace);
613b411c 1461
f0a4feb0 1462 if (n_mounts > 0) {
5f7a690a
LP		 1463 m = mounts = new0(MountEntry, n_mounts);
1464 if (!mounts)
1465 return -ENOMEM;
1466
d18aff04 1467 r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix);
613b411c 1468 if (r < 0)
f0a4feb0 1469 goto finish;
613b411c 1470
d18aff04 1471 r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix);
613b411c 1472 if (r < 0)
f0a4feb0 1473 goto finish;
613b411c 1474
d18aff04 1475 r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix);
613b411c 1476 if (r < 0)
f0a4feb0 1477 goto finish;
7ff7394d 1478
6c47cd7d
LP		 1479 r = append_empty_dir_mounts(&m, empty_directories);
1480 if (r < 0)
1481 goto finish;
1482
d2d6c096
LP		 1483 r = append_bind_mounts(&m, bind_mounts, n_bind_mounts);
1484 if (r < 0)
1485 goto finish;
1486
2abd4e38
YW		 1487 r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems);
1488 if (r < 0)
1489 goto finish;
1490
613b411c 1491 if (tmp_dir) {
56a13a49
ZJS		 1492 bool ro = streq(tmp_dir, RUN_SYSTEMD_EMPTY);
1493
34de407a 1494 *(m++) = (MountEntry) {
5327c910 1495 .path_const = "/tmp",
56a13a49 1496 .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP,
89bd586c 1497 .source_const = tmp_dir,
5327c910 1498 };
613b411c 1499 }
7ff7394d 1500
613b411c 1501 if (var_tmp_dir) {
56a13a49
ZJS		 1502 bool ro = streq(var_tmp_dir, RUN_SYSTEMD_EMPTY);
1503
34de407a 1504 *(m++) = (MountEntry) {
5327c910 1505 .path_const = "/var/tmp",
56a13a49 1506 .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP,
89bd586c 1507 .source_const = var_tmp_dir,
5327c910 1508 };
7ff7394d 1509 }
ac0930c8 1510
b3d13314
LB		 1511 r = append_mount_images(&m, mount_images, n_mount_images);
1512 if (r < 0)
1513 goto finish;
1514
c575770b 1515 if (ns_info->private_dev) {
34de407a 1516 *(m++) = (MountEntry) {
5327c910
LP		 1517 .path_const = "/dev",
1518 .mode = PRIVATE_DEV,
9ce4e4b0 1519 .flags = DEV_MOUNT_OPTIONS,
5327c910 1520 };
7f112f50
LP		 1521 }
1522
c575770b 1523 if (ns_info->protect_kernel_tunables) {
cbc056c8
ZJS		 1524 r = append_static_mounts(&m,
1525 protect_kernel_tunables_table,
1526 ELEMENTSOF(protect_kernel_tunables_table),
1527 ns_info->ignore_protect_paths);
c575770b 1528 if (r < 0)
f0a4feb0 1529 goto finish;
c575770b
DH		 1530 }
1531
1532 if (ns_info->protect_kernel_modules) {
cbc056c8
ZJS		 1533 r = append_static_mounts(&m,
1534 protect_kernel_modules_table,
1535 ELEMENTSOF(protect_kernel_modules_table),
1536 ns_info->ignore_protect_paths);
c575770b 1537 if (r < 0)
f0a4feb0 1538 goto finish;
c575770b 1539 }
59eeb84b 1540
94a7b275 1541 if (ns_info->protect_kernel_logs) {
cbc056c8
ZJS		 1542 r = append_static_mounts(&m,
1543 protect_kernel_logs_table,
1544 ELEMENTSOF(protect_kernel_logs_table),
1545 ns_info->ignore_protect_paths);
94a7b275
KK		 1546 if (r < 0)
1547 goto finish;
1548 }
1549
c575770b 1550 if (ns_info->protect_control_groups) {
34de407a 1551 *(m++) = (MountEntry) {
5327c910
LP		 1552 .path_const = "/sys/fs/cgroup",
1553 .mode = READONLY,
1554 };
59eeb84b
LP		 1555 }
1556
52b3d652 1557 r = append_protect_home(&m, ns_info->protect_home, ns_info->ignore_protect_paths);
b6c432ca 1558 if (r < 0)
f0a4feb0 1559 goto finish;
417116f2 1560
52b3d652 1561 r = append_protect_system(&m, ns_info->protect_system, false);
f471b2af 1562 if (r < 0)
f0a4feb0 1563 goto finish;
417116f2 1564
9b68367b 1565 if (namespace_info_mount_apivfs(ns_info)) {
cbc056c8
ZJS		 1566 r = append_static_mounts(&m,
1567 apivfs_table,
1568 ELEMENTSOF(apivfs_table),
1569 ns_info->ignore_protect_paths);
5d997827
LP		 1570 if (r < 0)
1571 goto finish;
1572 }
1573
aecd5ac6
TM		 1574 if (ns_info->protect_hostname) {
1575 *(m++) = (MountEntry) {
1576 .path_const = "/proc/sys/kernel/hostname",
1577 .mode = READONLY,
1578 };
1579 *(m++) = (MountEntry) {
1580 .path_const = "/proc/sys/kernel/domainname",
1581 .mode = READONLY,
1582 };
1583 }
1584
91dd5f7c
LP		 1585 if (log_namespace) {
1586 _cleanup_free_ char *q;
1587
1588 q = strjoin("/run/systemd/journal.", log_namespace);
1589 if (!q) {
1590 r = -ENOMEM;
1591 goto finish;
1592 }
1593
1594 *(m++) = (MountEntry) {
1595 .path_const = "/run/systemd/journal",
1596 .mode = BIND_MOUNT_RECURSIVE,
1597 .read_only = true,
1598 .source_malloc = TAKE_PTR(q),
1599 };
1600 }
1601
f0a4feb0 1602 assert(mounts + n_mounts == m);
ac0930c8 1603
5327c910 1604 /* Prepend the root directory where that's necessary */
e908468b 1605 r = prefix_where_needed(mounts, n_mounts, root);
5327c910
LP		 1606 if (r < 0)
1607 goto finish;
1608
839f1877 1609 normalize_mounts(root, mounts, &n_mounts);
15ae422b
LP		 1610 }
1611
1beab8b0
LP		 1612 /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */
1613
d944dc95 1614 if (unshare(CLONE_NEWNS) < 0) {
763a260a 1615 r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m");
1beab8b0 1616 if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS))
cbc056c8
ZJS		 1617 /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter
1618 * in place that doesn't allow us to create namespaces (or a missing cap), then
1619 * propagate a recognizable error back, which the caller can use to detect this case
1620 * (and only this) and optionally continue without namespacing applied. */
1beab8b0
LP		 1621 r = -ENOANO;
1622
d944dc95
LP		 1623 goto finish;
1624 }
1e4e94c8 1625
9b68367b
YW		 1626 /* Remount / as SLAVE so that nothing now mounted in the namespace
1627 * shows up in the parent */
1628 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
763a260a 1629 r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m");
9b68367b 1630 goto finish;
ee818b89
AC		 1631 }
1632
915e6d16 1633 if (root_image) {
e908468b 1634 /* A root image is specified, mount it to the right place */
2d3a5a73 1635 r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags);
763a260a
YW		 1636 if (r < 0) {
1637 log_debug_errno(r, "Failed to mount root image: %m");
915e6d16 1638 goto finish;
763a260a 1639 }
915e6d16 1640
07ce7407
TM		 1641 if (decrypted_image) {
1642 r = decrypted_image_relinquish(decrypted_image);
763a260a
YW		 1643 if (r < 0) {
1644 log_debug_errno(r, "Failed to relinquish decrypted image: %m");
07ce7407 1645 goto finish;
763a260a 1646 }
07ce7407 1647 }
78ebe980 1648
915e6d16
LP		 1649 loop_device_relinquish(loop_device);
1650
1651 } else if (root_directory) {
1652
e908468b
LP		 1653 /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */
1654 r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW);
763a260a
YW		 1655 if (r < 0) {
1656 log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root);
d944dc95 1657 goto finish;
763a260a 1658 }
8f1ad200 1659 if (r == 0) {
e908468b 1660 if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) {
763a260a 1661 r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root);
8f1ad200
LP		 1662 goto finish;
1663 }
d944dc95 1664 }
e908468b 1665
9b68367b 1666 } else {
e908468b
LP		 1667
1668 /* Let's mount the main root directory to the root directory to use */
1669 if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) {
763a260a 1670 r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root);
e908468b
LP		 1671 goto finish;
1672 }
ee818b89 1673 }
c2c13f2d 1674
4e0c20de
LP		 1675 /* Try to set up the new root directory before mounting anything else there. */
1676 if (root_image || root_directory)
1677 (void) base_filesystem_create(root, UID_INVALID, GID_INVALID);
1678
f0a4feb0 1679 if (n_mounts > 0) {
ac9de0b3 1680 _cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
6b000af4 1681 _cleanup_free_ char **deny_list = NULL;
da6053d0 1682 size_t j;
6b7c9f8b 1683
cbc056c8
ZJS		 1684 /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of
1685 * /proc. For example, this is the case with the option: 'InaccessiblePaths=/proc'. */
ac9de0b3
TR		 1686 proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
1687 if (!proc_self_mountinfo) {
763a260a 1688 r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m");
7cc5ef5f
ZJS		 1689 if (error_path)
1690 *error_path = strdup("/proc/self/mountinfo");
ac9de0b3
TR		 1691 goto finish;
1692 }
1693
088696fe
LP		 1694 /* First round, establish all mounts we need */
1695 for (;;) {
1696 bool again = false;
1697
1698 for (m = mounts; m < mounts + n_mounts; ++m) {
1699
1700 if (m->applied)
1701 continue;
1702
1703 r = follow_symlink(root, m);
7cc5ef5f
ZJS		 1704 if (r < 0) {
1705 if (error_path && mount_entry_path(m))
1706 *error_path = strdup(mount_entry_path(m));
088696fe 1707 goto finish;
7cc5ef5f 1708 }
088696fe 1709 if (r == 0) {
cbc056c8
ZJS		 1710 /* We hit a symlinked mount point. The entry got rewritten and might
1711 * point to a very different place now. Let's normalize the changed
1712 * list, and start from the beginning. After all to mount the entry
1713 * at the new location we might need some other mounts first */
088696fe
LP		 1714 again = true;
1715 break;
1716 }
1717
1718 r = apply_mount(root, m);
7cc5ef5f
ZJS		 1719 if (r < 0) {
1720 if (error_path && mount_entry_path(m))
1721 *error_path = strdup(mount_entry_path(m));
088696fe 1722 goto finish;
7cc5ef5f 1723 }
088696fe
LP		 1724
1725 m->applied = true;
1726 }
1727
1728 if (!again)
1729 break;
1730
839f1877 1731 normalize_mounts(root, mounts, &n_mounts);
c2c13f2d 1732 }
15ae422b 1733
6b000af4
LP		 1734 /* Create a deny list we can pass to bind_mount_recursive() */
1735 deny_list = new(char*, n_mounts+1);
1736 if (!deny_list) {
5f7a690a
LP		 1737 r = -ENOMEM;
1738 goto finish;
1739 }
f0a4feb0 1740 for (j = 0; j < n_mounts; j++)
6b000af4
LP		 1741 deny_list[j] = (char*) mount_entry_path(mounts+j);
1742 deny_list[j] = NULL;
6b7c9f8b
LP		 1743
1744 /* Second round, flip the ro bits if necessary. */
f0a4feb0 1745 for (m = mounts; m < mounts + n_mounts; ++m) {
6b000af4 1746 r = make_read_only(m, deny_list, proc_self_mountinfo);
7cc5ef5f
ZJS		 1747 if (r < 0) {
1748 if (error_path && mount_entry_path(m))
1749 *error_path = strdup(mount_entry_path(m));
d944dc95 1750 goto finish;
7cc5ef5f 1751 }
c2c13f2d 1752 }
15ae422b
LP		 1753 }
1754
9b68367b
YW		 1755 /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
1756 r = mount_move_root(root);
763a260a
YW		 1757 if (r < 0) {
1758 log_debug_errno(r, "Failed to mount root with MS_MOVE: %m");
9b68367b 1759 goto finish;
763a260a 1760 }
ee818b89 1761
55fe7432 1762 /* Remount / as the desired mode. Note that this will not
c2c13f2d
LP		 1763 * reestablish propagation from our side to the host, since
1764 * what's disconnected is disconnected. */
d944dc95 1765 if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
763a260a 1766 r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m");
d944dc95
LP		 1767 goto finish;
1768 }
15ae422b 1769
d944dc95 1770 r = 0;
15ae422b 1771
d944dc95 1772finish:
0cd41757
LP		 1773 if (n_mounts > 0)
1774 for (m = mounts; m < mounts + n_mounts; m++)
1775 mount_entry_done(m);
613b411c 1776
5f7a690a
LP		 1777 free(mounts);
1778
613b411c
LP		 1779 return r;
1780}
1781
da6053d0
LP		 1782void bind_mount_free_many(BindMount *b, size_t n) {
1783 size_t i;
d2d6c096
LP		 1784
1785 assert(b || n == 0);
1786
1787 for (i = 0; i < n; i++) {
1788 free(b[i].source);
1789 free(b[i].destination);
1790 }
1791
1792 free(b);
1793}
1794
da6053d0 1795int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) {
d2d6c096
LP		 1796 _cleanup_free_ char *s = NULL, *d = NULL;
1797 BindMount *c;
1798
1799 assert(b);
1800 assert(n);
1801 assert(item);
1802
1803 s = strdup(item->source);
1804 if (!s)
1805 return -ENOMEM;
1806
1807 d = strdup(item->destination);
1808 if (!d)
1809 return -ENOMEM;
1810
aa484f35 1811 c = reallocarray(*b, *n + 1, sizeof(BindMount));
d2d6c096
LP		 1812 if (!c)
1813 return -ENOMEM;
1814
1815 *b = c;
1816
1817 c[(*n) ++] = (BindMount) {
1cc6c93a
YW		 1818 .source = TAKE_PTR(s),
1819 .destination = TAKE_PTR(d),
d2d6c096 1820 .read_only = item->read_only,
9ce4e4b0 1821 .nosuid = item->nosuid,
d2d6c096
LP		 1822 .recursive = item->recursive,
1823 .ignore_enoent = item->ignore_enoent,
1824 };
1825
d2d6c096
LP		 1826 return 0;
1827}
1828
b3d13314
LB		 1829MountImage* mount_image_free_many(MountImage *m, size_t *n) {
1830 size_t i;
1831
1832 assert(n);
1833 assert(m || *n == 0);
1834
1835 for (i = 0; i < *n; i++) {
1836 free(m[i].source);
1837 free(m[i].destination);
427353f6 1838 mount_options_free_all(m[i].mount_options);
b3d13314
LB		 1839 }
1840
1841 free(m);
1842 *n = 0;
1843 return NULL;
1844}
1845
1846int mount_image_add(MountImage **m, size_t *n, const MountImage *item) {
1847 _cleanup_free_ char *s = NULL, *d = NULL;
427353f6
LB		 1848 _cleanup_(mount_options_free_allp) MountOptions *options = NULL;
1849 MountOptions *i;
b3d13314
LB		 1850 MountImage *c;
1851
1852 assert(m);
1853 assert(n);
1854 assert(item);
1855
1856 s = strdup(item->source);
1857 if (!s)
1858 return -ENOMEM;
1859
1860 d = strdup(item->destination);
1861 if (!d)
1862 return -ENOMEM;
1863
427353f6
LB		 1864 LIST_FOREACH(mount_options, i, item->mount_options) {
1865 _cleanup_(mount_options_free_allp) MountOptions *o;
1866
1867 o = new(MountOptions, 1);
1868 if (!o)
1869 return -ENOMEM;
1870
1871 *o = (MountOptions) {
1872 .partition_designator = i->partition_designator,
1873 .options = strdup(i->options),
1874 };
1875 if (!o->options)
1876 return -ENOMEM;
1877
1878 LIST_APPEND(mount_options, options, TAKE_PTR(o));
1879 }
1880
b3d13314
LB		 1881 c = reallocarray(*m, *n + 1, sizeof(MountImage));
1882 if (!c)
1883 return -ENOMEM;
1884
1885 *m = c;
1886
1887 c[(*n) ++] = (MountImage) {
1888 .source = TAKE_PTR(s),
1889 .destination = TAKE_PTR(d),
427353f6 1890 .mount_options = TAKE_PTR(options),
b3d13314
LB		 1891 .ignore_enoent = item->ignore_enoent,
1892 };
1893
1894 return 0;
1895}
1896
da6053d0
LP		 1897void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) {
1898 size_t i;
2abd4e38
YW		 1899
1900 assert(t || n == 0);
1901
1902 for (i = 0; i < n; i++) {
1903 free(t[i].path);
1904 free(t[i].options);
1905 }
1906
1907 free(t);
1908}
1909
1910int temporary_filesystem_add(
1911 TemporaryFileSystem **t,
da6053d0 1912 size_t *n,
2abd4e38
YW		 1913 const char *path,
1914 const char *options) {
1915
1916 _cleanup_free_ char *p = NULL, *o = NULL;
1917 TemporaryFileSystem *c;
1918
1919 assert(t);
1920 assert(n);
1921 assert(path);
1922
1923 p = strdup(path);
1924 if (!p)
1925 return -ENOMEM;
1926
1927 if (!isempty(options)) {
1928 o = strdup(options);
1929 if (!o)
1930 return -ENOMEM;
1931 }
1932
aa484f35 1933 c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem));
2abd4e38
YW		 1934 if (!c)
1935 return -ENOMEM;
1936
1937 *t = c;
1938
1939 c[(*n) ++] = (TemporaryFileSystem) {
1cc6c93a
YW		 1940 .path = TAKE_PTR(p),
1941 .options = TAKE_PTR(o),
2abd4e38
YW		 1942 };
1943
2abd4e38
YW		 1944 return 0;
1945}
1946
a652f050
JR		 1947static int make_tmp_prefix(const char *prefix) {
1948 _cleanup_free_ char *t = NULL;
1949 int r;
1950
1951 /* Don't do anything unless we know the dir is actually missing */
1952 r = access(prefix, F_OK);
1953 if (r >= 0)
1954 return 0;
1955 if (errno != ENOENT)
1956 return -errno;
1957
1958 r = mkdir_parents(prefix, 0755);
1959 if (r < 0)
1960 return r;
1961
1962 r = tempfn_random(prefix, NULL, &t);
1963 if (r < 0)
1964 return r;
1965
1966 if (mkdir(t, 0777) < 0)
1967 return -errno;
1968
1969 if (chmod(t, 01777) < 0) {
1970 r = -errno;
1971 (void) rmdir(t);
1972 return r;
1973 }
1974
1975 if (rename(t, prefix) < 0) {
1976 r = -errno;
1977 (void) rmdir(t);
1978 return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */
1979 }
1980
1981 return 0;
1982
1983}
1984
56a13a49
ZJS		 1985static int make_tmp_subdir(const char *parent, char **ret) {
1986 _cleanup_free_ char *y = NULL;
1987
3f181262
LP		 1988 y = path_join(parent, "/tmp");
1989 if (!y)
1990 return -ENOMEM;
56a13a49 1991
3f181262 1992 RUN_WITH_UMASK(0000) {
56a13a49
ZJS		 1993 if (mkdir(y, 0777 | S_ISVTX) < 0)
1994 return -errno;
1995 }
1996
1997 if (ret)
1998 *ret = TAKE_PTR(y);
1999 return 0;
2000}
2001
2002static int setup_one_tmp_dir(const char *id, const char *prefix, char **path, char **tmp_path) {
613b411c 2003 _cleanup_free_ char *x = NULL;
6b46ea73
LP		 2004 char bid[SD_ID128_STRING_MAX];
2005 sd_id128_t boot_id;
56a13a49 2006 bool rw = true;
6b46ea73 2007 int r;
613b411c
LP		 2008
2009 assert(id);
2010 assert(prefix);
2011 assert(path);
2012
6b46ea73
LP		 2013 /* We include the boot id in the directory so that after a
2014 * reboot we can easily identify obsolete directories. */
2015
2016 r = sd_id128_get_boot(&boot_id);
2017 if (r < 0)
2018 return r;
2019
605405c6 2020 x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX");
613b411c
LP		 2021 if (!x)
2022 return -ENOMEM;
2023
a652f050
JR		 2024 r = make_tmp_prefix(prefix);
2025 if (r < 0)
2026 return r;
2027
613b411c 2028 RUN_WITH_UMASK(0077)
56a13a49
ZJS		 2029 if (!mkdtemp(x)) {
2030 if (errno == EROFS || ERRNO_IS_DISK_SPACE(errno))
2031 rw = false;
2032 else
2033 return -errno;
2034 }
613b411c 2035
56a13a49
ZJS		 2036 if (rw) {
2037 r = make_tmp_subdir(x, tmp_path);
2038 if (r < 0)
2039 return r;
2040 } else {
2041 /* Trouble: we failed to create the directory. Instead of failing, let's simulate /tmp being
2042 * read-only. This way the service will get the EROFS result as if it was writing to the real
2043 * file system. */
2044 r = mkdir_p(RUN_SYSTEMD_EMPTY, 0500);
2045 if (r < 0)
2046 return r;
613b411c 2047
3f181262
LP		 2048 r = free_and_strdup(&x, RUN_SYSTEMD_EMPTY);
2049 if (r < 0)
2050 return r;
c17ec25e 2051 }
15ae422b 2052
1cc6c93a 2053 *path = TAKE_PTR(x);
613b411c
LP		 2054 return 0;
2055}
2056
2057int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) {
56a13a49
ZJS		 2058 _cleanup_(namespace_cleanup_tmpdirp) char *a = NULL;
2059 _cleanup_(rmdir_and_freep) char *a_tmp = NULL;
2060 char *b;
613b411c
LP		 2061 int r;
2062
2063 assert(id);
2064 assert(tmp_dir);
2065 assert(var_tmp_dir);
2066
56a13a49 2067 r = setup_one_tmp_dir(id, "/tmp", &a, &a_tmp);
613b411c
LP		 2068 if (r < 0)
2069 return r;
2070
56a13a49
ZJS		 2071 r = setup_one_tmp_dir(id, "/var/tmp", &b, NULL);
2072 if (r < 0)
613b411c 2073 return r;
613b411c 2074
56a13a49
ZJS		 2075 a_tmp = mfree(a_tmp); /* avoid rmdir */
2076 *tmp_dir = TAKE_PTR(a);
2077 *var_tmp_dir = TAKE_PTR(b);
613b411c
LP		 2078
2079 return 0;
2080}
2081
2caa38e9 2082int setup_netns(const int netns_storage_socket[static 2]) {
613b411c 2083 _cleanup_close_ int netns = -1;
3ee897d6 2084 int r, q;
613b411c
LP		 2085
2086 assert(netns_storage_socket);
2087 assert(netns_storage_socket[0] >= 0);
2088 assert(netns_storage_socket[1] >= 0);
2089
2090 /* We use the passed socketpair as a storage buffer for our
76cd584b
LP		 2091 * namespace reference fd. Whatever process runs this first
2092 * shall create a new namespace, all others should just join
2093 * it. To serialize that we use a file lock on the socket
2094 * pair.
613b411c
LP		 2095 *
2096 * It's a bit crazy, but hey, works great! */
2097
2098 if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
2099 return -errno;
2100
3ee897d6
LP		 2101 netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
2102 if (netns == -EAGAIN) {
44ffcbae 2103 /* Nothing stored yet, so let's create a new namespace. */
613b411c
LP		 2104
2105 if (unshare(CLONE_NEWNET) < 0) {
2106 r = -errno;
2107 goto fail;
2108 }
2109
44ffcbae 2110 (void) loopback_setup();
613b411c
LP		 2111
2112 netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY);
2113 if (netns < 0) {
2114 r = -errno;
2115 goto fail;
2116 }
2117
2118 r = 1;
613b411c 2119
3ee897d6
LP		 2120 } else if (netns < 0) {
2121 r = netns;
2122 goto fail;
613b411c 2123
3ee897d6
LP		 2124 } else {
2125 /* Yay, found something, so let's join the namespace */
613b411c
LP		 2126 if (setns(netns, CLONE_NEWNET) < 0) {
2127 r = -errno;
2128 goto fail;
2129 }
2130
2131 r = 0;
2132 }
2133
3ee897d6
LP		 2134 q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
2135 if (q < 0) {
2136 r = q;
613b411c
LP		 2137 goto fail;
2138 }
2139
2140fail:
fe048ce5 2141 (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
15ae422b
LP		 2142 return r;
2143}
417116f2 2144
2caa38e9 2145int open_netns_path(const int netns_storage_socket[static 2], const char *path) {
51af7fb2
LP		 2146 _cleanup_close_ int netns = -1;
2147 int q, r;
2148
2149 assert(netns_storage_socket);
2150 assert(netns_storage_socket[0] >= 0);
2151 assert(netns_storage_socket[1] >= 0);
2152 assert(path);
2153
2154 /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in
2155 * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a
2156 * new anonymous netns if needed. */
2157
2158 if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0)
2159 return -errno;
2160
2161 netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT);
2162 if (netns == -EAGAIN) {
2163 /* Nothing stored yet. Open the file from the file system. */
2164
2165 netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC);
2166 if (netns < 0) {
2167 r = -errno;
2168 goto fail;
2169 }
2170
2171 r = fd_is_network_ns(netns);
2172 if (r == 0) { /* Not a netns? Refuse early. */
2173 r = -EINVAL;
2174 goto fail;
2175 }
2176 if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */
2177 goto fail;
2178
2179 r = 1;
2180
2181 } else if (netns < 0) {
2182 r = netns;
2183 goto fail;
2184 } else
2185 r = 0; /* Already allocated */
2186
2187 q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT);
2188 if (q < 0) {
2189 r = q;
2190 goto fail;
2191 }
2192
2193fail:
2194 (void) lockf(netns_storage_socket[0], F_ULOCK, 0);
2195 return r;
2196}
2197
6e2d7c4f
MS		 2198bool ns_type_supported(NamespaceType type) {
2199 const char *t, *ns_proc;
2200
0fa5b831
LP		 2201 t = namespace_type_to_string(type);
2202 if (!t) /* Don't know how to translate this? Then it's not supported */
6e2d7c4f
MS		 2203 return false;
2204
6e2d7c4f 2205 ns_proc = strjoina("/proc/self/ns/", t);
6e2d7c4f
MS		 2206 return access(ns_proc, F_OK) == 0;
2207}
2208
1b8689f9 2209static const char *const protect_home_table[_PROTECT_HOME_MAX] = {
cbc056c8
ZJS		 2210 [PROTECT_HOME_NO] = "no",
2211 [PROTECT_HOME_YES] = "yes",
1b8689f9 2212 [PROTECT_HOME_READ_ONLY] = "read-only",
cbc056c8 2213 [PROTECT_HOME_TMPFS] = "tmpfs",
417116f2
LP		 2214};
2215
1e8c7bd5 2216DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES);
5e1c6154 2217
1b8689f9 2218static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = {
cbc056c8
ZJS		 2219 [PROTECT_SYSTEM_NO] = "no",
2220 [PROTECT_SYSTEM_YES] = "yes",
2221 [PROTECT_SYSTEM_FULL] = "full",
3f815163 2222 [PROTECT_SYSTEM_STRICT] = "strict",
1b8689f9
LP		 2223};
2224
1e8c7bd5 2225DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES);
03c791aa 2226
6e2d7c4f 2227static const char* const namespace_type_table[] = {
cbc056c8 2228 [NAMESPACE_MOUNT] = "mnt",
6e2d7c4f 2229 [NAMESPACE_CGROUP] = "cgroup",
cbc056c8
ZJS		 2230 [NAMESPACE_UTS] = "uts",
2231 [NAMESPACE_IPC] = "ipc",
2232 [NAMESPACE_USER] = "user",
2233 [NAMESPACE_PID] = "pid",
2234 [NAMESPACE_NET] = "net",
6e2d7c4f
MS		 2235};
2236
2237DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType);
Unnamed repository; edit this file 'description' to name the repository.