]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
15ae422b LP |
2 | /*** |
3 | This file is part of systemd. | |
4 | ||
5 | Copyright 2010 Lennart Poettering | |
15ae422b LP |
6 | ***/ |
7 | ||
8 | #include <errno.h> | |
07630cea | 9 | #include <sched.h> |
15ae422b | 10 | #include <stdio.h> |
07630cea LP |
11 | #include <string.h> |
12 | #include <sys/mount.h> | |
15ae422b | 13 | #include <sys/stat.h> |
07630cea | 14 | #include <unistd.h> |
25e870b5 | 15 | #include <linux/fs.h> |
15ae422b | 16 | |
b5efdb8a | 17 | #include "alloc-util.h" |
10404d52 | 18 | #include "base-filesystem.h" |
7f112f50 | 19 | #include "dev-setup.h" |
3ffd4af2 | 20 | #include "fd-util.h" |
d944dc95 | 21 | #include "fs-util.h" |
e908468b | 22 | #include "label.h" |
915e6d16 | 23 | #include "loop-util.h" |
07630cea LP |
24 | #include "loopback-setup.h" |
25 | #include "missing.h" | |
26 | #include "mkdir.h" | |
4349cd7c | 27 | #include "mount-util.h" |
3ffd4af2 | 28 | #include "namespace.h" |
07630cea | 29 | #include "path-util.h" |
d7b8eec7 | 30 | #include "selinux-util.h" |
2583fbea | 31 | #include "socket-util.h" |
36ce7110 | 32 | #include "stat-util.h" |
8b43440b | 33 | #include "string-table.h" |
07630cea LP |
34 | #include "string-util.h" |
35 | #include "strv.h" | |
affb60b1 | 36 | #include "umask-util.h" |
ee104e11 | 37 | #include "user-util.h" |
07630cea | 38 | #include "util.h" |
15ae422b | 39 | |
737ba3c8 | 40 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
41 | ||
c17ec25e | 42 | typedef enum MountMode { |
15ae422b LP |
43 | /* This is ordered by priority! */ |
44 | INACCESSIBLE, | |
d2d6c096 LP |
45 | BIND_MOUNT, |
46 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 47 | PRIVATE_TMP, |
7f112f50 | 48 | PRIVATE_DEV, |
5d997827 | 49 | BIND_DEV, |
6c47cd7d | 50 | EMPTY_DIR, |
5d997827 LP |
51 | SYSFS, |
52 | PROCFS, | |
53 | READONLY, | |
59eeb84b | 54 | READWRITE, |
2abd4e38 | 55 | TMPFS, |
c17ec25e | 56 | } MountMode; |
15ae422b | 57 | |
34de407a | 58 | typedef 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? */ |
55fe7432 | 64 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
d2d6c096 LP |
65 | const char *source_const; /* The source path, for bind mounts */ |
66 | char *source_malloc; | |
2abd4e38 YW |
67 | const char *options_const;/* Mount options for tmpfs */ |
68 | char *options_malloc; | |
69 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
34de407a | 70 | } MountEntry; |
15ae422b | 71 | |
5d997827 LP |
72 | /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted |
73 | * something there already. These mounts are hence overriden by any other explicitly configured mounts. */ | |
74 | static const MountEntry apivfs_table[] = { | |
75 | { "/proc", PROCFS, false }, | |
76 | { "/dev", BIND_DEV, false }, | |
77 | { "/sys", SYSFS, false }, | |
78 | }; | |
f471b2af | 79 | |
11a30cec | 80 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 81 | static const MountEntry protect_kernel_tunables_table[] = { |
c6232fb0 LP |
82 | { "/proc/sys", READONLY, false }, |
83 | { "/proc/sysrq-trigger", READONLY, true }, | |
84 | { "/proc/latency_stats", READONLY, true }, | |
85 | { "/proc/mtrr", READONLY, true }, | |
aa70f38b | 86 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ |
c6232fb0 LP |
87 | { "/proc/acpi", READONLY, true }, |
88 | { "/proc/timer_stats", READONLY, true }, | |
89 | { "/proc/asound", READONLY, true }, | |
90 | { "/proc/bus", READONLY, true }, | |
91 | { "/proc/fs", READONLY, true }, | |
92 | { "/proc/irq", READONLY, true }, | |
93 | { "/sys", READONLY, false }, | |
94 | { "/sys/kernel/debug", READONLY, true }, | |
95 | { "/sys/kernel/tracing", READONLY, true }, | |
13a141f0 | 96 | { "/sys/fs/bpf", READONLY, true }, |
c6232fb0 | 97 | { "/sys/fs/cgroup", READWRITE, false }, /* READONLY is set by ProtectControlGroups= option */ |
3a0bf6d6 | 98 | { "/sys/fs/selinux", READWRITE, true }, |
11a30cec DH |
99 | }; |
100 | ||
c575770b | 101 | /* ProtectKernelModules= option */ |
34de407a | 102 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 103 | #if HAVE_SPLIT_USR |
c6232fb0 | 104 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 105 | #endif |
c6232fb0 | 106 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
107 | }; |
108 | ||
b6c432ca DH |
109 | /* |
110 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
111 | * system should be protected by ProtectSystem= | |
112 | */ | |
34de407a | 113 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
114 | { "/home", READONLY, true }, |
115 | { "/run/user", READONLY, true }, | |
116 | { "/root", READONLY, true }, | |
b6c432ca DH |
117 | }; |
118 | ||
e4da7d8c YW |
119 | /* ProtectHome=tmpfs table */ |
120 | static const MountEntry protect_home_tmpfs_table[] = { | |
121 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
122 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
123 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME }, | |
124 | }; | |
125 | ||
b6c432ca | 126 | /* ProtectHome=yes table */ |
34de407a | 127 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
128 | { "/home", INACCESSIBLE, true }, |
129 | { "/run/user", INACCESSIBLE, true }, | |
130 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
131 | }; |
132 | ||
f471b2af | 133 | /* ProtectSystem=yes table */ |
34de407a | 134 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
135 | { "/usr", READONLY, false }, |
136 | { "/boot", READONLY, true }, | |
137 | { "/efi", READONLY, true }, | |
7486f305 AB |
138 | #if HAVE_SPLIT_USR |
139 | { "/lib", READONLY, true }, | |
140 | { "/lib64", READONLY, true }, | |
141 | { "/bin", READONLY, true }, | |
671f0f8d | 142 | # if HAVE_SPLIT_BIN |
7486f305 | 143 | { "/sbin", READONLY, true }, |
671f0f8d | 144 | # endif |
7486f305 | 145 | #endif |
f471b2af DH |
146 | }; |
147 | ||
148 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 149 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
150 | { "/usr", READONLY, false }, |
151 | { "/boot", READONLY, true }, | |
152 | { "/efi", READONLY, true }, | |
153 | { "/etc", READONLY, false }, | |
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 | /* | |
165 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
166 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
167 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
168 | * protect those, and these options should be fully orthogonal. | |
169 | * (And of course /home and friends are also left writable, as ProtectHome= | |
170 | * shall manage those, orthogonally). | |
171 | */ | |
34de407a | 172 | static const MountEntry protect_system_strict_table[] = { |
ddbe0412 LP |
173 | { "/", READONLY, false }, |
174 | { "/proc", READWRITE, false }, /* ProtectKernelTunables= */ | |
175 | { "/sys", READWRITE, false }, /* ProtectKernelTunables= */ | |
176 | { "/dev", READWRITE, false }, /* PrivateDevices= */ | |
177 | { "/home", READWRITE, true }, /* ProtectHome= */ | |
178 | { "/run/user", READWRITE, true }, /* ProtectHome= */ | |
179 | { "/root", READWRITE, true }, /* ProtectHome= */ | |
f471b2af DH |
180 | }; |
181 | ||
34de407a | 182 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
183 | assert(p); |
184 | ||
5327c910 LP |
185 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
186 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 187 | |
5327c910 | 188 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
189 | } |
190 | ||
34de407a | 191 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
192 | assert(p); |
193 | ||
194 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
195 | } | |
196 | ||
d2d6c096 LP |
197 | static const char *mount_entry_source(const MountEntry *p) { |
198 | assert(p); | |
199 | ||
200 | return p->source_malloc ?: p->source_const; | |
201 | } | |
202 | ||
2abd4e38 YW |
203 | static const char *mount_entry_options(const MountEntry *p) { |
204 | assert(p); | |
205 | ||
206 | return p->options_malloc ?: p->options_const; | |
207 | } | |
208 | ||
1eb7e08e LP |
209 | static void mount_entry_done(MountEntry *p) { |
210 | assert(p); | |
211 | ||
212 | p->path_malloc = mfree(p->path_malloc); | |
213 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 214 | p->options_malloc = mfree(p->options_malloc); |
1eb7e08e LP |
215 | } |
216 | ||
d18aff04 | 217 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
218 | char **i; |
219 | ||
613b411c LP |
220 | assert(p); |
221 | ||
5327c910 LP |
222 | /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */ |
223 | ||
15ae422b | 224 | STRV_FOREACH(i, strv) { |
5327c910 LP |
225 | bool ignore = false, needs_prefix = false; |
226 | const char *e = *i; | |
15ae422b | 227 | |
5327c910 LP |
228 | /* Look for any prefixes */ |
229 | if (startswith(e, "-")) { | |
230 | e++; | |
9c94d52e | 231 | ignore = true; |
ea92ae33 | 232 | } |
5327c910 LP |
233 | if (startswith(e, "+")) { |
234 | e++; | |
235 | needs_prefix = true; | |
236 | } | |
ea92ae33 | 237 | |
5327c910 | 238 | if (!path_is_absolute(e)) |
15ae422b LP |
239 | return -EINVAL; |
240 | ||
34de407a | 241 | *((*p)++) = (MountEntry) { |
5327c910 LP |
242 | .path_const = e, |
243 | .mode = mode, | |
244 | .ignore = ignore, | |
d18aff04 | 245 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 246 | }; |
15ae422b LP |
247 | } |
248 | ||
249 | return 0; | |
250 | } | |
251 | ||
6c47cd7d LP |
252 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
253 | char **i; | |
254 | ||
255 | assert(p); | |
256 | ||
257 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
258 | * "/private/" boundary directories for DynamicUser=1. */ | |
259 | ||
260 | STRV_FOREACH(i, strv) { | |
261 | ||
262 | *((*p)++) = (MountEntry) { | |
263 | .path_const = *i, | |
264 | .mode = EMPTY_DIR, | |
265 | .ignore = false, | |
266 | .has_prefix = false, | |
267 | .read_only = true, | |
2abd4e38 YW |
268 | .options_const = "mode=755", |
269 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, | |
6c47cd7d LP |
270 | }; |
271 | } | |
272 | ||
273 | return 0; | |
274 | } | |
275 | ||
d2d6c096 LP |
276 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, unsigned n) { |
277 | unsigned i; | |
278 | ||
279 | assert(p); | |
280 | ||
281 | for (i = 0; i < n; i++) { | |
282 | const BindMount *b = binds + i; | |
283 | ||
284 | *((*p)++) = (MountEntry) { | |
285 | .path_const = b->destination, | |
286 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
287 | .read_only = b->read_only, | |
288 | .source_const = b->source, | |
4ca763a9 | 289 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
290 | }; |
291 | } | |
292 | ||
293 | return 0; | |
294 | } | |
295 | ||
2abd4e38 YW |
296 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, unsigned n) { |
297 | unsigned i; | |
298 | int r; | |
299 | ||
300 | assert(p); | |
301 | ||
302 | for (i = 0; i < n; i++) { | |
303 | const TemporaryFileSystem *t = tmpfs + i; | |
304 | _cleanup_free_ char *o = NULL, *str = NULL; | |
305 | unsigned long flags = MS_NODEV|MS_STRICTATIME; | |
306 | bool ro = false; | |
307 | ||
308 | if (!path_is_absolute(t->path)) | |
309 | return -EINVAL; | |
310 | ||
311 | if (!isempty(t->options)) { | |
312 | str = strjoin("mode=0755,", t->options); | |
313 | if (!str) | |
314 | return -ENOMEM; | |
315 | ||
316 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); | |
317 | if (r < 0) | |
318 | return r; | |
319 | ||
320 | ro = !!(flags & MS_RDONLY); | |
321 | if (ro) | |
322 | flags ^= MS_RDONLY; | |
323 | } | |
324 | ||
325 | *((*p)++) = (MountEntry) { | |
326 | .path_const = t->path, | |
327 | .mode = TMPFS, | |
328 | .read_only = ro, | |
329 | .options_malloc = o, | |
330 | .flags = flags, | |
331 | }; | |
332 | ||
333 | o = NULL; | |
334 | } | |
335 | ||
336 | return 0; | |
337 | } | |
338 | ||
34de407a | 339 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, unsigned n, bool ignore_protect) { |
f471b2af | 340 | unsigned i; |
11a30cec DH |
341 | |
342 | assert(p); | |
f471b2af | 343 | assert(mounts); |
11a30cec | 344 | |
5327c910 | 345 | /* Adds a list of static pre-defined entries */ |
f471b2af | 346 | |
5327c910 | 347 | for (i = 0; i < n; i++) |
34de407a LP |
348 | *((*p)++) = (MountEntry) { |
349 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
350 | .mode = mounts[i].mode, |
351 | .ignore = mounts[i].ignore || ignore_protect, | |
352 | }; | |
f471b2af DH |
353 | |
354 | return 0; | |
355 | } | |
356 | ||
34de407a | 357 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
358 | assert(p); |
359 | ||
5327c910 | 360 | switch (protect_home) { |
b6c432ca | 361 | |
5327c910 | 362 | case PROTECT_HOME_NO: |
b6c432ca DH |
363 | return 0; |
364 | ||
b6c432ca | 365 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
366 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
367 | ||
e4da7d8c YW |
368 | case PROTECT_HOME_TMPFS: |
369 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
370 | ||
b6c432ca | 371 | case PROTECT_HOME_YES: |
5327c910 LP |
372 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
373 | ||
b6c432ca | 374 | default: |
5327c910 | 375 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 376 | } |
b6c432ca DH |
377 | } |
378 | ||
34de407a | 379 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
380 | assert(p); |
381 | ||
5327c910 LP |
382 | switch (protect_system) { |
383 | ||
384 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
385 | return 0; |
386 | ||
f471b2af | 387 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
388 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
389 | ||
f471b2af | 390 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
391 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
392 | ||
f471b2af | 393 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
394 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
395 | ||
f471b2af | 396 | default: |
5327c910 | 397 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 398 | } |
11a30cec DH |
399 | } |
400 | ||
c17ec25e | 401 | static int mount_path_compare(const void *a, const void *b) { |
34de407a | 402 | const MountEntry *p = a, *q = b; |
a0827e2b | 403 | int d; |
15ae422b | 404 | |
6ee1a919 | 405 | /* If the paths are not equal, then order prefixes first */ |
34de407a | 406 | d = path_compare(mount_entry_path(p), mount_entry_path(q)); |
6ee1a919 LP |
407 | if (d != 0) |
408 | return d; | |
15ae422b | 409 | |
6ee1a919 LP |
410 | /* If the paths are equal, check the mode */ |
411 | if (p->mode < q->mode) | |
412 | return -1; | |
15ae422b | 413 | |
6ee1a919 LP |
414 | if (p->mode > q->mode) |
415 | return 1; | |
15ae422b | 416 | |
6ee1a919 | 417 | return 0; |
15ae422b LP |
418 | } |
419 | ||
34de407a | 420 | static int prefix_where_needed(MountEntry *m, unsigned n, const char *root_directory) { |
5327c910 LP |
421 | unsigned i; |
422 | ||
423 | /* Prefixes all paths in the bind mount table with the root directory if it is specified and the entry needs | |
424 | * that. */ | |
425 | ||
426 | if (!root_directory) | |
427 | return 0; | |
428 | ||
429 | for (i = 0; i < n; i++) { | |
430 | char *s; | |
431 | ||
432 | if (m[i].has_prefix) | |
433 | continue; | |
434 | ||
34de407a | 435 | s = prefix_root(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
436 | if (!s) |
437 | return -ENOMEM; | |
438 | ||
e282f51f | 439 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
440 | m[i].has_prefix = true; |
441 | } | |
442 | ||
443 | return 0; | |
444 | } | |
445 | ||
34de407a LP |
446 | static void drop_duplicates(MountEntry *m, unsigned *n) { |
447 | MountEntry *f, *t, *previous; | |
15ae422b | 448 | |
c17ec25e | 449 | assert(m); |
15ae422b | 450 | assert(n); |
15ae422b | 451 | |
fe3c2583 LP |
452 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
453 | ||
1d54cd5d | 454 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 455 | |
fe3c2583 LP |
456 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
457 | * above. */ | |
34de407a LP |
458 | if (previous && path_equal(mount_entry_path(f), mount_entry_path(previous))) { |
459 | log_debug("%s is duplicate.", mount_entry_path(f)); | |
460 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ | |
1eb7e08e | 461 | mount_entry_done(f); |
15ae422b | 462 | continue; |
fe3c2583 | 463 | } |
15ae422b | 464 | |
e2d7c1a0 | 465 | *t = *f; |
15ae422b | 466 | previous = t; |
fe3c2583 LP |
467 | t++; |
468 | } | |
469 | ||
470 | *n = t - m; | |
471 | } | |
472 | ||
34de407a LP |
473 | static void drop_inaccessible(MountEntry *m, unsigned *n) { |
474 | MountEntry *f, *t; | |
fe3c2583 LP |
475 | const char *clear = NULL; |
476 | ||
477 | assert(m); | |
478 | assert(n); | |
479 | ||
480 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
481 | * ordered already. */ | |
482 | ||
1d54cd5d | 483 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
484 | |
485 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
486 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
487 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
488 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 489 | mount_entry_done(f); |
fe3c2583 LP |
490 | continue; |
491 | } | |
15ae422b | 492 | |
34de407a | 493 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
494 | |
495 | *t = *f; | |
15ae422b LP |
496 | t++; |
497 | } | |
498 | ||
c17ec25e | 499 | *n = t - m; |
15ae422b LP |
500 | } |
501 | ||
34de407a LP |
502 | static void drop_nop(MountEntry *m, unsigned *n) { |
503 | MountEntry *f, *t; | |
7648a565 LP |
504 | |
505 | assert(m); | |
506 | assert(n); | |
507 | ||
508 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
509 | * list is ordered by prefixes. */ | |
510 | ||
1d54cd5d | 511 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 LP |
512 | |
513 | /* Only suppress such subtrees for READONLY and READWRITE entries */ | |
514 | if (IN_SET(f->mode, READONLY, READWRITE)) { | |
34de407a | 515 | MountEntry *p; |
7648a565 LP |
516 | bool found = false; |
517 | ||
518 | /* Now let's find the first parent of the entry we are looking at. */ | |
519 | for (p = t-1; p >= m; p--) { | |
34de407a | 520 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
521 | found = true; |
522 | break; | |
523 | } | |
524 | } | |
525 | ||
526 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
527 | if (found && p->mode == f->mode) { | |
34de407a | 528 | log_debug("%s is redundant by %s", mount_entry_path(f), mount_entry_path(p)); |
1eb7e08e | 529 | mount_entry_done(f); |
7648a565 LP |
530 | continue; |
531 | } | |
532 | } | |
533 | ||
534 | *t = *f; | |
535 | t++; | |
536 | } | |
537 | ||
538 | *n = t - m; | |
539 | } | |
540 | ||
34de407a LP |
541 | static void drop_outside_root(const char *root_directory, MountEntry *m, unsigned *n) { |
542 | MountEntry *f, *t; | |
cd2902c9 LP |
543 | |
544 | assert(m); | |
545 | assert(n); | |
546 | ||
1d54cd5d | 547 | /* Nothing to do */ |
cd2902c9 LP |
548 | if (!root_directory) |
549 | return; | |
550 | ||
551 | /* Drops all mounts that are outside of the root directory. */ | |
552 | ||
1d54cd5d | 553 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 554 | |
34de407a LP |
555 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
556 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 557 | mount_entry_done(f); |
cd2902c9 LP |
558 | continue; |
559 | } | |
560 | ||
561 | *t = *f; | |
562 | t++; | |
563 | } | |
564 | ||
565 | *n = t - m; | |
566 | } | |
567 | ||
16498617 | 568 | static int clone_device_node(const char *d, const char *temporary_mount, bool *make_devnode) { |
6f7f3a33 | 569 | const char *dn; |
b5e99f23 ДГ |
570 | struct stat st; |
571 | int r; | |
572 | ||
414b304b | 573 | if (stat(d, &st) < 0) { |
b5e99f23 | 574 | if (errno == ENOENT) |
af984e13 | 575 | return -ENXIO; |
b5e99f23 ДГ |
576 | return -errno; |
577 | } | |
578 | ||
579 | if (!S_ISBLK(st.st_mode) && | |
580 | !S_ISCHR(st.st_mode)) | |
581 | return -EINVAL; | |
582 | ||
583 | if (st.st_rdev == 0) | |
af984e13 | 584 | return -ENXIO; |
b5e99f23 | 585 | |
6f7f3a33 | 586 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 587 | |
16498617 CB |
588 | if (*make_devnode) { |
589 | mac_selinux_create_file_prepare(d, st.st_mode); | |
590 | r = mknod(dn, st.st_mode, st.st_rdev); | |
591 | mac_selinux_create_file_clear(); | |
592 | ||
593 | if (r == 0) | |
af984e13 | 594 | return 0; |
16498617 CB |
595 | if (errno != EPERM) |
596 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
597 | ||
598 | *make_devnode = false; | |
599 | } | |
600 | ||
601 | /* We're about to fallback to bind-mounting the device | |
602 | * node. So create a dummy bind-mount target. */ | |
603 | mac_selinux_create_file_prepare(d, 0); | |
604 | r = mknod(dn, S_IFREG, 0); | |
b5e99f23 | 605 | mac_selinux_create_file_clear(); |
b5e99f23 | 606 | |
16498617 CB |
607 | if (r < 0 && errno != EEXIST) |
608 | return log_debug_errno(errno, "mknod fallback failed for %s: %m", d); | |
609 | ||
610 | /* Fallback to bind-mounting: | |
611 | * The assumption here is that all used device nodes carry standard | |
612 | * properties. Specifically, the devices nodes we bind-mount should | |
613 | * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx) | |
614 | * and should not carry ACLs. */ | |
615 | if (mount(d, dn, NULL, MS_BIND, NULL) < 0) | |
616 | return log_debug_errno(errno, "mount failed for %s: %m", d); | |
b5e99f23 | 617 | |
af984e13 | 618 | return 0; |
b5e99f23 ДГ |
619 | } |
620 | ||
5d997827 | 621 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
622 | static const char devnodes[] = |
623 | "/dev/null\0" | |
624 | "/dev/zero\0" | |
625 | "/dev/full\0" | |
626 | "/dev/random\0" | |
627 | "/dev/urandom\0" | |
628 | "/dev/tty\0"; | |
629 | ||
2b85f4e1 | 630 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 631 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 632 | bool can_mknod = true; |
7f112f50 LP |
633 | _cleanup_umask_ mode_t u; |
634 | int r; | |
635 | ||
636 | assert(m); | |
637 | ||
638 | u = umask(0000); | |
639 | ||
2b85f4e1 LP |
640 | if (!mkdtemp(temporary_mount)) |
641 | return -errno; | |
642 | ||
63c372cb | 643 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 644 | (void) mkdir(dev, 0755); |
737ba3c8 | 645 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2b85f4e1 LP |
646 | r = -errno; |
647 | goto fail; | |
648 | } | |
649 | ||
63c372cb | 650 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 651 | (void) mkdir(devpts, 0755); |
2b85f4e1 LP |
652 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
653 | r = -errno; | |
654 | goto fail; | |
655 | } | |
656 | ||
414b304b ДГ |
657 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx |
658 | * when /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible | |
659 | * thus, in that case make a clone | |
660 | * | |
661 | * in nspawn and other containers it will be a symlink, in that case make it a symlink | |
662 | */ | |
36ce7110 LP |
663 | r = is_symlink("/dev/ptmx"); |
664 | if (r < 0) | |
3164e3cb | 665 | goto fail; |
36ce7110 | 666 | if (r > 0) { |
414b304b ДГ |
667 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
668 | if (symlink("pts/ptmx", devptmx) < 0) { | |
669 | r = -errno; | |
670 | goto fail; | |
671 | } | |
672 | } else { | |
16498617 | 673 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
674 | if (r < 0) |
675 | goto fail; | |
414b304b | 676 | } |
e06b6479 | 677 | |
63c372cb | 678 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 679 | (void) mkdir(devshm, 0755); |
2b85f4e1 LP |
680 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
681 | if (r < 0) { | |
682 | r = -errno; | |
683 | goto fail; | |
684 | } | |
685 | ||
63c372cb | 686 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 687 | (void) mkdir(devmqueue, 0755); |
3164e3cb | 688 | (void) mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 689 | |
63c372cb | 690 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 691 | (void) mkdir(devhugepages, 0755); |
3164e3cb | 692 | (void) mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 693 | |
63c372cb | 694 | devlog = strjoina(temporary_mount, "/dev/log"); |
3164e3cb | 695 | (void) symlink("/run/systemd/journal/dev-log", devlog); |
82d25240 | 696 | |
7f112f50 | 697 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 698 | r = clone_device_node(d, temporary_mount, &can_mknod); |
af984e13 ZJS |
699 | /* ENXIO means the the *source* is not a device file, skip creation in that case */ |
700 | if (r < 0 && r != -ENXIO) | |
2b85f4e1 | 701 | goto fail; |
7f112f50 LP |
702 | } |
703 | ||
03cfe0d5 | 704 | dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
7f112f50 | 705 | |
ee818b89 AC |
706 | /* Create the /dev directory if missing. It is more likely to be |
707 | * missing when the service is started with RootDirectory. This is | |
708 | * consistent with mount units creating the mount points when missing. | |
709 | */ | |
34de407a | 710 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 711 | |
9e5f8252 | 712 | /* Unmount everything in old /dev */ |
34de407a LP |
713 | umount_recursive(mount_entry_path(m), 0); |
714 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { | |
2b85f4e1 LP |
715 | r = -errno; |
716 | goto fail; | |
717 | } | |
7f112f50 | 718 | |
2b85f4e1 LP |
719 | rmdir(dev); |
720 | rmdir(temporary_mount); | |
7f112f50 | 721 | |
2b85f4e1 | 722 | return 0; |
7f112f50 | 723 | |
2b85f4e1 LP |
724 | fail: |
725 | if (devpts) | |
726 | umount(devpts); | |
7f112f50 | 727 | |
2b85f4e1 LP |
728 | if (devshm) |
729 | umount(devshm); | |
7f112f50 | 730 | |
2b85f4e1 LP |
731 | if (devhugepages) |
732 | umount(devhugepages); | |
7f112f50 | 733 | |
2b85f4e1 LP |
734 | if (devmqueue) |
735 | umount(devmqueue); | |
7f112f50 | 736 | |
d267c5aa ZJS |
737 | umount(dev); |
738 | rmdir(dev); | |
2b85f4e1 | 739 | rmdir(temporary_mount); |
7f112f50 | 740 | |
2b85f4e1 | 741 | return r; |
7f112f50 LP |
742 | } |
743 | ||
2a2969fd | 744 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
745 | int r; |
746 | ||
747 | assert(m); | |
748 | ||
749 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
750 | * /dev. This is only used when RootDirectory= is set. */ | |
751 | ||
645767d6 LP |
752 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
753 | ||
5d997827 LP |
754 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
755 | if (r < 0) | |
756 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
757 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
758 | return 0; | |
759 | ||
760 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
761 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
762 | ||
763 | return 1; | |
764 | } | |
765 | ||
2a2969fd | 766 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
767 | int r; |
768 | ||
769 | assert(m); | |
770 | ||
645767d6 LP |
771 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
772 | ||
5d997827 LP |
773 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
774 | if (r < 0) | |
775 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
776 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
777 | return 0; | |
778 | ||
779 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
780 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
781 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
782 | ||
783 | return 1; | |
784 | } | |
785 | ||
2a2969fd | 786 | static int mount_procfs(const MountEntry *m) { |
5d997827 LP |
787 | int r; |
788 | ||
789 | assert(m); | |
790 | ||
645767d6 LP |
791 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
792 | ||
5d997827 LP |
793 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
794 | if (r < 0) | |
795 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
796 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
797 | return 0; | |
798 | ||
799 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
800 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
801 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
802 | ||
803 | return 1; | |
804 | } | |
805 | ||
2abd4e38 | 806 | static int mount_tmpfs(const MountEntry *m) { |
6c47cd7d LP |
807 | assert(m); |
808 | ||
2abd4e38 | 809 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d LP |
810 | |
811 | (void) mkdir_p_label(mount_entry_path(m), 0755); | |
812 | (void) umount_recursive(mount_entry_path(m), 0); | |
813 | ||
2abd4e38 | 814 | if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0) |
6c47cd7d LP |
815 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); |
816 | ||
817 | return 1; | |
818 | } | |
819 | ||
d2d6c096 LP |
820 | static int mount_entry_chase( |
821 | const char *root_directory, | |
2a2969fd | 822 | const MountEntry *m, |
d2d6c096 | 823 | const char *path, |
4ca763a9 | 824 | bool chase_nonexistent, |
d2d6c096 LP |
825 | char **location) { |
826 | ||
8fceda93 LP |
827 | char *chased; |
828 | int r; | |
829 | ||
830 | assert(m); | |
831 | ||
832 | /* Since mount() will always follow symlinks and we need to take the different root directory into account we | |
d2d6c096 LP |
833 | * chase the symlinks on our own first. This is called for the destination path, as well as the source path (if |
834 | * that applies). The result is stored in "location". */ | |
8fceda93 | 835 | |
62570f6f | 836 | r = chase_symlinks(path, root_directory, CHASE_TRAIL_SLASH | (chase_nonexistent ? CHASE_NONEXISTENT : 0), &chased); |
8fceda93 | 837 | if (r == -ENOENT && m->ignore) { |
d2d6c096 | 838 | log_debug_errno(r, "Path %s does not exist, ignoring.", path); |
8fceda93 LP |
839 | return 0; |
840 | } | |
841 | if (r < 0) | |
d2d6c096 | 842 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", path); |
8fceda93 | 843 | |
d2d6c096 | 844 | log_debug("Followed symlinks %s → %s.", path, chased); |
8fceda93 | 845 | |
d2d6c096 LP |
846 | free(*location); |
847 | *location = chased; | |
8fceda93 LP |
848 | |
849 | return 1; | |
850 | } | |
851 | ||
ac0930c8 | 852 | static int apply_mount( |
8fceda93 | 853 | const char *root_directory, |
89bd586c | 854 | MountEntry *m) { |
ac0930c8 | 855 | |
a227a4be | 856 | bool rbind = true, make = false; |
15ae422b | 857 | const char *what; |
15ae422b | 858 | int r; |
15ae422b | 859 | |
c17ec25e | 860 | assert(m); |
15ae422b | 861 | |
4ca763a9 | 862 | r = mount_entry_chase(root_directory, m, mount_entry_path(m), !IN_SET(m->mode, INACCESSIBLE, READONLY, READWRITE), &m->path_malloc); |
8fceda93 LP |
863 | if (r <= 0) |
864 | return r; | |
865 | ||
34de407a | 866 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 867 | |
c17ec25e | 868 | switch (m->mode) { |
15ae422b | 869 | |
160cfdbe LP |
870 | case INACCESSIBLE: { |
871 | struct stat target; | |
6d313367 LP |
872 | |
873 | /* First, get rid of everything that is below if there | |
874 | * is anything... Then, overmount it with an | |
c4b41707 | 875 | * inaccessible path. */ |
34de407a | 876 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 877 | |
34de407a LP |
878 | if (lstat(mount_entry_path(m), &target) < 0) |
879 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); | |
15ae422b | 880 | |
c4b41707 | 881 | what = mode_to_inaccessible_node(target.st_mode); |
5fd7cf6f LP |
882 | if (!what) { |
883 | log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 AP |
884 | return -ELOOP; |
885 | } | |
886 | break; | |
160cfdbe | 887 | } |
fe3c2583 | 888 | |
15ae422b | 889 | case READONLY: |
15ae422b | 890 | case READWRITE: |
8fceda93 | 891 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
d944dc95 | 892 | if (r < 0) |
34de407a | 893 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
894 | 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. */ |
895 | return 0; | |
6b7c9f8b | 896 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 897 | what = mount_entry_path(m); |
6b7c9f8b | 898 | break; |
15ae422b | 899 | |
d2d6c096 LP |
900 | case BIND_MOUNT: |
901 | rbind = false; | |
d2d6c096 | 902 | |
4831981d | 903 | _fallthrough_; |
d2d6c096 LP |
904 | case BIND_MOUNT_RECURSIVE: |
905 | /* Also chase the source mount */ | |
5d997827 | 906 | |
4ca763a9 | 907 | r = mount_entry_chase(root_directory, m, mount_entry_source(m), false, &m->source_malloc); |
d2d6c096 LP |
908 | if (r <= 0) |
909 | return r; | |
910 | ||
911 | what = mount_entry_source(m); | |
a227a4be | 912 | make = true; |
d2d6c096 LP |
913 | break; |
914 | ||
6c47cd7d | 915 | case EMPTY_DIR: |
2abd4e38 YW |
916 | case TMPFS: |
917 | return mount_tmpfs(m); | |
6c47cd7d | 918 | |
ac0930c8 | 919 | case PRIVATE_TMP: |
89bd586c | 920 | what = mount_entry_source(m); |
a227a4be | 921 | make = true; |
15ae422b | 922 | break; |
e364ad06 | 923 | |
d6797c92 | 924 | case PRIVATE_DEV: |
5d997827 LP |
925 | return mount_private_dev(m); |
926 | ||
927 | case BIND_DEV: | |
928 | return mount_bind_dev(m); | |
929 | ||
930 | case SYSFS: | |
931 | return mount_sysfs(m); | |
932 | ||
933 | case PROCFS: | |
934 | return mount_procfs(m); | |
d6797c92 | 935 | |
e364ad06 LP |
936 | default: |
937 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
938 | } |
939 | ||
ac0930c8 | 940 | assert(what); |
15ae422b | 941 | |
a227a4be LP |
942 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) { |
943 | bool try_again = false; | |
944 | r = -errno; | |
945 | ||
946 | if (r == -ENOENT && make) { | |
947 | struct stat st; | |
948 | ||
949 | /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */ | |
950 | ||
951 | if (stat(what, &st) >= 0) { | |
952 | ||
953 | (void) mkdir_parents(mount_entry_path(m), 0755); | |
954 | ||
955 | if (S_ISDIR(st.st_mode)) | |
956 | try_again = mkdir(mount_entry_path(m), 0755) >= 0; | |
957 | else | |
958 | try_again = touch(mount_entry_path(m)) >= 0; | |
959 | } | |
960 | } | |
961 | ||
962 | if (try_again) { | |
963 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) | |
964 | r = -errno; | |
965 | else | |
966 | r = 0; | |
967 | } | |
968 | ||
969 | if (r < 0) | |
970 | return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); | |
971 | } | |
6b7c9f8b | 972 | |
34de407a | 973 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 974 | return 0; |
ac0930c8 | 975 | } |
15ae422b | 976 | |
2a2969fd | 977 | static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) { |
6b7c9f8b | 978 | int r = 0; |
15ae422b | 979 | |
c17ec25e | 980 | assert(m); |
ac9de0b3 | 981 | assert(proc_self_mountinfo); |
ac0930c8 | 982 | |
2abd4e38 YW |
983 | if (mount_entry_read_only(m)) { |
984 | if (IN_SET(m->mode, EMPTY_DIR, TMPFS)) { | |
985 | /* Make superblock readonly */ | |
986 | if (mount(NULL, mount_entry_path(m), NULL, MS_REMOUNT | MS_RDONLY | m->flags, mount_entry_options(m)) < 0) | |
987 | r = -errno; | |
988 | } else | |
989 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), true, blacklist, proc_self_mountinfo); | |
990 | } else if (m->mode == PRIVATE_DEV) { | |
991 | /* Superblock can be readonly but the submounts can't */ | |
34de407a | 992 | if (mount(NULL, mount_entry_path(m), NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0) |
6b7c9f8b | 993 | r = -errno; |
737ba3c8 | 994 | } else |
6b7c9f8b LP |
995 | return 0; |
996 | ||
997 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
998 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
999 | * read-only mounts already applied. */ | |
ac0930c8 | 1000 | |
8fceda93 LP |
1001 | if (r == -ENOENT && m->ignore) |
1002 | r = 0; | |
5327c910 | 1003 | |
1d54cd5d | 1004 | return r; |
d944dc95 LP |
1005 | } |
1006 | ||
bb0ff3fb | 1007 | static bool namespace_info_mount_apivfs(const char *root_directory, const NamespaceInfo *ns_info) { |
5d997827 LP |
1008 | assert(ns_info); |
1009 | ||
9c988f93 DH |
1010 | /* |
1011 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1012 | * since to protect the API VFS mounts, they need to be around in the | |
1013 | * first place... and RootDirectory= or RootImage= need to be set. | |
1014 | */ | |
5d997827 | 1015 | |
9c988f93 DH |
1016 | /* root_directory should point to a mount point */ |
1017 | return root_directory && | |
1018 | (ns_info->mount_apivfs || | |
1019 | ns_info->protect_control_groups || | |
1020 | ns_info->protect_kernel_tunables); | |
5d997827 LP |
1021 | } |
1022 | ||
2652c6c1 | 1023 | static unsigned namespace_calculate_mounts( |
9c988f93 | 1024 | const char* root_directory, |
bb0ff3fb | 1025 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1026 | char** read_write_paths, |
1027 | char** read_only_paths, | |
1028 | char** inaccessible_paths, | |
6c47cd7d | 1029 | char** empty_directories, |
d2d6c096 | 1030 | unsigned n_bind_mounts, |
2abd4e38 | 1031 | unsigned n_temporary_filesystems, |
2652c6c1 DH |
1032 | const char* tmp_dir, |
1033 | const char* var_tmp_dir, | |
2652c6c1 DH |
1034 | ProtectHome protect_home, |
1035 | ProtectSystem protect_system) { | |
1036 | ||
b6c432ca | 1037 | unsigned protect_home_cnt; |
f471b2af DH |
1038 | unsigned protect_system_cnt = |
1039 | (protect_system == PROTECT_SYSTEM_STRICT ? | |
1040 | ELEMENTSOF(protect_system_strict_table) : | |
1041 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
1042 | ELEMENTSOF(protect_system_full_table) : | |
1043 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
1044 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
1045 | ||
b6c432ca DH |
1046 | protect_home_cnt = |
1047 | (protect_home == PROTECT_HOME_YES ? | |
1048 | ELEMENTSOF(protect_home_yes_table) : | |
1049 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
e4da7d8c YW |
1050 | ELEMENTSOF(protect_home_read_only_table) : |
1051 | ((protect_home == PROTECT_HOME_TMPFS) ? | |
1052 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); | |
b6c432ca | 1053 | |
2652c6c1 DH |
1054 | return !!tmp_dir + !!var_tmp_dir + |
1055 | strv_length(read_write_paths) + | |
1056 | strv_length(read_only_paths) + | |
1057 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1058 | strv_length(empty_directories) + |
d2d6c096 | 1059 | n_bind_mounts + |
2abd4e38 | 1060 | n_temporary_filesystems + |
c575770b DH |
1061 | ns_info->private_dev + |
1062 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
1063 | (ns_info->protect_control_groups ? 1 : 0) + | |
1064 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + | |
5d997827 | 1065 | protect_home_cnt + protect_system_cnt + |
9c988f93 | 1066 | (namespace_info_mount_apivfs(root_directory, ns_info) ? ELEMENTSOF(apivfs_table) : 0); |
2652c6c1 DH |
1067 | } |
1068 | ||
613b411c | 1069 | int setup_namespace( |
ee818b89 | 1070 | const char* root_directory, |
915e6d16 | 1071 | const char* root_image, |
bb0ff3fb | 1072 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1073 | char** read_write_paths, |
1074 | char** read_only_paths, | |
1075 | char** inaccessible_paths, | |
6c47cd7d | 1076 | char** empty_directories, |
d2d6c096 LP |
1077 | const BindMount *bind_mounts, |
1078 | unsigned n_bind_mounts, | |
2abd4e38 YW |
1079 | const TemporaryFileSystem *temporary_filesystems, |
1080 | unsigned n_temporary_filesystems, | |
a004cb4c LP |
1081 | const char* tmp_dir, |
1082 | const char* var_tmp_dir, | |
1b8689f9 LP |
1083 | ProtectHome protect_home, |
1084 | ProtectSystem protect_system, | |
915e6d16 LP |
1085 | unsigned long mount_flags, |
1086 | DissectImageFlags dissect_image_flags) { | |
15ae422b | 1087 | |
915e6d16 | 1088 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1089 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1090 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 1091 | _cleanup_free_ void *root_hash = NULL; |
34de407a | 1092 | MountEntry *m, *mounts = NULL; |
78ebe980 | 1093 | size_t root_hash_size = 0; |
d944dc95 | 1094 | bool make_slave = false; |
e908468b | 1095 | const char *root; |
f0a4feb0 | 1096 | unsigned n_mounts; |
d18aff04 | 1097 | bool require_prefix = false; |
c17ec25e | 1098 | int r = 0; |
15ae422b | 1099 | |
915e6d16 LP |
1100 | assert(ns_info); |
1101 | ||
613b411c | 1102 | if (mount_flags == 0) |
c17ec25e | 1103 | mount_flags = MS_SHARED; |
ac0930c8 | 1104 | |
915e6d16 LP |
1105 | if (root_image) { |
1106 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1107 | ||
1108 | if (protect_system == PROTECT_SYSTEM_STRICT && strv_isempty(read_write_paths)) | |
1109 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; | |
1110 | ||
1111 | r = loop_device_make_by_path(root_image, | |
1112 | dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR, | |
1113 | &loop_device); | |
1114 | if (r < 0) | |
1115 | return r; | |
1116 | ||
78ebe980 LP |
1117 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
1118 | if (r < 0) | |
1119 | return r; | |
1120 | ||
1121 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
1122 | if (r < 0) | |
1123 | return r; | |
1124 | ||
1125 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 LP |
1126 | if (r < 0) |
1127 | return r; | |
915e6d16 LP |
1128 | } |
1129 | ||
e908468b LP |
1130 | if (root_directory) |
1131 | root = root_directory; | |
2abd4e38 | 1132 | else if (root_image || n_bind_mounts > 0 || n_temporary_filesystems > 0) { |
e908468b LP |
1133 | |
1134 | /* If we are booting from an image, create a mount point for the image, if it's still missing. We use | |
1135 | * the same mount point for all images, which is safe, since they all live in their own namespaces | |
1136 | * after all, and hence won't see each other. We also use such a root directory whenever there are bind | |
1137 | * mounts configured, so that their source mounts are never obstructed by mounts we already applied | |
1138 | * while we are applying them. */ | |
1139 | ||
1140 | root = "/run/systemd/unit-root"; | |
1141 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1142 | require_prefix = true; |
e908468b LP |
1143 | } else |
1144 | root = NULL; | |
1145 | ||
cfbeb4ef | 1146 | n_mounts = namespace_calculate_mounts( |
e908468b | 1147 | root, |
cfbeb4ef LP |
1148 | ns_info, |
1149 | read_write_paths, | |
1150 | read_only_paths, | |
1151 | inaccessible_paths, | |
6c47cd7d | 1152 | empty_directories, |
f5c52a77 | 1153 | n_bind_mounts, |
2abd4e38 | 1154 | n_temporary_filesystems, |
cfbeb4ef LP |
1155 | tmp_dir, var_tmp_dir, |
1156 | protect_home, protect_system); | |
613b411c | 1157 | |
2652c6c1 | 1158 | /* Set mount slave mode */ |
e908468b | 1159 | if (root || n_mounts > 0) |
d944dc95 LP |
1160 | make_slave = true; |
1161 | ||
f0a4feb0 | 1162 | if (n_mounts > 0) { |
34de407a | 1163 | m = mounts = (MountEntry *) alloca0(n_mounts * sizeof(MountEntry)); |
d18aff04 | 1164 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1165 | if (r < 0) |
f0a4feb0 | 1166 | goto finish; |
613b411c | 1167 | |
d18aff04 | 1168 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1169 | if (r < 0) |
f0a4feb0 | 1170 | goto finish; |
613b411c | 1171 | |
d18aff04 | 1172 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1173 | if (r < 0) |
f0a4feb0 | 1174 | goto finish; |
7ff7394d | 1175 | |
6c47cd7d LP |
1176 | r = append_empty_dir_mounts(&m, empty_directories); |
1177 | if (r < 0) | |
1178 | goto finish; | |
1179 | ||
d2d6c096 LP |
1180 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1181 | if (r < 0) | |
1182 | goto finish; | |
1183 | ||
2abd4e38 YW |
1184 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1185 | if (r < 0) | |
1186 | goto finish; | |
1187 | ||
613b411c | 1188 | if (tmp_dir) { |
34de407a | 1189 | *(m++) = (MountEntry) { |
5327c910 LP |
1190 | .path_const = "/tmp", |
1191 | .mode = PRIVATE_TMP, | |
89bd586c | 1192 | .source_const = tmp_dir, |
5327c910 | 1193 | }; |
613b411c | 1194 | } |
7ff7394d | 1195 | |
613b411c | 1196 | if (var_tmp_dir) { |
34de407a | 1197 | *(m++) = (MountEntry) { |
5327c910 | 1198 | .path_const = "/var/tmp", |
89bd586c YW |
1199 | .mode = PRIVATE_TMP, |
1200 | .source_const = var_tmp_dir, | |
5327c910 | 1201 | }; |
7ff7394d | 1202 | } |
ac0930c8 | 1203 | |
c575770b | 1204 | if (ns_info->private_dev) { |
34de407a | 1205 | *(m++) = (MountEntry) { |
5327c910 LP |
1206 | .path_const = "/dev", |
1207 | .mode = PRIVATE_DEV, | |
1208 | }; | |
7f112f50 LP |
1209 | } |
1210 | ||
c575770b | 1211 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 1212 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 1213 | if (r < 0) |
f0a4feb0 | 1214 | goto finish; |
c575770b DH |
1215 | } |
1216 | ||
1217 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 1218 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 1219 | if (r < 0) |
f0a4feb0 | 1220 | goto finish; |
c575770b | 1221 | } |
59eeb84b | 1222 | |
c575770b | 1223 | if (ns_info->protect_control_groups) { |
34de407a | 1224 | *(m++) = (MountEntry) { |
5327c910 LP |
1225 | .path_const = "/sys/fs/cgroup", |
1226 | .mode = READONLY, | |
1227 | }; | |
59eeb84b LP |
1228 | } |
1229 | ||
5327c910 | 1230 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1231 | if (r < 0) |
f0a4feb0 | 1232 | goto finish; |
417116f2 | 1233 | |
5327c910 | 1234 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1235 | if (r < 0) |
f0a4feb0 | 1236 | goto finish; |
417116f2 | 1237 | |
e908468b | 1238 | if (namespace_info_mount_apivfs(root, ns_info)) { |
5d997827 LP |
1239 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); |
1240 | if (r < 0) | |
1241 | goto finish; | |
1242 | } | |
1243 | ||
f0a4feb0 | 1244 | assert(mounts + n_mounts == m); |
ac0930c8 | 1245 | |
5327c910 | 1246 | /* Prepend the root directory where that's necessary */ |
e908468b | 1247 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1248 | if (r < 0) |
1249 | goto finish; | |
1250 | ||
34de407a | 1251 | qsort(mounts, n_mounts, sizeof(MountEntry), mount_path_compare); |
fe3c2583 | 1252 | |
f0a4feb0 | 1253 | drop_duplicates(mounts, &n_mounts); |
e908468b | 1254 | drop_outside_root(root, mounts, &n_mounts); |
f0a4feb0 DH |
1255 | drop_inaccessible(mounts, &n_mounts); |
1256 | drop_nop(mounts, &n_mounts); | |
15ae422b LP |
1257 | } |
1258 | ||
d944dc95 LP |
1259 | if (unshare(CLONE_NEWNS) < 0) { |
1260 | r = -errno; | |
1261 | goto finish; | |
1262 | } | |
1e4e94c8 | 1263 | |
d944dc95 | 1264 | if (make_slave) { |
c2c13f2d LP |
1265 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1266 | shows up in the parent */ | |
d944dc95 LP |
1267 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { |
1268 | r = -errno; | |
1269 | goto finish; | |
1270 | } | |
ee818b89 AC |
1271 | } |
1272 | ||
915e6d16 | 1273 | if (root_image) { |
e908468b | 1274 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1275 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
915e6d16 LP |
1276 | if (r < 0) |
1277 | goto finish; | |
1278 | ||
07ce7407 TM |
1279 | if (decrypted_image) { |
1280 | r = decrypted_image_relinquish(decrypted_image); | |
1281 | if (r < 0) | |
1282 | goto finish; | |
1283 | } | |
78ebe980 | 1284 | |
915e6d16 LP |
1285 | loop_device_relinquish(loop_device); |
1286 | ||
1287 | } else if (root_directory) { | |
1288 | ||
e908468b LP |
1289 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1290 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
8f1ad200 | 1291 | if (r < 0) |
d944dc95 | 1292 | goto finish; |
8f1ad200 | 1293 | if (r == 0) { |
e908468b | 1294 | if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) { |
8f1ad200 LP |
1295 | r = -errno; |
1296 | goto finish; | |
1297 | } | |
d944dc95 | 1298 | } |
e908468b LP |
1299 | |
1300 | } else if (root) { | |
1301 | ||
1302 | /* Let's mount the main root directory to the root directory to use */ | |
1303 | if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
1304 | r = -errno; | |
1305 | goto finish; | |
1306 | } | |
ee818b89 | 1307 | } |
c2c13f2d | 1308 | |
4e0c20de LP |
1309 | /* Try to set up the new root directory before mounting anything else there. */ |
1310 | if (root_image || root_directory) | |
1311 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1312 | ||
f0a4feb0 | 1313 | if (n_mounts > 0) { |
ac9de0b3 | 1314 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b7c9f8b LP |
1315 | char **blacklist; |
1316 | unsigned j; | |
1317 | ||
ac9de0b3 TR |
1318 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc. |
1319 | * For example, this is the case with the option: 'InaccessiblePaths=/proc' */ | |
1320 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); | |
1321 | if (!proc_self_mountinfo) { | |
1322 | r = -errno; | |
1323 | goto finish; | |
1324 | } | |
1325 | ||
6b7c9f8b | 1326 | /* First round, add in all special mounts we need */ |
f0a4feb0 | 1327 | for (m = mounts; m < mounts + n_mounts; ++m) { |
89bd586c | 1328 | r = apply_mount(root, m); |
c2c13f2d | 1329 | if (r < 0) |
d944dc95 | 1330 | goto finish; |
c2c13f2d | 1331 | } |
15ae422b | 1332 | |
6b7c9f8b | 1333 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
f0a4feb0 DH |
1334 | blacklist = newa(char*, n_mounts+1); |
1335 | for (j = 0; j < n_mounts; j++) | |
34de407a | 1336 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1337 | blacklist[j] = NULL; |
1338 | ||
1339 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1340 | for (m = mounts; m < mounts + n_mounts; ++m) { |
ac9de0b3 | 1341 | r = make_read_only(m, blacklist, proc_self_mountinfo); |
c2c13f2d | 1342 | if (r < 0) |
d944dc95 | 1343 | goto finish; |
c2c13f2d | 1344 | } |
15ae422b LP |
1345 | } |
1346 | ||
e908468b | 1347 | if (root) { |
ee818b89 | 1348 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
e908468b | 1349 | r = mount_move_root(root); |
d944dc95 LP |
1350 | if (r < 0) |
1351 | goto finish; | |
ee818b89 AC |
1352 | } |
1353 | ||
55fe7432 | 1354 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1355 | * reestablish propagation from our side to the host, since |
1356 | * what's disconnected is disconnected. */ | |
d944dc95 LP |
1357 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
1358 | r = -errno; | |
1359 | goto finish; | |
1360 | } | |
15ae422b | 1361 | |
d944dc95 | 1362 | r = 0; |
15ae422b | 1363 | |
d944dc95 | 1364 | finish: |
f0a4feb0 | 1365 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1366 | mount_entry_done(m); |
613b411c LP |
1367 | |
1368 | return r; | |
1369 | } | |
1370 | ||
d2d6c096 LP |
1371 | void bind_mount_free_many(BindMount *b, unsigned n) { |
1372 | unsigned i; | |
1373 | ||
1374 | assert(b || n == 0); | |
1375 | ||
1376 | for (i = 0; i < n; i++) { | |
1377 | free(b[i].source); | |
1378 | free(b[i].destination); | |
1379 | } | |
1380 | ||
1381 | free(b); | |
1382 | } | |
1383 | ||
1384 | int bind_mount_add(BindMount **b, unsigned *n, const BindMount *item) { | |
1385 | _cleanup_free_ char *s = NULL, *d = NULL; | |
1386 | BindMount *c; | |
1387 | ||
1388 | assert(b); | |
1389 | assert(n); | |
1390 | assert(item); | |
1391 | ||
1392 | s = strdup(item->source); | |
1393 | if (!s) | |
1394 | return -ENOMEM; | |
1395 | ||
1396 | d = strdup(item->destination); | |
1397 | if (!d) | |
1398 | return -ENOMEM; | |
1399 | ||
aa484f35 | 1400 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
1401 | if (!c) |
1402 | return -ENOMEM; | |
1403 | ||
1404 | *b = c; | |
1405 | ||
1406 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
1407 | .source = TAKE_PTR(s), |
1408 | .destination = TAKE_PTR(d), | |
d2d6c096 LP |
1409 | .read_only = item->read_only, |
1410 | .recursive = item->recursive, | |
1411 | .ignore_enoent = item->ignore_enoent, | |
1412 | }; | |
1413 | ||
d2d6c096 LP |
1414 | return 0; |
1415 | } | |
1416 | ||
2abd4e38 YW |
1417 | void temporary_filesystem_free_many(TemporaryFileSystem *t, unsigned n) { |
1418 | unsigned i; | |
1419 | ||
1420 | assert(t || n == 0); | |
1421 | ||
1422 | for (i = 0; i < n; i++) { | |
1423 | free(t[i].path); | |
1424 | free(t[i].options); | |
1425 | } | |
1426 | ||
1427 | free(t); | |
1428 | } | |
1429 | ||
1430 | int temporary_filesystem_add( | |
1431 | TemporaryFileSystem **t, | |
1432 | unsigned *n, | |
1433 | const char *path, | |
1434 | const char *options) { | |
1435 | ||
1436 | _cleanup_free_ char *p = NULL, *o = NULL; | |
1437 | TemporaryFileSystem *c; | |
1438 | ||
1439 | assert(t); | |
1440 | assert(n); | |
1441 | assert(path); | |
1442 | ||
1443 | p = strdup(path); | |
1444 | if (!p) | |
1445 | return -ENOMEM; | |
1446 | ||
1447 | if (!isempty(options)) { | |
1448 | o = strdup(options); | |
1449 | if (!o) | |
1450 | return -ENOMEM; | |
1451 | } | |
1452 | ||
aa484f35 | 1453 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
1454 | if (!c) |
1455 | return -ENOMEM; | |
1456 | ||
1457 | *t = c; | |
1458 | ||
1459 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
1460 | .path = TAKE_PTR(p), |
1461 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
1462 | }; |
1463 | ||
2abd4e38 YW |
1464 | return 0; |
1465 | } | |
1466 | ||
613b411c LP |
1467 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1468 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1469 | char bid[SD_ID128_STRING_MAX]; |
1470 | sd_id128_t boot_id; | |
1471 | int r; | |
613b411c LP |
1472 | |
1473 | assert(id); | |
1474 | assert(prefix); | |
1475 | assert(path); | |
1476 | ||
6b46ea73 LP |
1477 | /* We include the boot id in the directory so that after a |
1478 | * reboot we can easily identify obsolete directories. */ | |
1479 | ||
1480 | r = sd_id128_get_boot(&boot_id); | |
1481 | if (r < 0) | |
1482 | return r; | |
1483 | ||
605405c6 | 1484 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1485 | if (!x) |
1486 | return -ENOMEM; | |
1487 | ||
1488 | RUN_WITH_UMASK(0077) | |
1489 | if (!mkdtemp(x)) | |
1490 | return -errno; | |
1491 | ||
1492 | RUN_WITH_UMASK(0000) { | |
1493 | char *y; | |
1494 | ||
63c372cb | 1495 | y = strjoina(x, "/tmp"); |
613b411c LP |
1496 | |
1497 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1498 | return -errno; | |
c17ec25e | 1499 | } |
15ae422b | 1500 | |
1cc6c93a | 1501 | *path = TAKE_PTR(x); |
613b411c LP |
1502 | |
1503 | return 0; | |
1504 | } | |
1505 | ||
1506 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1507 | char *a, *b; | |
1508 | int r; | |
1509 | ||
1510 | assert(id); | |
1511 | assert(tmp_dir); | |
1512 | assert(var_tmp_dir); | |
1513 | ||
1514 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1515 | if (r < 0) | |
1516 | return r; | |
1517 | ||
1518 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1519 | if (r < 0) { | |
1520 | char *t; | |
1521 | ||
63c372cb | 1522 | t = strjoina(a, "/tmp"); |
613b411c LP |
1523 | rmdir(t); |
1524 | rmdir(a); | |
1525 | ||
1526 | free(a); | |
1527 | return r; | |
1528 | } | |
1529 | ||
1530 | *tmp_dir = a; | |
1531 | *var_tmp_dir = b; | |
1532 | ||
1533 | return 0; | |
1534 | } | |
1535 | ||
1536 | int setup_netns(int netns_storage_socket[2]) { | |
1537 | _cleanup_close_ int netns = -1; | |
3ee897d6 | 1538 | int r, q; |
613b411c LP |
1539 | |
1540 | assert(netns_storage_socket); | |
1541 | assert(netns_storage_socket[0] >= 0); | |
1542 | assert(netns_storage_socket[1] >= 0); | |
1543 | ||
1544 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1545 | * namespace reference fd. Whatever process runs this first |
1546 | * shall create a new namespace, all others should just join | |
1547 | * it. To serialize that we use a file lock on the socket | |
1548 | * pair. | |
613b411c LP |
1549 | * |
1550 | * It's a bit crazy, but hey, works great! */ | |
1551 | ||
1552 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1553 | return -errno; | |
1554 | ||
3ee897d6 LP |
1555 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1556 | if (netns == -EAGAIN) { | |
613b411c LP |
1557 | /* Nothing stored yet, so let's create a new namespace */ |
1558 | ||
1559 | if (unshare(CLONE_NEWNET) < 0) { | |
1560 | r = -errno; | |
1561 | goto fail; | |
1562 | } | |
1563 | ||
1564 | loopback_setup(); | |
1565 | ||
1566 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1567 | if (netns < 0) { | |
1568 | r = -errno; | |
1569 | goto fail; | |
1570 | } | |
1571 | ||
1572 | r = 1; | |
613b411c | 1573 | |
3ee897d6 LP |
1574 | } else if (netns < 0) { |
1575 | r = netns; | |
1576 | goto fail; | |
613b411c | 1577 | |
3ee897d6 LP |
1578 | } else { |
1579 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1580 | if (setns(netns, CLONE_NEWNET) < 0) { |
1581 | r = -errno; | |
1582 | goto fail; | |
1583 | } | |
1584 | ||
1585 | r = 0; | |
1586 | } | |
1587 | ||
3ee897d6 LP |
1588 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1589 | if (q < 0) { | |
1590 | r = q; | |
613b411c LP |
1591 | goto fail; |
1592 | } | |
1593 | ||
1594 | fail: | |
fe048ce5 | 1595 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1596 | return r; |
1597 | } | |
417116f2 | 1598 | |
6e2d7c4f MS |
1599 | bool ns_type_supported(NamespaceType type) { |
1600 | const char *t, *ns_proc; | |
1601 | ||
0fa5b831 LP |
1602 | t = namespace_type_to_string(type); |
1603 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
1604 | return false; |
1605 | ||
6e2d7c4f | 1606 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
1607 | return access(ns_proc, F_OK) == 0; |
1608 | } | |
1609 | ||
1b8689f9 LP |
1610 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1611 | [PROTECT_HOME_NO] = "no", | |
1612 | [PROTECT_HOME_YES] = "yes", | |
1613 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
e4da7d8c | 1614 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
1615 | }; |
1616 | ||
1b8689f9 LP |
1617 | DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome); |
1618 | ||
5e1c6154 YW |
1619 | ProtectHome parse_protect_home_or_bool(const char *s) { |
1620 | int r; | |
1621 | ||
1622 | r = parse_boolean(s); | |
1623 | if (r > 0) | |
1624 | return PROTECT_HOME_YES; | |
1625 | if (r == 0) | |
1626 | return PROTECT_HOME_NO; | |
1627 | ||
1628 | return protect_home_from_string(s); | |
1629 | } | |
1630 | ||
1b8689f9 LP |
1631 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
1632 | [PROTECT_SYSTEM_NO] = "no", | |
1633 | [PROTECT_SYSTEM_YES] = "yes", | |
1634 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1635 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1636 | }; |
1637 | ||
1638 | DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem); | |
6e2d7c4f | 1639 | |
03c791aa YW |
1640 | ProtectSystem parse_protect_system_or_bool(const char *s) { |
1641 | int r; | |
1642 | ||
1643 | r = parse_boolean(s); | |
1644 | if (r > 0) | |
1645 | return PROTECT_SYSTEM_YES; | |
1646 | if (r == 0) | |
1647 | return PROTECT_SYSTEM_NO; | |
1648 | ||
1649 | return protect_system_from_string(s); | |
1650 | } | |
1651 | ||
6e2d7c4f MS |
1652 | static const char* const namespace_type_table[] = { |
1653 | [NAMESPACE_MOUNT] = "mnt", | |
1654 | [NAMESPACE_CGROUP] = "cgroup", | |
1655 | [NAMESPACE_UTS] = "uts", | |
1656 | [NAMESPACE_IPC] = "ipc", | |
1657 | [NAMESPACE_USER] = "user", | |
1658 | [NAMESPACE_PID] = "pid", | |
1659 | [NAMESPACE_NET] = "net", | |
1660 | }; | |
1661 | ||
1662 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); |