]>
Commit | Line | Data |
---|---|---|
15ae422b LP |
1 | /*** |
2 | This file is part of systemd. | |
3 | ||
4 | Copyright 2010 Lennart Poettering | |
5 | ||
6 | systemd is free software; you can redistribute it and/or modify it | |
5430f7f2 LP |
7 | under the terms of the GNU Lesser General Public License as published by |
8 | the Free Software Foundation; either version 2.1 of the License, or | |
15ae422b LP |
9 | (at your option) any later version. |
10 | ||
11 | systemd is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
5430f7f2 | 14 | Lesser General Public License for more details. |
15ae422b | 15 | |
5430f7f2 | 16 | You should have received a copy of the GNU Lesser General Public License |
15ae422b LP |
17 | along with systemd; If not, see <http://www.gnu.org/licenses/>. |
18 | ***/ | |
19 | ||
20 | #include <errno.h> | |
07630cea | 21 | #include <sched.h> |
15ae422b | 22 | #include <stdio.h> |
07630cea LP |
23 | #include <string.h> |
24 | #include <sys/mount.h> | |
15ae422b | 25 | #include <sys/stat.h> |
07630cea | 26 | #include <unistd.h> |
25e870b5 | 27 | #include <linux/fs.h> |
15ae422b | 28 | |
b5efdb8a | 29 | #include "alloc-util.h" |
7f112f50 | 30 | #include "dev-setup.h" |
3ffd4af2 | 31 | #include "fd-util.h" |
d944dc95 | 32 | #include "fs-util.h" |
915e6d16 | 33 | #include "loop-util.h" |
07630cea LP |
34 | #include "loopback-setup.h" |
35 | #include "missing.h" | |
36 | #include "mkdir.h" | |
4349cd7c | 37 | #include "mount-util.h" |
3ffd4af2 | 38 | #include "namespace.h" |
07630cea | 39 | #include "path-util.h" |
d7b8eec7 | 40 | #include "selinux-util.h" |
2583fbea | 41 | #include "socket-util.h" |
8b43440b | 42 | #include "string-table.h" |
07630cea LP |
43 | #include "string-util.h" |
44 | #include "strv.h" | |
affb60b1 | 45 | #include "umask-util.h" |
ee104e11 | 46 | #include "user-util.h" |
07630cea | 47 | #include "util.h" |
15ae422b | 48 | |
737ba3c8 | 49 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
50 | ||
c17ec25e | 51 | typedef enum MountMode { |
15ae422b LP |
52 | /* This is ordered by priority! */ |
53 | INACCESSIBLE, | |
d2d6c096 LP |
54 | BIND_MOUNT, |
55 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 LP |
56 | PRIVATE_TMP, |
57 | PRIVATE_VAR_TMP, | |
7f112f50 | 58 | PRIVATE_DEV, |
5d997827 LP |
59 | BIND_DEV, |
60 | SYSFS, | |
61 | PROCFS, | |
62 | READONLY, | |
59eeb84b | 63 | READWRITE, |
c17ec25e | 64 | } MountMode; |
15ae422b | 65 | |
34de407a | 66 | typedef struct MountEntry { |
5327c910 | 67 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 68 | MountMode mode:5; |
5327c910 LP |
69 | bool ignore:1; /* Ignore if path does not exist? */ |
70 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 71 | bool read_only:1; /* Shall this mount point be read-only? */ |
5327c910 | 72 | char *path_malloc; /* Use this instead of 'path' if we had to allocate memory */ |
d2d6c096 LP |
73 | const char *source_const; /* The source path, for bind mounts */ |
74 | char *source_malloc; | |
34de407a | 75 | } MountEntry; |
15ae422b | 76 | |
5d997827 LP |
77 | /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted |
78 | * something there already. These mounts are hence overriden by any other explicitly configured mounts. */ | |
79 | static const MountEntry apivfs_table[] = { | |
80 | { "/proc", PROCFS, false }, | |
81 | { "/dev", BIND_DEV, false }, | |
82 | { "/sys", SYSFS, false }, | |
83 | }; | |
f471b2af | 84 | |
11a30cec | 85 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 86 | static const MountEntry protect_kernel_tunables_table[] = { |
c6232fb0 LP |
87 | { "/proc/sys", READONLY, false }, |
88 | { "/proc/sysrq-trigger", READONLY, true }, | |
89 | { "/proc/latency_stats", READONLY, true }, | |
90 | { "/proc/mtrr", READONLY, true }, | |
aa70f38b | 91 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ |
c6232fb0 LP |
92 | { "/proc/acpi", READONLY, true }, |
93 | { "/proc/timer_stats", READONLY, true }, | |
94 | { "/proc/asound", READONLY, true }, | |
95 | { "/proc/bus", READONLY, true }, | |
96 | { "/proc/fs", READONLY, true }, | |
97 | { "/proc/irq", READONLY, true }, | |
98 | { "/sys", READONLY, false }, | |
99 | { "/sys/kernel/debug", READONLY, true }, | |
100 | { "/sys/kernel/tracing", READONLY, true }, | |
101 | { "/sys/fs/cgroup", READWRITE, false }, /* READONLY is set by ProtectControlGroups= option */ | |
11a30cec DH |
102 | }; |
103 | ||
c575770b | 104 | /* ProtectKernelModules= option */ |
34de407a | 105 | static const MountEntry protect_kernel_modules_table[] = { |
c575770b | 106 | #ifdef HAVE_SPLIT_USR |
c6232fb0 | 107 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 108 | #endif |
c6232fb0 | 109 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
110 | }; |
111 | ||
b6c432ca DH |
112 | /* |
113 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
114 | * system should be protected by ProtectSystem= | |
115 | */ | |
34de407a | 116 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
117 | { "/home", READONLY, true }, |
118 | { "/run/user", READONLY, true }, | |
119 | { "/root", READONLY, true }, | |
b6c432ca DH |
120 | }; |
121 | ||
122 | /* ProtectHome=yes table */ | |
34de407a | 123 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
124 | { "/home", INACCESSIBLE, true }, |
125 | { "/run/user", INACCESSIBLE, true }, | |
126 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
127 | }; |
128 | ||
f471b2af | 129 | /* ProtectSystem=yes table */ |
34de407a | 130 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
131 | { "/usr", READONLY, false }, |
132 | { "/boot", READONLY, true }, | |
133 | { "/efi", READONLY, true }, | |
f471b2af DH |
134 | }; |
135 | ||
136 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 137 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
138 | { "/usr", READONLY, false }, |
139 | { "/boot", READONLY, true }, | |
140 | { "/efi", READONLY, true }, | |
141 | { "/etc", READONLY, false }, | |
f471b2af DH |
142 | }; |
143 | ||
144 | /* | |
145 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
146 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
147 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
148 | * protect those, and these options should be fully orthogonal. | |
149 | * (And of course /home and friends are also left writable, as ProtectHome= | |
150 | * shall manage those, orthogonally). | |
151 | */ | |
34de407a | 152 | static const MountEntry protect_system_strict_table[] = { |
ddbe0412 LP |
153 | { "/", READONLY, false }, |
154 | { "/proc", READWRITE, false }, /* ProtectKernelTunables= */ | |
155 | { "/sys", READWRITE, false }, /* ProtectKernelTunables= */ | |
156 | { "/dev", READWRITE, false }, /* PrivateDevices= */ | |
157 | { "/home", READWRITE, true }, /* ProtectHome= */ | |
158 | { "/run/user", READWRITE, true }, /* ProtectHome= */ | |
159 | { "/root", READWRITE, true }, /* ProtectHome= */ | |
f471b2af DH |
160 | }; |
161 | ||
34de407a | 162 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
163 | assert(p); |
164 | ||
5327c910 LP |
165 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
166 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 167 | |
5327c910 | 168 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
169 | } |
170 | ||
34de407a | 171 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
172 | assert(p); |
173 | ||
174 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
175 | } | |
176 | ||
d2d6c096 LP |
177 | static const char *mount_entry_source(const MountEntry *p) { |
178 | assert(p); | |
179 | ||
180 | return p->source_malloc ?: p->source_const; | |
181 | } | |
182 | ||
1eb7e08e LP |
183 | static void mount_entry_done(MountEntry *p) { |
184 | assert(p); | |
185 | ||
186 | p->path_malloc = mfree(p->path_malloc); | |
187 | p->source_malloc = mfree(p->source_malloc); | |
188 | } | |
189 | ||
34de407a | 190 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode) { |
15ae422b LP |
191 | char **i; |
192 | ||
613b411c LP |
193 | assert(p); |
194 | ||
5327c910 LP |
195 | /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */ |
196 | ||
15ae422b | 197 | STRV_FOREACH(i, strv) { |
5327c910 LP |
198 | bool ignore = false, needs_prefix = false; |
199 | const char *e = *i; | |
15ae422b | 200 | |
5327c910 LP |
201 | /* Look for any prefixes */ |
202 | if (startswith(e, "-")) { | |
203 | e++; | |
9c94d52e | 204 | ignore = true; |
ea92ae33 | 205 | } |
5327c910 LP |
206 | if (startswith(e, "+")) { |
207 | e++; | |
208 | needs_prefix = true; | |
209 | } | |
ea92ae33 | 210 | |
5327c910 | 211 | if (!path_is_absolute(e)) |
15ae422b LP |
212 | return -EINVAL; |
213 | ||
34de407a | 214 | *((*p)++) = (MountEntry) { |
5327c910 LP |
215 | .path_const = e, |
216 | .mode = mode, | |
217 | .ignore = ignore, | |
218 | .has_prefix = !needs_prefix, | |
219 | }; | |
15ae422b LP |
220 | } |
221 | ||
222 | return 0; | |
223 | } | |
224 | ||
d2d6c096 LP |
225 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, unsigned n) { |
226 | unsigned i; | |
227 | ||
228 | assert(p); | |
229 | ||
230 | for (i = 0; i < n; i++) { | |
231 | const BindMount *b = binds + i; | |
232 | ||
233 | *((*p)++) = (MountEntry) { | |
234 | .path_const = b->destination, | |
235 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
236 | .read_only = b->read_only, | |
237 | .source_const = b->source, | |
238 | }; | |
239 | } | |
240 | ||
241 | return 0; | |
242 | } | |
243 | ||
34de407a | 244 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, unsigned n, bool ignore_protect) { |
f471b2af | 245 | unsigned i; |
11a30cec DH |
246 | |
247 | assert(p); | |
f471b2af | 248 | assert(mounts); |
11a30cec | 249 | |
5327c910 | 250 | /* Adds a list of static pre-defined entries */ |
f471b2af | 251 | |
5327c910 | 252 | for (i = 0; i < n; i++) |
34de407a LP |
253 | *((*p)++) = (MountEntry) { |
254 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
255 | .mode = mounts[i].mode, |
256 | .ignore = mounts[i].ignore || ignore_protect, | |
257 | }; | |
f471b2af DH |
258 | |
259 | return 0; | |
260 | } | |
261 | ||
34de407a | 262 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
263 | assert(p); |
264 | ||
5327c910 | 265 | switch (protect_home) { |
b6c432ca | 266 | |
5327c910 | 267 | case PROTECT_HOME_NO: |
b6c432ca DH |
268 | return 0; |
269 | ||
b6c432ca | 270 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
271 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
272 | ||
b6c432ca | 273 | case PROTECT_HOME_YES: |
5327c910 LP |
274 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
275 | ||
b6c432ca | 276 | default: |
5327c910 | 277 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 278 | } |
b6c432ca DH |
279 | } |
280 | ||
34de407a | 281 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
282 | assert(p); |
283 | ||
5327c910 LP |
284 | switch (protect_system) { |
285 | ||
286 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
287 | return 0; |
288 | ||
f471b2af | 289 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
290 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
291 | ||
f471b2af | 292 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
293 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
294 | ||
f471b2af | 295 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
296 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
297 | ||
f471b2af | 298 | default: |
5327c910 | 299 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 300 | } |
11a30cec DH |
301 | } |
302 | ||
c17ec25e | 303 | static int mount_path_compare(const void *a, const void *b) { |
34de407a | 304 | const MountEntry *p = a, *q = b; |
a0827e2b | 305 | int d; |
15ae422b | 306 | |
6ee1a919 | 307 | /* If the paths are not equal, then order prefixes first */ |
34de407a | 308 | d = path_compare(mount_entry_path(p), mount_entry_path(q)); |
6ee1a919 LP |
309 | if (d != 0) |
310 | return d; | |
15ae422b | 311 | |
6ee1a919 LP |
312 | /* If the paths are equal, check the mode */ |
313 | if (p->mode < q->mode) | |
314 | return -1; | |
15ae422b | 315 | |
6ee1a919 LP |
316 | if (p->mode > q->mode) |
317 | return 1; | |
15ae422b | 318 | |
6ee1a919 | 319 | return 0; |
15ae422b LP |
320 | } |
321 | ||
34de407a | 322 | static int prefix_where_needed(MountEntry *m, unsigned n, const char *root_directory) { |
5327c910 LP |
323 | unsigned i; |
324 | ||
325 | /* Prefixes all paths in the bind mount table with the root directory if it is specified and the entry needs | |
326 | * that. */ | |
327 | ||
328 | if (!root_directory) | |
329 | return 0; | |
330 | ||
331 | for (i = 0; i < n; i++) { | |
332 | char *s; | |
333 | ||
334 | if (m[i].has_prefix) | |
335 | continue; | |
336 | ||
34de407a | 337 | s = prefix_root(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
338 | if (!s) |
339 | return -ENOMEM; | |
340 | ||
341 | free(m[i].path_malloc); | |
342 | m[i].path_malloc = s; | |
343 | ||
344 | m[i].has_prefix = true; | |
345 | } | |
346 | ||
347 | return 0; | |
348 | } | |
349 | ||
34de407a LP |
350 | static void drop_duplicates(MountEntry *m, unsigned *n) { |
351 | MountEntry *f, *t, *previous; | |
15ae422b | 352 | |
c17ec25e | 353 | assert(m); |
15ae422b | 354 | assert(n); |
15ae422b | 355 | |
fe3c2583 LP |
356 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
357 | ||
1d54cd5d | 358 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 359 | |
fe3c2583 LP |
360 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
361 | * above. */ | |
34de407a LP |
362 | if (previous && path_equal(mount_entry_path(f), mount_entry_path(previous))) { |
363 | log_debug("%s is duplicate.", mount_entry_path(f)); | |
364 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ | |
1eb7e08e | 365 | mount_entry_done(f); |
15ae422b | 366 | continue; |
fe3c2583 | 367 | } |
15ae422b | 368 | |
e2d7c1a0 | 369 | *t = *f; |
15ae422b | 370 | previous = t; |
fe3c2583 LP |
371 | t++; |
372 | } | |
373 | ||
374 | *n = t - m; | |
375 | } | |
376 | ||
34de407a LP |
377 | static void drop_inaccessible(MountEntry *m, unsigned *n) { |
378 | MountEntry *f, *t; | |
fe3c2583 LP |
379 | const char *clear = NULL; |
380 | ||
381 | assert(m); | |
382 | assert(n); | |
383 | ||
384 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
385 | * ordered already. */ | |
386 | ||
1d54cd5d | 387 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
388 | |
389 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
390 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
391 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
392 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 393 | mount_entry_done(f); |
fe3c2583 LP |
394 | continue; |
395 | } | |
15ae422b | 396 | |
34de407a | 397 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
398 | |
399 | *t = *f; | |
15ae422b LP |
400 | t++; |
401 | } | |
402 | ||
c17ec25e | 403 | *n = t - m; |
15ae422b LP |
404 | } |
405 | ||
34de407a LP |
406 | static void drop_nop(MountEntry *m, unsigned *n) { |
407 | MountEntry *f, *t; | |
7648a565 LP |
408 | |
409 | assert(m); | |
410 | assert(n); | |
411 | ||
412 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
413 | * list is ordered by prefixes. */ | |
414 | ||
1d54cd5d | 415 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 LP |
416 | |
417 | /* Only suppress such subtrees for READONLY and READWRITE entries */ | |
418 | if (IN_SET(f->mode, READONLY, READWRITE)) { | |
34de407a | 419 | MountEntry *p; |
7648a565 LP |
420 | bool found = false; |
421 | ||
422 | /* Now let's find the first parent of the entry we are looking at. */ | |
423 | for (p = t-1; p >= m; p--) { | |
34de407a | 424 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
425 | found = true; |
426 | break; | |
427 | } | |
428 | } | |
429 | ||
430 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
431 | if (found && p->mode == f->mode) { | |
34de407a | 432 | log_debug("%s is redundant by %s", mount_entry_path(f), mount_entry_path(p)); |
1eb7e08e | 433 | mount_entry_done(f); |
7648a565 LP |
434 | continue; |
435 | } | |
436 | } | |
437 | ||
438 | *t = *f; | |
439 | t++; | |
440 | } | |
441 | ||
442 | *n = t - m; | |
443 | } | |
444 | ||
34de407a LP |
445 | static void drop_outside_root(const char *root_directory, MountEntry *m, unsigned *n) { |
446 | MountEntry *f, *t; | |
cd2902c9 LP |
447 | |
448 | assert(m); | |
449 | assert(n); | |
450 | ||
1d54cd5d | 451 | /* Nothing to do */ |
cd2902c9 LP |
452 | if (!root_directory) |
453 | return; | |
454 | ||
455 | /* Drops all mounts that are outside of the root directory. */ | |
456 | ||
1d54cd5d | 457 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 458 | |
34de407a LP |
459 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
460 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 461 | mount_entry_done(f); |
cd2902c9 LP |
462 | continue; |
463 | } | |
464 | ||
465 | *t = *f; | |
466 | t++; | |
467 | } | |
468 | ||
469 | *n = t - m; | |
470 | } | |
471 | ||
5d997827 | 472 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
473 | static const char devnodes[] = |
474 | "/dev/null\0" | |
475 | "/dev/zero\0" | |
476 | "/dev/full\0" | |
477 | "/dev/random\0" | |
478 | "/dev/urandom\0" | |
479 | "/dev/tty\0"; | |
480 | ||
2b85f4e1 | 481 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 482 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
7f112f50 LP |
483 | _cleanup_umask_ mode_t u; |
484 | int r; | |
485 | ||
486 | assert(m); | |
487 | ||
488 | u = umask(0000); | |
489 | ||
2b85f4e1 LP |
490 | if (!mkdtemp(temporary_mount)) |
491 | return -errno; | |
492 | ||
63c372cb | 493 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 494 | (void) mkdir(dev, 0755); |
737ba3c8 | 495 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2b85f4e1 LP |
496 | r = -errno; |
497 | goto fail; | |
498 | } | |
499 | ||
63c372cb | 500 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 501 | (void) mkdir(devpts, 0755); |
2b85f4e1 LP |
502 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
503 | r = -errno; | |
504 | goto fail; | |
505 | } | |
506 | ||
63c372cb | 507 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
3164e3cb ZJS |
508 | if (symlink("pts/ptmx", devptmx) < 0) { |
509 | r = -errno; | |
510 | goto fail; | |
511 | } | |
e06b6479 | 512 | |
63c372cb | 513 | devshm = strjoina(temporary_mount, "/dev/shm"); |
dc751688 | 514 | (void) mkdir(devshm, 01777); |
2b85f4e1 LP |
515 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
516 | if (r < 0) { | |
517 | r = -errno; | |
518 | goto fail; | |
519 | } | |
520 | ||
63c372cb | 521 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 522 | (void) mkdir(devmqueue, 0755); |
3164e3cb | 523 | (void) mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 524 | |
63c372cb | 525 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 526 | (void) mkdir(devhugepages, 0755); |
3164e3cb | 527 | (void) mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 528 | |
63c372cb | 529 | devlog = strjoina(temporary_mount, "/dev/log"); |
3164e3cb | 530 | (void) symlink("/run/systemd/journal/dev-log", devlog); |
82d25240 | 531 | |
7f112f50 | 532 | NULSTR_FOREACH(d, devnodes) { |
2b85f4e1 LP |
533 | _cleanup_free_ char *dn = NULL; |
534 | struct stat st; | |
535 | ||
536 | r = stat(d, &st); | |
7f112f50 | 537 | if (r < 0) { |
2b85f4e1 LP |
538 | |
539 | if (errno == ENOENT) | |
540 | continue; | |
541 | ||
542 | r = -errno; | |
543 | goto fail; | |
7f112f50 LP |
544 | } |
545 | ||
2b85f4e1 LP |
546 | if (!S_ISBLK(st.st_mode) && |
547 | !S_ISCHR(st.st_mode)) { | |
548 | r = -EINVAL; | |
549 | goto fail; | |
550 | } | |
551 | ||
552 | if (st.st_rdev == 0) | |
553 | continue; | |
554 | ||
555 | dn = strappend(temporary_mount, d); | |
556 | if (!dn) { | |
557 | r = -ENOMEM; | |
558 | goto fail; | |
559 | } | |
560 | ||
ecabcf8b | 561 | mac_selinux_create_file_prepare(d, st.st_mode); |
2b85f4e1 | 562 | r = mknod(dn, st.st_mode, st.st_rdev); |
ecabcf8b | 563 | mac_selinux_create_file_clear(); |
dd078a1e | 564 | |
2b85f4e1 LP |
565 | if (r < 0) { |
566 | r = -errno; | |
567 | goto fail; | |
568 | } | |
7f112f50 LP |
569 | } |
570 | ||
03cfe0d5 | 571 | dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
7f112f50 | 572 | |
ee818b89 AC |
573 | /* Create the /dev directory if missing. It is more likely to be |
574 | * missing when the service is started with RootDirectory. This is | |
575 | * consistent with mount units creating the mount points when missing. | |
576 | */ | |
34de407a | 577 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 578 | |
9e5f8252 | 579 | /* Unmount everything in old /dev */ |
34de407a LP |
580 | umount_recursive(mount_entry_path(m), 0); |
581 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { | |
2b85f4e1 LP |
582 | r = -errno; |
583 | goto fail; | |
584 | } | |
7f112f50 | 585 | |
2b85f4e1 LP |
586 | rmdir(dev); |
587 | rmdir(temporary_mount); | |
7f112f50 | 588 | |
2b85f4e1 | 589 | return 0; |
7f112f50 | 590 | |
2b85f4e1 LP |
591 | fail: |
592 | if (devpts) | |
593 | umount(devpts); | |
7f112f50 | 594 | |
2b85f4e1 LP |
595 | if (devshm) |
596 | umount(devshm); | |
7f112f50 | 597 | |
2b85f4e1 LP |
598 | if (devhugepages) |
599 | umount(devhugepages); | |
7f112f50 | 600 | |
2b85f4e1 LP |
601 | if (devmqueue) |
602 | umount(devmqueue); | |
7f112f50 | 603 | |
d267c5aa ZJS |
604 | umount(dev); |
605 | rmdir(dev); | |
2b85f4e1 | 606 | rmdir(temporary_mount); |
7f112f50 | 607 | |
2b85f4e1 | 608 | return r; |
7f112f50 LP |
609 | } |
610 | ||
5d997827 LP |
611 | static int mount_bind_dev(MountEntry *m) { |
612 | int r; | |
613 | ||
614 | assert(m); | |
615 | ||
616 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
617 | * /dev. This is only used when RootDirectory= is set. */ | |
618 | ||
619 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); | |
620 | if (r < 0) | |
621 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
622 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
623 | return 0; | |
624 | ||
625 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
626 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
627 | ||
628 | return 1; | |
629 | } | |
630 | ||
631 | static int mount_sysfs(MountEntry *m) { | |
632 | int r; | |
633 | ||
634 | assert(m); | |
635 | ||
636 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); | |
637 | if (r < 0) | |
638 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
639 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
640 | return 0; | |
641 | ||
642 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
643 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
644 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
645 | ||
646 | return 1; | |
647 | } | |
648 | ||
649 | static int mount_procfs(MountEntry *m) { | |
650 | int r; | |
651 | ||
652 | assert(m); | |
653 | ||
654 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); | |
655 | if (r < 0) | |
656 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
657 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
658 | return 0; | |
659 | ||
660 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
661 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
662 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
663 | ||
664 | return 1; | |
665 | } | |
666 | ||
d2d6c096 LP |
667 | static int mount_entry_chase( |
668 | const char *root_directory, | |
669 | MountEntry *m, | |
670 | const char *path, | |
671 | char **location) { | |
672 | ||
8fceda93 LP |
673 | char *chased; |
674 | int r; | |
675 | ||
676 | assert(m); | |
677 | ||
678 | /* Since mount() will always follow symlinks and we need to take the different root directory into account we | |
d2d6c096 LP |
679 | * chase the symlinks on our own first. This is called for the destination path, as well as the source path (if |
680 | * that applies). The result is stored in "location". */ | |
8fceda93 | 681 | |
d2d6c096 | 682 | r = chase_symlinks(path, root_directory, 0, &chased); |
8fceda93 | 683 | if (r == -ENOENT && m->ignore) { |
d2d6c096 | 684 | log_debug_errno(r, "Path %s does not exist, ignoring.", path); |
8fceda93 LP |
685 | return 0; |
686 | } | |
687 | if (r < 0) | |
d2d6c096 | 688 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", path); |
8fceda93 | 689 | |
d2d6c096 | 690 | log_debug("Followed symlinks %s → %s.", path, chased); |
8fceda93 | 691 | |
d2d6c096 LP |
692 | free(*location); |
693 | *location = chased; | |
8fceda93 LP |
694 | |
695 | return 1; | |
696 | } | |
697 | ||
ac0930c8 | 698 | static int apply_mount( |
8fceda93 | 699 | const char *root_directory, |
34de407a | 700 | MountEntry *m, |
ac0930c8 | 701 | const char *tmp_dir, |
c17ec25e | 702 | const char *var_tmp_dir) { |
ac0930c8 | 703 | |
15ae422b | 704 | const char *what; |
d2d6c096 | 705 | bool rbind = true; |
15ae422b | 706 | int r; |
15ae422b | 707 | |
c17ec25e | 708 | assert(m); |
15ae422b | 709 | |
d2d6c096 | 710 | r = mount_entry_chase(root_directory, m, mount_entry_path(m), &m->path_malloc); |
8fceda93 LP |
711 | if (r <= 0) |
712 | return r; | |
713 | ||
34de407a | 714 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 715 | |
c17ec25e | 716 | switch (m->mode) { |
15ae422b | 717 | |
160cfdbe LP |
718 | case INACCESSIBLE: { |
719 | struct stat target; | |
6d313367 LP |
720 | |
721 | /* First, get rid of everything that is below if there | |
722 | * is anything... Then, overmount it with an | |
c4b41707 | 723 | * inaccessible path. */ |
34de407a | 724 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 725 | |
34de407a LP |
726 | if (lstat(mount_entry_path(m), &target) < 0) |
727 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); | |
15ae422b | 728 | |
c4b41707 | 729 | what = mode_to_inaccessible_node(target.st_mode); |
5fd7cf6f LP |
730 | if (!what) { |
731 | log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 AP |
732 | return -ELOOP; |
733 | } | |
734 | break; | |
160cfdbe | 735 | } |
fe3c2583 | 736 | |
15ae422b | 737 | case READONLY: |
15ae422b | 738 | case READWRITE: |
8fceda93 | 739 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
d944dc95 | 740 | if (r < 0) |
34de407a | 741 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
742 | 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. */ |
743 | return 0; | |
6b7c9f8b | 744 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 745 | what = mount_entry_path(m); |
6b7c9f8b | 746 | break; |
15ae422b | 747 | |
d2d6c096 LP |
748 | case BIND_MOUNT: |
749 | rbind = false; | |
750 | /* fallthrough */ | |
751 | ||
752 | case BIND_MOUNT_RECURSIVE: | |
753 | /* Also chase the source mount */ | |
5d997827 | 754 | |
d2d6c096 LP |
755 | r = mount_entry_chase(root_directory, m, mount_entry_source(m), &m->source_malloc); |
756 | if (r <= 0) | |
757 | return r; | |
758 | ||
759 | what = mount_entry_source(m); | |
760 | break; | |
761 | ||
ac0930c8 LP |
762 | case PRIVATE_TMP: |
763 | what = tmp_dir; | |
764 | break; | |
765 | ||
766 | case PRIVATE_VAR_TMP: | |
767 | what = var_tmp_dir; | |
15ae422b | 768 | break; |
e364ad06 | 769 | |
d6797c92 | 770 | case PRIVATE_DEV: |
5d997827 LP |
771 | return mount_private_dev(m); |
772 | ||
773 | case BIND_DEV: | |
774 | return mount_bind_dev(m); | |
775 | ||
776 | case SYSFS: | |
777 | return mount_sysfs(m); | |
778 | ||
779 | case PROCFS: | |
780 | return mount_procfs(m); | |
d6797c92 | 781 | |
e364ad06 LP |
782 | default: |
783 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
784 | } |
785 | ||
ac0930c8 | 786 | assert(what); |
15ae422b | 787 | |
d2d6c096 | 788 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) |
34de407a | 789 | return log_debug_errno(errno, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); |
6b7c9f8b | 790 | |
34de407a | 791 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 792 | return 0; |
ac0930c8 | 793 | } |
15ae422b | 794 | |
34de407a | 795 | static int make_read_only(MountEntry *m, char **blacklist) { |
6b7c9f8b | 796 | int r = 0; |
15ae422b | 797 | |
c17ec25e | 798 | assert(m); |
ac0930c8 | 799 | |
34de407a LP |
800 | if (mount_entry_read_only(m)) |
801 | r = bind_remount_recursive(mount_entry_path(m), true, blacklist); | |
13e785f7 | 802 | else if (m->mode == PRIVATE_DEV) { /* Superblock can be readonly but the submounts can't */ |
34de407a | 803 | if (mount(NULL, mount_entry_path(m), NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0) |
6b7c9f8b | 804 | r = -errno; |
737ba3c8 | 805 | } else |
6b7c9f8b LP |
806 | return 0; |
807 | ||
808 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
809 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
810 | * read-only mounts already applied. */ | |
ac0930c8 | 811 | |
8fceda93 LP |
812 | if (r == -ENOENT && m->ignore) |
813 | r = 0; | |
5327c910 | 814 | |
1d54cd5d | 815 | return r; |
d944dc95 LP |
816 | } |
817 | ||
5d997827 LP |
818 | static bool namespace_info_mount_apivfs(const NameSpaceInfo *ns_info) { |
819 | assert(ns_info); | |
820 | ||
821 | /* ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, since to protect the API VFS mounts, | |
822 | * they need to be around in the first place... */ | |
823 | ||
824 | return ns_info->mount_apivfs || | |
825 | ns_info->protect_control_groups || | |
826 | ns_info->protect_kernel_tunables; | |
827 | } | |
828 | ||
2652c6c1 | 829 | static unsigned namespace_calculate_mounts( |
c575770b | 830 | const NameSpaceInfo *ns_info, |
2652c6c1 DH |
831 | char** read_write_paths, |
832 | char** read_only_paths, | |
833 | char** inaccessible_paths, | |
d2d6c096 LP |
834 | const BindMount *bind_mounts, |
835 | unsigned n_bind_mounts, | |
2652c6c1 DH |
836 | const char* tmp_dir, |
837 | const char* var_tmp_dir, | |
2652c6c1 DH |
838 | ProtectHome protect_home, |
839 | ProtectSystem protect_system) { | |
840 | ||
b6c432ca | 841 | unsigned protect_home_cnt; |
f471b2af DH |
842 | unsigned protect_system_cnt = |
843 | (protect_system == PROTECT_SYSTEM_STRICT ? | |
844 | ELEMENTSOF(protect_system_strict_table) : | |
845 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
846 | ELEMENTSOF(protect_system_full_table) : | |
847 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
848 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
849 | ||
b6c432ca DH |
850 | protect_home_cnt = |
851 | (protect_home == PROTECT_HOME_YES ? | |
852 | ELEMENTSOF(protect_home_yes_table) : | |
853 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
854 | ELEMENTSOF(protect_home_read_only_table) : 0)); | |
855 | ||
2652c6c1 DH |
856 | return !!tmp_dir + !!var_tmp_dir + |
857 | strv_length(read_write_paths) + | |
858 | strv_length(read_only_paths) + | |
859 | strv_length(inaccessible_paths) + | |
d2d6c096 | 860 | n_bind_mounts + |
c575770b DH |
861 | ns_info->private_dev + |
862 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
863 | (ns_info->protect_control_groups ? 1 : 0) + | |
864 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + | |
5d997827 LP |
865 | protect_home_cnt + protect_system_cnt + |
866 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0); | |
2652c6c1 DH |
867 | } |
868 | ||
613b411c | 869 | int setup_namespace( |
ee818b89 | 870 | const char* root_directory, |
915e6d16 | 871 | const char* root_image, |
c575770b | 872 | const NameSpaceInfo *ns_info, |
2a624c36 AP |
873 | char** read_write_paths, |
874 | char** read_only_paths, | |
875 | char** inaccessible_paths, | |
d2d6c096 LP |
876 | const BindMount *bind_mounts, |
877 | unsigned n_bind_mounts, | |
a004cb4c LP |
878 | const char* tmp_dir, |
879 | const char* var_tmp_dir, | |
1b8689f9 LP |
880 | ProtectHome protect_home, |
881 | ProtectSystem protect_system, | |
915e6d16 LP |
882 | unsigned long mount_flags, |
883 | DissectImageFlags dissect_image_flags) { | |
15ae422b | 884 | |
915e6d16 | 885 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 886 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 887 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 888 | _cleanup_free_ void *root_hash = NULL; |
34de407a | 889 | MountEntry *m, *mounts = NULL; |
78ebe980 | 890 | size_t root_hash_size = 0; |
d944dc95 | 891 | bool make_slave = false; |
f0a4feb0 | 892 | unsigned n_mounts; |
c17ec25e | 893 | int r = 0; |
15ae422b | 894 | |
915e6d16 LP |
895 | assert(ns_info); |
896 | ||
613b411c | 897 | if (mount_flags == 0) |
c17ec25e | 898 | mount_flags = MS_SHARED; |
ac0930c8 | 899 | |
915e6d16 LP |
900 | if (root_image) { |
901 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
902 | ||
903 | if (protect_system == PROTECT_SYSTEM_STRICT && strv_isempty(read_write_paths)) | |
904 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; | |
905 | ||
906 | r = loop_device_make_by_path(root_image, | |
907 | dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR, | |
908 | &loop_device); | |
909 | if (r < 0) | |
910 | return r; | |
911 | ||
78ebe980 LP |
912 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
913 | if (r < 0) | |
914 | return r; | |
915 | ||
916 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
917 | if (r < 0) | |
918 | return r; | |
919 | ||
920 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 LP |
921 | if (r < 0) |
922 | return r; | |
923 | ||
924 | if (!root_directory) { | |
925 | /* Create a mount point for the image, if it's still missing. We use the same mount point for | |
926 | * all images, which is safe, since they all live in their own namespaces after all, and hence | |
927 | * won't see each other. */ | |
928 | root_directory = "/run/systemd/unit-root"; | |
929 | (void) mkdir(root_directory, 0700); | |
930 | } | |
931 | } | |
932 | ||
cfbeb4ef LP |
933 | n_mounts = namespace_calculate_mounts( |
934 | ns_info, | |
935 | read_write_paths, | |
936 | read_only_paths, | |
937 | inaccessible_paths, | |
d2d6c096 | 938 | bind_mounts, n_bind_mounts, |
cfbeb4ef LP |
939 | tmp_dir, var_tmp_dir, |
940 | protect_home, protect_system); | |
613b411c | 941 | |
2652c6c1 | 942 | /* Set mount slave mode */ |
f0a4feb0 | 943 | if (root_directory || n_mounts > 0) |
d944dc95 LP |
944 | make_slave = true; |
945 | ||
f0a4feb0 | 946 | if (n_mounts > 0) { |
34de407a | 947 | m = mounts = (MountEntry *) alloca0(n_mounts * sizeof(MountEntry)); |
5327c910 | 948 | r = append_access_mounts(&m, read_write_paths, READWRITE); |
613b411c | 949 | if (r < 0) |
f0a4feb0 | 950 | goto finish; |
613b411c | 951 | |
5327c910 | 952 | r = append_access_mounts(&m, read_only_paths, READONLY); |
613b411c | 953 | if (r < 0) |
f0a4feb0 | 954 | goto finish; |
613b411c | 955 | |
5327c910 | 956 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE); |
613b411c | 957 | if (r < 0) |
f0a4feb0 | 958 | goto finish; |
7ff7394d | 959 | |
d2d6c096 LP |
960 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
961 | if (r < 0) | |
962 | goto finish; | |
963 | ||
613b411c | 964 | if (tmp_dir) { |
34de407a | 965 | *(m++) = (MountEntry) { |
5327c910 LP |
966 | .path_const = "/tmp", |
967 | .mode = PRIVATE_TMP, | |
968 | }; | |
613b411c | 969 | } |
7ff7394d | 970 | |
613b411c | 971 | if (var_tmp_dir) { |
34de407a | 972 | *(m++) = (MountEntry) { |
5327c910 LP |
973 | .path_const = "/var/tmp", |
974 | .mode = PRIVATE_VAR_TMP, | |
975 | }; | |
7ff7394d | 976 | } |
ac0930c8 | 977 | |
c575770b | 978 | if (ns_info->private_dev) { |
34de407a | 979 | *(m++) = (MountEntry) { |
5327c910 LP |
980 | .path_const = "/dev", |
981 | .mode = PRIVATE_DEV, | |
982 | }; | |
7f112f50 LP |
983 | } |
984 | ||
c575770b | 985 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 986 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 987 | if (r < 0) |
f0a4feb0 | 988 | goto finish; |
c575770b DH |
989 | } |
990 | ||
991 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 992 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 993 | if (r < 0) |
f0a4feb0 | 994 | goto finish; |
c575770b | 995 | } |
59eeb84b | 996 | |
c575770b | 997 | if (ns_info->protect_control_groups) { |
34de407a | 998 | *(m++) = (MountEntry) { |
5327c910 LP |
999 | .path_const = "/sys/fs/cgroup", |
1000 | .mode = READONLY, | |
1001 | }; | |
59eeb84b LP |
1002 | } |
1003 | ||
5327c910 | 1004 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1005 | if (r < 0) |
f0a4feb0 | 1006 | goto finish; |
417116f2 | 1007 | |
5327c910 | 1008 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1009 | if (r < 0) |
f0a4feb0 | 1010 | goto finish; |
417116f2 | 1011 | |
5d997827 LP |
1012 | if (namespace_info_mount_apivfs(ns_info)) { |
1013 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); | |
1014 | if (r < 0) | |
1015 | goto finish; | |
1016 | } | |
1017 | ||
f0a4feb0 | 1018 | assert(mounts + n_mounts == m); |
ac0930c8 | 1019 | |
5327c910 LP |
1020 | /* Prepend the root directory where that's necessary */ |
1021 | r = prefix_where_needed(mounts, n_mounts, root_directory); | |
1022 | if (r < 0) | |
1023 | goto finish; | |
1024 | ||
34de407a | 1025 | qsort(mounts, n_mounts, sizeof(MountEntry), mount_path_compare); |
fe3c2583 | 1026 | |
f0a4feb0 DH |
1027 | drop_duplicates(mounts, &n_mounts); |
1028 | drop_outside_root(root_directory, mounts, &n_mounts); | |
1029 | drop_inaccessible(mounts, &n_mounts); | |
1030 | drop_nop(mounts, &n_mounts); | |
15ae422b LP |
1031 | } |
1032 | ||
d944dc95 LP |
1033 | if (unshare(CLONE_NEWNS) < 0) { |
1034 | r = -errno; | |
1035 | goto finish; | |
1036 | } | |
1e4e94c8 | 1037 | |
d944dc95 | 1038 | if (make_slave) { |
c2c13f2d LP |
1039 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1040 | shows up in the parent */ | |
d944dc95 LP |
1041 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { |
1042 | r = -errno; | |
1043 | goto finish; | |
1044 | } | |
ee818b89 AC |
1045 | } |
1046 | ||
915e6d16 LP |
1047 | if (root_image) { |
1048 | r = dissected_image_mount(dissected_image, root_directory, dissect_image_flags); | |
1049 | if (r < 0) | |
1050 | goto finish; | |
1051 | ||
78ebe980 LP |
1052 | r = decrypted_image_relinquish(decrypted_image); |
1053 | if (r < 0) | |
1054 | goto finish; | |
1055 | ||
915e6d16 LP |
1056 | loop_device_relinquish(loop_device); |
1057 | ||
1058 | } else if (root_directory) { | |
1059 | ||
8f1ad200 | 1060 | /* Turn directory into bind mount, if it isn't one yet */ |
e1873695 | 1061 | r = path_is_mount_point(root_directory, NULL, AT_SYMLINK_FOLLOW); |
8f1ad200 | 1062 | if (r < 0) |
d944dc95 | 1063 | goto finish; |
8f1ad200 LP |
1064 | if (r == 0) { |
1065 | if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
1066 | r = -errno; | |
1067 | goto finish; | |
1068 | } | |
d944dc95 | 1069 | } |
ee818b89 | 1070 | } |
c2c13f2d | 1071 | |
f0a4feb0 | 1072 | if (n_mounts > 0) { |
6b7c9f8b LP |
1073 | char **blacklist; |
1074 | unsigned j; | |
1075 | ||
1076 | /* First round, add in all special mounts we need */ | |
f0a4feb0 | 1077 | for (m = mounts; m < mounts + n_mounts; ++m) { |
8fceda93 | 1078 | r = apply_mount(root_directory, m, tmp_dir, var_tmp_dir); |
c2c13f2d | 1079 | if (r < 0) |
d944dc95 | 1080 | goto finish; |
c2c13f2d | 1081 | } |
15ae422b | 1082 | |
6b7c9f8b | 1083 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
f0a4feb0 DH |
1084 | blacklist = newa(char*, n_mounts+1); |
1085 | for (j = 0; j < n_mounts; j++) | |
34de407a | 1086 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1087 | blacklist[j] = NULL; |
1088 | ||
1089 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1090 | for (m = mounts; m < mounts + n_mounts; ++m) { |
6b7c9f8b | 1091 | r = make_read_only(m, blacklist); |
c2c13f2d | 1092 | if (r < 0) |
d944dc95 | 1093 | goto finish; |
c2c13f2d | 1094 | } |
15ae422b LP |
1095 | } |
1096 | ||
ee818b89 AC |
1097 | if (root_directory) { |
1098 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ | |
1099 | r = mount_move_root(root_directory); | |
d944dc95 LP |
1100 | if (r < 0) |
1101 | goto finish; | |
ee818b89 AC |
1102 | } |
1103 | ||
c2c13f2d LP |
1104 | /* Remount / as the desired mode. Not that this will not |
1105 | * reestablish propagation from our side to the host, since | |
1106 | * what's disconnected is disconnected. */ | |
d944dc95 LP |
1107 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
1108 | r = -errno; | |
1109 | goto finish; | |
1110 | } | |
15ae422b | 1111 | |
d944dc95 | 1112 | r = 0; |
15ae422b | 1113 | |
d944dc95 | 1114 | finish: |
f0a4feb0 | 1115 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1116 | mount_entry_done(m); |
613b411c LP |
1117 | |
1118 | return r; | |
1119 | } | |
1120 | ||
d2d6c096 LP |
1121 | void bind_mount_free_many(BindMount *b, unsigned n) { |
1122 | unsigned i; | |
1123 | ||
1124 | assert(b || n == 0); | |
1125 | ||
1126 | for (i = 0; i < n; i++) { | |
1127 | free(b[i].source); | |
1128 | free(b[i].destination); | |
1129 | } | |
1130 | ||
1131 | free(b); | |
1132 | } | |
1133 | ||
1134 | int bind_mount_add(BindMount **b, unsigned *n, const BindMount *item) { | |
1135 | _cleanup_free_ char *s = NULL, *d = NULL; | |
1136 | BindMount *c; | |
1137 | ||
1138 | assert(b); | |
1139 | assert(n); | |
1140 | assert(item); | |
1141 | ||
1142 | s = strdup(item->source); | |
1143 | if (!s) | |
1144 | return -ENOMEM; | |
1145 | ||
1146 | d = strdup(item->destination); | |
1147 | if (!d) | |
1148 | return -ENOMEM; | |
1149 | ||
1150 | c = realloc_multiply(*b, sizeof(BindMount), *n + 1); | |
1151 | if (!c) | |
1152 | return -ENOMEM; | |
1153 | ||
1154 | *b = c; | |
1155 | ||
1156 | c[(*n) ++] = (BindMount) { | |
1157 | .source = s, | |
1158 | .destination = d, | |
1159 | .read_only = item->read_only, | |
1160 | .recursive = item->recursive, | |
1161 | .ignore_enoent = item->ignore_enoent, | |
1162 | }; | |
1163 | ||
1164 | s = d = NULL; | |
1165 | return 0; | |
1166 | } | |
1167 | ||
613b411c LP |
1168 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1169 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1170 | char bid[SD_ID128_STRING_MAX]; |
1171 | sd_id128_t boot_id; | |
1172 | int r; | |
613b411c LP |
1173 | |
1174 | assert(id); | |
1175 | assert(prefix); | |
1176 | assert(path); | |
1177 | ||
6b46ea73 LP |
1178 | /* We include the boot id in the directory so that after a |
1179 | * reboot we can easily identify obsolete directories. */ | |
1180 | ||
1181 | r = sd_id128_get_boot(&boot_id); | |
1182 | if (r < 0) | |
1183 | return r; | |
1184 | ||
605405c6 | 1185 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1186 | if (!x) |
1187 | return -ENOMEM; | |
1188 | ||
1189 | RUN_WITH_UMASK(0077) | |
1190 | if (!mkdtemp(x)) | |
1191 | return -errno; | |
1192 | ||
1193 | RUN_WITH_UMASK(0000) { | |
1194 | char *y; | |
1195 | ||
63c372cb | 1196 | y = strjoina(x, "/tmp"); |
613b411c LP |
1197 | |
1198 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1199 | return -errno; | |
c17ec25e | 1200 | } |
15ae422b | 1201 | |
613b411c LP |
1202 | *path = x; |
1203 | x = NULL; | |
1204 | ||
1205 | return 0; | |
1206 | } | |
1207 | ||
1208 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1209 | char *a, *b; | |
1210 | int r; | |
1211 | ||
1212 | assert(id); | |
1213 | assert(tmp_dir); | |
1214 | assert(var_tmp_dir); | |
1215 | ||
1216 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1217 | if (r < 0) | |
1218 | return r; | |
1219 | ||
1220 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1221 | if (r < 0) { | |
1222 | char *t; | |
1223 | ||
63c372cb | 1224 | t = strjoina(a, "/tmp"); |
613b411c LP |
1225 | rmdir(t); |
1226 | rmdir(a); | |
1227 | ||
1228 | free(a); | |
1229 | return r; | |
1230 | } | |
1231 | ||
1232 | *tmp_dir = a; | |
1233 | *var_tmp_dir = b; | |
1234 | ||
1235 | return 0; | |
1236 | } | |
1237 | ||
1238 | int setup_netns(int netns_storage_socket[2]) { | |
1239 | _cleanup_close_ int netns = -1; | |
3ee897d6 | 1240 | int r, q; |
613b411c LP |
1241 | |
1242 | assert(netns_storage_socket); | |
1243 | assert(netns_storage_socket[0] >= 0); | |
1244 | assert(netns_storage_socket[1] >= 0); | |
1245 | ||
1246 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1247 | * namespace reference fd. Whatever process runs this first |
1248 | * shall create a new namespace, all others should just join | |
1249 | * it. To serialize that we use a file lock on the socket | |
1250 | * pair. | |
613b411c LP |
1251 | * |
1252 | * It's a bit crazy, but hey, works great! */ | |
1253 | ||
1254 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1255 | return -errno; | |
1256 | ||
3ee897d6 LP |
1257 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1258 | if (netns == -EAGAIN) { | |
613b411c LP |
1259 | /* Nothing stored yet, so let's create a new namespace */ |
1260 | ||
1261 | if (unshare(CLONE_NEWNET) < 0) { | |
1262 | r = -errno; | |
1263 | goto fail; | |
1264 | } | |
1265 | ||
1266 | loopback_setup(); | |
1267 | ||
1268 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1269 | if (netns < 0) { | |
1270 | r = -errno; | |
1271 | goto fail; | |
1272 | } | |
1273 | ||
1274 | r = 1; | |
613b411c | 1275 | |
3ee897d6 LP |
1276 | } else if (netns < 0) { |
1277 | r = netns; | |
1278 | goto fail; | |
613b411c | 1279 | |
3ee897d6 LP |
1280 | } else { |
1281 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1282 | if (setns(netns, CLONE_NEWNET) < 0) { |
1283 | r = -errno; | |
1284 | goto fail; | |
1285 | } | |
1286 | ||
1287 | r = 0; | |
1288 | } | |
1289 | ||
3ee897d6 LP |
1290 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1291 | if (q < 0) { | |
1292 | r = q; | |
613b411c LP |
1293 | goto fail; |
1294 | } | |
1295 | ||
1296 | fail: | |
fe048ce5 | 1297 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1298 | return r; |
1299 | } | |
417116f2 | 1300 | |
1b8689f9 LP |
1301 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1302 | [PROTECT_HOME_NO] = "no", | |
1303 | [PROTECT_HOME_YES] = "yes", | |
1304 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
417116f2 LP |
1305 | }; |
1306 | ||
1b8689f9 LP |
1307 | DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome); |
1308 | ||
1309 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { | |
1310 | [PROTECT_SYSTEM_NO] = "no", | |
1311 | [PROTECT_SYSTEM_YES] = "yes", | |
1312 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1313 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1314 | }; |
1315 | ||
1316 | DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem); |