]>
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" |
10404d52 | 30 | #include "base-filesystem.h" |
7f112f50 | 31 | #include "dev-setup.h" |
3ffd4af2 | 32 | #include "fd-util.h" |
d944dc95 | 33 | #include "fs-util.h" |
915e6d16 | 34 | #include "loop-util.h" |
07630cea LP |
35 | #include "loopback-setup.h" |
36 | #include "missing.h" | |
37 | #include "mkdir.h" | |
4349cd7c | 38 | #include "mount-util.h" |
3ffd4af2 | 39 | #include "namespace.h" |
07630cea | 40 | #include "path-util.h" |
d7b8eec7 | 41 | #include "selinux-util.h" |
2583fbea | 42 | #include "socket-util.h" |
8b43440b | 43 | #include "string-table.h" |
07630cea LP |
44 | #include "string-util.h" |
45 | #include "strv.h" | |
affb60b1 | 46 | #include "umask-util.h" |
ee104e11 | 47 | #include "user-util.h" |
07630cea | 48 | #include "util.h" |
15ae422b | 49 | |
737ba3c8 | 50 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
51 | ||
c17ec25e | 52 | typedef enum MountMode { |
15ae422b LP |
53 | /* This is ordered by priority! */ |
54 | INACCESSIBLE, | |
d2d6c096 LP |
55 | BIND_MOUNT, |
56 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 LP |
57 | PRIVATE_TMP, |
58 | PRIVATE_VAR_TMP, | |
7f112f50 | 59 | PRIVATE_DEV, |
5d997827 LP |
60 | BIND_DEV, |
61 | SYSFS, | |
62 | PROCFS, | |
63 | READONLY, | |
59eeb84b | 64 | READWRITE, |
c17ec25e | 65 | } MountMode; |
15ae422b | 66 | |
34de407a | 67 | typedef struct MountEntry { |
5327c910 | 68 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 69 | MountMode mode:5; |
5327c910 LP |
70 | bool ignore:1; /* Ignore if path does not exist? */ |
71 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 72 | bool read_only:1; /* Shall this mount point be read-only? */ |
5327c910 | 73 | char *path_malloc; /* Use this instead of 'path' if we had to allocate memory */ |
d2d6c096 LP |
74 | const char *source_const; /* The source path, for bind mounts */ |
75 | char *source_malloc; | |
34de407a | 76 | } MountEntry; |
15ae422b | 77 | |
5d997827 LP |
78 | /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted |
79 | * something there already. These mounts are hence overriden by any other explicitly configured mounts. */ | |
80 | static const MountEntry apivfs_table[] = { | |
81 | { "/proc", PROCFS, false }, | |
82 | { "/dev", BIND_DEV, false }, | |
83 | { "/sys", SYSFS, false }, | |
84 | }; | |
f471b2af | 85 | |
11a30cec | 86 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 87 | static const MountEntry protect_kernel_tunables_table[] = { |
c6232fb0 LP |
88 | { "/proc/sys", READONLY, false }, |
89 | { "/proc/sysrq-trigger", READONLY, true }, | |
90 | { "/proc/latency_stats", READONLY, true }, | |
91 | { "/proc/mtrr", READONLY, true }, | |
aa70f38b | 92 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ |
c6232fb0 LP |
93 | { "/proc/acpi", READONLY, true }, |
94 | { "/proc/timer_stats", READONLY, true }, | |
95 | { "/proc/asound", READONLY, true }, | |
96 | { "/proc/bus", READONLY, true }, | |
97 | { "/proc/fs", READONLY, true }, | |
98 | { "/proc/irq", READONLY, true }, | |
99 | { "/sys", READONLY, false }, | |
100 | { "/sys/kernel/debug", READONLY, true }, | |
101 | { "/sys/kernel/tracing", READONLY, true }, | |
102 | { "/sys/fs/cgroup", READWRITE, false }, /* READONLY is set by ProtectControlGroups= option */ | |
3a0bf6d6 | 103 | { "/sys/fs/selinux", READWRITE, true }, |
11a30cec DH |
104 | }; |
105 | ||
c575770b | 106 | /* ProtectKernelModules= option */ |
34de407a | 107 | static const MountEntry protect_kernel_modules_table[] = { |
c575770b | 108 | #ifdef HAVE_SPLIT_USR |
c6232fb0 | 109 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 110 | #endif |
c6232fb0 | 111 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
112 | }; |
113 | ||
b6c432ca DH |
114 | /* |
115 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
116 | * system should be protected by ProtectSystem= | |
117 | */ | |
34de407a | 118 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
119 | { "/home", READONLY, true }, |
120 | { "/run/user", READONLY, true }, | |
121 | { "/root", READONLY, true }, | |
b6c432ca DH |
122 | }; |
123 | ||
124 | /* ProtectHome=yes table */ | |
34de407a | 125 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
126 | { "/home", INACCESSIBLE, true }, |
127 | { "/run/user", INACCESSIBLE, true }, | |
128 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
129 | }; |
130 | ||
f471b2af | 131 | /* ProtectSystem=yes table */ |
34de407a | 132 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
133 | { "/usr", READONLY, false }, |
134 | { "/boot", READONLY, true }, | |
135 | { "/efi", READONLY, true }, | |
f471b2af DH |
136 | }; |
137 | ||
138 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 139 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
140 | { "/usr", READONLY, false }, |
141 | { "/boot", READONLY, true }, | |
142 | { "/efi", READONLY, true }, | |
143 | { "/etc", READONLY, false }, | |
f471b2af DH |
144 | }; |
145 | ||
146 | /* | |
147 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
148 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
149 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
150 | * protect those, and these options should be fully orthogonal. | |
151 | * (And of course /home and friends are also left writable, as ProtectHome= | |
152 | * shall manage those, orthogonally). | |
153 | */ | |
34de407a | 154 | static const MountEntry protect_system_strict_table[] = { |
ddbe0412 LP |
155 | { "/", READONLY, false }, |
156 | { "/proc", READWRITE, false }, /* ProtectKernelTunables= */ | |
157 | { "/sys", READWRITE, false }, /* ProtectKernelTunables= */ | |
158 | { "/dev", READWRITE, false }, /* PrivateDevices= */ | |
159 | { "/home", READWRITE, true }, /* ProtectHome= */ | |
160 | { "/run/user", READWRITE, true }, /* ProtectHome= */ | |
161 | { "/root", READWRITE, true }, /* ProtectHome= */ | |
f471b2af DH |
162 | }; |
163 | ||
34de407a | 164 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
165 | assert(p); |
166 | ||
5327c910 LP |
167 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
168 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 169 | |
5327c910 | 170 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
171 | } |
172 | ||
34de407a | 173 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
174 | assert(p); |
175 | ||
176 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
177 | } | |
178 | ||
d2d6c096 LP |
179 | static const char *mount_entry_source(const MountEntry *p) { |
180 | assert(p); | |
181 | ||
182 | return p->source_malloc ?: p->source_const; | |
183 | } | |
184 | ||
1eb7e08e LP |
185 | static void mount_entry_done(MountEntry *p) { |
186 | assert(p); | |
187 | ||
188 | p->path_malloc = mfree(p->path_malloc); | |
189 | p->source_malloc = mfree(p->source_malloc); | |
190 | } | |
191 | ||
34de407a | 192 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode) { |
15ae422b LP |
193 | char **i; |
194 | ||
613b411c LP |
195 | assert(p); |
196 | ||
5327c910 LP |
197 | /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */ |
198 | ||
15ae422b | 199 | STRV_FOREACH(i, strv) { |
5327c910 LP |
200 | bool ignore = false, needs_prefix = false; |
201 | const char *e = *i; | |
15ae422b | 202 | |
5327c910 LP |
203 | /* Look for any prefixes */ |
204 | if (startswith(e, "-")) { | |
205 | e++; | |
9c94d52e | 206 | ignore = true; |
ea92ae33 | 207 | } |
5327c910 LP |
208 | if (startswith(e, "+")) { |
209 | e++; | |
210 | needs_prefix = true; | |
211 | } | |
ea92ae33 | 212 | |
5327c910 | 213 | if (!path_is_absolute(e)) |
15ae422b LP |
214 | return -EINVAL; |
215 | ||
34de407a | 216 | *((*p)++) = (MountEntry) { |
5327c910 LP |
217 | .path_const = e, |
218 | .mode = mode, | |
219 | .ignore = ignore, | |
220 | .has_prefix = !needs_prefix, | |
221 | }; | |
15ae422b LP |
222 | } |
223 | ||
224 | return 0; | |
225 | } | |
226 | ||
d2d6c096 LP |
227 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, unsigned n) { |
228 | unsigned i; | |
229 | ||
230 | assert(p); | |
231 | ||
232 | for (i = 0; i < n; i++) { | |
233 | const BindMount *b = binds + i; | |
234 | ||
235 | *((*p)++) = (MountEntry) { | |
236 | .path_const = b->destination, | |
237 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
238 | .read_only = b->read_only, | |
239 | .source_const = b->source, | |
240 | }; | |
241 | } | |
242 | ||
243 | return 0; | |
244 | } | |
245 | ||
34de407a | 246 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, unsigned n, bool ignore_protect) { |
f471b2af | 247 | unsigned i; |
11a30cec DH |
248 | |
249 | assert(p); | |
f471b2af | 250 | assert(mounts); |
11a30cec | 251 | |
5327c910 | 252 | /* Adds a list of static pre-defined entries */ |
f471b2af | 253 | |
5327c910 | 254 | for (i = 0; i < n; i++) |
34de407a LP |
255 | *((*p)++) = (MountEntry) { |
256 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
257 | .mode = mounts[i].mode, |
258 | .ignore = mounts[i].ignore || ignore_protect, | |
259 | }; | |
f471b2af DH |
260 | |
261 | return 0; | |
262 | } | |
263 | ||
34de407a | 264 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
265 | assert(p); |
266 | ||
5327c910 | 267 | switch (protect_home) { |
b6c432ca | 268 | |
5327c910 | 269 | case PROTECT_HOME_NO: |
b6c432ca DH |
270 | return 0; |
271 | ||
b6c432ca | 272 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
273 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
274 | ||
b6c432ca | 275 | case PROTECT_HOME_YES: |
5327c910 LP |
276 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
277 | ||
b6c432ca | 278 | default: |
5327c910 | 279 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 280 | } |
b6c432ca DH |
281 | } |
282 | ||
34de407a | 283 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
284 | assert(p); |
285 | ||
5327c910 LP |
286 | switch (protect_system) { |
287 | ||
288 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
289 | return 0; |
290 | ||
f471b2af | 291 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
292 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
293 | ||
f471b2af | 294 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
295 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
296 | ||
f471b2af | 297 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
298 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
299 | ||
f471b2af | 300 | default: |
5327c910 | 301 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 302 | } |
11a30cec DH |
303 | } |
304 | ||
c17ec25e | 305 | static int mount_path_compare(const void *a, const void *b) { |
34de407a | 306 | const MountEntry *p = a, *q = b; |
a0827e2b | 307 | int d; |
15ae422b | 308 | |
6ee1a919 | 309 | /* If the paths are not equal, then order prefixes first */ |
34de407a | 310 | d = path_compare(mount_entry_path(p), mount_entry_path(q)); |
6ee1a919 LP |
311 | if (d != 0) |
312 | return d; | |
15ae422b | 313 | |
6ee1a919 LP |
314 | /* If the paths are equal, check the mode */ |
315 | if (p->mode < q->mode) | |
316 | return -1; | |
15ae422b | 317 | |
6ee1a919 LP |
318 | if (p->mode > q->mode) |
319 | return 1; | |
15ae422b | 320 | |
6ee1a919 | 321 | return 0; |
15ae422b LP |
322 | } |
323 | ||
34de407a | 324 | static int prefix_where_needed(MountEntry *m, unsigned n, const char *root_directory) { |
5327c910 LP |
325 | unsigned i; |
326 | ||
327 | /* Prefixes all paths in the bind mount table with the root directory if it is specified and the entry needs | |
328 | * that. */ | |
329 | ||
330 | if (!root_directory) | |
331 | return 0; | |
332 | ||
333 | for (i = 0; i < n; i++) { | |
334 | char *s; | |
335 | ||
336 | if (m[i].has_prefix) | |
337 | continue; | |
338 | ||
34de407a | 339 | s = prefix_root(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
340 | if (!s) |
341 | return -ENOMEM; | |
342 | ||
343 | free(m[i].path_malloc); | |
344 | m[i].path_malloc = s; | |
345 | ||
346 | m[i].has_prefix = true; | |
347 | } | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
34de407a LP |
352 | static void drop_duplicates(MountEntry *m, unsigned *n) { |
353 | MountEntry *f, *t, *previous; | |
15ae422b | 354 | |
c17ec25e | 355 | assert(m); |
15ae422b | 356 | assert(n); |
15ae422b | 357 | |
fe3c2583 LP |
358 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
359 | ||
1d54cd5d | 360 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 361 | |
fe3c2583 LP |
362 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
363 | * above. */ | |
34de407a LP |
364 | if (previous && path_equal(mount_entry_path(f), mount_entry_path(previous))) { |
365 | log_debug("%s is duplicate.", mount_entry_path(f)); | |
366 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ | |
1eb7e08e | 367 | mount_entry_done(f); |
15ae422b | 368 | continue; |
fe3c2583 | 369 | } |
15ae422b | 370 | |
e2d7c1a0 | 371 | *t = *f; |
15ae422b | 372 | previous = t; |
fe3c2583 LP |
373 | t++; |
374 | } | |
375 | ||
376 | *n = t - m; | |
377 | } | |
378 | ||
34de407a LP |
379 | static void drop_inaccessible(MountEntry *m, unsigned *n) { |
380 | MountEntry *f, *t; | |
fe3c2583 LP |
381 | const char *clear = NULL; |
382 | ||
383 | assert(m); | |
384 | assert(n); | |
385 | ||
386 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
387 | * ordered already. */ | |
388 | ||
1d54cd5d | 389 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
390 | |
391 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
392 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
393 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
394 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 395 | mount_entry_done(f); |
fe3c2583 LP |
396 | continue; |
397 | } | |
15ae422b | 398 | |
34de407a | 399 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
400 | |
401 | *t = *f; | |
15ae422b LP |
402 | t++; |
403 | } | |
404 | ||
c17ec25e | 405 | *n = t - m; |
15ae422b LP |
406 | } |
407 | ||
34de407a LP |
408 | static void drop_nop(MountEntry *m, unsigned *n) { |
409 | MountEntry *f, *t; | |
7648a565 LP |
410 | |
411 | assert(m); | |
412 | assert(n); | |
413 | ||
414 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
415 | * list is ordered by prefixes. */ | |
416 | ||
1d54cd5d | 417 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 LP |
418 | |
419 | /* Only suppress such subtrees for READONLY and READWRITE entries */ | |
420 | if (IN_SET(f->mode, READONLY, READWRITE)) { | |
34de407a | 421 | MountEntry *p; |
7648a565 LP |
422 | bool found = false; |
423 | ||
424 | /* Now let's find the first parent of the entry we are looking at. */ | |
425 | for (p = t-1; p >= m; p--) { | |
34de407a | 426 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
427 | found = true; |
428 | break; | |
429 | } | |
430 | } | |
431 | ||
432 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
433 | if (found && p->mode == f->mode) { | |
34de407a | 434 | log_debug("%s is redundant by %s", mount_entry_path(f), mount_entry_path(p)); |
1eb7e08e | 435 | mount_entry_done(f); |
7648a565 LP |
436 | continue; |
437 | } | |
438 | } | |
439 | ||
440 | *t = *f; | |
441 | t++; | |
442 | } | |
443 | ||
444 | *n = t - m; | |
445 | } | |
446 | ||
34de407a LP |
447 | static void drop_outside_root(const char *root_directory, MountEntry *m, unsigned *n) { |
448 | MountEntry *f, *t; | |
cd2902c9 LP |
449 | |
450 | assert(m); | |
451 | assert(n); | |
452 | ||
1d54cd5d | 453 | /* Nothing to do */ |
cd2902c9 LP |
454 | if (!root_directory) |
455 | return; | |
456 | ||
457 | /* Drops all mounts that are outside of the root directory. */ | |
458 | ||
1d54cd5d | 459 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 460 | |
34de407a LP |
461 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
462 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 463 | mount_entry_done(f); |
cd2902c9 LP |
464 | continue; |
465 | } | |
466 | ||
467 | *t = *f; | |
468 | t++; | |
469 | } | |
470 | ||
471 | *n = t - m; | |
472 | } | |
473 | ||
5d997827 | 474 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
475 | static const char devnodes[] = |
476 | "/dev/null\0" | |
477 | "/dev/zero\0" | |
478 | "/dev/full\0" | |
479 | "/dev/random\0" | |
480 | "/dev/urandom\0" | |
481 | "/dev/tty\0"; | |
482 | ||
2b85f4e1 | 483 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 484 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
7f112f50 LP |
485 | _cleanup_umask_ mode_t u; |
486 | int r; | |
487 | ||
488 | assert(m); | |
489 | ||
490 | u = umask(0000); | |
491 | ||
2b85f4e1 LP |
492 | if (!mkdtemp(temporary_mount)) |
493 | return -errno; | |
494 | ||
63c372cb | 495 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 496 | (void) mkdir(dev, 0755); |
737ba3c8 | 497 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2b85f4e1 LP |
498 | r = -errno; |
499 | goto fail; | |
500 | } | |
501 | ||
63c372cb | 502 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 503 | (void) mkdir(devpts, 0755); |
2b85f4e1 LP |
504 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
505 | r = -errno; | |
506 | goto fail; | |
507 | } | |
508 | ||
63c372cb | 509 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
3164e3cb ZJS |
510 | if (symlink("pts/ptmx", devptmx) < 0) { |
511 | r = -errno; | |
512 | goto fail; | |
513 | } | |
e06b6479 | 514 | |
63c372cb | 515 | devshm = strjoina(temporary_mount, "/dev/shm"); |
dc751688 | 516 | (void) mkdir(devshm, 01777); |
2b85f4e1 LP |
517 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
518 | if (r < 0) { | |
519 | r = -errno; | |
520 | goto fail; | |
521 | } | |
522 | ||
63c372cb | 523 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 524 | (void) mkdir(devmqueue, 0755); |
3164e3cb | 525 | (void) mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 526 | |
63c372cb | 527 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 528 | (void) mkdir(devhugepages, 0755); |
3164e3cb | 529 | (void) mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 530 | |
63c372cb | 531 | devlog = strjoina(temporary_mount, "/dev/log"); |
3164e3cb | 532 | (void) symlink("/run/systemd/journal/dev-log", devlog); |
82d25240 | 533 | |
7f112f50 | 534 | NULSTR_FOREACH(d, devnodes) { |
2b85f4e1 LP |
535 | _cleanup_free_ char *dn = NULL; |
536 | struct stat st; | |
537 | ||
538 | r = stat(d, &st); | |
7f112f50 | 539 | if (r < 0) { |
2b85f4e1 LP |
540 | |
541 | if (errno == ENOENT) | |
542 | continue; | |
543 | ||
544 | r = -errno; | |
545 | goto fail; | |
7f112f50 LP |
546 | } |
547 | ||
2b85f4e1 LP |
548 | if (!S_ISBLK(st.st_mode) && |
549 | !S_ISCHR(st.st_mode)) { | |
550 | r = -EINVAL; | |
551 | goto fail; | |
552 | } | |
553 | ||
554 | if (st.st_rdev == 0) | |
555 | continue; | |
556 | ||
557 | dn = strappend(temporary_mount, d); | |
558 | if (!dn) { | |
559 | r = -ENOMEM; | |
560 | goto fail; | |
561 | } | |
562 | ||
ecabcf8b | 563 | mac_selinux_create_file_prepare(d, st.st_mode); |
2b85f4e1 | 564 | r = mknod(dn, st.st_mode, st.st_rdev); |
ecabcf8b | 565 | mac_selinux_create_file_clear(); |
dd078a1e | 566 | |
2b85f4e1 LP |
567 | if (r < 0) { |
568 | r = -errno; | |
569 | goto fail; | |
570 | } | |
7f112f50 LP |
571 | } |
572 | ||
03cfe0d5 | 573 | dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
7f112f50 | 574 | |
ee818b89 AC |
575 | /* Create the /dev directory if missing. It is more likely to be |
576 | * missing when the service is started with RootDirectory. This is | |
577 | * consistent with mount units creating the mount points when missing. | |
578 | */ | |
34de407a | 579 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 580 | |
9e5f8252 | 581 | /* Unmount everything in old /dev */ |
34de407a LP |
582 | umount_recursive(mount_entry_path(m), 0); |
583 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { | |
2b85f4e1 LP |
584 | r = -errno; |
585 | goto fail; | |
586 | } | |
7f112f50 | 587 | |
2b85f4e1 LP |
588 | rmdir(dev); |
589 | rmdir(temporary_mount); | |
7f112f50 | 590 | |
2b85f4e1 | 591 | return 0; |
7f112f50 | 592 | |
2b85f4e1 LP |
593 | fail: |
594 | if (devpts) | |
595 | umount(devpts); | |
7f112f50 | 596 | |
2b85f4e1 LP |
597 | if (devshm) |
598 | umount(devshm); | |
7f112f50 | 599 | |
2b85f4e1 LP |
600 | if (devhugepages) |
601 | umount(devhugepages); | |
7f112f50 | 602 | |
2b85f4e1 LP |
603 | if (devmqueue) |
604 | umount(devmqueue); | |
7f112f50 | 605 | |
d267c5aa ZJS |
606 | umount(dev); |
607 | rmdir(dev); | |
2b85f4e1 | 608 | rmdir(temporary_mount); |
7f112f50 | 609 | |
2b85f4e1 | 610 | return r; |
7f112f50 LP |
611 | } |
612 | ||
5d997827 LP |
613 | static int mount_bind_dev(MountEntry *m) { |
614 | int r; | |
615 | ||
616 | assert(m); | |
617 | ||
618 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
619 | * /dev. This is only used when RootDirectory= is set. */ | |
620 | ||
645767d6 LP |
621 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
622 | ||
5d997827 LP |
623 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
624 | if (r < 0) | |
625 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
626 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
627 | return 0; | |
628 | ||
629 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
630 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
631 | ||
632 | return 1; | |
633 | } | |
634 | ||
635 | static int mount_sysfs(MountEntry *m) { | |
636 | int r; | |
637 | ||
638 | assert(m); | |
639 | ||
645767d6 LP |
640 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
641 | ||
5d997827 LP |
642 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
643 | if (r < 0) | |
644 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
645 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
646 | return 0; | |
647 | ||
648 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
649 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
650 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
651 | ||
652 | return 1; | |
653 | } | |
654 | ||
655 | static int mount_procfs(MountEntry *m) { | |
656 | int r; | |
657 | ||
658 | assert(m); | |
659 | ||
645767d6 LP |
660 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
661 | ||
5d997827 LP |
662 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
663 | if (r < 0) | |
664 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
665 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
666 | return 0; | |
667 | ||
668 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
669 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
670 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
671 | ||
672 | return 1; | |
673 | } | |
674 | ||
d2d6c096 LP |
675 | static int mount_entry_chase( |
676 | const char *root_directory, | |
677 | MountEntry *m, | |
678 | const char *path, | |
679 | char **location) { | |
680 | ||
8fceda93 LP |
681 | char *chased; |
682 | int r; | |
683 | ||
684 | assert(m); | |
685 | ||
686 | /* Since mount() will always follow symlinks and we need to take the different root directory into account we | |
d2d6c096 LP |
687 | * chase the symlinks on our own first. This is called for the destination path, as well as the source path (if |
688 | * that applies). The result is stored in "location". */ | |
8fceda93 | 689 | |
d2d6c096 | 690 | r = chase_symlinks(path, root_directory, 0, &chased); |
8fceda93 | 691 | if (r == -ENOENT && m->ignore) { |
d2d6c096 | 692 | log_debug_errno(r, "Path %s does not exist, ignoring.", path); |
8fceda93 LP |
693 | return 0; |
694 | } | |
695 | if (r < 0) | |
d2d6c096 | 696 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", path); |
8fceda93 | 697 | |
d2d6c096 | 698 | log_debug("Followed symlinks %s → %s.", path, chased); |
8fceda93 | 699 | |
d2d6c096 LP |
700 | free(*location); |
701 | *location = chased; | |
8fceda93 LP |
702 | |
703 | return 1; | |
704 | } | |
705 | ||
ac0930c8 | 706 | static int apply_mount( |
8fceda93 | 707 | const char *root_directory, |
34de407a | 708 | MountEntry *m, |
ac0930c8 | 709 | const char *tmp_dir, |
c17ec25e | 710 | const char *var_tmp_dir) { |
ac0930c8 | 711 | |
15ae422b | 712 | const char *what; |
d2d6c096 | 713 | bool rbind = true; |
15ae422b | 714 | int r; |
15ae422b | 715 | |
c17ec25e | 716 | assert(m); |
15ae422b | 717 | |
d2d6c096 | 718 | r = mount_entry_chase(root_directory, m, mount_entry_path(m), &m->path_malloc); |
8fceda93 LP |
719 | if (r <= 0) |
720 | return r; | |
721 | ||
34de407a | 722 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 723 | |
c17ec25e | 724 | switch (m->mode) { |
15ae422b | 725 | |
160cfdbe LP |
726 | case INACCESSIBLE: { |
727 | struct stat target; | |
6d313367 LP |
728 | |
729 | /* First, get rid of everything that is below if there | |
730 | * is anything... Then, overmount it with an | |
c4b41707 | 731 | * inaccessible path. */ |
34de407a | 732 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 733 | |
34de407a LP |
734 | if (lstat(mount_entry_path(m), &target) < 0) |
735 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); | |
15ae422b | 736 | |
c4b41707 | 737 | what = mode_to_inaccessible_node(target.st_mode); |
5fd7cf6f LP |
738 | if (!what) { |
739 | log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 AP |
740 | return -ELOOP; |
741 | } | |
742 | break; | |
160cfdbe | 743 | } |
fe3c2583 | 744 | |
15ae422b | 745 | case READONLY: |
15ae422b | 746 | case READWRITE: |
8fceda93 | 747 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
d944dc95 | 748 | if (r < 0) |
34de407a | 749 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
750 | 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. */ |
751 | return 0; | |
6b7c9f8b | 752 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 753 | what = mount_entry_path(m); |
6b7c9f8b | 754 | break; |
15ae422b | 755 | |
d2d6c096 LP |
756 | case BIND_MOUNT: |
757 | rbind = false; | |
758 | /* fallthrough */ | |
759 | ||
760 | case BIND_MOUNT_RECURSIVE: | |
761 | /* Also chase the source mount */ | |
5d997827 | 762 | |
d2d6c096 LP |
763 | r = mount_entry_chase(root_directory, m, mount_entry_source(m), &m->source_malloc); |
764 | if (r <= 0) | |
765 | return r; | |
766 | ||
767 | what = mount_entry_source(m); | |
768 | break; | |
769 | ||
ac0930c8 LP |
770 | case PRIVATE_TMP: |
771 | what = tmp_dir; | |
772 | break; | |
773 | ||
774 | case PRIVATE_VAR_TMP: | |
775 | what = var_tmp_dir; | |
15ae422b | 776 | break; |
e364ad06 | 777 | |
d6797c92 | 778 | case PRIVATE_DEV: |
5d997827 LP |
779 | return mount_private_dev(m); |
780 | ||
781 | case BIND_DEV: | |
782 | return mount_bind_dev(m); | |
783 | ||
784 | case SYSFS: | |
785 | return mount_sysfs(m); | |
786 | ||
787 | case PROCFS: | |
788 | return mount_procfs(m); | |
d6797c92 | 789 | |
e364ad06 LP |
790 | default: |
791 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
792 | } |
793 | ||
ac0930c8 | 794 | assert(what); |
15ae422b | 795 | |
d2d6c096 | 796 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) |
34de407a | 797 | return log_debug_errno(errno, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); |
6b7c9f8b | 798 | |
34de407a | 799 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 800 | return 0; |
ac0930c8 | 801 | } |
15ae422b | 802 | |
ac9de0b3 | 803 | static int make_read_only(MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) { |
6b7c9f8b | 804 | int r = 0; |
15ae422b | 805 | |
c17ec25e | 806 | assert(m); |
ac9de0b3 | 807 | assert(proc_self_mountinfo); |
ac0930c8 | 808 | |
34de407a | 809 | if (mount_entry_read_only(m)) |
ac9de0b3 | 810 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), true, blacklist, proc_self_mountinfo); |
13e785f7 | 811 | else if (m->mode == PRIVATE_DEV) { /* Superblock can be readonly but the submounts can't */ |
34de407a | 812 | if (mount(NULL, mount_entry_path(m), NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0) |
6b7c9f8b | 813 | r = -errno; |
737ba3c8 | 814 | } else |
6b7c9f8b LP |
815 | return 0; |
816 | ||
817 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
818 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
819 | * read-only mounts already applied. */ | |
ac0930c8 | 820 | |
8fceda93 LP |
821 | if (r == -ENOENT && m->ignore) |
822 | r = 0; | |
5327c910 | 823 | |
1d54cd5d | 824 | return r; |
d944dc95 LP |
825 | } |
826 | ||
9c988f93 | 827 | static bool namespace_info_mount_apivfs(const char *root_directory, const NameSpaceInfo *ns_info) { |
5d997827 LP |
828 | assert(ns_info); |
829 | ||
9c988f93 DH |
830 | /* |
831 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
832 | * since to protect the API VFS mounts, they need to be around in the | |
833 | * first place... and RootDirectory= or RootImage= need to be set. | |
834 | */ | |
5d997827 | 835 | |
9c988f93 DH |
836 | /* root_directory should point to a mount point */ |
837 | return root_directory && | |
838 | (ns_info->mount_apivfs || | |
839 | ns_info->protect_control_groups || | |
840 | ns_info->protect_kernel_tunables); | |
5d997827 LP |
841 | } |
842 | ||
2652c6c1 | 843 | static unsigned namespace_calculate_mounts( |
9c988f93 | 844 | const char* root_directory, |
c575770b | 845 | const NameSpaceInfo *ns_info, |
2652c6c1 DH |
846 | char** read_write_paths, |
847 | char** read_only_paths, | |
848 | char** inaccessible_paths, | |
d2d6c096 LP |
849 | const BindMount *bind_mounts, |
850 | unsigned n_bind_mounts, | |
2652c6c1 DH |
851 | const char* tmp_dir, |
852 | const char* var_tmp_dir, | |
2652c6c1 DH |
853 | ProtectHome protect_home, |
854 | ProtectSystem protect_system) { | |
855 | ||
b6c432ca | 856 | unsigned protect_home_cnt; |
f471b2af DH |
857 | unsigned protect_system_cnt = |
858 | (protect_system == PROTECT_SYSTEM_STRICT ? | |
859 | ELEMENTSOF(protect_system_strict_table) : | |
860 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
861 | ELEMENTSOF(protect_system_full_table) : | |
862 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
863 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
864 | ||
b6c432ca DH |
865 | protect_home_cnt = |
866 | (protect_home == PROTECT_HOME_YES ? | |
867 | ELEMENTSOF(protect_home_yes_table) : | |
868 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
869 | ELEMENTSOF(protect_home_read_only_table) : 0)); | |
870 | ||
2652c6c1 DH |
871 | return !!tmp_dir + !!var_tmp_dir + |
872 | strv_length(read_write_paths) + | |
873 | strv_length(read_only_paths) + | |
874 | strv_length(inaccessible_paths) + | |
d2d6c096 | 875 | n_bind_mounts + |
c575770b DH |
876 | ns_info->private_dev + |
877 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
878 | (ns_info->protect_control_groups ? 1 : 0) + | |
879 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + | |
5d997827 | 880 | protect_home_cnt + protect_system_cnt + |
9c988f93 | 881 | (namespace_info_mount_apivfs(root_directory, ns_info) ? ELEMENTSOF(apivfs_table) : 0); |
2652c6c1 DH |
882 | } |
883 | ||
613b411c | 884 | int setup_namespace( |
ee818b89 | 885 | const char* root_directory, |
915e6d16 | 886 | const char* root_image, |
c575770b | 887 | const NameSpaceInfo *ns_info, |
2a624c36 AP |
888 | char** read_write_paths, |
889 | char** read_only_paths, | |
890 | char** inaccessible_paths, | |
d2d6c096 LP |
891 | const BindMount *bind_mounts, |
892 | unsigned n_bind_mounts, | |
a004cb4c LP |
893 | const char* tmp_dir, |
894 | const char* var_tmp_dir, | |
1b8689f9 LP |
895 | ProtectHome protect_home, |
896 | ProtectSystem protect_system, | |
915e6d16 LP |
897 | unsigned long mount_flags, |
898 | DissectImageFlags dissect_image_flags) { | |
15ae422b | 899 | |
915e6d16 | 900 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 901 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 902 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 903 | _cleanup_free_ void *root_hash = NULL; |
34de407a | 904 | MountEntry *m, *mounts = NULL; |
78ebe980 | 905 | size_t root_hash_size = 0; |
d944dc95 | 906 | bool make_slave = false; |
f0a4feb0 | 907 | unsigned n_mounts; |
c17ec25e | 908 | int r = 0; |
15ae422b | 909 | |
915e6d16 LP |
910 | assert(ns_info); |
911 | ||
613b411c | 912 | if (mount_flags == 0) |
c17ec25e | 913 | mount_flags = MS_SHARED; |
ac0930c8 | 914 | |
915e6d16 LP |
915 | if (root_image) { |
916 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
917 | ||
918 | if (protect_system == PROTECT_SYSTEM_STRICT && strv_isempty(read_write_paths)) | |
919 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; | |
920 | ||
921 | r = loop_device_make_by_path(root_image, | |
922 | dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR, | |
923 | &loop_device); | |
924 | if (r < 0) | |
925 | return r; | |
926 | ||
78ebe980 LP |
927 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
928 | if (r < 0) | |
929 | return r; | |
930 | ||
931 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
932 | if (r < 0) | |
933 | return r; | |
934 | ||
935 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 LP |
936 | if (r < 0) |
937 | return r; | |
938 | ||
939 | if (!root_directory) { | |
940 | /* Create a mount point for the image, if it's still missing. We use the same mount point for | |
941 | * all images, which is safe, since they all live in their own namespaces after all, and hence | |
942 | * won't see each other. */ | |
943 | root_directory = "/run/systemd/unit-root"; | |
944 | (void) mkdir(root_directory, 0700); | |
945 | } | |
946 | } | |
947 | ||
cfbeb4ef | 948 | n_mounts = namespace_calculate_mounts( |
9c988f93 | 949 | root_directory, |
cfbeb4ef LP |
950 | ns_info, |
951 | read_write_paths, | |
952 | read_only_paths, | |
953 | inaccessible_paths, | |
d2d6c096 | 954 | bind_mounts, n_bind_mounts, |
cfbeb4ef LP |
955 | tmp_dir, var_tmp_dir, |
956 | protect_home, protect_system); | |
613b411c | 957 | |
2652c6c1 | 958 | /* Set mount slave mode */ |
f0a4feb0 | 959 | if (root_directory || n_mounts > 0) |
d944dc95 LP |
960 | make_slave = true; |
961 | ||
f0a4feb0 | 962 | if (n_mounts > 0) { |
34de407a | 963 | m = mounts = (MountEntry *) alloca0(n_mounts * sizeof(MountEntry)); |
5327c910 | 964 | r = append_access_mounts(&m, read_write_paths, READWRITE); |
613b411c | 965 | if (r < 0) |
f0a4feb0 | 966 | goto finish; |
613b411c | 967 | |
5327c910 | 968 | r = append_access_mounts(&m, read_only_paths, READONLY); |
613b411c | 969 | if (r < 0) |
f0a4feb0 | 970 | goto finish; |
613b411c | 971 | |
5327c910 | 972 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE); |
613b411c | 973 | if (r < 0) |
f0a4feb0 | 974 | goto finish; |
7ff7394d | 975 | |
d2d6c096 LP |
976 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
977 | if (r < 0) | |
978 | goto finish; | |
979 | ||
613b411c | 980 | if (tmp_dir) { |
34de407a | 981 | *(m++) = (MountEntry) { |
5327c910 LP |
982 | .path_const = "/tmp", |
983 | .mode = PRIVATE_TMP, | |
984 | }; | |
613b411c | 985 | } |
7ff7394d | 986 | |
613b411c | 987 | if (var_tmp_dir) { |
34de407a | 988 | *(m++) = (MountEntry) { |
5327c910 LP |
989 | .path_const = "/var/tmp", |
990 | .mode = PRIVATE_VAR_TMP, | |
991 | }; | |
7ff7394d | 992 | } |
ac0930c8 | 993 | |
c575770b | 994 | if (ns_info->private_dev) { |
34de407a | 995 | *(m++) = (MountEntry) { |
5327c910 LP |
996 | .path_const = "/dev", |
997 | .mode = PRIVATE_DEV, | |
998 | }; | |
7f112f50 LP |
999 | } |
1000 | ||
c575770b | 1001 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 1002 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 1003 | if (r < 0) |
f0a4feb0 | 1004 | goto finish; |
c575770b DH |
1005 | } |
1006 | ||
1007 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 1008 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 1009 | if (r < 0) |
f0a4feb0 | 1010 | goto finish; |
c575770b | 1011 | } |
59eeb84b | 1012 | |
c575770b | 1013 | if (ns_info->protect_control_groups) { |
34de407a | 1014 | *(m++) = (MountEntry) { |
5327c910 LP |
1015 | .path_const = "/sys/fs/cgroup", |
1016 | .mode = READONLY, | |
1017 | }; | |
59eeb84b LP |
1018 | } |
1019 | ||
5327c910 | 1020 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1021 | if (r < 0) |
f0a4feb0 | 1022 | goto finish; |
417116f2 | 1023 | |
5327c910 | 1024 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1025 | if (r < 0) |
f0a4feb0 | 1026 | goto finish; |
417116f2 | 1027 | |
9c988f93 | 1028 | if (namespace_info_mount_apivfs(root_directory, ns_info)) { |
5d997827 LP |
1029 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); |
1030 | if (r < 0) | |
1031 | goto finish; | |
1032 | } | |
1033 | ||
f0a4feb0 | 1034 | assert(mounts + n_mounts == m); |
ac0930c8 | 1035 | |
5327c910 LP |
1036 | /* Prepend the root directory where that's necessary */ |
1037 | r = prefix_where_needed(mounts, n_mounts, root_directory); | |
1038 | if (r < 0) | |
1039 | goto finish; | |
1040 | ||
34de407a | 1041 | qsort(mounts, n_mounts, sizeof(MountEntry), mount_path_compare); |
fe3c2583 | 1042 | |
f0a4feb0 DH |
1043 | drop_duplicates(mounts, &n_mounts); |
1044 | drop_outside_root(root_directory, mounts, &n_mounts); | |
1045 | drop_inaccessible(mounts, &n_mounts); | |
1046 | drop_nop(mounts, &n_mounts); | |
15ae422b LP |
1047 | } |
1048 | ||
d944dc95 LP |
1049 | if (unshare(CLONE_NEWNS) < 0) { |
1050 | r = -errno; | |
1051 | goto finish; | |
1052 | } | |
1e4e94c8 | 1053 | |
d944dc95 | 1054 | if (make_slave) { |
c2c13f2d LP |
1055 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1056 | shows up in the parent */ | |
d944dc95 LP |
1057 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { |
1058 | r = -errno; | |
1059 | goto finish; | |
1060 | } | |
ee818b89 AC |
1061 | } |
1062 | ||
10404d52 DH |
1063 | /* Try to set up the new root directory before mounting anything there */ |
1064 | if (root_directory) | |
1065 | (void) base_filesystem_create(root_directory, UID_INVALID, GID_INVALID); | |
1066 | ||
915e6d16 LP |
1067 | if (root_image) { |
1068 | r = dissected_image_mount(dissected_image, root_directory, dissect_image_flags); | |
1069 | if (r < 0) | |
1070 | goto finish; | |
1071 | ||
07ce7407 TM |
1072 | if (decrypted_image) { |
1073 | r = decrypted_image_relinquish(decrypted_image); | |
1074 | if (r < 0) | |
1075 | goto finish; | |
1076 | } | |
78ebe980 | 1077 | |
915e6d16 LP |
1078 | loop_device_relinquish(loop_device); |
1079 | ||
1080 | } else if (root_directory) { | |
1081 | ||
8f1ad200 | 1082 | /* Turn directory into bind mount, if it isn't one yet */ |
e1873695 | 1083 | r = path_is_mount_point(root_directory, NULL, AT_SYMLINK_FOLLOW); |
8f1ad200 | 1084 | if (r < 0) |
d944dc95 | 1085 | goto finish; |
8f1ad200 LP |
1086 | if (r == 0) { |
1087 | if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
1088 | r = -errno; | |
1089 | goto finish; | |
1090 | } | |
d944dc95 | 1091 | } |
ee818b89 | 1092 | } |
c2c13f2d | 1093 | |
f0a4feb0 | 1094 | if (n_mounts > 0) { |
ac9de0b3 | 1095 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b7c9f8b LP |
1096 | char **blacklist; |
1097 | unsigned j; | |
1098 | ||
ac9de0b3 TR |
1099 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc. |
1100 | * For example, this is the case with the option: 'InaccessiblePaths=/proc' */ | |
1101 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); | |
1102 | if (!proc_self_mountinfo) { | |
1103 | r = -errno; | |
1104 | goto finish; | |
1105 | } | |
1106 | ||
6b7c9f8b | 1107 | /* First round, add in all special mounts we need */ |
f0a4feb0 | 1108 | for (m = mounts; m < mounts + n_mounts; ++m) { |
8fceda93 | 1109 | r = apply_mount(root_directory, m, tmp_dir, var_tmp_dir); |
c2c13f2d | 1110 | if (r < 0) |
d944dc95 | 1111 | goto finish; |
c2c13f2d | 1112 | } |
15ae422b | 1113 | |
6b7c9f8b | 1114 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
f0a4feb0 DH |
1115 | blacklist = newa(char*, n_mounts+1); |
1116 | for (j = 0; j < n_mounts; j++) | |
34de407a | 1117 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1118 | blacklist[j] = NULL; |
1119 | ||
1120 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1121 | for (m = mounts; m < mounts + n_mounts; ++m) { |
ac9de0b3 | 1122 | r = make_read_only(m, blacklist, proc_self_mountinfo); |
c2c13f2d | 1123 | if (r < 0) |
d944dc95 | 1124 | goto finish; |
c2c13f2d | 1125 | } |
15ae422b LP |
1126 | } |
1127 | ||
ee818b89 AC |
1128 | if (root_directory) { |
1129 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ | |
1130 | r = mount_move_root(root_directory); | |
d944dc95 LP |
1131 | if (r < 0) |
1132 | goto finish; | |
ee818b89 AC |
1133 | } |
1134 | ||
c2c13f2d LP |
1135 | /* Remount / as the desired mode. Not that this will not |
1136 | * reestablish propagation from our side to the host, since | |
1137 | * what's disconnected is disconnected. */ | |
d944dc95 LP |
1138 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
1139 | r = -errno; | |
1140 | goto finish; | |
1141 | } | |
15ae422b | 1142 | |
d944dc95 | 1143 | r = 0; |
15ae422b | 1144 | |
d944dc95 | 1145 | finish: |
f0a4feb0 | 1146 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1147 | mount_entry_done(m); |
613b411c LP |
1148 | |
1149 | return r; | |
1150 | } | |
1151 | ||
d2d6c096 LP |
1152 | void bind_mount_free_many(BindMount *b, unsigned n) { |
1153 | unsigned i; | |
1154 | ||
1155 | assert(b || n == 0); | |
1156 | ||
1157 | for (i = 0; i < n; i++) { | |
1158 | free(b[i].source); | |
1159 | free(b[i].destination); | |
1160 | } | |
1161 | ||
1162 | free(b); | |
1163 | } | |
1164 | ||
1165 | int bind_mount_add(BindMount **b, unsigned *n, const BindMount *item) { | |
1166 | _cleanup_free_ char *s = NULL, *d = NULL; | |
1167 | BindMount *c; | |
1168 | ||
1169 | assert(b); | |
1170 | assert(n); | |
1171 | assert(item); | |
1172 | ||
1173 | s = strdup(item->source); | |
1174 | if (!s) | |
1175 | return -ENOMEM; | |
1176 | ||
1177 | d = strdup(item->destination); | |
1178 | if (!d) | |
1179 | return -ENOMEM; | |
1180 | ||
1181 | c = realloc_multiply(*b, sizeof(BindMount), *n + 1); | |
1182 | if (!c) | |
1183 | return -ENOMEM; | |
1184 | ||
1185 | *b = c; | |
1186 | ||
1187 | c[(*n) ++] = (BindMount) { | |
1188 | .source = s, | |
1189 | .destination = d, | |
1190 | .read_only = item->read_only, | |
1191 | .recursive = item->recursive, | |
1192 | .ignore_enoent = item->ignore_enoent, | |
1193 | }; | |
1194 | ||
1195 | s = d = NULL; | |
1196 | return 0; | |
1197 | } | |
1198 | ||
613b411c LP |
1199 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1200 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1201 | char bid[SD_ID128_STRING_MAX]; |
1202 | sd_id128_t boot_id; | |
1203 | int r; | |
613b411c LP |
1204 | |
1205 | assert(id); | |
1206 | assert(prefix); | |
1207 | assert(path); | |
1208 | ||
6b46ea73 LP |
1209 | /* We include the boot id in the directory so that after a |
1210 | * reboot we can easily identify obsolete directories. */ | |
1211 | ||
1212 | r = sd_id128_get_boot(&boot_id); | |
1213 | if (r < 0) | |
1214 | return r; | |
1215 | ||
605405c6 | 1216 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1217 | if (!x) |
1218 | return -ENOMEM; | |
1219 | ||
1220 | RUN_WITH_UMASK(0077) | |
1221 | if (!mkdtemp(x)) | |
1222 | return -errno; | |
1223 | ||
1224 | RUN_WITH_UMASK(0000) { | |
1225 | char *y; | |
1226 | ||
63c372cb | 1227 | y = strjoina(x, "/tmp"); |
613b411c LP |
1228 | |
1229 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1230 | return -errno; | |
c17ec25e | 1231 | } |
15ae422b | 1232 | |
613b411c LP |
1233 | *path = x; |
1234 | x = NULL; | |
1235 | ||
1236 | return 0; | |
1237 | } | |
1238 | ||
1239 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1240 | char *a, *b; | |
1241 | int r; | |
1242 | ||
1243 | assert(id); | |
1244 | assert(tmp_dir); | |
1245 | assert(var_tmp_dir); | |
1246 | ||
1247 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1248 | if (r < 0) | |
1249 | return r; | |
1250 | ||
1251 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1252 | if (r < 0) { | |
1253 | char *t; | |
1254 | ||
63c372cb | 1255 | t = strjoina(a, "/tmp"); |
613b411c LP |
1256 | rmdir(t); |
1257 | rmdir(a); | |
1258 | ||
1259 | free(a); | |
1260 | return r; | |
1261 | } | |
1262 | ||
1263 | *tmp_dir = a; | |
1264 | *var_tmp_dir = b; | |
1265 | ||
1266 | return 0; | |
1267 | } | |
1268 | ||
1269 | int setup_netns(int netns_storage_socket[2]) { | |
1270 | _cleanup_close_ int netns = -1; | |
3ee897d6 | 1271 | int r, q; |
613b411c LP |
1272 | |
1273 | assert(netns_storage_socket); | |
1274 | assert(netns_storage_socket[0] >= 0); | |
1275 | assert(netns_storage_socket[1] >= 0); | |
1276 | ||
1277 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1278 | * namespace reference fd. Whatever process runs this first |
1279 | * shall create a new namespace, all others should just join | |
1280 | * it. To serialize that we use a file lock on the socket | |
1281 | * pair. | |
613b411c LP |
1282 | * |
1283 | * It's a bit crazy, but hey, works great! */ | |
1284 | ||
1285 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1286 | return -errno; | |
1287 | ||
3ee897d6 LP |
1288 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1289 | if (netns == -EAGAIN) { | |
613b411c LP |
1290 | /* Nothing stored yet, so let's create a new namespace */ |
1291 | ||
1292 | if (unshare(CLONE_NEWNET) < 0) { | |
1293 | r = -errno; | |
1294 | goto fail; | |
1295 | } | |
1296 | ||
1297 | loopback_setup(); | |
1298 | ||
1299 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1300 | if (netns < 0) { | |
1301 | r = -errno; | |
1302 | goto fail; | |
1303 | } | |
1304 | ||
1305 | r = 1; | |
613b411c | 1306 | |
3ee897d6 LP |
1307 | } else if (netns < 0) { |
1308 | r = netns; | |
1309 | goto fail; | |
613b411c | 1310 | |
3ee897d6 LP |
1311 | } else { |
1312 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1313 | if (setns(netns, CLONE_NEWNET) < 0) { |
1314 | r = -errno; | |
1315 | goto fail; | |
1316 | } | |
1317 | ||
1318 | r = 0; | |
1319 | } | |
1320 | ||
3ee897d6 LP |
1321 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1322 | if (q < 0) { | |
1323 | r = q; | |
613b411c LP |
1324 | goto fail; |
1325 | } | |
1326 | ||
1327 | fail: | |
fe048ce5 | 1328 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1329 | return r; |
1330 | } | |
417116f2 | 1331 | |
1b8689f9 LP |
1332 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1333 | [PROTECT_HOME_NO] = "no", | |
1334 | [PROTECT_HOME_YES] = "yes", | |
1335 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
417116f2 LP |
1336 | }; |
1337 | ||
1b8689f9 LP |
1338 | DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome); |
1339 | ||
1340 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { | |
1341 | [PROTECT_SYSTEM_NO] = "no", | |
1342 | [PROTECT_SYSTEM_YES] = "yes", | |
1343 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1344 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1345 | }; |
1346 | ||
1347 | DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem); |