]>
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" |
07630cea LP |
33 | #include "loopback-setup.h" |
34 | #include "missing.h" | |
35 | #include "mkdir.h" | |
4349cd7c | 36 | #include "mount-util.h" |
3ffd4af2 | 37 | #include "namespace.h" |
07630cea | 38 | #include "path-util.h" |
d7b8eec7 | 39 | #include "selinux-util.h" |
2583fbea | 40 | #include "socket-util.h" |
8b43440b | 41 | #include "string-table.h" |
07630cea LP |
42 | #include "string-util.h" |
43 | #include "strv.h" | |
affb60b1 | 44 | #include "umask-util.h" |
ee104e11 | 45 | #include "user-util.h" |
07630cea | 46 | #include "util.h" |
15ae422b | 47 | |
737ba3c8 | 48 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
49 | ||
c17ec25e | 50 | typedef enum MountMode { |
15ae422b LP |
51 | /* This is ordered by priority! */ |
52 | INACCESSIBLE, | |
53 | READONLY, | |
ac0930c8 LP |
54 | PRIVATE_TMP, |
55 | PRIVATE_VAR_TMP, | |
7f112f50 | 56 | PRIVATE_DEV, |
59eeb84b | 57 | READWRITE, |
c17ec25e | 58 | } MountMode; |
15ae422b | 59 | |
c17ec25e | 60 | typedef struct BindMount { |
d944dc95 LP |
61 | const char *path; /* stack memory, doesn't need to be freed explicitly */ |
62 | char *chased; /* malloc()ed memory, needs to be freed */ | |
c17ec25e | 63 | MountMode mode; |
ea92ae33 | 64 | bool ignore; |
c17ec25e | 65 | } BindMount; |
15ae422b | 66 | |
c17ec25e | 67 | static int append_mounts(BindMount **p, char **strv, MountMode mode) { |
15ae422b LP |
68 | char **i; |
69 | ||
613b411c LP |
70 | assert(p); |
71 | ||
15ae422b LP |
72 | STRV_FOREACH(i, strv) { |
73 | ||
ea92ae33 MW |
74 | (*p)->ignore = false; |
75 | ||
94828d2d | 76 | if ((mode == INACCESSIBLE || mode == READONLY || mode == READWRITE) && (*i)[0] == '-') { |
ea92ae33 MW |
77 | (*p)->ignore = true; |
78 | (*i)++; | |
79 | } | |
80 | ||
15ae422b LP |
81 | if (!path_is_absolute(*i)) |
82 | return -EINVAL; | |
83 | ||
84 | (*p)->path = *i; | |
85 | (*p)->mode = mode; | |
86 | (*p)++; | |
87 | } | |
88 | ||
89 | return 0; | |
90 | } | |
91 | ||
c17ec25e MS |
92 | static int mount_path_compare(const void *a, const void *b) { |
93 | const BindMount *p = a, *q = b; | |
a0827e2b | 94 | int d; |
15ae422b | 95 | |
6ee1a919 | 96 | /* If the paths are not equal, then order prefixes first */ |
a0827e2b | 97 | d = path_compare(p->path, q->path); |
6ee1a919 LP |
98 | if (d != 0) |
99 | return d; | |
15ae422b | 100 | |
6ee1a919 LP |
101 | /* If the paths are equal, check the mode */ |
102 | if (p->mode < q->mode) | |
103 | return -1; | |
15ae422b | 104 | |
6ee1a919 LP |
105 | if (p->mode > q->mode) |
106 | return 1; | |
15ae422b | 107 | |
6ee1a919 | 108 | return 0; |
15ae422b LP |
109 | } |
110 | ||
c17ec25e MS |
111 | static void drop_duplicates(BindMount *m, unsigned *n) { |
112 | BindMount *f, *t, *previous; | |
15ae422b | 113 | |
c17ec25e | 114 | assert(m); |
15ae422b | 115 | assert(n); |
15ae422b | 116 | |
fe3c2583 LP |
117 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
118 | ||
c17ec25e | 119 | for (f = m, t = m, previous = NULL; f < m+*n; f++) { |
15ae422b | 120 | |
fe3c2583 LP |
121 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
122 | * above. */ | |
123 | if (previous && path_equal(f->path, previous->path)) { | |
124 | log_debug("%s is duplicate.", f->path); | |
15ae422b | 125 | continue; |
fe3c2583 | 126 | } |
15ae422b | 127 | |
e2d7c1a0 | 128 | *t = *f; |
15ae422b | 129 | previous = t; |
fe3c2583 LP |
130 | t++; |
131 | } | |
132 | ||
133 | *n = t - m; | |
134 | } | |
135 | ||
136 | static void drop_inaccessible(BindMount *m, unsigned *n) { | |
137 | BindMount *f, *t; | |
138 | const char *clear = NULL; | |
139 | ||
140 | assert(m); | |
141 | assert(n); | |
142 | ||
143 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
144 | * ordered already. */ | |
145 | ||
146 | for (f = m, t = m; f < m+*n; f++) { | |
147 | ||
148 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
149 | * it, as inaccessible paths really should drop the entire subtree. */ | |
150 | if (clear && path_startswith(f->path, clear)) { | |
151 | log_debug("%s is masked by %s.", f->path, clear); | |
152 | continue; | |
153 | } | |
15ae422b | 154 | |
fe3c2583 LP |
155 | clear = f->mode == INACCESSIBLE ? f->path : NULL; |
156 | ||
157 | *t = *f; | |
15ae422b LP |
158 | t++; |
159 | } | |
160 | ||
c17ec25e | 161 | *n = t - m; |
15ae422b LP |
162 | } |
163 | ||
7648a565 LP |
164 | static void drop_nop(BindMount *m, unsigned *n) { |
165 | BindMount *f, *t; | |
166 | ||
167 | assert(m); | |
168 | assert(n); | |
169 | ||
170 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
171 | * list is ordered by prefixes. */ | |
172 | ||
173 | for (f = m, t = m; f < m+*n; f++) { | |
174 | ||
175 | /* Only suppress such subtrees for READONLY and READWRITE entries */ | |
176 | if (IN_SET(f->mode, READONLY, READWRITE)) { | |
177 | BindMount *p; | |
178 | bool found = false; | |
179 | ||
180 | /* Now let's find the first parent of the entry we are looking at. */ | |
181 | for (p = t-1; p >= m; p--) { | |
182 | if (path_startswith(f->path, p->path)) { | |
183 | found = true; | |
184 | break; | |
185 | } | |
186 | } | |
187 | ||
188 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
189 | if (found && p->mode == f->mode) { | |
190 | log_debug("%s is redundant by %s", f->path, p->path); | |
191 | continue; | |
192 | } | |
193 | } | |
194 | ||
195 | *t = *f; | |
196 | t++; | |
197 | } | |
198 | ||
199 | *n = t - m; | |
200 | } | |
201 | ||
7f112f50 LP |
202 | static int mount_dev(BindMount *m) { |
203 | static const char devnodes[] = | |
204 | "/dev/null\0" | |
205 | "/dev/zero\0" | |
206 | "/dev/full\0" | |
207 | "/dev/random\0" | |
208 | "/dev/urandom\0" | |
209 | "/dev/tty\0"; | |
210 | ||
2b85f4e1 | 211 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 212 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
7f112f50 LP |
213 | _cleanup_umask_ mode_t u; |
214 | int r; | |
215 | ||
216 | assert(m); | |
217 | ||
218 | u = umask(0000); | |
219 | ||
2b85f4e1 LP |
220 | if (!mkdtemp(temporary_mount)) |
221 | return -errno; | |
222 | ||
63c372cb | 223 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 224 | (void) mkdir(dev, 0755); |
737ba3c8 | 225 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2b85f4e1 LP |
226 | r = -errno; |
227 | goto fail; | |
228 | } | |
229 | ||
63c372cb | 230 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 231 | (void) mkdir(devpts, 0755); |
2b85f4e1 LP |
232 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
233 | r = -errno; | |
234 | goto fail; | |
235 | } | |
236 | ||
63c372cb | 237 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
3164e3cb ZJS |
238 | if (symlink("pts/ptmx", devptmx) < 0) { |
239 | r = -errno; | |
240 | goto fail; | |
241 | } | |
e06b6479 | 242 | |
63c372cb | 243 | devshm = strjoina(temporary_mount, "/dev/shm"); |
dc751688 | 244 | (void) mkdir(devshm, 01777); |
2b85f4e1 LP |
245 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
246 | if (r < 0) { | |
247 | r = -errno; | |
248 | goto fail; | |
249 | } | |
250 | ||
63c372cb | 251 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 252 | (void) mkdir(devmqueue, 0755); |
3164e3cb | 253 | (void) mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 254 | |
63c372cb | 255 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 256 | (void) mkdir(devhugepages, 0755); |
3164e3cb | 257 | (void) mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 258 | |
63c372cb | 259 | devlog = strjoina(temporary_mount, "/dev/log"); |
3164e3cb | 260 | (void) symlink("/run/systemd/journal/dev-log", devlog); |
82d25240 | 261 | |
7f112f50 | 262 | NULSTR_FOREACH(d, devnodes) { |
2b85f4e1 LP |
263 | _cleanup_free_ char *dn = NULL; |
264 | struct stat st; | |
265 | ||
266 | r = stat(d, &st); | |
7f112f50 | 267 | if (r < 0) { |
2b85f4e1 LP |
268 | |
269 | if (errno == ENOENT) | |
270 | continue; | |
271 | ||
272 | r = -errno; | |
273 | goto fail; | |
7f112f50 LP |
274 | } |
275 | ||
2b85f4e1 LP |
276 | if (!S_ISBLK(st.st_mode) && |
277 | !S_ISCHR(st.st_mode)) { | |
278 | r = -EINVAL; | |
279 | goto fail; | |
280 | } | |
281 | ||
282 | if (st.st_rdev == 0) | |
283 | continue; | |
284 | ||
285 | dn = strappend(temporary_mount, d); | |
286 | if (!dn) { | |
287 | r = -ENOMEM; | |
288 | goto fail; | |
289 | } | |
290 | ||
ecabcf8b | 291 | mac_selinux_create_file_prepare(d, st.st_mode); |
2b85f4e1 | 292 | r = mknod(dn, st.st_mode, st.st_rdev); |
ecabcf8b | 293 | mac_selinux_create_file_clear(); |
dd078a1e | 294 | |
2b85f4e1 LP |
295 | if (r < 0) { |
296 | r = -errno; | |
297 | goto fail; | |
298 | } | |
7f112f50 LP |
299 | } |
300 | ||
03cfe0d5 | 301 | dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
7f112f50 | 302 | |
ee818b89 AC |
303 | /* Create the /dev directory if missing. It is more likely to be |
304 | * missing when the service is started with RootDirectory. This is | |
305 | * consistent with mount units creating the mount points when missing. | |
306 | */ | |
307 | (void) mkdir_p_label(m->path, 0755); | |
308 | ||
9e5f8252 | 309 | /* Unmount everything in old /dev */ |
310 | umount_recursive(m->path, 0); | |
ee818b89 | 311 | if (mount(dev, m->path, NULL, MS_MOVE, NULL) < 0) { |
2b85f4e1 LP |
312 | r = -errno; |
313 | goto fail; | |
314 | } | |
7f112f50 | 315 | |
2b85f4e1 LP |
316 | rmdir(dev); |
317 | rmdir(temporary_mount); | |
7f112f50 | 318 | |
2b85f4e1 | 319 | return 0; |
7f112f50 | 320 | |
2b85f4e1 LP |
321 | fail: |
322 | if (devpts) | |
323 | umount(devpts); | |
7f112f50 | 324 | |
2b85f4e1 LP |
325 | if (devshm) |
326 | umount(devshm); | |
7f112f50 | 327 | |
2b85f4e1 LP |
328 | if (devhugepages) |
329 | umount(devhugepages); | |
7f112f50 | 330 | |
2b85f4e1 LP |
331 | if (devmqueue) |
332 | umount(devmqueue); | |
7f112f50 | 333 | |
d267c5aa ZJS |
334 | umount(dev); |
335 | rmdir(dev); | |
2b85f4e1 | 336 | rmdir(temporary_mount); |
7f112f50 | 337 | |
2b85f4e1 | 338 | return r; |
7f112f50 LP |
339 | } |
340 | ||
ac0930c8 | 341 | static int apply_mount( |
c17ec25e | 342 | BindMount *m, |
ac0930c8 | 343 | const char *tmp_dir, |
c17ec25e | 344 | const char *var_tmp_dir) { |
ac0930c8 | 345 | |
15ae422b | 346 | const char *what; |
15ae422b | 347 | int r; |
15ae422b | 348 | |
c17ec25e | 349 | assert(m); |
15ae422b | 350 | |
fe3c2583 LP |
351 | log_debug("Applying namespace mount on %s", m->path); |
352 | ||
c17ec25e | 353 | switch (m->mode) { |
15ae422b | 354 | |
160cfdbe LP |
355 | case INACCESSIBLE: { |
356 | struct stat target; | |
6d313367 LP |
357 | |
358 | /* First, get rid of everything that is below if there | |
359 | * is anything... Then, overmount it with an | |
c4b41707 | 360 | * inaccessible path. */ |
fe3c2583 | 361 | (void) umount_recursive(m->path, 0); |
6d313367 | 362 | |
d944dc95 | 363 | if (lstat(m->path, &target) < 0) |
160cfdbe | 364 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", m->path); |
15ae422b | 365 | |
c4b41707 | 366 | what = mode_to_inaccessible_node(target.st_mode); |
5fd7cf6f LP |
367 | if (!what) { |
368 | log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 AP |
369 | return -ELOOP; |
370 | } | |
371 | break; | |
160cfdbe | 372 | } |
fe3c2583 | 373 | |
15ae422b | 374 | case READONLY: |
15ae422b | 375 | case READWRITE: |
6b7c9f8b LP |
376 | |
377 | r = path_is_mount_point(m->path, 0); | |
d944dc95 | 378 | if (r < 0) |
6b7c9f8b | 379 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", m->path); |
6b7c9f8b LP |
380 | 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. */ |
381 | return 0; | |
382 | ||
383 | /* This isn't a mount point yet, let's make it one. */ | |
384 | what = m->path; | |
385 | break; | |
15ae422b | 386 | |
ac0930c8 LP |
387 | case PRIVATE_TMP: |
388 | what = tmp_dir; | |
389 | break; | |
390 | ||
391 | case PRIVATE_VAR_TMP: | |
392 | what = var_tmp_dir; | |
15ae422b | 393 | break; |
e364ad06 | 394 | |
d6797c92 LP |
395 | case PRIVATE_DEV: |
396 | return mount_dev(m); | |
397 | ||
e364ad06 LP |
398 | default: |
399 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
400 | } |
401 | ||
ac0930c8 | 402 | assert(what); |
15ae422b | 403 | |
d944dc95 | 404 | if (mount(what, m->path, NULL, MS_BIND|MS_REC, NULL) < 0) |
5fd7cf6f | 405 | return log_debug_errno(errno, "Failed to mount %s to %s: %m", what, m->path); |
6b7c9f8b LP |
406 | |
407 | log_debug("Successfully mounted %s to %s", what, m->path); | |
408 | return 0; | |
ac0930c8 | 409 | } |
15ae422b | 410 | |
6b7c9f8b LP |
411 | static int make_read_only(BindMount *m, char **blacklist) { |
412 | int r = 0; | |
15ae422b | 413 | |
c17ec25e | 414 | assert(m); |
ac0930c8 | 415 | |
d6797c92 | 416 | if (IN_SET(m->mode, INACCESSIBLE, READONLY)) |
6b7c9f8b LP |
417 | r = bind_remount_recursive(m->path, true, blacklist); |
418 | else if (m->mode == PRIVATE_DEV) { /* Can be readonly but the submounts can't*/ | |
419 | if (mount(NULL, m->path, NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0) | |
420 | r = -errno; | |
737ba3c8 | 421 | } else |
6b7c9f8b LP |
422 | return 0; |
423 | ||
424 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
425 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
426 | * read-only mounts already applied. */ | |
ac0930c8 | 427 | |
d6797c92 | 428 | return r; |
15ae422b LP |
429 | } |
430 | ||
d944dc95 LP |
431 | static int chase_all_symlinks(const char *root_directory, BindMount *m, unsigned *n) { |
432 | BindMount *f, *t; | |
433 | int r; | |
434 | ||
435 | assert(m); | |
436 | assert(n); | |
437 | ||
438 | /* Since mount() will always follow symlinks and we need to take the different root directory into account we | |
439 | * chase the symlinks on our own first. This call wil do so for all entries and remove all entries where we | |
440 | * can't resolve the path, and which have been marked for such removal. */ | |
441 | ||
442 | for (f = m, t = m; f < m+*n; f++) { | |
443 | ||
444 | r = chase_symlinks(f->path, root_directory, &f->chased); | |
445 | if (r == -ENOENT && f->ignore) /* Doesn't exist? Then remove it! */ | |
446 | continue; | |
447 | if (r < 0) | |
448 | return log_debug_errno(r, "Failed to chase symlinks for %s: %m", f->path); | |
449 | ||
450 | if (path_equal(f->path, f->chased)) | |
451 | f->chased = mfree(f->chased); | |
452 | else { | |
453 | log_debug("Chased %s → %s", f->path, f->chased); | |
454 | f->path = f->chased; | |
455 | } | |
456 | ||
457 | *t = *f; | |
458 | t++; | |
459 | } | |
460 | ||
461 | *n = t - m; | |
462 | return 0; | |
463 | } | |
464 | ||
613b411c | 465 | int setup_namespace( |
ee818b89 | 466 | const char* root_directory, |
2a624c36 AP |
467 | char** read_write_paths, |
468 | char** read_only_paths, | |
469 | char** inaccessible_paths, | |
a004cb4c LP |
470 | const char* tmp_dir, |
471 | const char* var_tmp_dir, | |
7f112f50 | 472 | bool private_dev, |
59eeb84b LP |
473 | bool protect_sysctl, |
474 | bool protect_cgroups, | |
1b8689f9 LP |
475 | ProtectHome protect_home, |
476 | ProtectSystem protect_system, | |
e6547662 | 477 | unsigned long mount_flags) { |
15ae422b | 478 | |
7ff7394d | 479 | BindMount *m, *mounts = NULL; |
d944dc95 | 480 | bool make_slave = false; |
613b411c | 481 | unsigned n; |
c17ec25e | 482 | int r = 0; |
15ae422b | 483 | |
613b411c | 484 | if (mount_flags == 0) |
c17ec25e | 485 | mount_flags = MS_SHARED; |
ac0930c8 | 486 | |
9ca6ff50 | 487 | n = !!tmp_dir + !!var_tmp_dir + |
2a624c36 AP |
488 | strv_length(read_write_paths) + |
489 | strv_length(read_only_paths) + | |
490 | strv_length(inaccessible_paths) + | |
417116f2 | 491 | private_dev + |
59eeb84b LP |
492 | (protect_sysctl ? 3 : 0) + |
493 | (protect_cgroups != protect_sysctl) + | |
3f815163 LP |
494 | (protect_home != PROTECT_HOME_NO || protect_system == PROTECT_SYSTEM_STRICT ? 3 : 0) + |
495 | (protect_system == PROTECT_SYSTEM_STRICT ? | |
496 | (2 + !private_dev + !protect_sysctl) : | |
497 | ((protect_system != PROTECT_SYSTEM_NO ? 3 : 0) + | |
498 | (protect_system == PROTECT_SYSTEM_FULL ? 1 : 0))); | |
613b411c | 499 | |
d944dc95 LP |
500 | if (root_directory || n > 0) |
501 | make_slave = true; | |
502 | ||
613b411c | 503 | if (n > 0) { |
002b2268 | 504 | m = mounts = (BindMount *) alloca0(n * sizeof(BindMount)); |
2a624c36 | 505 | r = append_mounts(&m, read_write_paths, READWRITE); |
613b411c LP |
506 | if (r < 0) |
507 | return r; | |
508 | ||
2a624c36 | 509 | r = append_mounts(&m, read_only_paths, READONLY); |
613b411c LP |
510 | if (r < 0) |
511 | return r; | |
512 | ||
2a624c36 | 513 | r = append_mounts(&m, inaccessible_paths, INACCESSIBLE); |
613b411c | 514 | if (r < 0) |
7ff7394d ZJS |
515 | return r; |
516 | ||
613b411c | 517 | if (tmp_dir) { |
ee818b89 | 518 | m->path = prefix_roota(root_directory, "/tmp"); |
7ff7394d ZJS |
519 | m->mode = PRIVATE_TMP; |
520 | m++; | |
613b411c | 521 | } |
7ff7394d | 522 | |
613b411c | 523 | if (var_tmp_dir) { |
ee818b89 | 524 | m->path = prefix_roota(root_directory, "/var/tmp"); |
7ff7394d ZJS |
525 | m->mode = PRIVATE_VAR_TMP; |
526 | m++; | |
527 | } | |
ac0930c8 | 528 | |
7f112f50 | 529 | if (private_dev) { |
ee818b89 | 530 | m->path = prefix_roota(root_directory, "/dev"); |
7f112f50 LP |
531 | m->mode = PRIVATE_DEV; |
532 | m++; | |
533 | } | |
534 | ||
59eeb84b LP |
535 | if (protect_sysctl) { |
536 | m->path = prefix_roota(root_directory, "/proc/sys"); | |
537 | m->mode = READONLY; | |
538 | m++; | |
539 | ||
540 | m->path = prefix_roota(root_directory, "/proc/sysrq-trigger"); | |
541 | m->mode = READONLY; | |
542 | m->ignore = true; /* Not always compiled into the kernel */ | |
543 | m++; | |
544 | ||
545 | m->path = prefix_roota(root_directory, "/sys"); | |
546 | m->mode = READONLY; | |
547 | m++; | |
548 | } | |
549 | ||
550 | if (protect_cgroups != protect_sysctl) { | |
551 | m->path = prefix_roota(root_directory, "/sys/fs/cgroup"); | |
552 | m->mode = protect_cgroups ? READONLY : READWRITE; | |
553 | m++; | |
554 | } | |
555 | ||
3f815163 | 556 | if (protect_home != PROTECT_HOME_NO || protect_system == PROTECT_SYSTEM_STRICT) { |
ee818b89 AC |
557 | const char *home_dir, *run_user_dir, *root_dir; |
558 | ||
3f815163 LP |
559 | /* If protection of $HOME and $XDG_RUNTIME_DIR is requested, then go for it. If we are in |
560 | * strict system protection mode, then also add entries for these directories, but mark them | |
561 | * writable. This is because we want ProtectHome= and ProtectSystem= to be fully orthogonal. */ | |
562 | ||
ee818b89 AC |
563 | home_dir = prefix_roota(root_directory, "/home"); |
564 | home_dir = strjoina("-", home_dir); | |
565 | run_user_dir = prefix_roota(root_directory, "/run/user"); | |
566 | run_user_dir = strjoina("-", run_user_dir); | |
567 | root_dir = prefix_roota(root_directory, "/root"); | |
568 | root_dir = strjoina("-", root_dir); | |
569 | ||
570 | r = append_mounts(&m, STRV_MAKE(home_dir, run_user_dir, root_dir), | |
3f815163 LP |
571 | protect_home == PROTECT_HOME_READ_ONLY ? READONLY : |
572 | protect_home == PROTECT_HOME_YES ? INACCESSIBLE : READWRITE); | |
417116f2 LP |
573 | if (r < 0) |
574 | return r; | |
575 | } | |
576 | ||
3f815163 LP |
577 | if (protect_system == PROTECT_SYSTEM_STRICT) { |
578 | /* In strict mode, we mount everything read-only, except for /proc, /dev, /sys which are the | |
579 | * kernel API VFS, which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
580 | * protect those, and these options should be fully orthogonal. (And of course /home and | |
581 | * friends are also left writable, as ProtectHome= shall manage those, orthogonally, see | |
582 | * above). */ | |
583 | ||
584 | m->path = prefix_roota(root_directory, "/"); | |
585 | m->mode = READONLY; | |
586 | m++; | |
587 | ||
588 | m->path = prefix_roota(root_directory, "/proc"); | |
589 | m->mode = READWRITE; | |
590 | m++; | |
591 | ||
592 | if (!private_dev) { | |
593 | m->path = prefix_roota(root_directory, "/dev"); | |
594 | m->mode = READWRITE; | |
595 | m++; | |
596 | } | |
597 | if (!protect_sysctl) { | |
598 | m->path = prefix_roota(root_directory, "/sys"); | |
599 | m->mode = READWRITE; | |
600 | m++; | |
601 | } | |
602 | ||
603 | } else if (protect_system != PROTECT_SYSTEM_NO) { | |
604 | const char *usr_dir, *boot_dir, *efi_dir, *etc_dir; | |
605 | ||
606 | /* In any other mode we simply mark the relevant three directories ready-only. */ | |
ee818b89 | 607 | |
d38e01dc | 608 | usr_dir = prefix_roota(root_directory, "/usr"); |
ee818b89 AC |
609 | boot_dir = prefix_roota(root_directory, "/boot"); |
610 | boot_dir = strjoina("-", boot_dir); | |
3f815163 LP |
611 | efi_dir = prefix_roota(root_directory, "/efi"); |
612 | efi_dir = strjoina("-", efi_dir); | |
ee818b89 AC |
613 | etc_dir = prefix_roota(root_directory, "/etc"); |
614 | ||
615 | r = append_mounts(&m, protect_system == PROTECT_SYSTEM_FULL | |
3f815163 LP |
616 | ? STRV_MAKE(usr_dir, boot_dir, efi_dir, etc_dir) |
617 | : STRV_MAKE(usr_dir, boot_dir, efi_dir), READONLY); | |
417116f2 LP |
618 | if (r < 0) |
619 | return r; | |
620 | } | |
621 | ||
7ff7394d | 622 | assert(mounts + n == m); |
ac0930c8 | 623 | |
d944dc95 LP |
624 | /* Resolve symlinks manually first, as mount() will always follow them relative to the host's |
625 | * root. Moreover we want to suppress duplicates based on the resolved paths. This of course is a bit | |
626 | * racy. */ | |
627 | r = chase_all_symlinks(root_directory, mounts, &n); | |
628 | if (r < 0) | |
629 | goto finish; | |
630 | ||
7ff7394d | 631 | qsort(mounts, n, sizeof(BindMount), mount_path_compare); |
fe3c2583 | 632 | |
7ff7394d | 633 | drop_duplicates(mounts, &n); |
fe3c2583 | 634 | drop_inaccessible(mounts, &n); |
7648a565 | 635 | drop_nop(mounts, &n); |
15ae422b LP |
636 | } |
637 | ||
d944dc95 LP |
638 | if (unshare(CLONE_NEWNS) < 0) { |
639 | r = -errno; | |
640 | goto finish; | |
641 | } | |
1e4e94c8 | 642 | |
d944dc95 | 643 | if (make_slave) { |
c2c13f2d LP |
644 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
645 | shows up in the parent */ | |
d944dc95 LP |
646 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { |
647 | r = -errno; | |
648 | goto finish; | |
649 | } | |
ee818b89 AC |
650 | } |
651 | ||
652 | if (root_directory) { | |
653 | /* Turn directory into bind mount */ | |
d944dc95 LP |
654 | if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0) { |
655 | r = -errno; | |
656 | goto finish; | |
657 | } | |
ee818b89 | 658 | } |
c2c13f2d | 659 | |
ee818b89 | 660 | if (n > 0) { |
6b7c9f8b LP |
661 | char **blacklist; |
662 | unsigned j; | |
663 | ||
664 | /* First round, add in all special mounts we need */ | |
c2c13f2d LP |
665 | for (m = mounts; m < mounts + n; ++m) { |
666 | r = apply_mount(m, tmp_dir, var_tmp_dir); | |
667 | if (r < 0) | |
d944dc95 | 668 | goto finish; |
c2c13f2d | 669 | } |
15ae422b | 670 | |
6b7c9f8b LP |
671 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
672 | blacklist = newa(char*, n+1); | |
673 | for (j = 0; j < n; j++) | |
674 | blacklist[j] = (char*) mounts[j].path; | |
675 | blacklist[j] = NULL; | |
676 | ||
677 | /* Second round, flip the ro bits if necessary. */ | |
c2c13f2d | 678 | for (m = mounts; m < mounts + n; ++m) { |
6b7c9f8b | 679 | r = make_read_only(m, blacklist); |
c2c13f2d | 680 | if (r < 0) |
d944dc95 | 681 | goto finish; |
c2c13f2d | 682 | } |
15ae422b LP |
683 | } |
684 | ||
ee818b89 AC |
685 | if (root_directory) { |
686 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ | |
687 | r = mount_move_root(root_directory); | |
d944dc95 LP |
688 | if (r < 0) |
689 | goto finish; | |
ee818b89 AC |
690 | } |
691 | ||
c2c13f2d LP |
692 | /* Remount / as the desired mode. Not that this will not |
693 | * reestablish propagation from our side to the host, since | |
694 | * what's disconnected is disconnected. */ | |
d944dc95 LP |
695 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
696 | r = -errno; | |
697 | goto finish; | |
698 | } | |
15ae422b | 699 | |
d944dc95 | 700 | r = 0; |
15ae422b | 701 | |
d944dc95 LP |
702 | finish: |
703 | for (m = mounts; m < mounts + n; m++) | |
704 | free(m->chased); | |
613b411c LP |
705 | |
706 | return r; | |
707 | } | |
708 | ||
709 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { | |
710 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
711 | char bid[SD_ID128_STRING_MAX]; |
712 | sd_id128_t boot_id; | |
713 | int r; | |
613b411c LP |
714 | |
715 | assert(id); | |
716 | assert(prefix); | |
717 | assert(path); | |
718 | ||
6b46ea73 LP |
719 | /* We include the boot id in the directory so that after a |
720 | * reboot we can easily identify obsolete directories. */ | |
721 | ||
722 | r = sd_id128_get_boot(&boot_id); | |
723 | if (r < 0) | |
724 | return r; | |
725 | ||
726 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX", NULL); | |
613b411c LP |
727 | if (!x) |
728 | return -ENOMEM; | |
729 | ||
730 | RUN_WITH_UMASK(0077) | |
731 | if (!mkdtemp(x)) | |
732 | return -errno; | |
733 | ||
734 | RUN_WITH_UMASK(0000) { | |
735 | char *y; | |
736 | ||
63c372cb | 737 | y = strjoina(x, "/tmp"); |
613b411c LP |
738 | |
739 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
740 | return -errno; | |
c17ec25e | 741 | } |
15ae422b | 742 | |
613b411c LP |
743 | *path = x; |
744 | x = NULL; | |
745 | ||
746 | return 0; | |
747 | } | |
748 | ||
749 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
750 | char *a, *b; | |
751 | int r; | |
752 | ||
753 | assert(id); | |
754 | assert(tmp_dir); | |
755 | assert(var_tmp_dir); | |
756 | ||
757 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
758 | if (r < 0) | |
759 | return r; | |
760 | ||
761 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
762 | if (r < 0) { | |
763 | char *t; | |
764 | ||
63c372cb | 765 | t = strjoina(a, "/tmp"); |
613b411c LP |
766 | rmdir(t); |
767 | rmdir(a); | |
768 | ||
769 | free(a); | |
770 | return r; | |
771 | } | |
772 | ||
773 | *tmp_dir = a; | |
774 | *var_tmp_dir = b; | |
775 | ||
776 | return 0; | |
777 | } | |
778 | ||
779 | int setup_netns(int netns_storage_socket[2]) { | |
780 | _cleanup_close_ int netns = -1; | |
3ee897d6 | 781 | int r, q; |
613b411c LP |
782 | |
783 | assert(netns_storage_socket); | |
784 | assert(netns_storage_socket[0] >= 0); | |
785 | assert(netns_storage_socket[1] >= 0); | |
786 | ||
787 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
788 | * namespace reference fd. Whatever process runs this first |
789 | * shall create a new namespace, all others should just join | |
790 | * it. To serialize that we use a file lock on the socket | |
791 | * pair. | |
613b411c LP |
792 | * |
793 | * It's a bit crazy, but hey, works great! */ | |
794 | ||
795 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
796 | return -errno; | |
797 | ||
3ee897d6 LP |
798 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
799 | if (netns == -EAGAIN) { | |
613b411c LP |
800 | /* Nothing stored yet, so let's create a new namespace */ |
801 | ||
802 | if (unshare(CLONE_NEWNET) < 0) { | |
803 | r = -errno; | |
804 | goto fail; | |
805 | } | |
806 | ||
807 | loopback_setup(); | |
808 | ||
809 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
810 | if (netns < 0) { | |
811 | r = -errno; | |
812 | goto fail; | |
813 | } | |
814 | ||
815 | r = 1; | |
613b411c | 816 | |
3ee897d6 LP |
817 | } else if (netns < 0) { |
818 | r = netns; | |
819 | goto fail; | |
613b411c | 820 | |
3ee897d6 LP |
821 | } else { |
822 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
823 | if (setns(netns, CLONE_NEWNET) < 0) { |
824 | r = -errno; | |
825 | goto fail; | |
826 | } | |
827 | ||
828 | r = 0; | |
829 | } | |
830 | ||
3ee897d6 LP |
831 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
832 | if (q < 0) { | |
833 | r = q; | |
613b411c LP |
834 | goto fail; |
835 | } | |
836 | ||
837 | fail: | |
fe048ce5 | 838 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
839 | return r; |
840 | } | |
417116f2 | 841 | |
1b8689f9 LP |
842 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
843 | [PROTECT_HOME_NO] = "no", | |
844 | [PROTECT_HOME_YES] = "yes", | |
845 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
417116f2 LP |
846 | }; |
847 | ||
1b8689f9 LP |
848 | DEFINE_STRING_TABLE_LOOKUP(protect_home, ProtectHome); |
849 | ||
850 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { | |
851 | [PROTECT_SYSTEM_NO] = "no", | |
852 | [PROTECT_SYSTEM_YES] = "yes", | |
853 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 854 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
855 | }; |
856 | ||
857 | DEFINE_STRING_TABLE_LOOKUP(protect_system, ProtectSystem); |