]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
15ae422b LP |
2 | |
3 | #include <errno.h> | |
07630cea | 4 | #include <sched.h> |
15ae422b | 5 | #include <stdio.h> |
07630cea LP |
6 | #include <string.h> |
7 | #include <sys/mount.h> | |
15ae422b | 8 | #include <sys/stat.h> |
07630cea | 9 | #include <unistd.h> |
25e870b5 | 10 | #include <linux/fs.h> |
15ae422b | 11 | |
b5efdb8a | 12 | #include "alloc-util.h" |
10404d52 | 13 | #include "base-filesystem.h" |
7f112f50 | 14 | #include "dev-setup.h" |
3ffd4af2 | 15 | #include "fd-util.h" |
d944dc95 | 16 | #include "fs-util.h" |
e908468b | 17 | #include "label.h" |
915e6d16 | 18 | #include "loop-util.h" |
07630cea LP |
19 | #include "loopback-setup.h" |
20 | #include "missing.h" | |
21 | #include "mkdir.h" | |
4349cd7c | 22 | #include "mount-util.h" |
3ffd4af2 | 23 | #include "namespace.h" |
07630cea | 24 | #include "path-util.h" |
d7b8eec7 | 25 | #include "selinux-util.h" |
2583fbea | 26 | #include "socket-util.h" |
36ce7110 | 27 | #include "stat-util.h" |
8b43440b | 28 | #include "string-table.h" |
07630cea LP |
29 | #include "string-util.h" |
30 | #include "strv.h" | |
affb60b1 | 31 | #include "umask-util.h" |
ee104e11 | 32 | #include "user-util.h" |
07630cea | 33 | #include "util.h" |
15ae422b | 34 | |
737ba3c8 | 35 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
36 | ||
c17ec25e | 37 | typedef enum MountMode { |
15ae422b LP |
38 | /* This is ordered by priority! */ |
39 | INACCESSIBLE, | |
d2d6c096 LP |
40 | BIND_MOUNT, |
41 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 42 | PRIVATE_TMP, |
7f112f50 | 43 | PRIVATE_DEV, |
5d997827 | 44 | BIND_DEV, |
6c47cd7d | 45 | EMPTY_DIR, |
5d997827 LP |
46 | SYSFS, |
47 | PROCFS, | |
48 | READONLY, | |
59eeb84b | 49 | READWRITE, |
2abd4e38 | 50 | TMPFS, |
c17ec25e | 51 | } MountMode; |
15ae422b | 52 | |
34de407a | 53 | typedef struct MountEntry { |
5327c910 | 54 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 55 | MountMode mode:5; |
5327c910 LP |
56 | bool ignore:1; /* Ignore if path does not exist? */ |
57 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 58 | bool read_only:1; /* Shall this mount point be read-only? */ |
088696fe | 59 | bool applied:1; /* Already applied */ |
55fe7432 | 60 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
d2d6c096 LP |
61 | const char *source_const; /* The source path, for bind mounts */ |
62 | char *source_malloc; | |
2abd4e38 YW |
63 | const char *options_const;/* Mount options for tmpfs */ |
64 | char *options_malloc; | |
65 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 66 | unsigned n_followed; |
34de407a | 67 | } MountEntry; |
15ae422b | 68 | |
5d997827 | 69 | /* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted |
3fe91079 | 70 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
71 | static const MountEntry apivfs_table[] = { |
72 | { "/proc", PROCFS, false }, | |
73 | { "/dev", BIND_DEV, false }, | |
74 | { "/sys", SYSFS, false }, | |
75 | }; | |
f471b2af | 76 | |
11a30cec | 77 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 78 | static const MountEntry protect_kernel_tunables_table[] = { |
c6232fb0 | 79 | { "/proc/acpi", READONLY, true }, |
4e2c0a22 | 80 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ |
c6232fb0 LP |
81 | { "/proc/asound", READONLY, true }, |
82 | { "/proc/bus", READONLY, true }, | |
83 | { "/proc/fs", READONLY, true }, | |
84 | { "/proc/irq", READONLY, true }, | |
4e2c0a22 LP |
85 | { "/proc/kallsyms", INACCESSIBLE, true }, |
86 | { "/proc/kcore", INACCESSIBLE, true }, | |
87 | { "/proc/latency_stats", READONLY, true }, | |
88 | { "/proc/mtrr", READONLY, true }, | |
89 | { "/proc/scsi", READONLY, true }, | |
90 | { "/proc/sys", READONLY, false }, | |
91 | { "/proc/sysrq-trigger", READONLY, true }, | |
92 | { "/proc/timer_stats", READONLY, true }, | |
c6232fb0 | 93 | { "/sys", READONLY, false }, |
13a141f0 | 94 | { "/sys/fs/bpf", READONLY, true }, |
c6232fb0 | 95 | { "/sys/fs/cgroup", READWRITE, false }, /* READONLY is set by ProtectControlGroups= option */ |
3a0bf6d6 | 96 | { "/sys/fs/selinux", READWRITE, true }, |
4e2c0a22 LP |
97 | { "/sys/kernel/debug", READONLY, true }, |
98 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
99 | }; |
100 | ||
c575770b | 101 | /* ProtectKernelModules= option */ |
34de407a | 102 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 103 | #if HAVE_SPLIT_USR |
c6232fb0 | 104 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 105 | #endif |
c6232fb0 | 106 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
107 | }; |
108 | ||
b6c432ca DH |
109 | /* |
110 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
111 | * system should be protected by ProtectSystem= | |
112 | */ | |
34de407a | 113 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
114 | { "/home", READONLY, true }, |
115 | { "/run/user", READONLY, true }, | |
116 | { "/root", READONLY, true }, | |
b6c432ca DH |
117 | }; |
118 | ||
e4da7d8c YW |
119 | /* ProtectHome=tmpfs table */ |
120 | static const MountEntry protect_home_tmpfs_table[] = { | |
121 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
122 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
123 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME }, | |
124 | }; | |
125 | ||
b6c432ca | 126 | /* ProtectHome=yes table */ |
34de407a | 127 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
128 | { "/home", INACCESSIBLE, true }, |
129 | { "/run/user", INACCESSIBLE, true }, | |
130 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
131 | }; |
132 | ||
f471b2af | 133 | /* ProtectSystem=yes table */ |
34de407a | 134 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
135 | { "/usr", READONLY, false }, |
136 | { "/boot", READONLY, true }, | |
137 | { "/efi", READONLY, true }, | |
7486f305 AB |
138 | #if HAVE_SPLIT_USR |
139 | { "/lib", READONLY, true }, | |
140 | { "/lib64", READONLY, true }, | |
141 | { "/bin", READONLY, true }, | |
671f0f8d | 142 | # if HAVE_SPLIT_BIN |
7486f305 | 143 | { "/sbin", READONLY, true }, |
671f0f8d | 144 | # endif |
7486f305 | 145 | #endif |
f471b2af DH |
146 | }; |
147 | ||
148 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 149 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
150 | { "/usr", READONLY, false }, |
151 | { "/boot", READONLY, true }, | |
152 | { "/efi", READONLY, true }, | |
153 | { "/etc", READONLY, false }, | |
7486f305 AB |
154 | #if HAVE_SPLIT_USR |
155 | { "/lib", READONLY, true }, | |
156 | { "/lib64", READONLY, true }, | |
157 | { "/bin", READONLY, true }, | |
671f0f8d | 158 | # if HAVE_SPLIT_BIN |
7486f305 | 159 | { "/sbin", READONLY, true }, |
671f0f8d | 160 | # endif |
7486f305 | 161 | #endif |
f471b2af DH |
162 | }; |
163 | ||
164 | /* | |
165 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
166 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
167 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
168 | * protect those, and these options should be fully orthogonal. | |
169 | * (And of course /home and friends are also left writable, as ProtectHome= | |
170 | * shall manage those, orthogonally). | |
171 | */ | |
34de407a | 172 | static const MountEntry protect_system_strict_table[] = { |
ddbe0412 LP |
173 | { "/", READONLY, false }, |
174 | { "/proc", READWRITE, false }, /* ProtectKernelTunables= */ | |
175 | { "/sys", READWRITE, false }, /* ProtectKernelTunables= */ | |
176 | { "/dev", READWRITE, false }, /* PrivateDevices= */ | |
177 | { "/home", READWRITE, true }, /* ProtectHome= */ | |
178 | { "/run/user", READWRITE, true }, /* ProtectHome= */ | |
179 | { "/root", READWRITE, true }, /* ProtectHome= */ | |
f471b2af DH |
180 | }; |
181 | ||
34de407a | 182 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
183 | assert(p); |
184 | ||
5327c910 LP |
185 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
186 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 187 | |
5327c910 | 188 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
189 | } |
190 | ||
34de407a | 191 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
192 | assert(p); |
193 | ||
194 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
195 | } | |
196 | ||
d2d6c096 LP |
197 | static const char *mount_entry_source(const MountEntry *p) { |
198 | assert(p); | |
199 | ||
200 | return p->source_malloc ?: p->source_const; | |
201 | } | |
202 | ||
2abd4e38 YW |
203 | static const char *mount_entry_options(const MountEntry *p) { |
204 | assert(p); | |
205 | ||
206 | return p->options_malloc ?: p->options_const; | |
207 | } | |
208 | ||
1eb7e08e LP |
209 | static void mount_entry_done(MountEntry *p) { |
210 | assert(p); | |
211 | ||
212 | p->path_malloc = mfree(p->path_malloc); | |
213 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 214 | p->options_malloc = mfree(p->options_malloc); |
1eb7e08e LP |
215 | } |
216 | ||
d18aff04 | 217 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
218 | char **i; |
219 | ||
613b411c LP |
220 | assert(p); |
221 | ||
5327c910 LP |
222 | /* Adds a list of user-supplied READWRITE/READONLY/INACCESSIBLE entries */ |
223 | ||
15ae422b | 224 | STRV_FOREACH(i, strv) { |
5327c910 LP |
225 | bool ignore = false, needs_prefix = false; |
226 | const char *e = *i; | |
15ae422b | 227 | |
5327c910 LP |
228 | /* Look for any prefixes */ |
229 | if (startswith(e, "-")) { | |
230 | e++; | |
9c94d52e | 231 | ignore = true; |
ea92ae33 | 232 | } |
5327c910 LP |
233 | if (startswith(e, "+")) { |
234 | e++; | |
235 | needs_prefix = true; | |
236 | } | |
ea92ae33 | 237 | |
763a260a YW |
238 | if (!path_is_absolute(e)) { |
239 | log_debug("Path is not absolute: %s", e); | |
15ae422b | 240 | return -EINVAL; |
763a260a | 241 | } |
15ae422b | 242 | |
34de407a | 243 | *((*p)++) = (MountEntry) { |
5327c910 LP |
244 | .path_const = e, |
245 | .mode = mode, | |
246 | .ignore = ignore, | |
d18aff04 | 247 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 248 | }; |
15ae422b LP |
249 | } |
250 | ||
251 | return 0; | |
252 | } | |
253 | ||
6c47cd7d LP |
254 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
255 | char **i; | |
256 | ||
257 | assert(p); | |
258 | ||
259 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
260 | * "/private/" boundary directories for DynamicUser=1. */ | |
261 | ||
262 | STRV_FOREACH(i, strv) { | |
263 | ||
264 | *((*p)++) = (MountEntry) { | |
265 | .path_const = *i, | |
266 | .mode = EMPTY_DIR, | |
267 | .ignore = false, | |
6c47cd7d | 268 | .read_only = true, |
2abd4e38 YW |
269 | .options_const = "mode=755", |
270 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, | |
6c47cd7d LP |
271 | }; |
272 | } | |
273 | ||
274 | return 0; | |
275 | } | |
276 | ||
da6053d0 LP |
277 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
278 | size_t i; | |
d2d6c096 LP |
279 | |
280 | assert(p); | |
281 | ||
282 | for (i = 0; i < n; i++) { | |
283 | const BindMount *b = binds + i; | |
284 | ||
285 | *((*p)++) = (MountEntry) { | |
286 | .path_const = b->destination, | |
287 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
288 | .read_only = b->read_only, | |
289 | .source_const = b->source, | |
4ca763a9 | 290 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
291 | }; |
292 | } | |
293 | ||
294 | return 0; | |
295 | } | |
296 | ||
da6053d0 LP |
297 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
298 | size_t i; | |
2abd4e38 YW |
299 | int r; |
300 | ||
301 | assert(p); | |
302 | ||
303 | for (i = 0; i < n; i++) { | |
304 | const TemporaryFileSystem *t = tmpfs + i; | |
305 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 306 | unsigned long flags; |
2abd4e38 YW |
307 | bool ro = false; |
308 | ||
763a260a YW |
309 | if (!path_is_absolute(t->path)) { |
310 | log_debug("Path is not absolute: %s", t->path); | |
2abd4e38 | 311 | return -EINVAL; |
763a260a | 312 | } |
2abd4e38 | 313 | |
ad8e66dc AJ |
314 | str = strjoin("mode=0755,", t->options); |
315 | if (!str) | |
316 | return -ENOMEM; | |
2abd4e38 | 317 | |
ad8e66dc AJ |
318 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
319 | if (r < 0) | |
320 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 321 | |
ad8e66dc AJ |
322 | ro = flags & MS_RDONLY; |
323 | if (ro) | |
324 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
325 | |
326 | *((*p)++) = (MountEntry) { | |
327 | .path_const = t->path, | |
328 | .mode = TMPFS, | |
329 | .read_only = ro, | |
ad8e66dc | 330 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
331 | .flags = flags, |
332 | }; | |
2abd4e38 YW |
333 | } |
334 | ||
335 | return 0; | |
336 | } | |
337 | ||
da6053d0 LP |
338 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
339 | size_t i; | |
11a30cec DH |
340 | |
341 | assert(p); | |
f471b2af | 342 | assert(mounts); |
11a30cec | 343 | |
5327c910 | 344 | /* Adds a list of static pre-defined entries */ |
f471b2af | 345 | |
5327c910 | 346 | for (i = 0; i < n; i++) |
34de407a LP |
347 | *((*p)++) = (MountEntry) { |
348 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
349 | .mode = mounts[i].mode, |
350 | .ignore = mounts[i].ignore || ignore_protect, | |
351 | }; | |
f471b2af DH |
352 | |
353 | return 0; | |
354 | } | |
355 | ||
34de407a | 356 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
357 | assert(p); |
358 | ||
5327c910 | 359 | switch (protect_home) { |
b6c432ca | 360 | |
5327c910 | 361 | case PROTECT_HOME_NO: |
b6c432ca DH |
362 | return 0; |
363 | ||
b6c432ca | 364 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
365 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
366 | ||
e4da7d8c YW |
367 | case PROTECT_HOME_TMPFS: |
368 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
369 | ||
b6c432ca | 370 | case PROTECT_HOME_YES: |
5327c910 LP |
371 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
372 | ||
b6c432ca | 373 | default: |
5327c910 | 374 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 375 | } |
b6c432ca DH |
376 | } |
377 | ||
34de407a | 378 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
379 | assert(p); |
380 | ||
5327c910 LP |
381 | switch (protect_system) { |
382 | ||
383 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
384 | return 0; |
385 | ||
f471b2af | 386 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
387 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
388 | ||
f471b2af | 389 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
390 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
391 | ||
f471b2af | 392 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
393 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
394 | ||
f471b2af | 395 | default: |
5327c910 | 396 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 397 | } |
11a30cec DH |
398 | } |
399 | ||
93bab288 | 400 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 401 | int d; |
15ae422b | 402 | |
6ee1a919 | 403 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 404 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
405 | if (d != 0) |
406 | return d; | |
15ae422b | 407 | |
6ee1a919 | 408 | /* If the paths are equal, check the mode */ |
93bab288 | 409 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
410 | } |
411 | ||
da6053d0 LP |
412 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
413 | size_t i; | |
5327c910 | 414 | |
4a756839 | 415 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 LP |
416 | |
417 | for (i = 0; i < n; i++) { | |
418 | char *s; | |
419 | ||
420 | if (m[i].has_prefix) | |
421 | continue; | |
422 | ||
34de407a | 423 | s = prefix_root(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
424 | if (!s) |
425 | return -ENOMEM; | |
426 | ||
e282f51f | 427 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
428 | m[i].has_prefix = true; |
429 | } | |
430 | ||
431 | return 0; | |
432 | } | |
433 | ||
da6053d0 | 434 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 435 | MountEntry *f, *t, *previous; |
15ae422b | 436 | |
c17ec25e | 437 | assert(m); |
15ae422b | 438 | assert(n); |
15ae422b | 439 | |
fe3c2583 LP |
440 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
441 | ||
1d54cd5d | 442 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 443 | |
fe3c2583 | 444 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
445 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
446 | if (previous && | |
447 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
448 | !f->applied && !previous->applied) { | |
34de407a LP |
449 | log_debug("%s is duplicate.", mount_entry_path(f)); |
450 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ | |
1eb7e08e | 451 | mount_entry_done(f); |
15ae422b | 452 | continue; |
fe3c2583 | 453 | } |
15ae422b | 454 | |
e2d7c1a0 | 455 | *t = *f; |
15ae422b | 456 | previous = t; |
fe3c2583 LP |
457 | t++; |
458 | } | |
459 | ||
460 | *n = t - m; | |
461 | } | |
462 | ||
da6053d0 | 463 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 464 | MountEntry *f, *t; |
fe3c2583 LP |
465 | const char *clear = NULL; |
466 | ||
467 | assert(m); | |
468 | assert(n); | |
469 | ||
470 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
471 | * ordered already. */ | |
472 | ||
1d54cd5d | 473 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
474 | |
475 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
476 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
477 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
478 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 479 | mount_entry_done(f); |
fe3c2583 LP |
480 | continue; |
481 | } | |
15ae422b | 482 | |
34de407a | 483 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
484 | |
485 | *t = *f; | |
15ae422b LP |
486 | t++; |
487 | } | |
488 | ||
c17ec25e | 489 | *n = t - m; |
15ae422b LP |
490 | } |
491 | ||
da6053d0 | 492 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 493 | MountEntry *f, *t; |
7648a565 LP |
494 | |
495 | assert(m); | |
496 | assert(n); | |
497 | ||
498 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
499 | * list is ordered by prefixes. */ | |
500 | ||
1d54cd5d | 501 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 LP |
502 | |
503 | /* Only suppress such subtrees for READONLY and READWRITE entries */ | |
504 | if (IN_SET(f->mode, READONLY, READWRITE)) { | |
34de407a | 505 | MountEntry *p; |
7648a565 LP |
506 | bool found = false; |
507 | ||
508 | /* Now let's find the first parent of the entry we are looking at. */ | |
509 | for (p = t-1; p >= m; p--) { | |
34de407a | 510 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
511 | found = true; |
512 | break; | |
513 | } | |
514 | } | |
515 | ||
516 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
517 | if (found && p->mode == f->mode) { | |
34de407a | 518 | log_debug("%s is redundant by %s", mount_entry_path(f), mount_entry_path(p)); |
1eb7e08e | 519 | mount_entry_done(f); |
7648a565 LP |
520 | continue; |
521 | } | |
522 | } | |
523 | ||
524 | *t = *f; | |
525 | t++; | |
526 | } | |
527 | ||
528 | *n = t - m; | |
529 | } | |
530 | ||
da6053d0 | 531 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 532 | MountEntry *f, *t; |
cd2902c9 LP |
533 | |
534 | assert(m); | |
535 | assert(n); | |
536 | ||
1d54cd5d | 537 | /* Nothing to do */ |
cd2902c9 LP |
538 | if (!root_directory) |
539 | return; | |
540 | ||
541 | /* Drops all mounts that are outside of the root directory. */ | |
542 | ||
1d54cd5d | 543 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 544 | |
34de407a LP |
545 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
546 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 547 | mount_entry_done(f); |
cd2902c9 LP |
548 | continue; |
549 | } | |
550 | ||
551 | *t = *f; | |
552 | t++; | |
553 | } | |
554 | ||
555 | *n = t - m; | |
556 | } | |
557 | ||
b2a60844 LP |
558 | static int clone_device_node( |
559 | const char *d, | |
560 | const char *temporary_mount, | |
561 | bool *make_devnode) { | |
562 | ||
563 | _cleanup_free_ char *sl = NULL; | |
564 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
565 | struct stat st; |
566 | int r; | |
567 | ||
414b304b | 568 | if (stat(d, &st) < 0) { |
b2a60844 LP |
569 | if (errno == ENOENT) { |
570 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 571 | return -ENXIO; |
b2a60844 LP |
572 | } |
573 | ||
574 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
575 | } |
576 | ||
577 | if (!S_ISBLK(st.st_mode) && | |
b2a60844 LP |
578 | !S_ISCHR(st.st_mode)) { |
579 | log_debug("Device node '%s' to clone is not a device node, ignoring.", d); | |
b5e99f23 | 580 | return -EINVAL; |
b2a60844 | 581 | } |
b5e99f23 | 582 | |
6f7f3a33 | 583 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 584 | |
b2a60844 | 585 | /* First, try to create device node properly */ |
16498617 CB |
586 | if (*make_devnode) { |
587 | mac_selinux_create_file_prepare(d, st.st_mode); | |
588 | r = mknod(dn, st.st_mode, st.st_rdev); | |
589 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
590 | if (r >= 0) |
591 | goto add_symlink; | |
16498617 CB |
592 | if (errno != EPERM) |
593 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
594 | ||
b2a60844 | 595 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
596 | *make_devnode = false; |
597 | } | |
598 | ||
599 | /* We're about to fallback to bind-mounting the device | |
600 | * node. So create a dummy bind-mount target. */ | |
601 | mac_selinux_create_file_prepare(d, 0); | |
602 | r = mknod(dn, S_IFREG, 0); | |
b5e99f23 | 603 | mac_selinux_create_file_clear(); |
16498617 | 604 | if (r < 0 && errno != EEXIST) |
b2a60844 | 605 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 CB |
606 | |
607 | /* Fallback to bind-mounting: | |
608 | * The assumption here is that all used device nodes carry standard | |
609 | * properties. Specifically, the devices nodes we bind-mount should | |
610 | * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx) | |
611 | * and should not carry ACLs. */ | |
612 | if (mount(d, dn, NULL, MS_BIND, NULL) < 0) | |
b2a60844 LP |
613 | return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d); |
614 | ||
615 | add_symlink: | |
616 | bn = path_startswith(d, "/dev/"); | |
617 | if (!bn) | |
618 | return 0; | |
619 | ||
620 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
621 | if (asprintf(&sl, "%s/dev/%s/%u:%u", temporary_mount, S_ISCHR(st.st_mode) ? "char" : "block", major(st.st_rdev), minor(st.st_rdev)) < 0) | |
622 | return log_oom(); | |
623 | ||
624 | (void) mkdir_parents(sl, 0755); | |
625 | ||
626 | t = strjoina("../", bn); | |
627 | ||
628 | if (symlink(t, sl) < 0) | |
2e4a4fae | 629 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 630 | |
af984e13 | 631 | return 0; |
b5e99f23 ДГ |
632 | } |
633 | ||
5d997827 | 634 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
635 | static const char devnodes[] = |
636 | "/dev/null\0" | |
637 | "/dev/zero\0" | |
638 | "/dev/full\0" | |
639 | "/dev/random\0" | |
640 | "/dev/urandom\0" | |
641 | "/dev/tty\0"; | |
642 | ||
2b85f4e1 | 643 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 644 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 645 | bool can_mknod = true; |
7f112f50 LP |
646 | _cleanup_umask_ mode_t u; |
647 | int r; | |
648 | ||
649 | assert(m); | |
650 | ||
651 | u = umask(0000); | |
652 | ||
2b85f4e1 | 653 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 654 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 655 | |
63c372cb | 656 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 657 | (void) mkdir(dev, 0755); |
737ba3c8 | 658 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2e4a4fae | 659 | r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev); |
2b85f4e1 LP |
660 | goto fail; |
661 | } | |
662 | ||
63c372cb | 663 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 664 | (void) mkdir(devpts, 0755); |
2b85f4e1 | 665 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
2e4a4fae | 666 | r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts); |
2b85f4e1 LP |
667 | goto fail; |
668 | } | |
669 | ||
2e4a4fae YW |
670 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx. |
671 | * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible. | |
672 | * Thus, in that case make a clone. | |
673 | * In nspawn and other containers it will be a symlink, in that case make it a symlink. */ | |
36ce7110 | 674 | r = is_symlink("/dev/ptmx"); |
2e4a4fae YW |
675 | if (r < 0) { |
676 | log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m"); | |
3164e3cb | 677 | goto fail; |
2e4a4fae | 678 | } else if (r > 0) { |
414b304b ДГ |
679 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
680 | if (symlink("pts/ptmx", devptmx) < 0) { | |
2e4a4fae | 681 | r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx); |
414b304b ДГ |
682 | goto fail; |
683 | } | |
684 | } else { | |
16498617 | 685 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
686 | if (r < 0) |
687 | goto fail; | |
414b304b | 688 | } |
e06b6479 | 689 | |
63c372cb | 690 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 691 | (void) mkdir(devshm, 0755); |
2b85f4e1 LP |
692 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
693 | if (r < 0) { | |
2e4a4fae | 694 | r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm); |
2b85f4e1 LP |
695 | goto fail; |
696 | } | |
697 | ||
63c372cb | 698 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 699 | (void) mkdir(devmqueue, 0755); |
2e4a4fae YW |
700 | if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0) |
701 | log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue); | |
2b85f4e1 | 702 | |
63c372cb | 703 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 704 | (void) mkdir(devhugepages, 0755); |
2e4a4fae YW |
705 | if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0) |
706 | log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages); | |
2b85f4e1 | 707 | |
63c372cb | 708 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
709 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
710 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 711 | |
7f112f50 | 712 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 713 | r = clone_device_node(d, temporary_mount, &can_mknod); |
af984e13 ZJS |
714 | /* ENXIO means the the *source* is not a device file, skip creation in that case */ |
715 | if (r < 0 && r != -ENXIO) | |
2b85f4e1 | 716 | goto fail; |
7f112f50 LP |
717 | } |
718 | ||
2e4a4fae YW |
719 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
720 | if (r < 0) | |
721 | log_debug_errno(r, "Failed to setup basic device tree at '%s', ignoring: %m", temporary_mount); | |
7f112f50 | 722 | |
ee818b89 AC |
723 | /* Create the /dev directory if missing. It is more likely to be |
724 | * missing when the service is started with RootDirectory. This is | |
725 | * consistent with mount units creating the mount points when missing. | |
726 | */ | |
34de407a | 727 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 728 | |
9e5f8252 | 729 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
730 | r = umount_recursive(mount_entry_path(m), 0); |
731 | if (r < 0) | |
732 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
733 | ||
34de407a | 734 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { |
2e4a4fae | 735 | r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m)); |
2b85f4e1 LP |
736 | goto fail; |
737 | } | |
7f112f50 | 738 | |
2b85f4e1 LP |
739 | rmdir(dev); |
740 | rmdir(temporary_mount); | |
7f112f50 | 741 | |
2b85f4e1 | 742 | return 0; |
7f112f50 | 743 | |
2b85f4e1 LP |
744 | fail: |
745 | if (devpts) | |
746 | umount(devpts); | |
7f112f50 | 747 | |
2b85f4e1 LP |
748 | if (devshm) |
749 | umount(devshm); | |
7f112f50 | 750 | |
2b85f4e1 LP |
751 | if (devhugepages) |
752 | umount(devhugepages); | |
7f112f50 | 753 | |
2b85f4e1 LP |
754 | if (devmqueue) |
755 | umount(devmqueue); | |
7f112f50 | 756 | |
d267c5aa ZJS |
757 | umount(dev); |
758 | rmdir(dev); | |
2b85f4e1 | 759 | rmdir(temporary_mount); |
7f112f50 | 760 | |
2b85f4e1 | 761 | return r; |
7f112f50 LP |
762 | } |
763 | ||
2a2969fd | 764 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
765 | int r; |
766 | ||
767 | assert(m); | |
768 | ||
769 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
770 | * /dev. This is only used when RootDirectory= is set. */ | |
771 | ||
645767d6 LP |
772 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
773 | ||
5d997827 LP |
774 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
775 | if (r < 0) | |
776 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
777 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
778 | return 0; | |
779 | ||
780 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
781 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
782 | ||
783 | return 1; | |
784 | } | |
785 | ||
2a2969fd | 786 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
787 | int r; |
788 | ||
789 | assert(m); | |
790 | ||
645767d6 LP |
791 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
792 | ||
5d997827 LP |
793 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
794 | if (r < 0) | |
795 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
796 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
797 | return 0; | |
798 | ||
799 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
800 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
801 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
802 | ||
803 | return 1; | |
804 | } | |
805 | ||
2a2969fd | 806 | static int mount_procfs(const MountEntry *m) { |
5d997827 LP |
807 | int r; |
808 | ||
809 | assert(m); | |
810 | ||
645767d6 LP |
811 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
812 | ||
5d997827 LP |
813 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
814 | if (r < 0) | |
815 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
816 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
817 | return 0; | |
818 | ||
819 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
820 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
821 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
822 | ||
823 | return 1; | |
824 | } | |
825 | ||
2abd4e38 | 826 | static int mount_tmpfs(const MountEntry *m) { |
6c47cd7d LP |
827 | assert(m); |
828 | ||
2abd4e38 | 829 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d LP |
830 | |
831 | (void) mkdir_p_label(mount_entry_path(m), 0755); | |
832 | (void) umount_recursive(mount_entry_path(m), 0); | |
833 | ||
2abd4e38 | 834 | if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0) |
6c47cd7d LP |
835 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); |
836 | ||
837 | return 1; | |
838 | } | |
839 | ||
088696fe | 840 | static int follow_symlink( |
d2d6c096 | 841 | const char *root_directory, |
088696fe | 842 | MountEntry *m) { |
d2d6c096 | 843 | |
088696fe | 844 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
845 | int r; |
846 | ||
088696fe LP |
847 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
848 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
849 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
850 | * end and already have a fully normalized name. */ | |
8fceda93 | 851 | |
088696fe LP |
852 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target); |
853 | if (r < 0) | |
854 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
855 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
856 | return 1; | |
8fceda93 | 857 | |
088696fe LP |
858 | if (m->n_followed >= CHASE_SYMLINKS_MAX) { /* put a boundary on things */ |
859 | log_debug("Symlink loop on '%s'.", mount_entry_path(m)); | |
860 | return -ELOOP; | |
8fceda93 | 861 | } |
8fceda93 | 862 | |
088696fe | 863 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 864 | |
088696fe LP |
865 | free_and_replace(m->path_malloc, target); |
866 | m->has_prefix = true; | |
8fceda93 | 867 | |
088696fe LP |
868 | m->n_followed ++; |
869 | ||
870 | return 0; | |
8fceda93 LP |
871 | } |
872 | ||
ac0930c8 | 873 | static int apply_mount( |
8fceda93 | 874 | const char *root_directory, |
89bd586c | 875 | MountEntry *m) { |
ac0930c8 | 876 | |
a227a4be | 877 | bool rbind = true, make = false; |
15ae422b | 878 | const char *what; |
15ae422b | 879 | int r; |
15ae422b | 880 | |
c17ec25e | 881 | assert(m); |
15ae422b | 882 | |
34de407a | 883 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 884 | |
c17ec25e | 885 | switch (m->mode) { |
15ae422b | 886 | |
160cfdbe LP |
887 | case INACCESSIBLE: { |
888 | struct stat target; | |
6d313367 LP |
889 | |
890 | /* First, get rid of everything that is below if there | |
891 | * is anything... Then, overmount it with an | |
c4b41707 | 892 | * inaccessible path. */ |
34de407a | 893 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 894 | |
088696fe LP |
895 | if (lstat(mount_entry_path(m), &target) < 0) { |
896 | if (errno == ENOENT && m->ignore) | |
897 | return 0; | |
898 | ||
34de407a | 899 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); |
088696fe | 900 | } |
15ae422b | 901 | |
c4b41707 | 902 | what = mode_to_inaccessible_node(target.st_mode); |
5fd7cf6f LP |
903 | if (!what) { |
904 | log_debug("File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 AP |
905 | return -ELOOP; |
906 | } | |
907 | break; | |
160cfdbe | 908 | } |
fe3c2583 | 909 | |
15ae422b | 910 | case READONLY: |
15ae422b | 911 | case READWRITE: |
8fceda93 | 912 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
913 | if (r == -ENOENT && m->ignore) |
914 | return 0; | |
d944dc95 | 915 | if (r < 0) |
34de407a | 916 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
917 | 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. */ |
918 | return 0; | |
6b7c9f8b | 919 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 920 | what = mount_entry_path(m); |
6b7c9f8b | 921 | break; |
15ae422b | 922 | |
d2d6c096 LP |
923 | case BIND_MOUNT: |
924 | rbind = false; | |
d2d6c096 | 925 | |
4831981d | 926 | _fallthrough_; |
088696fe LP |
927 | case BIND_MOUNT_RECURSIVE: { |
928 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 929 | |
088696fe LP |
930 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind |
931 | * mount source paths are always relative to the host root, hence we pass NULL as root directory to | |
932 | * chase_symlinks() here. */ | |
933 | ||
934 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased); | |
935 | if (r == -ENOENT && m->ignore) { | |
936 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
937 | return 0; | |
938 | } | |
939 | if (r < 0) | |
940 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
941 | ||
942 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
943 | ||
944 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
945 | |
946 | what = mount_entry_source(m); | |
a227a4be | 947 | make = true; |
d2d6c096 | 948 | break; |
088696fe | 949 | } |
d2d6c096 | 950 | |
6c47cd7d | 951 | case EMPTY_DIR: |
2abd4e38 YW |
952 | case TMPFS: |
953 | return mount_tmpfs(m); | |
6c47cd7d | 954 | |
ac0930c8 | 955 | case PRIVATE_TMP: |
89bd586c | 956 | what = mount_entry_source(m); |
a227a4be | 957 | make = true; |
15ae422b | 958 | break; |
e364ad06 | 959 | |
d6797c92 | 960 | case PRIVATE_DEV: |
5d997827 LP |
961 | return mount_private_dev(m); |
962 | ||
963 | case BIND_DEV: | |
964 | return mount_bind_dev(m); | |
965 | ||
966 | case SYSFS: | |
967 | return mount_sysfs(m); | |
968 | ||
969 | case PROCFS: | |
970 | return mount_procfs(m); | |
d6797c92 | 971 | |
e364ad06 LP |
972 | default: |
973 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
974 | } |
975 | ||
ac0930c8 | 976 | assert(what); |
15ae422b | 977 | |
a227a4be LP |
978 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) { |
979 | bool try_again = false; | |
980 | r = -errno; | |
981 | ||
982 | if (r == -ENOENT && make) { | |
983 | struct stat st; | |
984 | ||
985 | /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */ | |
986 | ||
e8717862 LP |
987 | if (stat(what, &st) < 0) |
988 | log_debug_errno(errno, "Mount point source '%s' is not accessible: %m", what); | |
989 | else { | |
990 | int q; | |
a227a4be LP |
991 | |
992 | (void) mkdir_parents(mount_entry_path(m), 0755); | |
993 | ||
994 | if (S_ISDIR(st.st_mode)) | |
e8717862 | 995 | q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0; |
a227a4be | 996 | else |
e8717862 LP |
997 | q = touch(mount_entry_path(m)); |
998 | ||
999 | if (q < 0) | |
1000 | log_debug_errno(q, "Failed to create destination mount point node '%s': %m", mount_entry_path(m)); | |
1001 | else | |
1002 | try_again = true; | |
a227a4be LP |
1003 | } |
1004 | } | |
1005 | ||
1006 | if (try_again) { | |
1007 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) | |
1008 | r = -errno; | |
1009 | else | |
1010 | r = 0; | |
1011 | } | |
1012 | ||
1013 | if (r < 0) | |
1014 | return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); | |
1015 | } | |
6b7c9f8b | 1016 | |
34de407a | 1017 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1018 | return 0; |
ac0930c8 | 1019 | } |
15ae422b | 1020 | |
69338c3d AJ |
1021 | /* Change the per-mount readonly flag on an existing mount */ |
1022 | static int remount_bind_readonly(const char *path, unsigned long orig_flags) { | |
1023 | int r; | |
1024 | ||
1025 | r = mount(NULL, path, NULL, MS_REMOUNT | MS_BIND | MS_RDONLY | orig_flags, NULL); | |
1026 | ||
1027 | return r < 0 ? -errno : 0; | |
1028 | } | |
1029 | ||
2a2969fd | 1030 | static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) { |
763a260a | 1031 | bool submounts = false; |
6b7c9f8b | 1032 | int r = 0; |
15ae422b | 1033 | |
c17ec25e | 1034 | assert(m); |
ac9de0b3 | 1035 | assert(proc_self_mountinfo); |
ac0930c8 | 1036 | |
2abd4e38 YW |
1037 | if (mount_entry_read_only(m)) { |
1038 | if (IN_SET(m->mode, EMPTY_DIR, TMPFS)) { | |
69338c3d | 1039 | r = remount_bind_readonly(mount_entry_path(m), m->flags); |
763a260a YW |
1040 | } else { |
1041 | submounts = true; | |
2abd4e38 | 1042 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), true, blacklist, proc_self_mountinfo); |
763a260a | 1043 | } |
2abd4e38 | 1044 | } else if (m->mode == PRIVATE_DEV) { |
69338c3d AJ |
1045 | /* Set /dev readonly, but not submounts like /dev/shm. Also, we only set the per-mount read-only flag. |
1046 | * We can't set it on the superblock, if we are inside a user namespace and running Linux <= 4.17. */ | |
1047 | r = remount_bind_readonly(mount_entry_path(m), DEV_MOUNT_OPTIONS); | |
737ba3c8 | 1048 | } else |
6b7c9f8b LP |
1049 | return 0; |
1050 | ||
1051 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
1052 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
1053 | * read-only mounts already applied. */ | |
ac0930c8 | 1054 | |
8fceda93 LP |
1055 | if (r == -ENOENT && m->ignore) |
1056 | r = 0; | |
5327c910 | 1057 | |
763a260a YW |
1058 | if (r < 0) |
1059 | return log_debug_errno(r, "Failed to re-mount '%s'%s read-only: %m", mount_entry_path(m), | |
1060 | submounts ? " and its submounts" : ""); | |
1061 | ||
1062 | return 0; | |
d944dc95 LP |
1063 | } |
1064 | ||
9b68367b | 1065 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1066 | assert(ns_info); |
1067 | ||
9c988f93 DH |
1068 | /* |
1069 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1070 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1071 | * first place... |
9c988f93 | 1072 | */ |
5d997827 | 1073 | |
9b68367b YW |
1074 | return ns_info->mount_apivfs || |
1075 | ns_info->protect_control_groups || | |
1076 | ns_info->protect_kernel_tunables; | |
5d997827 LP |
1077 | } |
1078 | ||
da6053d0 | 1079 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1080 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1081 | char** read_write_paths, |
1082 | char** read_only_paths, | |
1083 | char** inaccessible_paths, | |
6c47cd7d | 1084 | char** empty_directories, |
da6053d0 LP |
1085 | size_t n_bind_mounts, |
1086 | size_t n_temporary_filesystems, | |
2652c6c1 DH |
1087 | const char* tmp_dir, |
1088 | const char* var_tmp_dir, | |
2652c6c1 DH |
1089 | ProtectHome protect_home, |
1090 | ProtectSystem protect_system) { | |
1091 | ||
da6053d0 LP |
1092 | size_t protect_home_cnt; |
1093 | size_t protect_system_cnt = | |
f471b2af DH |
1094 | (protect_system == PROTECT_SYSTEM_STRICT ? |
1095 | ELEMENTSOF(protect_system_strict_table) : | |
1096 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
1097 | ELEMENTSOF(protect_system_full_table) : | |
1098 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
1099 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
1100 | ||
b6c432ca DH |
1101 | protect_home_cnt = |
1102 | (protect_home == PROTECT_HOME_YES ? | |
1103 | ELEMENTSOF(protect_home_yes_table) : | |
1104 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
e4da7d8c YW |
1105 | ELEMENTSOF(protect_home_read_only_table) : |
1106 | ((protect_home == PROTECT_HOME_TMPFS) ? | |
1107 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); | |
b6c432ca | 1108 | |
2652c6c1 DH |
1109 | return !!tmp_dir + !!var_tmp_dir + |
1110 | strv_length(read_write_paths) + | |
1111 | strv_length(read_only_paths) + | |
1112 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1113 | strv_length(empty_directories) + |
d2d6c096 | 1114 | n_bind_mounts + |
2abd4e38 | 1115 | n_temporary_filesystems + |
c575770b DH |
1116 | ns_info->private_dev + |
1117 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
1118 | (ns_info->protect_control_groups ? 1 : 0) + | |
1119 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + | |
5d997827 | 1120 | protect_home_cnt + protect_system_cnt + |
9b68367b | 1121 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0); |
2652c6c1 DH |
1122 | } |
1123 | ||
da6053d0 | 1124 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1125 | assert(root_directory); |
f8b64b57 LP |
1126 | assert(n_mounts); |
1127 | assert(mounts || *n_mounts == 0); | |
1128 | ||
93bab288 | 1129 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1130 | |
1131 | drop_duplicates(mounts, n_mounts); | |
1132 | drop_outside_root(root_directory, mounts, n_mounts); | |
1133 | drop_inaccessible(mounts, n_mounts); | |
1134 | drop_nop(mounts, n_mounts); | |
1135 | } | |
1136 | ||
613b411c | 1137 | int setup_namespace( |
ee818b89 | 1138 | const char* root_directory, |
915e6d16 | 1139 | const char* root_image, |
bb0ff3fb | 1140 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1141 | char** read_write_paths, |
1142 | char** read_only_paths, | |
1143 | char** inaccessible_paths, | |
6c47cd7d | 1144 | char** empty_directories, |
d2d6c096 | 1145 | const BindMount *bind_mounts, |
da6053d0 | 1146 | size_t n_bind_mounts, |
2abd4e38 | 1147 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1148 | size_t n_temporary_filesystems, |
a004cb4c LP |
1149 | const char* tmp_dir, |
1150 | const char* var_tmp_dir, | |
1b8689f9 LP |
1151 | ProtectHome protect_home, |
1152 | ProtectSystem protect_system, | |
915e6d16 LP |
1153 | unsigned long mount_flags, |
1154 | DissectImageFlags dissect_image_flags) { | |
15ae422b | 1155 | |
915e6d16 | 1156 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1157 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1158 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 1159 | _cleanup_free_ void *root_hash = NULL; |
34de407a | 1160 | MountEntry *m, *mounts = NULL; |
9b68367b | 1161 | size_t n_mounts, root_hash_size = 0; |
d18aff04 | 1162 | bool require_prefix = false; |
9b68367b | 1163 | const char *root; |
c17ec25e | 1164 | int r = 0; |
15ae422b | 1165 | |
915e6d16 LP |
1166 | assert(ns_info); |
1167 | ||
613b411c | 1168 | if (mount_flags == 0) |
c17ec25e | 1169 | mount_flags = MS_SHARED; |
ac0930c8 | 1170 | |
915e6d16 LP |
1171 | if (root_image) { |
1172 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1173 | ||
c9ef8573 LP |
1174 | if (protect_system == PROTECT_SYSTEM_STRICT && |
1175 | protect_home != PROTECT_HOME_NO && | |
1176 | strv_isempty(read_write_paths)) | |
915e6d16 LP |
1177 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1178 | ||
1179 | r = loop_device_make_by_path(root_image, | |
1180 | dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR, | |
1181 | &loop_device); | |
1182 | if (r < 0) | |
763a260a | 1183 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
915e6d16 | 1184 | |
78ebe980 LP |
1185 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
1186 | if (r < 0) | |
763a260a | 1187 | return log_debug_errno(r, "Failed to load root hash: %m"); |
78ebe980 LP |
1188 | |
1189 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
1190 | if (r < 0) | |
763a260a | 1191 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 LP |
1192 | |
1193 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 | 1194 | if (r < 0) |
763a260a | 1195 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1196 | } |
1197 | ||
e908468b LP |
1198 | if (root_directory) |
1199 | root = root_directory; | |
0722b359 JS |
1200 | else { |
1201 | /* Always create the mount namespace in a temporary directory, instead of operating | |
1202 | * directly in the root. The temporary directory prevents any mounts from being | |
1203 | * potentially obscured my other mounts we already applied. | |
1204 | * We use the same mount point for all images, which is safe, since they all live | |
1205 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1206 | |
1207 | root = "/run/systemd/unit-root"; | |
1208 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1209 | require_prefix = true; |
0722b359 | 1210 | } |
e908468b | 1211 | |
cfbeb4ef LP |
1212 | n_mounts = namespace_calculate_mounts( |
1213 | ns_info, | |
1214 | read_write_paths, | |
1215 | read_only_paths, | |
1216 | inaccessible_paths, | |
6c47cd7d | 1217 | empty_directories, |
f5c52a77 | 1218 | n_bind_mounts, |
2abd4e38 | 1219 | n_temporary_filesystems, |
cfbeb4ef LP |
1220 | tmp_dir, var_tmp_dir, |
1221 | protect_home, protect_system); | |
613b411c | 1222 | |
f0a4feb0 | 1223 | if (n_mounts > 0) { |
34de407a | 1224 | m = mounts = (MountEntry *) alloca0(n_mounts * sizeof(MountEntry)); |
d18aff04 | 1225 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1226 | if (r < 0) |
f0a4feb0 | 1227 | goto finish; |
613b411c | 1228 | |
d18aff04 | 1229 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1230 | if (r < 0) |
f0a4feb0 | 1231 | goto finish; |
613b411c | 1232 | |
d18aff04 | 1233 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1234 | if (r < 0) |
f0a4feb0 | 1235 | goto finish; |
7ff7394d | 1236 | |
6c47cd7d LP |
1237 | r = append_empty_dir_mounts(&m, empty_directories); |
1238 | if (r < 0) | |
1239 | goto finish; | |
1240 | ||
d2d6c096 LP |
1241 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1242 | if (r < 0) | |
1243 | goto finish; | |
1244 | ||
2abd4e38 YW |
1245 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1246 | if (r < 0) | |
1247 | goto finish; | |
1248 | ||
613b411c | 1249 | if (tmp_dir) { |
34de407a | 1250 | *(m++) = (MountEntry) { |
5327c910 LP |
1251 | .path_const = "/tmp", |
1252 | .mode = PRIVATE_TMP, | |
89bd586c | 1253 | .source_const = tmp_dir, |
5327c910 | 1254 | }; |
613b411c | 1255 | } |
7ff7394d | 1256 | |
613b411c | 1257 | if (var_tmp_dir) { |
34de407a | 1258 | *(m++) = (MountEntry) { |
5327c910 | 1259 | .path_const = "/var/tmp", |
89bd586c YW |
1260 | .mode = PRIVATE_TMP, |
1261 | .source_const = var_tmp_dir, | |
5327c910 | 1262 | }; |
7ff7394d | 1263 | } |
ac0930c8 | 1264 | |
c575770b | 1265 | if (ns_info->private_dev) { |
34de407a | 1266 | *(m++) = (MountEntry) { |
5327c910 LP |
1267 | .path_const = "/dev", |
1268 | .mode = PRIVATE_DEV, | |
1269 | }; | |
7f112f50 LP |
1270 | } |
1271 | ||
c575770b | 1272 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 1273 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 1274 | if (r < 0) |
f0a4feb0 | 1275 | goto finish; |
c575770b DH |
1276 | } |
1277 | ||
1278 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 1279 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 1280 | if (r < 0) |
f0a4feb0 | 1281 | goto finish; |
c575770b | 1282 | } |
59eeb84b | 1283 | |
c575770b | 1284 | if (ns_info->protect_control_groups) { |
34de407a | 1285 | *(m++) = (MountEntry) { |
5327c910 LP |
1286 | .path_const = "/sys/fs/cgroup", |
1287 | .mode = READONLY, | |
1288 | }; | |
59eeb84b LP |
1289 | } |
1290 | ||
5327c910 | 1291 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1292 | if (r < 0) |
f0a4feb0 | 1293 | goto finish; |
417116f2 | 1294 | |
5327c910 | 1295 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1296 | if (r < 0) |
f0a4feb0 | 1297 | goto finish; |
417116f2 | 1298 | |
9b68367b | 1299 | if (namespace_info_mount_apivfs(ns_info)) { |
5d997827 LP |
1300 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); |
1301 | if (r < 0) | |
1302 | goto finish; | |
1303 | } | |
1304 | ||
f0a4feb0 | 1305 | assert(mounts + n_mounts == m); |
ac0930c8 | 1306 | |
5327c910 | 1307 | /* Prepend the root directory where that's necessary */ |
e908468b | 1308 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1309 | if (r < 0) |
1310 | goto finish; | |
1311 | ||
839f1877 | 1312 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1313 | } |
1314 | ||
1beab8b0 LP |
1315 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1316 | ||
d944dc95 | 1317 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1318 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 LP |
1319 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
1320 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place | |
1321 | * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable | |
1322 | * error back, which the caller can use to detect this case (and only this) and optionally | |
1323 | * continue without namespacing applied. */ | |
1324 | r = -ENOANO; | |
1325 | ||
d944dc95 LP |
1326 | goto finish; |
1327 | } | |
1e4e94c8 | 1328 | |
9b68367b YW |
1329 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1330 | * shows up in the parent */ | |
1331 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1332 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1333 | goto finish; |
ee818b89 AC |
1334 | } |
1335 | ||
915e6d16 | 1336 | if (root_image) { |
e908468b | 1337 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1338 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1339 | if (r < 0) { |
1340 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1341 | goto finish; |
763a260a | 1342 | } |
915e6d16 | 1343 | |
07ce7407 TM |
1344 | if (decrypted_image) { |
1345 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1346 | if (r < 0) { |
1347 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1348 | goto finish; |
763a260a | 1349 | } |
07ce7407 | 1350 | } |
78ebe980 | 1351 | |
915e6d16 LP |
1352 | loop_device_relinquish(loop_device); |
1353 | ||
1354 | } else if (root_directory) { | |
1355 | ||
e908468b LP |
1356 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1357 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1358 | if (r < 0) { |
1359 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1360 | goto finish; |
763a260a | 1361 | } |
8f1ad200 | 1362 | if (r == 0) { |
e908468b | 1363 | if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) { |
763a260a | 1364 | r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root); |
8f1ad200 LP |
1365 | goto finish; |
1366 | } | |
d944dc95 | 1367 | } |
e908468b | 1368 | |
9b68367b | 1369 | } else { |
e908468b LP |
1370 | |
1371 | /* Let's mount the main root directory to the root directory to use */ | |
1372 | if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
763a260a | 1373 | r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root); |
e908468b LP |
1374 | goto finish; |
1375 | } | |
ee818b89 | 1376 | } |
c2c13f2d | 1377 | |
4e0c20de LP |
1378 | /* Try to set up the new root directory before mounting anything else there. */ |
1379 | if (root_image || root_directory) | |
1380 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1381 | ||
f0a4feb0 | 1382 | if (n_mounts > 0) { |
ac9de0b3 | 1383 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b7c9f8b | 1384 | char **blacklist; |
da6053d0 | 1385 | size_t j; |
6b7c9f8b | 1386 | |
ac9de0b3 TR |
1387 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc. |
1388 | * For example, this is the case with the option: 'InaccessiblePaths=/proc' */ | |
1389 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); | |
1390 | if (!proc_self_mountinfo) { | |
763a260a | 1391 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
ac9de0b3 TR |
1392 | goto finish; |
1393 | } | |
1394 | ||
088696fe LP |
1395 | /* First round, establish all mounts we need */ |
1396 | for (;;) { | |
1397 | bool again = false; | |
1398 | ||
1399 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1400 | ||
1401 | if (m->applied) | |
1402 | continue; | |
1403 | ||
1404 | r = follow_symlink(root, m); | |
1405 | if (r < 0) | |
1406 | goto finish; | |
1407 | if (r == 0) { | |
1408 | /* We hit a symlinked mount point. The entry got rewritten and might point to a | |
1409 | * very different place now. Let's normalize the changed list, and start from | |
1410 | * the beginning. After all to mount the entry at the new location we might | |
1411 | * need some other mounts first */ | |
1412 | again = true; | |
1413 | break; | |
1414 | } | |
1415 | ||
1416 | r = apply_mount(root, m); | |
1417 | if (r < 0) | |
1418 | goto finish; | |
1419 | ||
1420 | m->applied = true; | |
1421 | } | |
1422 | ||
1423 | if (!again) | |
1424 | break; | |
1425 | ||
839f1877 | 1426 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1427 | } |
15ae422b | 1428 | |
6b7c9f8b | 1429 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
f0a4feb0 DH |
1430 | blacklist = newa(char*, n_mounts+1); |
1431 | for (j = 0; j < n_mounts; j++) | |
34de407a | 1432 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1433 | blacklist[j] = NULL; |
1434 | ||
1435 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1436 | for (m = mounts; m < mounts + n_mounts; ++m) { |
ac9de0b3 | 1437 | r = make_read_only(m, blacklist, proc_self_mountinfo); |
c2c13f2d | 1438 | if (r < 0) |
d944dc95 | 1439 | goto finish; |
c2c13f2d | 1440 | } |
15ae422b LP |
1441 | } |
1442 | ||
9b68367b YW |
1443 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1444 | r = mount_move_root(root); | |
763a260a YW |
1445 | if (r < 0) { |
1446 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1447 | goto finish; |
763a260a | 1448 | } |
ee818b89 | 1449 | |
55fe7432 | 1450 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1451 | * reestablish propagation from our side to the host, since |
1452 | * what's disconnected is disconnected. */ | |
d944dc95 | 1453 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1454 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1455 | goto finish; |
1456 | } | |
15ae422b | 1457 | |
d944dc95 | 1458 | r = 0; |
15ae422b | 1459 | |
d944dc95 | 1460 | finish: |
f0a4feb0 | 1461 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1462 | mount_entry_done(m); |
613b411c LP |
1463 | |
1464 | return r; | |
1465 | } | |
1466 | ||
da6053d0 LP |
1467 | void bind_mount_free_many(BindMount *b, size_t n) { |
1468 | size_t i; | |
d2d6c096 LP |
1469 | |
1470 | assert(b || n == 0); | |
1471 | ||
1472 | for (i = 0; i < n; i++) { | |
1473 | free(b[i].source); | |
1474 | free(b[i].destination); | |
1475 | } | |
1476 | ||
1477 | free(b); | |
1478 | } | |
1479 | ||
da6053d0 | 1480 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
1481 | _cleanup_free_ char *s = NULL, *d = NULL; |
1482 | BindMount *c; | |
1483 | ||
1484 | assert(b); | |
1485 | assert(n); | |
1486 | assert(item); | |
1487 | ||
1488 | s = strdup(item->source); | |
1489 | if (!s) | |
1490 | return -ENOMEM; | |
1491 | ||
1492 | d = strdup(item->destination); | |
1493 | if (!d) | |
1494 | return -ENOMEM; | |
1495 | ||
aa484f35 | 1496 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
1497 | if (!c) |
1498 | return -ENOMEM; | |
1499 | ||
1500 | *b = c; | |
1501 | ||
1502 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
1503 | .source = TAKE_PTR(s), |
1504 | .destination = TAKE_PTR(d), | |
d2d6c096 LP |
1505 | .read_only = item->read_only, |
1506 | .recursive = item->recursive, | |
1507 | .ignore_enoent = item->ignore_enoent, | |
1508 | }; | |
1509 | ||
d2d6c096 LP |
1510 | return 0; |
1511 | } | |
1512 | ||
da6053d0 LP |
1513 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
1514 | size_t i; | |
2abd4e38 YW |
1515 | |
1516 | assert(t || n == 0); | |
1517 | ||
1518 | for (i = 0; i < n; i++) { | |
1519 | free(t[i].path); | |
1520 | free(t[i].options); | |
1521 | } | |
1522 | ||
1523 | free(t); | |
1524 | } | |
1525 | ||
1526 | int temporary_filesystem_add( | |
1527 | TemporaryFileSystem **t, | |
da6053d0 | 1528 | size_t *n, |
2abd4e38 YW |
1529 | const char *path, |
1530 | const char *options) { | |
1531 | ||
1532 | _cleanup_free_ char *p = NULL, *o = NULL; | |
1533 | TemporaryFileSystem *c; | |
1534 | ||
1535 | assert(t); | |
1536 | assert(n); | |
1537 | assert(path); | |
1538 | ||
1539 | p = strdup(path); | |
1540 | if (!p) | |
1541 | return -ENOMEM; | |
1542 | ||
1543 | if (!isempty(options)) { | |
1544 | o = strdup(options); | |
1545 | if (!o) | |
1546 | return -ENOMEM; | |
1547 | } | |
1548 | ||
aa484f35 | 1549 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
1550 | if (!c) |
1551 | return -ENOMEM; | |
1552 | ||
1553 | *t = c; | |
1554 | ||
1555 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
1556 | .path = TAKE_PTR(p), |
1557 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
1558 | }; |
1559 | ||
2abd4e38 YW |
1560 | return 0; |
1561 | } | |
1562 | ||
613b411c LP |
1563 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1564 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1565 | char bid[SD_ID128_STRING_MAX]; |
1566 | sd_id128_t boot_id; | |
1567 | int r; | |
613b411c LP |
1568 | |
1569 | assert(id); | |
1570 | assert(prefix); | |
1571 | assert(path); | |
1572 | ||
6b46ea73 LP |
1573 | /* We include the boot id in the directory so that after a |
1574 | * reboot we can easily identify obsolete directories. */ | |
1575 | ||
1576 | r = sd_id128_get_boot(&boot_id); | |
1577 | if (r < 0) | |
1578 | return r; | |
1579 | ||
605405c6 | 1580 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1581 | if (!x) |
1582 | return -ENOMEM; | |
1583 | ||
1584 | RUN_WITH_UMASK(0077) | |
1585 | if (!mkdtemp(x)) | |
1586 | return -errno; | |
1587 | ||
1588 | RUN_WITH_UMASK(0000) { | |
1589 | char *y; | |
1590 | ||
63c372cb | 1591 | y = strjoina(x, "/tmp"); |
613b411c LP |
1592 | |
1593 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1594 | return -errno; | |
c17ec25e | 1595 | } |
15ae422b | 1596 | |
1cc6c93a | 1597 | *path = TAKE_PTR(x); |
613b411c LP |
1598 | |
1599 | return 0; | |
1600 | } | |
1601 | ||
1602 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1603 | char *a, *b; | |
1604 | int r; | |
1605 | ||
1606 | assert(id); | |
1607 | assert(tmp_dir); | |
1608 | assert(var_tmp_dir); | |
1609 | ||
1610 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1611 | if (r < 0) | |
1612 | return r; | |
1613 | ||
1614 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1615 | if (r < 0) { | |
1616 | char *t; | |
1617 | ||
63c372cb | 1618 | t = strjoina(a, "/tmp"); |
613b411c LP |
1619 | rmdir(t); |
1620 | rmdir(a); | |
1621 | ||
1622 | free(a); | |
1623 | return r; | |
1624 | } | |
1625 | ||
1626 | *tmp_dir = a; | |
1627 | *var_tmp_dir = b; | |
1628 | ||
1629 | return 0; | |
1630 | } | |
1631 | ||
1632 | int setup_netns(int netns_storage_socket[2]) { | |
1633 | _cleanup_close_ int netns = -1; | |
3ee897d6 | 1634 | int r, q; |
613b411c LP |
1635 | |
1636 | assert(netns_storage_socket); | |
1637 | assert(netns_storage_socket[0] >= 0); | |
1638 | assert(netns_storage_socket[1] >= 0); | |
1639 | ||
1640 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1641 | * namespace reference fd. Whatever process runs this first |
1642 | * shall create a new namespace, all others should just join | |
1643 | * it. To serialize that we use a file lock on the socket | |
1644 | * pair. | |
613b411c LP |
1645 | * |
1646 | * It's a bit crazy, but hey, works great! */ | |
1647 | ||
1648 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1649 | return -errno; | |
1650 | ||
3ee897d6 LP |
1651 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1652 | if (netns == -EAGAIN) { | |
613b411c LP |
1653 | /* Nothing stored yet, so let's create a new namespace */ |
1654 | ||
1655 | if (unshare(CLONE_NEWNET) < 0) { | |
1656 | r = -errno; | |
1657 | goto fail; | |
1658 | } | |
1659 | ||
1660 | loopback_setup(); | |
1661 | ||
1662 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1663 | if (netns < 0) { | |
1664 | r = -errno; | |
1665 | goto fail; | |
1666 | } | |
1667 | ||
1668 | r = 1; | |
613b411c | 1669 | |
3ee897d6 LP |
1670 | } else if (netns < 0) { |
1671 | r = netns; | |
1672 | goto fail; | |
613b411c | 1673 | |
3ee897d6 LP |
1674 | } else { |
1675 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1676 | if (setns(netns, CLONE_NEWNET) < 0) { |
1677 | r = -errno; | |
1678 | goto fail; | |
1679 | } | |
1680 | ||
1681 | r = 0; | |
1682 | } | |
1683 | ||
3ee897d6 LP |
1684 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1685 | if (q < 0) { | |
1686 | r = q; | |
613b411c LP |
1687 | goto fail; |
1688 | } | |
1689 | ||
1690 | fail: | |
fe048ce5 | 1691 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1692 | return r; |
1693 | } | |
417116f2 | 1694 | |
6e2d7c4f MS |
1695 | bool ns_type_supported(NamespaceType type) { |
1696 | const char *t, *ns_proc; | |
1697 | ||
0fa5b831 LP |
1698 | t = namespace_type_to_string(type); |
1699 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
1700 | return false; |
1701 | ||
6e2d7c4f | 1702 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
1703 | return access(ns_proc, F_OK) == 0; |
1704 | } | |
1705 | ||
1b8689f9 LP |
1706 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1707 | [PROTECT_HOME_NO] = "no", | |
1708 | [PROTECT_HOME_YES] = "yes", | |
1709 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
e4da7d8c | 1710 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
1711 | }; |
1712 | ||
1e8c7bd5 | 1713 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 1714 | |
1b8689f9 LP |
1715 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
1716 | [PROTECT_SYSTEM_NO] = "no", | |
1717 | [PROTECT_SYSTEM_YES] = "yes", | |
1718 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1719 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1720 | }; |
1721 | ||
1e8c7bd5 | 1722 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 1723 | |
6e2d7c4f MS |
1724 | static const char* const namespace_type_table[] = { |
1725 | [NAMESPACE_MOUNT] = "mnt", | |
1726 | [NAMESPACE_CGROUP] = "cgroup", | |
1727 | [NAMESPACE_UTS] = "uts", | |
1728 | [NAMESPACE_IPC] = "ipc", | |
1729 | [NAMESPACE_USER] = "user", | |
1730 | [NAMESPACE_PID] = "pid", | |
1731 | [NAMESPACE_NET] = "net", | |
1732 | }; | |
1733 | ||
1734 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); |