]>
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" |
049af8ad | 23 | #include "mountpoint-util.h" |
0cb8e3d1 | 24 | #include "namespace-util.h" |
3ffd4af2 | 25 | #include "namespace.h" |
07630cea | 26 | #include "path-util.h" |
d7b8eec7 | 27 | #include "selinux-util.h" |
2583fbea | 28 | #include "socket-util.h" |
36ce7110 | 29 | #include "stat-util.h" |
8b43440b | 30 | #include "string-table.h" |
07630cea LP |
31 | #include "string-util.h" |
32 | #include "strv.h" | |
affb60b1 | 33 | #include "umask-util.h" |
ee104e11 | 34 | #include "user-util.h" |
07630cea | 35 | #include "util.h" |
15ae422b | 36 | |
737ba3c8 | 37 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
38 | ||
c17ec25e | 39 | typedef enum MountMode { |
15ae422b LP |
40 | /* This is ordered by priority! */ |
41 | INACCESSIBLE, | |
d2d6c096 LP |
42 | BIND_MOUNT, |
43 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 44 | PRIVATE_TMP, |
7f112f50 | 45 | PRIVATE_DEV, |
5d997827 | 46 | BIND_DEV, |
6c47cd7d | 47 | EMPTY_DIR, |
5d997827 LP |
48 | SYSFS, |
49 | PROCFS, | |
50 | READONLY, | |
59eeb84b | 51 | READWRITE, |
2abd4e38 | 52 | TMPFS, |
1e05071d | 53 | READWRITE_IMPLICIT, /* Should have the lowest priority. */ |
5beb8688 | 54 | _MOUNT_MODE_MAX, |
c17ec25e | 55 | } MountMode; |
15ae422b | 56 | |
34de407a | 57 | typedef struct MountEntry { |
5327c910 | 58 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 59 | MountMode mode:5; |
5327c910 LP |
60 | bool ignore:1; /* Ignore if path does not exist? */ |
61 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 62 | bool read_only:1; /* Shall this mount point be read-only? */ |
088696fe | 63 | bool applied:1; /* Already applied */ |
55fe7432 | 64 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
d2d6c096 LP |
65 | const char *source_const; /* The source path, for bind mounts */ |
66 | char *source_malloc; | |
2abd4e38 YW |
67 | const char *options_const;/* Mount options for tmpfs */ |
68 | char *options_malloc; | |
69 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 70 | unsigned n_followed; |
34de407a | 71 | } MountEntry; |
15ae422b | 72 | |
5d997827 | 73 | /* 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 | 74 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
75 | static const MountEntry apivfs_table[] = { |
76 | { "/proc", PROCFS, false }, | |
77 | { "/dev", BIND_DEV, false }, | |
78 | { "/sys", SYSFS, false }, | |
79 | }; | |
f471b2af | 80 | |
11a30cec | 81 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 82 | static const MountEntry protect_kernel_tunables_table[] = { |
1e05071d YW |
83 | { "/proc/acpi", READONLY, true }, |
84 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ | |
85 | { "/proc/asound", READONLY, true }, | |
86 | { "/proc/bus", READONLY, true }, | |
87 | { "/proc/fs", READONLY, true }, | |
88 | { "/proc/irq", READONLY, true }, | |
89 | { "/proc/kallsyms", INACCESSIBLE, true }, | |
90 | { "/proc/kcore", INACCESSIBLE, true }, | |
91 | { "/proc/latency_stats", READONLY, true }, | |
92 | { "/proc/mtrr", READONLY, true }, | |
93 | { "/proc/scsi", READONLY, true }, | |
94 | { "/proc/sys", READONLY, false }, | |
95 | { "/proc/sysrq-trigger", READONLY, true }, | |
96 | { "/proc/timer_stats", READONLY, true }, | |
97 | { "/sys", READONLY, false }, | |
98 | { "/sys/fs/bpf", READONLY, true }, | |
99 | { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */ | |
100 | { "/sys/fs/selinux", READWRITE_IMPLICIT, true }, | |
101 | { "/sys/kernel/debug", READONLY, true }, | |
102 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
103 | }; |
104 | ||
c575770b | 105 | /* ProtectKernelModules= option */ |
34de407a | 106 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 107 | #if HAVE_SPLIT_USR |
c6232fb0 | 108 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 109 | #endif |
c6232fb0 | 110 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
111 | }; |
112 | ||
b6c432ca DH |
113 | /* |
114 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
115 | * system should be protected by ProtectSystem= | |
116 | */ | |
34de407a | 117 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
118 | { "/home", READONLY, true }, |
119 | { "/run/user", READONLY, true }, | |
120 | { "/root", READONLY, true }, | |
b6c432ca DH |
121 | }; |
122 | ||
e4da7d8c YW |
123 | /* ProtectHome=tmpfs table */ |
124 | static const MountEntry protect_home_tmpfs_table[] = { | |
125 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
126 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
127 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME }, | |
128 | }; | |
129 | ||
b6c432ca | 130 | /* ProtectHome=yes table */ |
34de407a | 131 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
132 | { "/home", INACCESSIBLE, true }, |
133 | { "/run/user", INACCESSIBLE, true }, | |
134 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
135 | }; |
136 | ||
f471b2af | 137 | /* ProtectSystem=yes table */ |
34de407a | 138 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
139 | { "/usr", READONLY, false }, |
140 | { "/boot", READONLY, true }, | |
141 | { "/efi", READONLY, true }, | |
7486f305 AB |
142 | #if HAVE_SPLIT_USR |
143 | { "/lib", READONLY, true }, | |
144 | { "/lib64", READONLY, true }, | |
145 | { "/bin", READONLY, true }, | |
671f0f8d | 146 | # if HAVE_SPLIT_BIN |
7486f305 | 147 | { "/sbin", READONLY, true }, |
671f0f8d | 148 | # endif |
7486f305 | 149 | #endif |
f471b2af DH |
150 | }; |
151 | ||
152 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 153 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
154 | { "/usr", READONLY, false }, |
155 | { "/boot", READONLY, true }, | |
156 | { "/efi", READONLY, true }, | |
157 | { "/etc", READONLY, false }, | |
7486f305 AB |
158 | #if HAVE_SPLIT_USR |
159 | { "/lib", READONLY, true }, | |
160 | { "/lib64", READONLY, true }, | |
161 | { "/bin", READONLY, true }, | |
671f0f8d | 162 | # if HAVE_SPLIT_BIN |
7486f305 | 163 | { "/sbin", READONLY, true }, |
671f0f8d | 164 | # endif |
7486f305 | 165 | #endif |
f471b2af DH |
166 | }; |
167 | ||
168 | /* | |
169 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
170 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
171 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
172 | * protect those, and these options should be fully orthogonal. | |
173 | * (And of course /home and friends are also left writable, as ProtectHome= | |
174 | * shall manage those, orthogonally). | |
175 | */ | |
34de407a | 176 | static const MountEntry protect_system_strict_table[] = { |
1e05071d YW |
177 | { "/", READONLY, false }, |
178 | { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
179 | { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
180 | { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */ | |
181 | { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
182 | { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
183 | { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
f471b2af DH |
184 | }; |
185 | ||
5beb8688 YW |
186 | static const char * const mount_mode_table[_MOUNT_MODE_MAX] = { |
187 | [INACCESSIBLE] = "inaccessible", | |
188 | [BIND_MOUNT] = "bind", | |
189 | [BIND_MOUNT_RECURSIVE] = "rbind", | |
190 | [PRIVATE_TMP] = "private-tmp", | |
191 | [PRIVATE_DEV] = "private-dev", | |
192 | [BIND_DEV] = "bind-dev", | |
193 | [EMPTY_DIR] = "empty", | |
194 | [SYSFS] = "sysfs", | |
195 | [PROCFS] = "procfs", | |
196 | [READONLY] = "read-only", | |
197 | [READWRITE] = "read-write", | |
198 | [TMPFS] = "tmpfs", | |
199 | [READWRITE_IMPLICIT] = "rw-implicit", | |
200 | }; | |
201 | ||
202 | DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode); | |
203 | ||
34de407a | 204 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
205 | assert(p); |
206 | ||
5327c910 LP |
207 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
208 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 209 | |
5327c910 | 210 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
211 | } |
212 | ||
34de407a | 213 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
214 | assert(p); |
215 | ||
216 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
217 | } | |
218 | ||
d2d6c096 LP |
219 | static const char *mount_entry_source(const MountEntry *p) { |
220 | assert(p); | |
221 | ||
222 | return p->source_malloc ?: p->source_const; | |
223 | } | |
224 | ||
2abd4e38 YW |
225 | static const char *mount_entry_options(const MountEntry *p) { |
226 | assert(p); | |
227 | ||
228 | return p->options_malloc ?: p->options_const; | |
229 | } | |
230 | ||
1eb7e08e LP |
231 | static void mount_entry_done(MountEntry *p) { |
232 | assert(p); | |
233 | ||
234 | p->path_malloc = mfree(p->path_malloc); | |
235 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 236 | p->options_malloc = mfree(p->options_malloc); |
1eb7e08e LP |
237 | } |
238 | ||
d18aff04 | 239 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
240 | char **i; |
241 | ||
613b411c LP |
242 | assert(p); |
243 | ||
1e05071d | 244 | /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */ |
5327c910 | 245 | |
15ae422b | 246 | STRV_FOREACH(i, strv) { |
5327c910 LP |
247 | bool ignore = false, needs_prefix = false; |
248 | const char *e = *i; | |
15ae422b | 249 | |
5327c910 LP |
250 | /* Look for any prefixes */ |
251 | if (startswith(e, "-")) { | |
252 | e++; | |
9c94d52e | 253 | ignore = true; |
ea92ae33 | 254 | } |
5327c910 LP |
255 | if (startswith(e, "+")) { |
256 | e++; | |
257 | needs_prefix = true; | |
258 | } | |
ea92ae33 | 259 | |
baaa35ad ZJS |
260 | if (!path_is_absolute(e)) |
261 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
262 | "Path is not absolute: %s", e); | |
15ae422b | 263 | |
34de407a | 264 | *((*p)++) = (MountEntry) { |
5327c910 LP |
265 | .path_const = e, |
266 | .mode = mode, | |
267 | .ignore = ignore, | |
d18aff04 | 268 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 269 | }; |
15ae422b LP |
270 | } |
271 | ||
272 | return 0; | |
273 | } | |
274 | ||
6c47cd7d LP |
275 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
276 | char **i; | |
277 | ||
278 | assert(p); | |
279 | ||
280 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
281 | * "/private/" boundary directories for DynamicUser=1. */ | |
282 | ||
283 | STRV_FOREACH(i, strv) { | |
284 | ||
285 | *((*p)++) = (MountEntry) { | |
286 | .path_const = *i, | |
287 | .mode = EMPTY_DIR, | |
288 | .ignore = false, | |
6c47cd7d | 289 | .read_only = true, |
2abd4e38 YW |
290 | .options_const = "mode=755", |
291 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, | |
6c47cd7d LP |
292 | }; |
293 | } | |
294 | ||
295 | return 0; | |
296 | } | |
297 | ||
da6053d0 LP |
298 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
299 | size_t i; | |
d2d6c096 LP |
300 | |
301 | assert(p); | |
302 | ||
303 | for (i = 0; i < n; i++) { | |
304 | const BindMount *b = binds + i; | |
305 | ||
306 | *((*p)++) = (MountEntry) { | |
307 | .path_const = b->destination, | |
308 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
309 | .read_only = b->read_only, | |
310 | .source_const = b->source, | |
4ca763a9 | 311 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
312 | }; |
313 | } | |
314 | ||
315 | return 0; | |
316 | } | |
317 | ||
da6053d0 LP |
318 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
319 | size_t i; | |
2abd4e38 YW |
320 | int r; |
321 | ||
322 | assert(p); | |
323 | ||
324 | for (i = 0; i < n; i++) { | |
325 | const TemporaryFileSystem *t = tmpfs + i; | |
326 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 327 | unsigned long flags; |
2abd4e38 YW |
328 | bool ro = false; |
329 | ||
baaa35ad ZJS |
330 | if (!path_is_absolute(t->path)) |
331 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
332 | "Path is not absolute: %s", | |
333 | t->path); | |
2abd4e38 | 334 | |
ad8e66dc AJ |
335 | str = strjoin("mode=0755,", t->options); |
336 | if (!str) | |
337 | return -ENOMEM; | |
2abd4e38 | 338 | |
ad8e66dc AJ |
339 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
340 | if (r < 0) | |
341 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 342 | |
ad8e66dc AJ |
343 | ro = flags & MS_RDONLY; |
344 | if (ro) | |
345 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
346 | |
347 | *((*p)++) = (MountEntry) { | |
348 | .path_const = t->path, | |
349 | .mode = TMPFS, | |
350 | .read_only = ro, | |
ad8e66dc | 351 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
352 | .flags = flags, |
353 | }; | |
2abd4e38 YW |
354 | } |
355 | ||
356 | return 0; | |
357 | } | |
358 | ||
da6053d0 LP |
359 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
360 | size_t i; | |
11a30cec DH |
361 | |
362 | assert(p); | |
f471b2af | 363 | assert(mounts); |
11a30cec | 364 | |
5327c910 | 365 | /* Adds a list of static pre-defined entries */ |
f471b2af | 366 | |
5327c910 | 367 | for (i = 0; i < n; i++) |
34de407a LP |
368 | *((*p)++) = (MountEntry) { |
369 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
370 | .mode = mounts[i].mode, |
371 | .ignore = mounts[i].ignore || ignore_protect, | |
372 | }; | |
f471b2af DH |
373 | |
374 | return 0; | |
375 | } | |
376 | ||
34de407a | 377 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
378 | assert(p); |
379 | ||
5327c910 | 380 | switch (protect_home) { |
b6c432ca | 381 | |
5327c910 | 382 | case PROTECT_HOME_NO: |
b6c432ca DH |
383 | return 0; |
384 | ||
b6c432ca | 385 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
386 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
387 | ||
e4da7d8c YW |
388 | case PROTECT_HOME_TMPFS: |
389 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
390 | ||
b6c432ca | 391 | case PROTECT_HOME_YES: |
5327c910 LP |
392 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
393 | ||
b6c432ca | 394 | default: |
5327c910 | 395 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 396 | } |
b6c432ca DH |
397 | } |
398 | ||
34de407a | 399 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
400 | assert(p); |
401 | ||
5327c910 LP |
402 | switch (protect_system) { |
403 | ||
404 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
405 | return 0; |
406 | ||
f471b2af | 407 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
408 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
409 | ||
f471b2af | 410 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
411 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
412 | ||
f471b2af | 413 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
414 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
415 | ||
f471b2af | 416 | default: |
5327c910 | 417 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 418 | } |
11a30cec DH |
419 | } |
420 | ||
93bab288 | 421 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 422 | int d; |
15ae422b | 423 | |
6ee1a919 | 424 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 425 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
426 | if (d != 0) |
427 | return d; | |
15ae422b | 428 | |
6ee1a919 | 429 | /* If the paths are equal, check the mode */ |
93bab288 | 430 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
431 | } |
432 | ||
da6053d0 LP |
433 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
434 | size_t i; | |
5327c910 | 435 | |
4a756839 | 436 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 LP |
437 | |
438 | for (i = 0; i < n; i++) { | |
439 | char *s; | |
440 | ||
441 | if (m[i].has_prefix) | |
442 | continue; | |
443 | ||
34de407a | 444 | s = prefix_root(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
445 | if (!s) |
446 | return -ENOMEM; | |
447 | ||
e282f51f | 448 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
449 | m[i].has_prefix = true; |
450 | } | |
451 | ||
452 | return 0; | |
453 | } | |
454 | ||
da6053d0 | 455 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 456 | MountEntry *f, *t, *previous; |
15ae422b | 457 | |
c17ec25e | 458 | assert(m); |
15ae422b | 459 | assert(n); |
15ae422b | 460 | |
fe3c2583 LP |
461 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
462 | ||
1d54cd5d | 463 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 464 | |
fe3c2583 | 465 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
466 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
467 | if (previous && | |
468 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
469 | !f->applied && !previous->applied) { | |
5beb8688 | 470 | log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode)); |
34de407a | 471 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ |
1eb7e08e | 472 | mount_entry_done(f); |
15ae422b | 473 | continue; |
fe3c2583 | 474 | } |
15ae422b | 475 | |
e2d7c1a0 | 476 | *t = *f; |
15ae422b | 477 | previous = t; |
fe3c2583 LP |
478 | t++; |
479 | } | |
480 | ||
481 | *n = t - m; | |
482 | } | |
483 | ||
da6053d0 | 484 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 485 | MountEntry *f, *t; |
fe3c2583 LP |
486 | const char *clear = NULL; |
487 | ||
488 | assert(m); | |
489 | assert(n); | |
490 | ||
491 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
492 | * ordered already. */ | |
493 | ||
1d54cd5d | 494 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
495 | |
496 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
497 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
498 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
499 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 500 | mount_entry_done(f); |
fe3c2583 LP |
501 | continue; |
502 | } | |
15ae422b | 503 | |
34de407a | 504 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
505 | |
506 | *t = *f; | |
15ae422b LP |
507 | t++; |
508 | } | |
509 | ||
c17ec25e | 510 | *n = t - m; |
15ae422b LP |
511 | } |
512 | ||
da6053d0 | 513 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 514 | MountEntry *f, *t; |
7648a565 LP |
515 | |
516 | assert(m); | |
517 | assert(n); | |
518 | ||
519 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
520 | * list is ordered by prefixes. */ | |
521 | ||
1d54cd5d | 522 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 | 523 | |
1e05071d YW |
524 | /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */ |
525 | if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) { | |
34de407a | 526 | MountEntry *p; |
7648a565 LP |
527 | bool found = false; |
528 | ||
529 | /* Now let's find the first parent of the entry we are looking at. */ | |
530 | for (p = t-1; p >= m; p--) { | |
34de407a | 531 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
532 | found = true; |
533 | break; | |
534 | } | |
535 | } | |
536 | ||
537 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
538 | if (found && p->mode == f->mode) { | |
5beb8688 YW |
539 | log_debug("%s (%s) is made redundant by %s (%s)", |
540 | mount_entry_path(f), mount_mode_to_string(f->mode), | |
541 | mount_entry_path(p), mount_mode_to_string(p->mode)); | |
1eb7e08e | 542 | mount_entry_done(f); |
7648a565 LP |
543 | continue; |
544 | } | |
545 | } | |
546 | ||
547 | *t = *f; | |
548 | t++; | |
549 | } | |
550 | ||
551 | *n = t - m; | |
552 | } | |
553 | ||
da6053d0 | 554 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 555 | MountEntry *f, *t; |
cd2902c9 LP |
556 | |
557 | assert(m); | |
558 | assert(n); | |
559 | ||
1d54cd5d | 560 | /* Nothing to do */ |
cd2902c9 LP |
561 | if (!root_directory) |
562 | return; | |
563 | ||
564 | /* Drops all mounts that are outside of the root directory. */ | |
565 | ||
1d54cd5d | 566 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 567 | |
34de407a LP |
568 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
569 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 570 | mount_entry_done(f); |
cd2902c9 LP |
571 | continue; |
572 | } | |
573 | ||
574 | *t = *f; | |
575 | t++; | |
576 | } | |
577 | ||
578 | *n = t - m; | |
579 | } | |
580 | ||
b2a60844 LP |
581 | static int clone_device_node( |
582 | const char *d, | |
583 | const char *temporary_mount, | |
584 | bool *make_devnode) { | |
585 | ||
586 | _cleanup_free_ char *sl = NULL; | |
587 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
588 | struct stat st; |
589 | int r; | |
590 | ||
414b304b | 591 | if (stat(d, &st) < 0) { |
b2a60844 LP |
592 | if (errno == ENOENT) { |
593 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 594 | return -ENXIO; |
b2a60844 LP |
595 | } |
596 | ||
597 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
598 | } |
599 | ||
600 | if (!S_ISBLK(st.st_mode) && | |
baaa35ad ZJS |
601 | !S_ISCHR(st.st_mode)) |
602 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
603 | "Device node '%s' to clone is not a device node, ignoring.", | |
604 | d); | |
b5e99f23 | 605 | |
6f7f3a33 | 606 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 607 | |
b2a60844 | 608 | /* First, try to create device node properly */ |
16498617 CB |
609 | if (*make_devnode) { |
610 | mac_selinux_create_file_prepare(d, st.st_mode); | |
611 | r = mknod(dn, st.st_mode, st.st_rdev); | |
612 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
613 | if (r >= 0) |
614 | goto add_symlink; | |
16498617 CB |
615 | if (errno != EPERM) |
616 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
617 | ||
b2a60844 | 618 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
619 | *make_devnode = false; |
620 | } | |
621 | ||
622 | /* We're about to fallback to bind-mounting the device | |
623 | * node. So create a dummy bind-mount target. */ | |
624 | mac_selinux_create_file_prepare(d, 0); | |
625 | r = mknod(dn, S_IFREG, 0); | |
b5e99f23 | 626 | mac_selinux_create_file_clear(); |
16498617 | 627 | if (r < 0 && errno != EEXIST) |
b2a60844 | 628 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 CB |
629 | |
630 | /* Fallback to bind-mounting: | |
631 | * The assumption here is that all used device nodes carry standard | |
632 | * properties. Specifically, the devices nodes we bind-mount should | |
633 | * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx) | |
634 | * and should not carry ACLs. */ | |
635 | if (mount(d, dn, NULL, MS_BIND, NULL) < 0) | |
b2a60844 LP |
636 | return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d); |
637 | ||
638 | add_symlink: | |
639 | bn = path_startswith(d, "/dev/"); | |
640 | if (!bn) | |
641 | return 0; | |
642 | ||
643 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
644 | 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) | |
645 | return log_oom(); | |
646 | ||
647 | (void) mkdir_parents(sl, 0755); | |
648 | ||
649 | t = strjoina("../", bn); | |
650 | ||
651 | if (symlink(t, sl) < 0) | |
2e4a4fae | 652 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 653 | |
af984e13 | 654 | return 0; |
b5e99f23 ДГ |
655 | } |
656 | ||
5d997827 | 657 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
658 | static const char devnodes[] = |
659 | "/dev/null\0" | |
660 | "/dev/zero\0" | |
661 | "/dev/full\0" | |
662 | "/dev/random\0" | |
663 | "/dev/urandom\0" | |
664 | "/dev/tty\0"; | |
665 | ||
2b85f4e1 | 666 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 667 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 668 | bool can_mknod = true; |
7f112f50 LP |
669 | _cleanup_umask_ mode_t u; |
670 | int r; | |
671 | ||
672 | assert(m); | |
673 | ||
674 | u = umask(0000); | |
675 | ||
2b85f4e1 | 676 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 677 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 678 | |
63c372cb | 679 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 680 | (void) mkdir(dev, 0755); |
737ba3c8 | 681 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2e4a4fae | 682 | r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev); |
2b85f4e1 LP |
683 | goto fail; |
684 | } | |
685 | ||
63c372cb | 686 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 687 | (void) mkdir(devpts, 0755); |
2b85f4e1 | 688 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
2e4a4fae | 689 | r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts); |
2b85f4e1 LP |
690 | goto fail; |
691 | } | |
692 | ||
2e4a4fae YW |
693 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx. |
694 | * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible. | |
695 | * Thus, in that case make a clone. | |
696 | * In nspawn and other containers it will be a symlink, in that case make it a symlink. */ | |
36ce7110 | 697 | r = is_symlink("/dev/ptmx"); |
2e4a4fae YW |
698 | if (r < 0) { |
699 | log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m"); | |
3164e3cb | 700 | goto fail; |
2e4a4fae | 701 | } else if (r > 0) { |
414b304b ДГ |
702 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
703 | if (symlink("pts/ptmx", devptmx) < 0) { | |
2e4a4fae | 704 | r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx); |
414b304b ДГ |
705 | goto fail; |
706 | } | |
707 | } else { | |
16498617 | 708 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
709 | if (r < 0) |
710 | goto fail; | |
414b304b | 711 | } |
e06b6479 | 712 | |
63c372cb | 713 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 714 | (void) mkdir(devshm, 0755); |
2b85f4e1 LP |
715 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
716 | if (r < 0) { | |
2e4a4fae | 717 | r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm); |
2b85f4e1 LP |
718 | goto fail; |
719 | } | |
720 | ||
63c372cb | 721 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 722 | (void) mkdir(devmqueue, 0755); |
2e4a4fae YW |
723 | if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0) |
724 | log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue); | |
2b85f4e1 | 725 | |
63c372cb | 726 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 727 | (void) mkdir(devhugepages, 0755); |
2e4a4fae YW |
728 | if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0) |
729 | log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages); | |
2b85f4e1 | 730 | |
63c372cb | 731 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
732 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
733 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 734 | |
7f112f50 | 735 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 736 | r = clone_device_node(d, temporary_mount, &can_mknod); |
af984e13 ZJS |
737 | /* ENXIO means the the *source* is not a device file, skip creation in that case */ |
738 | if (r < 0 && r != -ENXIO) | |
2b85f4e1 | 739 | goto fail; |
7f112f50 LP |
740 | } |
741 | ||
2e4a4fae YW |
742 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
743 | if (r < 0) | |
744 | log_debug_errno(r, "Failed to setup basic device tree at '%s', ignoring: %m", temporary_mount); | |
7f112f50 | 745 | |
ee818b89 AC |
746 | /* Create the /dev directory if missing. It is more likely to be |
747 | * missing when the service is started with RootDirectory. This is | |
748 | * consistent with mount units creating the mount points when missing. | |
749 | */ | |
34de407a | 750 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 751 | |
9e5f8252 | 752 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
753 | r = umount_recursive(mount_entry_path(m), 0); |
754 | if (r < 0) | |
755 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
756 | ||
34de407a | 757 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { |
2e4a4fae | 758 | r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m)); |
2b85f4e1 LP |
759 | goto fail; |
760 | } | |
7f112f50 | 761 | |
2b85f4e1 LP |
762 | rmdir(dev); |
763 | rmdir(temporary_mount); | |
7f112f50 | 764 | |
2b85f4e1 | 765 | return 0; |
7f112f50 | 766 | |
2b85f4e1 LP |
767 | fail: |
768 | if (devpts) | |
769 | umount(devpts); | |
7f112f50 | 770 | |
2b85f4e1 LP |
771 | if (devshm) |
772 | umount(devshm); | |
7f112f50 | 773 | |
2b85f4e1 LP |
774 | if (devhugepages) |
775 | umount(devhugepages); | |
7f112f50 | 776 | |
2b85f4e1 LP |
777 | if (devmqueue) |
778 | umount(devmqueue); | |
7f112f50 | 779 | |
d267c5aa ZJS |
780 | umount(dev); |
781 | rmdir(dev); | |
2b85f4e1 | 782 | rmdir(temporary_mount); |
7f112f50 | 783 | |
2b85f4e1 | 784 | return r; |
7f112f50 LP |
785 | } |
786 | ||
2a2969fd | 787 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
788 | int r; |
789 | ||
790 | assert(m); | |
791 | ||
792 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
793 | * /dev. This is only used when RootDirectory= is set. */ | |
794 | ||
645767d6 LP |
795 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
796 | ||
5d997827 LP |
797 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
798 | if (r < 0) | |
799 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
800 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
801 | return 0; | |
802 | ||
803 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
804 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
805 | ||
806 | return 1; | |
807 | } | |
808 | ||
2a2969fd | 809 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
810 | int r; |
811 | ||
812 | assert(m); | |
813 | ||
645767d6 LP |
814 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
815 | ||
5d997827 LP |
816 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
817 | if (r < 0) | |
818 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
819 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
820 | return 0; | |
821 | ||
822 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
823 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
824 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
825 | ||
826 | return 1; | |
827 | } | |
828 | ||
2a2969fd | 829 | static int mount_procfs(const MountEntry *m) { |
5d997827 LP |
830 | int r; |
831 | ||
832 | assert(m); | |
833 | ||
645767d6 LP |
834 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
835 | ||
5d997827 LP |
836 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
837 | if (r < 0) | |
838 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
839 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
840 | return 0; | |
841 | ||
842 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
843 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
844 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
845 | ||
846 | return 1; | |
847 | } | |
848 | ||
2abd4e38 | 849 | static int mount_tmpfs(const MountEntry *m) { |
6c47cd7d LP |
850 | assert(m); |
851 | ||
2abd4e38 | 852 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d LP |
853 | |
854 | (void) mkdir_p_label(mount_entry_path(m), 0755); | |
855 | (void) umount_recursive(mount_entry_path(m), 0); | |
856 | ||
2abd4e38 | 857 | if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0) |
6c47cd7d LP |
858 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); |
859 | ||
860 | return 1; | |
861 | } | |
862 | ||
088696fe | 863 | static int follow_symlink( |
d2d6c096 | 864 | const char *root_directory, |
088696fe | 865 | MountEntry *m) { |
d2d6c096 | 866 | |
088696fe | 867 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
868 | int r; |
869 | ||
088696fe LP |
870 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
871 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
872 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
873 | * end and already have a fully normalized name. */ | |
8fceda93 | 874 | |
088696fe LP |
875 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target); |
876 | if (r < 0) | |
877 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
878 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
879 | return 1; | |
8fceda93 | 880 | |
baaa35ad ZJS |
881 | if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */ |
882 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
883 | "Symlink loop on '%s'.", | |
884 | mount_entry_path(m)); | |
8fceda93 | 885 | |
088696fe | 886 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 887 | |
088696fe LP |
888 | free_and_replace(m->path_malloc, target); |
889 | m->has_prefix = true; | |
8fceda93 | 890 | |
088696fe LP |
891 | m->n_followed ++; |
892 | ||
893 | return 0; | |
8fceda93 LP |
894 | } |
895 | ||
ac0930c8 | 896 | static int apply_mount( |
8fceda93 | 897 | const char *root_directory, |
89bd586c | 898 | MountEntry *m) { |
ac0930c8 | 899 | |
a227a4be | 900 | bool rbind = true, make = false; |
15ae422b | 901 | const char *what; |
15ae422b | 902 | int r; |
15ae422b | 903 | |
c17ec25e | 904 | assert(m); |
15ae422b | 905 | |
34de407a | 906 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 907 | |
c17ec25e | 908 | switch (m->mode) { |
15ae422b | 909 | |
160cfdbe LP |
910 | case INACCESSIBLE: { |
911 | struct stat target; | |
6d313367 LP |
912 | |
913 | /* First, get rid of everything that is below if there | |
914 | * is anything... Then, overmount it with an | |
c4b41707 | 915 | * inaccessible path. */ |
34de407a | 916 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 917 | |
088696fe LP |
918 | if (lstat(mount_entry_path(m), &target) < 0) { |
919 | if (errno == ENOENT && m->ignore) | |
920 | return 0; | |
921 | ||
34de407a | 922 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); |
088696fe | 923 | } |
15ae422b | 924 | |
c4b41707 | 925 | what = mode_to_inaccessible_node(target.st_mode); |
baaa35ad ZJS |
926 | if (!what) |
927 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
928 | "File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
c4b41707 | 929 | break; |
160cfdbe | 930 | } |
fe3c2583 | 931 | |
15ae422b | 932 | case READONLY: |
15ae422b | 933 | case READWRITE: |
1e05071d | 934 | case READWRITE_IMPLICIT: |
8fceda93 | 935 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
936 | if (r == -ENOENT && m->ignore) |
937 | return 0; | |
d944dc95 | 938 | if (r < 0) |
34de407a | 939 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
940 | 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. */ |
941 | return 0; | |
6b7c9f8b | 942 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 943 | what = mount_entry_path(m); |
6b7c9f8b | 944 | break; |
15ae422b | 945 | |
d2d6c096 LP |
946 | case BIND_MOUNT: |
947 | rbind = false; | |
d2d6c096 | 948 | |
4831981d | 949 | _fallthrough_; |
088696fe LP |
950 | case BIND_MOUNT_RECURSIVE: { |
951 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 952 | |
088696fe LP |
953 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind |
954 | * mount source paths are always relative to the host root, hence we pass NULL as root directory to | |
955 | * chase_symlinks() here. */ | |
956 | ||
957 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased); | |
958 | if (r == -ENOENT && m->ignore) { | |
959 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
960 | return 0; | |
961 | } | |
962 | if (r < 0) | |
963 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
964 | ||
965 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
966 | ||
967 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
968 | |
969 | what = mount_entry_source(m); | |
a227a4be | 970 | make = true; |
d2d6c096 | 971 | break; |
088696fe | 972 | } |
d2d6c096 | 973 | |
6c47cd7d | 974 | case EMPTY_DIR: |
2abd4e38 YW |
975 | case TMPFS: |
976 | return mount_tmpfs(m); | |
6c47cd7d | 977 | |
ac0930c8 | 978 | case PRIVATE_TMP: |
89bd586c | 979 | what = mount_entry_source(m); |
a227a4be | 980 | make = true; |
15ae422b | 981 | break; |
e364ad06 | 982 | |
d6797c92 | 983 | case PRIVATE_DEV: |
5d997827 LP |
984 | return mount_private_dev(m); |
985 | ||
986 | case BIND_DEV: | |
987 | return mount_bind_dev(m); | |
988 | ||
989 | case SYSFS: | |
990 | return mount_sysfs(m); | |
991 | ||
992 | case PROCFS: | |
993 | return mount_procfs(m); | |
d6797c92 | 994 | |
e364ad06 LP |
995 | default: |
996 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
997 | } |
998 | ||
ac0930c8 | 999 | assert(what); |
15ae422b | 1000 | |
a227a4be LP |
1001 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) { |
1002 | bool try_again = false; | |
1003 | r = -errno; | |
1004 | ||
1005 | if (r == -ENOENT && make) { | |
1006 | struct stat st; | |
1007 | ||
1008 | /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */ | |
1009 | ||
e8717862 LP |
1010 | if (stat(what, &st) < 0) |
1011 | log_debug_errno(errno, "Mount point source '%s' is not accessible: %m", what); | |
1012 | else { | |
1013 | int q; | |
a227a4be LP |
1014 | |
1015 | (void) mkdir_parents(mount_entry_path(m), 0755); | |
1016 | ||
1017 | if (S_ISDIR(st.st_mode)) | |
e8717862 | 1018 | q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0; |
a227a4be | 1019 | else |
e8717862 LP |
1020 | q = touch(mount_entry_path(m)); |
1021 | ||
1022 | if (q < 0) | |
1023 | log_debug_errno(q, "Failed to create destination mount point node '%s': %m", mount_entry_path(m)); | |
1024 | else | |
1025 | try_again = true; | |
a227a4be LP |
1026 | } |
1027 | } | |
1028 | ||
1029 | if (try_again) { | |
1030 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) | |
1031 | r = -errno; | |
1032 | else | |
1033 | r = 0; | |
1034 | } | |
1035 | ||
1036 | if (r < 0) | |
1037 | return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); | |
1038 | } | |
6b7c9f8b | 1039 | |
34de407a | 1040 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1041 | return 0; |
ac0930c8 | 1042 | } |
15ae422b | 1043 | |
69338c3d AJ |
1044 | /* Change the per-mount readonly flag on an existing mount */ |
1045 | static int remount_bind_readonly(const char *path, unsigned long orig_flags) { | |
1046 | int r; | |
1047 | ||
1048 | r = mount(NULL, path, NULL, MS_REMOUNT | MS_BIND | MS_RDONLY | orig_flags, NULL); | |
1049 | ||
1050 | return r < 0 ? -errno : 0; | |
1051 | } | |
1052 | ||
2a2969fd | 1053 | static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) { |
763a260a | 1054 | bool submounts = false; |
6b7c9f8b | 1055 | int r = 0; |
15ae422b | 1056 | |
c17ec25e | 1057 | assert(m); |
ac9de0b3 | 1058 | assert(proc_self_mountinfo); |
ac0930c8 | 1059 | |
2abd4e38 YW |
1060 | if (mount_entry_read_only(m)) { |
1061 | if (IN_SET(m->mode, EMPTY_DIR, TMPFS)) { | |
69338c3d | 1062 | r = remount_bind_readonly(mount_entry_path(m), m->flags); |
763a260a YW |
1063 | } else { |
1064 | submounts = true; | |
2abd4e38 | 1065 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), true, blacklist, proc_self_mountinfo); |
763a260a | 1066 | } |
2abd4e38 | 1067 | } else if (m->mode == PRIVATE_DEV) { |
69338c3d AJ |
1068 | /* Set /dev readonly, but not submounts like /dev/shm. Also, we only set the per-mount read-only flag. |
1069 | * We can't set it on the superblock, if we are inside a user namespace and running Linux <= 4.17. */ | |
1070 | r = remount_bind_readonly(mount_entry_path(m), DEV_MOUNT_OPTIONS); | |
737ba3c8 | 1071 | } else |
6b7c9f8b LP |
1072 | return 0; |
1073 | ||
1074 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked read-only | |
1075 | * already stays this way. This improves compatibility with container managers, where we won't attempt to undo | |
1076 | * read-only mounts already applied. */ | |
ac0930c8 | 1077 | |
8fceda93 LP |
1078 | if (r == -ENOENT && m->ignore) |
1079 | r = 0; | |
5327c910 | 1080 | |
763a260a YW |
1081 | if (r < 0) |
1082 | return log_debug_errno(r, "Failed to re-mount '%s'%s read-only: %m", mount_entry_path(m), | |
1083 | submounts ? " and its submounts" : ""); | |
1084 | ||
1085 | return 0; | |
d944dc95 LP |
1086 | } |
1087 | ||
9b68367b | 1088 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1089 | assert(ns_info); |
1090 | ||
9c988f93 DH |
1091 | /* |
1092 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1093 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1094 | * first place... |
9c988f93 | 1095 | */ |
5d997827 | 1096 | |
9b68367b YW |
1097 | return ns_info->mount_apivfs || |
1098 | ns_info->protect_control_groups || | |
1099 | ns_info->protect_kernel_tunables; | |
5d997827 LP |
1100 | } |
1101 | ||
da6053d0 | 1102 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1103 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1104 | char** read_write_paths, |
1105 | char** read_only_paths, | |
1106 | char** inaccessible_paths, | |
6c47cd7d | 1107 | char** empty_directories, |
da6053d0 LP |
1108 | size_t n_bind_mounts, |
1109 | size_t n_temporary_filesystems, | |
2652c6c1 DH |
1110 | const char* tmp_dir, |
1111 | const char* var_tmp_dir, | |
2652c6c1 DH |
1112 | ProtectHome protect_home, |
1113 | ProtectSystem protect_system) { | |
1114 | ||
da6053d0 LP |
1115 | size_t protect_home_cnt; |
1116 | size_t protect_system_cnt = | |
f471b2af DH |
1117 | (protect_system == PROTECT_SYSTEM_STRICT ? |
1118 | ELEMENTSOF(protect_system_strict_table) : | |
1119 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
1120 | ELEMENTSOF(protect_system_full_table) : | |
1121 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
1122 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
1123 | ||
b6c432ca DH |
1124 | protect_home_cnt = |
1125 | (protect_home == PROTECT_HOME_YES ? | |
1126 | ELEMENTSOF(protect_home_yes_table) : | |
1127 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
e4da7d8c YW |
1128 | ELEMENTSOF(protect_home_read_only_table) : |
1129 | ((protect_home == PROTECT_HOME_TMPFS) ? | |
1130 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); | |
b6c432ca | 1131 | |
2652c6c1 DH |
1132 | return !!tmp_dir + !!var_tmp_dir + |
1133 | strv_length(read_write_paths) + | |
1134 | strv_length(read_only_paths) + | |
1135 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1136 | strv_length(empty_directories) + |
d2d6c096 | 1137 | n_bind_mounts + |
2abd4e38 | 1138 | n_temporary_filesystems + |
c575770b DH |
1139 | ns_info->private_dev + |
1140 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
1141 | (ns_info->protect_control_groups ? 1 : 0) + | |
1142 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + | |
5d997827 | 1143 | protect_home_cnt + protect_system_cnt + |
aecd5ac6 | 1144 | (ns_info->protect_hostname ? 2 : 0) + |
9b68367b | 1145 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0); |
2652c6c1 DH |
1146 | } |
1147 | ||
da6053d0 | 1148 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1149 | assert(root_directory); |
f8b64b57 LP |
1150 | assert(n_mounts); |
1151 | assert(mounts || *n_mounts == 0); | |
1152 | ||
93bab288 | 1153 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1154 | |
1155 | drop_duplicates(mounts, n_mounts); | |
1156 | drop_outside_root(root_directory, mounts, n_mounts); | |
1157 | drop_inaccessible(mounts, n_mounts); | |
1158 | drop_nop(mounts, n_mounts); | |
1159 | } | |
1160 | ||
613b411c | 1161 | int setup_namespace( |
ee818b89 | 1162 | const char* root_directory, |
915e6d16 | 1163 | const char* root_image, |
bb0ff3fb | 1164 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1165 | char** read_write_paths, |
1166 | char** read_only_paths, | |
1167 | char** inaccessible_paths, | |
6c47cd7d | 1168 | char** empty_directories, |
d2d6c096 | 1169 | const BindMount *bind_mounts, |
da6053d0 | 1170 | size_t n_bind_mounts, |
2abd4e38 | 1171 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1172 | size_t n_temporary_filesystems, |
a004cb4c LP |
1173 | const char* tmp_dir, |
1174 | const char* var_tmp_dir, | |
1b8689f9 LP |
1175 | ProtectHome protect_home, |
1176 | ProtectSystem protect_system, | |
915e6d16 LP |
1177 | unsigned long mount_flags, |
1178 | DissectImageFlags dissect_image_flags) { | |
15ae422b | 1179 | |
915e6d16 | 1180 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1181 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1182 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 1183 | _cleanup_free_ void *root_hash = NULL; |
34de407a | 1184 | MountEntry *m, *mounts = NULL; |
9b68367b | 1185 | size_t n_mounts, root_hash_size = 0; |
d18aff04 | 1186 | bool require_prefix = false; |
9b68367b | 1187 | const char *root; |
c17ec25e | 1188 | int r = 0; |
15ae422b | 1189 | |
915e6d16 LP |
1190 | assert(ns_info); |
1191 | ||
613b411c | 1192 | if (mount_flags == 0) |
c17ec25e | 1193 | mount_flags = MS_SHARED; |
ac0930c8 | 1194 | |
915e6d16 LP |
1195 | if (root_image) { |
1196 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1197 | ||
c9ef8573 LP |
1198 | if (protect_system == PROTECT_SYSTEM_STRICT && |
1199 | protect_home != PROTECT_HOME_NO && | |
1200 | strv_isempty(read_write_paths)) | |
915e6d16 LP |
1201 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1202 | ||
1203 | r = loop_device_make_by_path(root_image, | |
1204 | dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR, | |
1205 | &loop_device); | |
1206 | if (r < 0) | |
763a260a | 1207 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
915e6d16 | 1208 | |
78ebe980 LP |
1209 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
1210 | if (r < 0) | |
763a260a | 1211 | return log_debug_errno(r, "Failed to load root hash: %m"); |
78ebe980 LP |
1212 | |
1213 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
1214 | if (r < 0) | |
763a260a | 1215 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 LP |
1216 | |
1217 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 | 1218 | if (r < 0) |
763a260a | 1219 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1220 | } |
1221 | ||
e908468b LP |
1222 | if (root_directory) |
1223 | root = root_directory; | |
0722b359 JS |
1224 | else { |
1225 | /* Always create the mount namespace in a temporary directory, instead of operating | |
1226 | * directly in the root. The temporary directory prevents any mounts from being | |
1227 | * potentially obscured my other mounts we already applied. | |
1228 | * We use the same mount point for all images, which is safe, since they all live | |
1229 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1230 | |
1231 | root = "/run/systemd/unit-root"; | |
1232 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1233 | require_prefix = true; |
0722b359 | 1234 | } |
e908468b | 1235 | |
cfbeb4ef LP |
1236 | n_mounts = namespace_calculate_mounts( |
1237 | ns_info, | |
1238 | read_write_paths, | |
1239 | read_only_paths, | |
1240 | inaccessible_paths, | |
6c47cd7d | 1241 | empty_directories, |
f5c52a77 | 1242 | n_bind_mounts, |
2abd4e38 | 1243 | n_temporary_filesystems, |
cfbeb4ef LP |
1244 | tmp_dir, var_tmp_dir, |
1245 | protect_home, protect_system); | |
613b411c | 1246 | |
f0a4feb0 | 1247 | if (n_mounts > 0) { |
34de407a | 1248 | m = mounts = (MountEntry *) alloca0(n_mounts * sizeof(MountEntry)); |
d18aff04 | 1249 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1250 | if (r < 0) |
f0a4feb0 | 1251 | goto finish; |
613b411c | 1252 | |
d18aff04 | 1253 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1254 | if (r < 0) |
f0a4feb0 | 1255 | goto finish; |
613b411c | 1256 | |
d18aff04 | 1257 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1258 | if (r < 0) |
f0a4feb0 | 1259 | goto finish; |
7ff7394d | 1260 | |
6c47cd7d LP |
1261 | r = append_empty_dir_mounts(&m, empty_directories); |
1262 | if (r < 0) | |
1263 | goto finish; | |
1264 | ||
d2d6c096 LP |
1265 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1266 | if (r < 0) | |
1267 | goto finish; | |
1268 | ||
2abd4e38 YW |
1269 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1270 | if (r < 0) | |
1271 | goto finish; | |
1272 | ||
613b411c | 1273 | if (tmp_dir) { |
34de407a | 1274 | *(m++) = (MountEntry) { |
5327c910 LP |
1275 | .path_const = "/tmp", |
1276 | .mode = PRIVATE_TMP, | |
89bd586c | 1277 | .source_const = tmp_dir, |
5327c910 | 1278 | }; |
613b411c | 1279 | } |
7ff7394d | 1280 | |
613b411c | 1281 | if (var_tmp_dir) { |
34de407a | 1282 | *(m++) = (MountEntry) { |
5327c910 | 1283 | .path_const = "/var/tmp", |
89bd586c YW |
1284 | .mode = PRIVATE_TMP, |
1285 | .source_const = var_tmp_dir, | |
5327c910 | 1286 | }; |
7ff7394d | 1287 | } |
ac0930c8 | 1288 | |
c575770b | 1289 | if (ns_info->private_dev) { |
34de407a | 1290 | *(m++) = (MountEntry) { |
5327c910 LP |
1291 | .path_const = "/dev", |
1292 | .mode = PRIVATE_DEV, | |
1293 | }; | |
7f112f50 LP |
1294 | } |
1295 | ||
c575770b | 1296 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 1297 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 1298 | if (r < 0) |
f0a4feb0 | 1299 | goto finish; |
c575770b DH |
1300 | } |
1301 | ||
1302 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 1303 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 1304 | if (r < 0) |
f0a4feb0 | 1305 | goto finish; |
c575770b | 1306 | } |
59eeb84b | 1307 | |
c575770b | 1308 | if (ns_info->protect_control_groups) { |
34de407a | 1309 | *(m++) = (MountEntry) { |
5327c910 LP |
1310 | .path_const = "/sys/fs/cgroup", |
1311 | .mode = READONLY, | |
1312 | }; | |
59eeb84b LP |
1313 | } |
1314 | ||
5327c910 | 1315 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1316 | if (r < 0) |
f0a4feb0 | 1317 | goto finish; |
417116f2 | 1318 | |
5327c910 | 1319 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1320 | if (r < 0) |
f0a4feb0 | 1321 | goto finish; |
417116f2 | 1322 | |
9b68367b | 1323 | if (namespace_info_mount_apivfs(ns_info)) { |
5d997827 LP |
1324 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); |
1325 | if (r < 0) | |
1326 | goto finish; | |
1327 | } | |
1328 | ||
aecd5ac6 TM |
1329 | if (ns_info->protect_hostname) { |
1330 | *(m++) = (MountEntry) { | |
1331 | .path_const = "/proc/sys/kernel/hostname", | |
1332 | .mode = READONLY, | |
1333 | }; | |
1334 | *(m++) = (MountEntry) { | |
1335 | .path_const = "/proc/sys/kernel/domainname", | |
1336 | .mode = READONLY, | |
1337 | }; | |
1338 | } | |
1339 | ||
f0a4feb0 | 1340 | assert(mounts + n_mounts == m); |
ac0930c8 | 1341 | |
5327c910 | 1342 | /* Prepend the root directory where that's necessary */ |
e908468b | 1343 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1344 | if (r < 0) |
1345 | goto finish; | |
1346 | ||
839f1877 | 1347 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1348 | } |
1349 | ||
1beab8b0 LP |
1350 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1351 | ||
d944dc95 | 1352 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1353 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 LP |
1354 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
1355 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place | |
1356 | * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable | |
1357 | * error back, which the caller can use to detect this case (and only this) and optionally | |
1358 | * continue without namespacing applied. */ | |
1359 | r = -ENOANO; | |
1360 | ||
d944dc95 LP |
1361 | goto finish; |
1362 | } | |
1e4e94c8 | 1363 | |
9b68367b YW |
1364 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1365 | * shows up in the parent */ | |
1366 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1367 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1368 | goto finish; |
ee818b89 AC |
1369 | } |
1370 | ||
915e6d16 | 1371 | if (root_image) { |
e908468b | 1372 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1373 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1374 | if (r < 0) { |
1375 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1376 | goto finish; |
763a260a | 1377 | } |
915e6d16 | 1378 | |
07ce7407 TM |
1379 | if (decrypted_image) { |
1380 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1381 | if (r < 0) { |
1382 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1383 | goto finish; |
763a260a | 1384 | } |
07ce7407 | 1385 | } |
78ebe980 | 1386 | |
915e6d16 LP |
1387 | loop_device_relinquish(loop_device); |
1388 | ||
1389 | } else if (root_directory) { | |
1390 | ||
e908468b LP |
1391 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1392 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1393 | if (r < 0) { |
1394 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1395 | goto finish; |
763a260a | 1396 | } |
8f1ad200 | 1397 | if (r == 0) { |
e908468b | 1398 | if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) { |
763a260a | 1399 | r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root); |
8f1ad200 LP |
1400 | goto finish; |
1401 | } | |
d944dc95 | 1402 | } |
e908468b | 1403 | |
9b68367b | 1404 | } else { |
e908468b LP |
1405 | |
1406 | /* Let's mount the main root directory to the root directory to use */ | |
1407 | if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
763a260a | 1408 | r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root); |
e908468b LP |
1409 | goto finish; |
1410 | } | |
ee818b89 | 1411 | } |
c2c13f2d | 1412 | |
4e0c20de LP |
1413 | /* Try to set up the new root directory before mounting anything else there. */ |
1414 | if (root_image || root_directory) | |
1415 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1416 | ||
f0a4feb0 | 1417 | if (n_mounts > 0) { |
ac9de0b3 | 1418 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b7c9f8b | 1419 | char **blacklist; |
da6053d0 | 1420 | size_t j; |
6b7c9f8b | 1421 | |
ac9de0b3 TR |
1422 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc. |
1423 | * For example, this is the case with the option: 'InaccessiblePaths=/proc' */ | |
1424 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); | |
1425 | if (!proc_self_mountinfo) { | |
763a260a | 1426 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
ac9de0b3 TR |
1427 | goto finish; |
1428 | } | |
1429 | ||
088696fe LP |
1430 | /* First round, establish all mounts we need */ |
1431 | for (;;) { | |
1432 | bool again = false; | |
1433 | ||
1434 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1435 | ||
1436 | if (m->applied) | |
1437 | continue; | |
1438 | ||
1439 | r = follow_symlink(root, m); | |
1440 | if (r < 0) | |
1441 | goto finish; | |
1442 | if (r == 0) { | |
1443 | /* We hit a symlinked mount point. The entry got rewritten and might point to a | |
1444 | * very different place now. Let's normalize the changed list, and start from | |
1445 | * the beginning. After all to mount the entry at the new location we might | |
1446 | * need some other mounts first */ | |
1447 | again = true; | |
1448 | break; | |
1449 | } | |
1450 | ||
1451 | r = apply_mount(root, m); | |
1452 | if (r < 0) | |
1453 | goto finish; | |
1454 | ||
1455 | m->applied = true; | |
1456 | } | |
1457 | ||
1458 | if (!again) | |
1459 | break; | |
1460 | ||
839f1877 | 1461 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1462 | } |
15ae422b | 1463 | |
6b7c9f8b | 1464 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
f0a4feb0 DH |
1465 | blacklist = newa(char*, n_mounts+1); |
1466 | for (j = 0; j < n_mounts; j++) | |
34de407a | 1467 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1468 | blacklist[j] = NULL; |
1469 | ||
1470 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1471 | for (m = mounts; m < mounts + n_mounts; ++m) { |
ac9de0b3 | 1472 | r = make_read_only(m, blacklist, proc_self_mountinfo); |
c2c13f2d | 1473 | if (r < 0) |
d944dc95 | 1474 | goto finish; |
c2c13f2d | 1475 | } |
15ae422b LP |
1476 | } |
1477 | ||
9b68367b YW |
1478 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1479 | r = mount_move_root(root); | |
763a260a YW |
1480 | if (r < 0) { |
1481 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1482 | goto finish; |
763a260a | 1483 | } |
ee818b89 | 1484 | |
55fe7432 | 1485 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1486 | * reestablish propagation from our side to the host, since |
1487 | * what's disconnected is disconnected. */ | |
d944dc95 | 1488 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1489 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1490 | goto finish; |
1491 | } | |
15ae422b | 1492 | |
d944dc95 | 1493 | r = 0; |
15ae422b | 1494 | |
d944dc95 | 1495 | finish: |
f0a4feb0 | 1496 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1497 | mount_entry_done(m); |
613b411c LP |
1498 | |
1499 | return r; | |
1500 | } | |
1501 | ||
da6053d0 LP |
1502 | void bind_mount_free_many(BindMount *b, size_t n) { |
1503 | size_t i; | |
d2d6c096 LP |
1504 | |
1505 | assert(b || n == 0); | |
1506 | ||
1507 | for (i = 0; i < n; i++) { | |
1508 | free(b[i].source); | |
1509 | free(b[i].destination); | |
1510 | } | |
1511 | ||
1512 | free(b); | |
1513 | } | |
1514 | ||
da6053d0 | 1515 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
1516 | _cleanup_free_ char *s = NULL, *d = NULL; |
1517 | BindMount *c; | |
1518 | ||
1519 | assert(b); | |
1520 | assert(n); | |
1521 | assert(item); | |
1522 | ||
1523 | s = strdup(item->source); | |
1524 | if (!s) | |
1525 | return -ENOMEM; | |
1526 | ||
1527 | d = strdup(item->destination); | |
1528 | if (!d) | |
1529 | return -ENOMEM; | |
1530 | ||
aa484f35 | 1531 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
1532 | if (!c) |
1533 | return -ENOMEM; | |
1534 | ||
1535 | *b = c; | |
1536 | ||
1537 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
1538 | .source = TAKE_PTR(s), |
1539 | .destination = TAKE_PTR(d), | |
d2d6c096 LP |
1540 | .read_only = item->read_only, |
1541 | .recursive = item->recursive, | |
1542 | .ignore_enoent = item->ignore_enoent, | |
1543 | }; | |
1544 | ||
d2d6c096 LP |
1545 | return 0; |
1546 | } | |
1547 | ||
da6053d0 LP |
1548 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
1549 | size_t i; | |
2abd4e38 YW |
1550 | |
1551 | assert(t || n == 0); | |
1552 | ||
1553 | for (i = 0; i < n; i++) { | |
1554 | free(t[i].path); | |
1555 | free(t[i].options); | |
1556 | } | |
1557 | ||
1558 | free(t); | |
1559 | } | |
1560 | ||
1561 | int temporary_filesystem_add( | |
1562 | TemporaryFileSystem **t, | |
da6053d0 | 1563 | size_t *n, |
2abd4e38 YW |
1564 | const char *path, |
1565 | const char *options) { | |
1566 | ||
1567 | _cleanup_free_ char *p = NULL, *o = NULL; | |
1568 | TemporaryFileSystem *c; | |
1569 | ||
1570 | assert(t); | |
1571 | assert(n); | |
1572 | assert(path); | |
1573 | ||
1574 | p = strdup(path); | |
1575 | if (!p) | |
1576 | return -ENOMEM; | |
1577 | ||
1578 | if (!isempty(options)) { | |
1579 | o = strdup(options); | |
1580 | if (!o) | |
1581 | return -ENOMEM; | |
1582 | } | |
1583 | ||
aa484f35 | 1584 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
1585 | if (!c) |
1586 | return -ENOMEM; | |
1587 | ||
1588 | *t = c; | |
1589 | ||
1590 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
1591 | .path = TAKE_PTR(p), |
1592 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
1593 | }; |
1594 | ||
2abd4e38 YW |
1595 | return 0; |
1596 | } | |
1597 | ||
613b411c LP |
1598 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1599 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1600 | char bid[SD_ID128_STRING_MAX]; |
1601 | sd_id128_t boot_id; | |
1602 | int r; | |
613b411c LP |
1603 | |
1604 | assert(id); | |
1605 | assert(prefix); | |
1606 | assert(path); | |
1607 | ||
6b46ea73 LP |
1608 | /* We include the boot id in the directory so that after a |
1609 | * reboot we can easily identify obsolete directories. */ | |
1610 | ||
1611 | r = sd_id128_get_boot(&boot_id); | |
1612 | if (r < 0) | |
1613 | return r; | |
1614 | ||
605405c6 | 1615 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1616 | if (!x) |
1617 | return -ENOMEM; | |
1618 | ||
1619 | RUN_WITH_UMASK(0077) | |
1620 | if (!mkdtemp(x)) | |
1621 | return -errno; | |
1622 | ||
1623 | RUN_WITH_UMASK(0000) { | |
1624 | char *y; | |
1625 | ||
63c372cb | 1626 | y = strjoina(x, "/tmp"); |
613b411c LP |
1627 | |
1628 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1629 | return -errno; | |
c17ec25e | 1630 | } |
15ae422b | 1631 | |
1cc6c93a | 1632 | *path = TAKE_PTR(x); |
613b411c LP |
1633 | |
1634 | return 0; | |
1635 | } | |
1636 | ||
1637 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1638 | char *a, *b; | |
1639 | int r; | |
1640 | ||
1641 | assert(id); | |
1642 | assert(tmp_dir); | |
1643 | assert(var_tmp_dir); | |
1644 | ||
1645 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1646 | if (r < 0) | |
1647 | return r; | |
1648 | ||
1649 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1650 | if (r < 0) { | |
1651 | char *t; | |
1652 | ||
63c372cb | 1653 | t = strjoina(a, "/tmp"); |
613b411c LP |
1654 | rmdir(t); |
1655 | rmdir(a); | |
1656 | ||
1657 | free(a); | |
1658 | return r; | |
1659 | } | |
1660 | ||
1661 | *tmp_dir = a; | |
1662 | *var_tmp_dir = b; | |
1663 | ||
1664 | return 0; | |
1665 | } | |
1666 | ||
3042bbeb | 1667 | int setup_netns(int netns_storage_socket[static 2]) { |
613b411c | 1668 | _cleanup_close_ int netns = -1; |
3ee897d6 | 1669 | int r, q; |
613b411c LP |
1670 | |
1671 | assert(netns_storage_socket); | |
1672 | assert(netns_storage_socket[0] >= 0); | |
1673 | assert(netns_storage_socket[1] >= 0); | |
1674 | ||
1675 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1676 | * namespace reference fd. Whatever process runs this first |
1677 | * shall create a new namespace, all others should just join | |
1678 | * it. To serialize that we use a file lock on the socket | |
1679 | * pair. | |
613b411c LP |
1680 | * |
1681 | * It's a bit crazy, but hey, works great! */ | |
1682 | ||
1683 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1684 | return -errno; | |
1685 | ||
3ee897d6 LP |
1686 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1687 | if (netns == -EAGAIN) { | |
44ffcbae | 1688 | /* Nothing stored yet, so let's create a new namespace. */ |
613b411c LP |
1689 | |
1690 | if (unshare(CLONE_NEWNET) < 0) { | |
1691 | r = -errno; | |
1692 | goto fail; | |
1693 | } | |
1694 | ||
44ffcbae | 1695 | (void) loopback_setup(); |
613b411c LP |
1696 | |
1697 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1698 | if (netns < 0) { | |
1699 | r = -errno; | |
1700 | goto fail; | |
1701 | } | |
1702 | ||
1703 | r = 1; | |
613b411c | 1704 | |
3ee897d6 LP |
1705 | } else if (netns < 0) { |
1706 | r = netns; | |
1707 | goto fail; | |
613b411c | 1708 | |
3ee897d6 LP |
1709 | } else { |
1710 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1711 | if (setns(netns, CLONE_NEWNET) < 0) { |
1712 | r = -errno; | |
1713 | goto fail; | |
1714 | } | |
1715 | ||
1716 | r = 0; | |
1717 | } | |
1718 | ||
3ee897d6 LP |
1719 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1720 | if (q < 0) { | |
1721 | r = q; | |
613b411c LP |
1722 | goto fail; |
1723 | } | |
1724 | ||
1725 | fail: | |
fe048ce5 | 1726 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1727 | return r; |
1728 | } | |
417116f2 | 1729 | |
51af7fb2 LP |
1730 | int open_netns_path(int netns_storage_socket[static 2], const char *path) { |
1731 | _cleanup_close_ int netns = -1; | |
1732 | int q, r; | |
1733 | ||
1734 | assert(netns_storage_socket); | |
1735 | assert(netns_storage_socket[0] >= 0); | |
1736 | assert(netns_storage_socket[1] >= 0); | |
1737 | assert(path); | |
1738 | ||
1739 | /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in | |
1740 | * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a | |
1741 | * new anonymous netns if needed. */ | |
1742 | ||
1743 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1744 | return -errno; | |
1745 | ||
1746 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); | |
1747 | if (netns == -EAGAIN) { | |
1748 | /* Nothing stored yet. Open the file from the file system. */ | |
1749 | ||
1750 | netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
1751 | if (netns < 0) { | |
1752 | r = -errno; | |
1753 | goto fail; | |
1754 | } | |
1755 | ||
1756 | r = fd_is_network_ns(netns); | |
1757 | if (r == 0) { /* Not a netns? Refuse early. */ | |
1758 | r = -EINVAL; | |
1759 | goto fail; | |
1760 | } | |
1761 | if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */ | |
1762 | goto fail; | |
1763 | ||
1764 | r = 1; | |
1765 | ||
1766 | } else if (netns < 0) { | |
1767 | r = netns; | |
1768 | goto fail; | |
1769 | } else | |
1770 | r = 0; /* Already allocated */ | |
1771 | ||
1772 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); | |
1773 | if (q < 0) { | |
1774 | r = q; | |
1775 | goto fail; | |
1776 | } | |
1777 | ||
1778 | fail: | |
1779 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); | |
1780 | return r; | |
1781 | } | |
1782 | ||
6e2d7c4f MS |
1783 | bool ns_type_supported(NamespaceType type) { |
1784 | const char *t, *ns_proc; | |
1785 | ||
0fa5b831 LP |
1786 | t = namespace_type_to_string(type); |
1787 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
1788 | return false; |
1789 | ||
6e2d7c4f | 1790 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
1791 | return access(ns_proc, F_OK) == 0; |
1792 | } | |
1793 | ||
1b8689f9 LP |
1794 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1795 | [PROTECT_HOME_NO] = "no", | |
1796 | [PROTECT_HOME_YES] = "yes", | |
1797 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
e4da7d8c | 1798 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
1799 | }; |
1800 | ||
1e8c7bd5 | 1801 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 1802 | |
1b8689f9 LP |
1803 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
1804 | [PROTECT_SYSTEM_NO] = "no", | |
1805 | [PROTECT_SYSTEM_YES] = "yes", | |
1806 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1807 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1808 | }; |
1809 | ||
1e8c7bd5 | 1810 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 1811 | |
6e2d7c4f MS |
1812 | static const char* const namespace_type_table[] = { |
1813 | [NAMESPACE_MOUNT] = "mnt", | |
1814 | [NAMESPACE_CGROUP] = "cgroup", | |
1815 | [NAMESPACE_UTS] = "uts", | |
1816 | [NAMESPACE_IPC] = "ipc", | |
1817 | [NAMESPACE_USER] = "user", | |
1818 | [NAMESPACE_PID] = "pid", | |
1819 | [NAMESPACE_NET] = "net", | |
1820 | }; | |
1821 | ||
1822 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); |