]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
15ae422b LP |
2 | |
3 | #include <errno.h> | |
e08f94ac | 4 | #include <linux/loop.h> |
07630cea | 5 | #include <sched.h> |
15ae422b | 6 | #include <stdio.h> |
07630cea | 7 | #include <sys/mount.h> |
07630cea | 8 | #include <unistd.h> |
25e870b5 | 9 | #include <linux/fs.h> |
15ae422b | 10 | |
b5efdb8a | 11 | #include "alloc-util.h" |
10404d52 | 12 | #include "base-filesystem.h" |
7f112f50 | 13 | #include "dev-setup.h" |
3ffd4af2 | 14 | #include "fd-util.h" |
e5f10caf | 15 | #include "format-util.h" |
d944dc95 | 16 | #include "fs-util.h" |
e908468b | 17 | #include "label.h" |
915e6d16 | 18 | #include "loop-util.h" |
07630cea | 19 | #include "loopback-setup.h" |
07630cea | 20 | #include "mkdir.h" |
4349cd7c | 21 | #include "mount-util.h" |
049af8ad | 22 | #include "mountpoint-util.h" |
0cb8e3d1 | 23 | #include "namespace-util.h" |
3ffd4af2 | 24 | #include "namespace.h" |
d8b4d14d | 25 | #include "nulstr-util.h" |
07630cea | 26 | #include "path-util.h" |
d7b8eec7 | 27 | #include "selinux-util.h" |
2583fbea | 28 | #include "socket-util.h" |
760877e9 | 29 | #include "sort-util.h" |
36ce7110 | 30 | #include "stat-util.h" |
8b43440b | 31 | #include "string-table.h" |
07630cea LP |
32 | #include "string-util.h" |
33 | #include "strv.h" | |
a652f050 | 34 | #include "tmpfile-util.h" |
affb60b1 | 35 | #include "umask-util.h" |
ee104e11 | 36 | #include "user-util.h" |
15ae422b | 37 | |
737ba3c8 | 38 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
39 | ||
c17ec25e | 40 | typedef enum MountMode { |
15ae422b LP |
41 | /* This is ordered by priority! */ |
42 | INACCESSIBLE, | |
d2d6c096 LP |
43 | BIND_MOUNT, |
44 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 45 | PRIVATE_TMP, |
7f112f50 | 46 | PRIVATE_DEV, |
5d997827 | 47 | BIND_DEV, |
6c47cd7d | 48 | EMPTY_DIR, |
5d997827 LP |
49 | SYSFS, |
50 | PROCFS, | |
51 | READONLY, | |
59eeb84b | 52 | READWRITE, |
2abd4e38 | 53 | TMPFS, |
1e05071d | 54 | READWRITE_IMPLICIT, /* Should have the lowest priority. */ |
5beb8688 | 55 | _MOUNT_MODE_MAX, |
c17ec25e | 56 | } MountMode; |
15ae422b | 57 | |
34de407a | 58 | typedef struct MountEntry { |
5327c910 | 59 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 60 | MountMode mode:5; |
5327c910 LP |
61 | bool ignore:1; /* Ignore if path does not exist? */ |
62 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 63 | bool read_only:1; /* Shall this mount point be read-only? */ |
9ce4e4b0 | 64 | bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */ |
088696fe | 65 | bool applied:1; /* Already applied */ |
55fe7432 | 66 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
d2d6c096 LP |
67 | const char *source_const; /* The source path, for bind mounts */ |
68 | char *source_malloc; | |
2abd4e38 YW |
69 | const char *options_const;/* Mount options for tmpfs */ |
70 | char *options_malloc; | |
71 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 72 | unsigned n_followed; |
34de407a | 73 | } MountEntry; |
15ae422b | 74 | |
5d997827 | 75 | /* 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 | 76 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
77 | static const MountEntry apivfs_table[] = { |
78 | { "/proc", PROCFS, false }, | |
79 | { "/dev", BIND_DEV, false }, | |
80 | { "/sys", SYSFS, false }, | |
81 | }; | |
f471b2af | 82 | |
11a30cec | 83 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 84 | static const MountEntry protect_kernel_tunables_table[] = { |
1e05071d YW |
85 | { "/proc/acpi", READONLY, true }, |
86 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ | |
87 | { "/proc/asound", READONLY, true }, | |
88 | { "/proc/bus", READONLY, true }, | |
89 | { "/proc/fs", READONLY, true }, | |
90 | { "/proc/irq", READONLY, true }, | |
91 | { "/proc/kallsyms", INACCESSIBLE, true }, | |
92 | { "/proc/kcore", INACCESSIBLE, true }, | |
93 | { "/proc/latency_stats", READONLY, true }, | |
94 | { "/proc/mtrr", READONLY, true }, | |
95 | { "/proc/scsi", READONLY, true }, | |
96 | { "/proc/sys", READONLY, false }, | |
97 | { "/proc/sysrq-trigger", READONLY, true }, | |
98 | { "/proc/timer_stats", READONLY, true }, | |
99 | { "/sys", READONLY, false }, | |
100 | { "/sys/fs/bpf", READONLY, true }, | |
101 | { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */ | |
102 | { "/sys/fs/selinux", READWRITE_IMPLICIT, true }, | |
103 | { "/sys/kernel/debug", READONLY, true }, | |
104 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
105 | }; |
106 | ||
c575770b | 107 | /* ProtectKernelModules= option */ |
34de407a | 108 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 109 | #if HAVE_SPLIT_USR |
c6232fb0 | 110 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 111 | #endif |
c6232fb0 | 112 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
113 | }; |
114 | ||
94a7b275 KK |
115 | /* ProtectKernelLogs= option */ |
116 | static const MountEntry protect_kernel_logs_table[] = { | |
117 | { "/proc/kmsg", INACCESSIBLE, true }, | |
118 | { "/dev/kmsg", INACCESSIBLE, true }, | |
119 | }; | |
120 | ||
b6c432ca DH |
121 | /* |
122 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
123 | * system should be protected by ProtectSystem= | |
124 | */ | |
34de407a | 125 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
126 | { "/home", READONLY, true }, |
127 | { "/run/user", READONLY, true }, | |
128 | { "/root", READONLY, true }, | |
b6c432ca DH |
129 | }; |
130 | ||
e4da7d8c YW |
131 | /* ProtectHome=tmpfs table */ |
132 | static const MountEntry protect_home_tmpfs_table[] = { | |
133 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
134 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755", .flags = MS_NODEV|MS_STRICTATIME }, | |
135 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700", .flags = MS_NODEV|MS_STRICTATIME }, | |
136 | }; | |
137 | ||
b6c432ca | 138 | /* ProtectHome=yes table */ |
34de407a | 139 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
140 | { "/home", INACCESSIBLE, true }, |
141 | { "/run/user", INACCESSIBLE, true }, | |
142 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
143 | }; |
144 | ||
f471b2af | 145 | /* ProtectSystem=yes table */ |
34de407a | 146 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
147 | { "/usr", READONLY, false }, |
148 | { "/boot", READONLY, true }, | |
149 | { "/efi", READONLY, true }, | |
7486f305 AB |
150 | #if HAVE_SPLIT_USR |
151 | { "/lib", READONLY, true }, | |
152 | { "/lib64", READONLY, true }, | |
153 | { "/bin", READONLY, true }, | |
671f0f8d | 154 | # if HAVE_SPLIT_BIN |
7486f305 | 155 | { "/sbin", READONLY, true }, |
671f0f8d | 156 | # endif |
7486f305 | 157 | #endif |
f471b2af DH |
158 | }; |
159 | ||
160 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 161 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
162 | { "/usr", READONLY, false }, |
163 | { "/boot", READONLY, true }, | |
164 | { "/efi", READONLY, true }, | |
165 | { "/etc", READONLY, false }, | |
7486f305 AB |
166 | #if HAVE_SPLIT_USR |
167 | { "/lib", READONLY, true }, | |
168 | { "/lib64", READONLY, true }, | |
169 | { "/bin", READONLY, true }, | |
671f0f8d | 170 | # if HAVE_SPLIT_BIN |
7486f305 | 171 | { "/sbin", READONLY, true }, |
671f0f8d | 172 | # endif |
7486f305 | 173 | #endif |
f471b2af DH |
174 | }; |
175 | ||
176 | /* | |
177 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
178 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
179 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
180 | * protect those, and these options should be fully orthogonal. | |
181 | * (And of course /home and friends are also left writable, as ProtectHome= | |
182 | * shall manage those, orthogonally). | |
183 | */ | |
34de407a | 184 | static const MountEntry protect_system_strict_table[] = { |
1e05071d YW |
185 | { "/", READONLY, false }, |
186 | { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
187 | { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
188 | { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */ | |
189 | { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
190 | { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
191 | { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
f471b2af DH |
192 | }; |
193 | ||
5beb8688 YW |
194 | static const char * const mount_mode_table[_MOUNT_MODE_MAX] = { |
195 | [INACCESSIBLE] = "inaccessible", | |
196 | [BIND_MOUNT] = "bind", | |
197 | [BIND_MOUNT_RECURSIVE] = "rbind", | |
198 | [PRIVATE_TMP] = "private-tmp", | |
199 | [PRIVATE_DEV] = "private-dev", | |
200 | [BIND_DEV] = "bind-dev", | |
201 | [EMPTY_DIR] = "empty", | |
202 | [SYSFS] = "sysfs", | |
203 | [PROCFS] = "procfs", | |
204 | [READONLY] = "read-only", | |
205 | [READWRITE] = "read-write", | |
206 | [TMPFS] = "tmpfs", | |
207 | [READWRITE_IMPLICIT] = "rw-implicit", | |
208 | }; | |
209 | ||
210 | DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode); | |
211 | ||
34de407a | 212 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
213 | assert(p); |
214 | ||
5327c910 LP |
215 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
216 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 217 | |
5327c910 | 218 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
219 | } |
220 | ||
34de407a | 221 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
222 | assert(p); |
223 | ||
224 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE); | |
225 | } | |
226 | ||
d2d6c096 LP |
227 | static const char *mount_entry_source(const MountEntry *p) { |
228 | assert(p); | |
229 | ||
230 | return p->source_malloc ?: p->source_const; | |
231 | } | |
232 | ||
2abd4e38 YW |
233 | static const char *mount_entry_options(const MountEntry *p) { |
234 | assert(p); | |
235 | ||
236 | return p->options_malloc ?: p->options_const; | |
237 | } | |
238 | ||
1eb7e08e LP |
239 | static void mount_entry_done(MountEntry *p) { |
240 | assert(p); | |
241 | ||
242 | p->path_malloc = mfree(p->path_malloc); | |
243 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 244 | p->options_malloc = mfree(p->options_malloc); |
1eb7e08e LP |
245 | } |
246 | ||
d18aff04 | 247 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
248 | char **i; |
249 | ||
613b411c LP |
250 | assert(p); |
251 | ||
1e05071d | 252 | /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */ |
5327c910 | 253 | |
15ae422b | 254 | STRV_FOREACH(i, strv) { |
5327c910 LP |
255 | bool ignore = false, needs_prefix = false; |
256 | const char *e = *i; | |
15ae422b | 257 | |
5327c910 LP |
258 | /* Look for any prefixes */ |
259 | if (startswith(e, "-")) { | |
260 | e++; | |
9c94d52e | 261 | ignore = true; |
ea92ae33 | 262 | } |
5327c910 LP |
263 | if (startswith(e, "+")) { |
264 | e++; | |
265 | needs_prefix = true; | |
266 | } | |
ea92ae33 | 267 | |
baaa35ad ZJS |
268 | if (!path_is_absolute(e)) |
269 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
270 | "Path is not absolute: %s", e); | |
15ae422b | 271 | |
34de407a | 272 | *((*p)++) = (MountEntry) { |
5327c910 LP |
273 | .path_const = e, |
274 | .mode = mode, | |
275 | .ignore = ignore, | |
d18aff04 | 276 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 277 | }; |
15ae422b LP |
278 | } |
279 | ||
280 | return 0; | |
281 | } | |
282 | ||
6c47cd7d LP |
283 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
284 | char **i; | |
285 | ||
286 | assert(p); | |
287 | ||
288 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
289 | * "/private/" boundary directories for DynamicUser=1. */ | |
290 | ||
291 | STRV_FOREACH(i, strv) { | |
292 | ||
293 | *((*p)++) = (MountEntry) { | |
294 | .path_const = *i, | |
295 | .mode = EMPTY_DIR, | |
296 | .ignore = false, | |
6c47cd7d | 297 | .read_only = true, |
2abd4e38 YW |
298 | .options_const = "mode=755", |
299 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, | |
6c47cd7d LP |
300 | }; |
301 | } | |
302 | ||
303 | return 0; | |
304 | } | |
305 | ||
da6053d0 LP |
306 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
307 | size_t i; | |
d2d6c096 LP |
308 | |
309 | assert(p); | |
310 | ||
311 | for (i = 0; i < n; i++) { | |
312 | const BindMount *b = binds + i; | |
313 | ||
314 | *((*p)++) = (MountEntry) { | |
315 | .path_const = b->destination, | |
316 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
317 | .read_only = b->read_only, | |
9ce4e4b0 | 318 | .nosuid = b->nosuid, |
d2d6c096 | 319 | .source_const = b->source, |
4ca763a9 | 320 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
321 | }; |
322 | } | |
323 | ||
324 | return 0; | |
325 | } | |
326 | ||
da6053d0 LP |
327 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
328 | size_t i; | |
2abd4e38 YW |
329 | int r; |
330 | ||
331 | assert(p); | |
332 | ||
333 | for (i = 0; i < n; i++) { | |
334 | const TemporaryFileSystem *t = tmpfs + i; | |
335 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 336 | unsigned long flags; |
2abd4e38 YW |
337 | bool ro = false; |
338 | ||
baaa35ad ZJS |
339 | if (!path_is_absolute(t->path)) |
340 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
341 | "Path is not absolute: %s", | |
342 | t->path); | |
2abd4e38 | 343 | |
ad8e66dc AJ |
344 | str = strjoin("mode=0755,", t->options); |
345 | if (!str) | |
346 | return -ENOMEM; | |
2abd4e38 | 347 | |
ad8e66dc AJ |
348 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
349 | if (r < 0) | |
350 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 351 | |
ad8e66dc AJ |
352 | ro = flags & MS_RDONLY; |
353 | if (ro) | |
354 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
355 | |
356 | *((*p)++) = (MountEntry) { | |
357 | .path_const = t->path, | |
358 | .mode = TMPFS, | |
359 | .read_only = ro, | |
ad8e66dc | 360 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
361 | .flags = flags, |
362 | }; | |
2abd4e38 YW |
363 | } |
364 | ||
365 | return 0; | |
366 | } | |
367 | ||
da6053d0 LP |
368 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
369 | size_t i; | |
11a30cec DH |
370 | |
371 | assert(p); | |
f471b2af | 372 | assert(mounts); |
11a30cec | 373 | |
5327c910 | 374 | /* Adds a list of static pre-defined entries */ |
f471b2af | 375 | |
5327c910 | 376 | for (i = 0; i < n; i++) |
34de407a LP |
377 | *((*p)++) = (MountEntry) { |
378 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
379 | .mode = mounts[i].mode, |
380 | .ignore = mounts[i].ignore || ignore_protect, | |
381 | }; | |
f471b2af DH |
382 | |
383 | return 0; | |
384 | } | |
385 | ||
34de407a | 386 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
387 | assert(p); |
388 | ||
5327c910 | 389 | switch (protect_home) { |
b6c432ca | 390 | |
5327c910 | 391 | case PROTECT_HOME_NO: |
b6c432ca DH |
392 | return 0; |
393 | ||
b6c432ca | 394 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
395 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
396 | ||
e4da7d8c YW |
397 | case PROTECT_HOME_TMPFS: |
398 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
399 | ||
b6c432ca | 400 | case PROTECT_HOME_YES: |
5327c910 LP |
401 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
402 | ||
b6c432ca | 403 | default: |
5327c910 | 404 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 405 | } |
b6c432ca DH |
406 | } |
407 | ||
34de407a | 408 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
409 | assert(p); |
410 | ||
5327c910 LP |
411 | switch (protect_system) { |
412 | ||
413 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
414 | return 0; |
415 | ||
f471b2af | 416 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
417 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
418 | ||
f471b2af | 419 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
420 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
421 | ||
f471b2af | 422 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
423 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
424 | ||
f471b2af | 425 | default: |
5327c910 | 426 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 427 | } |
11a30cec DH |
428 | } |
429 | ||
93bab288 | 430 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 431 | int d; |
15ae422b | 432 | |
6ee1a919 | 433 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 434 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
435 | if (d != 0) |
436 | return d; | |
15ae422b | 437 | |
6ee1a919 | 438 | /* If the paths are equal, check the mode */ |
93bab288 | 439 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
440 | } |
441 | ||
da6053d0 LP |
442 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
443 | size_t i; | |
5327c910 | 444 | |
4a756839 | 445 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 LP |
446 | |
447 | for (i = 0; i < n; i++) { | |
448 | char *s; | |
449 | ||
450 | if (m[i].has_prefix) | |
451 | continue; | |
452 | ||
c6134d3e | 453 | s = path_join(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
454 | if (!s) |
455 | return -ENOMEM; | |
456 | ||
e282f51f | 457 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
458 | m[i].has_prefix = true; |
459 | } | |
460 | ||
461 | return 0; | |
462 | } | |
463 | ||
da6053d0 | 464 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 465 | MountEntry *f, *t, *previous; |
15ae422b | 466 | |
c17ec25e | 467 | assert(m); |
15ae422b | 468 | assert(n); |
15ae422b | 469 | |
fe3c2583 LP |
470 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
471 | ||
1d54cd5d | 472 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 473 | |
fe3c2583 | 474 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
475 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
476 | if (previous && | |
477 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
478 | !f->applied && !previous->applied) { | |
5beb8688 | 479 | log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode)); |
34de407a | 480 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ |
1eb7e08e | 481 | mount_entry_done(f); |
15ae422b | 482 | continue; |
fe3c2583 | 483 | } |
15ae422b | 484 | |
e2d7c1a0 | 485 | *t = *f; |
15ae422b | 486 | previous = t; |
fe3c2583 LP |
487 | t++; |
488 | } | |
489 | ||
490 | *n = t - m; | |
491 | } | |
492 | ||
da6053d0 | 493 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 494 | MountEntry *f, *t; |
fe3c2583 LP |
495 | const char *clear = NULL; |
496 | ||
497 | assert(m); | |
498 | assert(n); | |
499 | ||
500 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
501 | * ordered already. */ | |
502 | ||
1d54cd5d | 503 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
504 | |
505 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
506 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
507 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
508 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 509 | mount_entry_done(f); |
fe3c2583 LP |
510 | continue; |
511 | } | |
15ae422b | 512 | |
34de407a | 513 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
514 | |
515 | *t = *f; | |
15ae422b LP |
516 | t++; |
517 | } | |
518 | ||
c17ec25e | 519 | *n = t - m; |
15ae422b LP |
520 | } |
521 | ||
da6053d0 | 522 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 523 | MountEntry *f, *t; |
7648a565 LP |
524 | |
525 | assert(m); | |
526 | assert(n); | |
527 | ||
528 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
529 | * list is ordered by prefixes. */ | |
530 | ||
1d54cd5d | 531 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 | 532 | |
1e05071d YW |
533 | /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */ |
534 | if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) { | |
34de407a | 535 | MountEntry *p; |
7648a565 LP |
536 | bool found = false; |
537 | ||
538 | /* Now let's find the first parent of the entry we are looking at. */ | |
539 | for (p = t-1; p >= m; p--) { | |
34de407a | 540 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
541 | found = true; |
542 | break; | |
543 | } | |
544 | } | |
545 | ||
546 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
547 | if (found && p->mode == f->mode) { | |
5beb8688 YW |
548 | log_debug("%s (%s) is made redundant by %s (%s)", |
549 | mount_entry_path(f), mount_mode_to_string(f->mode), | |
550 | mount_entry_path(p), mount_mode_to_string(p->mode)); | |
1eb7e08e | 551 | mount_entry_done(f); |
7648a565 LP |
552 | continue; |
553 | } | |
554 | } | |
555 | ||
556 | *t = *f; | |
557 | t++; | |
558 | } | |
559 | ||
560 | *n = t - m; | |
561 | } | |
562 | ||
da6053d0 | 563 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 564 | MountEntry *f, *t; |
cd2902c9 LP |
565 | |
566 | assert(m); | |
567 | assert(n); | |
568 | ||
1d54cd5d | 569 | /* Nothing to do */ |
cd2902c9 LP |
570 | if (!root_directory) |
571 | return; | |
572 | ||
573 | /* Drops all mounts that are outside of the root directory. */ | |
574 | ||
1d54cd5d | 575 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 576 | |
34de407a LP |
577 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
578 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 579 | mount_entry_done(f); |
cd2902c9 LP |
580 | continue; |
581 | } | |
582 | ||
583 | *t = *f; | |
584 | t++; | |
585 | } | |
586 | ||
587 | *n = t - m; | |
588 | } | |
589 | ||
b2a60844 LP |
590 | static int clone_device_node( |
591 | const char *d, | |
592 | const char *temporary_mount, | |
593 | bool *make_devnode) { | |
594 | ||
595 | _cleanup_free_ char *sl = NULL; | |
596 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
597 | struct stat st; |
598 | int r; | |
599 | ||
414b304b | 600 | if (stat(d, &st) < 0) { |
b2a60844 LP |
601 | if (errno == ENOENT) { |
602 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 603 | return -ENXIO; |
b2a60844 LP |
604 | } |
605 | ||
606 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
607 | } |
608 | ||
609 | if (!S_ISBLK(st.st_mode) && | |
baaa35ad ZJS |
610 | !S_ISCHR(st.st_mode)) |
611 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
612 | "Device node '%s' to clone is not a device node, ignoring.", | |
613 | d); | |
b5e99f23 | 614 | |
6f7f3a33 | 615 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 616 | |
b2a60844 | 617 | /* First, try to create device node properly */ |
16498617 CB |
618 | if (*make_devnode) { |
619 | mac_selinux_create_file_prepare(d, st.st_mode); | |
620 | r = mknod(dn, st.st_mode, st.st_rdev); | |
621 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
622 | if (r >= 0) |
623 | goto add_symlink; | |
16498617 CB |
624 | if (errno != EPERM) |
625 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
626 | ||
b2a60844 | 627 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
628 | *make_devnode = false; |
629 | } | |
630 | ||
631 | /* We're about to fallback to bind-mounting the device | |
1acf344d CG |
632 | * node. So create a dummy bind-mount target. |
633 | * Do not prepare device-node SELinux label (see issue 13762) */ | |
16498617 | 634 | r = mknod(dn, S_IFREG, 0); |
16498617 | 635 | if (r < 0 && errno != EEXIST) |
b2a60844 | 636 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 CB |
637 | |
638 | /* Fallback to bind-mounting: | |
639 | * The assumption here is that all used device nodes carry standard | |
640 | * properties. Specifically, the devices nodes we bind-mount should | |
641 | * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx) | |
642 | * and should not carry ACLs. */ | |
643 | if (mount(d, dn, NULL, MS_BIND, NULL) < 0) | |
b2a60844 LP |
644 | return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d); |
645 | ||
646 | add_symlink: | |
647 | bn = path_startswith(d, "/dev/"); | |
648 | if (!bn) | |
649 | return 0; | |
650 | ||
651 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
652 | 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) | |
653 | return log_oom(); | |
654 | ||
655 | (void) mkdir_parents(sl, 0755); | |
656 | ||
657 | t = strjoina("../", bn); | |
658 | ||
659 | if (symlink(t, sl) < 0) | |
2e4a4fae | 660 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 661 | |
af984e13 | 662 | return 0; |
b5e99f23 ДГ |
663 | } |
664 | ||
5d997827 | 665 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
666 | static const char devnodes[] = |
667 | "/dev/null\0" | |
668 | "/dev/zero\0" | |
669 | "/dev/full\0" | |
670 | "/dev/random\0" | |
671 | "/dev/urandom\0" | |
672 | "/dev/tty\0"; | |
673 | ||
2b85f4e1 | 674 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 675 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 676 | bool can_mknod = true; |
7f112f50 LP |
677 | _cleanup_umask_ mode_t u; |
678 | int r; | |
679 | ||
680 | assert(m); | |
681 | ||
682 | u = umask(0000); | |
683 | ||
2b85f4e1 | 684 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 685 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 686 | |
63c372cb | 687 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 688 | (void) mkdir(dev, 0755); |
737ba3c8 | 689 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755") < 0) { |
2e4a4fae | 690 | r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev); |
2b85f4e1 LP |
691 | goto fail; |
692 | } | |
c3151977 TM |
693 | r = label_fix_container(dev, "/dev", 0); |
694 | if (r < 0) { | |
695 | log_debug_errno(errno, "Failed to fix label of '%s' as /dev: %m", dev); | |
696 | goto fail; | |
697 | } | |
2b85f4e1 | 698 | |
63c372cb | 699 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 700 | (void) mkdir(devpts, 0755); |
2b85f4e1 | 701 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
2e4a4fae | 702 | r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts); |
2b85f4e1 LP |
703 | goto fail; |
704 | } | |
705 | ||
2e4a4fae YW |
706 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx. |
707 | * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible. | |
708 | * Thus, in that case make a clone. | |
709 | * In nspawn and other containers it will be a symlink, in that case make it a symlink. */ | |
36ce7110 | 710 | r = is_symlink("/dev/ptmx"); |
2e4a4fae YW |
711 | if (r < 0) { |
712 | log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m"); | |
3164e3cb | 713 | goto fail; |
2e4a4fae | 714 | } else if (r > 0) { |
414b304b ДГ |
715 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
716 | if (symlink("pts/ptmx", devptmx) < 0) { | |
2e4a4fae | 717 | r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx); |
414b304b ДГ |
718 | goto fail; |
719 | } | |
720 | } else { | |
16498617 | 721 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
722 | if (r < 0) |
723 | goto fail; | |
414b304b | 724 | } |
e06b6479 | 725 | |
63c372cb | 726 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 727 | (void) mkdir(devshm, 0755); |
2b85f4e1 LP |
728 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
729 | if (r < 0) { | |
2e4a4fae | 730 | r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm); |
2b85f4e1 LP |
731 | goto fail; |
732 | } | |
733 | ||
63c372cb | 734 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 735 | (void) mkdir(devmqueue, 0755); |
2e4a4fae YW |
736 | if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0) |
737 | log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue); | |
2b85f4e1 | 738 | |
63c372cb | 739 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 740 | (void) mkdir(devhugepages, 0755); |
2e4a4fae YW |
741 | if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0) |
742 | log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages); | |
2b85f4e1 | 743 | |
63c372cb | 744 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
745 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
746 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 747 | |
7f112f50 | 748 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 749 | r = clone_device_node(d, temporary_mount, &can_mknod); |
af984e13 ZJS |
750 | /* ENXIO means the the *source* is not a device file, skip creation in that case */ |
751 | if (r < 0 && r != -ENXIO) | |
2b85f4e1 | 752 | goto fail; |
7f112f50 LP |
753 | } |
754 | ||
2e4a4fae YW |
755 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
756 | if (r < 0) | |
105a1a36 | 757 | log_debug_errno(r, "Failed to set up basic device tree at '%s', ignoring: %m", temporary_mount); |
7f112f50 | 758 | |
ee818b89 AC |
759 | /* Create the /dev directory if missing. It is more likely to be |
760 | * missing when the service is started with RootDirectory. This is | |
761 | * consistent with mount units creating the mount points when missing. | |
762 | */ | |
34de407a | 763 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 764 | |
9e5f8252 | 765 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
766 | r = umount_recursive(mount_entry_path(m), 0); |
767 | if (r < 0) | |
768 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
769 | ||
34de407a | 770 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { |
2e4a4fae | 771 | r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m)); |
2b85f4e1 LP |
772 | goto fail; |
773 | } | |
7f112f50 | 774 | |
1019a48f LP |
775 | (void) rmdir(dev); |
776 | (void) rmdir(temporary_mount); | |
7f112f50 | 777 | |
2b85f4e1 | 778 | return 0; |
7f112f50 | 779 | |
2b85f4e1 LP |
780 | fail: |
781 | if (devpts) | |
1019a48f | 782 | (void) umount(devpts); |
7f112f50 | 783 | |
2b85f4e1 | 784 | if (devshm) |
1019a48f | 785 | (void) umount(devshm); |
7f112f50 | 786 | |
2b85f4e1 | 787 | if (devhugepages) |
1019a48f | 788 | (void) umount(devhugepages); |
7f112f50 | 789 | |
2b85f4e1 | 790 | if (devmqueue) |
1019a48f | 791 | (void) umount(devmqueue); |
7f112f50 | 792 | |
1019a48f LP |
793 | (void) umount(dev); |
794 | (void) rmdir(dev); | |
795 | (void) rmdir(temporary_mount); | |
7f112f50 | 796 | |
2b85f4e1 | 797 | return r; |
7f112f50 LP |
798 | } |
799 | ||
2a2969fd | 800 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
801 | int r; |
802 | ||
803 | assert(m); | |
804 | ||
805 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
806 | * /dev. This is only used when RootDirectory= is set. */ | |
807 | ||
645767d6 LP |
808 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
809 | ||
5d997827 LP |
810 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
811 | if (r < 0) | |
812 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
813 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
814 | return 0; | |
815 | ||
816 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
817 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
818 | ||
819 | return 1; | |
820 | } | |
821 | ||
2a2969fd | 822 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
823 | int r; |
824 | ||
825 | assert(m); | |
826 | ||
645767d6 LP |
827 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
828 | ||
5d997827 LP |
829 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
830 | if (r < 0) | |
831 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
832 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
833 | return 0; | |
834 | ||
835 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
836 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
837 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
838 | ||
839 | return 1; | |
840 | } | |
841 | ||
2a2969fd | 842 | static int mount_procfs(const MountEntry *m) { |
5d997827 LP |
843 | int r; |
844 | ||
845 | assert(m); | |
846 | ||
645767d6 LP |
847 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
848 | ||
5d997827 LP |
849 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
850 | if (r < 0) | |
851 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
852 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
853 | return 0; | |
854 | ||
855 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
856 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
857 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
858 | ||
859 | return 1; | |
860 | } | |
861 | ||
2abd4e38 | 862 | static int mount_tmpfs(const MountEntry *m) { |
6c47cd7d LP |
863 | assert(m); |
864 | ||
2abd4e38 | 865 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d LP |
866 | |
867 | (void) mkdir_p_label(mount_entry_path(m), 0755); | |
868 | (void) umount_recursive(mount_entry_path(m), 0); | |
869 | ||
2abd4e38 | 870 | if (mount("tmpfs", mount_entry_path(m), "tmpfs", m->flags, mount_entry_options(m)) < 0) |
6c47cd7d LP |
871 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); |
872 | ||
873 | return 1; | |
874 | } | |
875 | ||
088696fe | 876 | static int follow_symlink( |
d2d6c096 | 877 | const char *root_directory, |
088696fe | 878 | MountEntry *m) { |
d2d6c096 | 879 | |
088696fe | 880 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
881 | int r; |
882 | ||
088696fe LP |
883 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
884 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
885 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
886 | * end and already have a fully normalized name. */ | |
8fceda93 | 887 | |
a5648b80 | 888 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL); |
088696fe LP |
889 | if (r < 0) |
890 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
891 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
892 | return 1; | |
8fceda93 | 893 | |
baaa35ad ZJS |
894 | if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */ |
895 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
896 | "Symlink loop on '%s'.", | |
897 | mount_entry_path(m)); | |
8fceda93 | 898 | |
088696fe | 899 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 900 | |
088696fe LP |
901 | free_and_replace(m->path_malloc, target); |
902 | m->has_prefix = true; | |
8fceda93 | 903 | |
088696fe LP |
904 | m->n_followed ++; |
905 | ||
906 | return 0; | |
8fceda93 LP |
907 | } |
908 | ||
ac0930c8 | 909 | static int apply_mount( |
8fceda93 | 910 | const char *root_directory, |
89bd586c | 911 | MountEntry *m) { |
ac0930c8 | 912 | |
e5f10caf | 913 | _cleanup_free_ char *inaccessible = NULL; |
a227a4be | 914 | bool rbind = true, make = false; |
15ae422b | 915 | const char *what; |
15ae422b | 916 | int r; |
15ae422b | 917 | |
c17ec25e | 918 | assert(m); |
15ae422b | 919 | |
34de407a | 920 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 921 | |
c17ec25e | 922 | switch (m->mode) { |
15ae422b | 923 | |
160cfdbe | 924 | case INACCESSIBLE: { |
e5f10caf AZ |
925 | _cleanup_free_ char *tmp = NULL; |
926 | const char *runtime_dir; | |
160cfdbe | 927 | struct stat target; |
6d313367 LP |
928 | |
929 | /* First, get rid of everything that is below if there | |
930 | * is anything... Then, overmount it with an | |
c4b41707 | 931 | * inaccessible path. */ |
34de407a | 932 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 933 | |
088696fe LP |
934 | if (lstat(mount_entry_path(m), &target) < 0) { |
935 | if (errno == ENOENT && m->ignore) | |
936 | return 0; | |
937 | ||
34de407a | 938 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", mount_entry_path(m)); |
088696fe | 939 | } |
15ae422b | 940 | |
e5f10caf AZ |
941 | if (geteuid() == 0) |
942 | runtime_dir = "/run/systemd"; | |
943 | else { | |
944 | if (asprintf(&tmp, "/run/user/"UID_FMT, geteuid()) < 0) | |
945 | log_oom(); | |
946 | ||
947 | runtime_dir = tmp; | |
948 | } | |
949 | ||
950 | r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible); | |
951 | if (r < 0) | |
baaa35ad ZJS |
952 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), |
953 | "File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
e5f10caf | 954 | what = inaccessible; |
c4b41707 | 955 | break; |
160cfdbe | 956 | } |
fe3c2583 | 957 | |
15ae422b | 958 | case READONLY: |
15ae422b | 959 | case READWRITE: |
1e05071d | 960 | case READWRITE_IMPLICIT: |
8fceda93 | 961 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
962 | if (r == -ENOENT && m->ignore) |
963 | return 0; | |
d944dc95 | 964 | if (r < 0) |
34de407a | 965 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", mount_entry_path(m)); |
6b7c9f8b LP |
966 | 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. */ |
967 | return 0; | |
6b7c9f8b | 968 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 969 | what = mount_entry_path(m); |
6b7c9f8b | 970 | break; |
15ae422b | 971 | |
d2d6c096 LP |
972 | case BIND_MOUNT: |
973 | rbind = false; | |
d2d6c096 | 974 | |
4831981d | 975 | _fallthrough_; |
088696fe LP |
976 | case BIND_MOUNT_RECURSIVE: { |
977 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 978 | |
088696fe LP |
979 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note that bind |
980 | * mount source paths are always relative to the host root, hence we pass NULL as root directory to | |
981 | * chase_symlinks() here. */ | |
982 | ||
a5648b80 | 983 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL); |
088696fe LP |
984 | if (r == -ENOENT && m->ignore) { |
985 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
986 | return 0; | |
987 | } | |
988 | if (r < 0) | |
989 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
990 | ||
991 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
992 | ||
993 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
994 | |
995 | what = mount_entry_source(m); | |
a227a4be | 996 | make = true; |
d2d6c096 | 997 | break; |
088696fe | 998 | } |
d2d6c096 | 999 | |
6c47cd7d | 1000 | case EMPTY_DIR: |
2abd4e38 YW |
1001 | case TMPFS: |
1002 | return mount_tmpfs(m); | |
6c47cd7d | 1003 | |
ac0930c8 | 1004 | case PRIVATE_TMP: |
89bd586c | 1005 | what = mount_entry_source(m); |
a227a4be | 1006 | make = true; |
15ae422b | 1007 | break; |
e364ad06 | 1008 | |
d6797c92 | 1009 | case PRIVATE_DEV: |
5d997827 LP |
1010 | return mount_private_dev(m); |
1011 | ||
1012 | case BIND_DEV: | |
1013 | return mount_bind_dev(m); | |
1014 | ||
1015 | case SYSFS: | |
1016 | return mount_sysfs(m); | |
1017 | ||
1018 | case PROCFS: | |
1019 | return mount_procfs(m); | |
d6797c92 | 1020 | |
e364ad06 LP |
1021 | default: |
1022 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
1023 | } |
1024 | ||
ac0930c8 | 1025 | assert(what); |
15ae422b | 1026 | |
a227a4be LP |
1027 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) { |
1028 | bool try_again = false; | |
1029 | r = -errno; | |
1030 | ||
1031 | if (r == -ENOENT && make) { | |
1032 | struct stat st; | |
1033 | ||
1034 | /* Hmm, either the source or the destination are missing. Let's see if we can create the destination, then try again */ | |
1035 | ||
e8717862 LP |
1036 | if (stat(what, &st) < 0) |
1037 | log_debug_errno(errno, "Mount point source '%s' is not accessible: %m", what); | |
1038 | else { | |
1039 | int q; | |
a227a4be LP |
1040 | |
1041 | (void) mkdir_parents(mount_entry_path(m), 0755); | |
1042 | ||
1043 | if (S_ISDIR(st.st_mode)) | |
e8717862 | 1044 | q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0; |
a227a4be | 1045 | else |
e8717862 LP |
1046 | q = touch(mount_entry_path(m)); |
1047 | ||
1048 | if (q < 0) | |
1049 | log_debug_errno(q, "Failed to create destination mount point node '%s': %m", mount_entry_path(m)); | |
1050 | else | |
1051 | try_again = true; | |
a227a4be LP |
1052 | } |
1053 | } | |
1054 | ||
1055 | if (try_again) { | |
1056 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) | |
1057 | r = -errno; | |
1058 | else | |
1059 | r = 0; | |
1060 | } | |
1061 | ||
1062 | if (r < 0) | |
1063 | return log_debug_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); | |
1064 | } | |
6b7c9f8b | 1065 | |
34de407a | 1066 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1067 | return 0; |
ac0930c8 | 1068 | } |
15ae422b | 1069 | |
2a2969fd | 1070 | static int make_read_only(const MountEntry *m, char **blacklist, FILE *proc_self_mountinfo) { |
9ce4e4b0 | 1071 | unsigned long new_flags = 0, flags_mask = 0; |
763a260a | 1072 | bool submounts = false; |
6b7c9f8b | 1073 | int r = 0; |
15ae422b | 1074 | |
c17ec25e | 1075 | assert(m); |
ac9de0b3 | 1076 | assert(proc_self_mountinfo); |
ac0930c8 | 1077 | |
9ce4e4b0 LP |
1078 | if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) { |
1079 | new_flags |= MS_RDONLY; | |
1080 | flags_mask |= MS_RDONLY; | |
1081 | } | |
1082 | ||
1083 | if (m->nosuid) { | |
1084 | new_flags |= MS_NOSUID; | |
1085 | flags_mask |= MS_NOSUID; | |
1086 | } | |
1087 | ||
1088 | if (flags_mask == 0) /* No Change? */ | |
6b7c9f8b LP |
1089 | return 0; |
1090 | ||
9ce4e4b0 LP |
1091 | /* We generally apply these changes recursively, except for /dev, and the cases we know there's |
1092 | * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the | |
1093 | * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace | |
1094 | * and running Linux <= 4.17. */ | |
1095 | submounts = | |
1096 | mount_entry_read_only(m) && | |
1097 | !IN_SET(m->mode, EMPTY_DIR, TMPFS); | |
1098 | if (submounts) | |
1099 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, blacklist, proc_self_mountinfo); | |
1100 | else | |
7cce68e1 | 1101 | r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo); |
9ce4e4b0 | 1102 | |
867189b5 LP |
1103 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked |
1104 | * read-only already stays this way. This improves compatibility with container managers, where we | |
1105 | * won't attempt to undo read-only mounts already applied. */ | |
ac0930c8 | 1106 | |
8fceda93 | 1107 | if (r == -ENOENT && m->ignore) |
867189b5 | 1108 | return 0; |
763a260a | 1109 | if (r < 0) |
9ce4e4b0 | 1110 | return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m), |
763a260a | 1111 | submounts ? " and its submounts" : ""); |
763a260a | 1112 | return 0; |
d944dc95 LP |
1113 | } |
1114 | ||
9b68367b | 1115 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1116 | assert(ns_info); |
1117 | ||
9c988f93 DH |
1118 | /* |
1119 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1120 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1121 | * first place... |
9c988f93 | 1122 | */ |
5d997827 | 1123 | |
9b68367b YW |
1124 | return ns_info->mount_apivfs || |
1125 | ns_info->protect_control_groups || | |
1126 | ns_info->protect_kernel_tunables; | |
5d997827 LP |
1127 | } |
1128 | ||
da6053d0 | 1129 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1130 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1131 | char** read_write_paths, |
1132 | char** read_only_paths, | |
1133 | char** inaccessible_paths, | |
6c47cd7d | 1134 | char** empty_directories, |
da6053d0 LP |
1135 | size_t n_bind_mounts, |
1136 | size_t n_temporary_filesystems, | |
2652c6c1 DH |
1137 | const char* tmp_dir, |
1138 | const char* var_tmp_dir, | |
91dd5f7c | 1139 | const char* log_namespace, |
2652c6c1 DH |
1140 | ProtectHome protect_home, |
1141 | ProtectSystem protect_system) { | |
1142 | ||
da6053d0 LP |
1143 | size_t protect_home_cnt; |
1144 | size_t protect_system_cnt = | |
f471b2af DH |
1145 | (protect_system == PROTECT_SYSTEM_STRICT ? |
1146 | ELEMENTSOF(protect_system_strict_table) : | |
1147 | ((protect_system == PROTECT_SYSTEM_FULL) ? | |
1148 | ELEMENTSOF(protect_system_full_table) : | |
1149 | ((protect_system == PROTECT_SYSTEM_YES) ? | |
1150 | ELEMENTSOF(protect_system_yes_table) : 0))); | |
1151 | ||
b6c432ca DH |
1152 | protect_home_cnt = |
1153 | (protect_home == PROTECT_HOME_YES ? | |
1154 | ELEMENTSOF(protect_home_yes_table) : | |
1155 | ((protect_home == PROTECT_HOME_READ_ONLY) ? | |
e4da7d8c YW |
1156 | ELEMENTSOF(protect_home_read_only_table) : |
1157 | ((protect_home == PROTECT_HOME_TMPFS) ? | |
1158 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); | |
b6c432ca | 1159 | |
2652c6c1 DH |
1160 | return !!tmp_dir + !!var_tmp_dir + |
1161 | strv_length(read_write_paths) + | |
1162 | strv_length(read_only_paths) + | |
1163 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1164 | strv_length(empty_directories) + |
d2d6c096 | 1165 | n_bind_mounts + |
2abd4e38 | 1166 | n_temporary_filesystems + |
c575770b DH |
1167 | ns_info->private_dev + |
1168 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
c575770b | 1169 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + |
94a7b275 KK |
1170 | (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) + |
1171 | (ns_info->protect_control_groups ? 1 : 0) + | |
5d997827 | 1172 | protect_home_cnt + protect_system_cnt + |
aecd5ac6 | 1173 | (ns_info->protect_hostname ? 2 : 0) + |
91dd5f7c LP |
1174 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) + |
1175 | !!log_namespace; | |
2652c6c1 DH |
1176 | } |
1177 | ||
da6053d0 | 1178 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1179 | assert(root_directory); |
f8b64b57 LP |
1180 | assert(n_mounts); |
1181 | assert(mounts || *n_mounts == 0); | |
1182 | ||
93bab288 | 1183 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1184 | |
1185 | drop_duplicates(mounts, n_mounts); | |
1186 | drop_outside_root(root_directory, mounts, n_mounts); | |
1187 | drop_inaccessible(mounts, n_mounts); | |
1188 | drop_nop(mounts, n_mounts); | |
1189 | } | |
1190 | ||
c8c535d5 LP |
1191 | static bool root_read_only( |
1192 | char **read_only_paths, | |
1193 | ProtectSystem protect_system) { | |
1194 | ||
1195 | /* Determine whether the root directory is going to be read-only given the configured settings. */ | |
1196 | ||
1197 | if (protect_system == PROTECT_SYSTEM_STRICT) | |
1198 | return true; | |
1199 | ||
de46b2be | 1200 | if (prefixed_path_strv_contains(read_only_paths, "/")) |
c8c535d5 LP |
1201 | return true; |
1202 | ||
1203 | return false; | |
1204 | } | |
1205 | ||
1206 | static bool home_read_only( | |
1207 | char** read_only_paths, | |
1208 | char** inaccessible_paths, | |
1209 | char** empty_directories, | |
1210 | const BindMount *bind_mounts, | |
1211 | size_t n_bind_mounts, | |
1212 | const TemporaryFileSystem *temporary_filesystems, | |
1213 | size_t n_temporary_filesystems, | |
1214 | ProtectHome protect_home) { | |
1215 | ||
1216 | size_t i; | |
1217 | ||
1218 | /* Determine whether the /home directory is going to be read-only given the configured settings. Yes, | |
1219 | * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple | |
1220 | * settings. */ | |
1221 | ||
1222 | if (protect_home != PROTECT_HOME_NO) | |
1223 | return true; | |
1224 | ||
de46b2be TM |
1225 | if (prefixed_path_strv_contains(read_only_paths, "/home") || |
1226 | prefixed_path_strv_contains(inaccessible_paths, "/home") || | |
1227 | prefixed_path_strv_contains(empty_directories, "/home")) | |
c8c535d5 LP |
1228 | return true; |
1229 | ||
1230 | for (i = 0; i < n_temporary_filesystems; i++) | |
1231 | if (path_equal(temporary_filesystems[i].path, "/home")) | |
1232 | return true; | |
1233 | ||
1234 | /* If /home is overmounted with some dir from the host it's not writable. */ | |
1235 | for (i = 0; i < n_bind_mounts; i++) | |
1236 | if (path_equal(bind_mounts[i].destination, "/home")) | |
1237 | return true; | |
1238 | ||
1239 | return false; | |
1240 | } | |
1241 | ||
613b411c | 1242 | int setup_namespace( |
ee818b89 | 1243 | const char* root_directory, |
915e6d16 | 1244 | const char* root_image, |
bb0ff3fb | 1245 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1246 | char** read_write_paths, |
1247 | char** read_only_paths, | |
1248 | char** inaccessible_paths, | |
6c47cd7d | 1249 | char** empty_directories, |
d2d6c096 | 1250 | const BindMount *bind_mounts, |
da6053d0 | 1251 | size_t n_bind_mounts, |
2abd4e38 | 1252 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1253 | size_t n_temporary_filesystems, |
a004cb4c LP |
1254 | const char* tmp_dir, |
1255 | const char* var_tmp_dir, | |
91dd5f7c | 1256 | const char *log_namespace, |
1b8689f9 LP |
1257 | ProtectHome protect_home, |
1258 | ProtectSystem protect_system, | |
915e6d16 | 1259 | unsigned long mount_flags, |
7cc5ef5f ZJS |
1260 | DissectImageFlags dissect_image_flags, |
1261 | char **error_path) { | |
15ae422b | 1262 | |
915e6d16 | 1263 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1264 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1265 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
78ebe980 | 1266 | _cleanup_free_ void *root_hash = NULL; |
5f7a690a | 1267 | MountEntry *m = NULL, *mounts = NULL; |
9b68367b | 1268 | size_t n_mounts, root_hash_size = 0; |
d18aff04 | 1269 | bool require_prefix = false; |
9b68367b | 1270 | const char *root; |
c17ec25e | 1271 | int r = 0; |
15ae422b | 1272 | |
915e6d16 LP |
1273 | assert(ns_info); |
1274 | ||
613b411c | 1275 | if (mount_flags == 0) |
c17ec25e | 1276 | mount_flags = MS_SHARED; |
ac0930c8 | 1277 | |
915e6d16 LP |
1278 | if (root_image) { |
1279 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1280 | ||
c8c535d5 LP |
1281 | /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */ |
1282 | if (root_read_only(read_only_paths, | |
1283 | protect_system) && | |
1284 | home_read_only(read_only_paths, inaccessible_paths, empty_directories, | |
1285 | bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems, | |
1286 | protect_home) && | |
c9ef8573 | 1287 | strv_isempty(read_write_paths)) |
915e6d16 LP |
1288 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1289 | ||
1290 | r = loop_device_make_by_path(root_image, | |
b0a94268 | 1291 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, |
e08f94ac | 1292 | LO_FLAGS_PARTSCAN, |
915e6d16 LP |
1293 | &loop_device); |
1294 | if (r < 0) | |
763a260a | 1295 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
915e6d16 | 1296 | |
78ebe980 LP |
1297 | r = root_hash_load(root_image, &root_hash, &root_hash_size); |
1298 | if (r < 0) | |
763a260a | 1299 | return log_debug_errno(r, "Failed to load root hash: %m"); |
78ebe980 LP |
1300 | |
1301 | r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image); | |
1302 | if (r < 0) | |
763a260a | 1303 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 LP |
1304 | |
1305 | r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image); | |
915e6d16 | 1306 | if (r < 0) |
763a260a | 1307 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1308 | } |
1309 | ||
e908468b LP |
1310 | if (root_directory) |
1311 | root = root_directory; | |
0722b359 JS |
1312 | else { |
1313 | /* Always create the mount namespace in a temporary directory, instead of operating | |
1314 | * directly in the root. The temporary directory prevents any mounts from being | |
1315 | * potentially obscured my other mounts we already applied. | |
1316 | * We use the same mount point for all images, which is safe, since they all live | |
1317 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1318 | |
1319 | root = "/run/systemd/unit-root"; | |
1320 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1321 | require_prefix = true; |
0722b359 | 1322 | } |
e908468b | 1323 | |
cfbeb4ef LP |
1324 | n_mounts = namespace_calculate_mounts( |
1325 | ns_info, | |
1326 | read_write_paths, | |
1327 | read_only_paths, | |
1328 | inaccessible_paths, | |
6c47cd7d | 1329 | empty_directories, |
f5c52a77 | 1330 | n_bind_mounts, |
2abd4e38 | 1331 | n_temporary_filesystems, |
cfbeb4ef | 1332 | tmp_dir, var_tmp_dir, |
91dd5f7c | 1333 | log_namespace, |
cfbeb4ef | 1334 | protect_home, protect_system); |
613b411c | 1335 | |
f0a4feb0 | 1336 | if (n_mounts > 0) { |
5f7a690a LP |
1337 | m = mounts = new0(MountEntry, n_mounts); |
1338 | if (!mounts) | |
1339 | return -ENOMEM; | |
1340 | ||
d18aff04 | 1341 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1342 | if (r < 0) |
f0a4feb0 | 1343 | goto finish; |
613b411c | 1344 | |
d18aff04 | 1345 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1346 | if (r < 0) |
f0a4feb0 | 1347 | goto finish; |
613b411c | 1348 | |
d18aff04 | 1349 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1350 | if (r < 0) |
f0a4feb0 | 1351 | goto finish; |
7ff7394d | 1352 | |
6c47cd7d LP |
1353 | r = append_empty_dir_mounts(&m, empty_directories); |
1354 | if (r < 0) | |
1355 | goto finish; | |
1356 | ||
d2d6c096 LP |
1357 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1358 | if (r < 0) | |
1359 | goto finish; | |
1360 | ||
2abd4e38 YW |
1361 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1362 | if (r < 0) | |
1363 | goto finish; | |
1364 | ||
613b411c | 1365 | if (tmp_dir) { |
34de407a | 1366 | *(m++) = (MountEntry) { |
5327c910 LP |
1367 | .path_const = "/tmp", |
1368 | .mode = PRIVATE_TMP, | |
89bd586c | 1369 | .source_const = tmp_dir, |
5327c910 | 1370 | }; |
613b411c | 1371 | } |
7ff7394d | 1372 | |
613b411c | 1373 | if (var_tmp_dir) { |
34de407a | 1374 | *(m++) = (MountEntry) { |
5327c910 | 1375 | .path_const = "/var/tmp", |
89bd586c YW |
1376 | .mode = PRIVATE_TMP, |
1377 | .source_const = var_tmp_dir, | |
5327c910 | 1378 | }; |
7ff7394d | 1379 | } |
ac0930c8 | 1380 | |
c575770b | 1381 | if (ns_info->private_dev) { |
34de407a | 1382 | *(m++) = (MountEntry) { |
5327c910 LP |
1383 | .path_const = "/dev", |
1384 | .mode = PRIVATE_DEV, | |
9ce4e4b0 | 1385 | .flags = DEV_MOUNT_OPTIONS, |
5327c910 | 1386 | }; |
7f112f50 LP |
1387 | } |
1388 | ||
c575770b | 1389 | if (ns_info->protect_kernel_tunables) { |
5327c910 | 1390 | r = append_static_mounts(&m, protect_kernel_tunables_table, ELEMENTSOF(protect_kernel_tunables_table), ns_info->ignore_protect_paths); |
c575770b | 1391 | if (r < 0) |
f0a4feb0 | 1392 | goto finish; |
c575770b DH |
1393 | } |
1394 | ||
1395 | if (ns_info->protect_kernel_modules) { | |
5327c910 | 1396 | r = append_static_mounts(&m, protect_kernel_modules_table, ELEMENTSOF(protect_kernel_modules_table), ns_info->ignore_protect_paths); |
c575770b | 1397 | if (r < 0) |
f0a4feb0 | 1398 | goto finish; |
c575770b | 1399 | } |
59eeb84b | 1400 | |
94a7b275 KK |
1401 | if (ns_info->protect_kernel_logs) { |
1402 | r = append_static_mounts(&m, protect_kernel_logs_table, ELEMENTSOF(protect_kernel_logs_table), ns_info->ignore_protect_paths); | |
1403 | if (r < 0) | |
1404 | goto finish; | |
1405 | } | |
1406 | ||
c575770b | 1407 | if (ns_info->protect_control_groups) { |
34de407a | 1408 | *(m++) = (MountEntry) { |
5327c910 LP |
1409 | .path_const = "/sys/fs/cgroup", |
1410 | .mode = READONLY, | |
1411 | }; | |
59eeb84b LP |
1412 | } |
1413 | ||
5327c910 | 1414 | r = append_protect_home(&m, protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1415 | if (r < 0) |
f0a4feb0 | 1416 | goto finish; |
417116f2 | 1417 | |
5327c910 | 1418 | r = append_protect_system(&m, protect_system, false); |
f471b2af | 1419 | if (r < 0) |
f0a4feb0 | 1420 | goto finish; |
417116f2 | 1421 | |
9b68367b | 1422 | if (namespace_info_mount_apivfs(ns_info)) { |
5d997827 LP |
1423 | r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths); |
1424 | if (r < 0) | |
1425 | goto finish; | |
1426 | } | |
1427 | ||
aecd5ac6 TM |
1428 | if (ns_info->protect_hostname) { |
1429 | *(m++) = (MountEntry) { | |
1430 | .path_const = "/proc/sys/kernel/hostname", | |
1431 | .mode = READONLY, | |
1432 | }; | |
1433 | *(m++) = (MountEntry) { | |
1434 | .path_const = "/proc/sys/kernel/domainname", | |
1435 | .mode = READONLY, | |
1436 | }; | |
1437 | } | |
1438 | ||
91dd5f7c LP |
1439 | if (log_namespace) { |
1440 | _cleanup_free_ char *q; | |
1441 | ||
1442 | q = strjoin("/run/systemd/journal.", log_namespace); | |
1443 | if (!q) { | |
1444 | r = -ENOMEM; | |
1445 | goto finish; | |
1446 | } | |
1447 | ||
1448 | *(m++) = (MountEntry) { | |
1449 | .path_const = "/run/systemd/journal", | |
1450 | .mode = BIND_MOUNT_RECURSIVE, | |
1451 | .read_only = true, | |
1452 | .source_malloc = TAKE_PTR(q), | |
1453 | }; | |
1454 | } | |
1455 | ||
f0a4feb0 | 1456 | assert(mounts + n_mounts == m); |
ac0930c8 | 1457 | |
5327c910 | 1458 | /* Prepend the root directory where that's necessary */ |
e908468b | 1459 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1460 | if (r < 0) |
1461 | goto finish; | |
1462 | ||
839f1877 | 1463 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1464 | } |
1465 | ||
1beab8b0 LP |
1466 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1467 | ||
d944dc95 | 1468 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1469 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 LP |
1470 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
1471 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter in place | |
1472 | * that doesn't allow us to create namespaces (or a missing cap), then propagate a recognizable | |
1473 | * error back, which the caller can use to detect this case (and only this) and optionally | |
1474 | * continue without namespacing applied. */ | |
1475 | r = -ENOANO; | |
1476 | ||
d944dc95 LP |
1477 | goto finish; |
1478 | } | |
1e4e94c8 | 1479 | |
9b68367b YW |
1480 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1481 | * shows up in the parent */ | |
1482 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1483 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1484 | goto finish; |
ee818b89 AC |
1485 | } |
1486 | ||
915e6d16 | 1487 | if (root_image) { |
e908468b | 1488 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1489 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1490 | if (r < 0) { |
1491 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1492 | goto finish; |
763a260a | 1493 | } |
915e6d16 | 1494 | |
07ce7407 TM |
1495 | if (decrypted_image) { |
1496 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1497 | if (r < 0) { |
1498 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1499 | goto finish; |
763a260a | 1500 | } |
07ce7407 | 1501 | } |
78ebe980 | 1502 | |
915e6d16 LP |
1503 | loop_device_relinquish(loop_device); |
1504 | ||
1505 | } else if (root_directory) { | |
1506 | ||
e908468b LP |
1507 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1508 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1509 | if (r < 0) { |
1510 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1511 | goto finish; |
763a260a | 1512 | } |
8f1ad200 | 1513 | if (r == 0) { |
e908468b | 1514 | if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) { |
763a260a | 1515 | r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root); |
8f1ad200 LP |
1516 | goto finish; |
1517 | } | |
d944dc95 | 1518 | } |
e908468b | 1519 | |
9b68367b | 1520 | } else { |
e908468b LP |
1521 | |
1522 | /* Let's mount the main root directory to the root directory to use */ | |
1523 | if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
763a260a | 1524 | r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root); |
e908468b LP |
1525 | goto finish; |
1526 | } | |
ee818b89 | 1527 | } |
c2c13f2d | 1528 | |
4e0c20de LP |
1529 | /* Try to set up the new root directory before mounting anything else there. */ |
1530 | if (root_image || root_directory) | |
1531 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1532 | ||
f0a4feb0 | 1533 | if (n_mounts > 0) { |
ac9de0b3 | 1534 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
5f7a690a | 1535 | _cleanup_free_ char **blacklist = NULL; |
da6053d0 | 1536 | size_t j; |
6b7c9f8b | 1537 | |
ac9de0b3 TR |
1538 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of /proc. |
1539 | * For example, this is the case with the option: 'InaccessiblePaths=/proc' */ | |
1540 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); | |
1541 | if (!proc_self_mountinfo) { | |
763a260a | 1542 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
7cc5ef5f ZJS |
1543 | if (error_path) |
1544 | *error_path = strdup("/proc/self/mountinfo"); | |
ac9de0b3 TR |
1545 | goto finish; |
1546 | } | |
1547 | ||
088696fe LP |
1548 | /* First round, establish all mounts we need */ |
1549 | for (;;) { | |
1550 | bool again = false; | |
1551 | ||
1552 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1553 | ||
1554 | if (m->applied) | |
1555 | continue; | |
1556 | ||
1557 | r = follow_symlink(root, m); | |
7cc5ef5f ZJS |
1558 | if (r < 0) { |
1559 | if (error_path && mount_entry_path(m)) | |
1560 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1561 | goto finish; |
7cc5ef5f | 1562 | } |
088696fe LP |
1563 | if (r == 0) { |
1564 | /* We hit a symlinked mount point. The entry got rewritten and might point to a | |
1565 | * very different place now. Let's normalize the changed list, and start from | |
1566 | * the beginning. After all to mount the entry at the new location we might | |
1567 | * need some other mounts first */ | |
1568 | again = true; | |
1569 | break; | |
1570 | } | |
1571 | ||
1572 | r = apply_mount(root, m); | |
7cc5ef5f ZJS |
1573 | if (r < 0) { |
1574 | if (error_path && mount_entry_path(m)) | |
1575 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1576 | goto finish; |
7cc5ef5f | 1577 | } |
088696fe LP |
1578 | |
1579 | m->applied = true; | |
1580 | } | |
1581 | ||
1582 | if (!again) | |
1583 | break; | |
1584 | ||
839f1877 | 1585 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1586 | } |
15ae422b | 1587 | |
6b7c9f8b | 1588 | /* Create a blacklist we can pass to bind_mount_recursive() */ |
5f7a690a LP |
1589 | blacklist = new(char*, n_mounts+1); |
1590 | if (!blacklist) { | |
1591 | r = -ENOMEM; | |
1592 | goto finish; | |
1593 | } | |
f0a4feb0 | 1594 | for (j = 0; j < n_mounts; j++) |
34de407a | 1595 | blacklist[j] = (char*) mount_entry_path(mounts+j); |
6b7c9f8b LP |
1596 | blacklist[j] = NULL; |
1597 | ||
1598 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1599 | for (m = mounts; m < mounts + n_mounts; ++m) { |
ac9de0b3 | 1600 | r = make_read_only(m, blacklist, proc_self_mountinfo); |
7cc5ef5f ZJS |
1601 | if (r < 0) { |
1602 | if (error_path && mount_entry_path(m)) | |
1603 | *error_path = strdup(mount_entry_path(m)); | |
d944dc95 | 1604 | goto finish; |
7cc5ef5f | 1605 | } |
c2c13f2d | 1606 | } |
15ae422b LP |
1607 | } |
1608 | ||
9b68367b YW |
1609 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1610 | r = mount_move_root(root); | |
763a260a YW |
1611 | if (r < 0) { |
1612 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1613 | goto finish; |
763a260a | 1614 | } |
ee818b89 | 1615 | |
55fe7432 | 1616 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1617 | * reestablish propagation from our side to the host, since |
1618 | * what's disconnected is disconnected. */ | |
d944dc95 | 1619 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1620 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1621 | goto finish; |
1622 | } | |
15ae422b | 1623 | |
d944dc95 | 1624 | r = 0; |
15ae422b | 1625 | |
d944dc95 | 1626 | finish: |
f0a4feb0 | 1627 | for (m = mounts; m < mounts + n_mounts; m++) |
1eb7e08e | 1628 | mount_entry_done(m); |
613b411c | 1629 | |
5f7a690a LP |
1630 | free(mounts); |
1631 | ||
613b411c LP |
1632 | return r; |
1633 | } | |
1634 | ||
da6053d0 LP |
1635 | void bind_mount_free_many(BindMount *b, size_t n) { |
1636 | size_t i; | |
d2d6c096 LP |
1637 | |
1638 | assert(b || n == 0); | |
1639 | ||
1640 | for (i = 0; i < n; i++) { | |
1641 | free(b[i].source); | |
1642 | free(b[i].destination); | |
1643 | } | |
1644 | ||
1645 | free(b); | |
1646 | } | |
1647 | ||
da6053d0 | 1648 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
1649 | _cleanup_free_ char *s = NULL, *d = NULL; |
1650 | BindMount *c; | |
1651 | ||
1652 | assert(b); | |
1653 | assert(n); | |
1654 | assert(item); | |
1655 | ||
1656 | s = strdup(item->source); | |
1657 | if (!s) | |
1658 | return -ENOMEM; | |
1659 | ||
1660 | d = strdup(item->destination); | |
1661 | if (!d) | |
1662 | return -ENOMEM; | |
1663 | ||
aa484f35 | 1664 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
1665 | if (!c) |
1666 | return -ENOMEM; | |
1667 | ||
1668 | *b = c; | |
1669 | ||
1670 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
1671 | .source = TAKE_PTR(s), |
1672 | .destination = TAKE_PTR(d), | |
d2d6c096 | 1673 | .read_only = item->read_only, |
9ce4e4b0 | 1674 | .nosuid = item->nosuid, |
d2d6c096 LP |
1675 | .recursive = item->recursive, |
1676 | .ignore_enoent = item->ignore_enoent, | |
1677 | }; | |
1678 | ||
d2d6c096 LP |
1679 | return 0; |
1680 | } | |
1681 | ||
da6053d0 LP |
1682 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
1683 | size_t i; | |
2abd4e38 YW |
1684 | |
1685 | assert(t || n == 0); | |
1686 | ||
1687 | for (i = 0; i < n; i++) { | |
1688 | free(t[i].path); | |
1689 | free(t[i].options); | |
1690 | } | |
1691 | ||
1692 | free(t); | |
1693 | } | |
1694 | ||
1695 | int temporary_filesystem_add( | |
1696 | TemporaryFileSystem **t, | |
da6053d0 | 1697 | size_t *n, |
2abd4e38 YW |
1698 | const char *path, |
1699 | const char *options) { | |
1700 | ||
1701 | _cleanup_free_ char *p = NULL, *o = NULL; | |
1702 | TemporaryFileSystem *c; | |
1703 | ||
1704 | assert(t); | |
1705 | assert(n); | |
1706 | assert(path); | |
1707 | ||
1708 | p = strdup(path); | |
1709 | if (!p) | |
1710 | return -ENOMEM; | |
1711 | ||
1712 | if (!isempty(options)) { | |
1713 | o = strdup(options); | |
1714 | if (!o) | |
1715 | return -ENOMEM; | |
1716 | } | |
1717 | ||
aa484f35 | 1718 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
1719 | if (!c) |
1720 | return -ENOMEM; | |
1721 | ||
1722 | *t = c; | |
1723 | ||
1724 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
1725 | .path = TAKE_PTR(p), |
1726 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
1727 | }; |
1728 | ||
2abd4e38 YW |
1729 | return 0; |
1730 | } | |
1731 | ||
a652f050 JR |
1732 | static int make_tmp_prefix(const char *prefix) { |
1733 | _cleanup_free_ char *t = NULL; | |
1734 | int r; | |
1735 | ||
1736 | /* Don't do anything unless we know the dir is actually missing */ | |
1737 | r = access(prefix, F_OK); | |
1738 | if (r >= 0) | |
1739 | return 0; | |
1740 | if (errno != ENOENT) | |
1741 | return -errno; | |
1742 | ||
1743 | r = mkdir_parents(prefix, 0755); | |
1744 | if (r < 0) | |
1745 | return r; | |
1746 | ||
1747 | r = tempfn_random(prefix, NULL, &t); | |
1748 | if (r < 0) | |
1749 | return r; | |
1750 | ||
1751 | if (mkdir(t, 0777) < 0) | |
1752 | return -errno; | |
1753 | ||
1754 | if (chmod(t, 01777) < 0) { | |
1755 | r = -errno; | |
1756 | (void) rmdir(t); | |
1757 | return r; | |
1758 | } | |
1759 | ||
1760 | if (rename(t, prefix) < 0) { | |
1761 | r = -errno; | |
1762 | (void) rmdir(t); | |
1763 | return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */ | |
1764 | } | |
1765 | ||
1766 | return 0; | |
1767 | ||
1768 | } | |
1769 | ||
613b411c LP |
1770 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path) { |
1771 | _cleanup_free_ char *x = NULL; | |
6b46ea73 LP |
1772 | char bid[SD_ID128_STRING_MAX]; |
1773 | sd_id128_t boot_id; | |
1774 | int r; | |
613b411c LP |
1775 | |
1776 | assert(id); | |
1777 | assert(prefix); | |
1778 | assert(path); | |
1779 | ||
6b46ea73 LP |
1780 | /* We include the boot id in the directory so that after a |
1781 | * reboot we can easily identify obsolete directories. */ | |
1782 | ||
1783 | r = sd_id128_get_boot(&boot_id); | |
1784 | if (r < 0) | |
1785 | return r; | |
1786 | ||
605405c6 | 1787 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
1788 | if (!x) |
1789 | return -ENOMEM; | |
1790 | ||
a652f050 JR |
1791 | r = make_tmp_prefix(prefix); |
1792 | if (r < 0) | |
1793 | return r; | |
1794 | ||
613b411c LP |
1795 | RUN_WITH_UMASK(0077) |
1796 | if (!mkdtemp(x)) | |
1797 | return -errno; | |
1798 | ||
1799 | RUN_WITH_UMASK(0000) { | |
1800 | char *y; | |
1801 | ||
63c372cb | 1802 | y = strjoina(x, "/tmp"); |
613b411c LP |
1803 | |
1804 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
1805 | return -errno; | |
c17ec25e | 1806 | } |
15ae422b | 1807 | |
1cc6c93a | 1808 | *path = TAKE_PTR(x); |
613b411c LP |
1809 | |
1810 | return 0; | |
1811 | } | |
1812 | ||
1813 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
1814 | char *a, *b; | |
1815 | int r; | |
1816 | ||
1817 | assert(id); | |
1818 | assert(tmp_dir); | |
1819 | assert(var_tmp_dir); | |
1820 | ||
1821 | r = setup_one_tmp_dir(id, "/tmp", &a); | |
1822 | if (r < 0) | |
1823 | return r; | |
1824 | ||
1825 | r = setup_one_tmp_dir(id, "/var/tmp", &b); | |
1826 | if (r < 0) { | |
1827 | char *t; | |
1828 | ||
63c372cb | 1829 | t = strjoina(a, "/tmp"); |
6990fb6b LP |
1830 | (void) rmdir(t); |
1831 | (void) rmdir(a); | |
613b411c LP |
1832 | |
1833 | free(a); | |
1834 | return r; | |
1835 | } | |
1836 | ||
1837 | *tmp_dir = a; | |
1838 | *var_tmp_dir = b; | |
1839 | ||
1840 | return 0; | |
1841 | } | |
1842 | ||
2caa38e9 | 1843 | int setup_netns(const int netns_storage_socket[static 2]) { |
613b411c | 1844 | _cleanup_close_ int netns = -1; |
3ee897d6 | 1845 | int r, q; |
613b411c LP |
1846 | |
1847 | assert(netns_storage_socket); | |
1848 | assert(netns_storage_socket[0] >= 0); | |
1849 | assert(netns_storage_socket[1] >= 0); | |
1850 | ||
1851 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
1852 | * namespace reference fd. Whatever process runs this first |
1853 | * shall create a new namespace, all others should just join | |
1854 | * it. To serialize that we use a file lock on the socket | |
1855 | * pair. | |
613b411c LP |
1856 | * |
1857 | * It's a bit crazy, but hey, works great! */ | |
1858 | ||
1859 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1860 | return -errno; | |
1861 | ||
3ee897d6 LP |
1862 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
1863 | if (netns == -EAGAIN) { | |
44ffcbae | 1864 | /* Nothing stored yet, so let's create a new namespace. */ |
613b411c LP |
1865 | |
1866 | if (unshare(CLONE_NEWNET) < 0) { | |
1867 | r = -errno; | |
1868 | goto fail; | |
1869 | } | |
1870 | ||
44ffcbae | 1871 | (void) loopback_setup(); |
613b411c LP |
1872 | |
1873 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1874 | if (netns < 0) { | |
1875 | r = -errno; | |
1876 | goto fail; | |
1877 | } | |
1878 | ||
1879 | r = 1; | |
613b411c | 1880 | |
3ee897d6 LP |
1881 | } else if (netns < 0) { |
1882 | r = netns; | |
1883 | goto fail; | |
613b411c | 1884 | |
3ee897d6 LP |
1885 | } else { |
1886 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
1887 | if (setns(netns, CLONE_NEWNET) < 0) { |
1888 | r = -errno; | |
1889 | goto fail; | |
1890 | } | |
1891 | ||
1892 | r = 0; | |
1893 | } | |
1894 | ||
3ee897d6 LP |
1895 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
1896 | if (q < 0) { | |
1897 | r = q; | |
613b411c LP |
1898 | goto fail; |
1899 | } | |
1900 | ||
1901 | fail: | |
fe048ce5 | 1902 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
1903 | return r; |
1904 | } | |
417116f2 | 1905 | |
2caa38e9 | 1906 | int open_netns_path(const int netns_storage_socket[static 2], const char *path) { |
51af7fb2 LP |
1907 | _cleanup_close_ int netns = -1; |
1908 | int q, r; | |
1909 | ||
1910 | assert(netns_storage_socket); | |
1911 | assert(netns_storage_socket[0] >= 0); | |
1912 | assert(netns_storage_socket[1] >= 0); | |
1913 | assert(path); | |
1914 | ||
1915 | /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in | |
1916 | * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a | |
1917 | * new anonymous netns if needed. */ | |
1918 | ||
1919 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
1920 | return -errno; | |
1921 | ||
1922 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); | |
1923 | if (netns == -EAGAIN) { | |
1924 | /* Nothing stored yet. Open the file from the file system. */ | |
1925 | ||
1926 | netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
1927 | if (netns < 0) { | |
1928 | r = -errno; | |
1929 | goto fail; | |
1930 | } | |
1931 | ||
1932 | r = fd_is_network_ns(netns); | |
1933 | if (r == 0) { /* Not a netns? Refuse early. */ | |
1934 | r = -EINVAL; | |
1935 | goto fail; | |
1936 | } | |
1937 | if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */ | |
1938 | goto fail; | |
1939 | ||
1940 | r = 1; | |
1941 | ||
1942 | } else if (netns < 0) { | |
1943 | r = netns; | |
1944 | goto fail; | |
1945 | } else | |
1946 | r = 0; /* Already allocated */ | |
1947 | ||
1948 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); | |
1949 | if (q < 0) { | |
1950 | r = q; | |
1951 | goto fail; | |
1952 | } | |
1953 | ||
1954 | fail: | |
1955 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); | |
1956 | return r; | |
1957 | } | |
1958 | ||
6e2d7c4f MS |
1959 | bool ns_type_supported(NamespaceType type) { |
1960 | const char *t, *ns_proc; | |
1961 | ||
0fa5b831 LP |
1962 | t = namespace_type_to_string(type); |
1963 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
1964 | return false; |
1965 | ||
6e2d7c4f | 1966 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
1967 | return access(ns_proc, F_OK) == 0; |
1968 | } | |
1969 | ||
1b8689f9 LP |
1970 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
1971 | [PROTECT_HOME_NO] = "no", | |
1972 | [PROTECT_HOME_YES] = "yes", | |
1973 | [PROTECT_HOME_READ_ONLY] = "read-only", | |
e4da7d8c | 1974 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
1975 | }; |
1976 | ||
1e8c7bd5 | 1977 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 1978 | |
1b8689f9 LP |
1979 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
1980 | [PROTECT_SYSTEM_NO] = "no", | |
1981 | [PROTECT_SYSTEM_YES] = "yes", | |
1982 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 1983 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
1984 | }; |
1985 | ||
1e8c7bd5 | 1986 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 1987 | |
6e2d7c4f MS |
1988 | static const char* const namespace_type_table[] = { |
1989 | [NAMESPACE_MOUNT] = "mnt", | |
1990 | [NAMESPACE_CGROUP] = "cgroup", | |
1991 | [NAMESPACE_UTS] = "uts", | |
1992 | [NAMESPACE_IPC] = "ipc", | |
1993 | [NAMESPACE_USER] = "user", | |
1994 | [NAMESPACE_PID] = "pid", | |
1995 | [NAMESPACE_NET] = "net", | |
1996 | }; | |
1997 | ||
1998 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); |