]>
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" |
b3d13314 | 18 | #include "list.h" |
915e6d16 | 19 | #include "loop-util.h" |
07630cea | 20 | #include "loopback-setup.h" |
07630cea | 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" |
d8b4d14d | 26 | #include "nulstr-util.h" |
07630cea | 27 | #include "path-util.h" |
d7b8eec7 | 28 | #include "selinux-util.h" |
2583fbea | 29 | #include "socket-util.h" |
760877e9 | 30 | #include "sort-util.h" |
36ce7110 | 31 | #include "stat-util.h" |
8b43440b | 32 | #include "string-table.h" |
07630cea LP |
33 | #include "string-util.h" |
34 | #include "strv.h" | |
a652f050 | 35 | #include "tmpfile-util.h" |
affb60b1 | 36 | #include "umask-util.h" |
ee104e11 | 37 | #include "user-util.h" |
15ae422b | 38 | |
737ba3c8 | 39 | #define DEV_MOUNT_OPTIONS (MS_NOSUID|MS_STRICTATIME|MS_NOEXEC) |
40 | ||
c17ec25e | 41 | typedef enum MountMode { |
15ae422b LP |
42 | /* This is ordered by priority! */ |
43 | INACCESSIBLE, | |
b3d13314 | 44 | MOUNT_IMAGES, |
d2d6c096 LP |
45 | BIND_MOUNT, |
46 | BIND_MOUNT_RECURSIVE, | |
ac0930c8 | 47 | PRIVATE_TMP, |
56a13a49 | 48 | PRIVATE_TMP_READONLY, |
7f112f50 | 49 | PRIVATE_DEV, |
5d997827 | 50 | BIND_DEV, |
6c47cd7d | 51 | EMPTY_DIR, |
5d997827 LP |
52 | SYSFS, |
53 | PROCFS, | |
54 | READONLY, | |
59eeb84b | 55 | READWRITE, |
2abd4e38 | 56 | TMPFS, |
1e05071d | 57 | READWRITE_IMPLICIT, /* Should have the lowest priority. */ |
5beb8688 | 58 | _MOUNT_MODE_MAX, |
c17ec25e | 59 | } MountMode; |
15ae422b | 60 | |
34de407a | 61 | typedef struct MountEntry { |
5327c910 | 62 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 63 | MountMode mode:5; |
5327c910 LP |
64 | bool ignore:1; /* Ignore if path does not exist? */ |
65 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 66 | bool read_only:1; /* Shall this mount point be read-only? */ |
9ce4e4b0 | 67 | bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */ |
088696fe | 68 | bool applied:1; /* Already applied */ |
55fe7432 | 69 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
b3d13314 | 70 | const char *source_const; /* The source path, for bind mounts or images */ |
d2d6c096 | 71 | char *source_malloc; |
2abd4e38 YW |
72 | const char *options_const;/* Mount options for tmpfs */ |
73 | char *options_malloc; | |
74 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 75 | unsigned n_followed; |
427353f6 | 76 | LIST_HEAD(MountOptions, image_options); |
34de407a | 77 | } MountEntry; |
15ae422b | 78 | |
5d997827 | 79 | /* 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 | 80 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
81 | static const MountEntry apivfs_table[] = { |
82 | { "/proc", PROCFS, false }, | |
83 | { "/dev", BIND_DEV, false }, | |
84 | { "/sys", SYSFS, false }, | |
85 | }; | |
f471b2af | 86 | |
11a30cec | 87 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 88 | static const MountEntry protect_kernel_tunables_table[] = { |
1e05071d YW |
89 | { "/proc/acpi", READONLY, true }, |
90 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ | |
91 | { "/proc/asound", READONLY, true }, | |
92 | { "/proc/bus", READONLY, true }, | |
93 | { "/proc/fs", READONLY, true }, | |
94 | { "/proc/irq", READONLY, true }, | |
95 | { "/proc/kallsyms", INACCESSIBLE, true }, | |
96 | { "/proc/kcore", INACCESSIBLE, true }, | |
97 | { "/proc/latency_stats", READONLY, true }, | |
98 | { "/proc/mtrr", READONLY, true }, | |
99 | { "/proc/scsi", READONLY, true }, | |
100 | { "/proc/sys", READONLY, false }, | |
101 | { "/proc/sysrq-trigger", READONLY, true }, | |
102 | { "/proc/timer_stats", READONLY, true }, | |
103 | { "/sys", READONLY, false }, | |
104 | { "/sys/fs/bpf", READONLY, true }, | |
105 | { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */ | |
106 | { "/sys/fs/selinux", READWRITE_IMPLICIT, true }, | |
107 | { "/sys/kernel/debug", READONLY, true }, | |
108 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
109 | }; |
110 | ||
c575770b | 111 | /* ProtectKernelModules= option */ |
34de407a | 112 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 113 | #if HAVE_SPLIT_USR |
c6232fb0 | 114 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 115 | #endif |
c6232fb0 | 116 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
117 | }; |
118 | ||
94a7b275 KK |
119 | /* ProtectKernelLogs= option */ |
120 | static const MountEntry protect_kernel_logs_table[] = { | |
121 | { "/proc/kmsg", INACCESSIBLE, true }, | |
122 | { "/dev/kmsg", INACCESSIBLE, true }, | |
123 | }; | |
124 | ||
b6c432ca DH |
125 | /* |
126 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
127 | * system should be protected by ProtectSystem= | |
128 | */ | |
34de407a | 129 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
130 | { "/home", READONLY, true }, |
131 | { "/run/user", READONLY, true }, | |
132 | { "/root", READONLY, true }, | |
b6c432ca DH |
133 | }; |
134 | ||
e4da7d8c YW |
135 | /* ProtectHome=tmpfs table */ |
136 | static const MountEntry protect_home_tmpfs_table[] = { | |
7d85383e TM |
137 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, |
138 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
139 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
e4da7d8c YW |
140 | }; |
141 | ||
b6c432ca | 142 | /* ProtectHome=yes table */ |
34de407a | 143 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
144 | { "/home", INACCESSIBLE, true }, |
145 | { "/run/user", INACCESSIBLE, true }, | |
146 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
147 | }; |
148 | ||
f471b2af | 149 | /* ProtectSystem=yes table */ |
34de407a | 150 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
151 | { "/usr", READONLY, false }, |
152 | { "/boot", READONLY, true }, | |
153 | { "/efi", READONLY, true }, | |
7486f305 AB |
154 | #if HAVE_SPLIT_USR |
155 | { "/lib", READONLY, true }, | |
156 | { "/lib64", READONLY, true }, | |
157 | { "/bin", READONLY, true }, | |
671f0f8d | 158 | # if HAVE_SPLIT_BIN |
7486f305 | 159 | { "/sbin", READONLY, true }, |
671f0f8d | 160 | # endif |
7486f305 | 161 | #endif |
f471b2af DH |
162 | }; |
163 | ||
164 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 165 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
166 | { "/usr", READONLY, false }, |
167 | { "/boot", READONLY, true }, | |
168 | { "/efi", READONLY, true }, | |
169 | { "/etc", READONLY, false }, | |
7486f305 AB |
170 | #if HAVE_SPLIT_USR |
171 | { "/lib", READONLY, true }, | |
172 | { "/lib64", READONLY, true }, | |
173 | { "/bin", READONLY, true }, | |
671f0f8d | 174 | # if HAVE_SPLIT_BIN |
7486f305 | 175 | { "/sbin", READONLY, true }, |
671f0f8d | 176 | # endif |
7486f305 | 177 | #endif |
f471b2af DH |
178 | }; |
179 | ||
180 | /* | |
181 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
182 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
183 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
184 | * protect those, and these options should be fully orthogonal. | |
185 | * (And of course /home and friends are also left writable, as ProtectHome= | |
186 | * shall manage those, orthogonally). | |
187 | */ | |
34de407a | 188 | static const MountEntry protect_system_strict_table[] = { |
1e05071d YW |
189 | { "/", READONLY, false }, |
190 | { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
191 | { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
192 | { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */ | |
193 | { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
194 | { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
195 | { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
f471b2af DH |
196 | }; |
197 | ||
5beb8688 YW |
198 | static const char * const mount_mode_table[_MOUNT_MODE_MAX] = { |
199 | [INACCESSIBLE] = "inaccessible", | |
200 | [BIND_MOUNT] = "bind", | |
201 | [BIND_MOUNT_RECURSIVE] = "rbind", | |
202 | [PRIVATE_TMP] = "private-tmp", | |
203 | [PRIVATE_DEV] = "private-dev", | |
204 | [BIND_DEV] = "bind-dev", | |
205 | [EMPTY_DIR] = "empty", | |
206 | [SYSFS] = "sysfs", | |
207 | [PROCFS] = "procfs", | |
208 | [READONLY] = "read-only", | |
209 | [READWRITE] = "read-write", | |
210 | [TMPFS] = "tmpfs", | |
b3d13314 | 211 | [MOUNT_IMAGES] = "mount-images", |
5beb8688 YW |
212 | [READWRITE_IMPLICIT] = "rw-implicit", |
213 | }; | |
214 | ||
215 | DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode); | |
216 | ||
34de407a | 217 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
218 | assert(p); |
219 | ||
5327c910 LP |
220 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
221 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 222 | |
5327c910 | 223 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
224 | } |
225 | ||
34de407a | 226 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
227 | assert(p); |
228 | ||
56a13a49 | 229 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE, PRIVATE_TMP_READONLY); |
cfbeb4ef LP |
230 | } |
231 | ||
d2d6c096 LP |
232 | static const char *mount_entry_source(const MountEntry *p) { |
233 | assert(p); | |
234 | ||
235 | return p->source_malloc ?: p->source_const; | |
236 | } | |
237 | ||
2abd4e38 YW |
238 | static const char *mount_entry_options(const MountEntry *p) { |
239 | assert(p); | |
240 | ||
241 | return p->options_malloc ?: p->options_const; | |
242 | } | |
243 | ||
1eb7e08e LP |
244 | static void mount_entry_done(MountEntry *p) { |
245 | assert(p); | |
246 | ||
247 | p->path_malloc = mfree(p->path_malloc); | |
248 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 249 | p->options_malloc = mfree(p->options_malloc); |
427353f6 | 250 | p->image_options = mount_options_free_all(p->image_options); |
1eb7e08e LP |
251 | } |
252 | ||
d18aff04 | 253 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
254 | char **i; |
255 | ||
613b411c LP |
256 | assert(p); |
257 | ||
1e05071d | 258 | /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */ |
5327c910 | 259 | |
15ae422b | 260 | STRV_FOREACH(i, strv) { |
5327c910 LP |
261 | bool ignore = false, needs_prefix = false; |
262 | const char *e = *i; | |
15ae422b | 263 | |
5327c910 LP |
264 | /* Look for any prefixes */ |
265 | if (startswith(e, "-")) { | |
266 | e++; | |
9c94d52e | 267 | ignore = true; |
ea92ae33 | 268 | } |
5327c910 LP |
269 | if (startswith(e, "+")) { |
270 | e++; | |
271 | needs_prefix = true; | |
272 | } | |
ea92ae33 | 273 | |
baaa35ad ZJS |
274 | if (!path_is_absolute(e)) |
275 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
276 | "Path is not absolute: %s", e); | |
15ae422b | 277 | |
34de407a | 278 | *((*p)++) = (MountEntry) { |
5327c910 LP |
279 | .path_const = e, |
280 | .mode = mode, | |
281 | .ignore = ignore, | |
d18aff04 | 282 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 283 | }; |
15ae422b LP |
284 | } |
285 | ||
286 | return 0; | |
287 | } | |
288 | ||
6c47cd7d LP |
289 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
290 | char **i; | |
291 | ||
292 | assert(p); | |
293 | ||
294 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
295 | * "/private/" boundary directories for DynamicUser=1. */ | |
296 | ||
297 | STRV_FOREACH(i, strv) { | |
298 | ||
299 | *((*p)++) = (MountEntry) { | |
300 | .path_const = *i, | |
301 | .mode = EMPTY_DIR, | |
302 | .ignore = false, | |
6c47cd7d | 303 | .read_only = true, |
7d85383e | 304 | .options_const = "mode=755" TMPFS_LIMITS_EMPTY_OR_ALMOST, |
2abd4e38 | 305 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, |
6c47cd7d LP |
306 | }; |
307 | } | |
308 | ||
309 | return 0; | |
310 | } | |
311 | ||
da6053d0 LP |
312 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
313 | size_t i; | |
d2d6c096 LP |
314 | |
315 | assert(p); | |
316 | ||
317 | for (i = 0; i < n; i++) { | |
318 | const BindMount *b = binds + i; | |
319 | ||
320 | *((*p)++) = (MountEntry) { | |
321 | .path_const = b->destination, | |
322 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
323 | .read_only = b->read_only, | |
9ce4e4b0 | 324 | .nosuid = b->nosuid, |
d2d6c096 | 325 | .source_const = b->source, |
4ca763a9 | 326 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
327 | }; |
328 | } | |
329 | ||
330 | return 0; | |
331 | } | |
332 | ||
b3d13314 LB |
333 | static int append_mount_images(MountEntry **p, const MountImage *mount_images, size_t n) { |
334 | assert(p); | |
335 | ||
336 | for (size_t i = 0; i < n; i++) { | |
337 | const MountImage *m = mount_images + i; | |
338 | ||
339 | *((*p)++) = (MountEntry) { | |
340 | .path_const = m->destination, | |
341 | .mode = MOUNT_IMAGES, | |
342 | .source_const = m->source, | |
427353f6 | 343 | .image_options = m->mount_options, |
b3d13314 LB |
344 | .ignore = m->ignore_enoent, |
345 | }; | |
346 | } | |
347 | ||
348 | return 0; | |
349 | } | |
350 | ||
da6053d0 | 351 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
2abd4e38 YW |
352 | assert(p); |
353 | ||
b67ec8e5 | 354 | for (size_t i = 0; i < n; i++) { |
2abd4e38 YW |
355 | const TemporaryFileSystem *t = tmpfs + i; |
356 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 357 | unsigned long flags; |
2abd4e38 | 358 | bool ro = false; |
b67ec8e5 | 359 | int r; |
2abd4e38 | 360 | |
baaa35ad ZJS |
361 | if (!path_is_absolute(t->path)) |
362 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
363 | "Path is not absolute: %s", | |
364 | t->path); | |
2abd4e38 | 365 | |
b67ec8e5 | 366 | str = strjoin("mode=0755" NESTED_TMPFS_LIMITS ",", t->options); |
ad8e66dc AJ |
367 | if (!str) |
368 | return -ENOMEM; | |
2abd4e38 | 369 | |
ad8e66dc AJ |
370 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
371 | if (r < 0) | |
372 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 373 | |
ad8e66dc AJ |
374 | ro = flags & MS_RDONLY; |
375 | if (ro) | |
376 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
377 | |
378 | *((*p)++) = (MountEntry) { | |
379 | .path_const = t->path, | |
380 | .mode = TMPFS, | |
381 | .read_only = ro, | |
ad8e66dc | 382 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
383 | .flags = flags, |
384 | }; | |
2abd4e38 YW |
385 | } |
386 | ||
387 | return 0; | |
388 | } | |
389 | ||
da6053d0 LP |
390 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
391 | size_t i; | |
11a30cec DH |
392 | |
393 | assert(p); | |
f471b2af | 394 | assert(mounts); |
11a30cec | 395 | |
5327c910 | 396 | /* Adds a list of static pre-defined entries */ |
f471b2af | 397 | |
5327c910 | 398 | for (i = 0; i < n; i++) |
34de407a LP |
399 | *((*p)++) = (MountEntry) { |
400 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
401 | .mode = mounts[i].mode, |
402 | .ignore = mounts[i].ignore || ignore_protect, | |
403 | }; | |
f471b2af DH |
404 | |
405 | return 0; | |
406 | } | |
407 | ||
34de407a | 408 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
409 | assert(p); |
410 | ||
5327c910 | 411 | switch (protect_home) { |
b6c432ca | 412 | |
5327c910 | 413 | case PROTECT_HOME_NO: |
b6c432ca DH |
414 | return 0; |
415 | ||
b6c432ca | 416 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
417 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
418 | ||
e4da7d8c YW |
419 | case PROTECT_HOME_TMPFS: |
420 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
421 | ||
b6c432ca | 422 | case PROTECT_HOME_YES: |
5327c910 LP |
423 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
424 | ||
b6c432ca | 425 | default: |
5327c910 | 426 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 427 | } |
b6c432ca DH |
428 | } |
429 | ||
34de407a | 430 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
431 | assert(p); |
432 | ||
5327c910 LP |
433 | switch (protect_system) { |
434 | ||
435 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
436 | return 0; |
437 | ||
f471b2af | 438 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
439 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
440 | ||
f471b2af | 441 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
442 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
443 | ||
f471b2af | 444 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
445 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
446 | ||
f471b2af | 447 | default: |
5327c910 | 448 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 449 | } |
11a30cec DH |
450 | } |
451 | ||
93bab288 | 452 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 453 | int d; |
15ae422b | 454 | |
6ee1a919 | 455 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 456 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
457 | if (d != 0) |
458 | return d; | |
15ae422b | 459 | |
6ee1a919 | 460 | /* If the paths are equal, check the mode */ |
93bab288 | 461 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
462 | } |
463 | ||
da6053d0 LP |
464 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
465 | size_t i; | |
5327c910 | 466 | |
4a756839 | 467 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 LP |
468 | |
469 | for (i = 0; i < n; i++) { | |
470 | char *s; | |
471 | ||
472 | if (m[i].has_prefix) | |
473 | continue; | |
474 | ||
c6134d3e | 475 | s = path_join(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
476 | if (!s) |
477 | return -ENOMEM; | |
478 | ||
e282f51f | 479 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
480 | m[i].has_prefix = true; |
481 | } | |
482 | ||
483 | return 0; | |
484 | } | |
485 | ||
da6053d0 | 486 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 487 | MountEntry *f, *t, *previous; |
15ae422b | 488 | |
c17ec25e | 489 | assert(m); |
15ae422b | 490 | assert(n); |
15ae422b | 491 | |
fe3c2583 LP |
492 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
493 | ||
1d54cd5d | 494 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 495 | |
fe3c2583 | 496 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
497 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
498 | if (previous && | |
499 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
500 | !f->applied && !previous->applied) { | |
5beb8688 | 501 | log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode)); |
34de407a | 502 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ |
1eb7e08e | 503 | mount_entry_done(f); |
15ae422b | 504 | continue; |
fe3c2583 | 505 | } |
15ae422b | 506 | |
e2d7c1a0 | 507 | *t = *f; |
15ae422b | 508 | previous = t; |
fe3c2583 LP |
509 | t++; |
510 | } | |
511 | ||
512 | *n = t - m; | |
513 | } | |
514 | ||
da6053d0 | 515 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 516 | MountEntry *f, *t; |
fe3c2583 LP |
517 | const char *clear = NULL; |
518 | ||
519 | assert(m); | |
520 | assert(n); | |
521 | ||
522 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
523 | * ordered already. */ | |
524 | ||
1d54cd5d | 525 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
526 | |
527 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
528 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
529 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
530 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 531 | mount_entry_done(f); |
fe3c2583 LP |
532 | continue; |
533 | } | |
15ae422b | 534 | |
34de407a | 535 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
536 | |
537 | *t = *f; | |
15ae422b LP |
538 | t++; |
539 | } | |
540 | ||
c17ec25e | 541 | *n = t - m; |
15ae422b LP |
542 | } |
543 | ||
da6053d0 | 544 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 545 | MountEntry *f, *t; |
7648a565 LP |
546 | |
547 | assert(m); | |
548 | assert(n); | |
549 | ||
550 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
551 | * list is ordered by prefixes. */ | |
552 | ||
1d54cd5d | 553 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 | 554 | |
1e05071d YW |
555 | /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */ |
556 | if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) { | |
34de407a | 557 | MountEntry *p; |
7648a565 LP |
558 | bool found = false; |
559 | ||
560 | /* Now let's find the first parent of the entry we are looking at. */ | |
561 | for (p = t-1; p >= m; p--) { | |
34de407a | 562 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
563 | found = true; |
564 | break; | |
565 | } | |
566 | } | |
567 | ||
568 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
569 | if (found && p->mode == f->mode) { | |
5beb8688 YW |
570 | log_debug("%s (%s) is made redundant by %s (%s)", |
571 | mount_entry_path(f), mount_mode_to_string(f->mode), | |
572 | mount_entry_path(p), mount_mode_to_string(p->mode)); | |
1eb7e08e | 573 | mount_entry_done(f); |
7648a565 LP |
574 | continue; |
575 | } | |
576 | } | |
577 | ||
578 | *t = *f; | |
579 | t++; | |
580 | } | |
581 | ||
582 | *n = t - m; | |
583 | } | |
584 | ||
da6053d0 | 585 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 586 | MountEntry *f, *t; |
cd2902c9 LP |
587 | |
588 | assert(m); | |
589 | assert(n); | |
590 | ||
1d54cd5d | 591 | /* Nothing to do */ |
cd2902c9 LP |
592 | if (!root_directory) |
593 | return; | |
594 | ||
595 | /* Drops all mounts that are outside of the root directory. */ | |
596 | ||
1d54cd5d | 597 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 598 | |
34de407a LP |
599 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
600 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 601 | mount_entry_done(f); |
cd2902c9 LP |
602 | continue; |
603 | } | |
604 | ||
605 | *t = *f; | |
606 | t++; | |
607 | } | |
608 | ||
609 | *n = t - m; | |
610 | } | |
611 | ||
b2a60844 LP |
612 | static int clone_device_node( |
613 | const char *d, | |
614 | const char *temporary_mount, | |
615 | bool *make_devnode) { | |
616 | ||
617 | _cleanup_free_ char *sl = NULL; | |
618 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
619 | struct stat st; |
620 | int r; | |
621 | ||
414b304b | 622 | if (stat(d, &st) < 0) { |
b2a60844 LP |
623 | if (errno == ENOENT) { |
624 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 625 | return -ENXIO; |
b2a60844 LP |
626 | } |
627 | ||
628 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
629 | } |
630 | ||
631 | if (!S_ISBLK(st.st_mode) && | |
baaa35ad ZJS |
632 | !S_ISCHR(st.st_mode)) |
633 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
634 | "Device node '%s' to clone is not a device node, ignoring.", | |
635 | d); | |
b5e99f23 | 636 | |
6f7f3a33 | 637 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 638 | |
b2a60844 | 639 | /* First, try to create device node properly */ |
16498617 CB |
640 | if (*make_devnode) { |
641 | mac_selinux_create_file_prepare(d, st.st_mode); | |
642 | r = mknod(dn, st.st_mode, st.st_rdev); | |
643 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
644 | if (r >= 0) |
645 | goto add_symlink; | |
16498617 CB |
646 | if (errno != EPERM) |
647 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
648 | ||
b2a60844 | 649 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
650 | *make_devnode = false; |
651 | } | |
652 | ||
2aed63f4 | 653 | /* We're about to fall back to bind-mounting the device |
1acf344d CG |
654 | * node. So create a dummy bind-mount target. |
655 | * Do not prepare device-node SELinux label (see issue 13762) */ | |
16498617 | 656 | r = mknod(dn, S_IFREG, 0); |
16498617 | 657 | if (r < 0 && errno != EEXIST) |
b2a60844 | 658 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 CB |
659 | |
660 | /* Fallback to bind-mounting: | |
661 | * The assumption here is that all used device nodes carry standard | |
662 | * properties. Specifically, the devices nodes we bind-mount should | |
663 | * either be owned by root:root or root:tty (e.g. /dev/tty, /dev/ptmx) | |
664 | * and should not carry ACLs. */ | |
665 | if (mount(d, dn, NULL, MS_BIND, NULL) < 0) | |
b2a60844 LP |
666 | return log_debug_errno(errno, "Bind mounting failed for '%s': %m", d); |
667 | ||
668 | add_symlink: | |
669 | bn = path_startswith(d, "/dev/"); | |
670 | if (!bn) | |
671 | return 0; | |
672 | ||
673 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
cbc056c8 ZJS |
674 | if (asprintf(&sl, "%s/dev/%s/%u:%u", |
675 | temporary_mount, | |
676 | S_ISCHR(st.st_mode) ? "char" : "block", | |
677 | major(st.st_rdev), minor(st.st_rdev)) < 0) | |
b2a60844 LP |
678 | return log_oom(); |
679 | ||
680 | (void) mkdir_parents(sl, 0755); | |
681 | ||
682 | t = strjoina("../", bn); | |
b2a60844 | 683 | if (symlink(t, sl) < 0) |
2e4a4fae | 684 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 685 | |
af984e13 | 686 | return 0; |
b5e99f23 ДГ |
687 | } |
688 | ||
5d997827 | 689 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
690 | static const char devnodes[] = |
691 | "/dev/null\0" | |
692 | "/dev/zero\0" | |
693 | "/dev/full\0" | |
694 | "/dev/random\0" | |
695 | "/dev/urandom\0" | |
696 | "/dev/tty\0"; | |
697 | ||
2b85f4e1 | 698 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 699 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 700 | bool can_mknod = true; |
7f112f50 LP |
701 | _cleanup_umask_ mode_t u; |
702 | int r; | |
703 | ||
704 | assert(m); | |
705 | ||
706 | u = umask(0000); | |
707 | ||
2b85f4e1 | 708 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 709 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 710 | |
63c372cb | 711 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 712 | (void) mkdir(dev, 0755); |
7d85383e | 713 | if (mount("tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755" TMPFS_LIMITS_DEV) < 0) { |
2e4a4fae | 714 | r = log_debug_errno(errno, "Failed to mount tmpfs on '%s': %m", dev); |
2b85f4e1 LP |
715 | goto fail; |
716 | } | |
c3151977 TM |
717 | r = label_fix_container(dev, "/dev", 0); |
718 | if (r < 0) { | |
719 | log_debug_errno(errno, "Failed to fix label of '%s' as /dev: %m", dev); | |
720 | goto fail; | |
721 | } | |
2b85f4e1 | 722 | |
63c372cb | 723 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 724 | (void) mkdir(devpts, 0755); |
2b85f4e1 | 725 | if (mount("/dev/pts", devpts, NULL, MS_BIND, NULL) < 0) { |
2e4a4fae | 726 | r = log_debug_errno(errno, "Failed to bind mount /dev/pts on '%s': %m", devpts); |
2b85f4e1 LP |
727 | goto fail; |
728 | } | |
729 | ||
2e4a4fae YW |
730 | /* /dev/ptmx can either be a device node or a symlink to /dev/pts/ptmx. |
731 | * When /dev/ptmx a device node, /dev/pts/ptmx has 000 permissions making it inaccessible. | |
732 | * Thus, in that case make a clone. | |
733 | * In nspawn and other containers it will be a symlink, in that case make it a symlink. */ | |
36ce7110 | 734 | r = is_symlink("/dev/ptmx"); |
2e4a4fae YW |
735 | if (r < 0) { |
736 | log_debug_errno(r, "Failed to detect whether /dev/ptmx is a symlink or not: %m"); | |
3164e3cb | 737 | goto fail; |
2e4a4fae | 738 | } else if (r > 0) { |
414b304b ДГ |
739 | devptmx = strjoina(temporary_mount, "/dev/ptmx"); |
740 | if (symlink("pts/ptmx", devptmx) < 0) { | |
2e4a4fae | 741 | r = log_debug_errno(errno, "Failed to create a symlink '%s' to pts/ptmx: %m", devptmx); |
414b304b ДГ |
742 | goto fail; |
743 | } | |
744 | } else { | |
16498617 | 745 | r = clone_device_node("/dev/ptmx", temporary_mount, &can_mknod); |
152c475f LP |
746 | if (r < 0) |
747 | goto fail; | |
414b304b | 748 | } |
e06b6479 | 749 | |
63c372cb | 750 | devshm = strjoina(temporary_mount, "/dev/shm"); |
8d953682 | 751 | (void) mkdir(devshm, 0755); |
2b85f4e1 LP |
752 | r = mount("/dev/shm", devshm, NULL, MS_BIND, NULL); |
753 | if (r < 0) { | |
2e4a4fae | 754 | r = log_debug_errno(errno, "Failed to bind mount /dev/shm on '%s': %m", devshm); |
2b85f4e1 LP |
755 | goto fail; |
756 | } | |
757 | ||
63c372cb | 758 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 759 | (void) mkdir(devmqueue, 0755); |
2e4a4fae YW |
760 | if (mount("/dev/mqueue", devmqueue, NULL, MS_BIND, NULL) < 0) |
761 | log_debug_errno(errno, "Failed to bind mount /dev/mqueue on '%s', ignoring: %m", devmqueue); | |
2b85f4e1 | 762 | |
63c372cb | 763 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 764 | (void) mkdir(devhugepages, 0755); |
2e4a4fae YW |
765 | if (mount("/dev/hugepages", devhugepages, NULL, MS_BIND, NULL) < 0) |
766 | log_debug_errno(errno, "Failed to bind mount /dev/hugepages on '%s', ignoring: %m", devhugepages); | |
2b85f4e1 | 767 | |
63c372cb | 768 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
769 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
770 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 771 | |
7f112f50 | 772 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 773 | r = clone_device_node(d, temporary_mount, &can_mknod); |
37b22b3b | 774 | /* ENXIO means the *source* is not a device file, skip creation in that case */ |
af984e13 | 775 | if (r < 0 && r != -ENXIO) |
2b85f4e1 | 776 | goto fail; |
7f112f50 LP |
777 | } |
778 | ||
2e4a4fae YW |
779 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
780 | if (r < 0) | |
105a1a36 | 781 | log_debug_errno(r, "Failed to set up basic device tree at '%s', ignoring: %m", temporary_mount); |
7f112f50 | 782 | |
ee818b89 AC |
783 | /* Create the /dev directory if missing. It is more likely to be |
784 | * missing when the service is started with RootDirectory. This is | |
785 | * consistent with mount units creating the mount points when missing. | |
786 | */ | |
34de407a | 787 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 788 | |
9e5f8252 | 789 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
790 | r = umount_recursive(mount_entry_path(m), 0); |
791 | if (r < 0) | |
792 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
793 | ||
34de407a | 794 | if (mount(dev, mount_entry_path(m), NULL, MS_MOVE, NULL) < 0) { |
2e4a4fae | 795 | r = log_debug_errno(errno, "Failed to move mount point '%s' to '%s': %m", dev, mount_entry_path(m)); |
2b85f4e1 LP |
796 | goto fail; |
797 | } | |
7f112f50 | 798 | |
1019a48f LP |
799 | (void) rmdir(dev); |
800 | (void) rmdir(temporary_mount); | |
7f112f50 | 801 | |
2b85f4e1 | 802 | return 0; |
7f112f50 | 803 | |
2b85f4e1 LP |
804 | fail: |
805 | if (devpts) | |
1019a48f | 806 | (void) umount(devpts); |
7f112f50 | 807 | |
2b85f4e1 | 808 | if (devshm) |
1019a48f | 809 | (void) umount(devshm); |
7f112f50 | 810 | |
2b85f4e1 | 811 | if (devhugepages) |
1019a48f | 812 | (void) umount(devhugepages); |
7f112f50 | 813 | |
2b85f4e1 | 814 | if (devmqueue) |
1019a48f | 815 | (void) umount(devmqueue); |
7f112f50 | 816 | |
1019a48f LP |
817 | (void) umount(dev); |
818 | (void) rmdir(dev); | |
819 | (void) rmdir(temporary_mount); | |
7f112f50 | 820 | |
2b85f4e1 | 821 | return r; |
7f112f50 LP |
822 | } |
823 | ||
2a2969fd | 824 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
825 | int r; |
826 | ||
827 | assert(m); | |
828 | ||
829 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
830 | * /dev. This is only used when RootDirectory= is set. */ | |
831 | ||
645767d6 LP |
832 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
833 | ||
5d997827 LP |
834 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
835 | if (r < 0) | |
836 | return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m"); | |
837 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
838 | return 0; | |
839 | ||
840 | if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
841 | return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m)); | |
842 | ||
843 | return 1; | |
844 | } | |
845 | ||
2a2969fd | 846 | static int mount_sysfs(const MountEntry *m) { |
5d997827 LP |
847 | int r; |
848 | ||
849 | assert(m); | |
850 | ||
645767d6 LP |
851 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
852 | ||
5d997827 LP |
853 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
854 | if (r < 0) | |
855 | return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m"); | |
856 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
857 | return 0; | |
858 | ||
859 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
860 | if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0) | |
861 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
862 | ||
863 | return 1; | |
864 | } | |
865 | ||
2a2969fd | 866 | static int mount_procfs(const MountEntry *m) { |
5d997827 LP |
867 | int r; |
868 | ||
869 | assert(m); | |
870 | ||
645767d6 LP |
871 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
872 | ||
5d997827 LP |
873 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); |
874 | if (r < 0) | |
875 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
876 | if (r > 0) /* make this a NOP if /proc is already a mount point */ | |
877 | return 0; | |
878 | ||
879 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */ | |
880 | if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0) | |
881 | return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m)); | |
882 | ||
883 | return 1; | |
884 | } | |
885 | ||
2abd4e38 | 886 | static int mount_tmpfs(const MountEntry *m) { |
df6b900a | 887 | const char *entry_path, *inner_path; |
abad72be | 888 | int r; |
abad72be | 889 | |
6c47cd7d LP |
890 | assert(m); |
891 | ||
df6b900a LP |
892 | entry_path = mount_entry_path(m); |
893 | inner_path = m->path_const; | |
894 | ||
2abd4e38 | 895 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d | 896 | |
abad72be CG |
897 | (void) mkdir_p_label(entry_path, 0755); |
898 | (void) umount_recursive(entry_path, 0); | |
6c47cd7d | 899 | |
abad72be CG |
900 | if (mount("tmpfs", entry_path, "tmpfs", m->flags, mount_entry_options(m)) < 0) |
901 | return log_debug_errno(errno, "Failed to mount %s: %m", entry_path); | |
902 | ||
df6b900a | 903 | r = label_fix_container(entry_path, inner_path, 0); |
abad72be | 904 | if (r < 0) |
df6b900a | 905 | return log_debug_errno(r, "Failed to fix label of '%s' as '%s': %m", entry_path, inner_path); |
6c47cd7d LP |
906 | |
907 | return 1; | |
908 | } | |
909 | ||
b3d13314 LB |
910 | static int mount_images(const MountEntry *m) { |
911 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; | |
912 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; | |
913 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; | |
914 | _cleanup_free_ void *root_hash_decoded = NULL; | |
915 | _cleanup_free_ char *verity_data = NULL, *hash_sig = NULL; | |
916 | DissectImageFlags dissect_image_flags = m->read_only ? DISSECT_IMAGE_READ_ONLY : 0; | |
917 | size_t root_hash_size = 0; | |
918 | int r; | |
919 | ||
920 | r = verity_metadata_load(mount_entry_source(m), NULL, &root_hash_decoded, &root_hash_size, &verity_data, &hash_sig); | |
921 | if (r < 0) | |
922 | return log_debug_errno(r, "Failed to load root hash: %m"); | |
923 | dissect_image_flags |= verity_data ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0; | |
924 | ||
925 | r = loop_device_make_by_path(mount_entry_source(m), | |
926 | m->read_only ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, | |
927 | verity_data ? 0 : LO_FLAGS_PARTSCAN, | |
928 | &loop_device); | |
929 | if (r < 0) | |
930 | return log_debug_errno(r, "Failed to create loop device for image: %m"); | |
931 | ||
427353f6 | 932 | r = dissect_image(loop_device->fd, root_hash_decoded, root_hash_size, verity_data, m->image_options, dissect_image_flags, &dissected_image); |
b3d13314 LB |
933 | /* No partition table? Might be a single-filesystem image, try again */ |
934 | if (!verity_data && r < 0 && r == -ENOPKG) | |
427353f6 | 935 | r = dissect_image(loop_device->fd, root_hash_decoded, root_hash_size, verity_data, m->image_options, dissect_image_flags|DISSECT_IMAGE_NO_PARTITION_TABLE, &dissected_image); |
b3d13314 LB |
936 | if (r < 0) |
937 | return log_debug_errno(r, "Failed to dissect image: %m"); | |
938 | ||
939 | r = dissected_image_decrypt(dissected_image, NULL, root_hash_decoded, root_hash_size, verity_data, hash_sig, NULL, 0, dissect_image_flags, &decrypted_image); | |
940 | if (r < 0) | |
941 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); | |
942 | ||
943 | r = mkdir_p_label(mount_entry_path(m), 0755); | |
944 | if (r < 0) | |
945 | return log_debug_errno(r, "Failed to create destination directory %s: %m", mount_entry_path(m)); | |
946 | r = umount_recursive(mount_entry_path(m), 0); | |
947 | if (r < 0) | |
948 | return log_debug_errno(r, "Failed to umount under destination directory %s: %m", mount_entry_path(m)); | |
949 | ||
950 | r = dissected_image_mount(dissected_image, mount_entry_path(m), UID_INVALID, dissect_image_flags); | |
951 | if (r < 0) | |
952 | return log_debug_errno(r, "Failed to mount image: %m"); | |
953 | ||
954 | if (decrypted_image) { | |
955 | r = decrypted_image_relinquish(decrypted_image); | |
956 | if (r < 0) | |
957 | return log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
958 | } | |
959 | ||
960 | loop_device_relinquish(loop_device); | |
961 | ||
962 | return 1; | |
963 | } | |
964 | ||
088696fe | 965 | static int follow_symlink( |
d2d6c096 | 966 | const char *root_directory, |
088696fe | 967 | MountEntry *m) { |
d2d6c096 | 968 | |
088696fe | 969 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
970 | int r; |
971 | ||
088696fe LP |
972 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
973 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
974 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
975 | * end and already have a fully normalized name. */ | |
8fceda93 | 976 | |
a5648b80 | 977 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL); |
088696fe LP |
978 | if (r < 0) |
979 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
980 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
981 | return 1; | |
8fceda93 | 982 | |
baaa35ad ZJS |
983 | if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */ |
984 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
985 | "Symlink loop on '%s'.", | |
986 | mount_entry_path(m)); | |
8fceda93 | 987 | |
088696fe | 988 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 989 | |
088696fe LP |
990 | free_and_replace(m->path_malloc, target); |
991 | m->has_prefix = true; | |
8fceda93 | 992 | |
088696fe LP |
993 | m->n_followed ++; |
994 | ||
995 | return 0; | |
8fceda93 LP |
996 | } |
997 | ||
ac0930c8 | 998 | static int apply_mount( |
8fceda93 | 999 | const char *root_directory, |
89bd586c | 1000 | MountEntry *m) { |
ac0930c8 | 1001 | |
e5f10caf | 1002 | _cleanup_free_ char *inaccessible = NULL; |
a227a4be | 1003 | bool rbind = true, make = false; |
15ae422b | 1004 | const char *what; |
15ae422b | 1005 | int r; |
15ae422b | 1006 | |
c17ec25e | 1007 | assert(m); |
15ae422b | 1008 | |
34de407a | 1009 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 1010 | |
c17ec25e | 1011 | switch (m->mode) { |
15ae422b | 1012 | |
160cfdbe | 1013 | case INACCESSIBLE: { |
e5f10caf AZ |
1014 | _cleanup_free_ char *tmp = NULL; |
1015 | const char *runtime_dir; | |
160cfdbe | 1016 | struct stat target; |
6d313367 LP |
1017 | |
1018 | /* First, get rid of everything that is below if there | |
1019 | * is anything... Then, overmount it with an | |
c4b41707 | 1020 | * inaccessible path. */ |
34de407a | 1021 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 1022 | |
088696fe LP |
1023 | if (lstat(mount_entry_path(m), &target) < 0) { |
1024 | if (errno == ENOENT && m->ignore) | |
1025 | return 0; | |
1026 | ||
cbc056c8 ZJS |
1027 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", |
1028 | mount_entry_path(m)); | |
088696fe | 1029 | } |
15ae422b | 1030 | |
e5f10caf | 1031 | if (geteuid() == 0) |
48b747fa | 1032 | runtime_dir = "/run"; |
e5f10caf | 1033 | else { |
48b747fa LP |
1034 | if (asprintf(&tmp, "/run/user/" UID_FMT, geteuid()) < 0) |
1035 | return -ENOMEM; | |
e5f10caf AZ |
1036 | |
1037 | runtime_dir = tmp; | |
1038 | } | |
1039 | ||
1040 | r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible); | |
1041 | if (r < 0) | |
baaa35ad ZJS |
1042 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), |
1043 | "File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
e5f10caf | 1044 | what = inaccessible; |
c4b41707 | 1045 | break; |
160cfdbe | 1046 | } |
fe3c2583 | 1047 | |
15ae422b | 1048 | case READONLY: |
15ae422b | 1049 | case READWRITE: |
1e05071d | 1050 | case READWRITE_IMPLICIT: |
8fceda93 | 1051 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
1052 | if (r == -ENOENT && m->ignore) |
1053 | return 0; | |
d944dc95 | 1054 | if (r < 0) |
cbc056c8 ZJS |
1055 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", |
1056 | mount_entry_path(m)); | |
1057 | if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY | |
1058 | * bit for the mount point if needed. */ | |
6b7c9f8b | 1059 | return 0; |
6b7c9f8b | 1060 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 1061 | what = mount_entry_path(m); |
6b7c9f8b | 1062 | break; |
15ae422b | 1063 | |
d2d6c096 LP |
1064 | case BIND_MOUNT: |
1065 | rbind = false; | |
d2d6c096 | 1066 | |
4831981d | 1067 | _fallthrough_; |
088696fe LP |
1068 | case BIND_MOUNT_RECURSIVE: { |
1069 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 1070 | |
cbc056c8 ZJS |
1071 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note |
1072 | * that bind mount source paths are always relative to the host root, hence we pass NULL as | |
1073 | * root directory to chase_symlinks() here. */ | |
088696fe | 1074 | |
a5648b80 | 1075 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL); |
088696fe LP |
1076 | if (r == -ENOENT && m->ignore) { |
1077 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
1078 | return 0; | |
1079 | } | |
1080 | if (r < 0) | |
1081 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
1082 | ||
1083 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
1084 | ||
1085 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
1086 | |
1087 | what = mount_entry_source(m); | |
a227a4be | 1088 | make = true; |
d2d6c096 | 1089 | break; |
088696fe | 1090 | } |
d2d6c096 | 1091 | |
6c47cd7d | 1092 | case EMPTY_DIR: |
2abd4e38 YW |
1093 | case TMPFS: |
1094 | return mount_tmpfs(m); | |
6c47cd7d | 1095 | |
ac0930c8 | 1096 | case PRIVATE_TMP: |
56a13a49 | 1097 | case PRIVATE_TMP_READONLY: |
89bd586c | 1098 | what = mount_entry_source(m); |
a227a4be | 1099 | make = true; |
15ae422b | 1100 | break; |
e364ad06 | 1101 | |
d6797c92 | 1102 | case PRIVATE_DEV: |
5d997827 LP |
1103 | return mount_private_dev(m); |
1104 | ||
1105 | case BIND_DEV: | |
1106 | return mount_bind_dev(m); | |
1107 | ||
1108 | case SYSFS: | |
1109 | return mount_sysfs(m); | |
1110 | ||
1111 | case PROCFS: | |
1112 | return mount_procfs(m); | |
d6797c92 | 1113 | |
b3d13314 LB |
1114 | case MOUNT_IMAGES: |
1115 | return mount_images(m); | |
1116 | ||
e364ad06 LP |
1117 | default: |
1118 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
1119 | } |
1120 | ||
ac0930c8 | 1121 | assert(what); |
15ae422b | 1122 | |
a227a4be LP |
1123 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) { |
1124 | bool try_again = false; | |
1125 | r = -errno; | |
1126 | ||
1127 | if (r == -ENOENT && make) { | |
1128 | struct stat st; | |
1129 | ||
cbc056c8 ZJS |
1130 | /* Hmm, either the source or the destination are missing. Let's see if we can create |
1131 | the destination, then try again. */ | |
a227a4be | 1132 | |
e8717862 | 1133 | if (stat(what, &st) < 0) |
5dc60faa | 1134 | log_error_errno(errno, "Mount point source '%s' is not accessible: %m", what); |
e8717862 LP |
1135 | else { |
1136 | int q; | |
a227a4be LP |
1137 | |
1138 | (void) mkdir_parents(mount_entry_path(m), 0755); | |
1139 | ||
1140 | if (S_ISDIR(st.st_mode)) | |
e8717862 | 1141 | q = mkdir(mount_entry_path(m), 0755) < 0 ? -errno : 0; |
a227a4be | 1142 | else |
e8717862 LP |
1143 | q = touch(mount_entry_path(m)); |
1144 | ||
1145 | if (q < 0) | |
cbc056c8 ZJS |
1146 | log_error_errno(q, "Failed to create destination mount point node '%s': %m", |
1147 | mount_entry_path(m)); | |
e8717862 LP |
1148 | else |
1149 | try_again = true; | |
a227a4be LP |
1150 | } |
1151 | } | |
1152 | ||
1153 | if (try_again) { | |
1154 | if (mount(what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL) < 0) | |
1155 | r = -errno; | |
1156 | else | |
1157 | r = 0; | |
1158 | } | |
1159 | ||
1160 | if (r < 0) | |
5dc60faa | 1161 | return log_error_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); |
a227a4be | 1162 | } |
6b7c9f8b | 1163 | |
34de407a | 1164 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1165 | return 0; |
ac0930c8 | 1166 | } |
15ae422b | 1167 | |
6b000af4 | 1168 | static int make_read_only(const MountEntry *m, char **deny_list, FILE *proc_self_mountinfo) { |
9ce4e4b0 | 1169 | unsigned long new_flags = 0, flags_mask = 0; |
763a260a | 1170 | bool submounts = false; |
6b7c9f8b | 1171 | int r = 0; |
15ae422b | 1172 | |
c17ec25e | 1173 | assert(m); |
ac9de0b3 | 1174 | assert(proc_self_mountinfo); |
ac0930c8 | 1175 | |
9ce4e4b0 LP |
1176 | if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) { |
1177 | new_flags |= MS_RDONLY; | |
1178 | flags_mask |= MS_RDONLY; | |
1179 | } | |
1180 | ||
1181 | if (m->nosuid) { | |
1182 | new_flags |= MS_NOSUID; | |
1183 | flags_mask |= MS_NOSUID; | |
1184 | } | |
1185 | ||
1186 | if (flags_mask == 0) /* No Change? */ | |
6b7c9f8b LP |
1187 | return 0; |
1188 | ||
9ce4e4b0 LP |
1189 | /* We generally apply these changes recursively, except for /dev, and the cases we know there's |
1190 | * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the | |
1191 | * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace | |
1192 | * and running Linux <= 4.17. */ | |
1193 | submounts = | |
1194 | mount_entry_read_only(m) && | |
1195 | !IN_SET(m->mode, EMPTY_DIR, TMPFS); | |
1196 | if (submounts) | |
6b000af4 | 1197 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, deny_list, proc_self_mountinfo); |
9ce4e4b0 | 1198 | else |
7cce68e1 | 1199 | r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo); |
9ce4e4b0 | 1200 | |
867189b5 LP |
1201 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked |
1202 | * read-only already stays this way. This improves compatibility with container managers, where we | |
1203 | * won't attempt to undo read-only mounts already applied. */ | |
ac0930c8 | 1204 | |
8fceda93 | 1205 | if (r == -ENOENT && m->ignore) |
867189b5 | 1206 | return 0; |
763a260a | 1207 | if (r < 0) |
9ce4e4b0 | 1208 | return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m), |
763a260a | 1209 | submounts ? " and its submounts" : ""); |
763a260a | 1210 | return 0; |
d944dc95 LP |
1211 | } |
1212 | ||
9b68367b | 1213 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1214 | assert(ns_info); |
1215 | ||
9c988f93 DH |
1216 | /* |
1217 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1218 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1219 | * first place... |
9c988f93 | 1220 | */ |
5d997827 | 1221 | |
9b68367b YW |
1222 | return ns_info->mount_apivfs || |
1223 | ns_info->protect_control_groups || | |
1224 | ns_info->protect_kernel_tunables; | |
5d997827 LP |
1225 | } |
1226 | ||
da6053d0 | 1227 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1228 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1229 | char** read_write_paths, |
1230 | char** read_only_paths, | |
1231 | char** inaccessible_paths, | |
6c47cd7d | 1232 | char** empty_directories, |
da6053d0 LP |
1233 | size_t n_bind_mounts, |
1234 | size_t n_temporary_filesystems, | |
b3d13314 | 1235 | size_t n_mount_images, |
2652c6c1 DH |
1236 | const char* tmp_dir, |
1237 | const char* var_tmp_dir, | |
52b3d652 | 1238 | const char* log_namespace) { |
2652c6c1 | 1239 | |
da6053d0 LP |
1240 | size_t protect_home_cnt; |
1241 | size_t protect_system_cnt = | |
52b3d652 | 1242 | (ns_info->protect_system == PROTECT_SYSTEM_STRICT ? |
f471b2af | 1243 | ELEMENTSOF(protect_system_strict_table) : |
52b3d652 | 1244 | ((ns_info->protect_system == PROTECT_SYSTEM_FULL) ? |
f471b2af | 1245 | ELEMENTSOF(protect_system_full_table) : |
52b3d652 | 1246 | ((ns_info->protect_system == PROTECT_SYSTEM_YES) ? |
f471b2af DH |
1247 | ELEMENTSOF(protect_system_yes_table) : 0))); |
1248 | ||
b6c432ca | 1249 | protect_home_cnt = |
52b3d652 | 1250 | (ns_info->protect_home == PROTECT_HOME_YES ? |
b6c432ca | 1251 | ELEMENTSOF(protect_home_yes_table) : |
52b3d652 | 1252 | ((ns_info->protect_home == PROTECT_HOME_READ_ONLY) ? |
e4da7d8c | 1253 | ELEMENTSOF(protect_home_read_only_table) : |
52b3d652 | 1254 | ((ns_info->protect_home == PROTECT_HOME_TMPFS) ? |
e4da7d8c | 1255 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); |
b6c432ca | 1256 | |
2652c6c1 DH |
1257 | return !!tmp_dir + !!var_tmp_dir + |
1258 | strv_length(read_write_paths) + | |
1259 | strv_length(read_only_paths) + | |
1260 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1261 | strv_length(empty_directories) + |
d2d6c096 | 1262 | n_bind_mounts + |
b3d13314 | 1263 | n_mount_images + |
2abd4e38 | 1264 | n_temporary_filesystems + |
c575770b DH |
1265 | ns_info->private_dev + |
1266 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
c575770b | 1267 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + |
94a7b275 KK |
1268 | (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) + |
1269 | (ns_info->protect_control_groups ? 1 : 0) + | |
5d997827 | 1270 | protect_home_cnt + protect_system_cnt + |
aecd5ac6 | 1271 | (ns_info->protect_hostname ? 2 : 0) + |
91dd5f7c LP |
1272 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) + |
1273 | !!log_namespace; | |
2652c6c1 DH |
1274 | } |
1275 | ||
da6053d0 | 1276 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1277 | assert(root_directory); |
f8b64b57 LP |
1278 | assert(n_mounts); |
1279 | assert(mounts || *n_mounts == 0); | |
1280 | ||
93bab288 | 1281 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1282 | |
1283 | drop_duplicates(mounts, n_mounts); | |
1284 | drop_outside_root(root_directory, mounts, n_mounts); | |
1285 | drop_inaccessible(mounts, n_mounts); | |
1286 | drop_nop(mounts, n_mounts); | |
1287 | } | |
1288 | ||
c8c535d5 LP |
1289 | static bool root_read_only( |
1290 | char **read_only_paths, | |
1291 | ProtectSystem protect_system) { | |
1292 | ||
1293 | /* Determine whether the root directory is going to be read-only given the configured settings. */ | |
1294 | ||
1295 | if (protect_system == PROTECT_SYSTEM_STRICT) | |
1296 | return true; | |
1297 | ||
de46b2be | 1298 | if (prefixed_path_strv_contains(read_only_paths, "/")) |
c8c535d5 LP |
1299 | return true; |
1300 | ||
1301 | return false; | |
1302 | } | |
1303 | ||
1304 | static bool home_read_only( | |
1305 | char** read_only_paths, | |
1306 | char** inaccessible_paths, | |
1307 | char** empty_directories, | |
1308 | const BindMount *bind_mounts, | |
1309 | size_t n_bind_mounts, | |
1310 | const TemporaryFileSystem *temporary_filesystems, | |
1311 | size_t n_temporary_filesystems, | |
1312 | ProtectHome protect_home) { | |
1313 | ||
1314 | size_t i; | |
1315 | ||
1316 | /* Determine whether the /home directory is going to be read-only given the configured settings. Yes, | |
1317 | * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple | |
1318 | * settings. */ | |
1319 | ||
1320 | if (protect_home != PROTECT_HOME_NO) | |
1321 | return true; | |
1322 | ||
de46b2be TM |
1323 | if (prefixed_path_strv_contains(read_only_paths, "/home") || |
1324 | prefixed_path_strv_contains(inaccessible_paths, "/home") || | |
1325 | prefixed_path_strv_contains(empty_directories, "/home")) | |
c8c535d5 LP |
1326 | return true; |
1327 | ||
1328 | for (i = 0; i < n_temporary_filesystems; i++) | |
1329 | if (path_equal(temporary_filesystems[i].path, "/home")) | |
1330 | return true; | |
1331 | ||
1332 | /* If /home is overmounted with some dir from the host it's not writable. */ | |
1333 | for (i = 0; i < n_bind_mounts; i++) | |
1334 | if (path_equal(bind_mounts[i].destination, "/home")) | |
1335 | return true; | |
1336 | ||
1337 | return false; | |
1338 | } | |
1339 | ||
613b411c | 1340 | int setup_namespace( |
ee818b89 | 1341 | const char* root_directory, |
915e6d16 | 1342 | const char* root_image, |
18d73705 | 1343 | const MountOptions *root_image_options, |
bb0ff3fb | 1344 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1345 | char** read_write_paths, |
1346 | char** read_only_paths, | |
1347 | char** inaccessible_paths, | |
6c47cd7d | 1348 | char** empty_directories, |
d2d6c096 | 1349 | const BindMount *bind_mounts, |
da6053d0 | 1350 | size_t n_bind_mounts, |
2abd4e38 | 1351 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1352 | size_t n_temporary_filesystems, |
b3d13314 LB |
1353 | const MountImage *mount_images, |
1354 | size_t n_mount_images, | |
a004cb4c LP |
1355 | const char* tmp_dir, |
1356 | const char* var_tmp_dir, | |
91dd5f7c | 1357 | const char *log_namespace, |
915e6d16 | 1358 | unsigned long mount_flags, |
0389f4fa LB |
1359 | const void *root_hash, |
1360 | size_t root_hash_size, | |
1361 | const char *root_hash_path, | |
d4d55b0d LB |
1362 | const void *root_hash_sig, |
1363 | size_t root_hash_sig_size, | |
1364 | const char *root_hash_sig_path, | |
0389f4fa | 1365 | const char *root_verity, |
7cc5ef5f ZJS |
1366 | DissectImageFlags dissect_image_flags, |
1367 | char **error_path) { | |
15ae422b | 1368 | |
915e6d16 | 1369 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1370 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1371 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
0389f4fa | 1372 | _cleanup_free_ void *root_hash_decoded = NULL; |
c2923fdc | 1373 | _cleanup_free_ char *verity_data = NULL, *hash_sig_path = NULL; |
5f7a690a | 1374 | MountEntry *m = NULL, *mounts = NULL; |
0389f4fa | 1375 | size_t n_mounts; |
d18aff04 | 1376 | bool require_prefix = false; |
9b68367b | 1377 | const char *root; |
c17ec25e | 1378 | int r = 0; |
15ae422b | 1379 | |
915e6d16 LP |
1380 | assert(ns_info); |
1381 | ||
613b411c | 1382 | if (mount_flags == 0) |
c17ec25e | 1383 | mount_flags = MS_SHARED; |
ac0930c8 | 1384 | |
915e6d16 LP |
1385 | if (root_image) { |
1386 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1387 | ||
c8c535d5 LP |
1388 | /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */ |
1389 | if (root_read_only(read_only_paths, | |
52b3d652 | 1390 | ns_info->protect_system) && |
c8c535d5 LP |
1391 | home_read_only(read_only_paths, inaccessible_paths, empty_directories, |
1392 | bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems, | |
52b3d652 | 1393 | ns_info->protect_home) && |
c9ef8573 | 1394 | strv_isempty(read_write_paths)) |
915e6d16 LP |
1395 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1396 | ||
1397 | r = loop_device_make_by_path(root_image, | |
b0a94268 | 1398 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, |
e08f94ac | 1399 | LO_FLAGS_PARTSCAN, |
915e6d16 LP |
1400 | &loop_device); |
1401 | if (r < 0) | |
763a260a | 1402 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
915e6d16 | 1403 | |
cbc056c8 ZJS |
1404 | r = verity_metadata_load(root_image, |
1405 | root_hash_path, | |
1406 | root_hash ? NULL : &root_hash_decoded, | |
1407 | root_hash ? NULL : &root_hash_size, | |
1408 | root_verity ? NULL : &verity_data, | |
1409 | root_hash_sig || root_hash_sig_path ? NULL : &hash_sig_path); | |
78ebe980 | 1410 | if (r < 0) |
763a260a | 1411 | return log_debug_errno(r, "Failed to load root hash: %m"); |
0389f4fa | 1412 | dissect_image_flags |= root_verity || verity_data ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0; |
78ebe980 | 1413 | |
cbc056c8 ZJS |
1414 | r = dissect_image(loop_device->fd, |
1415 | root_hash ?: root_hash_decoded, | |
1416 | root_hash_size, | |
1417 | root_verity ?: verity_data, | |
18d73705 | 1418 | root_image_options, |
cbc056c8 ZJS |
1419 | dissect_image_flags, |
1420 | &dissected_image); | |
78ebe980 | 1421 | if (r < 0) |
763a260a | 1422 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 | 1423 | |
cbc056c8 ZJS |
1424 | r = dissected_image_decrypt(dissected_image, |
1425 | NULL, | |
1426 | root_hash ?: root_hash_decoded, | |
1427 | root_hash_size, | |
1428 | root_verity ?: verity_data, | |
1429 | root_hash_sig_path ?: hash_sig_path, | |
1430 | root_hash_sig, | |
1431 | root_hash_sig_size, | |
1432 | dissect_image_flags, | |
1433 | &decrypted_image); | |
915e6d16 | 1434 | if (r < 0) |
763a260a | 1435 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1436 | } |
1437 | ||
e908468b LP |
1438 | if (root_directory) |
1439 | root = root_directory; | |
0722b359 JS |
1440 | else { |
1441 | /* Always create the mount namespace in a temporary directory, instead of operating | |
1442 | * directly in the root. The temporary directory prevents any mounts from being | |
1443 | * potentially obscured my other mounts we already applied. | |
1444 | * We use the same mount point for all images, which is safe, since they all live | |
1445 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1446 | |
1447 | root = "/run/systemd/unit-root"; | |
1448 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1449 | require_prefix = true; |
0722b359 | 1450 | } |
e908468b | 1451 | |
cfbeb4ef LP |
1452 | n_mounts = namespace_calculate_mounts( |
1453 | ns_info, | |
1454 | read_write_paths, | |
1455 | read_only_paths, | |
1456 | inaccessible_paths, | |
6c47cd7d | 1457 | empty_directories, |
f5c52a77 | 1458 | n_bind_mounts, |
2abd4e38 | 1459 | n_temporary_filesystems, |
b3d13314 | 1460 | n_mount_images, |
cfbeb4ef | 1461 | tmp_dir, var_tmp_dir, |
52b3d652 | 1462 | log_namespace); |
613b411c | 1463 | |
f0a4feb0 | 1464 | if (n_mounts > 0) { |
5f7a690a LP |
1465 | m = mounts = new0(MountEntry, n_mounts); |
1466 | if (!mounts) | |
1467 | return -ENOMEM; | |
1468 | ||
d18aff04 | 1469 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1470 | if (r < 0) |
f0a4feb0 | 1471 | goto finish; |
613b411c | 1472 | |
d18aff04 | 1473 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1474 | if (r < 0) |
f0a4feb0 | 1475 | goto finish; |
613b411c | 1476 | |
d18aff04 | 1477 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1478 | if (r < 0) |
f0a4feb0 | 1479 | goto finish; |
7ff7394d | 1480 | |
6c47cd7d LP |
1481 | r = append_empty_dir_mounts(&m, empty_directories); |
1482 | if (r < 0) | |
1483 | goto finish; | |
1484 | ||
d2d6c096 LP |
1485 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1486 | if (r < 0) | |
1487 | goto finish; | |
1488 | ||
2abd4e38 YW |
1489 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1490 | if (r < 0) | |
1491 | goto finish; | |
1492 | ||
613b411c | 1493 | if (tmp_dir) { |
56a13a49 ZJS |
1494 | bool ro = streq(tmp_dir, RUN_SYSTEMD_EMPTY); |
1495 | ||
34de407a | 1496 | *(m++) = (MountEntry) { |
5327c910 | 1497 | .path_const = "/tmp", |
56a13a49 | 1498 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1499 | .source_const = tmp_dir, |
5327c910 | 1500 | }; |
613b411c | 1501 | } |
7ff7394d | 1502 | |
613b411c | 1503 | if (var_tmp_dir) { |
56a13a49 ZJS |
1504 | bool ro = streq(var_tmp_dir, RUN_SYSTEMD_EMPTY); |
1505 | ||
34de407a | 1506 | *(m++) = (MountEntry) { |
5327c910 | 1507 | .path_const = "/var/tmp", |
56a13a49 | 1508 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1509 | .source_const = var_tmp_dir, |
5327c910 | 1510 | }; |
7ff7394d | 1511 | } |
ac0930c8 | 1512 | |
b3d13314 LB |
1513 | r = append_mount_images(&m, mount_images, n_mount_images); |
1514 | if (r < 0) | |
1515 | goto finish; | |
1516 | ||
c575770b | 1517 | if (ns_info->private_dev) { |
34de407a | 1518 | *(m++) = (MountEntry) { |
5327c910 LP |
1519 | .path_const = "/dev", |
1520 | .mode = PRIVATE_DEV, | |
9ce4e4b0 | 1521 | .flags = DEV_MOUNT_OPTIONS, |
5327c910 | 1522 | }; |
7f112f50 LP |
1523 | } |
1524 | ||
c575770b | 1525 | if (ns_info->protect_kernel_tunables) { |
cbc056c8 ZJS |
1526 | r = append_static_mounts(&m, |
1527 | protect_kernel_tunables_table, | |
1528 | ELEMENTSOF(protect_kernel_tunables_table), | |
1529 | ns_info->ignore_protect_paths); | |
c575770b | 1530 | if (r < 0) |
f0a4feb0 | 1531 | goto finish; |
c575770b DH |
1532 | } |
1533 | ||
1534 | if (ns_info->protect_kernel_modules) { | |
cbc056c8 ZJS |
1535 | r = append_static_mounts(&m, |
1536 | protect_kernel_modules_table, | |
1537 | ELEMENTSOF(protect_kernel_modules_table), | |
1538 | ns_info->ignore_protect_paths); | |
c575770b | 1539 | if (r < 0) |
f0a4feb0 | 1540 | goto finish; |
c575770b | 1541 | } |
59eeb84b | 1542 | |
94a7b275 | 1543 | if (ns_info->protect_kernel_logs) { |
cbc056c8 ZJS |
1544 | r = append_static_mounts(&m, |
1545 | protect_kernel_logs_table, | |
1546 | ELEMENTSOF(protect_kernel_logs_table), | |
1547 | ns_info->ignore_protect_paths); | |
94a7b275 KK |
1548 | if (r < 0) |
1549 | goto finish; | |
1550 | } | |
1551 | ||
c575770b | 1552 | if (ns_info->protect_control_groups) { |
34de407a | 1553 | *(m++) = (MountEntry) { |
5327c910 LP |
1554 | .path_const = "/sys/fs/cgroup", |
1555 | .mode = READONLY, | |
1556 | }; | |
59eeb84b LP |
1557 | } |
1558 | ||
52b3d652 | 1559 | r = append_protect_home(&m, ns_info->protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1560 | if (r < 0) |
f0a4feb0 | 1561 | goto finish; |
417116f2 | 1562 | |
52b3d652 | 1563 | r = append_protect_system(&m, ns_info->protect_system, false); |
f471b2af | 1564 | if (r < 0) |
f0a4feb0 | 1565 | goto finish; |
417116f2 | 1566 | |
9b68367b | 1567 | if (namespace_info_mount_apivfs(ns_info)) { |
cbc056c8 ZJS |
1568 | r = append_static_mounts(&m, |
1569 | apivfs_table, | |
1570 | ELEMENTSOF(apivfs_table), | |
1571 | ns_info->ignore_protect_paths); | |
5d997827 LP |
1572 | if (r < 0) |
1573 | goto finish; | |
1574 | } | |
1575 | ||
aecd5ac6 TM |
1576 | if (ns_info->protect_hostname) { |
1577 | *(m++) = (MountEntry) { | |
1578 | .path_const = "/proc/sys/kernel/hostname", | |
1579 | .mode = READONLY, | |
1580 | }; | |
1581 | *(m++) = (MountEntry) { | |
1582 | .path_const = "/proc/sys/kernel/domainname", | |
1583 | .mode = READONLY, | |
1584 | }; | |
1585 | } | |
1586 | ||
91dd5f7c LP |
1587 | if (log_namespace) { |
1588 | _cleanup_free_ char *q; | |
1589 | ||
1590 | q = strjoin("/run/systemd/journal.", log_namespace); | |
1591 | if (!q) { | |
1592 | r = -ENOMEM; | |
1593 | goto finish; | |
1594 | } | |
1595 | ||
1596 | *(m++) = (MountEntry) { | |
1597 | .path_const = "/run/systemd/journal", | |
1598 | .mode = BIND_MOUNT_RECURSIVE, | |
1599 | .read_only = true, | |
1600 | .source_malloc = TAKE_PTR(q), | |
1601 | }; | |
1602 | } | |
1603 | ||
f0a4feb0 | 1604 | assert(mounts + n_mounts == m); |
ac0930c8 | 1605 | |
5327c910 | 1606 | /* Prepend the root directory where that's necessary */ |
e908468b | 1607 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1608 | if (r < 0) |
1609 | goto finish; | |
1610 | ||
839f1877 | 1611 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1612 | } |
1613 | ||
1beab8b0 LP |
1614 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1615 | ||
d944dc95 | 1616 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1617 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 | 1618 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
cbc056c8 ZJS |
1619 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter |
1620 | * in place that doesn't allow us to create namespaces (or a missing cap), then | |
1621 | * propagate a recognizable error back, which the caller can use to detect this case | |
1622 | * (and only this) and optionally continue without namespacing applied. */ | |
1beab8b0 LP |
1623 | r = -ENOANO; |
1624 | ||
d944dc95 LP |
1625 | goto finish; |
1626 | } | |
1e4e94c8 | 1627 | |
9b68367b YW |
1628 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1629 | * shows up in the parent */ | |
1630 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1631 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1632 | goto finish; |
ee818b89 AC |
1633 | } |
1634 | ||
915e6d16 | 1635 | if (root_image) { |
e908468b | 1636 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1637 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1638 | if (r < 0) { |
1639 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1640 | goto finish; |
763a260a | 1641 | } |
915e6d16 | 1642 | |
07ce7407 TM |
1643 | if (decrypted_image) { |
1644 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1645 | if (r < 0) { |
1646 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1647 | goto finish; |
763a260a | 1648 | } |
07ce7407 | 1649 | } |
78ebe980 | 1650 | |
915e6d16 LP |
1651 | loop_device_relinquish(loop_device); |
1652 | ||
1653 | } else if (root_directory) { | |
1654 | ||
e908468b LP |
1655 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1656 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1657 | if (r < 0) { |
1658 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1659 | goto finish; |
763a260a | 1660 | } |
8f1ad200 | 1661 | if (r == 0) { |
e908468b | 1662 | if (mount(root, root, NULL, MS_BIND|MS_REC, NULL) < 0) { |
763a260a | 1663 | r = log_debug_errno(errno, "Failed to bind mount '%s': %m", root); |
8f1ad200 LP |
1664 | goto finish; |
1665 | } | |
d944dc95 | 1666 | } |
e908468b | 1667 | |
9b68367b | 1668 | } else { |
e908468b LP |
1669 | |
1670 | /* Let's mount the main root directory to the root directory to use */ | |
1671 | if (mount("/", root, NULL, MS_BIND|MS_REC, NULL) < 0) { | |
763a260a | 1672 | r = log_debug_errno(errno, "Failed to bind mount '/' on '%s': %m", root); |
e908468b LP |
1673 | goto finish; |
1674 | } | |
ee818b89 | 1675 | } |
c2c13f2d | 1676 | |
4e0c20de LP |
1677 | /* Try to set up the new root directory before mounting anything else there. */ |
1678 | if (root_image || root_directory) | |
1679 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1680 | ||
f0a4feb0 | 1681 | if (n_mounts > 0) { |
ac9de0b3 | 1682 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b000af4 | 1683 | _cleanup_free_ char **deny_list = NULL; |
da6053d0 | 1684 | size_t j; |
6b7c9f8b | 1685 | |
cbc056c8 ZJS |
1686 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of |
1687 | * /proc. For example, this is the case with the option: 'InaccessiblePaths=/proc'. */ | |
ac9de0b3 TR |
1688 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); |
1689 | if (!proc_self_mountinfo) { | |
763a260a | 1690 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
7cc5ef5f ZJS |
1691 | if (error_path) |
1692 | *error_path = strdup("/proc/self/mountinfo"); | |
ac9de0b3 TR |
1693 | goto finish; |
1694 | } | |
1695 | ||
088696fe LP |
1696 | /* First round, establish all mounts we need */ |
1697 | for (;;) { | |
1698 | bool again = false; | |
1699 | ||
1700 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1701 | ||
1702 | if (m->applied) | |
1703 | continue; | |
1704 | ||
1705 | r = follow_symlink(root, m); | |
7cc5ef5f ZJS |
1706 | if (r < 0) { |
1707 | if (error_path && mount_entry_path(m)) | |
1708 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1709 | goto finish; |
7cc5ef5f | 1710 | } |
088696fe | 1711 | if (r == 0) { |
cbc056c8 ZJS |
1712 | /* We hit a symlinked mount point. The entry got rewritten and might |
1713 | * point to a very different place now. Let's normalize the changed | |
1714 | * list, and start from the beginning. After all to mount the entry | |
1715 | * at the new location we might need some other mounts first */ | |
088696fe LP |
1716 | again = true; |
1717 | break; | |
1718 | } | |
1719 | ||
1720 | r = apply_mount(root, m); | |
7cc5ef5f ZJS |
1721 | if (r < 0) { |
1722 | if (error_path && mount_entry_path(m)) | |
1723 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1724 | goto finish; |
7cc5ef5f | 1725 | } |
088696fe LP |
1726 | |
1727 | m->applied = true; | |
1728 | } | |
1729 | ||
1730 | if (!again) | |
1731 | break; | |
1732 | ||
839f1877 | 1733 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1734 | } |
15ae422b | 1735 | |
6b000af4 LP |
1736 | /* Create a deny list we can pass to bind_mount_recursive() */ |
1737 | deny_list = new(char*, n_mounts+1); | |
1738 | if (!deny_list) { | |
5f7a690a LP |
1739 | r = -ENOMEM; |
1740 | goto finish; | |
1741 | } | |
f0a4feb0 | 1742 | for (j = 0; j < n_mounts; j++) |
6b000af4 LP |
1743 | deny_list[j] = (char*) mount_entry_path(mounts+j); |
1744 | deny_list[j] = NULL; | |
6b7c9f8b LP |
1745 | |
1746 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1747 | for (m = mounts; m < mounts + n_mounts; ++m) { |
6b000af4 | 1748 | r = make_read_only(m, deny_list, proc_self_mountinfo); |
7cc5ef5f ZJS |
1749 | if (r < 0) { |
1750 | if (error_path && mount_entry_path(m)) | |
1751 | *error_path = strdup(mount_entry_path(m)); | |
d944dc95 | 1752 | goto finish; |
7cc5ef5f | 1753 | } |
c2c13f2d | 1754 | } |
15ae422b LP |
1755 | } |
1756 | ||
9b68367b YW |
1757 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1758 | r = mount_move_root(root); | |
763a260a YW |
1759 | if (r < 0) { |
1760 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1761 | goto finish; |
763a260a | 1762 | } |
ee818b89 | 1763 | |
55fe7432 | 1764 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1765 | * reestablish propagation from our side to the host, since |
1766 | * what's disconnected is disconnected. */ | |
d944dc95 | 1767 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1768 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1769 | goto finish; |
1770 | } | |
15ae422b | 1771 | |
d944dc95 | 1772 | r = 0; |
15ae422b | 1773 | |
d944dc95 | 1774 | finish: |
0cd41757 LP |
1775 | if (n_mounts > 0) |
1776 | for (m = mounts; m < mounts + n_mounts; m++) | |
1777 | mount_entry_done(m); | |
613b411c | 1778 | |
5f7a690a LP |
1779 | free(mounts); |
1780 | ||
613b411c LP |
1781 | return r; |
1782 | } | |
1783 | ||
da6053d0 LP |
1784 | void bind_mount_free_many(BindMount *b, size_t n) { |
1785 | size_t i; | |
d2d6c096 LP |
1786 | |
1787 | assert(b || n == 0); | |
1788 | ||
1789 | for (i = 0; i < n; i++) { | |
1790 | free(b[i].source); | |
1791 | free(b[i].destination); | |
1792 | } | |
1793 | ||
1794 | free(b); | |
1795 | } | |
1796 | ||
da6053d0 | 1797 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
1798 | _cleanup_free_ char *s = NULL, *d = NULL; |
1799 | BindMount *c; | |
1800 | ||
1801 | assert(b); | |
1802 | assert(n); | |
1803 | assert(item); | |
1804 | ||
1805 | s = strdup(item->source); | |
1806 | if (!s) | |
1807 | return -ENOMEM; | |
1808 | ||
1809 | d = strdup(item->destination); | |
1810 | if (!d) | |
1811 | return -ENOMEM; | |
1812 | ||
aa484f35 | 1813 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
1814 | if (!c) |
1815 | return -ENOMEM; | |
1816 | ||
1817 | *b = c; | |
1818 | ||
1819 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
1820 | .source = TAKE_PTR(s), |
1821 | .destination = TAKE_PTR(d), | |
d2d6c096 | 1822 | .read_only = item->read_only, |
9ce4e4b0 | 1823 | .nosuid = item->nosuid, |
d2d6c096 LP |
1824 | .recursive = item->recursive, |
1825 | .ignore_enoent = item->ignore_enoent, | |
1826 | }; | |
1827 | ||
d2d6c096 LP |
1828 | return 0; |
1829 | } | |
1830 | ||
b3d13314 LB |
1831 | MountImage* mount_image_free_many(MountImage *m, size_t *n) { |
1832 | size_t i; | |
1833 | ||
1834 | assert(n); | |
1835 | assert(m || *n == 0); | |
1836 | ||
1837 | for (i = 0; i < *n; i++) { | |
1838 | free(m[i].source); | |
1839 | free(m[i].destination); | |
427353f6 | 1840 | mount_options_free_all(m[i].mount_options); |
b3d13314 LB |
1841 | } |
1842 | ||
1843 | free(m); | |
1844 | *n = 0; | |
1845 | return NULL; | |
1846 | } | |
1847 | ||
1848 | int mount_image_add(MountImage **m, size_t *n, const MountImage *item) { | |
1849 | _cleanup_free_ char *s = NULL, *d = NULL; | |
427353f6 LB |
1850 | _cleanup_(mount_options_free_allp) MountOptions *options = NULL; |
1851 | MountOptions *i; | |
b3d13314 LB |
1852 | MountImage *c; |
1853 | ||
1854 | assert(m); | |
1855 | assert(n); | |
1856 | assert(item); | |
1857 | ||
1858 | s = strdup(item->source); | |
1859 | if (!s) | |
1860 | return -ENOMEM; | |
1861 | ||
1862 | d = strdup(item->destination); | |
1863 | if (!d) | |
1864 | return -ENOMEM; | |
1865 | ||
427353f6 LB |
1866 | LIST_FOREACH(mount_options, i, item->mount_options) { |
1867 | _cleanup_(mount_options_free_allp) MountOptions *o; | |
1868 | ||
1869 | o = new(MountOptions, 1); | |
1870 | if (!o) | |
1871 | return -ENOMEM; | |
1872 | ||
1873 | *o = (MountOptions) { | |
1874 | .partition_designator = i->partition_designator, | |
1875 | .options = strdup(i->options), | |
1876 | }; | |
1877 | if (!o->options) | |
1878 | return -ENOMEM; | |
1879 | ||
1880 | LIST_APPEND(mount_options, options, TAKE_PTR(o)); | |
1881 | } | |
1882 | ||
b3d13314 LB |
1883 | c = reallocarray(*m, *n + 1, sizeof(MountImage)); |
1884 | if (!c) | |
1885 | return -ENOMEM; | |
1886 | ||
1887 | *m = c; | |
1888 | ||
1889 | c[(*n) ++] = (MountImage) { | |
1890 | .source = TAKE_PTR(s), | |
1891 | .destination = TAKE_PTR(d), | |
427353f6 | 1892 | .mount_options = TAKE_PTR(options), |
b3d13314 LB |
1893 | .ignore_enoent = item->ignore_enoent, |
1894 | }; | |
1895 | ||
1896 | return 0; | |
1897 | } | |
1898 | ||
da6053d0 LP |
1899 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
1900 | size_t i; | |
2abd4e38 YW |
1901 | |
1902 | assert(t || n == 0); | |
1903 | ||
1904 | for (i = 0; i < n; i++) { | |
1905 | free(t[i].path); | |
1906 | free(t[i].options); | |
1907 | } | |
1908 | ||
1909 | free(t); | |
1910 | } | |
1911 | ||
1912 | int temporary_filesystem_add( | |
1913 | TemporaryFileSystem **t, | |
da6053d0 | 1914 | size_t *n, |
2abd4e38 YW |
1915 | const char *path, |
1916 | const char *options) { | |
1917 | ||
1918 | _cleanup_free_ char *p = NULL, *o = NULL; | |
1919 | TemporaryFileSystem *c; | |
1920 | ||
1921 | assert(t); | |
1922 | assert(n); | |
1923 | assert(path); | |
1924 | ||
1925 | p = strdup(path); | |
1926 | if (!p) | |
1927 | return -ENOMEM; | |
1928 | ||
1929 | if (!isempty(options)) { | |
1930 | o = strdup(options); | |
1931 | if (!o) | |
1932 | return -ENOMEM; | |
1933 | } | |
1934 | ||
aa484f35 | 1935 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
1936 | if (!c) |
1937 | return -ENOMEM; | |
1938 | ||
1939 | *t = c; | |
1940 | ||
1941 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
1942 | .path = TAKE_PTR(p), |
1943 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
1944 | }; |
1945 | ||
2abd4e38 YW |
1946 | return 0; |
1947 | } | |
1948 | ||
a652f050 JR |
1949 | static int make_tmp_prefix(const char *prefix) { |
1950 | _cleanup_free_ char *t = NULL; | |
1951 | int r; | |
1952 | ||
1953 | /* Don't do anything unless we know the dir is actually missing */ | |
1954 | r = access(prefix, F_OK); | |
1955 | if (r >= 0) | |
1956 | return 0; | |
1957 | if (errno != ENOENT) | |
1958 | return -errno; | |
1959 | ||
1960 | r = mkdir_parents(prefix, 0755); | |
1961 | if (r < 0) | |
1962 | return r; | |
1963 | ||
1964 | r = tempfn_random(prefix, NULL, &t); | |
1965 | if (r < 0) | |
1966 | return r; | |
1967 | ||
1968 | if (mkdir(t, 0777) < 0) | |
1969 | return -errno; | |
1970 | ||
1971 | if (chmod(t, 01777) < 0) { | |
1972 | r = -errno; | |
1973 | (void) rmdir(t); | |
1974 | return r; | |
1975 | } | |
1976 | ||
1977 | if (rename(t, prefix) < 0) { | |
1978 | r = -errno; | |
1979 | (void) rmdir(t); | |
1980 | return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */ | |
1981 | } | |
1982 | ||
1983 | return 0; | |
1984 | ||
1985 | } | |
1986 | ||
56a13a49 ZJS |
1987 | static int make_tmp_subdir(const char *parent, char **ret) { |
1988 | _cleanup_free_ char *y = NULL; | |
1989 | ||
3f181262 LP |
1990 | y = path_join(parent, "/tmp"); |
1991 | if (!y) | |
1992 | return -ENOMEM; | |
56a13a49 | 1993 | |
3f181262 | 1994 | RUN_WITH_UMASK(0000) { |
56a13a49 ZJS |
1995 | if (mkdir(y, 0777 | S_ISVTX) < 0) |
1996 | return -errno; | |
1997 | } | |
1998 | ||
1999 | if (ret) | |
2000 | *ret = TAKE_PTR(y); | |
2001 | return 0; | |
2002 | } | |
2003 | ||
2004 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path, char **tmp_path) { | |
613b411c | 2005 | _cleanup_free_ char *x = NULL; |
6b46ea73 LP |
2006 | char bid[SD_ID128_STRING_MAX]; |
2007 | sd_id128_t boot_id; | |
56a13a49 | 2008 | bool rw = true; |
6b46ea73 | 2009 | int r; |
613b411c LP |
2010 | |
2011 | assert(id); | |
2012 | assert(prefix); | |
2013 | assert(path); | |
2014 | ||
6b46ea73 LP |
2015 | /* We include the boot id in the directory so that after a |
2016 | * reboot we can easily identify obsolete directories. */ | |
2017 | ||
2018 | r = sd_id128_get_boot(&boot_id); | |
2019 | if (r < 0) | |
2020 | return r; | |
2021 | ||
605405c6 | 2022 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
2023 | if (!x) |
2024 | return -ENOMEM; | |
2025 | ||
a652f050 JR |
2026 | r = make_tmp_prefix(prefix); |
2027 | if (r < 0) | |
2028 | return r; | |
2029 | ||
613b411c | 2030 | RUN_WITH_UMASK(0077) |
56a13a49 ZJS |
2031 | if (!mkdtemp(x)) { |
2032 | if (errno == EROFS || ERRNO_IS_DISK_SPACE(errno)) | |
2033 | rw = false; | |
2034 | else | |
2035 | return -errno; | |
2036 | } | |
613b411c | 2037 | |
56a13a49 ZJS |
2038 | if (rw) { |
2039 | r = make_tmp_subdir(x, tmp_path); | |
2040 | if (r < 0) | |
2041 | return r; | |
2042 | } else { | |
2043 | /* Trouble: we failed to create the directory. Instead of failing, let's simulate /tmp being | |
2044 | * read-only. This way the service will get the EROFS result as if it was writing to the real | |
2045 | * file system. */ | |
2046 | r = mkdir_p(RUN_SYSTEMD_EMPTY, 0500); | |
2047 | if (r < 0) | |
2048 | return r; | |
613b411c | 2049 | |
3f181262 LP |
2050 | r = free_and_strdup(&x, RUN_SYSTEMD_EMPTY); |
2051 | if (r < 0) | |
2052 | return r; | |
c17ec25e | 2053 | } |
15ae422b | 2054 | |
1cc6c93a | 2055 | *path = TAKE_PTR(x); |
613b411c LP |
2056 | return 0; |
2057 | } | |
2058 | ||
2059 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
56a13a49 ZJS |
2060 | _cleanup_(namespace_cleanup_tmpdirp) char *a = NULL; |
2061 | _cleanup_(rmdir_and_freep) char *a_tmp = NULL; | |
2062 | char *b; | |
613b411c LP |
2063 | int r; |
2064 | ||
2065 | assert(id); | |
2066 | assert(tmp_dir); | |
2067 | assert(var_tmp_dir); | |
2068 | ||
56a13a49 | 2069 | r = setup_one_tmp_dir(id, "/tmp", &a, &a_tmp); |
613b411c LP |
2070 | if (r < 0) |
2071 | return r; | |
2072 | ||
56a13a49 ZJS |
2073 | r = setup_one_tmp_dir(id, "/var/tmp", &b, NULL); |
2074 | if (r < 0) | |
613b411c | 2075 | return r; |
613b411c | 2076 | |
56a13a49 ZJS |
2077 | a_tmp = mfree(a_tmp); /* avoid rmdir */ |
2078 | *tmp_dir = TAKE_PTR(a); | |
2079 | *var_tmp_dir = TAKE_PTR(b); | |
613b411c LP |
2080 | |
2081 | return 0; | |
2082 | } | |
2083 | ||
2caa38e9 | 2084 | int setup_netns(const int netns_storage_socket[static 2]) { |
613b411c | 2085 | _cleanup_close_ int netns = -1; |
3ee897d6 | 2086 | int r, q; |
613b411c LP |
2087 | |
2088 | assert(netns_storage_socket); | |
2089 | assert(netns_storage_socket[0] >= 0); | |
2090 | assert(netns_storage_socket[1] >= 0); | |
2091 | ||
2092 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
2093 | * namespace reference fd. Whatever process runs this first |
2094 | * shall create a new namespace, all others should just join | |
2095 | * it. To serialize that we use a file lock on the socket | |
2096 | * pair. | |
613b411c LP |
2097 | * |
2098 | * It's a bit crazy, but hey, works great! */ | |
2099 | ||
2100 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2101 | return -errno; | |
2102 | ||
3ee897d6 LP |
2103 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
2104 | if (netns == -EAGAIN) { | |
44ffcbae | 2105 | /* Nothing stored yet, so let's create a new namespace. */ |
613b411c LP |
2106 | |
2107 | if (unshare(CLONE_NEWNET) < 0) { | |
2108 | r = -errno; | |
2109 | goto fail; | |
2110 | } | |
2111 | ||
44ffcbae | 2112 | (void) loopback_setup(); |
613b411c LP |
2113 | |
2114 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
2115 | if (netns < 0) { | |
2116 | r = -errno; | |
2117 | goto fail; | |
2118 | } | |
2119 | ||
2120 | r = 1; | |
613b411c | 2121 | |
3ee897d6 LP |
2122 | } else if (netns < 0) { |
2123 | r = netns; | |
2124 | goto fail; | |
613b411c | 2125 | |
3ee897d6 LP |
2126 | } else { |
2127 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
2128 | if (setns(netns, CLONE_NEWNET) < 0) { |
2129 | r = -errno; | |
2130 | goto fail; | |
2131 | } | |
2132 | ||
2133 | r = 0; | |
2134 | } | |
2135 | ||
3ee897d6 LP |
2136 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
2137 | if (q < 0) { | |
2138 | r = q; | |
613b411c LP |
2139 | goto fail; |
2140 | } | |
2141 | ||
2142 | fail: | |
fe048ce5 | 2143 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
2144 | return r; |
2145 | } | |
417116f2 | 2146 | |
2caa38e9 | 2147 | int open_netns_path(const int netns_storage_socket[static 2], const char *path) { |
51af7fb2 LP |
2148 | _cleanup_close_ int netns = -1; |
2149 | int q, r; | |
2150 | ||
2151 | assert(netns_storage_socket); | |
2152 | assert(netns_storage_socket[0] >= 0); | |
2153 | assert(netns_storage_socket[1] >= 0); | |
2154 | assert(path); | |
2155 | ||
2156 | /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in | |
2157 | * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a | |
2158 | * new anonymous netns if needed. */ | |
2159 | ||
2160 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2161 | return -errno; | |
2162 | ||
2163 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); | |
2164 | if (netns == -EAGAIN) { | |
2165 | /* Nothing stored yet. Open the file from the file system. */ | |
2166 | ||
2167 | netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
2168 | if (netns < 0) { | |
2169 | r = -errno; | |
2170 | goto fail; | |
2171 | } | |
2172 | ||
2173 | r = fd_is_network_ns(netns); | |
2174 | if (r == 0) { /* Not a netns? Refuse early. */ | |
2175 | r = -EINVAL; | |
2176 | goto fail; | |
2177 | } | |
2178 | if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */ | |
2179 | goto fail; | |
2180 | ||
2181 | r = 1; | |
2182 | ||
2183 | } else if (netns < 0) { | |
2184 | r = netns; | |
2185 | goto fail; | |
2186 | } else | |
2187 | r = 0; /* Already allocated */ | |
2188 | ||
2189 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); | |
2190 | if (q < 0) { | |
2191 | r = q; | |
2192 | goto fail; | |
2193 | } | |
2194 | ||
2195 | fail: | |
2196 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); | |
2197 | return r; | |
2198 | } | |
2199 | ||
6e2d7c4f MS |
2200 | bool ns_type_supported(NamespaceType type) { |
2201 | const char *t, *ns_proc; | |
2202 | ||
0fa5b831 LP |
2203 | t = namespace_type_to_string(type); |
2204 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
2205 | return false; |
2206 | ||
6e2d7c4f | 2207 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
2208 | return access(ns_proc, F_OK) == 0; |
2209 | } | |
2210 | ||
1b8689f9 | 2211 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
cbc056c8 ZJS |
2212 | [PROTECT_HOME_NO] = "no", |
2213 | [PROTECT_HOME_YES] = "yes", | |
1b8689f9 | 2214 | [PROTECT_HOME_READ_ONLY] = "read-only", |
cbc056c8 | 2215 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
2216 | }; |
2217 | ||
1e8c7bd5 | 2218 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 2219 | |
1b8689f9 | 2220 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
cbc056c8 ZJS |
2221 | [PROTECT_SYSTEM_NO] = "no", |
2222 | [PROTECT_SYSTEM_YES] = "yes", | |
2223 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 2224 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
2225 | }; |
2226 | ||
1e8c7bd5 | 2227 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 2228 | |
6e2d7c4f | 2229 | static const char* const namespace_type_table[] = { |
cbc056c8 | 2230 | [NAMESPACE_MOUNT] = "mnt", |
6e2d7c4f | 2231 | [NAMESPACE_CGROUP] = "cgroup", |
cbc056c8 ZJS |
2232 | [NAMESPACE_UTS] = "uts", |
2233 | [NAMESPACE_IPC] = "ipc", | |
2234 | [NAMESPACE_USER] = "user", | |
2235 | [NAMESPACE_PID] = "pid", | |
2236 | [NAMESPACE_NET] = "net", | |
6e2d7c4f MS |
2237 | }; |
2238 | ||
2239 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); |