]>
Commit | Line | Data |
---|---|---|
db9ecf05 | 1 | /* SPDX-License-Identifier: LGPL-2.1-or-later */ |
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, | |
94293d65 | 54 | RUN, |
5d997827 | 55 | READONLY, |
59eeb84b | 56 | READWRITE, |
2abd4e38 | 57 | TMPFS, |
1e05071d | 58 | READWRITE_IMPLICIT, /* Should have the lowest priority. */ |
5beb8688 | 59 | _MOUNT_MODE_MAX, |
c17ec25e | 60 | } MountMode; |
15ae422b | 61 | |
34de407a | 62 | typedef struct MountEntry { |
5327c910 | 63 | const char *path_const; /* Memory allocated on stack or static */ |
cfbeb4ef | 64 | MountMode mode:5; |
5327c910 LP |
65 | bool ignore:1; /* Ignore if path does not exist? */ |
66 | bool has_prefix:1; /* Already is prefixed by the root dir? */ | |
cfbeb4ef | 67 | bool read_only:1; /* Shall this mount point be read-only? */ |
9ce4e4b0 | 68 | bool nosuid:1; /* Shall set MS_NOSUID on the mount itself */ |
088696fe | 69 | bool applied:1; /* Already applied */ |
55fe7432 | 70 | char *path_malloc; /* Use this instead of 'path_const' if we had to allocate memory */ |
b3d13314 | 71 | const char *source_const; /* The source path, for bind mounts or images */ |
d2d6c096 | 72 | char *source_malloc; |
2abd4e38 YW |
73 | const char *options_const;/* Mount options for tmpfs */ |
74 | char *options_malloc; | |
75 | unsigned long flags; /* Mount flags used by EMPTY_DIR and TMPFS. Do not include MS_RDONLY here, but please use read_only. */ | |
088696fe | 76 | unsigned n_followed; |
427353f6 | 77 | LIST_HEAD(MountOptions, image_options); |
34de407a | 78 | } MountEntry; |
15ae422b | 79 | |
94293d65 | 80 | /* If MountAPIVFS= is used, let's mount /sys, /proc, /dev and /run into the it, but only as a fallback if the user hasn't mounted |
3fe91079 | 81 | * something there already. These mounts are hence overridden by any other explicitly configured mounts. */ |
5d997827 LP |
82 | static const MountEntry apivfs_table[] = { |
83 | { "/proc", PROCFS, false }, | |
84 | { "/dev", BIND_DEV, false }, | |
85 | { "/sys", SYSFS, false }, | |
94293d65 | 86 | { "/run", RUN, false, .options_const = "mode=755" TMPFS_LIMITS_RUN, .flags = MS_NOSUID|MS_NODEV|MS_STRICTATIME }, |
5d997827 | 87 | }; |
f471b2af | 88 | |
11a30cec | 89 | /* ProtectKernelTunables= option and the related filesystem APIs */ |
34de407a | 90 | static const MountEntry protect_kernel_tunables_table[] = { |
1e05071d YW |
91 | { "/proc/acpi", READONLY, true }, |
92 | { "/proc/apm", READONLY, true }, /* Obsolete API, there's no point in permitting access to this, ever */ | |
93 | { "/proc/asound", READONLY, true }, | |
94 | { "/proc/bus", READONLY, true }, | |
95 | { "/proc/fs", READONLY, true }, | |
96 | { "/proc/irq", READONLY, true }, | |
97 | { "/proc/kallsyms", INACCESSIBLE, true }, | |
98 | { "/proc/kcore", INACCESSIBLE, true }, | |
99 | { "/proc/latency_stats", READONLY, true }, | |
100 | { "/proc/mtrr", READONLY, true }, | |
101 | { "/proc/scsi", READONLY, true }, | |
4e399953 | 102 | { "/proc/sys", READONLY, true }, |
1e05071d YW |
103 | { "/proc/sysrq-trigger", READONLY, true }, |
104 | { "/proc/timer_stats", READONLY, true }, | |
105 | { "/sys", READONLY, false }, | |
106 | { "/sys/fs/bpf", READONLY, true }, | |
107 | { "/sys/fs/cgroup", READWRITE_IMPLICIT, false }, /* READONLY is set by ProtectControlGroups= option */ | |
108 | { "/sys/fs/selinux", READWRITE_IMPLICIT, true }, | |
109 | { "/sys/kernel/debug", READONLY, true }, | |
110 | { "/sys/kernel/tracing", READONLY, true }, | |
11a30cec DH |
111 | }; |
112 | ||
c575770b | 113 | /* ProtectKernelModules= option */ |
34de407a | 114 | static const MountEntry protect_kernel_modules_table[] = { |
349cc4a5 | 115 | #if HAVE_SPLIT_USR |
c6232fb0 | 116 | { "/lib/modules", INACCESSIBLE, true }, |
c575770b | 117 | #endif |
c6232fb0 | 118 | { "/usr/lib/modules", INACCESSIBLE, true }, |
c575770b DH |
119 | }; |
120 | ||
94a7b275 KK |
121 | /* ProtectKernelLogs= option */ |
122 | static const MountEntry protect_kernel_logs_table[] = { | |
123 | { "/proc/kmsg", INACCESSIBLE, true }, | |
124 | { "/dev/kmsg", INACCESSIBLE, true }, | |
125 | }; | |
126 | ||
b6c432ca DH |
127 | /* |
128 | * ProtectHome=read-only table, protect $HOME and $XDG_RUNTIME_DIR and rest of | |
129 | * system should be protected by ProtectSystem= | |
130 | */ | |
34de407a | 131 | static const MountEntry protect_home_read_only_table[] = { |
c6232fb0 LP |
132 | { "/home", READONLY, true }, |
133 | { "/run/user", READONLY, true }, | |
134 | { "/root", READONLY, true }, | |
b6c432ca DH |
135 | }; |
136 | ||
e4da7d8c YW |
137 | /* ProtectHome=tmpfs table */ |
138 | static const MountEntry protect_home_tmpfs_table[] = { | |
7d85383e TM |
139 | { "/home", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, |
140 | { "/run/user", TMPFS, true, .read_only = true, .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
141 | { "/root", TMPFS, true, .read_only = true, .options_const = "mode=0700" TMPFS_LIMITS_EMPTY_OR_ALMOST, .flags = MS_NODEV|MS_STRICTATIME }, | |
e4da7d8c YW |
142 | }; |
143 | ||
b6c432ca | 144 | /* ProtectHome=yes table */ |
34de407a | 145 | static const MountEntry protect_home_yes_table[] = { |
c6232fb0 LP |
146 | { "/home", INACCESSIBLE, true }, |
147 | { "/run/user", INACCESSIBLE, true }, | |
148 | { "/root", INACCESSIBLE, true }, | |
b6c432ca DH |
149 | }; |
150 | ||
f471b2af | 151 | /* ProtectSystem=yes table */ |
34de407a | 152 | static const MountEntry protect_system_yes_table[] = { |
c6232fb0 LP |
153 | { "/usr", READONLY, false }, |
154 | { "/boot", READONLY, true }, | |
155 | { "/efi", READONLY, true }, | |
7486f305 AB |
156 | #if HAVE_SPLIT_USR |
157 | { "/lib", READONLY, true }, | |
158 | { "/lib64", READONLY, true }, | |
159 | { "/bin", READONLY, true }, | |
671f0f8d | 160 | # if HAVE_SPLIT_BIN |
7486f305 | 161 | { "/sbin", READONLY, true }, |
671f0f8d | 162 | # endif |
7486f305 | 163 | #endif |
f471b2af DH |
164 | }; |
165 | ||
166 | /* ProtectSystem=full includes ProtectSystem=yes */ | |
34de407a | 167 | static const MountEntry protect_system_full_table[] = { |
c6232fb0 LP |
168 | { "/usr", READONLY, false }, |
169 | { "/boot", READONLY, true }, | |
170 | { "/efi", READONLY, true }, | |
171 | { "/etc", READONLY, false }, | |
7486f305 AB |
172 | #if HAVE_SPLIT_USR |
173 | { "/lib", READONLY, true }, | |
174 | { "/lib64", READONLY, true }, | |
175 | { "/bin", READONLY, true }, | |
671f0f8d | 176 | # if HAVE_SPLIT_BIN |
7486f305 | 177 | { "/sbin", READONLY, true }, |
671f0f8d | 178 | # endif |
7486f305 | 179 | #endif |
f471b2af DH |
180 | }; |
181 | ||
182 | /* | |
183 | * ProtectSystem=strict table. In this strict mode, we mount everything | |
184 | * read-only, except for /proc, /dev, /sys which are the kernel API VFS, | |
185 | * which are left writable, but PrivateDevices= + ProtectKernelTunables= | |
186 | * protect those, and these options should be fully orthogonal. | |
187 | * (And of course /home and friends are also left writable, as ProtectHome= | |
188 | * shall manage those, orthogonally). | |
189 | */ | |
34de407a | 190 | static const MountEntry protect_system_strict_table[] = { |
1e05071d YW |
191 | { "/", READONLY, false }, |
192 | { "/proc", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
193 | { "/sys", READWRITE_IMPLICIT, false }, /* ProtectKernelTunables= */ | |
194 | { "/dev", READWRITE_IMPLICIT, false }, /* PrivateDevices= */ | |
195 | { "/home", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
196 | { "/run/user", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
197 | { "/root", READWRITE_IMPLICIT, true }, /* ProtectHome= */ | |
f471b2af DH |
198 | }; |
199 | ||
5beb8688 YW |
200 | static const char * const mount_mode_table[_MOUNT_MODE_MAX] = { |
201 | [INACCESSIBLE] = "inaccessible", | |
202 | [BIND_MOUNT] = "bind", | |
203 | [BIND_MOUNT_RECURSIVE] = "rbind", | |
204 | [PRIVATE_TMP] = "private-tmp", | |
205 | [PRIVATE_DEV] = "private-dev", | |
206 | [BIND_DEV] = "bind-dev", | |
207 | [EMPTY_DIR] = "empty", | |
208 | [SYSFS] = "sysfs", | |
209 | [PROCFS] = "procfs", | |
210 | [READONLY] = "read-only", | |
211 | [READWRITE] = "read-write", | |
212 | [TMPFS] = "tmpfs", | |
b3d13314 | 213 | [MOUNT_IMAGES] = "mount-images", |
5beb8688 YW |
214 | [READWRITE_IMPLICIT] = "rw-implicit", |
215 | }; | |
216 | ||
217 | DEFINE_PRIVATE_STRING_TABLE_LOOKUP_TO_STRING(mount_mode, MountMode); | |
218 | ||
34de407a | 219 | static const char *mount_entry_path(const MountEntry *p) { |
f0a4feb0 DH |
220 | assert(p); |
221 | ||
5327c910 LP |
222 | /* Returns the path of this bind mount. If the malloc()-allocated ->path_buffer field is set we return that, |
223 | * otherwise the stack/static ->path field is returned. */ | |
f0a4feb0 | 224 | |
5327c910 | 225 | return p->path_malloc ?: p->path_const; |
f0a4feb0 DH |
226 | } |
227 | ||
34de407a | 228 | static bool mount_entry_read_only(const MountEntry *p) { |
cfbeb4ef LP |
229 | assert(p); |
230 | ||
56a13a49 | 231 | return p->read_only || IN_SET(p->mode, READONLY, INACCESSIBLE, PRIVATE_TMP_READONLY); |
cfbeb4ef LP |
232 | } |
233 | ||
d2d6c096 LP |
234 | static const char *mount_entry_source(const MountEntry *p) { |
235 | assert(p); | |
236 | ||
237 | return p->source_malloc ?: p->source_const; | |
238 | } | |
239 | ||
2abd4e38 YW |
240 | static const char *mount_entry_options(const MountEntry *p) { |
241 | assert(p); | |
242 | ||
243 | return p->options_malloc ?: p->options_const; | |
244 | } | |
245 | ||
1eb7e08e LP |
246 | static void mount_entry_done(MountEntry *p) { |
247 | assert(p); | |
248 | ||
249 | p->path_malloc = mfree(p->path_malloc); | |
250 | p->source_malloc = mfree(p->source_malloc); | |
2abd4e38 | 251 | p->options_malloc = mfree(p->options_malloc); |
427353f6 | 252 | p->image_options = mount_options_free_all(p->image_options); |
1eb7e08e LP |
253 | } |
254 | ||
d18aff04 | 255 | static int append_access_mounts(MountEntry **p, char **strv, MountMode mode, bool forcibly_require_prefix) { |
15ae422b LP |
256 | char **i; |
257 | ||
613b411c LP |
258 | assert(p); |
259 | ||
1e05071d | 260 | /* Adds a list of user-supplied READWRITE/READWRITE_IMPLICIT/READONLY/INACCESSIBLE entries */ |
5327c910 | 261 | |
15ae422b | 262 | STRV_FOREACH(i, strv) { |
5327c910 LP |
263 | bool ignore = false, needs_prefix = false; |
264 | const char *e = *i; | |
15ae422b | 265 | |
5327c910 LP |
266 | /* Look for any prefixes */ |
267 | if (startswith(e, "-")) { | |
268 | e++; | |
9c94d52e | 269 | ignore = true; |
ea92ae33 | 270 | } |
5327c910 LP |
271 | if (startswith(e, "+")) { |
272 | e++; | |
273 | needs_prefix = true; | |
274 | } | |
ea92ae33 | 275 | |
baaa35ad ZJS |
276 | if (!path_is_absolute(e)) |
277 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
278 | "Path is not absolute: %s", e); | |
15ae422b | 279 | |
34de407a | 280 | *((*p)++) = (MountEntry) { |
5327c910 LP |
281 | .path_const = e, |
282 | .mode = mode, | |
283 | .ignore = ignore, | |
d18aff04 | 284 | .has_prefix = !needs_prefix && !forcibly_require_prefix, |
5327c910 | 285 | }; |
15ae422b LP |
286 | } |
287 | ||
288 | return 0; | |
289 | } | |
290 | ||
6c47cd7d LP |
291 | static int append_empty_dir_mounts(MountEntry **p, char **strv) { |
292 | char **i; | |
293 | ||
294 | assert(p); | |
295 | ||
296 | /* Adds tmpfs mounts to provide readable but empty directories. This is primarily used to implement the | |
297 | * "/private/" boundary directories for DynamicUser=1. */ | |
298 | ||
299 | STRV_FOREACH(i, strv) { | |
300 | ||
301 | *((*p)++) = (MountEntry) { | |
302 | .path_const = *i, | |
303 | .mode = EMPTY_DIR, | |
304 | .ignore = false, | |
6c47cd7d | 305 | .read_only = true, |
7d85383e | 306 | .options_const = "mode=755" TMPFS_LIMITS_EMPTY_OR_ALMOST, |
2abd4e38 | 307 | .flags = MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, |
6c47cd7d LP |
308 | }; |
309 | } | |
310 | ||
311 | return 0; | |
312 | } | |
313 | ||
da6053d0 LP |
314 | static int append_bind_mounts(MountEntry **p, const BindMount *binds, size_t n) { |
315 | size_t i; | |
d2d6c096 LP |
316 | |
317 | assert(p); | |
318 | ||
319 | for (i = 0; i < n; i++) { | |
320 | const BindMount *b = binds + i; | |
321 | ||
322 | *((*p)++) = (MountEntry) { | |
323 | .path_const = b->destination, | |
324 | .mode = b->recursive ? BIND_MOUNT_RECURSIVE : BIND_MOUNT, | |
325 | .read_only = b->read_only, | |
9ce4e4b0 | 326 | .nosuid = b->nosuid, |
d2d6c096 | 327 | .source_const = b->source, |
4ca763a9 | 328 | .ignore = b->ignore_enoent, |
d2d6c096 LP |
329 | }; |
330 | } | |
331 | ||
332 | return 0; | |
333 | } | |
334 | ||
b3d13314 LB |
335 | static int append_mount_images(MountEntry **p, const MountImage *mount_images, size_t n) { |
336 | assert(p); | |
337 | ||
338 | for (size_t i = 0; i < n; i++) { | |
339 | const MountImage *m = mount_images + i; | |
340 | ||
341 | *((*p)++) = (MountEntry) { | |
342 | .path_const = m->destination, | |
343 | .mode = MOUNT_IMAGES, | |
344 | .source_const = m->source, | |
427353f6 | 345 | .image_options = m->mount_options, |
b3d13314 LB |
346 | .ignore = m->ignore_enoent, |
347 | }; | |
348 | } | |
349 | ||
350 | return 0; | |
351 | } | |
352 | ||
da6053d0 | 353 | static int append_tmpfs_mounts(MountEntry **p, const TemporaryFileSystem *tmpfs, size_t n) { |
2abd4e38 YW |
354 | assert(p); |
355 | ||
b67ec8e5 | 356 | for (size_t i = 0; i < n; i++) { |
2abd4e38 YW |
357 | const TemporaryFileSystem *t = tmpfs + i; |
358 | _cleanup_free_ char *o = NULL, *str = NULL; | |
ad8e66dc | 359 | unsigned long flags; |
2abd4e38 | 360 | bool ro = false; |
b67ec8e5 | 361 | int r; |
2abd4e38 | 362 | |
baaa35ad ZJS |
363 | if (!path_is_absolute(t->path)) |
364 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
365 | "Path is not absolute: %s", | |
366 | t->path); | |
2abd4e38 | 367 | |
b67ec8e5 | 368 | str = strjoin("mode=0755" NESTED_TMPFS_LIMITS ",", t->options); |
ad8e66dc AJ |
369 | if (!str) |
370 | return -ENOMEM; | |
2abd4e38 | 371 | |
ad8e66dc AJ |
372 | r = mount_option_mangle(str, MS_NODEV|MS_STRICTATIME, &flags, &o); |
373 | if (r < 0) | |
374 | return log_debug_errno(r, "Failed to parse mount option '%s': %m", str); | |
2abd4e38 | 375 | |
ad8e66dc AJ |
376 | ro = flags & MS_RDONLY; |
377 | if (ro) | |
378 | flags ^= MS_RDONLY; | |
2abd4e38 YW |
379 | |
380 | *((*p)++) = (MountEntry) { | |
381 | .path_const = t->path, | |
382 | .mode = TMPFS, | |
383 | .read_only = ro, | |
ad8e66dc | 384 | .options_malloc = TAKE_PTR(o), |
2abd4e38 YW |
385 | .flags = flags, |
386 | }; | |
2abd4e38 YW |
387 | } |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
da6053d0 LP |
392 | static int append_static_mounts(MountEntry **p, const MountEntry *mounts, size_t n, bool ignore_protect) { |
393 | size_t i; | |
11a30cec DH |
394 | |
395 | assert(p); | |
f471b2af | 396 | assert(mounts); |
11a30cec | 397 | |
5327c910 | 398 | /* Adds a list of static pre-defined entries */ |
f471b2af | 399 | |
5327c910 | 400 | for (i = 0; i < n; i++) |
34de407a LP |
401 | *((*p)++) = (MountEntry) { |
402 | .path_const = mount_entry_path(mounts+i), | |
5327c910 LP |
403 | .mode = mounts[i].mode, |
404 | .ignore = mounts[i].ignore || ignore_protect, | |
405 | }; | |
f471b2af DH |
406 | |
407 | return 0; | |
408 | } | |
409 | ||
34de407a | 410 | static int append_protect_home(MountEntry **p, ProtectHome protect_home, bool ignore_protect) { |
c575770b DH |
411 | assert(p); |
412 | ||
5327c910 | 413 | switch (protect_home) { |
b6c432ca | 414 | |
5327c910 | 415 | case PROTECT_HOME_NO: |
b6c432ca DH |
416 | return 0; |
417 | ||
b6c432ca | 418 | case PROTECT_HOME_READ_ONLY: |
5327c910 LP |
419 | return append_static_mounts(p, protect_home_read_only_table, ELEMENTSOF(protect_home_read_only_table), ignore_protect); |
420 | ||
e4da7d8c YW |
421 | case PROTECT_HOME_TMPFS: |
422 | return append_static_mounts(p, protect_home_tmpfs_table, ELEMENTSOF(protect_home_tmpfs_table), ignore_protect); | |
423 | ||
b6c432ca | 424 | case PROTECT_HOME_YES: |
5327c910 LP |
425 | return append_static_mounts(p, protect_home_yes_table, ELEMENTSOF(protect_home_yes_table), ignore_protect); |
426 | ||
b6c432ca | 427 | default: |
5327c910 | 428 | assert_not_reached("Unexpected ProtectHome= value"); |
b6c432ca | 429 | } |
b6c432ca DH |
430 | } |
431 | ||
34de407a | 432 | static int append_protect_system(MountEntry **p, ProtectSystem protect_system, bool ignore_protect) { |
f471b2af DH |
433 | assert(p); |
434 | ||
5327c910 LP |
435 | switch (protect_system) { |
436 | ||
437 | case PROTECT_SYSTEM_NO: | |
f471b2af DH |
438 | return 0; |
439 | ||
f471b2af | 440 | case PROTECT_SYSTEM_STRICT: |
5327c910 LP |
441 | return append_static_mounts(p, protect_system_strict_table, ELEMENTSOF(protect_system_strict_table), ignore_protect); |
442 | ||
f471b2af | 443 | case PROTECT_SYSTEM_YES: |
5327c910 LP |
444 | return append_static_mounts(p, protect_system_yes_table, ELEMENTSOF(protect_system_yes_table), ignore_protect); |
445 | ||
f471b2af | 446 | case PROTECT_SYSTEM_FULL: |
5327c910 LP |
447 | return append_static_mounts(p, protect_system_full_table, ELEMENTSOF(protect_system_full_table), ignore_protect); |
448 | ||
f471b2af | 449 | default: |
5327c910 | 450 | assert_not_reached("Unexpected ProtectSystem= value"); |
f471b2af | 451 | } |
11a30cec DH |
452 | } |
453 | ||
93bab288 | 454 | static int mount_path_compare(const MountEntry *a, const MountEntry *b) { |
a0827e2b | 455 | int d; |
15ae422b | 456 | |
6ee1a919 | 457 | /* If the paths are not equal, then order prefixes first */ |
93bab288 | 458 | d = path_compare(mount_entry_path(a), mount_entry_path(b)); |
6ee1a919 LP |
459 | if (d != 0) |
460 | return d; | |
15ae422b | 461 | |
6ee1a919 | 462 | /* If the paths are equal, check the mode */ |
93bab288 | 463 | return CMP((int) a->mode, (int) b->mode); |
15ae422b LP |
464 | } |
465 | ||
da6053d0 LP |
466 | static int prefix_where_needed(MountEntry *m, size_t n, const char *root_directory) { |
467 | size_t i; | |
5327c910 | 468 | |
4a756839 | 469 | /* Prefixes all paths in the bind mount table with the root directory if the entry needs that. */ |
5327c910 LP |
470 | |
471 | for (i = 0; i < n; i++) { | |
472 | char *s; | |
473 | ||
474 | if (m[i].has_prefix) | |
475 | continue; | |
476 | ||
c6134d3e | 477 | s = path_join(root_directory, mount_entry_path(m+i)); |
5327c910 LP |
478 | if (!s) |
479 | return -ENOMEM; | |
480 | ||
e282f51f | 481 | free_and_replace(m[i].path_malloc, s); |
5327c910 LP |
482 | m[i].has_prefix = true; |
483 | } | |
484 | ||
485 | return 0; | |
486 | } | |
487 | ||
da6053d0 | 488 | static void drop_duplicates(MountEntry *m, size_t *n) { |
34de407a | 489 | MountEntry *f, *t, *previous; |
15ae422b | 490 | |
c17ec25e | 491 | assert(m); |
15ae422b | 492 | assert(n); |
15ae422b | 493 | |
fe3c2583 LP |
494 | /* Drops duplicate entries. Expects that the array is properly ordered already. */ |
495 | ||
1d54cd5d | 496 | for (f = m, t = m, previous = NULL; f < m + *n; f++) { |
15ae422b | 497 | |
fe3c2583 | 498 | /* The first one wins (which is the one with the more restrictive mode), see mount_path_compare() |
088696fe LP |
499 | * above. Note that we only drop duplicates that haven't been mounted yet. */ |
500 | if (previous && | |
501 | path_equal(mount_entry_path(f), mount_entry_path(previous)) && | |
502 | !f->applied && !previous->applied) { | |
5beb8688 | 503 | log_debug("%s (%s) is duplicate.", mount_entry_path(f), mount_mode_to_string(f->mode)); |
34de407a | 504 | previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */ |
1eb7e08e | 505 | mount_entry_done(f); |
15ae422b | 506 | continue; |
fe3c2583 | 507 | } |
15ae422b | 508 | |
e2d7c1a0 | 509 | *t = *f; |
15ae422b | 510 | previous = t; |
fe3c2583 LP |
511 | t++; |
512 | } | |
513 | ||
514 | *n = t - m; | |
515 | } | |
516 | ||
da6053d0 | 517 | static void drop_inaccessible(MountEntry *m, size_t *n) { |
34de407a | 518 | MountEntry *f, *t; |
fe3c2583 LP |
519 | const char *clear = NULL; |
520 | ||
521 | assert(m); | |
522 | assert(n); | |
523 | ||
524 | /* Drops all entries obstructed by another entry further up the tree. Expects that the array is properly | |
525 | * ordered already. */ | |
526 | ||
1d54cd5d | 527 | for (f = m, t = m; f < m + *n; f++) { |
fe3c2583 LP |
528 | |
529 | /* If we found a path set for INACCESSIBLE earlier, and this entry has it as prefix we should drop | |
530 | * it, as inaccessible paths really should drop the entire subtree. */ | |
34de407a LP |
531 | if (clear && path_startswith(mount_entry_path(f), clear)) { |
532 | log_debug("%s is masked by %s.", mount_entry_path(f), clear); | |
1eb7e08e | 533 | mount_entry_done(f); |
fe3c2583 LP |
534 | continue; |
535 | } | |
15ae422b | 536 | |
34de407a | 537 | clear = f->mode == INACCESSIBLE ? mount_entry_path(f) : NULL; |
fe3c2583 LP |
538 | |
539 | *t = *f; | |
15ae422b LP |
540 | t++; |
541 | } | |
542 | ||
c17ec25e | 543 | *n = t - m; |
15ae422b LP |
544 | } |
545 | ||
da6053d0 | 546 | static void drop_nop(MountEntry *m, size_t *n) { |
34de407a | 547 | MountEntry *f, *t; |
7648a565 LP |
548 | |
549 | assert(m); | |
550 | assert(n); | |
551 | ||
552 | /* Drops all entries which have an immediate parent that has the same type, as they are redundant. Assumes the | |
553 | * list is ordered by prefixes. */ | |
554 | ||
1d54cd5d | 555 | for (f = m, t = m; f < m + *n; f++) { |
7648a565 | 556 | |
1e05071d YW |
557 | /* Only suppress such subtrees for READONLY, READWRITE and READWRITE_IMPLICIT entries */ |
558 | if (IN_SET(f->mode, READONLY, READWRITE, READWRITE_IMPLICIT)) { | |
34de407a | 559 | MountEntry *p; |
7648a565 LP |
560 | bool found = false; |
561 | ||
562 | /* Now let's find the first parent of the entry we are looking at. */ | |
563 | for (p = t-1; p >= m; p--) { | |
34de407a | 564 | if (path_startswith(mount_entry_path(f), mount_entry_path(p))) { |
7648a565 LP |
565 | found = true; |
566 | break; | |
567 | } | |
568 | } | |
569 | ||
570 | /* We found it, let's see if it's the same mode, if so, we can drop this entry */ | |
571 | if (found && p->mode == f->mode) { | |
5beb8688 YW |
572 | log_debug("%s (%s) is made redundant by %s (%s)", |
573 | mount_entry_path(f), mount_mode_to_string(f->mode), | |
574 | mount_entry_path(p), mount_mode_to_string(p->mode)); | |
1eb7e08e | 575 | mount_entry_done(f); |
7648a565 LP |
576 | continue; |
577 | } | |
578 | } | |
579 | ||
580 | *t = *f; | |
581 | t++; | |
582 | } | |
583 | ||
584 | *n = t - m; | |
585 | } | |
586 | ||
da6053d0 | 587 | static void drop_outside_root(const char *root_directory, MountEntry *m, size_t *n) { |
34de407a | 588 | MountEntry *f, *t; |
cd2902c9 LP |
589 | |
590 | assert(m); | |
591 | assert(n); | |
592 | ||
1d54cd5d | 593 | /* Nothing to do */ |
cd2902c9 LP |
594 | if (!root_directory) |
595 | return; | |
596 | ||
597 | /* Drops all mounts that are outside of the root directory. */ | |
598 | ||
1d54cd5d | 599 | for (f = m, t = m; f < m + *n; f++) { |
cd2902c9 | 600 | |
34de407a LP |
601 | if (!path_startswith(mount_entry_path(f), root_directory)) { |
602 | log_debug("%s is outside of root directory.", mount_entry_path(f)); | |
1eb7e08e | 603 | mount_entry_done(f); |
cd2902c9 LP |
604 | continue; |
605 | } | |
606 | ||
607 | *t = *f; | |
608 | t++; | |
609 | } | |
610 | ||
611 | *n = t - m; | |
612 | } | |
613 | ||
b2a60844 LP |
614 | static int clone_device_node( |
615 | const char *d, | |
616 | const char *temporary_mount, | |
617 | bool *make_devnode) { | |
618 | ||
619 | _cleanup_free_ char *sl = NULL; | |
620 | const char *dn, *bn, *t; | |
b5e99f23 ДГ |
621 | struct stat st; |
622 | int r; | |
623 | ||
414b304b | 624 | if (stat(d, &st) < 0) { |
b2a60844 LP |
625 | if (errno == ENOENT) { |
626 | log_debug_errno(errno, "Device node '%s' to clone does not exist, ignoring.", d); | |
af984e13 | 627 | return -ENXIO; |
b2a60844 LP |
628 | } |
629 | ||
630 | return log_debug_errno(errno, "Failed to stat() device node '%s' to clone, ignoring: %m", d); | |
b5e99f23 ДГ |
631 | } |
632 | ||
633 | if (!S_ISBLK(st.st_mode) && | |
baaa35ad ZJS |
634 | !S_ISCHR(st.st_mode)) |
635 | return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), | |
636 | "Device node '%s' to clone is not a device node, ignoring.", | |
637 | d); | |
b5e99f23 | 638 | |
6f7f3a33 | 639 | dn = strjoina(temporary_mount, d); |
b5e99f23 | 640 | |
b2a60844 | 641 | /* First, try to create device node properly */ |
16498617 CB |
642 | if (*make_devnode) { |
643 | mac_selinux_create_file_prepare(d, st.st_mode); | |
644 | r = mknod(dn, st.st_mode, st.st_rdev); | |
645 | mac_selinux_create_file_clear(); | |
b2a60844 LP |
646 | if (r >= 0) |
647 | goto add_symlink; | |
16498617 CB |
648 | if (errno != EPERM) |
649 | return log_debug_errno(errno, "mknod failed for %s: %m", d); | |
650 | ||
b2a60844 | 651 | /* This didn't work, let's not try this again for the next iterations. */ |
16498617 CB |
652 | *make_devnode = false; |
653 | } | |
654 | ||
2aed63f4 | 655 | /* We're about to fall back to bind-mounting the device |
1acf344d CG |
656 | * node. So create a dummy bind-mount target. |
657 | * Do not prepare device-node SELinux label (see issue 13762) */ | |
16498617 | 658 | r = mknod(dn, S_IFREG, 0); |
16498617 | 659 | if (r < 0 && errno != EEXIST) |
b2a60844 | 660 | return log_debug_errno(errno, "mknod() fallback failed for '%s': %m", d); |
16498617 | 661 | |
21935150 LP |
662 | /* Fallback to bind-mounting: The assumption here is that all used device nodes carry standard |
663 | * properties. Specifically, the devices nodes we bind-mount should either be owned by root:root or | |
664 | * root:tty (e.g. /dev/tty, /dev/ptmx) and should not carry ACLs. */ | |
665 | r = mount_nofollow_verbose(LOG_DEBUG, d, dn, NULL, MS_BIND, NULL); | |
666 | if (r < 0) | |
667 | return r; | |
b2a60844 LP |
668 | |
669 | add_symlink: | |
670 | bn = path_startswith(d, "/dev/"); | |
671 | if (!bn) | |
672 | return 0; | |
673 | ||
674 | /* Create symlinks like /dev/char/1:9 → ../urandom */ | |
cbc056c8 ZJS |
675 | if (asprintf(&sl, "%s/dev/%s/%u:%u", |
676 | temporary_mount, | |
677 | S_ISCHR(st.st_mode) ? "char" : "block", | |
678 | major(st.st_rdev), minor(st.st_rdev)) < 0) | |
b2a60844 LP |
679 | return log_oom(); |
680 | ||
681 | (void) mkdir_parents(sl, 0755); | |
682 | ||
683 | t = strjoina("../", bn); | |
b2a60844 | 684 | if (symlink(t, sl) < 0) |
2e4a4fae | 685 | log_debug_errno(errno, "Failed to symlink '%s' to '%s', ignoring: %m", t, sl); |
b5e99f23 | 686 | |
af984e13 | 687 | return 0; |
b5e99f23 ДГ |
688 | } |
689 | ||
5d997827 | 690 | static int mount_private_dev(MountEntry *m) { |
7f112f50 LP |
691 | static const char devnodes[] = |
692 | "/dev/null\0" | |
693 | "/dev/zero\0" | |
694 | "/dev/full\0" | |
695 | "/dev/random\0" | |
696 | "/dev/urandom\0" | |
697 | "/dev/tty\0"; | |
698 | ||
2b85f4e1 | 699 | char temporary_mount[] = "/tmp/namespace-dev-XXXXXX"; |
63cc4c31 | 700 | const char *d, *dev = NULL, *devpts = NULL, *devshm = NULL, *devhugepages = NULL, *devmqueue = NULL, *devlog = NULL, *devptmx = NULL; |
16498617 | 701 | bool can_mknod = true; |
7f112f50 LP |
702 | _cleanup_umask_ mode_t u; |
703 | int r; | |
704 | ||
705 | assert(m); | |
706 | ||
707 | u = umask(0000); | |
708 | ||
2b85f4e1 | 709 | if (!mkdtemp(temporary_mount)) |
2e4a4fae | 710 | return log_debug_errno(errno, "Failed to create temporary directory '%s': %m", temporary_mount); |
2b85f4e1 | 711 | |
63c372cb | 712 | dev = strjoina(temporary_mount, "/dev"); |
dc751688 | 713 | (void) mkdir(dev, 0755); |
21935150 LP |
714 | r = mount_nofollow_verbose(LOG_DEBUG, "tmpfs", dev, "tmpfs", DEV_MOUNT_OPTIONS, "mode=755" TMPFS_LIMITS_DEV); |
715 | if (r < 0) | |
2b85f4e1 | 716 | goto fail; |
21935150 | 717 | |
c3151977 TM |
718 | r = label_fix_container(dev, "/dev", 0); |
719 | if (r < 0) { | |
720 | log_debug_errno(errno, "Failed to fix label of '%s' as /dev: %m", dev); | |
721 | goto fail; | |
722 | } | |
2b85f4e1 | 723 | |
63c372cb | 724 | devpts = strjoina(temporary_mount, "/dev/pts"); |
dc751688 | 725 | (void) mkdir(devpts, 0755); |
21935150 LP |
726 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev/pts", devpts, NULL, MS_BIND, NULL); |
727 | if (r < 0) | |
2b85f4e1 | 728 | goto fail; |
2b85f4e1 | 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); |
21935150 LP |
752 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev/shm", devshm, NULL, MS_BIND, NULL); |
753 | if (r < 0) | |
2b85f4e1 | 754 | goto fail; |
2b85f4e1 | 755 | |
63c372cb | 756 | devmqueue = strjoina(temporary_mount, "/dev/mqueue"); |
dc751688 | 757 | (void) mkdir(devmqueue, 0755); |
21935150 | 758 | (void) mount_nofollow_verbose(LOG_DEBUG, "/dev/mqueue", devmqueue, NULL, MS_BIND, NULL); |
2b85f4e1 | 759 | |
63c372cb | 760 | devhugepages = strjoina(temporary_mount, "/dev/hugepages"); |
dc751688 | 761 | (void) mkdir(devhugepages, 0755); |
21935150 | 762 | (void) mount_nofollow_verbose(LOG_DEBUG, "/dev/hugepages", devhugepages, NULL, MS_BIND, NULL); |
2b85f4e1 | 763 | |
63c372cb | 764 | devlog = strjoina(temporary_mount, "/dev/log"); |
2e4a4fae YW |
765 | if (symlink("/run/systemd/journal/dev-log", devlog) < 0) |
766 | log_debug_errno(errno, "Failed to create a symlink '%s' to /run/systemd/journal/dev-log, ignoring: %m", devlog); | |
82d25240 | 767 | |
7f112f50 | 768 | NULSTR_FOREACH(d, devnodes) { |
16498617 | 769 | r = clone_device_node(d, temporary_mount, &can_mknod); |
37b22b3b | 770 | /* ENXIO means the *source* is not a device file, skip creation in that case */ |
af984e13 | 771 | if (r < 0 && r != -ENXIO) |
2b85f4e1 | 772 | goto fail; |
7f112f50 LP |
773 | } |
774 | ||
2e4a4fae YW |
775 | r = dev_setup(temporary_mount, UID_INVALID, GID_INVALID); |
776 | if (r < 0) | |
105a1a36 | 777 | log_debug_errno(r, "Failed to set up basic device tree at '%s', ignoring: %m", temporary_mount); |
7f112f50 | 778 | |
ee818b89 AC |
779 | /* Create the /dev directory if missing. It is more likely to be |
780 | * missing when the service is started with RootDirectory. This is | |
781 | * consistent with mount units creating the mount points when missing. | |
782 | */ | |
34de407a | 783 | (void) mkdir_p_label(mount_entry_path(m), 0755); |
ee818b89 | 784 | |
9e5f8252 | 785 | /* Unmount everything in old /dev */ |
2e4a4fae YW |
786 | r = umount_recursive(mount_entry_path(m), 0); |
787 | if (r < 0) | |
788 | log_debug_errno(r, "Failed to unmount directories below '%s', ignoring: %m", mount_entry_path(m)); | |
789 | ||
21935150 LP |
790 | r = mount_nofollow_verbose(LOG_DEBUG, dev, mount_entry_path(m), NULL, MS_MOVE, NULL); |
791 | if (r < 0) | |
2b85f4e1 | 792 | goto fail; |
7f112f50 | 793 | |
1019a48f LP |
794 | (void) rmdir(dev); |
795 | (void) rmdir(temporary_mount); | |
7f112f50 | 796 | |
2b85f4e1 | 797 | return 0; |
7f112f50 | 798 | |
2b85f4e1 LP |
799 | fail: |
800 | if (devpts) | |
21935150 | 801 | (void) umount_verbose(LOG_DEBUG, devpts, UMOUNT_NOFOLLOW); |
7f112f50 | 802 | |
2b85f4e1 | 803 | if (devshm) |
21935150 | 804 | (void) umount_verbose(LOG_DEBUG, devshm, UMOUNT_NOFOLLOW); |
7f112f50 | 805 | |
2b85f4e1 | 806 | if (devhugepages) |
21935150 | 807 | (void) umount_verbose(LOG_DEBUG, devhugepages, UMOUNT_NOFOLLOW); |
7f112f50 | 808 | |
2b85f4e1 | 809 | if (devmqueue) |
21935150 | 810 | (void) umount_verbose(LOG_DEBUG, devmqueue, UMOUNT_NOFOLLOW); |
7f112f50 | 811 | |
21935150 | 812 | (void) umount_verbose(LOG_DEBUG, dev, UMOUNT_NOFOLLOW); |
1019a48f LP |
813 | (void) rmdir(dev); |
814 | (void) rmdir(temporary_mount); | |
7f112f50 | 815 | |
2b85f4e1 | 816 | return r; |
7f112f50 LP |
817 | } |
818 | ||
2a2969fd | 819 | static int mount_bind_dev(const MountEntry *m) { |
5d997827 LP |
820 | int r; |
821 | ||
822 | assert(m); | |
823 | ||
824 | /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's | |
825 | * /dev. This is only used when RootDirectory= is set. */ | |
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 /dev is already mounted: %m"); | |
832 | if (r > 0) /* make this a NOP if /dev is already a mount point */ | |
833 | return 0; | |
834 | ||
21935150 LP |
835 | r = mount_nofollow_verbose(LOG_DEBUG, "/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL); |
836 | if (r < 0) | |
837 | return r; | |
5d997827 LP |
838 | |
839 | return 1; | |
840 | } | |
841 | ||
2a2969fd | 842 | static int mount_sysfs(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 /sys is already mounted: %m"); | |
852 | if (r > 0) /* make this a NOP if /sys is already a mount point */ | |
853 | return 0; | |
854 | ||
855 | /* Bind mount the host's version so that we get all child mounts of it, too. */ | |
21935150 LP |
856 | r = mount_nofollow_verbose(LOG_DEBUG, "/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL); |
857 | if (r < 0) | |
858 | return r; | |
5d997827 LP |
859 | |
860 | return 1; | |
861 | } | |
862 | ||
4e399953 | 863 | static int mount_procfs(const MountEntry *m, const NamespaceInfo *ns_info) { |
61f8a7bd | 864 | _cleanup_free_ char *opts = NULL; |
4e399953 | 865 | const char *entry_path; |
61f8a7bd | 866 | int r, n; |
5d997827 LP |
867 | |
868 | assert(m); | |
4e399953 | 869 | assert(ns_info); |
5d997827 | 870 | |
4e399953 LP |
871 | if (ns_info->protect_proc != PROTECT_PROC_DEFAULT || |
872 | ns_info->proc_subset != PROC_SUBSET_ALL) { | |
4e399953 LP |
873 | |
874 | /* Starting with kernel 5.8 procfs' hidepid= logic is truly per-instance (previously it | |
875 | * pretended to be per-instance but actually was per-namespace), hence let's make use of it | |
876 | * if requested. To make sure this logic succeeds only on kernels where hidepid= is | |
877 | * per-instance, we'll exclusively use the textual value for hidepid=, since support was | |
878 | * added in the same commit: if it's supported it is thus also per-instance. */ | |
879 | ||
880 | opts = strjoin("hidepid=", | |
881 | ns_info->protect_proc == PROTECT_PROC_DEFAULT ? "off" : | |
882 | protect_proc_to_string(ns_info->protect_proc), | |
883 | ns_info->proc_subset == PROC_SUBSET_PID ? ",subset=pid" : ""); | |
884 | if (!opts) | |
885 | return -ENOMEM; | |
4e399953 LP |
886 | } |
887 | ||
61f8a7bd YW |
888 | entry_path = mount_entry_path(m); |
889 | (void) mkdir_p_label(entry_path, 0755); | |
890 | ||
891 | /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in | |
892 | * one. i.e we don't reuse existing mounts here under any condition, we want a new instance owned by | |
893 | * our user namespace and with our hidepid= settings applied. Hence, let's get rid of everything | |
894 | * mounted on /proc/ first. */ | |
895 | ||
896 | n = umount_recursive(entry_path, 0); | |
897 | ||
898 | r = mount_nofollow_verbose(LOG_DEBUG, "proc", entry_path, "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, opts); | |
899 | if (r == -EINVAL && opts) | |
900 | /* If this failed with EINVAL then this likely means the textual hidepid= stuff is | |
901 | * not supported by the kernel, and thus the per-instance hidepid= neither, which | |
902 | * means we really don't want to use it, since it would affect our host's /proc | |
903 | * mount. Hence let's gracefully fallback to a classic, unrestricted version. */ | |
904 | r = mount_nofollow_verbose(LOG_DEBUG, "proc", entry_path, "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL); | |
905 | if (r == -EPERM) { | |
24ee0f9d | 906 | /* When we do not have enough privileges to mount /proc, fallback to use existing /proc. */ |
61f8a7bd YW |
907 | |
908 | if (n > 0) | |
909 | /* /proc or some of sub-mounts are umounted in the above. Refuse incomplete tree. | |
910 | * Propagate the original error code returned by mount() in the above. */ | |
911 | return -EPERM; | |
912 | ||
913 | r = path_is_mount_point(entry_path, NULL, 0); | |
914 | if (r < 0) | |
915 | return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m"); | |
916 | if (r == 0) | |
917 | /* /proc is not mounted. Propagate the original error code. */ | |
918 | return -EPERM; | |
ad74f28a YW |
919 | } else if (r < 0) |
920 | return r; | |
5d997827 LP |
921 | |
922 | return 1; | |
923 | } | |
924 | ||
2abd4e38 | 925 | static int mount_tmpfs(const MountEntry *m) { |
df6b900a | 926 | const char *entry_path, *inner_path; |
abad72be | 927 | int r; |
abad72be | 928 | |
6c47cd7d LP |
929 | assert(m); |
930 | ||
df6b900a LP |
931 | entry_path = mount_entry_path(m); |
932 | inner_path = m->path_const; | |
933 | ||
2abd4e38 | 934 | /* First, get rid of everything that is below if there is anything. Then, overmount with our new tmpfs */ |
6c47cd7d | 935 | |
abad72be CG |
936 | (void) mkdir_p_label(entry_path, 0755); |
937 | (void) umount_recursive(entry_path, 0); | |
6c47cd7d | 938 | |
21935150 LP |
939 | r = mount_nofollow_verbose(LOG_DEBUG, "tmpfs", entry_path, "tmpfs", m->flags, mount_entry_options(m)); |
940 | if (r < 0) | |
941 | return r; | |
abad72be | 942 | |
df6b900a | 943 | r = label_fix_container(entry_path, inner_path, 0); |
abad72be | 944 | if (r < 0) |
df6b900a | 945 | return log_debug_errno(r, "Failed to fix label of '%s' as '%s': %m", entry_path, inner_path); |
6c47cd7d LP |
946 | |
947 | return 1; | |
948 | } | |
949 | ||
94293d65 LB |
950 | static int mount_run(const MountEntry *m) { |
951 | int r; | |
952 | ||
953 | assert(m); | |
954 | ||
955 | r = path_is_mount_point(mount_entry_path(m), NULL, 0); | |
956 | if (r < 0 && r != -ENOENT) | |
957 | return log_debug_errno(r, "Unable to determine whether /run is already mounted: %m"); | |
958 | if (r > 0) /* make this a NOP if /run is already a mount point */ | |
959 | return 0; | |
960 | ||
961 | return mount_tmpfs(m); | |
962 | } | |
963 | ||
b3d13314 LB |
964 | static int mount_images(const MountEntry *m) { |
965 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; | |
966 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; | |
967 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; | |
aee36b4e | 968 | _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT; |
89e62e0b | 969 | DissectImageFlags dissect_image_flags; |
b3d13314 LB |
970 | int r; |
971 | ||
89e62e0b LP |
972 | assert(m); |
973 | ||
974 | r = verity_settings_load(&verity, mount_entry_source(m), NULL, NULL); | |
b3d13314 LB |
975 | if (r < 0) |
976 | return log_debug_errno(r, "Failed to load root hash: %m"); | |
b3d13314 | 977 | |
89e62e0b LP |
978 | dissect_image_flags = |
979 | (m->read_only ? DISSECT_IMAGE_READ_ONLY : 0) | | |
980 | (verity.data_path ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0); | |
981 | ||
982 | r = loop_device_make_by_path( | |
983 | mount_entry_source(m), | |
984 | m->read_only ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, | |
985 | verity.data_path ? 0 : LO_FLAGS_PARTSCAN, | |
986 | &loop_device); | |
b3d13314 LB |
987 | if (r < 0) |
988 | return log_debug_errno(r, "Failed to create loop device for image: %m"); | |
989 | ||
89e62e0b LP |
990 | r = dissect_image( |
991 | loop_device->fd, | |
992 | &verity, | |
993 | m->image_options, | |
994 | dissect_image_flags, | |
995 | &dissected_image); | |
b3d13314 | 996 | /* No partition table? Might be a single-filesystem image, try again */ |
89e62e0b LP |
997 | if (!verity.data_path && r == -ENOPKG) |
998 | r = dissect_image( | |
999 | loop_device->fd, | |
1000 | &verity, | |
1001 | m->image_options, | |
1002 | dissect_image_flags|DISSECT_IMAGE_NO_PARTITION_TABLE, | |
1003 | &dissected_image); | |
b3d13314 LB |
1004 | if (r < 0) |
1005 | return log_debug_errno(r, "Failed to dissect image: %m"); | |
1006 | ||
89e62e0b LP |
1007 | r = dissected_image_decrypt( |
1008 | dissected_image, | |
1009 | NULL, | |
1010 | &verity, | |
1011 | dissect_image_flags, | |
1012 | &decrypted_image); | |
b3d13314 LB |
1013 | if (r < 0) |
1014 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); | |
1015 | ||
1016 | r = mkdir_p_label(mount_entry_path(m), 0755); | |
1017 | if (r < 0) | |
1018 | return log_debug_errno(r, "Failed to create destination directory %s: %m", mount_entry_path(m)); | |
1019 | r = umount_recursive(mount_entry_path(m), 0); | |
1020 | if (r < 0) | |
1021 | return log_debug_errno(r, "Failed to umount under destination directory %s: %m", mount_entry_path(m)); | |
1022 | ||
1023 | r = dissected_image_mount(dissected_image, mount_entry_path(m), UID_INVALID, dissect_image_flags); | |
1024 | if (r < 0) | |
1025 | return log_debug_errno(r, "Failed to mount image: %m"); | |
1026 | ||
1027 | if (decrypted_image) { | |
1028 | r = decrypted_image_relinquish(decrypted_image); | |
1029 | if (r < 0) | |
1030 | return log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
1031 | } | |
1032 | ||
1033 | loop_device_relinquish(loop_device); | |
1034 | ||
1035 | return 1; | |
1036 | } | |
1037 | ||
088696fe | 1038 | static int follow_symlink( |
d2d6c096 | 1039 | const char *root_directory, |
088696fe | 1040 | MountEntry *m) { |
d2d6c096 | 1041 | |
088696fe | 1042 | _cleanup_free_ char *target = NULL; |
8fceda93 LP |
1043 | int r; |
1044 | ||
088696fe LP |
1045 | /* Let's chase symlinks, but only one step at a time. That's because depending where the symlink points we |
1046 | * might need to change the order in which we mount stuff. Hence: let's normalize piecemeal, and do one step at | |
1047 | * a time by specifying CHASE_STEP. This function returns 0 if we resolved one step, and > 0 if we reached the | |
1048 | * end and already have a fully normalized name. */ | |
8fceda93 | 1049 | |
a5648b80 | 1050 | r = chase_symlinks(mount_entry_path(m), root_directory, CHASE_STEP|CHASE_NONEXISTENT, &target, NULL); |
088696fe LP |
1051 | if (r < 0) |
1052 | return log_debug_errno(r, "Failed to chase symlinks '%s': %m", mount_entry_path(m)); | |
1053 | if (r > 0) /* Reached the end, nothing more to resolve */ | |
1054 | return 1; | |
8fceda93 | 1055 | |
baaa35ad ZJS |
1056 | if (m->n_followed >= CHASE_SYMLINKS_MAX) /* put a boundary on things */ |
1057 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), | |
1058 | "Symlink loop on '%s'.", | |
1059 | mount_entry_path(m)); | |
8fceda93 | 1060 | |
088696fe | 1061 | log_debug("Followed mount entry path symlink %s → %s.", mount_entry_path(m), target); |
8fceda93 | 1062 | |
088696fe LP |
1063 | free_and_replace(m->path_malloc, target); |
1064 | m->has_prefix = true; | |
8fceda93 | 1065 | |
088696fe LP |
1066 | m->n_followed ++; |
1067 | ||
1068 | return 0; | |
8fceda93 LP |
1069 | } |
1070 | ||
ac0930c8 | 1071 | static int apply_mount( |
8fceda93 | 1072 | const char *root_directory, |
4e399953 LP |
1073 | MountEntry *m, |
1074 | const NamespaceInfo *ns_info) { | |
ac0930c8 | 1075 | |
e5f10caf | 1076 | _cleanup_free_ char *inaccessible = NULL; |
a227a4be | 1077 | bool rbind = true, make = false; |
15ae422b | 1078 | const char *what; |
15ae422b | 1079 | int r; |
15ae422b | 1080 | |
c17ec25e | 1081 | assert(m); |
4e399953 | 1082 | assert(ns_info); |
15ae422b | 1083 | |
34de407a | 1084 | log_debug("Applying namespace mount on %s", mount_entry_path(m)); |
fe3c2583 | 1085 | |
c17ec25e | 1086 | switch (m->mode) { |
15ae422b | 1087 | |
160cfdbe | 1088 | case INACCESSIBLE: { |
e5f10caf AZ |
1089 | _cleanup_free_ char *tmp = NULL; |
1090 | const char *runtime_dir; | |
160cfdbe | 1091 | struct stat target; |
6d313367 LP |
1092 | |
1093 | /* First, get rid of everything that is below if there | |
1094 | * is anything... Then, overmount it with an | |
c4b41707 | 1095 | * inaccessible path. */ |
34de407a | 1096 | (void) umount_recursive(mount_entry_path(m), 0); |
6d313367 | 1097 | |
088696fe LP |
1098 | if (lstat(mount_entry_path(m), &target) < 0) { |
1099 | if (errno == ENOENT && m->ignore) | |
1100 | return 0; | |
1101 | ||
cbc056c8 ZJS |
1102 | return log_debug_errno(errno, "Failed to lstat() %s to determine what to mount over it: %m", |
1103 | mount_entry_path(m)); | |
088696fe | 1104 | } |
15ae422b | 1105 | |
e5f10caf | 1106 | if (geteuid() == 0) |
48b747fa | 1107 | runtime_dir = "/run"; |
e5f10caf | 1108 | else { |
48b747fa LP |
1109 | if (asprintf(&tmp, "/run/user/" UID_FMT, geteuid()) < 0) |
1110 | return -ENOMEM; | |
e5f10caf AZ |
1111 | |
1112 | runtime_dir = tmp; | |
1113 | } | |
1114 | ||
1115 | r = mode_to_inaccessible_node(runtime_dir, target.st_mode, &inaccessible); | |
1116 | if (r < 0) | |
baaa35ad ZJS |
1117 | return log_debug_errno(SYNTHETIC_ERRNO(ELOOP), |
1118 | "File type not supported for inaccessible mounts. Note that symlinks are not allowed"); | |
e5f10caf | 1119 | what = inaccessible; |
c4b41707 | 1120 | break; |
160cfdbe | 1121 | } |
fe3c2583 | 1122 | |
15ae422b | 1123 | case READONLY: |
15ae422b | 1124 | case READWRITE: |
1e05071d | 1125 | case READWRITE_IMPLICIT: |
8fceda93 | 1126 | r = path_is_mount_point(mount_entry_path(m), root_directory, 0); |
088696fe LP |
1127 | if (r == -ENOENT && m->ignore) |
1128 | return 0; | |
d944dc95 | 1129 | if (r < 0) |
cbc056c8 ZJS |
1130 | return log_debug_errno(r, "Failed to determine whether %s is already a mount point: %m", |
1131 | mount_entry_path(m)); | |
1132 | if (r > 0) /* Nothing to do here, it is already a mount. We just later toggle the MS_RDONLY | |
1133 | * bit for the mount point if needed. */ | |
6b7c9f8b | 1134 | return 0; |
6b7c9f8b | 1135 | /* This isn't a mount point yet, let's make it one. */ |
34de407a | 1136 | what = mount_entry_path(m); |
6b7c9f8b | 1137 | break; |
15ae422b | 1138 | |
d2d6c096 LP |
1139 | case BIND_MOUNT: |
1140 | rbind = false; | |
d2d6c096 | 1141 | |
4831981d | 1142 | _fallthrough_; |
088696fe LP |
1143 | case BIND_MOUNT_RECURSIVE: { |
1144 | _cleanup_free_ char *chased = NULL; | |
5d997827 | 1145 | |
cbc056c8 ZJS |
1146 | /* Since mount() will always follow symlinks we chase the symlinks on our own first. Note |
1147 | * that bind mount source paths are always relative to the host root, hence we pass NULL as | |
1148 | * root directory to chase_symlinks() here. */ | |
088696fe | 1149 | |
a5648b80 | 1150 | r = chase_symlinks(mount_entry_source(m), NULL, CHASE_TRAIL_SLASH, &chased, NULL); |
088696fe LP |
1151 | if (r == -ENOENT && m->ignore) { |
1152 | log_debug_errno(r, "Path %s does not exist, ignoring.", mount_entry_source(m)); | |
1153 | return 0; | |
1154 | } | |
1155 | if (r < 0) | |
1156 | return log_debug_errno(r, "Failed to follow symlinks on %s: %m", mount_entry_source(m)); | |
1157 | ||
1158 | log_debug("Followed source symlinks %s → %s.", mount_entry_source(m), chased); | |
1159 | ||
1160 | free_and_replace(m->source_malloc, chased); | |
d2d6c096 LP |
1161 | |
1162 | what = mount_entry_source(m); | |
a227a4be | 1163 | make = true; |
d2d6c096 | 1164 | break; |
088696fe | 1165 | } |
d2d6c096 | 1166 | |
6c47cd7d | 1167 | case EMPTY_DIR: |
2abd4e38 YW |
1168 | case TMPFS: |
1169 | return mount_tmpfs(m); | |
6c47cd7d | 1170 | |
ac0930c8 | 1171 | case PRIVATE_TMP: |
56a13a49 | 1172 | case PRIVATE_TMP_READONLY: |
89bd586c | 1173 | what = mount_entry_source(m); |
a227a4be | 1174 | make = true; |
15ae422b | 1175 | break; |
e364ad06 | 1176 | |
d6797c92 | 1177 | case PRIVATE_DEV: |
5d997827 LP |
1178 | return mount_private_dev(m); |
1179 | ||
1180 | case BIND_DEV: | |
1181 | return mount_bind_dev(m); | |
1182 | ||
1183 | case SYSFS: | |
1184 | return mount_sysfs(m); | |
1185 | ||
1186 | case PROCFS: | |
4e399953 | 1187 | return mount_procfs(m, ns_info); |
d6797c92 | 1188 | |
94293d65 LB |
1189 | case RUN: |
1190 | return mount_run(m); | |
1191 | ||
b3d13314 LB |
1192 | case MOUNT_IMAGES: |
1193 | return mount_images(m); | |
1194 | ||
e364ad06 LP |
1195 | default: |
1196 | assert_not_reached("Unknown mode"); | |
15ae422b LP |
1197 | } |
1198 | ||
ac0930c8 | 1199 | assert(what); |
15ae422b | 1200 | |
21935150 LP |
1201 | r = mount_nofollow_verbose(LOG_DEBUG, what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL); |
1202 | if (r < 0) { | |
a227a4be | 1203 | bool try_again = false; |
a227a4be LP |
1204 | |
1205 | if (r == -ENOENT && make) { | |
8bab8029 | 1206 | int q; |
a227a4be | 1207 | |
cbc056c8 ZJS |
1208 | /* Hmm, either the source or the destination are missing. Let's see if we can create |
1209 | the destination, then try again. */ | |
a227a4be | 1210 | |
8bab8029 | 1211 | (void) mkdir_parents(mount_entry_path(m), 0755); |
a227a4be | 1212 | |
8bab8029 LB |
1213 | q = make_mount_point_inode_from_path(what, mount_entry_path(m), 0755); |
1214 | if (q < 0) | |
1215 | log_error_errno(q, "Failed to create destination mount point node '%s': %m", | |
1216 | mount_entry_path(m)); | |
1217 | else | |
1218 | try_again = true; | |
a227a4be LP |
1219 | } |
1220 | ||
21935150 LP |
1221 | if (try_again) |
1222 | r = mount_nofollow_verbose(LOG_DEBUG, what, mount_entry_path(m), NULL, MS_BIND|(rbind ? MS_REC : 0), NULL); | |
a227a4be | 1223 | if (r < 0) |
5dc60faa | 1224 | return log_error_errno(r, "Failed to mount %s to %s: %m", what, mount_entry_path(m)); |
a227a4be | 1225 | } |
6b7c9f8b | 1226 | |
34de407a | 1227 | log_debug("Successfully mounted %s to %s", what, mount_entry_path(m)); |
6b7c9f8b | 1228 | return 0; |
ac0930c8 | 1229 | } |
15ae422b | 1230 | |
6b000af4 | 1231 | static int make_read_only(const MountEntry *m, char **deny_list, FILE *proc_self_mountinfo) { |
9ce4e4b0 | 1232 | unsigned long new_flags = 0, flags_mask = 0; |
763a260a | 1233 | bool submounts = false; |
6b7c9f8b | 1234 | int r = 0; |
15ae422b | 1235 | |
c17ec25e | 1236 | assert(m); |
ac9de0b3 | 1237 | assert(proc_self_mountinfo); |
ac0930c8 | 1238 | |
9ce4e4b0 LP |
1239 | if (mount_entry_read_only(m) || m->mode == PRIVATE_DEV) { |
1240 | new_flags |= MS_RDONLY; | |
1241 | flags_mask |= MS_RDONLY; | |
1242 | } | |
1243 | ||
1244 | if (m->nosuid) { | |
1245 | new_flags |= MS_NOSUID; | |
1246 | flags_mask |= MS_NOSUID; | |
1247 | } | |
1248 | ||
1249 | if (flags_mask == 0) /* No Change? */ | |
6b7c9f8b LP |
1250 | return 0; |
1251 | ||
9ce4e4b0 LP |
1252 | /* We generally apply these changes recursively, except for /dev, and the cases we know there's |
1253 | * nothing further down. Set /dev readonly, but not submounts like /dev/shm. Also, we only set the | |
1254 | * per-mount read-only flag. We can't set it on the superblock, if we are inside a user namespace | |
1255 | * and running Linux <= 4.17. */ | |
1256 | submounts = | |
1257 | mount_entry_read_only(m) && | |
1258 | !IN_SET(m->mode, EMPTY_DIR, TMPFS); | |
1259 | if (submounts) | |
6b000af4 | 1260 | r = bind_remount_recursive_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, deny_list, proc_self_mountinfo); |
9ce4e4b0 | 1261 | else |
7cce68e1 | 1262 | r = bind_remount_one_with_mountinfo(mount_entry_path(m), new_flags, flags_mask, proc_self_mountinfo); |
9ce4e4b0 | 1263 | |
867189b5 LP |
1264 | /* Not that we only turn on the MS_RDONLY flag here, we never turn it off. Something that was marked |
1265 | * read-only already stays this way. This improves compatibility with container managers, where we | |
1266 | * won't attempt to undo read-only mounts already applied. */ | |
ac0930c8 | 1267 | |
8fceda93 | 1268 | if (r == -ENOENT && m->ignore) |
867189b5 | 1269 | return 0; |
763a260a | 1270 | if (r < 0) |
9ce4e4b0 | 1271 | return log_debug_errno(r, "Failed to re-mount '%s'%s: %m", mount_entry_path(m), |
763a260a | 1272 | submounts ? " and its submounts" : ""); |
763a260a | 1273 | return 0; |
d944dc95 LP |
1274 | } |
1275 | ||
9b68367b | 1276 | static bool namespace_info_mount_apivfs(const NamespaceInfo *ns_info) { |
5d997827 LP |
1277 | assert(ns_info); |
1278 | ||
9c988f93 DH |
1279 | /* |
1280 | * ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, | |
1281 | * since to protect the API VFS mounts, they need to be around in the | |
9b68367b | 1282 | * first place... |
9c988f93 | 1283 | */ |
5d997827 | 1284 | |
9b68367b YW |
1285 | return ns_info->mount_apivfs || |
1286 | ns_info->protect_control_groups || | |
4e399953 LP |
1287 | ns_info->protect_kernel_tunables || |
1288 | ns_info->protect_proc != PROTECT_PROC_DEFAULT || | |
1289 | ns_info->proc_subset != PROC_SUBSET_ALL; | |
5d997827 LP |
1290 | } |
1291 | ||
da6053d0 | 1292 | static size_t namespace_calculate_mounts( |
bb0ff3fb | 1293 | const NamespaceInfo *ns_info, |
2652c6c1 DH |
1294 | char** read_write_paths, |
1295 | char** read_only_paths, | |
1296 | char** inaccessible_paths, | |
6c47cd7d | 1297 | char** empty_directories, |
da6053d0 LP |
1298 | size_t n_bind_mounts, |
1299 | size_t n_temporary_filesystems, | |
b3d13314 | 1300 | size_t n_mount_images, |
2652c6c1 DH |
1301 | const char* tmp_dir, |
1302 | const char* var_tmp_dir, | |
bbb4e7f3 | 1303 | const char *creds_path, |
5e8deb94 | 1304 | const char* log_namespace, |
3bdc25a4 LP |
1305 | bool setup_propagate, |
1306 | const char* notify_socket) { | |
2652c6c1 | 1307 | |
da6053d0 LP |
1308 | size_t protect_home_cnt; |
1309 | size_t protect_system_cnt = | |
52b3d652 | 1310 | (ns_info->protect_system == PROTECT_SYSTEM_STRICT ? |
f471b2af | 1311 | ELEMENTSOF(protect_system_strict_table) : |
52b3d652 | 1312 | ((ns_info->protect_system == PROTECT_SYSTEM_FULL) ? |
f471b2af | 1313 | ELEMENTSOF(protect_system_full_table) : |
52b3d652 | 1314 | ((ns_info->protect_system == PROTECT_SYSTEM_YES) ? |
f471b2af DH |
1315 | ELEMENTSOF(protect_system_yes_table) : 0))); |
1316 | ||
b6c432ca | 1317 | protect_home_cnt = |
52b3d652 | 1318 | (ns_info->protect_home == PROTECT_HOME_YES ? |
b6c432ca | 1319 | ELEMENTSOF(protect_home_yes_table) : |
52b3d652 | 1320 | ((ns_info->protect_home == PROTECT_HOME_READ_ONLY) ? |
e4da7d8c | 1321 | ELEMENTSOF(protect_home_read_only_table) : |
52b3d652 | 1322 | ((ns_info->protect_home == PROTECT_HOME_TMPFS) ? |
e4da7d8c | 1323 | ELEMENTSOF(protect_home_tmpfs_table) : 0))); |
b6c432ca | 1324 | |
2652c6c1 DH |
1325 | return !!tmp_dir + !!var_tmp_dir + |
1326 | strv_length(read_write_paths) + | |
1327 | strv_length(read_only_paths) + | |
1328 | strv_length(inaccessible_paths) + | |
6c47cd7d | 1329 | strv_length(empty_directories) + |
d2d6c096 | 1330 | n_bind_mounts + |
b3d13314 | 1331 | n_mount_images + |
2abd4e38 | 1332 | n_temporary_filesystems + |
c575770b DH |
1333 | ns_info->private_dev + |
1334 | (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) + | |
c575770b | 1335 | (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) + |
94a7b275 KK |
1336 | (ns_info->protect_kernel_logs ? ELEMENTSOF(protect_kernel_logs_table) : 0) + |
1337 | (ns_info->protect_control_groups ? 1 : 0) + | |
5d997827 | 1338 | protect_home_cnt + protect_system_cnt + |
aecd5ac6 | 1339 | (ns_info->protect_hostname ? 2 : 0) + |
91dd5f7c | 1340 | (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0) + |
bbb4e7f3 | 1341 | (creds_path ? 2 : 1) + |
3bdc25a4 LP |
1342 | !!log_namespace + |
1343 | setup_propagate + /* /run/systemd/incoming */ | |
1344 | !!notify_socket; | |
2652c6c1 DH |
1345 | } |
1346 | ||
da6053d0 | 1347 | static void normalize_mounts(const char *root_directory, MountEntry *mounts, size_t *n_mounts) { |
9b68367b | 1348 | assert(root_directory); |
f8b64b57 LP |
1349 | assert(n_mounts); |
1350 | assert(mounts || *n_mounts == 0); | |
1351 | ||
93bab288 | 1352 | typesafe_qsort(mounts, *n_mounts, mount_path_compare); |
f8b64b57 LP |
1353 | |
1354 | drop_duplicates(mounts, n_mounts); | |
1355 | drop_outside_root(root_directory, mounts, n_mounts); | |
1356 | drop_inaccessible(mounts, n_mounts); | |
1357 | drop_nop(mounts, n_mounts); | |
1358 | } | |
1359 | ||
c8c535d5 LP |
1360 | static bool root_read_only( |
1361 | char **read_only_paths, | |
1362 | ProtectSystem protect_system) { | |
1363 | ||
1364 | /* Determine whether the root directory is going to be read-only given the configured settings. */ | |
1365 | ||
1366 | if (protect_system == PROTECT_SYSTEM_STRICT) | |
1367 | return true; | |
1368 | ||
de46b2be | 1369 | if (prefixed_path_strv_contains(read_only_paths, "/")) |
c8c535d5 LP |
1370 | return true; |
1371 | ||
1372 | return false; | |
1373 | } | |
1374 | ||
1375 | static bool home_read_only( | |
1376 | char** read_only_paths, | |
1377 | char** inaccessible_paths, | |
1378 | char** empty_directories, | |
1379 | const BindMount *bind_mounts, | |
1380 | size_t n_bind_mounts, | |
1381 | const TemporaryFileSystem *temporary_filesystems, | |
1382 | size_t n_temporary_filesystems, | |
1383 | ProtectHome protect_home) { | |
1384 | ||
1385 | size_t i; | |
1386 | ||
1387 | /* Determine whether the /home directory is going to be read-only given the configured settings. Yes, | |
1388 | * this is a bit sloppy, since we don't bother checking for cases where / is affected by multiple | |
1389 | * settings. */ | |
1390 | ||
1391 | if (protect_home != PROTECT_HOME_NO) | |
1392 | return true; | |
1393 | ||
de46b2be TM |
1394 | if (prefixed_path_strv_contains(read_only_paths, "/home") || |
1395 | prefixed_path_strv_contains(inaccessible_paths, "/home") || | |
1396 | prefixed_path_strv_contains(empty_directories, "/home")) | |
c8c535d5 LP |
1397 | return true; |
1398 | ||
1399 | for (i = 0; i < n_temporary_filesystems; i++) | |
1400 | if (path_equal(temporary_filesystems[i].path, "/home")) | |
1401 | return true; | |
1402 | ||
1403 | /* If /home is overmounted with some dir from the host it's not writable. */ | |
1404 | for (i = 0; i < n_bind_mounts; i++) | |
1405 | if (path_equal(bind_mounts[i].destination, "/home")) | |
1406 | return true; | |
1407 | ||
1408 | return false; | |
1409 | } | |
1410 | ||
89e62e0b LP |
1411 | static int verity_settings_prepare( |
1412 | VeritySettings *verity, | |
1413 | const char *root_image, | |
1414 | const void *root_hash, | |
1415 | size_t root_hash_size, | |
1416 | const char *root_hash_path, | |
1417 | const void *root_hash_sig, | |
1418 | size_t root_hash_sig_size, | |
1419 | const char *root_hash_sig_path, | |
1420 | const char *verity_data_path) { | |
1421 | ||
1422 | int r; | |
1423 | ||
1424 | assert(verity); | |
1425 | ||
1426 | if (root_hash) { | |
1427 | void *d; | |
1428 | ||
1429 | d = memdup(root_hash, root_hash_size); | |
1430 | if (!d) | |
1431 | return -ENOMEM; | |
1432 | ||
1433 | free_and_replace(verity->root_hash, d); | |
1434 | verity->root_hash_size = root_hash_size; | |
aee36b4e | 1435 | verity->designator = PARTITION_ROOT; |
89e62e0b LP |
1436 | } |
1437 | ||
1438 | if (root_hash_sig) { | |
1439 | void *d; | |
1440 | ||
1441 | d = memdup(root_hash_sig, root_hash_sig_size); | |
1442 | if (!d) | |
1443 | return -ENOMEM; | |
1444 | ||
1445 | free_and_replace(verity->root_hash_sig, d); | |
1446 | verity->root_hash_sig_size = root_hash_sig_size; | |
aee36b4e | 1447 | verity->designator = PARTITION_ROOT; |
89e62e0b LP |
1448 | } |
1449 | ||
1450 | if (verity_data_path) { | |
1451 | r = free_and_strdup(&verity->data_path, verity_data_path); | |
1452 | if (r < 0) | |
1453 | return r; | |
1454 | } | |
1455 | ||
1456 | r = verity_settings_load( | |
1457 | verity, | |
1458 | root_image, | |
1459 | root_hash_path, | |
1460 | root_hash_sig_path); | |
1461 | if (r < 0) | |
1462 | return log_debug_errno(r, "Failed to load root hash: %m"); | |
1463 | ||
1464 | return 0; | |
1465 | } | |
1466 | ||
613b411c | 1467 | int setup_namespace( |
ee818b89 | 1468 | const char* root_directory, |
915e6d16 | 1469 | const char* root_image, |
18d73705 | 1470 | const MountOptions *root_image_options, |
bb0ff3fb | 1471 | const NamespaceInfo *ns_info, |
2a624c36 AP |
1472 | char** read_write_paths, |
1473 | char** read_only_paths, | |
1474 | char** inaccessible_paths, | |
6c47cd7d | 1475 | char** empty_directories, |
d2d6c096 | 1476 | const BindMount *bind_mounts, |
da6053d0 | 1477 | size_t n_bind_mounts, |
2abd4e38 | 1478 | const TemporaryFileSystem *temporary_filesystems, |
da6053d0 | 1479 | size_t n_temporary_filesystems, |
b3d13314 LB |
1480 | const MountImage *mount_images, |
1481 | size_t n_mount_images, | |
a004cb4c LP |
1482 | const char* tmp_dir, |
1483 | const char* var_tmp_dir, | |
bbb4e7f3 | 1484 | const char *creds_path, |
91dd5f7c | 1485 | const char *log_namespace, |
915e6d16 | 1486 | unsigned long mount_flags, |
0389f4fa LB |
1487 | const void *root_hash, |
1488 | size_t root_hash_size, | |
1489 | const char *root_hash_path, | |
d4d55b0d LB |
1490 | const void *root_hash_sig, |
1491 | size_t root_hash_sig_size, | |
1492 | const char *root_hash_sig_path, | |
89e62e0b | 1493 | const char *verity_data_path, |
5e8deb94 LB |
1494 | const char *propagate_dir, |
1495 | const char *incoming_dir, | |
3bdc25a4 | 1496 | const char *notify_socket, |
7cc5ef5f ZJS |
1497 | DissectImageFlags dissect_image_flags, |
1498 | char **error_path) { | |
15ae422b | 1499 | |
915e6d16 | 1500 | _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; |
78ebe980 | 1501 | _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL; |
915e6d16 | 1502 | _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; |
aee36b4e | 1503 | _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT; |
5f7a690a | 1504 | MountEntry *m = NULL, *mounts = NULL; |
5e8deb94 | 1505 | bool require_prefix = false, setup_propagate = false; |
9b68367b | 1506 | const char *root; |
89e62e0b LP |
1507 | size_t n_mounts; |
1508 | int r; | |
15ae422b | 1509 | |
915e6d16 LP |
1510 | assert(ns_info); |
1511 | ||
5e8deb94 LB |
1512 | if (!isempty(propagate_dir) && !isempty(incoming_dir)) |
1513 | setup_propagate = true; | |
1514 | ||
613b411c | 1515 | if (mount_flags == 0) |
c17ec25e | 1516 | mount_flags = MS_SHARED; |
ac0930c8 | 1517 | |
915e6d16 LP |
1518 | if (root_image) { |
1519 | dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT; | |
1520 | ||
c8c535d5 LP |
1521 | /* Make the whole image read-only if we can determine that we only access it in a read-only fashion. */ |
1522 | if (root_read_only(read_only_paths, | |
52b3d652 | 1523 | ns_info->protect_system) && |
c8c535d5 LP |
1524 | home_read_only(read_only_paths, inaccessible_paths, empty_directories, |
1525 | bind_mounts, n_bind_mounts, temporary_filesystems, n_temporary_filesystems, | |
52b3d652 | 1526 | ns_info->protect_home) && |
c9ef8573 | 1527 | strv_isempty(read_write_paths)) |
915e6d16 LP |
1528 | dissect_image_flags |= DISSECT_IMAGE_READ_ONLY; |
1529 | ||
89e62e0b LP |
1530 | r = verity_settings_prepare( |
1531 | &verity, | |
1532 | root_image, | |
1533 | root_hash, root_hash_size, root_hash_path, | |
1534 | root_hash_sig, root_hash_sig_size, root_hash_sig_path, | |
1535 | verity_data_path); | |
915e6d16 | 1536 | if (r < 0) |
89e62e0b LP |
1537 | return r; |
1538 | ||
1539 | SET_FLAG(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE, verity.data_path); | |
915e6d16 | 1540 | |
89e62e0b LP |
1541 | r = loop_device_make_by_path( |
1542 | root_image, | |
1543 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_READ_ONLY) ? O_RDONLY : -1 /* < 0 means writable if possible, read-only as fallback */, | |
1544 | FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN, | |
1545 | &loop_device); | |
78ebe980 | 1546 | if (r < 0) |
89e62e0b LP |
1547 | return log_debug_errno(r, "Failed to create loop device for root image: %m"); |
1548 | ||
1549 | r = dissect_image( | |
1550 | loop_device->fd, | |
1551 | &verity, | |
1552 | root_image_options, | |
1553 | dissect_image_flags, | |
1554 | &dissected_image); | |
78ebe980 | 1555 | if (r < 0) |
763a260a | 1556 | return log_debug_errno(r, "Failed to dissect image: %m"); |
78ebe980 | 1557 | |
89e62e0b LP |
1558 | r = dissected_image_decrypt( |
1559 | dissected_image, | |
1560 | NULL, | |
1561 | &verity, | |
1562 | dissect_image_flags, | |
1563 | &decrypted_image); | |
915e6d16 | 1564 | if (r < 0) |
763a260a | 1565 | return log_debug_errno(r, "Failed to decrypt dissected image: %m"); |
915e6d16 LP |
1566 | } |
1567 | ||
e908468b LP |
1568 | if (root_directory) |
1569 | root = root_directory; | |
0722b359 | 1570 | else { |
77f16dbd DDM |
1571 | /* /run/systemd should have been created by PID 1 early on already, but in some cases, like |
1572 | * when running tests (test-execute), it might not have been created yet so let's make sure | |
1573 | * we create it if it doesn't already exist. */ | |
1574 | (void) mkdir_p_label("/run/systemd", 0755); | |
1575 | ||
0722b359 JS |
1576 | /* Always create the mount namespace in a temporary directory, instead of operating |
1577 | * directly in the root. The temporary directory prevents any mounts from being | |
1578 | * potentially obscured my other mounts we already applied. | |
1579 | * We use the same mount point for all images, which is safe, since they all live | |
1580 | * in their own namespaces after all, and hence won't see each other. */ | |
e908468b LP |
1581 | |
1582 | root = "/run/systemd/unit-root"; | |
1583 | (void) mkdir_label(root, 0700); | |
d18aff04 | 1584 | require_prefix = true; |
0722b359 | 1585 | } |
e908468b | 1586 | |
cfbeb4ef LP |
1587 | n_mounts = namespace_calculate_mounts( |
1588 | ns_info, | |
1589 | read_write_paths, | |
1590 | read_only_paths, | |
1591 | inaccessible_paths, | |
6c47cd7d | 1592 | empty_directories, |
f5c52a77 | 1593 | n_bind_mounts, |
2abd4e38 | 1594 | n_temporary_filesystems, |
b3d13314 | 1595 | n_mount_images, |
cfbeb4ef | 1596 | tmp_dir, var_tmp_dir, |
bbb4e7f3 | 1597 | creds_path, |
5e8deb94 | 1598 | log_namespace, |
3bdc25a4 LP |
1599 | setup_propagate, |
1600 | notify_socket); | |
613b411c | 1601 | |
f0a4feb0 | 1602 | if (n_mounts > 0) { |
5f7a690a LP |
1603 | m = mounts = new0(MountEntry, n_mounts); |
1604 | if (!mounts) | |
1605 | return -ENOMEM; | |
1606 | ||
d18aff04 | 1607 | r = append_access_mounts(&m, read_write_paths, READWRITE, require_prefix); |
613b411c | 1608 | if (r < 0) |
f0a4feb0 | 1609 | goto finish; |
613b411c | 1610 | |
d18aff04 | 1611 | r = append_access_mounts(&m, read_only_paths, READONLY, require_prefix); |
613b411c | 1612 | if (r < 0) |
f0a4feb0 | 1613 | goto finish; |
613b411c | 1614 | |
d18aff04 | 1615 | r = append_access_mounts(&m, inaccessible_paths, INACCESSIBLE, require_prefix); |
613b411c | 1616 | if (r < 0) |
f0a4feb0 | 1617 | goto finish; |
7ff7394d | 1618 | |
6c47cd7d LP |
1619 | r = append_empty_dir_mounts(&m, empty_directories); |
1620 | if (r < 0) | |
1621 | goto finish; | |
1622 | ||
d2d6c096 LP |
1623 | r = append_bind_mounts(&m, bind_mounts, n_bind_mounts); |
1624 | if (r < 0) | |
1625 | goto finish; | |
1626 | ||
2abd4e38 YW |
1627 | r = append_tmpfs_mounts(&m, temporary_filesystems, n_temporary_filesystems); |
1628 | if (r < 0) | |
1629 | goto finish; | |
1630 | ||
613b411c | 1631 | if (tmp_dir) { |
56a13a49 ZJS |
1632 | bool ro = streq(tmp_dir, RUN_SYSTEMD_EMPTY); |
1633 | ||
34de407a | 1634 | *(m++) = (MountEntry) { |
5327c910 | 1635 | .path_const = "/tmp", |
56a13a49 | 1636 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1637 | .source_const = tmp_dir, |
5327c910 | 1638 | }; |
613b411c | 1639 | } |
7ff7394d | 1640 | |
613b411c | 1641 | if (var_tmp_dir) { |
56a13a49 ZJS |
1642 | bool ro = streq(var_tmp_dir, RUN_SYSTEMD_EMPTY); |
1643 | ||
34de407a | 1644 | *(m++) = (MountEntry) { |
5327c910 | 1645 | .path_const = "/var/tmp", |
56a13a49 | 1646 | .mode = ro ? PRIVATE_TMP_READONLY : PRIVATE_TMP, |
89bd586c | 1647 | .source_const = var_tmp_dir, |
5327c910 | 1648 | }; |
7ff7394d | 1649 | } |
ac0930c8 | 1650 | |
b3d13314 LB |
1651 | r = append_mount_images(&m, mount_images, n_mount_images); |
1652 | if (r < 0) | |
1653 | goto finish; | |
1654 | ||
d46b79bb | 1655 | if (ns_info->private_dev) |
34de407a | 1656 | *(m++) = (MountEntry) { |
5327c910 LP |
1657 | .path_const = "/dev", |
1658 | .mode = PRIVATE_DEV, | |
9ce4e4b0 | 1659 | .flags = DEV_MOUNT_OPTIONS, |
5327c910 | 1660 | }; |
7f112f50 | 1661 | |
c575770b | 1662 | if (ns_info->protect_kernel_tunables) { |
cbc056c8 ZJS |
1663 | r = append_static_mounts(&m, |
1664 | protect_kernel_tunables_table, | |
1665 | ELEMENTSOF(protect_kernel_tunables_table), | |
1666 | ns_info->ignore_protect_paths); | |
c575770b | 1667 | if (r < 0) |
f0a4feb0 | 1668 | goto finish; |
c575770b DH |
1669 | } |
1670 | ||
1671 | if (ns_info->protect_kernel_modules) { | |
cbc056c8 ZJS |
1672 | r = append_static_mounts(&m, |
1673 | protect_kernel_modules_table, | |
1674 | ELEMENTSOF(protect_kernel_modules_table), | |
1675 | ns_info->ignore_protect_paths); | |
c575770b | 1676 | if (r < 0) |
f0a4feb0 | 1677 | goto finish; |
c575770b | 1678 | } |
59eeb84b | 1679 | |
94a7b275 | 1680 | if (ns_info->protect_kernel_logs) { |
cbc056c8 ZJS |
1681 | r = append_static_mounts(&m, |
1682 | protect_kernel_logs_table, | |
1683 | ELEMENTSOF(protect_kernel_logs_table), | |
1684 | ns_info->ignore_protect_paths); | |
94a7b275 KK |
1685 | if (r < 0) |
1686 | goto finish; | |
1687 | } | |
1688 | ||
d46b79bb | 1689 | if (ns_info->protect_control_groups) |
34de407a | 1690 | *(m++) = (MountEntry) { |
5327c910 LP |
1691 | .path_const = "/sys/fs/cgroup", |
1692 | .mode = READONLY, | |
1693 | }; | |
59eeb84b | 1694 | |
52b3d652 | 1695 | r = append_protect_home(&m, ns_info->protect_home, ns_info->ignore_protect_paths); |
b6c432ca | 1696 | if (r < 0) |
f0a4feb0 | 1697 | goto finish; |
417116f2 | 1698 | |
52b3d652 | 1699 | r = append_protect_system(&m, ns_info->protect_system, false); |
f471b2af | 1700 | if (r < 0) |
f0a4feb0 | 1701 | goto finish; |
417116f2 | 1702 | |
9b68367b | 1703 | if (namespace_info_mount_apivfs(ns_info)) { |
cbc056c8 ZJS |
1704 | r = append_static_mounts(&m, |
1705 | apivfs_table, | |
1706 | ELEMENTSOF(apivfs_table), | |
1707 | ns_info->ignore_protect_paths); | |
5d997827 LP |
1708 | if (r < 0) |
1709 | goto finish; | |
1710 | } | |
1711 | ||
aecd5ac6 TM |
1712 | if (ns_info->protect_hostname) { |
1713 | *(m++) = (MountEntry) { | |
1714 | .path_const = "/proc/sys/kernel/hostname", | |
1715 | .mode = READONLY, | |
1716 | }; | |
1717 | *(m++) = (MountEntry) { | |
1718 | .path_const = "/proc/sys/kernel/domainname", | |
1719 | .mode = READONLY, | |
1720 | }; | |
1721 | } | |
1722 | ||
bbb4e7f3 LP |
1723 | if (creds_path) { |
1724 | /* If our service has a credentials store configured, then bind that one in, but hide | |
1725 | * everything else. */ | |
1726 | ||
1727 | *(m++) = (MountEntry) { | |
1728 | .path_const = "/run/credentials", | |
1729 | .mode = TMPFS, | |
1730 | .read_only = true, | |
1731 | .options_const = "mode=0755" TMPFS_LIMITS_EMPTY_OR_ALMOST, | |
1732 | .flags = MS_NODEV|MS_STRICTATIME|MS_NOSUID|MS_NOEXEC, | |
1733 | }; | |
1734 | ||
1735 | *(m++) = (MountEntry) { | |
1736 | .path_const = creds_path, | |
1737 | .mode = BIND_MOUNT, | |
1738 | .read_only = true, | |
1739 | .source_const = creds_path, | |
1740 | }; | |
1741 | } else { | |
1742 | /* If our service has no credentials store configured, then make the whole | |
1743 | * credentials tree inaccessible wholesale. */ | |
1744 | ||
1745 | *(m++) = (MountEntry) { | |
1746 | .path_const = "/run/credentials", | |
1747 | .mode = INACCESSIBLE, | |
1748 | .ignore = true, | |
1749 | }; | |
1750 | } | |
1751 | ||
91dd5f7c LP |
1752 | if (log_namespace) { |
1753 | _cleanup_free_ char *q; | |
1754 | ||
1755 | q = strjoin("/run/systemd/journal.", log_namespace); | |
1756 | if (!q) { | |
1757 | r = -ENOMEM; | |
1758 | goto finish; | |
1759 | } | |
1760 | ||
1761 | *(m++) = (MountEntry) { | |
1762 | .path_const = "/run/systemd/journal", | |
1763 | .mode = BIND_MOUNT_RECURSIVE, | |
1764 | .read_only = true, | |
1765 | .source_malloc = TAKE_PTR(q), | |
1766 | }; | |
1767 | } | |
1768 | ||
5e8deb94 LB |
1769 | /* Will be used to add bind mounts at runtime */ |
1770 | if (setup_propagate) | |
1771 | *(m++) = (MountEntry) { | |
1772 | .source_const = propagate_dir, | |
1773 | .path_const = incoming_dir, | |
1774 | .mode = BIND_MOUNT, | |
1775 | .read_only = true, | |
1776 | }; | |
1777 | ||
3bdc25a4 LP |
1778 | if (notify_socket) |
1779 | *(m++) = (MountEntry) { | |
1780 | .path_const = notify_socket, | |
1781 | .source_const = notify_socket, | |
1782 | .mode = BIND_MOUNT, | |
1783 | .read_only = true, | |
1784 | }; | |
1785 | ||
f0a4feb0 | 1786 | assert(mounts + n_mounts == m); |
ac0930c8 | 1787 | |
5327c910 | 1788 | /* Prepend the root directory where that's necessary */ |
e908468b | 1789 | r = prefix_where_needed(mounts, n_mounts, root); |
5327c910 LP |
1790 | if (r < 0) |
1791 | goto finish; | |
1792 | ||
839f1877 | 1793 | normalize_mounts(root, mounts, &n_mounts); |
15ae422b LP |
1794 | } |
1795 | ||
1beab8b0 LP |
1796 | /* All above is just preparation, figuring out what to do. Let's now actually start doing something. */ |
1797 | ||
d944dc95 | 1798 | if (unshare(CLONE_NEWNS) < 0) { |
763a260a | 1799 | r = log_debug_errno(errno, "Failed to unshare the mount namespace: %m"); |
1beab8b0 | 1800 | if (IN_SET(r, -EACCES, -EPERM, -EOPNOTSUPP, -ENOSYS)) |
cbc056c8 ZJS |
1801 | /* If the kernel doesn't support namespaces, or when there's a MAC or seccomp filter |
1802 | * in place that doesn't allow us to create namespaces (or a missing cap), then | |
1803 | * propagate a recognizable error back, which the caller can use to detect this case | |
1804 | * (and only this) and optionally continue without namespacing applied. */ | |
1beab8b0 LP |
1805 | r = -ENOANO; |
1806 | ||
d944dc95 LP |
1807 | goto finish; |
1808 | } | |
1e4e94c8 | 1809 | |
5e8deb94 LB |
1810 | /* Create the source directory to allow runtime propagation of mounts */ |
1811 | if (setup_propagate) | |
1812 | (void) mkdir_p(propagate_dir, 0600); | |
1813 | ||
9b68367b YW |
1814 | /* Remount / as SLAVE so that nothing now mounted in the namespace |
1815 | * shows up in the parent */ | |
1816 | if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { | |
763a260a | 1817 | r = log_debug_errno(errno, "Failed to remount '/' as SLAVE: %m"); |
9b68367b | 1818 | goto finish; |
ee818b89 AC |
1819 | } |
1820 | ||
915e6d16 | 1821 | if (root_image) { |
e908468b | 1822 | /* A root image is specified, mount it to the right place */ |
2d3a5a73 | 1823 | r = dissected_image_mount(dissected_image, root, UID_INVALID, dissect_image_flags); |
763a260a YW |
1824 | if (r < 0) { |
1825 | log_debug_errno(r, "Failed to mount root image: %m"); | |
915e6d16 | 1826 | goto finish; |
763a260a | 1827 | } |
915e6d16 | 1828 | |
07ce7407 TM |
1829 | if (decrypted_image) { |
1830 | r = decrypted_image_relinquish(decrypted_image); | |
763a260a YW |
1831 | if (r < 0) { |
1832 | log_debug_errno(r, "Failed to relinquish decrypted image: %m"); | |
07ce7407 | 1833 | goto finish; |
763a260a | 1834 | } |
07ce7407 | 1835 | } |
78ebe980 | 1836 | |
915e6d16 LP |
1837 | loop_device_relinquish(loop_device); |
1838 | ||
1839 | } else if (root_directory) { | |
1840 | ||
e908468b LP |
1841 | /* A root directory is specified. Turn its directory into bind mount, if it isn't one yet. */ |
1842 | r = path_is_mount_point(root, NULL, AT_SYMLINK_FOLLOW); | |
763a260a YW |
1843 | if (r < 0) { |
1844 | log_debug_errno(r, "Failed to detect that %s is a mount point or not: %m", root); | |
d944dc95 | 1845 | goto finish; |
763a260a | 1846 | } |
8f1ad200 | 1847 | if (r == 0) { |
21935150 LP |
1848 | r = mount_nofollow_verbose(LOG_DEBUG, root, root, NULL, MS_BIND|MS_REC, NULL); |
1849 | if (r < 0) | |
8f1ad200 | 1850 | goto finish; |
d944dc95 | 1851 | } |
e908468b | 1852 | |
9b68367b | 1853 | } else { |
e908468b | 1854 | /* Let's mount the main root directory to the root directory to use */ |
21935150 LP |
1855 | r = mount_nofollow_verbose(LOG_DEBUG, "/", root, NULL, MS_BIND|MS_REC, NULL); |
1856 | if (r < 0) | |
e908468b | 1857 | goto finish; |
ee818b89 | 1858 | } |
c2c13f2d | 1859 | |
4e0c20de LP |
1860 | /* Try to set up the new root directory before mounting anything else there. */ |
1861 | if (root_image || root_directory) | |
1862 | (void) base_filesystem_create(root, UID_INVALID, GID_INVALID); | |
1863 | ||
f0a4feb0 | 1864 | if (n_mounts > 0) { |
ac9de0b3 | 1865 | _cleanup_fclose_ FILE *proc_self_mountinfo = NULL; |
6b000af4 | 1866 | _cleanup_free_ char **deny_list = NULL; |
da6053d0 | 1867 | size_t j; |
6b7c9f8b | 1868 | |
cbc056c8 ZJS |
1869 | /* Open /proc/self/mountinfo now as it may become unavailable if we mount anything on top of |
1870 | * /proc. For example, this is the case with the option: 'InaccessiblePaths=/proc'. */ | |
ac9de0b3 TR |
1871 | proc_self_mountinfo = fopen("/proc/self/mountinfo", "re"); |
1872 | if (!proc_self_mountinfo) { | |
763a260a | 1873 | r = log_debug_errno(errno, "Failed to open /proc/self/mountinfo: %m"); |
7cc5ef5f ZJS |
1874 | if (error_path) |
1875 | *error_path = strdup("/proc/self/mountinfo"); | |
ac9de0b3 TR |
1876 | goto finish; |
1877 | } | |
1878 | ||
088696fe LP |
1879 | /* First round, establish all mounts we need */ |
1880 | for (;;) { | |
1881 | bool again = false; | |
1882 | ||
1883 | for (m = mounts; m < mounts + n_mounts; ++m) { | |
1884 | ||
1885 | if (m->applied) | |
1886 | continue; | |
1887 | ||
1888 | r = follow_symlink(root, m); | |
7cc5ef5f ZJS |
1889 | if (r < 0) { |
1890 | if (error_path && mount_entry_path(m)) | |
1891 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1892 | goto finish; |
7cc5ef5f | 1893 | } |
088696fe | 1894 | if (r == 0) { |
cbc056c8 ZJS |
1895 | /* We hit a symlinked mount point. The entry got rewritten and might |
1896 | * point to a very different place now. Let's normalize the changed | |
1897 | * list, and start from the beginning. After all to mount the entry | |
1898 | * at the new location we might need some other mounts first */ | |
088696fe LP |
1899 | again = true; |
1900 | break; | |
1901 | } | |
1902 | ||
4e399953 | 1903 | r = apply_mount(root, m, ns_info); |
7cc5ef5f ZJS |
1904 | if (r < 0) { |
1905 | if (error_path && mount_entry_path(m)) | |
1906 | *error_path = strdup(mount_entry_path(m)); | |
088696fe | 1907 | goto finish; |
7cc5ef5f | 1908 | } |
088696fe LP |
1909 | |
1910 | m->applied = true; | |
1911 | } | |
1912 | ||
1913 | if (!again) | |
1914 | break; | |
1915 | ||
839f1877 | 1916 | normalize_mounts(root, mounts, &n_mounts); |
c2c13f2d | 1917 | } |
15ae422b | 1918 | |
6b000af4 LP |
1919 | /* Create a deny list we can pass to bind_mount_recursive() */ |
1920 | deny_list = new(char*, n_mounts+1); | |
1921 | if (!deny_list) { | |
5f7a690a LP |
1922 | r = -ENOMEM; |
1923 | goto finish; | |
1924 | } | |
f0a4feb0 | 1925 | for (j = 0; j < n_mounts; j++) |
6b000af4 LP |
1926 | deny_list[j] = (char*) mount_entry_path(mounts+j); |
1927 | deny_list[j] = NULL; | |
6b7c9f8b LP |
1928 | |
1929 | /* Second round, flip the ro bits if necessary. */ | |
f0a4feb0 | 1930 | for (m = mounts; m < mounts + n_mounts; ++m) { |
6b000af4 | 1931 | r = make_read_only(m, deny_list, proc_self_mountinfo); |
7cc5ef5f ZJS |
1932 | if (r < 0) { |
1933 | if (error_path && mount_entry_path(m)) | |
1934 | *error_path = strdup(mount_entry_path(m)); | |
d944dc95 | 1935 | goto finish; |
7cc5ef5f | 1936 | } |
c2c13f2d | 1937 | } |
15ae422b LP |
1938 | } |
1939 | ||
9b68367b YW |
1940 | /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */ |
1941 | r = mount_move_root(root); | |
763a260a YW |
1942 | if (r < 0) { |
1943 | log_debug_errno(r, "Failed to mount root with MS_MOVE: %m"); | |
9b68367b | 1944 | goto finish; |
763a260a | 1945 | } |
ee818b89 | 1946 | |
55fe7432 | 1947 | /* Remount / as the desired mode. Note that this will not |
c2c13f2d LP |
1948 | * reestablish propagation from our side to the host, since |
1949 | * what's disconnected is disconnected. */ | |
d944dc95 | 1950 | if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) { |
763a260a | 1951 | r = log_debug_errno(errno, "Failed to remount '/' with desired mount flags: %m"); |
d944dc95 LP |
1952 | goto finish; |
1953 | } | |
15ae422b | 1954 | |
5e8deb94 LB |
1955 | /* bind_mount_in_namespace() will MS_MOVE into that directory, and that's only |
1956 | * supported for non-shared mounts. This needs to happen after remounting / or it will fail. */ | |
1957 | if (setup_propagate) { | |
1958 | r = mount(NULL, incoming_dir, NULL, MS_SLAVE, NULL); | |
1959 | if (r < 0) { | |
1960 | log_error_errno(r, "Failed to remount %s with MS_SLAVE: %m", incoming_dir); | |
1961 | goto finish; | |
1962 | } | |
1963 | } | |
1964 | ||
d944dc95 | 1965 | r = 0; |
15ae422b | 1966 | |
d944dc95 | 1967 | finish: |
0cd41757 LP |
1968 | if (n_mounts > 0) |
1969 | for (m = mounts; m < mounts + n_mounts; m++) | |
1970 | mount_entry_done(m); | |
613b411c | 1971 | |
5f7a690a LP |
1972 | free(mounts); |
1973 | ||
613b411c LP |
1974 | return r; |
1975 | } | |
1976 | ||
da6053d0 LP |
1977 | void bind_mount_free_many(BindMount *b, size_t n) { |
1978 | size_t i; | |
d2d6c096 LP |
1979 | |
1980 | assert(b || n == 0); | |
1981 | ||
1982 | for (i = 0; i < n; i++) { | |
1983 | free(b[i].source); | |
1984 | free(b[i].destination); | |
1985 | } | |
1986 | ||
1987 | free(b); | |
1988 | } | |
1989 | ||
da6053d0 | 1990 | int bind_mount_add(BindMount **b, size_t *n, const BindMount *item) { |
d2d6c096 LP |
1991 | _cleanup_free_ char *s = NULL, *d = NULL; |
1992 | BindMount *c; | |
1993 | ||
1994 | assert(b); | |
1995 | assert(n); | |
1996 | assert(item); | |
1997 | ||
1998 | s = strdup(item->source); | |
1999 | if (!s) | |
2000 | return -ENOMEM; | |
2001 | ||
2002 | d = strdup(item->destination); | |
2003 | if (!d) | |
2004 | return -ENOMEM; | |
2005 | ||
aa484f35 | 2006 | c = reallocarray(*b, *n + 1, sizeof(BindMount)); |
d2d6c096 LP |
2007 | if (!c) |
2008 | return -ENOMEM; | |
2009 | ||
2010 | *b = c; | |
2011 | ||
2012 | c[(*n) ++] = (BindMount) { | |
1cc6c93a YW |
2013 | .source = TAKE_PTR(s), |
2014 | .destination = TAKE_PTR(d), | |
d2d6c096 | 2015 | .read_only = item->read_only, |
9ce4e4b0 | 2016 | .nosuid = item->nosuid, |
d2d6c096 LP |
2017 | .recursive = item->recursive, |
2018 | .ignore_enoent = item->ignore_enoent, | |
2019 | }; | |
2020 | ||
d2d6c096 LP |
2021 | return 0; |
2022 | } | |
2023 | ||
b3d13314 LB |
2024 | MountImage* mount_image_free_many(MountImage *m, size_t *n) { |
2025 | size_t i; | |
2026 | ||
2027 | assert(n); | |
2028 | assert(m || *n == 0); | |
2029 | ||
2030 | for (i = 0; i < *n; i++) { | |
2031 | free(m[i].source); | |
2032 | free(m[i].destination); | |
427353f6 | 2033 | mount_options_free_all(m[i].mount_options); |
b3d13314 LB |
2034 | } |
2035 | ||
2036 | free(m); | |
2037 | *n = 0; | |
2038 | return NULL; | |
2039 | } | |
2040 | ||
2041 | int mount_image_add(MountImage **m, size_t *n, const MountImage *item) { | |
2042 | _cleanup_free_ char *s = NULL, *d = NULL; | |
427353f6 LB |
2043 | _cleanup_(mount_options_free_allp) MountOptions *options = NULL; |
2044 | MountOptions *i; | |
b3d13314 LB |
2045 | MountImage *c; |
2046 | ||
2047 | assert(m); | |
2048 | assert(n); | |
2049 | assert(item); | |
2050 | ||
2051 | s = strdup(item->source); | |
2052 | if (!s) | |
2053 | return -ENOMEM; | |
2054 | ||
2055 | d = strdup(item->destination); | |
2056 | if (!d) | |
2057 | return -ENOMEM; | |
2058 | ||
427353f6 LB |
2059 | LIST_FOREACH(mount_options, i, item->mount_options) { |
2060 | _cleanup_(mount_options_free_allp) MountOptions *o; | |
2061 | ||
2062 | o = new(MountOptions, 1); | |
2063 | if (!o) | |
2064 | return -ENOMEM; | |
2065 | ||
2066 | *o = (MountOptions) { | |
2067 | .partition_designator = i->partition_designator, | |
2068 | .options = strdup(i->options), | |
2069 | }; | |
2070 | if (!o->options) | |
2071 | return -ENOMEM; | |
2072 | ||
2073 | LIST_APPEND(mount_options, options, TAKE_PTR(o)); | |
2074 | } | |
2075 | ||
b3d13314 LB |
2076 | c = reallocarray(*m, *n + 1, sizeof(MountImage)); |
2077 | if (!c) | |
2078 | return -ENOMEM; | |
2079 | ||
2080 | *m = c; | |
2081 | ||
2082 | c[(*n) ++] = (MountImage) { | |
2083 | .source = TAKE_PTR(s), | |
2084 | .destination = TAKE_PTR(d), | |
427353f6 | 2085 | .mount_options = TAKE_PTR(options), |
b3d13314 LB |
2086 | .ignore_enoent = item->ignore_enoent, |
2087 | }; | |
2088 | ||
2089 | return 0; | |
2090 | } | |
2091 | ||
da6053d0 LP |
2092 | void temporary_filesystem_free_many(TemporaryFileSystem *t, size_t n) { |
2093 | size_t i; | |
2abd4e38 YW |
2094 | |
2095 | assert(t || n == 0); | |
2096 | ||
2097 | for (i = 0; i < n; i++) { | |
2098 | free(t[i].path); | |
2099 | free(t[i].options); | |
2100 | } | |
2101 | ||
2102 | free(t); | |
2103 | } | |
2104 | ||
2105 | int temporary_filesystem_add( | |
2106 | TemporaryFileSystem **t, | |
da6053d0 | 2107 | size_t *n, |
2abd4e38 YW |
2108 | const char *path, |
2109 | const char *options) { | |
2110 | ||
2111 | _cleanup_free_ char *p = NULL, *o = NULL; | |
2112 | TemporaryFileSystem *c; | |
2113 | ||
2114 | assert(t); | |
2115 | assert(n); | |
2116 | assert(path); | |
2117 | ||
2118 | p = strdup(path); | |
2119 | if (!p) | |
2120 | return -ENOMEM; | |
2121 | ||
2122 | if (!isempty(options)) { | |
2123 | o = strdup(options); | |
2124 | if (!o) | |
2125 | return -ENOMEM; | |
2126 | } | |
2127 | ||
aa484f35 | 2128 | c = reallocarray(*t, *n + 1, sizeof(TemporaryFileSystem)); |
2abd4e38 YW |
2129 | if (!c) |
2130 | return -ENOMEM; | |
2131 | ||
2132 | *t = c; | |
2133 | ||
2134 | c[(*n) ++] = (TemporaryFileSystem) { | |
1cc6c93a YW |
2135 | .path = TAKE_PTR(p), |
2136 | .options = TAKE_PTR(o), | |
2abd4e38 YW |
2137 | }; |
2138 | ||
2abd4e38 YW |
2139 | return 0; |
2140 | } | |
2141 | ||
a652f050 JR |
2142 | static int make_tmp_prefix(const char *prefix) { |
2143 | _cleanup_free_ char *t = NULL; | |
2144 | int r; | |
2145 | ||
2146 | /* Don't do anything unless we know the dir is actually missing */ | |
2147 | r = access(prefix, F_OK); | |
2148 | if (r >= 0) | |
2149 | return 0; | |
2150 | if (errno != ENOENT) | |
2151 | return -errno; | |
2152 | ||
2153 | r = mkdir_parents(prefix, 0755); | |
2154 | if (r < 0) | |
2155 | return r; | |
2156 | ||
2157 | r = tempfn_random(prefix, NULL, &t); | |
2158 | if (r < 0) | |
2159 | return r; | |
2160 | ||
2161 | if (mkdir(t, 0777) < 0) | |
2162 | return -errno; | |
2163 | ||
2164 | if (chmod(t, 01777) < 0) { | |
2165 | r = -errno; | |
2166 | (void) rmdir(t); | |
2167 | return r; | |
2168 | } | |
2169 | ||
2170 | if (rename(t, prefix) < 0) { | |
2171 | r = -errno; | |
2172 | (void) rmdir(t); | |
2173 | return r == -EEXIST ? 0 : r; /* it's fine if someone else created the dir by now */ | |
2174 | } | |
2175 | ||
2176 | return 0; | |
2177 | ||
2178 | } | |
2179 | ||
56a13a49 | 2180 | static int setup_one_tmp_dir(const char *id, const char *prefix, char **path, char **tmp_path) { |
613b411c | 2181 | _cleanup_free_ char *x = NULL; |
19cd4e19 | 2182 | _cleanup_free_ char *y = NULL; |
6b46ea73 LP |
2183 | char bid[SD_ID128_STRING_MAX]; |
2184 | sd_id128_t boot_id; | |
56a13a49 | 2185 | bool rw = true; |
6b46ea73 | 2186 | int r; |
613b411c LP |
2187 | |
2188 | assert(id); | |
2189 | assert(prefix); | |
2190 | assert(path); | |
2191 | ||
6b46ea73 LP |
2192 | /* We include the boot id in the directory so that after a |
2193 | * reboot we can easily identify obsolete directories. */ | |
2194 | ||
2195 | r = sd_id128_get_boot(&boot_id); | |
2196 | if (r < 0) | |
2197 | return r; | |
2198 | ||
605405c6 | 2199 | x = strjoin(prefix, "/systemd-private-", sd_id128_to_string(boot_id, bid), "-", id, "-XXXXXX"); |
613b411c LP |
2200 | if (!x) |
2201 | return -ENOMEM; | |
2202 | ||
a652f050 JR |
2203 | r = make_tmp_prefix(prefix); |
2204 | if (r < 0) | |
2205 | return r; | |
2206 | ||
613b411c | 2207 | RUN_WITH_UMASK(0077) |
56a13a49 ZJS |
2208 | if (!mkdtemp(x)) { |
2209 | if (errno == EROFS || ERRNO_IS_DISK_SPACE(errno)) | |
2210 | rw = false; | |
2211 | else | |
2212 | return -errno; | |
2213 | } | |
613b411c | 2214 | |
56a13a49 | 2215 | if (rw) { |
19cd4e19 | 2216 | y = strjoin(x, "/tmp"); |
2217 | if (!y) | |
2218 | return -ENOMEM; | |
2219 | ||
2220 | RUN_WITH_UMASK(0000) { | |
2221 | if (mkdir(y, 0777 | S_ISVTX) < 0) | |
2222 | return -errno; | |
2223 | } | |
2224 | ||
2225 | r = label_fix_container(y, prefix, 0); | |
56a13a49 ZJS |
2226 | if (r < 0) |
2227 | return r; | |
19cd4e19 | 2228 | |
2229 | if (tmp_path) | |
2230 | *tmp_path = TAKE_PTR(y); | |
56a13a49 ZJS |
2231 | } else { |
2232 | /* Trouble: we failed to create the directory. Instead of failing, let's simulate /tmp being | |
2233 | * read-only. This way the service will get the EROFS result as if it was writing to the real | |
2234 | * file system. */ | |
2235 | r = mkdir_p(RUN_SYSTEMD_EMPTY, 0500); | |
2236 | if (r < 0) | |
2237 | return r; | |
613b411c | 2238 | |
3f181262 LP |
2239 | r = free_and_strdup(&x, RUN_SYSTEMD_EMPTY); |
2240 | if (r < 0) | |
2241 | return r; | |
c17ec25e | 2242 | } |
15ae422b | 2243 | |
1cc6c93a | 2244 | *path = TAKE_PTR(x); |
613b411c LP |
2245 | return 0; |
2246 | } | |
2247 | ||
2248 | int setup_tmp_dirs(const char *id, char **tmp_dir, char **var_tmp_dir) { | |
56a13a49 ZJS |
2249 | _cleanup_(namespace_cleanup_tmpdirp) char *a = NULL; |
2250 | _cleanup_(rmdir_and_freep) char *a_tmp = NULL; | |
2251 | char *b; | |
613b411c LP |
2252 | int r; |
2253 | ||
2254 | assert(id); | |
2255 | assert(tmp_dir); | |
2256 | assert(var_tmp_dir); | |
2257 | ||
56a13a49 | 2258 | r = setup_one_tmp_dir(id, "/tmp", &a, &a_tmp); |
613b411c LP |
2259 | if (r < 0) |
2260 | return r; | |
2261 | ||
56a13a49 ZJS |
2262 | r = setup_one_tmp_dir(id, "/var/tmp", &b, NULL); |
2263 | if (r < 0) | |
613b411c | 2264 | return r; |
613b411c | 2265 | |
56a13a49 ZJS |
2266 | a_tmp = mfree(a_tmp); /* avoid rmdir */ |
2267 | *tmp_dir = TAKE_PTR(a); | |
2268 | *var_tmp_dir = TAKE_PTR(b); | |
613b411c LP |
2269 | |
2270 | return 0; | |
2271 | } | |
2272 | ||
2caa38e9 | 2273 | int setup_netns(const int netns_storage_socket[static 2]) { |
613b411c | 2274 | _cleanup_close_ int netns = -1; |
3ee897d6 | 2275 | int r, q; |
613b411c LP |
2276 | |
2277 | assert(netns_storage_socket); | |
2278 | assert(netns_storage_socket[0] >= 0); | |
2279 | assert(netns_storage_socket[1] >= 0); | |
2280 | ||
2281 | /* We use the passed socketpair as a storage buffer for our | |
76cd584b LP |
2282 | * namespace reference fd. Whatever process runs this first |
2283 | * shall create a new namespace, all others should just join | |
2284 | * it. To serialize that we use a file lock on the socket | |
2285 | * pair. | |
613b411c LP |
2286 | * |
2287 | * It's a bit crazy, but hey, works great! */ | |
2288 | ||
2289 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2290 | return -errno; | |
2291 | ||
3ee897d6 LP |
2292 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); |
2293 | if (netns == -EAGAIN) { | |
44ffcbae | 2294 | /* Nothing stored yet, so let's create a new namespace. */ |
613b411c LP |
2295 | |
2296 | if (unshare(CLONE_NEWNET) < 0) { | |
2297 | r = -errno; | |
2298 | goto fail; | |
2299 | } | |
2300 | ||
44ffcbae | 2301 | (void) loopback_setup(); |
613b411c LP |
2302 | |
2303 | netns = open("/proc/self/ns/net", O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
2304 | if (netns < 0) { | |
2305 | r = -errno; | |
2306 | goto fail; | |
2307 | } | |
2308 | ||
2309 | r = 1; | |
613b411c | 2310 | |
3ee897d6 LP |
2311 | } else if (netns < 0) { |
2312 | r = netns; | |
2313 | goto fail; | |
613b411c | 2314 | |
3ee897d6 LP |
2315 | } else { |
2316 | /* Yay, found something, so let's join the namespace */ | |
613b411c LP |
2317 | if (setns(netns, CLONE_NEWNET) < 0) { |
2318 | r = -errno; | |
2319 | goto fail; | |
2320 | } | |
2321 | ||
2322 | r = 0; | |
2323 | } | |
2324 | ||
3ee897d6 LP |
2325 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); |
2326 | if (q < 0) { | |
2327 | r = q; | |
613b411c LP |
2328 | goto fail; |
2329 | } | |
2330 | ||
2331 | fail: | |
fe048ce5 | 2332 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); |
15ae422b LP |
2333 | return r; |
2334 | } | |
417116f2 | 2335 | |
2caa38e9 | 2336 | int open_netns_path(const int netns_storage_socket[static 2], const char *path) { |
51af7fb2 LP |
2337 | _cleanup_close_ int netns = -1; |
2338 | int q, r; | |
2339 | ||
2340 | assert(netns_storage_socket); | |
2341 | assert(netns_storage_socket[0] >= 0); | |
2342 | assert(netns_storage_socket[1] >= 0); | |
2343 | assert(path); | |
2344 | ||
2345 | /* If the storage socket doesn't contain a netns fd yet, open one via the file system and store it in | |
2346 | * it. This is supposed to be called ahead of time, i.e. before setup_netns() which will allocate a | |
2347 | * new anonymous netns if needed. */ | |
2348 | ||
2349 | if (lockf(netns_storage_socket[0], F_LOCK, 0) < 0) | |
2350 | return -errno; | |
2351 | ||
2352 | netns = receive_one_fd(netns_storage_socket[0], MSG_DONTWAIT); | |
2353 | if (netns == -EAGAIN) { | |
2354 | /* Nothing stored yet. Open the file from the file system. */ | |
2355 | ||
2356 | netns = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
2357 | if (netns < 0) { | |
2358 | r = -errno; | |
2359 | goto fail; | |
2360 | } | |
2361 | ||
2362 | r = fd_is_network_ns(netns); | |
2363 | if (r == 0) { /* Not a netns? Refuse early. */ | |
2364 | r = -EINVAL; | |
2365 | goto fail; | |
2366 | } | |
2367 | if (r < 0 && r != -EUCLEAN) /* EUCLEAN: we don't know */ | |
2368 | goto fail; | |
2369 | ||
2370 | r = 1; | |
2371 | ||
2372 | } else if (netns < 0) { | |
2373 | r = netns; | |
2374 | goto fail; | |
2375 | } else | |
2376 | r = 0; /* Already allocated */ | |
2377 | ||
2378 | q = send_one_fd(netns_storage_socket[1], netns, MSG_DONTWAIT); | |
2379 | if (q < 0) { | |
2380 | r = q; | |
2381 | goto fail; | |
2382 | } | |
2383 | ||
2384 | fail: | |
2385 | (void) lockf(netns_storage_socket[0], F_ULOCK, 0); | |
2386 | return r; | |
2387 | } | |
2388 | ||
6e2d7c4f MS |
2389 | bool ns_type_supported(NamespaceType type) { |
2390 | const char *t, *ns_proc; | |
2391 | ||
0fa5b831 LP |
2392 | t = namespace_type_to_string(type); |
2393 | if (!t) /* Don't know how to translate this? Then it's not supported */ | |
6e2d7c4f MS |
2394 | return false; |
2395 | ||
6e2d7c4f | 2396 | ns_proc = strjoina("/proc/self/ns/", t); |
6e2d7c4f MS |
2397 | return access(ns_proc, F_OK) == 0; |
2398 | } | |
2399 | ||
1b8689f9 | 2400 | static const char *const protect_home_table[_PROTECT_HOME_MAX] = { |
cbc056c8 ZJS |
2401 | [PROTECT_HOME_NO] = "no", |
2402 | [PROTECT_HOME_YES] = "yes", | |
1b8689f9 | 2403 | [PROTECT_HOME_READ_ONLY] = "read-only", |
cbc056c8 | 2404 | [PROTECT_HOME_TMPFS] = "tmpfs", |
417116f2 LP |
2405 | }; |
2406 | ||
1e8c7bd5 | 2407 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_home, ProtectHome, PROTECT_HOME_YES); |
5e1c6154 | 2408 | |
1b8689f9 | 2409 | static const char *const protect_system_table[_PROTECT_SYSTEM_MAX] = { |
cbc056c8 ZJS |
2410 | [PROTECT_SYSTEM_NO] = "no", |
2411 | [PROTECT_SYSTEM_YES] = "yes", | |
2412 | [PROTECT_SYSTEM_FULL] = "full", | |
3f815163 | 2413 | [PROTECT_SYSTEM_STRICT] = "strict", |
1b8689f9 LP |
2414 | }; |
2415 | ||
1e8c7bd5 | 2416 | DEFINE_STRING_TABLE_LOOKUP_WITH_BOOLEAN(protect_system, ProtectSystem, PROTECT_SYSTEM_YES); |
03c791aa | 2417 | |
6e2d7c4f | 2418 | static const char* const namespace_type_table[] = { |
cbc056c8 | 2419 | [NAMESPACE_MOUNT] = "mnt", |
6e2d7c4f | 2420 | [NAMESPACE_CGROUP] = "cgroup", |
cbc056c8 ZJS |
2421 | [NAMESPACE_UTS] = "uts", |
2422 | [NAMESPACE_IPC] = "ipc", | |
2423 | [NAMESPACE_USER] = "user", | |
2424 | [NAMESPACE_PID] = "pid", | |
2425 | [NAMESPACE_NET] = "net", | |
6e2d7c4f MS |
2426 | }; |
2427 | ||
2428 | DEFINE_STRING_TABLE_LOOKUP(namespace_type, NamespaceType); | |
4e399953 LP |
2429 | |
2430 | static const char* const protect_proc_table[_PROTECT_PROC_MAX] = { | |
2431 | [PROTECT_PROC_DEFAULT] = "default", | |
2432 | [PROTECT_PROC_NOACCESS] = "noaccess", | |
2433 | [PROTECT_PROC_INVISIBLE] = "invisible", | |
2434 | [PROTECT_PROC_PTRACEABLE] = "ptraceable", | |
2435 | }; | |
2436 | ||
2437 | DEFINE_STRING_TABLE_LOOKUP(protect_proc, ProtectProc); | |
2438 | ||
2439 | static const char* const proc_subset_table[_PROC_SUBSET_MAX] = { | |
2440 | [PROC_SUBSET_ALL] = "all", | |
2441 | [PROC_SUBSET_PID] = "pid", | |
2442 | }; | |
2443 | ||
2444 | DEFINE_STRING_TABLE_LOOKUP(proc_subset, ProcSubset); |