]>
Commit | Line | Data |
---|---|---|
db9ecf05 | 1 | /* SPDX-License-Identifier: LGPL-2.1-or-later */ |
cd61c3bf | 2 | |
a8fbdf54 | 3 | #include <errno.h> |
ebd93cb6 | 4 | #include <fcntl.h> |
e08f94ac LP |
5 | #include <linux/fs.h> |
6 | #include <linux/loop.h> | |
65ddc2c5 | 7 | #include <linux/magic.h> |
a8fbdf54 TA |
8 | #include <stdio.h> |
9 | #include <stdlib.h> | |
e306723e | 10 | #include <sys/file.h> |
204f52e3 | 11 | #include <sys/ioctl.h> |
a8fbdf54 TA |
12 | #include <sys/stat.h> |
13 | #include <unistd.h> | |
546dbec5 | 14 | |
b5efdb8a | 15 | #include "alloc-util.h" |
cd61c3bf | 16 | #include "btrfs-util.h" |
f4351959 | 17 | #include "chase-symlinks.h" |
c8b3094d | 18 | #include "chattr-util.h" |
ebd93cb6 | 19 | #include "copy.h" |
a0956174 | 20 | #include "dirent-util.h" |
57f1b61b | 21 | #include "discover-image.h" |
c7664c07 | 22 | #include "dissect-image.h" |
686d13b9 | 23 | #include "env-file.h" |
b6e953f2 | 24 | #include "env-util.h" |
3ffd4af2 | 25 | #include "fd-util.h" |
f4f15635 | 26 | #include "fs-util.h" |
a8fbdf54 | 27 | #include "hashmap.h" |
e2054217 | 28 | #include "hostname-setup.h" |
c7664c07 | 29 | #include "id128-util.h" |
64e89f56 | 30 | #include "lock-util.h" |
a8fbdf54 | 31 | #include "log.h" |
c7664c07 | 32 | #include "loop-util.h" |
546dbec5 | 33 | #include "macro.h" |
30535c16 | 34 | #include "mkdir.h" |
d8b4d14d | 35 | #include "nulstr-util.h" |
d58ad743 | 36 | #include "os-util.h" |
8e0b6570 | 37 | #include "path-util.h" |
c6878637 | 38 | #include "rm-rf.h" |
65ddc2c5 | 39 | #include "stat-util.h" |
8b43440b | 40 | #include "string-table.h" |
07630cea | 41 | #include "string-util.h" |
8e0b6570 | 42 | #include "strv.h" |
a8fbdf54 | 43 | #include "time-util.h" |
8e0b6570 | 44 | #include "utf8.h" |
89a5a90c | 45 | #include "xattr-util.h" |
cd61c3bf | 46 | |
5ef46e5f | 47 | static const char* const image_search_path[_IMAGE_CLASS_MAX] = { |
9bca4ae4 LP |
48 | [IMAGE_MACHINE] = "/etc/machines\0" /* only place symlinks here */ |
49 | "/run/machines\0" /* and here too */ | |
50 | "/var/lib/machines\0" /* the main place for images */ | |
51 | "/var/lib/container\0" /* legacy */ | |
52 | "/usr/local/lib/machines\0" | |
53 | "/usr/lib/machines\0", | |
5ef46e5f | 54 | |
9bca4ae4 LP |
55 | [IMAGE_PORTABLE] = "/etc/portables\0" /* only place symlinks here */ |
56 | "/run/portables\0" /* and here too */ | |
57 | "/var/lib/portables\0" /* the main place for images */ | |
58 | "/usr/local/lib/portables\0" | |
59 | "/usr/lib/portables\0", | |
60 | ||
61 | [IMAGE_EXTENSION] = "/etc/extensions\0" /* only place symlinks here */ | |
62 | "/run/extensions\0" /* and here too */ | |
63 | "/var/lib/extensions\0" /* the main place for images */ | |
64 | "/usr/local/lib/extensions\0" | |
65 | "/usr/lib/extensions\0", | |
5ef46e5f | 66 | }; |
c2ce6a3d | 67 | |
8301aa0b YW |
68 | static Image *image_free(Image *i) { |
69 | assert(i); | |
9614bb06 | 70 | |
cd61c3bf LP |
71 | free(i->name); |
72 | free(i->path); | |
c7664c07 LP |
73 | |
74 | free(i->hostname); | |
75 | strv_free(i->machine_info); | |
76 | strv_free(i->os_release); | |
bcf94222 | 77 | strv_free(i->extension_release); |
c7664c07 | 78 | |
6b430fdb | 79 | return mfree(i); |
cd61c3bf LP |
80 | } |
81 | ||
8301aa0b | 82 | DEFINE_TRIVIAL_REF_UNREF_FUNC(Image, image, image_free); |
b07ec5a1 YW |
83 | DEFINE_HASH_OPS_WITH_VALUE_DESTRUCTOR(image_hash_ops, char, string_hash_func, string_compare_func, |
84 | Image, image_unref); | |
9614bb06 | 85 | |
8e0b6570 LP |
86 | static char **image_settings_path(Image *image) { |
87 | _cleanup_strv_free_ char **l = NULL; | |
162f6477 LP |
88 | _cleanup_free_ char *fn = NULL; |
89 | size_t i = 0; | |
90 | int r; | |
8e0b6570 LP |
91 | |
92 | assert(image); | |
93 | ||
94 | l = new0(char*, 4); | |
95 | if (!l) | |
96 | return NULL; | |
97 | ||
162f6477 LP |
98 | fn = strjoin(image->name, ".nspawn"); |
99 | if (!fn) | |
100 | return NULL; | |
8e0b6570 | 101 | |
b910cc72 LP |
102 | FOREACH_STRING(s, "/etc/systemd/nspawn", "/run/systemd/nspawn") { |
103 | l[i] = path_join(s, fn); | |
8e0b6570 LP |
104 | if (!l[i]) |
105 | return NULL; | |
106 | ||
107 | i++; | |
108 | } | |
109 | ||
162f6477 LP |
110 | r = file_in_same_dir(image->path, fn, l + i); |
111 | if (r == -ENOMEM) | |
8e0b6570 | 112 | return NULL; |
162f6477 LP |
113 | if (r < 0) |
114 | log_debug_errno(r, "Failed to generate .nspawn settings path from image path, ignoring: %m"); | |
115 | ||
116 | strv_uniq(l); | |
8e0b6570 | 117 | |
ae2a15bc | 118 | return TAKE_PTR(l); |
8e0b6570 LP |
119 | } |
120 | ||
162f6477 LP |
121 | static int image_roothash_path(Image *image, char **ret) { |
122 | _cleanup_free_ char *fn = NULL; | |
bafbac4e LP |
123 | |
124 | assert(image); | |
125 | ||
162f6477 LP |
126 | fn = strjoin(image->name, ".roothash"); |
127 | if (!fn) | |
128 | return -ENOMEM; | |
bafbac4e | 129 | |
162f6477 | 130 | return file_in_same_dir(image->path, fn, ret); |
bafbac4e LP |
131 | } |
132 | ||
c2ce6a3d | 133 | static int image_new( |
cd61c3bf | 134 | ImageType t, |
3775e141 | 135 | ImageClass c, |
5fc7f358 | 136 | const char *pretty, |
cd61c3bf | 137 | const char *path, |
5fc7f358 | 138 | const char *filename, |
cd61c3bf | 139 | bool read_only, |
10f9c755 | 140 | usec_t crtime, |
cd61c3bf | 141 | usec_t mtime, |
c2ce6a3d | 142 | Image **ret) { |
cd61c3bf LP |
143 | |
144 | _cleanup_(image_unrefp) Image *i = NULL; | |
cd61c3bf | 145 | |
cd61c3bf LP |
146 | assert(t >= 0); |
147 | assert(t < _IMAGE_TYPE_MAX); | |
5fc7f358 LP |
148 | assert(pretty); |
149 | assert(filename); | |
c2ce6a3d | 150 | assert(ret); |
cd61c3bf | 151 | |
c2108701 | 152 | i = new(Image, 1); |
cd61c3bf LP |
153 | if (!i) |
154 | return -ENOMEM; | |
155 | ||
c2108701 LP |
156 | *i = (Image) { |
157 | .n_ref = 1, | |
158 | .type = t, | |
3775e141 | 159 | .class = c, |
c2108701 LP |
160 | .read_only = read_only, |
161 | .crtime = crtime, | |
162 | .mtime = mtime, | |
163 | .usage = UINT64_MAX, | |
164 | .usage_exclusive = UINT64_MAX, | |
165 | .limit = UINT64_MAX, | |
166 | .limit_exclusive = UINT64_MAX, | |
167 | }; | |
cd61c3bf | 168 | |
5fc7f358 | 169 | i->name = strdup(pretty); |
cd61c3bf LP |
170 | if (!i->name) |
171 | return -ENOMEM; | |
172 | ||
657ee2d8 | 173 | i->path = path_join(path, filename); |
5fc7f358 LP |
174 | if (!i->path) |
175 | return -ENOMEM; | |
176 | ||
4ff361cc | 177 | path_simplify(i->path); |
cd61c3bf | 178 | |
1cc6c93a | 179 | *ret = TAKE_PTR(i); |
c2ce6a3d | 180 | |
cd61c3bf LP |
181 | return 0; |
182 | } | |
183 | ||
4756c94e LP |
184 | static int extract_pretty(const char *path, const char *suffix, char **ret) { |
185 | _cleanup_free_ char *name = NULL; | |
186 | const char *p; | |
4756c94e LP |
187 | |
188 | assert(path); | |
189 | assert(ret); | |
190 | ||
191 | p = last_path_component(path); | |
4756c94e | 192 | |
e8bec624 | 193 | name = strdupcspn(p, "/"); |
4756c94e LP |
194 | if (!name) |
195 | return -ENOMEM; | |
196 | ||
197 | if (suffix) { | |
198 | char *e; | |
199 | ||
200 | e = endswith(name, suffix); | |
201 | if (!e) | |
202 | return -EINVAL; | |
203 | ||
204 | *e = 0; | |
205 | } | |
206 | ||
207 | if (!image_name_is_valid(name)) | |
208 | return -EINVAL; | |
209 | ||
210 | *ret = TAKE_PTR(name); | |
211 | return 0; | |
212 | } | |
213 | ||
5fc7f358 | 214 | static int image_make( |
3775e141 | 215 | ImageClass c, |
5fc7f358 LP |
216 | const char *pretty, |
217 | int dfd, | |
218 | const char *path, | |
219 | const char *filename, | |
3a6ce860 | 220 | const struct stat *st, |
5fc7f358 LP |
221 | Image **ret) { |
222 | ||
2570578d | 223 | _cleanup_free_ char *pretty_buffer = NULL, *parent = NULL; |
3a6ce860 | 224 | struct stat stbuf; |
5fc7f358 | 225 | bool read_only; |
cd61c3bf LP |
226 | int r; |
227 | ||
3a6ce860 | 228 | assert(dfd >= 0 || dfd == AT_FDCWD); |
bcb846f3 | 229 | assert(path || dfd == AT_FDCWD); |
5fc7f358 | 230 | assert(filename); |
cd61c3bf | 231 | |
eb38edce | 232 | /* We explicitly *do* follow symlinks here, since we want to allow symlinking trees, raw files and block |
3a6ce860 LP |
233 | * devices into /var/lib/machines/, and treat them normally. |
234 | * | |
235 | * This function returns -ENOENT if we can't find the image after all, and -EMEDIUMTYPE if it's not a file we | |
236 | * recognize. */ | |
cd61c3bf | 237 | |
3a6ce860 LP |
238 | if (!st) { |
239 | if (fstatat(dfd, filename, &stbuf, 0) < 0) | |
240 | return -errno; | |
241 | ||
242 | st = &stbuf; | |
243 | } | |
cd61c3bf | 244 | |
2570578d LP |
245 | if (!path) { |
246 | if (dfd == AT_FDCWD) | |
247 | (void) safe_getcwd(&parent); | |
248 | else | |
249 | (void) fd_get_path(dfd, &parent); | |
250 | } | |
bcb846f3 | 251 | |
5fc7f358 LP |
252 | read_only = |
253 | (path && path_startswith(path, "/usr")) || | |
08ff5529 | 254 | (faccessat(dfd, filename, W_OK, AT_EACCESS) < 0 && errno == EROFS); |
86e339c8 | 255 | |
3a6ce860 | 256 | if (S_ISDIR(st->st_mode)) { |
254d1313 | 257 | _cleanup_close_ int fd = -EBADF; |
01b72568 | 258 | unsigned file_attr = 0; |
02dcf215 | 259 | usec_t crtime = 0; |
cd61c3bf | 260 | |
c2ce6a3d | 261 | if (!ret) |
3a6ce860 | 262 | return 0; |
cd61c3bf | 263 | |
4756c94e LP |
264 | if (!pretty) { |
265 | r = extract_pretty(filename, NULL, &pretty_buffer); | |
266 | if (r < 0) | |
267 | return r; | |
268 | ||
269 | pretty = pretty_buffer; | |
270 | } | |
5fc7f358 | 271 | |
01b72568 LP |
272 | fd = openat(dfd, filename, O_CLOEXEC|O_NOCTTY|O_DIRECTORY); |
273 | if (fd < 0) | |
274 | return -errno; | |
275 | ||
674b04ff | 276 | if (btrfs_might_be_subvol(st)) { |
cd61c3bf | 277 | |
65ddc2c5 | 278 | r = fd_is_fs_type(fd, BTRFS_SUPER_MAGIC); |
21222ea5 LP |
279 | if (r < 0) |
280 | return r; | |
281 | if (r) { | |
10f9c755 | 282 | BtrfsSubvolInfo info; |
cd61c3bf | 283 | |
c2ce6a3d | 284 | /* It's a btrfs subvolume */ |
cd61c3bf | 285 | |
5bcd08db | 286 | r = btrfs_subvol_get_info_fd(fd, 0, &info); |
10f9c755 LP |
287 | if (r < 0) |
288 | return r; | |
c2ce6a3d LP |
289 | |
290 | r = image_new(IMAGE_SUBVOLUME, | |
3775e141 | 291 | c, |
5fc7f358 | 292 | pretty, |
c2ce6a3d | 293 | path, |
5fc7f358 LP |
294 | filename, |
295 | info.read_only || read_only, | |
10f9c755 | 296 | info.otime, |
c2ce6a3d | 297 | 0, |
c2ce6a3d LP |
298 | ret); |
299 | if (r < 0) | |
300 | return r; | |
301 | ||
5bcd08db LP |
302 | if (btrfs_quota_scan_ongoing(fd) == 0) { |
303 | BtrfsQuotaInfo quota; | |
b6b18498 | 304 | |
5bcd08db LP |
305 | r = btrfs_subvol_get_subtree_quota_fd(fd, 0, "a); |
306 | if (r >= 0) { | |
307 | (*ret)->usage = quota.referenced; | |
308 | (*ret)->usage_exclusive = quota.exclusive; | |
309 | ||
310 | (*ret)->limit = quota.referenced_max; | |
311 | (*ret)->limit_exclusive = quota.exclusive_max; | |
312 | } | |
b6b18498 LP |
313 | } |
314 | ||
3a6ce860 | 315 | return 0; |
cd61c3bf | 316 | } |
c2ce6a3d | 317 | } |
cd61c3bf | 318 | |
02dcf215 | 319 | /* Get directory creation time (not available everywhere, but that's OK */ |
12a7f04a | 320 | (void) fd_getcrtime(fd, &crtime); |
02dcf215 LP |
321 | |
322 | /* If the IMMUTABLE bit is set, we consider the directory read-only. Since the ioctl is not | |
01b72568 LP |
323 | * supported everywhere we ignore failures. */ |
324 | (void) read_attr_fd(fd, &file_attr); | |
cd61c3bf | 325 | |
01b72568 | 326 | /* It's just a normal directory. */ |
c2ce6a3d | 327 | r = image_new(IMAGE_DIRECTORY, |
3775e141 | 328 | c, |
5fc7f358 | 329 | pretty, |
c2ce6a3d | 330 | path, |
5fc7f358 | 331 | filename, |
01b72568 | 332 | read_only || (file_attr & FS_IMMUTABLE_FL), |
02dcf215 LP |
333 | crtime, |
334 | 0, /* we don't use mtime of stat() here, since it's not the time of last change of the tree, but only of the top-level dir */ | |
c2ce6a3d LP |
335 | ret); |
336 | if (r < 0) | |
337 | return r; | |
cd61c3bf | 338 | |
3a6ce860 | 339 | return 0; |
cd61c3bf | 340 | |
3a6ce860 | 341 | } else if (S_ISREG(st->st_mode) && endswith(filename, ".raw")) { |
10f9c755 | 342 | usec_t crtime = 0; |
cd61c3bf | 343 | |
aceac2f0 | 344 | /* It's a RAW disk image */ |
cd61c3bf | 345 | |
c2ce6a3d | 346 | if (!ret) |
3a6ce860 | 347 | return 0; |
cd61c3bf | 348 | |
c53e07e2 | 349 | (void) fd_getcrtime_at(dfd, filename, AT_SYMLINK_FOLLOW, &crtime); |
10f9c755 | 350 | |
4756c94e LP |
351 | if (!pretty) { |
352 | r = extract_pretty(filename, ".raw", &pretty_buffer); | |
353 | if (r < 0) | |
354 | return r; | |
355 | ||
356 | pretty = pretty_buffer; | |
357 | } | |
10f9c755 | 358 | |
aceac2f0 | 359 | r = image_new(IMAGE_RAW, |
3775e141 | 360 | c, |
5fc7f358 | 361 | pretty, |
c2ce6a3d | 362 | path, |
5fc7f358 | 363 | filename, |
3a6ce860 | 364 | !(st->st_mode & 0222) || read_only, |
10f9c755 | 365 | crtime, |
3a6ce860 | 366 | timespec_load(&st->st_mtim), |
c2ce6a3d LP |
367 | ret); |
368 | if (r < 0) | |
369 | return r; | |
cd61c3bf | 370 | |
3a6ce860 LP |
371 | (*ret)->usage = (*ret)->usage_exclusive = st->st_blocks * 512; |
372 | (*ret)->limit = (*ret)->limit_exclusive = st->st_size; | |
b6b18498 | 373 | |
3a6ce860 | 374 | return 0; |
eb38edce | 375 | |
3a6ce860 | 376 | } else if (S_ISBLK(st->st_mode)) { |
254d1313 | 377 | _cleanup_close_ int block_fd = -EBADF; |
eb38edce LP |
378 | uint64_t size = UINT64_MAX; |
379 | ||
380 | /* A block device */ | |
381 | ||
382 | if (!ret) | |
3a6ce860 | 383 | return 0; |
eb38edce | 384 | |
4756c94e LP |
385 | if (!pretty) { |
386 | r = extract_pretty(filename, NULL, &pretty_buffer); | |
387 | if (r < 0) | |
388 | return r; | |
389 | ||
390 | pretty = pretty_buffer; | |
391 | } | |
eb38edce LP |
392 | |
393 | block_fd = openat(dfd, filename, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); | |
394 | if (block_fd < 0) | |
2570578d | 395 | log_debug_errno(errno, "Failed to open block device %s/%s, ignoring: %m", path ?: strnull(parent), filename); |
eb38edce | 396 | else { |
3a6ce860 LP |
397 | /* Refresh stat data after opening the node */ |
398 | if (fstat(block_fd, &stbuf) < 0) | |
eb38edce | 399 | return -errno; |
3a6ce860 LP |
400 | st = &stbuf; |
401 | ||
402 | if (!S_ISBLK(st->st_mode)) /* Verify that what we opened is actually what we think it is */ | |
eb38edce LP |
403 | return -ENOTTY; |
404 | ||
405 | if (!read_only) { | |
406 | int state = 0; | |
407 | ||
408 | if (ioctl(block_fd, BLKROGET, &state) < 0) | |
2570578d | 409 | log_debug_errno(errno, "Failed to issue BLKROGET on device %s/%s, ignoring: %m", path ?: strnull(parent), filename); |
eb38edce LP |
410 | else if (state) |
411 | read_only = true; | |
412 | } | |
413 | ||
414 | if (ioctl(block_fd, BLKGETSIZE64, &size) < 0) | |
2570578d | 415 | log_debug_errno(errno, "Failed to issue BLKGETSIZE64 on device %s/%s, ignoring: %m", path ?: strnull(parent), filename); |
eb38edce LP |
416 | |
417 | block_fd = safe_close(block_fd); | |
418 | } | |
419 | ||
420 | r = image_new(IMAGE_BLOCK, | |
3775e141 | 421 | c, |
eb38edce LP |
422 | pretty, |
423 | path, | |
424 | filename, | |
3a6ce860 | 425 | !(st->st_mode & 0222) || read_only, |
eb38edce LP |
426 | 0, |
427 | 0, | |
428 | ret); | |
429 | if (r < 0) | |
430 | return r; | |
431 | ||
ed0cb346 | 432 | if (!IN_SET(size, 0, UINT64_MAX)) |
eb38edce LP |
433 | (*ret)->usage = (*ret)->usage_exclusive = (*ret)->limit = (*ret)->limit_exclusive = size; |
434 | ||
3a6ce860 | 435 | return 0; |
c2ce6a3d | 436 | } |
cd61c3bf | 437 | |
3a6ce860 | 438 | return -EMEDIUMTYPE; |
c2ce6a3d | 439 | } |
cd61c3bf | 440 | |
d577d4a4 LP |
441 | int image_find(ImageClass class, |
442 | const char *name, | |
443 | const char *root, | |
444 | Image **ret) { | |
445 | ||
c2ce6a3d | 446 | int r; |
cd61c3bf | 447 | |
5ef46e5f LP |
448 | assert(class >= 0); |
449 | assert(class < _IMAGE_CLASS_MAX); | |
c2ce6a3d | 450 | assert(name); |
cd61c3bf | 451 | |
c2ce6a3d LP |
452 | /* There are no images with invalid names */ |
453 | if (!image_name_is_valid(name)) | |
3a6ce860 | 454 | return -ENOENT; |
cd61c3bf | 455 | |
5ef46e5f | 456 | NULSTR_FOREACH(path, image_search_path[class]) { |
d577d4a4 | 457 | _cleanup_free_ char *resolved = NULL; |
c2ce6a3d | 458 | _cleanup_closedir_ DIR *d = NULL; |
3a6ce860 | 459 | struct stat st; |
d577d4a4 | 460 | int flags; |
cd61c3bf | 461 | |
d577d4a4 LP |
462 | r = chase_symlinks_and_opendir(path, root, CHASE_PREFIX_ROOT, &resolved, &d); |
463 | if (r == -ENOENT) | |
464 | continue; | |
465 | if (r < 0) | |
466 | return r; | |
cd61c3bf | 467 | |
d577d4a4 LP |
468 | /* As mentioned above, we follow symlinks on this fstatat(), because we want to permit people |
469 | * to symlink block devices into the search path. (For now, we disable that when operating | |
470 | * relative to some root directory.) */ | |
471 | flags = root ? AT_SYMLINK_NOFOLLOW : 0; | |
472 | if (fstatat(dirfd(d), name, &st, flags) < 0) { | |
aceac2f0 | 473 | _cleanup_free_ char *raw = NULL; |
5fc7f358 | 474 | |
3a6ce860 LP |
475 | if (errno != ENOENT) |
476 | return -errno; | |
477 | ||
b910cc72 | 478 | raw = strjoin(name, ".raw"); |
aceac2f0 | 479 | if (!raw) |
5fc7f358 LP |
480 | return -ENOMEM; |
481 | ||
d577d4a4 | 482 | if (fstatat(dirfd(d), raw, &st, flags) < 0) { |
3a6ce860 LP |
483 | if (errno == ENOENT) |
484 | continue; | |
485 | ||
486 | return -errno; | |
487 | } | |
488 | ||
489 | if (!S_ISREG(st.st_mode)) | |
5fc7f358 | 490 | continue; |
3a6ce860 | 491 | |
3775e141 | 492 | r = image_make(class, name, dirfd(d), resolved, raw, &st, ret); |
3a6ce860 LP |
493 | |
494 | } else { | |
495 | if (!S_ISDIR(st.st_mode) && !S_ISBLK(st.st_mode)) | |
496 | continue; | |
497 | ||
3775e141 | 498 | r = image_make(class, name, dirfd(d), resolved, name, &st, ret); |
5fc7f358 | 499 | } |
3a6ce860 LP |
500 | if (IN_SET(r, -ENOENT, -EMEDIUMTYPE)) |
501 | continue; | |
c2ce6a3d LP |
502 | if (r < 0) |
503 | return r; | |
cd61c3bf | 504 | |
cf604fd4 LP |
505 | if (ret) |
506 | (*ret)->discoverable = true; | |
507 | ||
c2ce6a3d LP |
508 | return 1; |
509 | } | |
510 | ||
cf604fd4 | 511 | if (class == IMAGE_MACHINE && streq(name, ".host")) { |
3775e141 | 512 | r = image_make(class, ".host", AT_FDCWD, NULL, empty_to_root(root), NULL, ret); |
cf604fd4 LP |
513 | if (r < 0) |
514 | return r; | |
515 | ||
516 | if (ret) | |
517 | (*ret)->discoverable = true; | |
518 | ||
519 | return r; | |
520 | } | |
5fc7f358 | 521 | |
3a6ce860 | 522 | return -ENOENT; |
c2ce6a3d LP |
523 | }; |
524 | ||
2ddf182b | 525 | int image_from_path(const char *path, Image **ret) { |
cf604fd4 LP |
526 | |
527 | /* Note that we don't set the 'discoverable' field of the returned object, because we don't check here whether | |
528 | * the image is in the image search path. And if it is we don't know if the path we used is actually not | |
3fe91079 | 529 | * overridden by another, different image earlier in the search path */ |
cf604fd4 | 530 | |
2ddf182b | 531 | if (path_equal(path, "/")) |
3775e141 | 532 | return image_make(IMAGE_MACHINE, ".host", AT_FDCWD, NULL, "/", NULL, ret); |
2ddf182b | 533 | |
3775e141 | 534 | return image_make(_IMAGE_CLASS_INVALID, NULL, AT_FDCWD, NULL, path, NULL, ret); |
2ddf182b LP |
535 | } |
536 | ||
d577d4a4 | 537 | int image_find_harder(ImageClass class, const char *name_or_path, const char *root, Image **ret) { |
2ddf182b | 538 | if (image_name_is_valid(name_or_path)) |
d577d4a4 | 539 | return image_find(class, name_or_path, root, ret); |
2ddf182b LP |
540 | |
541 | return image_from_path(name_or_path, ret); | |
542 | } | |
543 | ||
d577d4a4 LP |
544 | int image_discover( |
545 | ImageClass class, | |
546 | const char *root, | |
547 | Hashmap *h) { | |
548 | ||
c2ce6a3d LP |
549 | int r; |
550 | ||
5ef46e5f LP |
551 | assert(class >= 0); |
552 | assert(class < _IMAGE_CLASS_MAX); | |
c2ce6a3d LP |
553 | assert(h); |
554 | ||
5ef46e5f | 555 | NULSTR_FOREACH(path, image_search_path[class]) { |
d577d4a4 | 556 | _cleanup_free_ char *resolved = NULL; |
c2ce6a3d | 557 | _cleanup_closedir_ DIR *d = NULL; |
c2ce6a3d | 558 | |
d577d4a4 LP |
559 | r = chase_symlinks_and_opendir(path, root, CHASE_PREFIX_ROOT, &resolved, &d); |
560 | if (r == -ENOENT) | |
561 | continue; | |
562 | if (r < 0) | |
563 | return r; | |
c2ce6a3d LP |
564 | |
565 | FOREACH_DIRENT_ALL(de, d, return -errno) { | |
566 | _cleanup_(image_unrefp) Image *image = NULL; | |
1bf36bf9 | 567 | _cleanup_free_ char *truncated = NULL; |
3a6ce860 LP |
568 | const char *pretty; |
569 | struct stat st; | |
d577d4a4 | 570 | int flags; |
c2ce6a3d | 571 | |
1bf36bf9 | 572 | if (dot_or_dot_dot(de->d_name)) |
c2ce6a3d LP |
573 | continue; |
574 | ||
d577d4a4 LP |
575 | /* As mentioned above, we follow symlinks on this fstatat(), because we want to |
576 | * permit people to symlink block devices into the search path. */ | |
577 | flags = root ? AT_SYMLINK_NOFOLLOW : 0; | |
578 | if (fstatat(dirfd(d), de->d_name, &st, flags) < 0) { | |
3a6ce860 LP |
579 | if (errno == ENOENT) |
580 | continue; | |
581 | ||
582 | return -errno; | |
583 | } | |
584 | ||
585 | if (S_ISREG(st.st_mode)) { | |
586 | const char *e; | |
587 | ||
588 | e = endswith(de->d_name, ".raw"); | |
589 | if (!e) | |
590 | continue; | |
591 | ||
1bf36bf9 LP |
592 | truncated = strndup(de->d_name, e - de->d_name); |
593 | if (!truncated) | |
594 | return -ENOMEM; | |
595 | ||
596 | pretty = truncated; | |
3a6ce860 | 597 | } else if (S_ISDIR(st.st_mode) || S_ISBLK(st.st_mode)) |
1bf36bf9 | 598 | pretty = de->d_name; |
3a6ce860 LP |
599 | else |
600 | continue; | |
1bf36bf9 LP |
601 | |
602 | if (!image_name_is_valid(pretty)) | |
603 | continue; | |
604 | ||
605 | if (hashmap_contains(h, pretty)) | |
c2ce6a3d LP |
606 | continue; |
607 | ||
3775e141 | 608 | r = image_make(class, pretty, dirfd(d), resolved, de->d_name, &st, &image); |
3a6ce860 | 609 | if (IN_SET(r, -ENOENT, -EMEDIUMTYPE)) |
c2ce6a3d LP |
610 | continue; |
611 | if (r < 0) | |
612 | return r; | |
613 | ||
cf604fd4 LP |
614 | image->discoverable = true; |
615 | ||
c2ce6a3d LP |
616 | r = hashmap_put(h, image->name, image); |
617 | if (r < 0) | |
618 | return r; | |
619 | ||
620 | image = NULL; | |
cd61c3bf LP |
621 | } |
622 | } | |
623 | ||
5ef46e5f | 624 | if (class == IMAGE_MACHINE && !hashmap_contains(h, ".host")) { |
5fc7f358 LP |
625 | _cleanup_(image_unrefp) Image *image = NULL; |
626 | ||
3775e141 | 627 | r = image_make(IMAGE_MACHINE, ".host", AT_FDCWD, NULL, empty_to_root("/"), NULL, &image); |
5fc7f358 LP |
628 | if (r < 0) |
629 | return r; | |
630 | ||
cf604fd4 LP |
631 | image->discoverable = true; |
632 | ||
5fc7f358 LP |
633 | r = hashmap_put(h, image->name, image); |
634 | if (r < 0) | |
635 | return r; | |
636 | ||
637 | image = NULL; | |
5fc7f358 LP |
638 | } |
639 | ||
cd61c3bf LP |
640 | return 0; |
641 | } | |
642 | ||
08682124 | 643 | int image_remove(Image *i) { |
8e766630 | 644 | _cleanup_(release_lock_file) LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT; |
8e0b6570 | 645 | _cleanup_strv_free_ char **settings = NULL; |
bafbac4e | 646 | _cleanup_free_ char *roothash = NULL; |
30535c16 LP |
647 | int r; |
648 | ||
08682124 LP |
649 | assert(i); |
650 | ||
d94c2b06 | 651 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
08682124 LP |
652 | return -EROFS; |
653 | ||
8e0b6570 LP |
654 | settings = image_settings_path(i); |
655 | if (!settings) | |
656 | return -ENOMEM; | |
657 | ||
162f6477 LP |
658 | r = image_roothash_path(i, &roothash); |
659 | if (r < 0) | |
660 | return r; | |
bafbac4e | 661 | |
30535c16 LP |
662 | /* Make sure we don't interfere with a running nspawn */ |
663 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
664 | if (r < 0) | |
665 | return r; | |
666 | ||
ebd93cb6 LP |
667 | switch (i->type) { |
668 | ||
669 | case IMAGE_SUBVOLUME: | |
9fb0b9c7 LP |
670 | |
671 | /* Let's unlink first, maybe it is a symlink? If that works we are happy. Otherwise, let's get out the | |
672 | * big guns */ | |
673 | if (unlink(i->path) < 0) { | |
674 | r = btrfs_subvol_remove(i->path, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA); | |
675 | if (r < 0) | |
676 | return r; | |
677 | } | |
678 | ||
8e0b6570 | 679 | break; |
ebd93cb6 LP |
680 | |
681 | case IMAGE_DIRECTORY: | |
01b72568 | 682 | /* Allow deletion of read-only directories */ |
db9a4254 | 683 | (void) chattr_path(i->path, 0, FS_IMMUTABLE_FL, NULL); |
8e0b6570 LP |
684 | r = rm_rf(i->path, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME); |
685 | if (r < 0) | |
686 | return r; | |
687 | ||
688 | break; | |
01b72568 | 689 | |
eb38edce LP |
690 | case IMAGE_BLOCK: |
691 | ||
692 | /* If this is inside of /dev, then it's a real block device, hence let's not touch the device node | |
693 | * itself (but let's remove the stuff stored alongside it). If it's anywhere else, let's try to unlink | |
694 | * the thing (it's most likely a symlink after all). */ | |
695 | ||
696 | if (path_startswith(i->path, "/dev")) | |
697 | break; | |
698 | ||
4831981d | 699 | _fallthrough_; |
aceac2f0 | 700 | case IMAGE_RAW: |
41d1ed05 LP |
701 | if (unlink(i->path) < 0) |
702 | return -errno; | |
8e0b6570 | 703 | break; |
ebd93cb6 LP |
704 | |
705 | default: | |
15411c0c | 706 | return -EOPNOTSUPP; |
ebd93cb6 | 707 | } |
8e0b6570 | 708 | |
de010b0b | 709 | STRV_FOREACH(j, settings) |
8e0b6570 LP |
710 | if (unlink(*j) < 0 && errno != ENOENT) |
711 | log_debug_errno(errno, "Failed to unlink %s, ignoring: %m", *j); | |
8e0b6570 | 712 | |
bafbac4e LP |
713 | if (unlink(roothash) < 0 && errno != ENOENT) |
714 | log_debug_errno(errno, "Failed to unlink %s, ignoring: %m", roothash); | |
715 | ||
8e0b6570 LP |
716 | return 0; |
717 | } | |
718 | ||
bafbac4e | 719 | static int rename_auxiliary_file(const char *path, const char *new_name, const char *suffix) { |
162f6477 LP |
720 | _cleanup_free_ char *fn = NULL, *rs = NULL; |
721 | int r; | |
8e0b6570 | 722 | |
162f6477 LP |
723 | fn = strjoin(new_name, suffix); |
724 | if (!fn) | |
8e0b6570 LP |
725 | return -ENOMEM; |
726 | ||
162f6477 LP |
727 | r = file_in_same_dir(path, fn, &rs); |
728 | if (r < 0) | |
729 | return r; | |
730 | ||
8e0b6570 | 731 | return rename_noreplace(AT_FDCWD, path, AT_FDCWD, rs); |
ebd93cb6 LP |
732 | } |
733 | ||
734 | int image_rename(Image *i, const char *new_name) { | |
8e766630 | 735 | _cleanup_(release_lock_file) LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT, name_lock = LOCK_FILE_INIT; |
bafbac4e | 736 | _cleanup_free_ char *new_path = NULL, *nn = NULL, *roothash = NULL; |
8e0b6570 | 737 | _cleanup_strv_free_ char **settings = NULL; |
01b72568 | 738 | unsigned file_attr = 0; |
ebd93cb6 LP |
739 | int r; |
740 | ||
741 | assert(i); | |
742 | ||
743 | if (!image_name_is_valid(new_name)) | |
744 | return -EINVAL; | |
745 | ||
d94c2b06 | 746 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
ebd93cb6 LP |
747 | return -EROFS; |
748 | ||
8e0b6570 LP |
749 | settings = image_settings_path(i); |
750 | if (!settings) | |
751 | return -ENOMEM; | |
752 | ||
162f6477 LP |
753 | r = image_roothash_path(i, &roothash); |
754 | if (r < 0) | |
755 | return r; | |
bafbac4e | 756 | |
30535c16 LP |
757 | /* Make sure we don't interfere with a running nspawn */ |
758 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
759 | if (r < 0) | |
760 | return r; | |
761 | ||
762 | /* Make sure nobody takes the new name, between the time we | |
763 | * checked it is currently unused in all search paths, and the | |
f8e2f4d6 | 764 | * time we take possession of it */ |
30535c16 LP |
765 | r = image_name_lock(new_name, LOCK_EX|LOCK_NB, &name_lock); |
766 | if (r < 0) | |
767 | return r; | |
768 | ||
d577d4a4 | 769 | r = image_find(IMAGE_MACHINE, new_name, NULL, NULL); |
3a6ce860 | 770 | if (r >= 0) |
ebd93cb6 | 771 | return -EEXIST; |
3a6ce860 LP |
772 | if (r != -ENOENT) |
773 | return r; | |
ebd93cb6 LP |
774 | |
775 | switch (i->type) { | |
776 | ||
ebd93cb6 | 777 | case IMAGE_DIRECTORY: |
01b72568 LP |
778 | /* Turn of the immutable bit while we rename the image, so that we can rename it */ |
779 | (void) read_attr_path(i->path, &file_attr); | |
780 | ||
781 | if (file_attr & FS_IMMUTABLE_FL) | |
db9a4254 | 782 | (void) chattr_path(i->path, 0, FS_IMMUTABLE_FL, NULL); |
01b72568 | 783 | |
4831981d | 784 | _fallthrough_; |
01b72568 | 785 | case IMAGE_SUBVOLUME: |
162f6477 | 786 | r = file_in_same_dir(i->path, new_name, &new_path); |
ebd93cb6 LP |
787 | break; |
788 | ||
eb38edce LP |
789 | case IMAGE_BLOCK: |
790 | ||
791 | /* Refuse renaming raw block devices in /dev, the names are picked by udev after all. */ | |
792 | if (path_startswith(i->path, "/dev")) | |
793 | return -EROFS; | |
794 | ||
162f6477 | 795 | r = file_in_same_dir(i->path, new_name, &new_path); |
eb38edce LP |
796 | break; |
797 | ||
aceac2f0 | 798 | case IMAGE_RAW: { |
ebd93cb6 LP |
799 | const char *fn; |
800 | ||
63c372cb | 801 | fn = strjoina(new_name, ".raw"); |
162f6477 LP |
802 | |
803 | r = file_in_same_dir(i->path, fn, &new_path); | |
ebd93cb6 LP |
804 | break; |
805 | } | |
806 | ||
807 | default: | |
15411c0c | 808 | return -EOPNOTSUPP; |
ebd93cb6 | 809 | } |
162f6477 LP |
810 | if (r < 0) |
811 | return r; | |
ebd93cb6 LP |
812 | |
813 | nn = strdup(new_name); | |
814 | if (!nn) | |
815 | return -ENOMEM; | |
816 | ||
f85ef957 AC |
817 | r = rename_noreplace(AT_FDCWD, i->path, AT_FDCWD, new_path); |
818 | if (r < 0) | |
819 | return r; | |
ebd93cb6 | 820 | |
01b72568 LP |
821 | /* Restore the immutable bit, if it was set before */ |
822 | if (file_attr & FS_IMMUTABLE_FL) | |
db9a4254 | 823 | (void) chattr_path(new_path, FS_IMMUTABLE_FL, FS_IMMUTABLE_FL, NULL); |
01b72568 | 824 | |
f9ecfd3b DL |
825 | free_and_replace(i->path, new_path); |
826 | free_and_replace(i->name, nn); | |
ebd93cb6 | 827 | |
8e0b6570 | 828 | STRV_FOREACH(j, settings) { |
bafbac4e | 829 | r = rename_auxiliary_file(*j, new_name, ".nspawn"); |
8e0b6570 LP |
830 | if (r < 0 && r != -ENOENT) |
831 | log_debug_errno(r, "Failed to rename settings file %s, ignoring: %m", *j); | |
832 | } | |
833 | ||
bafbac4e LP |
834 | r = rename_auxiliary_file(roothash, new_name, ".roothash"); |
835 | if (r < 0 && r != -ENOENT) | |
836 | log_debug_errno(r, "Failed to rename roothash file %s, ignoring: %m", roothash); | |
837 | ||
ebd93cb6 LP |
838 | return 0; |
839 | } | |
840 | ||
bafbac4e | 841 | static int clone_auxiliary_file(const char *path, const char *new_name, const char *suffix) { |
162f6477 LP |
842 | _cleanup_free_ char *fn = NULL, *rs = NULL; |
843 | int r; | |
8e0b6570 | 844 | |
162f6477 LP |
845 | fn = strjoin(new_name, suffix); |
846 | if (!fn) | |
8e0b6570 LP |
847 | return -ENOMEM; |
848 | ||
162f6477 LP |
849 | r = file_in_same_dir(path, fn, &rs); |
850 | if (r < 0) | |
851 | return r; | |
852 | ||
7c2f5495 | 853 | return copy_file_atomic(path, rs, 0664, COPY_REFLINK); |
8e0b6570 LP |
854 | } |
855 | ||
ebd93cb6 | 856 | int image_clone(Image *i, const char *new_name, bool read_only) { |
8e766630 | 857 | _cleanup_(release_lock_file) LockFile name_lock = LOCK_FILE_INIT; |
8e0b6570 | 858 | _cleanup_strv_free_ char **settings = NULL; |
bafbac4e | 859 | _cleanup_free_ char *roothash = NULL; |
ebd93cb6 LP |
860 | const char *new_path; |
861 | int r; | |
862 | ||
863 | assert(i); | |
864 | ||
865 | if (!image_name_is_valid(new_name)) | |
866 | return -EINVAL; | |
867 | ||
8e0b6570 LP |
868 | settings = image_settings_path(i); |
869 | if (!settings) | |
870 | return -ENOMEM; | |
871 | ||
162f6477 LP |
872 | r = image_roothash_path(i, &roothash); |
873 | if (r < 0) | |
874 | return r; | |
bafbac4e | 875 | |
30535c16 LP |
876 | /* Make sure nobody takes the new name, between the time we |
877 | * checked it is currently unused in all search paths, and the | |
f8e2f4d6 | 878 | * time we take possession of it */ |
30535c16 LP |
879 | r = image_name_lock(new_name, LOCK_EX|LOCK_NB, &name_lock); |
880 | if (r < 0) | |
881 | return r; | |
882 | ||
d577d4a4 | 883 | r = image_find(IMAGE_MACHINE, new_name, NULL, NULL); |
3a6ce860 | 884 | if (r >= 0) |
ebd93cb6 | 885 | return -EEXIST; |
3a6ce860 LP |
886 | if (r != -ENOENT) |
887 | return r; | |
ebd93cb6 LP |
888 | |
889 | switch (i->type) { | |
890 | ||
891 | case IMAGE_SUBVOLUME: | |
892 | case IMAGE_DIRECTORY: | |
9a50e3ca | 893 | /* If we can we'll always try to create a new btrfs subvolume here, even if the source is a plain |
13e785f7 | 894 | * directory. */ |
9a50e3ca | 895 | |
63c372cb | 896 | new_path = strjoina("/var/lib/machines/", new_name); |
ebd93cb6 | 897 | |
17cbb288 LP |
898 | r = btrfs_subvol_snapshot(i->path, new_path, |
899 | (read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | | |
900 | BTRFS_SNAPSHOT_FALLBACK_COPY | | |
901 | BTRFS_SNAPSHOT_FALLBACK_DIRECTORY | | |
902 | BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE | | |
903 | BTRFS_SNAPSHOT_RECURSIVE | | |
904 | BTRFS_SNAPSHOT_QUOTA); | |
905 | if (r >= 0) | |
9a50e3ca | 906 | /* Enable "subtree" quotas for the copy, if we didn't copy any quota from the source. */ |
8120ee28 | 907 | (void) btrfs_subvol_auto_qgroup(new_path, 0, true); |
5bcd08db | 908 | |
ebd93cb6 LP |
909 | break; |
910 | ||
aceac2f0 | 911 | case IMAGE_RAW: |
63c372cb | 912 | new_path = strjoina("/var/lib/machines/", new_name, ".raw"); |
ebd93cb6 | 913 | |
7c2f5495 DDM |
914 | r = copy_file_atomic_full(i->path, new_path, read_only ? 0444 : 0644, FS_NOCOW_FL, FS_NOCOW_FL, |
915 | COPY_REFLINK|COPY_CRTIME, NULL, NULL); | |
ebd93cb6 LP |
916 | break; |
917 | ||
eb38edce | 918 | case IMAGE_BLOCK: |
ebd93cb6 | 919 | default: |
15411c0c | 920 | return -EOPNOTSUPP; |
ebd93cb6 LP |
921 | } |
922 | ||
923 | if (r < 0) | |
924 | return r; | |
925 | ||
8e0b6570 | 926 | STRV_FOREACH(j, settings) { |
bafbac4e | 927 | r = clone_auxiliary_file(*j, new_name, ".nspawn"); |
8e0b6570 LP |
928 | if (r < 0 && r != -ENOENT) |
929 | log_debug_errno(r, "Failed to clone settings %s, ignoring: %m", *j); | |
930 | } | |
931 | ||
bafbac4e LP |
932 | r = clone_auxiliary_file(roothash, new_name, ".roothash"); |
933 | if (r < 0 && r != -ENOENT) | |
934 | log_debug_errno(r, "Failed to clone root hash file %s, ignoring: %m", roothash); | |
935 | ||
ebd93cb6 LP |
936 | return 0; |
937 | } | |
938 | ||
939 | int image_read_only(Image *i, bool b) { | |
8e766630 | 940 | _cleanup_(release_lock_file) LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT; |
ebd93cb6 | 941 | int r; |
c7664c07 | 942 | |
ebd93cb6 LP |
943 | assert(i); |
944 | ||
d94c2b06 | 945 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
ebd93cb6 LP |
946 | return -EROFS; |
947 | ||
30535c16 LP |
948 | /* Make sure we don't interfere with a running nspawn */ |
949 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
950 | if (r < 0) | |
951 | return r; | |
952 | ||
ebd93cb6 LP |
953 | switch (i->type) { |
954 | ||
955 | case IMAGE_SUBVOLUME: | |
5bcd08db LP |
956 | |
957 | /* Note that we set the flag only on the top-level | |
958 | * subvolume of the image. */ | |
959 | ||
ebd93cb6 LP |
960 | r = btrfs_subvol_set_read_only(i->path, b); |
961 | if (r < 0) | |
962 | return r; | |
01b72568 LP |
963 | |
964 | break; | |
965 | ||
966 | case IMAGE_DIRECTORY: | |
967 | /* For simple directory trees we cannot use the access | |
968 | mode of the top-level directory, since it has an | |
969 | effect on the container itself. However, we can | |
970 | use the "immutable" flag, to at least make the | |
971 | top-level directory read-only. It's not as good as | |
972 | a read-only subvolume, but at least something, and | |
13e785f7 | 973 | we can read the value back. */ |
01b72568 | 974 | |
db9a4254 | 975 | r = chattr_path(i->path, b ? FS_IMMUTABLE_FL : 0, FS_IMMUTABLE_FL, NULL); |
01b72568 LP |
976 | if (r < 0) |
977 | return r; | |
978 | ||
ebd93cb6 LP |
979 | break; |
980 | ||
aceac2f0 | 981 | case IMAGE_RAW: { |
ebd93cb6 LP |
982 | struct stat st; |
983 | ||
984 | if (stat(i->path, &st) < 0) | |
985 | return -errno; | |
986 | ||
987 | if (chmod(i->path, (st.st_mode & 0444) | (b ? 0000 : 0200)) < 0) | |
988 | return -errno; | |
f2068bcc LP |
989 | |
990 | /* If the images is now read-only, it's a good time to | |
991 | * defrag it, given that no write patterns will | |
992 | * fragment it again. */ | |
993 | if (b) | |
994 | (void) btrfs_defrag(i->path); | |
ebd93cb6 LP |
995 | break; |
996 | } | |
997 | ||
eb38edce | 998 | case IMAGE_BLOCK: { |
254d1313 | 999 | _cleanup_close_ int fd = -EBADF; |
eb38edce LP |
1000 | struct stat st; |
1001 | int state = b; | |
1002 | ||
1003 | fd = open(i->path, O_CLOEXEC|O_RDONLY|O_NONBLOCK|O_NOCTTY); | |
1004 | if (fd < 0) | |
1005 | return -errno; | |
1006 | ||
1007 | if (fstat(fd, &st) < 0) | |
1008 | return -errno; | |
1009 | if (!S_ISBLK(st.st_mode)) | |
1010 | return -ENOTTY; | |
1011 | ||
1012 | if (ioctl(fd, BLKROSET, &state) < 0) | |
1013 | return -errno; | |
1014 | ||
1015 | break; | |
1016 | } | |
1017 | ||
ebd93cb6 | 1018 | default: |
15411c0c | 1019 | return -EOPNOTSUPP; |
ebd93cb6 LP |
1020 | } |
1021 | ||
1022 | return 0; | |
08682124 LP |
1023 | } |
1024 | ||
30535c16 LP |
1025 | int image_path_lock(const char *path, int operation, LockFile *global, LockFile *local) { |
1026 | _cleanup_free_ char *p = NULL; | |
1027 | LockFile t = LOCK_FILE_INIT; | |
1028 | struct stat st; | |
f25bed67 | 1029 | bool exclusive; |
30535c16 LP |
1030 | int r; |
1031 | ||
1032 | assert(path); | |
1033 | assert(global); | |
1034 | assert(local); | |
1035 | ||
f25bed67 LP |
1036 | /* Locks an image path. This actually creates two locks: one "local" one, next to the image path |
1037 | * itself, which might be shared via NFS. And another "global" one, in /run, that uses the | |
1038 | * device/inode number. This has the benefit that we can even lock a tree that is a mount point, | |
1039 | * correctly. */ | |
30535c16 | 1040 | |
30535c16 LP |
1041 | if (!path_is_absolute(path)) |
1042 | return -EINVAL; | |
1043 | ||
f25bed67 LP |
1044 | switch (operation & (LOCK_SH|LOCK_EX)) { |
1045 | case LOCK_SH: | |
1046 | exclusive = false; | |
1047 | break; | |
1048 | case LOCK_EX: | |
1049 | exclusive = true; | |
1050 | break; | |
1051 | default: | |
1052 | return -EINVAL; | |
1053 | } | |
1054 | ||
b6e953f2 LP |
1055 | if (getenv_bool("SYSTEMD_NSPAWN_LOCK") == 0) { |
1056 | *local = *global = (LockFile) LOCK_FILE_INIT; | |
1057 | return 0; | |
1058 | } | |
1059 | ||
f25bed67 LP |
1060 | /* Prohibit taking exclusive locks on the host image. We can't allow this, since we ourselves are |
1061 | * running off it after all, and we don't want any images to manipulate the host image. We make an | |
1062 | * exception for shared locks however: we allow those (and make them NOPs since there's no point in | |
1063 | * taking them if there can't be exclusive locks). Strictly speaking these are questionable as well, | |
1064 | * since it means changes made to the host might propagate to the container as they happen (and a | |
1065 | * shared lock kinda suggests that no changes happen at all while it is in place), but it's too | |
1066 | * useful not to allow read-only containers off the host root, hence let's support this, and trust | |
1067 | * the user to do the right thing with this. */ | |
1068 | if (path_equal(path, "/")) { | |
1069 | if (exclusive) | |
1070 | return -EBUSY; | |
1071 | ||
1072 | *local = *global = (LockFile) LOCK_FILE_INIT; | |
1073 | return 0; | |
1074 | } | |
b6e953f2 | 1075 | |
30535c16 | 1076 | if (stat(path, &st) >= 0) { |
eb38edce LP |
1077 | if (S_ISBLK(st.st_mode)) |
1078 | r = asprintf(&p, "/run/systemd/nspawn/locks/block-%u:%u", major(st.st_rdev), minor(st.st_rdev)); | |
1079 | else if (S_ISDIR(st.st_mode) || S_ISREG(st.st_mode)) | |
1080 | r = asprintf(&p, "/run/systemd/nspawn/locks/inode-%lu:%lu", (unsigned long) st.st_dev, (unsigned long) st.st_ino); | |
1081 | else | |
1082 | return -ENOTTY; | |
eb38edce | 1083 | if (r < 0) |
30535c16 LP |
1084 | return -ENOMEM; |
1085 | } | |
1086 | ||
f25bed67 LP |
1087 | /* For block devices we don't need the "local" lock, as the major/minor lock above should be |
1088 | * sufficient, since block devices are host local anyway. */ | |
1089 | if (!path_startswith(path, "/dev/")) { | |
eb38edce | 1090 | r = make_lock_file_for(path, operation, &t); |
8be17c9b | 1091 | if (r < 0) { |
f25bed67 | 1092 | if (!exclusive && r == -EROFS) |
771b7ead | 1093 | log_debug_errno(r, "Failed to create shared lock for '%s', ignoring: %m", path); |
8be17c9b LT |
1094 | else |
1095 | return r; | |
1096 | } | |
eb38edce | 1097 | } |
30535c16 LP |
1098 | |
1099 | if (p) { | |
6e5dcce4 | 1100 | (void) mkdir_p("/run/systemd/nspawn/locks", 0700); |
30535c16 LP |
1101 | |
1102 | r = make_lock_file(p, operation, global); | |
1103 | if (r < 0) { | |
1104 | release_lock_file(&t); | |
1105 | return r; | |
1106 | } | |
546dbec5 LP |
1107 | } else |
1108 | *global = (LockFile) LOCK_FILE_INIT; | |
30535c16 LP |
1109 | |
1110 | *local = t; | |
1111 | return 0; | |
1112 | } | |
1113 | ||
cb81cd80 | 1114 | int image_set_limit(Image *i, uint64_t referenced_max) { |
d6ce17c7 LP |
1115 | assert(i); |
1116 | ||
d94c2b06 | 1117 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
d6ce17c7 LP |
1118 | return -EROFS; |
1119 | ||
1120 | if (i->type != IMAGE_SUBVOLUME) | |
15411c0c | 1121 | return -EOPNOTSUPP; |
d6ce17c7 | 1122 | |
5bcd08db LP |
1123 | /* We set the quota both for the subvolume as well as for the |
1124 | * subtree. The latter is mostly for historical reasons, since | |
1125 | * we didn't use to have a concept of subtree quota, and hence | |
1126 | * only modified the subvolume quota. */ | |
1127 | ||
1128 | (void) btrfs_qgroup_set_limit(i->path, 0, referenced_max); | |
1129 | (void) btrfs_subvol_auto_qgroup(i->path, 0, true); | |
1130 | return btrfs_subvol_set_subtree_quota_limit(i->path, 0, referenced_max); | |
d6ce17c7 LP |
1131 | } |
1132 | ||
c7664c07 | 1133 | int image_read_metadata(Image *i) { |
8e766630 | 1134 | _cleanup_(release_lock_file) LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT; |
c7664c07 LP |
1135 | int r; |
1136 | ||
1137 | assert(i); | |
1138 | ||
1139 | r = image_path_lock(i->path, LOCK_SH|LOCK_NB, &global_lock, &local_lock); | |
1140 | if (r < 0) | |
1141 | return r; | |
1142 | ||
1143 | switch (i->type) { | |
1144 | ||
1145 | case IMAGE_SUBVOLUME: | |
1146 | case IMAGE_DIRECTORY: { | |
bcf94222 | 1147 | _cleanup_strv_free_ char **machine_info = NULL, **os_release = NULL, **extension_release = NULL; |
c7664c07 LP |
1148 | sd_id128_t machine_id = SD_ID128_NULL; |
1149 | _cleanup_free_ char *hostname = NULL; | |
1150 | _cleanup_free_ char *path = NULL; | |
1151 | ||
a5648b80 | 1152 | r = chase_symlinks("/etc/hostname", i->path, CHASE_PREFIX_ROOT|CHASE_TRAIL_SLASH, &path, NULL); |
c7664c07 LP |
1153 | if (r < 0 && r != -ENOENT) |
1154 | log_debug_errno(r, "Failed to chase /etc/hostname in image %s: %m", i->name); | |
1155 | else if (r >= 0) { | |
1156 | r = read_etc_hostname(path, &hostname); | |
1157 | if (r < 0) | |
1158 | log_debug_errno(errno, "Failed to read /etc/hostname of image %s: %m", i->name); | |
1159 | } | |
1160 | ||
1161 | path = mfree(path); | |
1162 | ||
a5648b80 | 1163 | r = chase_symlinks("/etc/machine-id", i->path, CHASE_PREFIX_ROOT|CHASE_TRAIL_SLASH, &path, NULL); |
c7664c07 LP |
1164 | if (r < 0 && r != -ENOENT) |
1165 | log_debug_errno(r, "Failed to chase /etc/machine-id in image %s: %m", i->name); | |
1166 | else if (r >= 0) { | |
254d1313 | 1167 | _cleanup_close_ int fd = -EBADF; |
c7664c07 LP |
1168 | |
1169 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY); | |
1170 | if (fd < 0) | |
1171 | log_debug_errno(errno, "Failed to open %s: %m", path); | |
1172 | else { | |
057bf780 | 1173 | r = id128_read_fd(fd, ID128_FORMAT_PLAIN, &machine_id); |
c7664c07 LP |
1174 | if (r < 0) |
1175 | log_debug_errno(r, "Image %s contains invalid machine ID.", i->name); | |
1176 | } | |
1177 | } | |
1178 | ||
1179 | path = mfree(path); | |
1180 | ||
a5648b80 | 1181 | r = chase_symlinks("/etc/machine-info", i->path, CHASE_PREFIX_ROOT|CHASE_TRAIL_SLASH, &path, NULL); |
c7664c07 LP |
1182 | if (r < 0 && r != -ENOENT) |
1183 | log_debug_errno(r, "Failed to chase /etc/machine-info in image %s: %m", i->name); | |
1184 | else if (r >= 0) { | |
aa8fbc74 | 1185 | r = load_env_file_pairs(NULL, path, &machine_info); |
c7664c07 LP |
1186 | if (r < 0) |
1187 | log_debug_errno(r, "Failed to parse machine-info data of %s: %m", i->name); | |
1188 | } | |
1189 | ||
d58ad743 LP |
1190 | r = load_os_release_pairs(i->path, &os_release); |
1191 | if (r < 0) | |
1192 | log_debug_errno(r, "Failed to read os-release in image, ignoring: %m"); | |
c7664c07 | 1193 | |
06768b90 | 1194 | r = load_extension_release_pairs(i->path, i->name, /* relax_extension_release_check= */ false, &extension_release); |
bcf94222 LB |
1195 | if (r < 0) |
1196 | log_debug_errno(r, "Failed to read extension-release in image, ignoring: %m"); | |
1197 | ||
c7664c07 LP |
1198 | free_and_replace(i->hostname, hostname); |
1199 | i->machine_id = machine_id; | |
1200 | strv_free_and_replace(i->machine_info, machine_info); | |
1201 | strv_free_and_replace(i->os_release, os_release); | |
bcf94222 | 1202 | strv_free_and_replace(i->extension_release, extension_release); |
c7664c07 LP |
1203 | |
1204 | break; | |
1205 | } | |
1206 | ||
1207 | case IMAGE_RAW: | |
1208 | case IMAGE_BLOCK: { | |
1209 | _cleanup_(loop_device_unrefp) LoopDevice *d = NULL; | |
1210 | _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; | |
1211 | ||
22ee78a8 | 1212 | r = loop_device_make_by_path(i->path, O_RDONLY, /* sector_size= */ UINT32_MAX, LO_FLAGS_PARTSCAN, LOCK_SH, &d); |
41bc4849 LP |
1213 | if (r < 0) |
1214 | return r; | |
1215 | ||
bad31660 YW |
1216 | r = dissect_loop_device( |
1217 | d, | |
d04faa4e | 1218 | NULL, NULL, |
4b5de5dd LP |
1219 | DISSECT_IMAGE_GENERIC_ROOT | |
1220 | DISSECT_IMAGE_REQUIRE_ROOT | | |
1221 | DISSECT_IMAGE_RELAX_VAR_CHECK | | |
f6f4ec79 | 1222 | DISSECT_IMAGE_READ_ONLY | |
73d88b80 LP |
1223 | DISSECT_IMAGE_USR_NO_ROOT | |
1224 | DISSECT_IMAGE_ADD_PARTITION_DEVICES | | |
1225 | DISSECT_IMAGE_PIN_PARTITION_DEVICES, | |
75dc190d | 1226 | &m); |
c7664c07 LP |
1227 | if (r < 0) |
1228 | return r; | |
1229 | ||
22847508 ZJS |
1230 | r = dissected_image_acquire_metadata(m, |
1231 | DISSECT_IMAGE_VALIDATE_OS | | |
1232 | DISSECT_IMAGE_VALIDATE_OS_EXT); | |
c7664c07 LP |
1233 | if (r < 0) |
1234 | return r; | |
1235 | ||
1236 | free_and_replace(i->hostname, m->hostname); | |
1237 | i->machine_id = m->machine_id; | |
1238 | strv_free_and_replace(i->machine_info, m->machine_info); | |
1239 | strv_free_and_replace(i->os_release, m->os_release); | |
bcf94222 | 1240 | strv_free_and_replace(i->extension_release, m->extension_release); |
c7664c07 LP |
1241 | |
1242 | break; | |
1243 | } | |
1244 | ||
1245 | default: | |
1246 | return -EOPNOTSUPP; | |
1247 | } | |
1248 | ||
1249 | i->metadata_valid = true; | |
1250 | ||
1251 | return 0; | |
1252 | } | |
1253 | ||
30535c16 | 1254 | int image_name_lock(const char *name, int operation, LockFile *ret) { |
99d97afc LP |
1255 | const char *p; |
1256 | ||
30535c16 LP |
1257 | assert(name); |
1258 | assert(ret); | |
1259 | ||
1260 | /* Locks an image name, regardless of the precise path used. */ | |
1261 | ||
99d97afc LP |
1262 | if (streq(name, ".host")) |
1263 | return -EBUSY; | |
1264 | ||
30535c16 LP |
1265 | if (!image_name_is_valid(name)) |
1266 | return -EINVAL; | |
1267 | ||
b6e953f2 LP |
1268 | if (getenv_bool("SYSTEMD_NSPAWN_LOCK") == 0) { |
1269 | *ret = (LockFile) LOCK_FILE_INIT; | |
1270 | return 0; | |
1271 | } | |
1272 | ||
6e5dcce4 | 1273 | (void) mkdir_p("/run/systemd/nspawn/locks", 0700); |
99d97afc LP |
1274 | |
1275 | p = strjoina("/run/systemd/nspawn/locks/name-", name); | |
30535c16 LP |
1276 | return make_lock_file(p, operation, ret); |
1277 | } | |
1278 | ||
d577d4a4 LP |
1279 | bool image_in_search_path( |
1280 | ImageClass class, | |
1281 | const char *root, | |
1282 | const char *image) { | |
1283 | ||
ace9ab19 LP |
1284 | assert(image); |
1285 | ||
1286 | NULSTR_FOREACH(path, image_search_path[class]) { | |
d577d4a4 | 1287 | const char *p, *q; |
ace9ab19 LP |
1288 | size_t k; |
1289 | ||
d577d4a4 LP |
1290 | if (!empty_or_root(root)) { |
1291 | q = path_startswith(path, root); | |
1292 | if (!q) | |
1293 | continue; | |
1294 | } else | |
1295 | q = path; | |
1296 | ||
1297 | p = path_startswith(q, path); | |
ace9ab19 LP |
1298 | if (!p) |
1299 | continue; | |
1300 | ||
1301 | /* Make sure there's a filename following */ | |
1302 | k = strcspn(p, "/"); | |
1303 | if (k == 0) | |
1304 | continue; | |
1305 | ||
1306 | p += k; | |
1307 | ||
1308 | /* Accept trailing slashes */ | |
1309 | if (p[strspn(p, "/")] == 0) | |
1310 | return true; | |
1311 | ||
1312 | } | |
1313 | ||
1314 | return false; | |
1315 | } | |
1316 | ||
cd61c3bf LP |
1317 | static const char* const image_type_table[_IMAGE_TYPE_MAX] = { |
1318 | [IMAGE_DIRECTORY] = "directory", | |
1319 | [IMAGE_SUBVOLUME] = "subvolume", | |
aceac2f0 | 1320 | [IMAGE_RAW] = "raw", |
eb38edce | 1321 | [IMAGE_BLOCK] = "block", |
cd61c3bf LP |
1322 | }; |
1323 | ||
1324 | DEFINE_STRING_TABLE_LOOKUP(image_type, ImageType); | |
e0cd1900 LP |
1325 | |
1326 | static const char* const image_class_table[_IMAGE_CLASS_MAX] = { | |
1327 | [IMAGE_MACHINE] = "machine", | |
1328 | [IMAGE_PORTABLE] = "portable", | |
1329 | [IMAGE_EXTENSION] = "extension", | |
1330 | }; | |
1331 | ||
1332 | DEFINE_STRING_TABLE_LOOKUP(image_class, ImageClass); |