]>
Commit | Line | Data |
---|---|---|
cd61c3bf LP |
1 | /*** |
2 | This file is part of systemd. | |
3 | ||
4 | Copyright 2013 Lennart Poettering | |
5 | ||
6 | systemd is free software; you can redistribute it and/or modify it | |
7 | under the terms of the GNU Lesser General Public License as published by | |
8 | the Free Software Foundation; either version 2.1 of the License, or | |
9 | (at your option) any later version. | |
10 | ||
11 | systemd is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | Lesser General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU Lesser General Public License | |
17 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
18 | ***/ | |
19 | ||
a8fbdf54 TA |
20 | #include <dirent.h> |
21 | #include <errno.h> | |
ebd93cb6 | 22 | #include <fcntl.h> |
a8fbdf54 TA |
23 | #include <stdio.h> |
24 | #include <stdlib.h> | |
25 | #include <string.h> | |
e306723e | 26 | #include <sys/file.h> |
a8fbdf54 TA |
27 | #include <sys/stat.h> |
28 | #include <unistd.h> | |
8e0b6570 | 29 | #include <linux/fs.h> |
546dbec5 | 30 | |
b5efdb8a | 31 | #include "alloc-util.h" |
cd61c3bf | 32 | #include "btrfs-util.h" |
c8b3094d | 33 | #include "chattr-util.h" |
ebd93cb6 | 34 | #include "copy.h" |
a0956174 | 35 | #include "dirent-util.h" |
b6e953f2 | 36 | #include "env-util.h" |
3ffd4af2 | 37 | #include "fd-util.h" |
f4f15635 | 38 | #include "fs-util.h" |
a8fbdf54 TA |
39 | #include "hashmap.h" |
40 | #include "lockfile-util.h" | |
41 | #include "log.h" | |
3ffd4af2 | 42 | #include "machine-image.h" |
546dbec5 | 43 | #include "macro.h" |
30535c16 | 44 | #include "mkdir.h" |
8e0b6570 | 45 | #include "path-util.h" |
c6878637 | 46 | #include "rm-rf.h" |
8b43440b | 47 | #include "string-table.h" |
07630cea | 48 | #include "string-util.h" |
8e0b6570 | 49 | #include "strv.h" |
a8fbdf54 | 50 | #include "time-util.h" |
8e0b6570 | 51 | #include "utf8.h" |
a8fbdf54 | 52 | #include "util.h" |
89a5a90c | 53 | #include "xattr-util.h" |
cd61c3bf | 54 | |
c2ce6a3d | 55 | static const char image_search_path[] = |
42c6f2c9 | 56 | "/var/lib/machines\0" |
7d105503 | 57 | "/var/lib/container\0" /* legacy */ |
42c6f2c9 LP |
58 | "/usr/local/lib/machines\0" |
59 | "/usr/lib/machines\0"; | |
c2ce6a3d | 60 | |
cd61c3bf LP |
61 | Image *image_unref(Image *i) { |
62 | if (!i) | |
63 | return NULL; | |
64 | ||
65 | free(i->name); | |
66 | free(i->path); | |
6b430fdb | 67 | return mfree(i); |
cd61c3bf LP |
68 | } |
69 | ||
8e0b6570 LP |
70 | static char **image_settings_path(Image *image) { |
71 | _cleanup_strv_free_ char **l = NULL; | |
72 | char **ret; | |
73 | const char *fn, *s; | |
74 | unsigned i = 0; | |
75 | ||
76 | assert(image); | |
77 | ||
78 | l = new0(char*, 4); | |
79 | if (!l) | |
80 | return NULL; | |
81 | ||
82 | fn = strjoina(image->name, ".nspawn"); | |
83 | ||
84 | FOREACH_STRING(s, "/etc/systemd/nspawn/", "/run/systemd/nspawn/") { | |
85 | l[i] = strappend(s, fn); | |
86 | if (!l[i]) | |
87 | return NULL; | |
88 | ||
89 | i++; | |
90 | } | |
91 | ||
92 | l[i] = file_in_same_dir(image->path, fn); | |
93 | if (!l[i]) | |
94 | return NULL; | |
95 | ||
96 | ret = l; | |
97 | l = NULL; | |
98 | ||
99 | return ret; | |
100 | } | |
101 | ||
c2ce6a3d | 102 | static int image_new( |
cd61c3bf | 103 | ImageType t, |
5fc7f358 | 104 | const char *pretty, |
cd61c3bf | 105 | const char *path, |
5fc7f358 | 106 | const char *filename, |
cd61c3bf | 107 | bool read_only, |
10f9c755 | 108 | usec_t crtime, |
cd61c3bf | 109 | usec_t mtime, |
c2ce6a3d | 110 | Image **ret) { |
cd61c3bf LP |
111 | |
112 | _cleanup_(image_unrefp) Image *i = NULL; | |
cd61c3bf | 113 | |
cd61c3bf LP |
114 | assert(t >= 0); |
115 | assert(t < _IMAGE_TYPE_MAX); | |
5fc7f358 LP |
116 | assert(pretty); |
117 | assert(filename); | |
c2ce6a3d | 118 | assert(ret); |
cd61c3bf | 119 | |
c2ce6a3d | 120 | i = new0(Image, 1); |
cd61c3bf LP |
121 | if (!i) |
122 | return -ENOMEM; | |
123 | ||
124 | i->type = t; | |
125 | i->read_only = read_only; | |
10f9c755 | 126 | i->crtime = crtime; |
cd61c3bf | 127 | i->mtime = mtime; |
c19de711 | 128 | i->usage = i->usage_exclusive = (uint64_t) -1; |
b6b18498 | 129 | i->limit = i->limit_exclusive = (uint64_t) -1; |
cd61c3bf | 130 | |
5fc7f358 | 131 | i->name = strdup(pretty); |
cd61c3bf LP |
132 | if (!i->name) |
133 | return -ENOMEM; | |
134 | ||
5fc7f358 | 135 | if (path) |
605405c6 | 136 | i->path = strjoin(path, "/", filename); |
5fc7f358 LP |
137 | else |
138 | i->path = strdup(filename); | |
ebeccf9e | 139 | |
5fc7f358 LP |
140 | if (!i->path) |
141 | return -ENOMEM; | |
142 | ||
143 | path_kill_slashes(i->path); | |
cd61c3bf | 144 | |
c2ce6a3d | 145 | *ret = i; |
cd61c3bf | 146 | i = NULL; |
c2ce6a3d | 147 | |
cd61c3bf LP |
148 | return 0; |
149 | } | |
150 | ||
5fc7f358 LP |
151 | static int image_make( |
152 | const char *pretty, | |
153 | int dfd, | |
154 | const char *path, | |
155 | const char *filename, | |
156 | Image **ret) { | |
157 | ||
c2ce6a3d | 158 | struct stat st; |
5fc7f358 | 159 | bool read_only; |
cd61c3bf LP |
160 | int r; |
161 | ||
5fc7f358 | 162 | assert(filename); |
cd61c3bf | 163 | |
c2ce6a3d | 164 | /* We explicitly *do* follow symlinks here, since we want to |
5f129649 | 165 | * allow symlinking trees into /var/lib/machines/, and treat |
c2ce6a3d | 166 | * them normally. */ |
cd61c3bf | 167 | |
5fc7f358 | 168 | if (fstatat(dfd, filename, &st, 0) < 0) |
c2ce6a3d | 169 | return -errno; |
cd61c3bf | 170 | |
5fc7f358 LP |
171 | read_only = |
172 | (path && path_startswith(path, "/usr")) || | |
08ff5529 | 173 | (faccessat(dfd, filename, W_OK, AT_EACCESS) < 0 && errno == EROFS); |
86e339c8 | 174 | |
c2ce6a3d | 175 | if (S_ISDIR(st.st_mode)) { |
01b72568 LP |
176 | _cleanup_close_ int fd = -1; |
177 | unsigned file_attr = 0; | |
cd61c3bf | 178 | |
c2ce6a3d LP |
179 | if (!ret) |
180 | return 1; | |
cd61c3bf | 181 | |
5fc7f358 LP |
182 | if (!pretty) |
183 | pretty = filename; | |
184 | ||
01b72568 LP |
185 | fd = openat(dfd, filename, O_CLOEXEC|O_NOCTTY|O_DIRECTORY); |
186 | if (fd < 0) | |
187 | return -errno; | |
188 | ||
c2ce6a3d LP |
189 | /* btrfs subvolumes have inode 256 */ |
190 | if (st.st_ino == 256) { | |
cd61c3bf | 191 | |
21222ea5 LP |
192 | r = btrfs_is_filesystem(fd); |
193 | if (r < 0) | |
194 | return r; | |
195 | if (r) { | |
10f9c755 | 196 | BtrfsSubvolInfo info; |
cd61c3bf | 197 | |
c2ce6a3d | 198 | /* It's a btrfs subvolume */ |
cd61c3bf | 199 | |
5bcd08db | 200 | r = btrfs_subvol_get_info_fd(fd, 0, &info); |
10f9c755 LP |
201 | if (r < 0) |
202 | return r; | |
c2ce6a3d LP |
203 | |
204 | r = image_new(IMAGE_SUBVOLUME, | |
5fc7f358 | 205 | pretty, |
c2ce6a3d | 206 | path, |
5fc7f358 LP |
207 | filename, |
208 | info.read_only || read_only, | |
10f9c755 | 209 | info.otime, |
c2ce6a3d | 210 | 0, |
c2ce6a3d LP |
211 | ret); |
212 | if (r < 0) | |
213 | return r; | |
214 | ||
5bcd08db LP |
215 | if (btrfs_quota_scan_ongoing(fd) == 0) { |
216 | BtrfsQuotaInfo quota; | |
b6b18498 | 217 | |
5bcd08db LP |
218 | r = btrfs_subvol_get_subtree_quota_fd(fd, 0, "a); |
219 | if (r >= 0) { | |
220 | (*ret)->usage = quota.referenced; | |
221 | (*ret)->usage_exclusive = quota.exclusive; | |
222 | ||
223 | (*ret)->limit = quota.referenced_max; | |
224 | (*ret)->limit_exclusive = quota.exclusive_max; | |
225 | } | |
b6b18498 LP |
226 | } |
227 | ||
c2ce6a3d | 228 | return 1; |
cd61c3bf | 229 | } |
c2ce6a3d | 230 | } |
cd61c3bf | 231 | |
01b72568 LP |
232 | /* If the IMMUTABLE bit is set, we consider the |
233 | * directory read-only. Since the ioctl is not | |
234 | * supported everywhere we ignore failures. */ | |
235 | (void) read_attr_fd(fd, &file_attr); | |
cd61c3bf | 236 | |
01b72568 | 237 | /* It's just a normal directory. */ |
c2ce6a3d | 238 | r = image_new(IMAGE_DIRECTORY, |
5fc7f358 | 239 | pretty, |
c2ce6a3d | 240 | path, |
5fc7f358 | 241 | filename, |
01b72568 | 242 | read_only || (file_attr & FS_IMMUTABLE_FL), |
c2ce6a3d LP |
243 | 0, |
244 | 0, | |
245 | ret); | |
246 | if (r < 0) | |
247 | return r; | |
cd61c3bf | 248 | |
c2ce6a3d | 249 | return 1; |
cd61c3bf | 250 | |
aceac2f0 | 251 | } else if (S_ISREG(st.st_mode) && endswith(filename, ".raw")) { |
10f9c755 | 252 | usec_t crtime = 0; |
cd61c3bf | 253 | |
aceac2f0 | 254 | /* It's a RAW disk image */ |
cd61c3bf | 255 | |
c2ce6a3d LP |
256 | if (!ret) |
257 | return 1; | |
cd61c3bf | 258 | |
5fc7f358 | 259 | fd_getcrtime_at(dfd, filename, &crtime, 0); |
10f9c755 | 260 | |
5fc7f358 LP |
261 | if (!pretty) |
262 | pretty = strndupa(filename, strlen(filename) - 4); | |
10f9c755 | 263 | |
aceac2f0 | 264 | r = image_new(IMAGE_RAW, |
5fc7f358 | 265 | pretty, |
c2ce6a3d | 266 | path, |
5fc7f358 LP |
267 | filename, |
268 | !(st.st_mode & 0222) || read_only, | |
10f9c755 | 269 | crtime, |
c2ce6a3d | 270 | timespec_load(&st.st_mtim), |
c2ce6a3d LP |
271 | ret); |
272 | if (r < 0) | |
273 | return r; | |
cd61c3bf | 274 | |
c19de711 | 275 | (*ret)->usage = (*ret)->usage_exclusive = st.st_blocks * 512; |
b6b18498 LP |
276 | (*ret)->limit = (*ret)->limit_exclusive = st.st_size; |
277 | ||
c2ce6a3d LP |
278 | return 1; |
279 | } | |
cd61c3bf | 280 | |
c2ce6a3d LP |
281 | return 0; |
282 | } | |
cd61c3bf | 283 | |
c2ce6a3d LP |
284 | int image_find(const char *name, Image **ret) { |
285 | const char *path; | |
286 | int r; | |
cd61c3bf | 287 | |
c2ce6a3d | 288 | assert(name); |
cd61c3bf | 289 | |
c2ce6a3d LP |
290 | /* There are no images with invalid names */ |
291 | if (!image_name_is_valid(name)) | |
292 | return 0; | |
cd61c3bf | 293 | |
c2ce6a3d LP |
294 | NULSTR_FOREACH(path, image_search_path) { |
295 | _cleanup_closedir_ DIR *d = NULL; | |
cd61c3bf | 296 | |
c2ce6a3d LP |
297 | d = opendir(path); |
298 | if (!d) { | |
299 | if (errno == ENOENT) | |
300 | continue; | |
cd61c3bf | 301 | |
c2ce6a3d LP |
302 | return -errno; |
303 | } | |
cd61c3bf | 304 | |
5fc7f358 LP |
305 | r = image_make(NULL, dirfd(d), path, name, ret); |
306 | if (r == 0 || r == -ENOENT) { | |
aceac2f0 | 307 | _cleanup_free_ char *raw = NULL; |
5fc7f358 | 308 | |
aceac2f0 LP |
309 | raw = strappend(name, ".raw"); |
310 | if (!raw) | |
5fc7f358 LP |
311 | return -ENOMEM; |
312 | ||
aceac2f0 | 313 | r = image_make(NULL, dirfd(d), path, raw, ret); |
5fc7f358 LP |
314 | if (r == 0 || r == -ENOENT) |
315 | continue; | |
316 | } | |
c2ce6a3d LP |
317 | if (r < 0) |
318 | return r; | |
cd61c3bf | 319 | |
c2ce6a3d LP |
320 | return 1; |
321 | } | |
322 | ||
5fc7f358 | 323 | if (streq(name, ".host")) |
27c88c4e | 324 | return image_make(".host", AT_FDCWD, NULL, "/", ret); |
5fc7f358 | 325 | |
c2ce6a3d LP |
326 | return 0; |
327 | }; | |
328 | ||
329 | int image_discover(Hashmap *h) { | |
330 | const char *path; | |
331 | int r; | |
332 | ||
333 | assert(h); | |
334 | ||
335 | NULSTR_FOREACH(path, image_search_path) { | |
336 | _cleanup_closedir_ DIR *d = NULL; | |
337 | struct dirent *de; | |
338 | ||
339 | d = opendir(path); | |
340 | if (!d) { | |
341 | if (errno == ENOENT) | |
a67a4c8c | 342 | continue; |
c2ce6a3d LP |
343 | |
344 | return -errno; | |
345 | } | |
346 | ||
347 | FOREACH_DIRENT_ALL(de, d, return -errno) { | |
348 | _cleanup_(image_unrefp) Image *image = NULL; | |
349 | ||
350 | if (!image_name_is_valid(de->d_name)) | |
351 | continue; | |
352 | ||
353 | if (hashmap_contains(h, de->d_name)) | |
354 | continue; | |
355 | ||
5fc7f358 | 356 | r = image_make(NULL, dirfd(d), path, de->d_name, &image); |
c2ce6a3d LP |
357 | if (r == 0 || r == -ENOENT) |
358 | continue; | |
359 | if (r < 0) | |
360 | return r; | |
361 | ||
362 | r = hashmap_put(h, image->name, image); | |
363 | if (r < 0) | |
364 | return r; | |
365 | ||
366 | image = NULL; | |
cd61c3bf LP |
367 | } |
368 | } | |
369 | ||
5fc7f358 LP |
370 | if (!hashmap_contains(h, ".host")) { |
371 | _cleanup_(image_unrefp) Image *image = NULL; | |
372 | ||
373 | r = image_make(".host", AT_FDCWD, NULL, "/", &image); | |
374 | if (r < 0) | |
375 | return r; | |
376 | ||
377 | r = hashmap_put(h, image->name, image); | |
378 | if (r < 0) | |
379 | return r; | |
380 | ||
381 | image = NULL; | |
382 | ||
383 | } | |
384 | ||
cd61c3bf LP |
385 | return 0; |
386 | } | |
387 | ||
388 | void image_hashmap_free(Hashmap *map) { | |
389 | Image *i; | |
390 | ||
391 | while ((i = hashmap_steal_first(map))) | |
392 | image_unref(i); | |
393 | ||
394 | hashmap_free(map); | |
395 | } | |
396 | ||
08682124 | 397 | int image_remove(Image *i) { |
30535c16 | 398 | _cleanup_release_lock_file_ LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT; |
8e0b6570 LP |
399 | _cleanup_strv_free_ char **settings = NULL; |
400 | char **j; | |
30535c16 LP |
401 | int r; |
402 | ||
08682124 LP |
403 | assert(i); |
404 | ||
d94c2b06 | 405 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
08682124 LP |
406 | return -EROFS; |
407 | ||
8e0b6570 LP |
408 | settings = image_settings_path(i); |
409 | if (!settings) | |
410 | return -ENOMEM; | |
411 | ||
30535c16 LP |
412 | /* Make sure we don't interfere with a running nspawn */ |
413 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
414 | if (r < 0) | |
415 | return r; | |
416 | ||
ebd93cb6 LP |
417 | switch (i->type) { |
418 | ||
419 | case IMAGE_SUBVOLUME: | |
5bcd08db | 420 | r = btrfs_subvol_remove(i->path, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA); |
8e0b6570 LP |
421 | if (r < 0) |
422 | return r; | |
423 | break; | |
ebd93cb6 LP |
424 | |
425 | case IMAGE_DIRECTORY: | |
01b72568 | 426 | /* Allow deletion of read-only directories */ |
a67d68b8 | 427 | (void) chattr_path(i->path, 0, FS_IMMUTABLE_FL); |
8e0b6570 LP |
428 | r = rm_rf(i->path, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME); |
429 | if (r < 0) | |
430 | return r; | |
431 | ||
432 | break; | |
01b72568 | 433 | |
aceac2f0 | 434 | case IMAGE_RAW: |
41d1ed05 LP |
435 | if (unlink(i->path) < 0) |
436 | return -errno; | |
8e0b6570 | 437 | break; |
ebd93cb6 LP |
438 | |
439 | default: | |
15411c0c | 440 | return -EOPNOTSUPP; |
ebd93cb6 | 441 | } |
8e0b6570 LP |
442 | |
443 | STRV_FOREACH(j, settings) { | |
444 | if (unlink(*j) < 0 && errno != ENOENT) | |
445 | log_debug_errno(errno, "Failed to unlink %s, ignoring: %m", *j); | |
446 | } | |
447 | ||
448 | return 0; | |
449 | } | |
450 | ||
451 | static int rename_settings_file(const char *path, const char *new_name) { | |
452 | _cleanup_free_ char *rs = NULL; | |
453 | const char *fn; | |
454 | ||
455 | fn = strjoina(new_name, ".nspawn"); | |
456 | ||
457 | rs = file_in_same_dir(path, fn); | |
458 | if (!rs) | |
459 | return -ENOMEM; | |
460 | ||
461 | return rename_noreplace(AT_FDCWD, path, AT_FDCWD, rs); | |
ebd93cb6 LP |
462 | } |
463 | ||
464 | int image_rename(Image *i, const char *new_name) { | |
30535c16 | 465 | _cleanup_release_lock_file_ LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT, name_lock = LOCK_FILE_INIT; |
ebd93cb6 | 466 | _cleanup_free_ char *new_path = NULL, *nn = NULL; |
8e0b6570 | 467 | _cleanup_strv_free_ char **settings = NULL; |
01b72568 | 468 | unsigned file_attr = 0; |
8e0b6570 | 469 | char **j; |
ebd93cb6 LP |
470 | int r; |
471 | ||
472 | assert(i); | |
473 | ||
474 | if (!image_name_is_valid(new_name)) | |
475 | return -EINVAL; | |
476 | ||
d94c2b06 | 477 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
ebd93cb6 LP |
478 | return -EROFS; |
479 | ||
8e0b6570 LP |
480 | settings = image_settings_path(i); |
481 | if (!settings) | |
482 | return -ENOMEM; | |
483 | ||
30535c16 LP |
484 | /* Make sure we don't interfere with a running nspawn */ |
485 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
486 | if (r < 0) | |
487 | return r; | |
488 | ||
489 | /* Make sure nobody takes the new name, between the time we | |
490 | * checked it is currently unused in all search paths, and the | |
f8e2f4d6 | 491 | * time we take possession of it */ |
30535c16 LP |
492 | r = image_name_lock(new_name, LOCK_EX|LOCK_NB, &name_lock); |
493 | if (r < 0) | |
494 | return r; | |
495 | ||
ebd93cb6 LP |
496 | r = image_find(new_name, NULL); |
497 | if (r < 0) | |
498 | return r; | |
499 | if (r > 0) | |
500 | return -EEXIST; | |
501 | ||
502 | switch (i->type) { | |
503 | ||
ebd93cb6 | 504 | case IMAGE_DIRECTORY: |
01b72568 LP |
505 | /* Turn of the immutable bit while we rename the image, so that we can rename it */ |
506 | (void) read_attr_path(i->path, &file_attr); | |
507 | ||
508 | if (file_attr & FS_IMMUTABLE_FL) | |
a67d68b8 | 509 | (void) chattr_path(i->path, 0, FS_IMMUTABLE_FL); |
01b72568 LP |
510 | |
511 | /* fall through */ | |
512 | ||
513 | case IMAGE_SUBVOLUME: | |
ebd93cb6 LP |
514 | new_path = file_in_same_dir(i->path, new_name); |
515 | break; | |
516 | ||
aceac2f0 | 517 | case IMAGE_RAW: { |
ebd93cb6 LP |
518 | const char *fn; |
519 | ||
63c372cb | 520 | fn = strjoina(new_name, ".raw"); |
ebd93cb6 LP |
521 | new_path = file_in_same_dir(i->path, fn); |
522 | break; | |
523 | } | |
524 | ||
525 | default: | |
15411c0c | 526 | return -EOPNOTSUPP; |
ebd93cb6 LP |
527 | } |
528 | ||
529 | if (!new_path) | |
530 | return -ENOMEM; | |
531 | ||
532 | nn = strdup(new_name); | |
533 | if (!nn) | |
534 | return -ENOMEM; | |
535 | ||
f85ef957 AC |
536 | r = rename_noreplace(AT_FDCWD, i->path, AT_FDCWD, new_path); |
537 | if (r < 0) | |
538 | return r; | |
ebd93cb6 | 539 | |
01b72568 LP |
540 | /* Restore the immutable bit, if it was set before */ |
541 | if (file_attr & FS_IMMUTABLE_FL) | |
a67d68b8 | 542 | (void) chattr_path(new_path, FS_IMMUTABLE_FL, FS_IMMUTABLE_FL); |
01b72568 | 543 | |
ebd93cb6 LP |
544 | free(i->path); |
545 | i->path = new_path; | |
546 | new_path = NULL; | |
547 | ||
548 | free(i->name); | |
549 | i->name = nn; | |
550 | nn = NULL; | |
551 | ||
8e0b6570 LP |
552 | STRV_FOREACH(j, settings) { |
553 | r = rename_settings_file(*j, new_name); | |
554 | if (r < 0 && r != -ENOENT) | |
555 | log_debug_errno(r, "Failed to rename settings file %s, ignoring: %m", *j); | |
556 | } | |
557 | ||
ebd93cb6 LP |
558 | return 0; |
559 | } | |
560 | ||
8e0b6570 LP |
561 | static int clone_settings_file(const char *path, const char *new_name) { |
562 | _cleanup_free_ char *rs = NULL; | |
563 | const char *fn; | |
564 | ||
565 | fn = strjoina(new_name, ".nspawn"); | |
566 | ||
567 | rs = file_in_same_dir(path, fn); | |
568 | if (!rs) | |
569 | return -ENOMEM; | |
570 | ||
571 | return copy_file_atomic(path, rs, 0664, false, 0); | |
572 | } | |
573 | ||
ebd93cb6 | 574 | int image_clone(Image *i, const char *new_name, bool read_only) { |
30535c16 | 575 | _cleanup_release_lock_file_ LockFile name_lock = LOCK_FILE_INIT; |
8e0b6570 | 576 | _cleanup_strv_free_ char **settings = NULL; |
ebd93cb6 | 577 | const char *new_path; |
8e0b6570 | 578 | char **j; |
ebd93cb6 LP |
579 | int r; |
580 | ||
581 | assert(i); | |
582 | ||
583 | if (!image_name_is_valid(new_name)) | |
584 | return -EINVAL; | |
585 | ||
8e0b6570 LP |
586 | settings = image_settings_path(i); |
587 | if (!settings) | |
588 | return -ENOMEM; | |
589 | ||
30535c16 LP |
590 | /* Make sure nobody takes the new name, between the time we |
591 | * checked it is currently unused in all search paths, and the | |
f8e2f4d6 | 592 | * time we take possession of it */ |
30535c16 LP |
593 | r = image_name_lock(new_name, LOCK_EX|LOCK_NB, &name_lock); |
594 | if (r < 0) | |
595 | return r; | |
596 | ||
ebd93cb6 LP |
597 | r = image_find(new_name, NULL); |
598 | if (r < 0) | |
599 | return r; | |
600 | if (r > 0) | |
601 | return -EEXIST; | |
602 | ||
603 | switch (i->type) { | |
604 | ||
605 | case IMAGE_SUBVOLUME: | |
606 | case IMAGE_DIRECTORY: | |
9a50e3ca LP |
607 | /* If we can we'll always try to create a new btrfs subvolume here, even if the source is a plain |
608 | * directory.*/ | |
609 | ||
63c372cb | 610 | new_path = strjoina("/var/lib/machines/", new_name); |
ebd93cb6 | 611 | |
5bcd08db | 612 | r = btrfs_subvol_snapshot(i->path, new_path, (read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA); |
9a50e3ca LP |
613 | if (r == -EOPNOTSUPP) { |
614 | /* No btrfs snapshots supported, create a normal directory then. */ | |
615 | ||
616 | r = copy_directory(i->path, new_path, false); | |
617 | if (r >= 0) | |
618 | (void) chattr_path(new_path, read_only ? FS_IMMUTABLE_FL : 0, FS_IMMUTABLE_FL); | |
619 | } else if (r >= 0) | |
620 | /* Enable "subtree" quotas for the copy, if we didn't copy any quota from the source. */ | |
8120ee28 | 621 | (void) btrfs_subvol_auto_qgroup(new_path, 0, true); |
5bcd08db | 622 | |
ebd93cb6 LP |
623 | break; |
624 | ||
aceac2f0 | 625 | case IMAGE_RAW: |
63c372cb | 626 | new_path = strjoina("/var/lib/machines/", new_name, ".raw"); |
ebd93cb6 | 627 | |
f2068bcc | 628 | r = copy_file_atomic(i->path, new_path, read_only ? 0444 : 0644, false, FS_NOCOW_FL); |
ebd93cb6 LP |
629 | break; |
630 | ||
631 | default: | |
15411c0c | 632 | return -EOPNOTSUPP; |
ebd93cb6 LP |
633 | } |
634 | ||
635 | if (r < 0) | |
636 | return r; | |
637 | ||
8e0b6570 LP |
638 | STRV_FOREACH(j, settings) { |
639 | r = clone_settings_file(*j, new_name); | |
640 | if (r < 0 && r != -ENOENT) | |
641 | log_debug_errno(r, "Failed to clone settings %s, ignoring: %m", *j); | |
642 | } | |
643 | ||
ebd93cb6 LP |
644 | return 0; |
645 | } | |
646 | ||
647 | int image_read_only(Image *i, bool b) { | |
30535c16 | 648 | _cleanup_release_lock_file_ LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT; |
ebd93cb6 LP |
649 | int r; |
650 | assert(i); | |
651 | ||
d94c2b06 | 652 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
ebd93cb6 LP |
653 | return -EROFS; |
654 | ||
30535c16 LP |
655 | /* Make sure we don't interfere with a running nspawn */ |
656 | r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock); | |
657 | if (r < 0) | |
658 | return r; | |
659 | ||
ebd93cb6 LP |
660 | switch (i->type) { |
661 | ||
662 | case IMAGE_SUBVOLUME: | |
5bcd08db LP |
663 | |
664 | /* Note that we set the flag only on the top-level | |
665 | * subvolume of the image. */ | |
666 | ||
ebd93cb6 LP |
667 | r = btrfs_subvol_set_read_only(i->path, b); |
668 | if (r < 0) | |
669 | return r; | |
01b72568 LP |
670 | |
671 | break; | |
672 | ||
673 | case IMAGE_DIRECTORY: | |
674 | /* For simple directory trees we cannot use the access | |
675 | mode of the top-level directory, since it has an | |
676 | effect on the container itself. However, we can | |
677 | use the "immutable" flag, to at least make the | |
678 | top-level directory read-only. It's not as good as | |
679 | a read-only subvolume, but at least something, and | |
680 | we can read the value back.*/ | |
681 | ||
a67d68b8 | 682 | r = chattr_path(i->path, b ? FS_IMMUTABLE_FL : 0, FS_IMMUTABLE_FL); |
01b72568 LP |
683 | if (r < 0) |
684 | return r; | |
685 | ||
ebd93cb6 LP |
686 | break; |
687 | ||
aceac2f0 | 688 | case IMAGE_RAW: { |
ebd93cb6 LP |
689 | struct stat st; |
690 | ||
691 | if (stat(i->path, &st) < 0) | |
692 | return -errno; | |
693 | ||
694 | if (chmod(i->path, (st.st_mode & 0444) | (b ? 0000 : 0200)) < 0) | |
695 | return -errno; | |
f2068bcc LP |
696 | |
697 | /* If the images is now read-only, it's a good time to | |
698 | * defrag it, given that no write patterns will | |
699 | * fragment it again. */ | |
700 | if (b) | |
701 | (void) btrfs_defrag(i->path); | |
ebd93cb6 LP |
702 | break; |
703 | } | |
704 | ||
ebd93cb6 | 705 | default: |
15411c0c | 706 | return -EOPNOTSUPP; |
ebd93cb6 LP |
707 | } |
708 | ||
709 | return 0; | |
08682124 LP |
710 | } |
711 | ||
30535c16 LP |
712 | int image_path_lock(const char *path, int operation, LockFile *global, LockFile *local) { |
713 | _cleanup_free_ char *p = NULL; | |
714 | LockFile t = LOCK_FILE_INIT; | |
715 | struct stat st; | |
716 | int r; | |
717 | ||
718 | assert(path); | |
719 | assert(global); | |
720 | assert(local); | |
721 | ||
722 | /* Locks an image path. This actually creates two locks: one | |
723 | * "local" one, next to the image path itself, which might be | |
724 | * shared via NFS. And another "global" one, in /run, that | |
725 | * uses the device/inode number. This has the benefit that we | |
726 | * can even lock a tree that is a mount point, correctly. */ | |
727 | ||
30535c16 LP |
728 | if (!path_is_absolute(path)) |
729 | return -EINVAL; | |
730 | ||
b6e953f2 LP |
731 | if (getenv_bool("SYSTEMD_NSPAWN_LOCK") == 0) { |
732 | *local = *global = (LockFile) LOCK_FILE_INIT; | |
733 | return 0; | |
734 | } | |
735 | ||
736 | if (path_equal(path, "/")) | |
737 | return -EBUSY; | |
738 | ||
30535c16 LP |
739 | if (stat(path, &st) >= 0) { |
740 | if (asprintf(&p, "/run/systemd/nspawn/locks/inode-%lu:%lu", (unsigned long) st.st_dev, (unsigned long) st.st_ino) < 0) | |
741 | return -ENOMEM; | |
742 | } | |
743 | ||
744 | r = make_lock_file_for(path, operation, &t); | |
745 | if (r < 0) | |
746 | return r; | |
747 | ||
748 | if (p) { | |
7e7cddb2 | 749 | mkdir_p("/run/systemd/nspawn/locks", 0700); |
30535c16 LP |
750 | |
751 | r = make_lock_file(p, operation, global); | |
752 | if (r < 0) { | |
753 | release_lock_file(&t); | |
754 | return r; | |
755 | } | |
546dbec5 LP |
756 | } else |
757 | *global = (LockFile) LOCK_FILE_INIT; | |
30535c16 LP |
758 | |
759 | *local = t; | |
760 | return 0; | |
761 | } | |
762 | ||
cb81cd80 | 763 | int image_set_limit(Image *i, uint64_t referenced_max) { |
d6ce17c7 LP |
764 | assert(i); |
765 | ||
d94c2b06 | 766 | if (IMAGE_IS_VENDOR(i) || IMAGE_IS_HOST(i)) |
d6ce17c7 LP |
767 | return -EROFS; |
768 | ||
769 | if (i->type != IMAGE_SUBVOLUME) | |
15411c0c | 770 | return -EOPNOTSUPP; |
d6ce17c7 | 771 | |
5bcd08db LP |
772 | /* We set the quota both for the subvolume as well as for the |
773 | * subtree. The latter is mostly for historical reasons, since | |
774 | * we didn't use to have a concept of subtree quota, and hence | |
775 | * only modified the subvolume quota. */ | |
776 | ||
777 | (void) btrfs_qgroup_set_limit(i->path, 0, referenced_max); | |
778 | (void) btrfs_subvol_auto_qgroup(i->path, 0, true); | |
779 | return btrfs_subvol_set_subtree_quota_limit(i->path, 0, referenced_max); | |
d6ce17c7 LP |
780 | } |
781 | ||
30535c16 LP |
782 | int image_name_lock(const char *name, int operation, LockFile *ret) { |
783 | const char *p; | |
784 | ||
785 | assert(name); | |
786 | assert(ret); | |
787 | ||
788 | /* Locks an image name, regardless of the precise path used. */ | |
789 | ||
790 | if (!image_name_is_valid(name)) | |
791 | return -EINVAL; | |
792 | ||
b6e953f2 LP |
793 | if (getenv_bool("SYSTEMD_NSPAWN_LOCK") == 0) { |
794 | *ret = (LockFile) LOCK_FILE_INIT; | |
795 | return 0; | |
796 | } | |
797 | ||
30535c16 LP |
798 | if (streq(name, ".host")) |
799 | return -EBUSY; | |
800 | ||
7e7cddb2 | 801 | mkdir_p("/run/systemd/nspawn/locks", 0700); |
63c372cb | 802 | p = strjoina("/run/systemd/nspawn/locks/name-", name); |
30535c16 LP |
803 | |
804 | return make_lock_file(p, operation, ret); | |
805 | } | |
806 | ||
807 | bool image_name_is_valid(const char *s) { | |
808 | if (!filename_is_valid(s)) | |
809 | return false; | |
810 | ||
811 | if (string_has_cc(s, NULL)) | |
812 | return false; | |
813 | ||
814 | if (!utf8_is_valid(s)) | |
815 | return false; | |
816 | ||
817 | /* Temporary files for atomically creating new files */ | |
818 | if (startswith(s, ".#")) | |
819 | return false; | |
820 | ||
821 | return true; | |
822 | } | |
823 | ||
cd61c3bf LP |
824 | static const char* const image_type_table[_IMAGE_TYPE_MAX] = { |
825 | [IMAGE_DIRECTORY] = "directory", | |
826 | [IMAGE_SUBVOLUME] = "subvolume", | |
aceac2f0 | 827 | [IMAGE_RAW] = "raw", |
cd61c3bf LP |
828 | }; |
829 | ||
830 | DEFINE_STRING_TABLE_LOOKUP(image_type, ImageType); |