]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
d7c7c334 LP |
2 | /*** |
3 | This file is part of systemd. | |
4 | ||
5 | Copyright 2014 Lennart Poettering | |
6 | ||
7 | systemd is free software; you can redistribute it and/or modify it | |
8 | under the terms of the GNU Lesser General Public License as published by | |
9 | the Free Software Foundation; either version 2.1 of the License, or | |
10 | (at your option) any later version. | |
11 | ||
12 | systemd is distributed in the hope that it will be useful, but | |
13 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 | Lesser General Public License for more details. | |
16 | ||
17 | You should have received a copy of the GNU Lesser General Public License | |
18 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
19 | ***/ | |
20 | ||
11c3a366 TA |
21 | #include <errno.h> |
22 | #include <fcntl.h> | |
23 | #include <inttypes.h> | |
17cbb288 | 24 | #include <linux/fs.h> |
11c3a366 TA |
25 | #include <linux/loop.h> |
26 | #include <stddef.h> | |
27 | #include <stdio.h> | |
d7c7c334 | 28 | #include <stdlib.h> |
11c3a366 TA |
29 | #include <string.h> |
30 | #include <sys/ioctl.h> | |
d7c7c334 | 31 | #include <sys/stat.h> |
11c3a366 TA |
32 | #include <sys/statfs.h> |
33 | #include <sys/sysmacros.h> | |
34 | #include <unistd.h> | |
35 | ||
349cc4a5 | 36 | #if HAVE_LINUX_BTRFS_H |
d7c7c334 LP |
37 | #include <linux/btrfs.h> |
38 | #endif | |
39 | ||
b5efdb8a | 40 | #include "alloc-util.h" |
07630cea | 41 | #include "btrfs-ctree.h" |
3ffd4af2 | 42 | #include "btrfs-util.h" |
17cbb288 | 43 | #include "chattr-util.h" |
07630cea | 44 | #include "copy.h" |
3ffd4af2 | 45 | #include "fd-util.h" |
07630cea | 46 | #include "fileio.h" |
a90fb858 | 47 | #include "io-util.h" |
07630cea | 48 | #include "macro.h" |
d7c7c334 | 49 | #include "missing.h" |
d7c7c334 | 50 | #include "path-util.h" |
17cbb288 | 51 | #include "rm-rf.h" |
d7b8eec7 LP |
52 | #include "selinux-util.h" |
53 | #include "smack-util.h" | |
11c3a366 | 54 | #include "sparse-endian.h" |
872a590e | 55 | #include "stat-util.h" |
07630cea | 56 | #include "string-util.h" |
93cc7779 | 57 | #include "time-util.h" |
07630cea | 58 | #include "util.h" |
d7c7c334 | 59 | |
62572894 LP |
60 | /* WARNING: Be careful with file system ioctls! When we get an fd, we |
61 | * need to make sure it either refers to only a regular file or | |
62 | * directory, or that it is located on btrfs, before invoking any | |
63 | * btrfs ioctls. The ioctl numbers are reused by some device drivers | |
64 | * (such as DRM), and hence might have bad effects when invoked on | |
65 | * device nodes (that reference drivers) rather than fds to normal | |
66 | * files or directories. */ | |
67 | ||
d7c7c334 LP |
68 | static int validate_subvolume_name(const char *name) { |
69 | ||
70 | if (!filename_is_valid(name)) | |
71 | return -EINVAL; | |
72 | ||
73 | if (strlen(name) > BTRFS_SUBVOL_NAME_MAX) | |
74 | return -E2BIG; | |
75 | ||
76 | return 0; | |
77 | } | |
78 | ||
79 | static int open_parent(const char *path, int flags) { | |
80 | _cleanup_free_ char *parent = NULL; | |
5f311f8c | 81 | int fd; |
d7c7c334 LP |
82 | |
83 | assert(path); | |
84 | ||
5f311f8c LP |
85 | parent = dirname_malloc(path); |
86 | if (!parent) | |
87 | return -ENOMEM; | |
d7c7c334 LP |
88 | |
89 | fd = open(parent, flags); | |
90 | if (fd < 0) | |
91 | return -errno; | |
92 | ||
93 | return fd; | |
94 | } | |
95 | ||
96 | static int extract_subvolume_name(const char *path, const char **subvolume) { | |
97 | const char *fn; | |
98 | int r; | |
99 | ||
100 | assert(path); | |
101 | assert(subvolume); | |
102 | ||
103 | fn = basename(path); | |
104 | ||
105 | r = validate_subvolume_name(fn); | |
106 | if (r < 0) | |
107 | return r; | |
108 | ||
109 | *subvolume = fn; | |
110 | return 0; | |
111 | } | |
112 | ||
21222ea5 | 113 | int btrfs_is_filesystem(int fd) { |
d7c7c334 LP |
114 | struct statfs sfs; |
115 | ||
21222ea5 LP |
116 | assert(fd >= 0); |
117 | ||
118 | if (fstatfs(fd, &sfs) < 0) | |
119 | return -errno; | |
120 | ||
121 | return F_TYPE_EQUAL(sfs.f_type, BTRFS_SUPER_MAGIC); | |
122 | } | |
123 | ||
2904e949 | 124 | int btrfs_is_subvol_fd(int fd) { |
21222ea5 LP |
125 | struct stat st; |
126 | ||
127 | assert(fd >= 0); | |
128 | ||
cd61c3bf LP |
129 | /* On btrfs subvolumes always have the inode 256 */ |
130 | ||
131 | if (fstat(fd, &st) < 0) | |
d7c7c334 LP |
132 | return -errno; |
133 | ||
cd61c3bf | 134 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) |
d7c7c334 LP |
135 | return 0; |
136 | ||
21222ea5 | 137 | return btrfs_is_filesystem(fd); |
d7c7c334 LP |
138 | } |
139 | ||
2904e949 LP |
140 | int btrfs_is_subvol(const char *path) { |
141 | _cleanup_close_ int fd = -1; | |
142 | ||
143 | assert(path); | |
144 | ||
145 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
146 | if (fd < 0) | |
147 | return -errno; | |
148 | ||
149 | return btrfs_is_subvol_fd(fd); | |
150 | } | |
151 | ||
d7c7c334 LP |
152 | int btrfs_subvol_make(const char *path) { |
153 | struct btrfs_ioctl_vol_args args = {}; | |
154 | _cleanup_close_ int fd = -1; | |
155 | const char *subvolume; | |
156 | int r; | |
157 | ||
158 | assert(path); | |
159 | ||
160 | r = extract_subvolume_name(path, &subvolume); | |
161 | if (r < 0) | |
162 | return r; | |
163 | ||
164 | fd = open_parent(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
165 | if (fd < 0) | |
166 | return fd; | |
167 | ||
168 | strncpy(args.name, subvolume, sizeof(args.name)-1); | |
169 | ||
170 | if (ioctl(fd, BTRFS_IOC_SUBVOL_CREATE, &args) < 0) | |
171 | return -errno; | |
172 | ||
173 | return 0; | |
174 | } | |
175 | ||
d7b8eec7 LP |
176 | int btrfs_subvol_make_label(const char *path) { |
177 | int r; | |
178 | ||
179 | assert(path); | |
180 | ||
181 | r = mac_selinux_create_file_prepare(path, S_IFDIR); | |
182 | if (r < 0) | |
183 | return r; | |
184 | ||
185 | r = btrfs_subvol_make(path); | |
186 | mac_selinux_create_file_clear(); | |
187 | ||
188 | if (r < 0) | |
189 | return r; | |
190 | ||
191 | return mac_smack_fix(path, false, false); | |
192 | } | |
193 | ||
0d6e763b | 194 | int btrfs_subvol_set_read_only_fd(int fd, bool b) { |
d7c7c334 | 195 | uint64_t flags, nflags; |
0d6e763b | 196 | struct stat st; |
d7c7c334 | 197 | |
0d6e763b LP |
198 | assert(fd >= 0); |
199 | ||
200 | if (fstat(fd, &st) < 0) | |
d7c7c334 LP |
201 | return -errno; |
202 | ||
0d6e763b LP |
203 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) |
204 | return -EINVAL; | |
205 | ||
d7c7c334 LP |
206 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) |
207 | return -errno; | |
208 | ||
209 | if (b) | |
210 | nflags = flags | BTRFS_SUBVOL_RDONLY; | |
211 | else | |
212 | nflags = flags & ~BTRFS_SUBVOL_RDONLY; | |
213 | ||
214 | if (flags == nflags) | |
215 | return 0; | |
216 | ||
217 | if (ioctl(fd, BTRFS_IOC_SUBVOL_SETFLAGS, &nflags) < 0) | |
218 | return -errno; | |
219 | ||
220 | return 0; | |
221 | } | |
222 | ||
0d6e763b LP |
223 | int btrfs_subvol_set_read_only(const char *path, bool b) { |
224 | _cleanup_close_ int fd = -1; | |
225 | ||
226 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
227 | if (fd < 0) | |
228 | return -errno; | |
229 | ||
230 | return btrfs_subvol_set_read_only_fd(fd, b); | |
231 | } | |
232 | ||
10f9c755 | 233 | int btrfs_subvol_get_read_only_fd(int fd) { |
cd61c3bf | 234 | uint64_t flags; |
62572894 LP |
235 | struct stat st; |
236 | ||
237 | assert(fd >= 0); | |
238 | ||
239 | if (fstat(fd, &st) < 0) | |
240 | return -errno; | |
241 | ||
242 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) | |
243 | return -EINVAL; | |
cd61c3bf LP |
244 | |
245 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) | |
246 | return -errno; | |
247 | ||
248 | return !!(flags & BTRFS_SUBVOL_RDONLY); | |
249 | } | |
250 | ||
d7c7c334 | 251 | int btrfs_reflink(int infd, int outfd) { |
62572894 | 252 | struct stat st; |
d7c7c334 LP |
253 | int r; |
254 | ||
255 | assert(infd >= 0); | |
256 | assert(outfd >= 0); | |
257 | ||
62572894 LP |
258 | /* Make sure we invoke the ioctl on a regular file, so that no |
259 | * device driver accidentally gets it. */ | |
260 | ||
261 | if (fstat(outfd, &st) < 0) | |
262 | return -errno; | |
263 | ||
264 | if (!S_ISREG(st.st_mode)) | |
265 | return -EINVAL; | |
266 | ||
d7c7c334 LP |
267 | r = ioctl(outfd, BTRFS_IOC_CLONE, infd); |
268 | if (r < 0) | |
269 | return -errno; | |
270 | ||
271 | return 0; | |
272 | } | |
273 | ||
1c7dd825 LP |
274 | int btrfs_clone_range(int infd, uint64_t in_offset, int outfd, uint64_t out_offset, uint64_t sz) { |
275 | struct btrfs_ioctl_clone_range_args args = { | |
276 | .src_fd = infd, | |
277 | .src_offset = in_offset, | |
278 | .src_length = sz, | |
279 | .dest_offset = out_offset, | |
280 | }; | |
62572894 | 281 | struct stat st; |
1c7dd825 LP |
282 | int r; |
283 | ||
284 | assert(infd >= 0); | |
285 | assert(outfd >= 0); | |
286 | assert(sz > 0); | |
287 | ||
62572894 LP |
288 | if (fstat(outfd, &st) < 0) |
289 | return -errno; | |
290 | ||
291 | if (!S_ISREG(st.st_mode)) | |
292 | return -EINVAL; | |
293 | ||
1c7dd825 LP |
294 | r = ioctl(outfd, BTRFS_IOC_CLONE_RANGE, &args); |
295 | if (r < 0) | |
296 | return -errno; | |
297 | ||
298 | return 0; | |
299 | } | |
300 | ||
efe02862 | 301 | int btrfs_get_block_device_fd(int fd, dev_t *dev) { |
d7c7c334 | 302 | struct btrfs_ioctl_fs_info_args fsi = {}; |
d7c7c334 | 303 | uint64_t id; |
62572894 | 304 | int r; |
d7c7c334 | 305 | |
efe02862 | 306 | assert(fd >= 0); |
d7c7c334 LP |
307 | assert(dev); |
308 | ||
62572894 LP |
309 | r = btrfs_is_filesystem(fd); |
310 | if (r < 0) | |
311 | return r; | |
312 | if (!r) | |
313 | return -ENOTTY; | |
314 | ||
d7c7c334 LP |
315 | if (ioctl(fd, BTRFS_IOC_FS_INFO, &fsi) < 0) |
316 | return -errno; | |
317 | ||
318 | /* We won't do this for btrfs RAID */ | |
319 | if (fsi.num_devices != 1) | |
320 | return 0; | |
321 | ||
322 | for (id = 1; id <= fsi.max_id; id++) { | |
323 | struct btrfs_ioctl_dev_info_args di = { | |
324 | .devid = id, | |
325 | }; | |
326 | struct stat st; | |
327 | ||
328 | if (ioctl(fd, BTRFS_IOC_DEV_INFO, &di) < 0) { | |
329 | if (errno == ENODEV) | |
330 | continue; | |
331 | ||
332 | return -errno; | |
333 | } | |
334 | ||
335 | if (stat((char*) di.path, &st) < 0) | |
336 | return -errno; | |
337 | ||
338 | if (!S_ISBLK(st.st_mode)) | |
339 | return -ENODEV; | |
340 | ||
341 | if (major(st.st_rdev) == 0) | |
342 | return -ENODEV; | |
343 | ||
344 | *dev = st.st_rdev; | |
345 | return 1; | |
346 | } | |
347 | ||
348 | return -ENODEV; | |
349 | } | |
10f9c755 | 350 | |
efe02862 LP |
351 | int btrfs_get_block_device(const char *path, dev_t *dev) { |
352 | _cleanup_close_ int fd = -1; | |
353 | ||
354 | assert(path); | |
355 | assert(dev); | |
356 | ||
357 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
358 | if (fd < 0) | |
359 | return -errno; | |
360 | ||
361 | return btrfs_get_block_device_fd(fd, dev); | |
362 | } | |
363 | ||
10f9c755 LP |
364 | int btrfs_subvol_get_id_fd(int fd, uint64_t *ret) { |
365 | struct btrfs_ioctl_ino_lookup_args args = { | |
366 | .objectid = BTRFS_FIRST_FREE_OBJECTID | |
367 | }; | |
62572894 | 368 | int r; |
10f9c755 LP |
369 | |
370 | assert(fd >= 0); | |
371 | assert(ret); | |
372 | ||
62572894 LP |
373 | r = btrfs_is_filesystem(fd); |
374 | if (r < 0) | |
375 | return r; | |
376 | if (!r) | |
377 | return -ENOTTY; | |
378 | ||
10f9c755 LP |
379 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args) < 0) |
380 | return -errno; | |
381 | ||
382 | *ret = args.treeid; | |
383 | return 0; | |
384 | } | |
385 | ||
90578cbd LP |
386 | int btrfs_subvol_get_id(int fd, const char *subvol, uint64_t *ret) { |
387 | _cleanup_close_ int subvol_fd = -1; | |
388 | ||
389 | assert(fd >= 0); | |
390 | assert(ret); | |
391 | ||
392 | subvol_fd = openat(fd, subvol, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
393 | if (subvol_fd < 0) | |
394 | return -errno; | |
395 | ||
396 | return btrfs_subvol_get_id_fd(subvol_fd, ret); | |
397 | } | |
398 | ||
5743a585 LP |
399 | static bool btrfs_ioctl_search_args_inc(struct btrfs_ioctl_search_args *args) { |
400 | assert(args); | |
401 | ||
402 | /* the objectid, type, offset together make up the btrfs key, | |
403 | * which is considered a single 136byte integer when | |
404 | * comparing. This call increases the counter by one, dealing | |
405 | * with the overflow between the overflows */ | |
406 | ||
407 | if (args->key.min_offset < (uint64_t) -1) { | |
408 | args->key.min_offset++; | |
409 | return true; | |
410 | } | |
411 | ||
412 | if (args->key.min_type < (uint8_t) -1) { | |
413 | args->key.min_type++; | |
414 | args->key.min_offset = 0; | |
415 | return true; | |
416 | } | |
417 | ||
418 | if (args->key.min_objectid < (uint64_t) -1) { | |
419 | args->key.min_objectid++; | |
420 | args->key.min_offset = 0; | |
421 | args->key.min_type = 0; | |
422 | return true; | |
423 | } | |
424 | ||
425 | return 0; | |
426 | } | |
427 | ||
428 | static void btrfs_ioctl_search_args_set(struct btrfs_ioctl_search_args *args, const struct btrfs_ioctl_search_header *h) { | |
429 | assert(args); | |
430 | assert(h); | |
431 | ||
432 | args->key.min_objectid = h->objectid; | |
433 | args->key.min_type = h->type; | |
434 | args->key.min_offset = h->offset; | |
435 | } | |
436 | ||
437 | static int btrfs_ioctl_search_args_compare(const struct btrfs_ioctl_search_args *args) { | |
438 | assert(args); | |
439 | ||
440 | /* Compare min and max */ | |
441 | ||
442 | if (args->key.min_objectid < args->key.max_objectid) | |
443 | return -1; | |
444 | if (args->key.min_objectid > args->key.max_objectid) | |
445 | return 1; | |
446 | ||
447 | if (args->key.min_type < args->key.max_type) | |
448 | return -1; | |
449 | if (args->key.min_type > args->key.max_type) | |
450 | return 1; | |
451 | ||
452 | if (args->key.min_offset < args->key.max_offset) | |
453 | return -1; | |
454 | if (args->key.min_offset > args->key.max_offset) | |
455 | return 1; | |
456 | ||
457 | return 0; | |
458 | } | |
459 | ||
460 | #define FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) \ | |
461 | for ((i) = 0, \ | |
462 | (sh) = (const struct btrfs_ioctl_search_header*) (args).buf; \ | |
463 | (i) < (args).key.nr_items; \ | |
464 | (i)++, \ | |
465 | (sh) = (const struct btrfs_ioctl_search_header*) ((uint8_t*) (sh) + sizeof(struct btrfs_ioctl_search_header) + (sh)->len)) | |
466 | ||
467 | #define BTRFS_IOCTL_SEARCH_HEADER_BODY(sh) \ | |
468 | ((void*) ((uint8_t*) sh + sizeof(struct btrfs_ioctl_search_header))) | |
469 | ||
5bcd08db | 470 | int btrfs_subvol_get_info_fd(int fd, uint64_t subvol_id, BtrfsSubvolInfo *ret) { |
10f9c755 LP |
471 | struct btrfs_ioctl_search_args args = { |
472 | /* Tree of tree roots */ | |
b6b18498 | 473 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, |
10f9c755 LP |
474 | |
475 | /* Look precisely for the subvolume items */ | |
476 | .key.min_type = BTRFS_ROOT_ITEM_KEY, | |
477 | .key.max_type = BTRFS_ROOT_ITEM_KEY, | |
478 | ||
10f9c755 LP |
479 | .key.min_offset = 0, |
480 | .key.max_offset = (uint64_t) -1, | |
5743a585 LP |
481 | |
482 | /* No restrictions on the other components */ | |
10f9c755 LP |
483 | .key.min_transid = 0, |
484 | .key.max_transid = (uint64_t) -1, | |
10f9c755 LP |
485 | }; |
486 | ||
b6b18498 | 487 | bool found = false; |
10f9c755 LP |
488 | int r; |
489 | ||
490 | assert(fd >= 0); | |
491 | assert(ret); | |
492 | ||
5bcd08db LP |
493 | if (subvol_id == 0) { |
494 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
495 | if (r < 0) | |
496 | return r; | |
497 | } else { | |
498 | r = btrfs_is_filesystem(fd); | |
499 | if (r < 0) | |
500 | return r; | |
501 | if (!r) | |
502 | return -ENOTTY; | |
503 | } | |
10f9c755 LP |
504 | |
505 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
10f9c755 | 506 | |
5743a585 | 507 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
508 | const struct btrfs_ioctl_search_header *sh; |
509 | unsigned i; | |
510 | ||
511 | args.key.nr_items = 256; | |
512 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
513 | return -errno; | |
514 | ||
515 | if (args.key.nr_items <= 0) | |
516 | break; | |
10f9c755 | 517 | |
5743a585 | 518 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
10f9c755 | 519 | |
b6b18498 LP |
520 | const struct btrfs_root_item *ri; |
521 | ||
5743a585 LP |
522 | /* Make sure we start the next search at least from this entry */ |
523 | btrfs_ioctl_search_args_set(&args, sh); | |
524 | ||
b6b18498 LP |
525 | if (sh->objectid != subvol_id) |
526 | continue; | |
527 | if (sh->type != BTRFS_ROOT_ITEM_KEY) | |
528 | continue; | |
5743a585 LP |
529 | |
530 | /* Older versions of the struct lacked the otime setting */ | |
b6b18498 LP |
531 | if (sh->len < offsetof(struct btrfs_root_item, otime) + sizeof(struct btrfs_timespec)) |
532 | continue; | |
10f9c755 | 533 | |
5743a585 | 534 | ri = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
10f9c755 | 535 | |
b6b18498 LP |
536 | ret->otime = (usec_t) le64toh(ri->otime.sec) * USEC_PER_SEC + |
537 | (usec_t) le32toh(ri->otime.nsec) / NSEC_PER_USEC; | |
10f9c755 | 538 | |
b6b18498 LP |
539 | ret->subvol_id = subvol_id; |
540 | ret->read_only = !!(le64toh(ri->flags) & BTRFS_ROOT_SUBVOL_RDONLY); | |
10f9c755 | 541 | |
b6b18498 LP |
542 | assert_cc(sizeof(ri->uuid) == sizeof(ret->uuid)); |
543 | memcpy(&ret->uuid, ri->uuid, sizeof(ret->uuid)); | |
544 | memcpy(&ret->parent_uuid, ri->parent_uuid, sizeof(ret->parent_uuid)); | |
545 | ||
546 | found = true; | |
547 | goto finish; | |
548 | } | |
549 | ||
5743a585 LP |
550 | /* Increase search key by one, to read the next item, if we can. */ |
551 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
552 | break; |
553 | } | |
554 | ||
555 | finish: | |
556 | if (!found) | |
557 | return -ENODATA; | |
558 | ||
559 | return 0; | |
560 | } | |
561 | ||
5bcd08db | 562 | int btrfs_qgroup_get_quota_fd(int fd, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
b6b18498 LP |
563 | |
564 | struct btrfs_ioctl_search_args args = { | |
565 | /* Tree of quota items */ | |
566 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
567 | ||
5743a585 LP |
568 | /* The object ID is always 0 */ |
569 | .key.min_objectid = 0, | |
570 | .key.max_objectid = 0, | |
571 | ||
b6b18498 LP |
572 | /* Look precisely for the quota items */ |
573 | .key.min_type = BTRFS_QGROUP_STATUS_KEY, | |
574 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
575 | ||
b6b18498 LP |
576 | /* No restrictions on the other components */ |
577 | .key.min_transid = 0, | |
578 | .key.max_transid = (uint64_t) -1, | |
579 | }; | |
580 | ||
b6b18498 LP |
581 | bool found_info = false, found_limit = false; |
582 | int r; | |
583 | ||
584 | assert(fd >= 0); | |
585 | assert(ret); | |
586 | ||
5bcd08db LP |
587 | if (qgroupid == 0) { |
588 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
589 | if (r < 0) | |
590 | return r; | |
591 | } else { | |
592 | r = btrfs_is_filesystem(fd); | |
593 | if (r < 0) | |
594 | return r; | |
595 | if (!r) | |
596 | return -ENOTTY; | |
597 | } | |
b6b18498 | 598 | |
5bcd08db | 599 | args.key.min_offset = args.key.max_offset = qgroupid; |
b6b18498 | 600 | |
5743a585 | 601 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
602 | const struct btrfs_ioctl_search_header *sh; |
603 | unsigned i; | |
604 | ||
605 | args.key.nr_items = 256; | |
12ee6186 LP |
606 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
607 | if (errno == ENOENT) /* quota tree is missing: quota disabled */ | |
608 | break; | |
609 | ||
b6b18498 | 610 | return -errno; |
12ee6186 | 611 | } |
b6b18498 LP |
612 | |
613 | if (args.key.nr_items <= 0) | |
614 | break; | |
615 | ||
5743a585 | 616 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
b6b18498 | 617 | |
5743a585 LP |
618 | /* Make sure we start the next search at least from this entry */ |
619 | btrfs_ioctl_search_args_set(&args, sh); | |
b6b18498 LP |
620 | |
621 | if (sh->objectid != 0) | |
622 | continue; | |
5bcd08db | 623 | if (sh->offset != qgroupid) |
b6b18498 LP |
624 | continue; |
625 | ||
b6b18498 | 626 | if (sh->type == BTRFS_QGROUP_INFO_KEY) { |
5743a585 | 627 | const struct btrfs_qgroup_info_item *qii = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 628 | |
cb81cd80 | 629 | ret->referenced = le64toh(qii->rfer); |
b6b18498 LP |
630 | ret->exclusive = le64toh(qii->excl); |
631 | ||
632 | found_info = true; | |
633 | ||
634 | } else if (sh->type == BTRFS_QGROUP_LIMIT_KEY) { | |
5743a585 | 635 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 636 | |
5bcd08db LP |
637 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_RFER) |
638 | ret->referenced_max = le64toh(qli->max_rfer); | |
639 | else | |
cb81cd80 | 640 | ret->referenced_max = (uint64_t) -1; |
5bcd08db LP |
641 | |
642 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_EXCL) | |
643 | ret->exclusive_max = le64toh(qli->max_excl); | |
644 | else | |
b6b18498 LP |
645 | ret->exclusive_max = (uint64_t) -1; |
646 | ||
647 | found_limit = true; | |
648 | } | |
649 | ||
650 | if (found_info && found_limit) | |
651 | goto finish; | |
652 | } | |
653 | ||
5743a585 LP |
654 | /* Increase search key by one, to read the next item, if we can. */ |
655 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
656 | break; |
657 | } | |
658 | ||
659 | finish: | |
660 | if (!found_limit && !found_info) | |
661 | return -ENODATA; | |
662 | ||
663 | if (!found_info) { | |
cb81cd80 | 664 | ret->referenced = (uint64_t) -1; |
b6b18498 LP |
665 | ret->exclusive = (uint64_t) -1; |
666 | } | |
667 | ||
668 | if (!found_limit) { | |
cb81cd80 | 669 | ret->referenced_max = (uint64_t) -1; |
b6b18498 LP |
670 | ret->exclusive_max = (uint64_t) -1; |
671 | } | |
10f9c755 LP |
672 | |
673 | return 0; | |
674 | } | |
f27a3864 | 675 | |
5bcd08db LP |
676 | int btrfs_qgroup_get_quota(const char *path, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
677 | _cleanup_close_ int fd = -1; | |
678 | ||
679 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
680 | if (fd < 0) | |
681 | return -errno; | |
682 | ||
683 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
684 | } | |
685 | ||
686 | int btrfs_subvol_find_subtree_qgroup(int fd, uint64_t subvol_id, uint64_t *ret) { | |
687 | uint64_t level, lowest = (uint64_t) -1, lowest_qgroupid = 0; | |
688 | _cleanup_free_ uint64_t *qgroups = NULL; | |
689 | int r, n, i; | |
690 | ||
691 | assert(fd >= 0); | |
692 | assert(ret); | |
693 | ||
694 | /* This finds the "subtree" qgroup for a specific | |
695 | * subvolume. This only works for subvolumes that have been | |
696 | * prepared with btrfs_subvol_auto_qgroup_fd() with | |
697 | * insert_intermediary_qgroup=true (or equivalent). For others | |
698 | * it will return the leaf qgroup instead. The two cases may | |
699 | * be distuingished via the return value, which is 1 in case | |
700 | * an appropriate "subtree" qgroup was found, and 0 | |
701 | * otherwise. */ | |
702 | ||
703 | if (subvol_id == 0) { | |
704 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
705 | if (r < 0) | |
706 | return r; | |
707 | } | |
708 | ||
709 | r = btrfs_qgroupid_split(subvol_id, &level, NULL); | |
710 | if (r < 0) | |
711 | return r; | |
712 | if (level != 0) /* Input must be a leaf qgroup */ | |
713 | return -EINVAL; | |
714 | ||
715 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
716 | if (n < 0) | |
717 | return n; | |
718 | ||
719 | for (i = 0; i < n; i++) { | |
720 | uint64_t id; | |
721 | ||
722 | r = btrfs_qgroupid_split(qgroups[i], &level, &id); | |
723 | if (r < 0) | |
724 | return r; | |
725 | ||
726 | if (id != subvol_id) | |
727 | continue; | |
728 | ||
729 | if (lowest == (uint64_t) -1 || level < lowest) { | |
730 | lowest_qgroupid = qgroups[i]; | |
731 | lowest = level; | |
732 | } | |
733 | } | |
734 | ||
735 | if (lowest == (uint64_t) -1) { | |
736 | /* No suitable higher-level qgroup found, let's return | |
737 | * the leaf qgroup instead, and indicate that with the | |
738 | * return value. */ | |
739 | ||
740 | *ret = subvol_id; | |
741 | return 0; | |
742 | } | |
743 | ||
744 | *ret = lowest_qgroupid; | |
745 | return 1; | |
746 | } | |
747 | ||
748 | int btrfs_subvol_get_subtree_quota_fd(int fd, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
749 | uint64_t qgroupid; | |
750 | int r; | |
751 | ||
752 | assert(fd >= 0); | |
753 | assert(ret); | |
754 | ||
755 | /* This determines the quota data of the qgroup with the | |
756 | * lowest level, that shares the id part with the specified | |
757 | * subvolume. This is useful for determining the quota data | |
758 | * for entire subvolume subtrees, as long as the subtrees have | |
759 | * been set up with btrfs_qgroup_subvol_auto_fd() or in a | |
760 | * compatible way */ | |
761 | ||
762 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
763 | if (r < 0) | |
764 | return r; | |
765 | ||
766 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
767 | } | |
768 | ||
769 | int btrfs_subvol_get_subtree_quota(const char *path, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
770 | _cleanup_close_ int fd = -1; | |
771 | ||
772 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
773 | if (fd < 0) | |
774 | return -errno; | |
775 | ||
776 | return btrfs_subvol_get_subtree_quota_fd(fd, subvol_id, ret); | |
777 | } | |
778 | ||
f27a3864 | 779 | int btrfs_defrag_fd(int fd) { |
62572894 LP |
780 | struct stat st; |
781 | ||
f27a3864 LP |
782 | assert(fd >= 0); |
783 | ||
62572894 LP |
784 | if (fstat(fd, &st) < 0) |
785 | return -errno; | |
786 | ||
787 | if (!S_ISREG(st.st_mode)) | |
788 | return -EINVAL; | |
789 | ||
f27a3864 LP |
790 | if (ioctl(fd, BTRFS_IOC_DEFRAG, NULL) < 0) |
791 | return -errno; | |
792 | ||
793 | return 0; | |
794 | } | |
795 | ||
796 | int btrfs_defrag(const char *p) { | |
797 | _cleanup_close_ int fd = -1; | |
798 | ||
799 | fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
800 | if (fd < 0) | |
801 | return -errno; | |
802 | ||
803 | return btrfs_defrag_fd(fd); | |
804 | } | |
754061ce LP |
805 | |
806 | int btrfs_quota_enable_fd(int fd, bool b) { | |
807 | struct btrfs_ioctl_quota_ctl_args args = { | |
808 | .cmd = b ? BTRFS_QUOTA_CTL_ENABLE : BTRFS_QUOTA_CTL_DISABLE, | |
809 | }; | |
62572894 | 810 | int r; |
754061ce LP |
811 | |
812 | assert(fd >= 0); | |
813 | ||
62572894 LP |
814 | r = btrfs_is_filesystem(fd); |
815 | if (r < 0) | |
816 | return r; | |
817 | if (!r) | |
818 | return -ENOTTY; | |
819 | ||
754061ce LP |
820 | if (ioctl(fd, BTRFS_IOC_QUOTA_CTL, &args) < 0) |
821 | return -errno; | |
822 | ||
823 | return 0; | |
824 | } | |
825 | ||
826 | int btrfs_quota_enable(const char *path, bool b) { | |
827 | _cleanup_close_ int fd = -1; | |
828 | ||
829 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
830 | if (fd < 0) | |
831 | return -errno; | |
832 | ||
833 | return btrfs_quota_enable_fd(fd, b); | |
834 | } | |
d6ce17c7 | 835 | |
5bcd08db LP |
836 | int btrfs_qgroup_set_limit_fd(int fd, uint64_t qgroupid, uint64_t referenced_max) { |
837 | ||
d6ce17c7 | 838 | struct btrfs_ioctl_qgroup_limit_args args = { |
5bcd08db | 839 | .lim.max_rfer = referenced_max, |
d6ce17c7 LP |
840 | .lim.flags = BTRFS_QGROUP_LIMIT_MAX_RFER, |
841 | }; | |
5bcd08db | 842 | unsigned c; |
62572894 | 843 | int r; |
d6ce17c7 LP |
844 | |
845 | assert(fd >= 0); | |
846 | ||
5bcd08db LP |
847 | if (qgroupid == 0) { |
848 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
849 | if (r < 0) | |
850 | return r; | |
851 | } else { | |
852 | r = btrfs_is_filesystem(fd); | |
853 | if (r < 0) | |
854 | return r; | |
855 | if (!r) | |
856 | return -ENOTTY; | |
857 | } | |
62572894 | 858 | |
5bcd08db LP |
859 | args.qgroupid = qgroupid; |
860 | ||
861 | for (c = 0;; c++) { | |
862 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &args) < 0) { | |
863 | ||
864 | if (errno == EBUSY && c < 10) { | |
865 | (void) btrfs_quota_scan_wait(fd); | |
866 | continue; | |
867 | } | |
868 | ||
869 | return -errno; | |
870 | } | |
871 | ||
872 | break; | |
873 | } | |
d6ce17c7 LP |
874 | |
875 | return 0; | |
876 | } | |
877 | ||
5bcd08db LP |
878 | int btrfs_qgroup_set_limit(const char *path, uint64_t qgroupid, uint64_t referenced_max) { |
879 | _cleanup_close_ int fd = -1; | |
880 | ||
881 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
882 | if (fd < 0) | |
883 | return -errno; | |
884 | ||
885 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
886 | } | |
887 | ||
888 | int btrfs_subvol_set_subtree_quota_limit_fd(int fd, uint64_t subvol_id, uint64_t referenced_max) { | |
889 | uint64_t qgroupid; | |
890 | int r; | |
891 | ||
892 | assert(fd >= 0); | |
893 | ||
894 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
895 | if (r < 0) | |
896 | return r; | |
897 | ||
898 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
899 | } | |
900 | ||
901 | int btrfs_subvol_set_subtree_quota_limit(const char *path, uint64_t subvol_id, uint64_t referenced_max) { | |
d6ce17c7 LP |
902 | _cleanup_close_ int fd = -1; |
903 | ||
904 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
905 | if (fd < 0) | |
906 | return -errno; | |
907 | ||
5bcd08db | 908 | return btrfs_subvol_set_subtree_quota_limit_fd(fd, subvol_id, referenced_max); |
d6ce17c7 | 909 | } |
efe02862 | 910 | |
26166c88 | 911 | int btrfs_resize_loopback_fd(int fd, uint64_t new_size, bool grow_only) { |
efe02862 LP |
912 | struct btrfs_ioctl_vol_args args = {}; |
913 | _cleanup_free_ char *p = NULL, *loop = NULL, *backing = NULL; | |
914 | _cleanup_close_ int loop_fd = -1, backing_fd = -1; | |
915 | struct stat st; | |
a7f7d1bd | 916 | dev_t dev = 0; |
efe02862 LP |
917 | int r; |
918 | ||
a90fb858 LP |
919 | /* In contrast to btrfs quota ioctls ftruncate() cannot make sense of "infinity" or file sizes > 2^31 */ |
920 | if (!FILE_SIZE_VALID(new_size)) | |
921 | return -EINVAL; | |
922 | ||
efe02862 LP |
923 | /* btrfs cannot handle file systems < 16M, hence use this as minimum */ |
924 | if (new_size < 16*1024*1024) | |
925 | new_size = 16*1024*1024; | |
926 | ||
927 | r = btrfs_get_block_device_fd(fd, &dev); | |
928 | if (r < 0) | |
929 | return r; | |
930 | if (r == 0) | |
931 | return -ENODEV; | |
932 | ||
933 | if (asprintf(&p, "/sys/dev/block/%u:%u/loop/backing_file", major(dev), minor(dev)) < 0) | |
934 | return -ENOMEM; | |
935 | r = read_one_line_file(p, &backing); | |
26166c88 LP |
936 | if (r == -ENOENT) |
937 | return -ENODEV; | |
efe02862 LP |
938 | if (r < 0) |
939 | return r; | |
940 | if (isempty(backing) || !path_is_absolute(backing)) | |
941 | return -ENODEV; | |
942 | ||
943 | backing_fd = open(backing, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
944 | if (backing_fd < 0) | |
945 | return -errno; | |
946 | ||
947 | if (fstat(backing_fd, &st) < 0) | |
948 | return -errno; | |
949 | if (!S_ISREG(st.st_mode)) | |
950 | return -ENODEV; | |
951 | ||
952 | if (new_size == (uint64_t) st.st_size) | |
953 | return 0; | |
954 | ||
26166c88 LP |
955 | if (grow_only && new_size < (uint64_t) st.st_size) |
956 | return -EINVAL; | |
957 | ||
efe02862 LP |
958 | if (asprintf(&loop, "/dev/block/%u:%u", major(dev), minor(dev)) < 0) |
959 | return -ENOMEM; | |
960 | loop_fd = open(loop, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
961 | if (loop_fd < 0) | |
962 | return -errno; | |
963 | ||
964 | if (snprintf(args.name, sizeof(args.name), "%" PRIu64, new_size) >= (int) sizeof(args.name)) | |
965 | return -EINVAL; | |
966 | ||
967 | if (new_size < (uint64_t) st.st_size) { | |
968 | /* Decrease size: first decrease btrfs size, then shorten loopback */ | |
969 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
970 | return -errno; | |
971 | } | |
972 | ||
973 | if (ftruncate(backing_fd, new_size) < 0) | |
974 | return -errno; | |
975 | ||
976 | if (ioctl(loop_fd, LOOP_SET_CAPACITY, 0) < 0) | |
977 | return -errno; | |
978 | ||
979 | if (new_size > (uint64_t) st.st_size) { | |
980 | /* Increase size: first enlarge loopback, then increase btrfs size */ | |
981 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
982 | return -errno; | |
983 | } | |
984 | ||
26166c88 LP |
985 | /* Make sure the free disk space is correctly updated for both file systems */ |
986 | (void) fsync(fd); | |
987 | (void) fsync(backing_fd); | |
988 | ||
989 | return 1; | |
efe02862 LP |
990 | } |
991 | ||
26166c88 | 992 | int btrfs_resize_loopback(const char *p, uint64_t new_size, bool grow_only) { |
efe02862 LP |
993 | _cleanup_close_ int fd = -1; |
994 | ||
995 | fd = open(p, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
996 | if (fd < 0) | |
997 | return -errno; | |
998 | ||
26166c88 | 999 | return btrfs_resize_loopback_fd(fd, new_size, grow_only); |
efe02862 | 1000 | } |
d9e2daaf | 1001 | |
5bcd08db | 1002 | int btrfs_qgroupid_make(uint64_t level, uint64_t id, uint64_t *ret) { |
3f952f92 LP |
1003 | assert(ret); |
1004 | ||
1005 | if (level >= (UINT64_C(1) << (64 - BTRFS_QGROUP_LEVEL_SHIFT))) | |
1006 | return -EINVAL; | |
1007 | ||
1008 | if (id >= (UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT)) | |
1009 | return -EINVAL; | |
1010 | ||
1011 | *ret = (level << BTRFS_QGROUP_LEVEL_SHIFT) | id; | |
1012 | return 0; | |
1013 | } | |
1014 | ||
5bcd08db LP |
1015 | int btrfs_qgroupid_split(uint64_t qgroupid, uint64_t *level, uint64_t *id) { |
1016 | assert(level || id); | |
1017 | ||
1018 | if (level) | |
1019 | *level = qgroupid >> BTRFS_QGROUP_LEVEL_SHIFT; | |
1020 | ||
1021 | if (id) | |
1022 | *id = qgroupid & ((UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT) - 1); | |
1023 | ||
1024 | return 0; | |
1025 | } | |
1026 | ||
1027 | static int qgroup_create_or_destroy(int fd, bool b, uint64_t qgroupid) { | |
3f952f92 LP |
1028 | |
1029 | struct btrfs_ioctl_qgroup_create_args args = { | |
1030 | .create = b, | |
5bcd08db | 1031 | .qgroupid = qgroupid, |
3f952f92 | 1032 | }; |
5bcd08db | 1033 | unsigned c; |
3f952f92 LP |
1034 | int r; |
1035 | ||
5bcd08db | 1036 | r = btrfs_is_filesystem(fd); |
3f952f92 LP |
1037 | if (r < 0) |
1038 | return r; | |
5bcd08db LP |
1039 | if (r == 0) |
1040 | return -ENOTTY; | |
1041 | ||
1042 | for (c = 0;; c++) { | |
1043 | if (ioctl(fd, BTRFS_IOC_QGROUP_CREATE, &args) < 0) { | |
1044 | ||
be6d467c LP |
1045 | /* If quota is not enabled, we get EINVAL. Turn this into a recognizable error */ |
1046 | if (errno == EINVAL) | |
1047 | return -ENOPROTOOPT; | |
1048 | ||
5bcd08db LP |
1049 | if (errno == EBUSY && c < 10) { |
1050 | (void) btrfs_quota_scan_wait(fd); | |
1051 | continue; | |
1052 | } | |
1053 | ||
1054 | return -errno; | |
1055 | } | |
3f952f92 | 1056 | |
5bcd08db LP |
1057 | break; |
1058 | } | |
1059 | ||
1060 | return 0; | |
1061 | } | |
1062 | ||
1063 | int btrfs_qgroup_create(int fd, uint64_t qgroupid) { | |
1064 | return qgroup_create_or_destroy(fd, true, qgroupid); | |
1065 | } | |
1066 | ||
1067 | int btrfs_qgroup_destroy(int fd, uint64_t qgroupid) { | |
1068 | return qgroup_create_or_destroy(fd, false, qgroupid); | |
1069 | } | |
1070 | ||
1071 | int btrfs_qgroup_destroy_recursive(int fd, uint64_t qgroupid) { | |
1072 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1073 | uint64_t subvol_id; | |
1074 | int i, n, r; | |
1075 | ||
1076 | /* Destroys the specified qgroup, but unassigns it from all | |
1077 | * its parents first. Also, it recursively destroys all | |
1078 | * qgroups it is assgined to that have the same id part of the | |
1079 | * qgroupid as the specified group. */ | |
1080 | ||
1081 | r = btrfs_qgroupid_split(qgroupid, NULL, &subvol_id); | |
1082 | if (r < 0) | |
1083 | return r; | |
1084 | ||
1085 | n = btrfs_qgroup_find_parents(fd, qgroupid, &qgroups); | |
1086 | if (n < 0) | |
1087 | return n; | |
1088 | ||
1089 | for (i = 0; i < n; i++) { | |
1090 | uint64_t id; | |
1091 | ||
1092 | r = btrfs_qgroupid_split(qgroups[i], NULL, &id); | |
1093 | if (r < 0) | |
1094 | return r; | |
1095 | ||
1096 | r = btrfs_qgroup_unassign(fd, qgroupid, qgroups[i]); | |
1097 | if (r < 0) | |
1098 | return r; | |
1099 | ||
1100 | if (id != subvol_id) | |
1101 | continue; | |
1102 | ||
1103 | /* The parent qgroupid shares the same id part with | |
1104 | * us? If so, destroy it too. */ | |
1105 | ||
1106 | (void) btrfs_qgroup_destroy_recursive(fd, qgroups[i]); | |
1107 | } | |
1108 | ||
1109 | return btrfs_qgroup_destroy(fd, qgroupid); | |
1110 | } | |
1111 | ||
1112 | int btrfs_quota_scan_start(int fd) { | |
1113 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1114 | ||
1115 | assert(fd >= 0); | |
1116 | ||
1117 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN, &args) < 0) | |
3f952f92 LP |
1118 | return -errno; |
1119 | ||
1120 | return 0; | |
1121 | } | |
1122 | ||
5bcd08db LP |
1123 | int btrfs_quota_scan_wait(int fd) { |
1124 | assert(fd >= 0); | |
1125 | ||
1126 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_WAIT) < 0) | |
1127 | return -errno; | |
1128 | ||
1129 | return 0; | |
3f952f92 LP |
1130 | } |
1131 | ||
5bcd08db LP |
1132 | int btrfs_quota_scan_ongoing(int fd) { |
1133 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1134 | ||
1135 | assert(fd >= 0); | |
1136 | ||
1137 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_STATUS, &args) < 0) | |
1138 | return -errno; | |
1139 | ||
1140 | return !!args.flags; | |
3f952f92 LP |
1141 | } |
1142 | ||
5bcd08db LP |
1143 | static int qgroup_assign_or_unassign(int fd, bool b, uint64_t child, uint64_t parent) { |
1144 | struct btrfs_ioctl_qgroup_assign_args args = { | |
1145 | .assign = b, | |
1146 | .src = child, | |
1147 | .dst = parent, | |
1148 | }; | |
1149 | unsigned c; | |
1150 | int r; | |
1151 | ||
1152 | r = btrfs_is_filesystem(fd); | |
1153 | if (r < 0) | |
1154 | return r; | |
1155 | if (r == 0) | |
1156 | return -ENOTTY; | |
1157 | ||
1158 | for (c = 0;; c++) { | |
1159 | r = ioctl(fd, BTRFS_IOC_QGROUP_ASSIGN, &args); | |
1160 | if (r < 0) { | |
1161 | if (errno == EBUSY && c < 10) { | |
1162 | (void) btrfs_quota_scan_wait(fd); | |
1163 | continue; | |
1164 | } | |
1165 | ||
1166 | return -errno; | |
1167 | } | |
1168 | ||
1169 | if (r == 0) | |
1170 | return 0; | |
1171 | ||
1172 | /* If the return value is > 0, we need to request a rescan */ | |
1173 | ||
1174 | (void) btrfs_quota_scan_start(fd); | |
1175 | return 1; | |
1176 | } | |
1177 | } | |
1178 | ||
1179 | int btrfs_qgroup_assign(int fd, uint64_t child, uint64_t parent) { | |
1180 | return qgroup_assign_or_unassign(fd, true, child, parent); | |
1181 | } | |
1182 | ||
1183 | int btrfs_qgroup_unassign(int fd, uint64_t child, uint64_t parent) { | |
1184 | return qgroup_assign_or_unassign(fd, false, child, parent); | |
1185 | } | |
1186 | ||
1187 | static int subvol_remove_children(int fd, const char *subvolume, uint64_t subvol_id, BtrfsRemoveFlags flags) { | |
d9e2daaf LP |
1188 | struct btrfs_ioctl_search_args args = { |
1189 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1190 | ||
1191 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1192 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1193 | ||
1194 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1195 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1196 | ||
1197 | .key.min_transid = 0, | |
1198 | .key.max_transid = (uint64_t) -1, | |
1199 | }; | |
1200 | ||
1201 | struct btrfs_ioctl_vol_args vol_args = {}; | |
1202 | _cleanup_close_ int subvol_fd = -1; | |
62572894 | 1203 | struct stat st; |
3986b258 | 1204 | bool made_writable = false; |
d9e2daaf LP |
1205 | int r; |
1206 | ||
1207 | assert(fd >= 0); | |
1208 | assert(subvolume); | |
1209 | ||
62572894 LP |
1210 | if (fstat(fd, &st) < 0) |
1211 | return -errno; | |
1212 | ||
1213 | if (!S_ISDIR(st.st_mode)) | |
1214 | return -EINVAL; | |
1215 | ||
f7c9f4a2 | 1216 | subvol_fd = openat(fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1217 | if (subvol_fd < 0) |
1218 | return -errno; | |
1219 | ||
1220 | if (subvol_id == 0) { | |
1221 | r = btrfs_subvol_get_id_fd(subvol_fd, &subvol_id); | |
1222 | if (r < 0) | |
1223 | return r; | |
1224 | } | |
1225 | ||
3f952f92 LP |
1226 | /* First, try to remove the subvolume. If it happens to be |
1227 | * already empty, this will just work. */ | |
1228 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
1229 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) >= 0) { | |
5bcd08db | 1230 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); /* for the leaf subvolumes, the qgroup id is identical to the subvol id */ |
3f952f92 LP |
1231 | return 0; |
1232 | } | |
5bcd08db | 1233 | if (!(flags & BTRFS_REMOVE_RECURSIVE) || errno != ENOTEMPTY) |
3f952f92 LP |
1234 | return -errno; |
1235 | ||
1236 | /* OK, the subvolume is not empty, let's look for child | |
1237 | * subvolumes, and remove them, first */ | |
1238 | ||
d9e2daaf LP |
1239 | args.key.min_offset = args.key.max_offset = subvol_id; |
1240 | ||
1241 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1242 | const struct btrfs_ioctl_search_header *sh; | |
1243 | unsigned i; | |
1244 | ||
1245 | args.key.nr_items = 256; | |
1246 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1247 | return -errno; | |
1248 | ||
1249 | if (args.key.nr_items <= 0) | |
1250 | break; | |
1251 | ||
1252 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1253 | _cleanup_free_ char *p = NULL; | |
1254 | const struct btrfs_root_ref *ref; | |
1255 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1256 | ||
1257 | btrfs_ioctl_search_args_set(&args, sh); | |
1258 | ||
1259 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1260 | continue; | |
1261 | if (sh->offset != subvol_id) | |
1262 | continue; | |
1263 | ||
1264 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1265 | ||
1266 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); | |
1267 | if (!p) | |
1268 | return -ENOMEM; | |
1269 | ||
1270 | zero(ino_args); | |
1271 | ino_args.treeid = subvol_id; | |
cbf21ecc | 1272 | ino_args.objectid = htole64(ref->dirid); |
d9e2daaf LP |
1273 | |
1274 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1275 | return -errno; | |
1276 | ||
3986b258 LP |
1277 | if (!made_writable) { |
1278 | r = btrfs_subvol_set_read_only_fd(subvol_fd, false); | |
1279 | if (r < 0) | |
1280 | return r; | |
1281 | ||
1282 | made_writable = true; | |
1283 | } | |
1284 | ||
d9e2daaf LP |
1285 | if (isempty(ino_args.name)) |
1286 | /* Subvolume is in the top-level | |
1287 | * directory of the subvolume. */ | |
5bcd08db | 1288 | r = subvol_remove_children(subvol_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1289 | else { |
1290 | _cleanup_close_ int child_fd = -1; | |
1291 | ||
1292 | /* Subvolume is somewhere further down, | |
1293 | * hence we need to open the | |
1294 | * containing directory first */ | |
1295 | ||
f7c9f4a2 | 1296 | child_fd = openat(subvol_fd, ino_args.name, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1297 | if (child_fd < 0) |
1298 | return -errno; | |
1299 | ||
5bcd08db | 1300 | r = subvol_remove_children(child_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1301 | } |
1302 | if (r < 0) | |
1303 | return r; | |
1304 | } | |
1305 | ||
1306 | /* Increase search key by one, to read the next item, if we can. */ | |
1307 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1308 | break; | |
1309 | } | |
1310 | ||
1311 | /* OK, the child subvolumes should all be gone now, let's try | |
1312 | * again to remove the subvolume */ | |
1313 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) < 0) | |
1314 | return -errno; | |
1315 | ||
5bcd08db | 1316 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); |
d9e2daaf LP |
1317 | return 0; |
1318 | } | |
1319 | ||
5bcd08db | 1320 | int btrfs_subvol_remove(const char *path, BtrfsRemoveFlags flags) { |
d9e2daaf LP |
1321 | _cleanup_close_ int fd = -1; |
1322 | const char *subvolume; | |
1323 | int r; | |
1324 | ||
1325 | assert(path); | |
1326 | ||
1327 | r = extract_subvolume_name(path, &subvolume); | |
1328 | if (r < 0) | |
1329 | return r; | |
1330 | ||
1331 | fd = open_parent(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1332 | if (fd < 0) | |
1333 | return fd; | |
1334 | ||
5bcd08db LP |
1335 | return subvol_remove_children(fd, subvolume, 0, flags); |
1336 | } | |
1337 | ||
1338 | int btrfs_subvol_remove_fd(int fd, const char *subvolume, BtrfsRemoveFlags flags) { | |
1339 | return subvol_remove_children(fd, subvolume, 0, flags); | |
1340 | } | |
1341 | ||
1342 | int btrfs_qgroup_copy_limits(int fd, uint64_t old_qgroupid, uint64_t new_qgroupid) { | |
1343 | ||
1344 | struct btrfs_ioctl_search_args args = { | |
1345 | /* Tree of quota items */ | |
1346 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1347 | ||
1348 | /* The object ID is always 0 */ | |
1349 | .key.min_objectid = 0, | |
1350 | .key.max_objectid = 0, | |
1351 | ||
1352 | /* Look precisely for the quota items */ | |
1353 | .key.min_type = BTRFS_QGROUP_LIMIT_KEY, | |
1354 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
1355 | ||
1356 | /* For our qgroup */ | |
1357 | .key.min_offset = old_qgroupid, | |
1358 | .key.max_offset = old_qgroupid, | |
1359 | ||
1360 | /* No restrictions on the other components */ | |
1361 | .key.min_transid = 0, | |
1362 | .key.max_transid = (uint64_t) -1, | |
1363 | }; | |
1364 | ||
1365 | int r; | |
1366 | ||
1367 | r = btrfs_is_filesystem(fd); | |
1368 | if (r < 0) | |
1369 | return r; | |
1370 | if (!r) | |
1371 | return -ENOTTY; | |
1372 | ||
1373 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1374 | const struct btrfs_ioctl_search_header *sh; | |
1375 | unsigned i; | |
1376 | ||
1377 | args.key.nr_items = 256; | |
12ee6186 LP |
1378 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1379 | if (errno == ENOENT) /* quota tree missing: quota is not enabled, hence nothing to copy */ | |
1380 | break; | |
1381 | ||
5bcd08db | 1382 | return -errno; |
12ee6186 | 1383 | } |
5bcd08db LP |
1384 | |
1385 | if (args.key.nr_items <= 0) | |
1386 | break; | |
1387 | ||
1388 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1389 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1390 | struct btrfs_ioctl_qgroup_limit_args qargs; | |
1391 | unsigned c; | |
1392 | ||
1393 | /* Make sure we start the next search at least from this entry */ | |
1394 | btrfs_ioctl_search_args_set(&args, sh); | |
1395 | ||
1396 | if (sh->objectid != 0) | |
1397 | continue; | |
1398 | if (sh->type != BTRFS_QGROUP_LIMIT_KEY) | |
1399 | continue; | |
1400 | if (sh->offset != old_qgroupid) | |
1401 | continue; | |
1402 | ||
1403 | /* We found the entry, now copy things over. */ | |
1404 | ||
1405 | qargs = (struct btrfs_ioctl_qgroup_limit_args) { | |
1406 | .qgroupid = new_qgroupid, | |
1407 | ||
1408 | .lim.max_rfer = le64toh(qli->max_rfer), | |
1409 | .lim.max_excl = le64toh(qli->max_excl), | |
1410 | .lim.rsv_rfer = le64toh(qli->rsv_rfer), | |
1411 | .lim.rsv_excl = le64toh(qli->rsv_excl), | |
1412 | ||
1413 | .lim.flags = le64toh(qli->flags) & (BTRFS_QGROUP_LIMIT_MAX_RFER| | |
1414 | BTRFS_QGROUP_LIMIT_MAX_EXCL| | |
1415 | BTRFS_QGROUP_LIMIT_RSV_RFER| | |
1416 | BTRFS_QGROUP_LIMIT_RSV_EXCL), | |
1417 | }; | |
1418 | ||
1419 | for (c = 0;; c++) { | |
1420 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &qargs) < 0) { | |
1421 | if (errno == EBUSY && c < 10) { | |
1422 | (void) btrfs_quota_scan_wait(fd); | |
1423 | continue; | |
1424 | } | |
1425 | return -errno; | |
1426 | } | |
1427 | ||
1428 | break; | |
1429 | } | |
1430 | ||
1431 | return 1; | |
1432 | } | |
1433 | ||
1434 | /* Increase search key by one, to read the next item, if we can. */ | |
1435 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1436 | break; | |
1437 | } | |
1438 | ||
1439 | return 0; | |
d9e2daaf LP |
1440 | } |
1441 | ||
5bcd08db LP |
1442 | static int copy_quota_hierarchy(int fd, uint64_t old_subvol_id, uint64_t new_subvol_id) { |
1443 | _cleanup_free_ uint64_t *old_qgroups = NULL, *old_parent_qgroups = NULL; | |
1444 | bool copy_from_parent = false, insert_intermediary_qgroup = false; | |
1445 | int n_old_qgroups, n_old_parent_qgroups, r, i; | |
1446 | uint64_t old_parent_id; | |
1447 | ||
1448 | assert(fd >= 0); | |
1449 | ||
1450 | /* Copies a reduced form of quota information from the old to | |
1451 | * the new subvolume. */ | |
1452 | ||
1453 | n_old_qgroups = btrfs_qgroup_find_parents(fd, old_subvol_id, &old_qgroups); | |
1454 | if (n_old_qgroups <= 0) /* Nothing to copy */ | |
1455 | return n_old_qgroups; | |
1456 | ||
1457 | r = btrfs_subvol_get_parent(fd, old_subvol_id, &old_parent_id); | |
08c77cf3 LP |
1458 | if (r == -ENXIO) |
1459 | /* We have no parent, hence nothing to copy. */ | |
1460 | n_old_parent_qgroups = 0; | |
1461 | else if (r < 0) | |
5bcd08db | 1462 | return r; |
08c77cf3 LP |
1463 | else { |
1464 | n_old_parent_qgroups = btrfs_qgroup_find_parents(fd, old_parent_id, &old_parent_qgroups); | |
1465 | if (n_old_parent_qgroups < 0) | |
1466 | return n_old_parent_qgroups; | |
1467 | } | |
5bcd08db LP |
1468 | |
1469 | for (i = 0; i < n_old_qgroups; i++) { | |
1470 | uint64_t id; | |
1471 | int j; | |
1472 | ||
1473 | r = btrfs_qgroupid_split(old_qgroups[i], NULL, &id); | |
1474 | if (r < 0) | |
1475 | return r; | |
1476 | ||
1477 | if (id == old_subvol_id) { | |
1478 | /* The old subvolume was member of a qgroup | |
1479 | * that had the same id, but a different level | |
1480 | * as it self. Let's set up something similar | |
1481 | * in the destination. */ | |
1482 | insert_intermediary_qgroup = true; | |
1483 | break; | |
1484 | } | |
1485 | ||
1486 | for (j = 0; j < n_old_parent_qgroups; j++) | |
1487 | if (old_parent_qgroups[j] == old_qgroups[i]) { | |
1488 | /* The old subvolume shared a common | |
1489 | * parent qgroup with its parent | |
1490 | * subvolume. Let's set up something | |
1491 | * similar in the destination. */ | |
1492 | copy_from_parent = true; | |
1493 | } | |
1494 | } | |
1495 | ||
1496 | if (!insert_intermediary_qgroup && !copy_from_parent) | |
1497 | return 0; | |
1498 | ||
1499 | return btrfs_subvol_auto_qgroup_fd(fd, new_subvol_id, insert_intermediary_qgroup); | |
1500 | } | |
1501 | ||
1502 | static int copy_subtree_quota_limits(int fd, uint64_t old_subvol, uint64_t new_subvol) { | |
1503 | uint64_t old_subtree_qgroup, new_subtree_qgroup; | |
1504 | bool changed; | |
1505 | int r; | |
1506 | ||
1507 | /* First copy the leaf limits */ | |
1508 | r = btrfs_qgroup_copy_limits(fd, old_subvol, new_subvol); | |
1509 | if (r < 0) | |
1510 | return r; | |
1511 | changed = r > 0; | |
1512 | ||
1513 | /* Then, try to copy the subtree limits, if there are any. */ | |
1514 | r = btrfs_subvol_find_subtree_qgroup(fd, old_subvol, &old_subtree_qgroup); | |
1515 | if (r < 0) | |
1516 | return r; | |
1517 | if (r == 0) | |
1518 | return changed; | |
1519 | ||
1520 | r = btrfs_subvol_find_subtree_qgroup(fd, new_subvol, &new_subtree_qgroup); | |
1521 | if (r < 0) | |
1522 | return r; | |
1523 | if (r == 0) | |
1524 | return changed; | |
1525 | ||
1526 | r = btrfs_qgroup_copy_limits(fd, old_subtree_qgroup, new_subtree_qgroup); | |
1527 | if (r != 0) | |
1528 | return r; | |
1529 | ||
1530 | return changed; | |
d9e2daaf | 1531 | } |
f70a17f8 | 1532 | |
90578cbd | 1533 | static int subvol_snapshot_children(int old_fd, int new_fd, const char *subvolume, uint64_t old_subvol_id, BtrfsSnapshotFlags flags) { |
f70a17f8 LP |
1534 | |
1535 | struct btrfs_ioctl_search_args args = { | |
1536 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1537 | ||
1538 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1539 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1540 | ||
1541 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1542 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1543 | ||
1544 | .key.min_transid = 0, | |
1545 | .key.max_transid = (uint64_t) -1, | |
1546 | }; | |
1547 | ||
1548 | struct btrfs_ioctl_vol_args_v2 vol_args = { | |
1549 | .flags = flags & BTRFS_SNAPSHOT_READ_ONLY ? BTRFS_SUBVOL_RDONLY : 0, | |
1550 | .fd = old_fd, | |
1551 | }; | |
ffb296b2 | 1552 | _cleanup_close_ int subvolume_fd = -1; |
90578cbd LP |
1553 | uint64_t new_subvol_id; |
1554 | int r; | |
f70a17f8 LP |
1555 | |
1556 | assert(old_fd >= 0); | |
1557 | assert(new_fd >= 0); | |
1558 | assert(subvolume); | |
1559 | ||
1560 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
f70a17f8 LP |
1561 | |
1562 | if (ioctl(new_fd, BTRFS_IOC_SNAP_CREATE_V2, &vol_args) < 0) | |
1563 | return -errno; | |
1564 | ||
5bcd08db LP |
1565 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE) && |
1566 | !(flags & BTRFS_SNAPSHOT_QUOTA)) | |
f70a17f8 LP |
1567 | return 0; |
1568 | ||
90578cbd LP |
1569 | if (old_subvol_id == 0) { |
1570 | r = btrfs_subvol_get_id_fd(old_fd, &old_subvol_id); | |
f70a17f8 LP |
1571 | if (r < 0) |
1572 | return r; | |
1573 | } | |
1574 | ||
90578cbd LP |
1575 | r = btrfs_subvol_get_id(new_fd, vol_args.name, &new_subvol_id); |
1576 | if (r < 0) | |
1577 | return r; | |
1578 | ||
5bcd08db LP |
1579 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1580 | (void) copy_quota_hierarchy(new_fd, old_subvol_id, new_subvol_id); | |
1581 | ||
1582 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE)) { | |
1583 | ||
1584 | if (flags & BTRFS_SNAPSHOT_QUOTA) | |
1585 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1586 | ||
1587 | return 0; | |
1588 | } | |
1589 | ||
90578cbd | 1590 | args.key.min_offset = args.key.max_offset = old_subvol_id; |
f70a17f8 LP |
1591 | |
1592 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1593 | const struct btrfs_ioctl_search_header *sh; | |
1594 | unsigned i; | |
1595 | ||
1596 | args.key.nr_items = 256; | |
1597 | if (ioctl(old_fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1598 | return -errno; | |
1599 | ||
1600 | if (args.key.nr_items <= 0) | |
1601 | break; | |
1602 | ||
1603 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1604 | _cleanup_free_ char *p = NULL, *c = NULL, *np = NULL; | |
1605 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1606 | const struct btrfs_root_ref *ref; | |
1607 | _cleanup_close_ int old_child_fd = -1, new_child_fd = -1; | |
1608 | ||
1609 | btrfs_ioctl_search_args_set(&args, sh); | |
1610 | ||
1611 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1612 | continue; | |
90578cbd LP |
1613 | |
1614 | /* Avoid finding the source subvolume a second | |
1615 | * time */ | |
1616 | if (sh->offset != old_subvol_id) | |
f70a17f8 LP |
1617 | continue; |
1618 | ||
90578cbd LP |
1619 | /* Avoid running into loops if the new |
1620 | * subvolume is below the old one. */ | |
1621 | if (sh->objectid == new_subvol_id) | |
1622 | continue; | |
f70a17f8 | 1623 | |
90578cbd | 1624 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
f70a17f8 LP |
1625 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); |
1626 | if (!p) | |
1627 | return -ENOMEM; | |
1628 | ||
1629 | zero(ino_args); | |
90578cbd | 1630 | ino_args.treeid = old_subvol_id; |
f70a17f8 LP |
1631 | ino_args.objectid = htole64(ref->dirid); |
1632 | ||
1633 | if (ioctl(old_fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1634 | return -errno; | |
1635 | ||
1636 | /* The kernel returns an empty name if the | |
1637 | * subvolume is in the top-level directory, | |
1638 | * and otherwise appends a slash, so that we | |
1639 | * can just concatenate easily here, without | |
1640 | * adding a slash. */ | |
1641 | c = strappend(ino_args.name, p); | |
1642 | if (!c) | |
1643 | return -ENOMEM; | |
1644 | ||
f7c9f4a2 | 1645 | old_child_fd = openat(old_fd, c, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1646 | if (old_child_fd < 0) |
1647 | return -errno; | |
1648 | ||
605405c6 | 1649 | np = strjoin(subvolume, "/", ino_args.name); |
f70a17f8 LP |
1650 | if (!np) |
1651 | return -ENOMEM; | |
1652 | ||
f7c9f4a2 | 1653 | new_child_fd = openat(new_fd, np, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1654 | if (new_child_fd < 0) |
1655 | return -errno; | |
1656 | ||
ffb296b2 LP |
1657 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { |
1658 | /* If the snapshot is read-only we | |
1659 | * need to mark it writable | |
1660 | * temporarily, to put the subsnapshot | |
1661 | * into place. */ | |
1662 | ||
1663 | if (subvolume_fd < 0) { | |
f7c9f4a2 | 1664 | subvolume_fd = openat(new_fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
ffb296b2 LP |
1665 | if (subvolume_fd < 0) |
1666 | return -errno; | |
1667 | } | |
1668 | ||
1669 | r = btrfs_subvol_set_read_only_fd(subvolume_fd, false); | |
1670 | if (r < 0) | |
1671 | return r; | |
1672 | } | |
1673 | ||
f70a17f8 | 1674 | /* When btrfs clones the subvolumes, child |
90578cbd | 1675 | * subvolumes appear as empty directories. Remove |
f70a17f8 LP |
1676 | * them, so that we can create a new snapshot |
1677 | * in their place */ | |
ffb296b2 LP |
1678 | if (unlinkat(new_child_fd, p, AT_REMOVEDIR) < 0) { |
1679 | int k = -errno; | |
1680 | ||
1681 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) | |
1682 | (void) btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1683 | ||
1684 | return k; | |
1685 | } | |
f70a17f8 LP |
1686 | |
1687 | r = subvol_snapshot_children(old_child_fd, new_child_fd, p, sh->objectid, flags & ~BTRFS_SNAPSHOT_FALLBACK_COPY); | |
ffb296b2 LP |
1688 | |
1689 | /* Restore the readonly flag */ | |
1690 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
1691 | int k; | |
1692 | ||
1693 | k = btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1694 | if (r >= 0 && k < 0) | |
1695 | return k; | |
1696 | } | |
1697 | ||
f70a17f8 LP |
1698 | if (r < 0) |
1699 | return r; | |
1700 | } | |
1701 | ||
1702 | /* Increase search key by one, to read the next item, if we can. */ | |
1703 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1704 | break; | |
1705 | } | |
1706 | ||
5bcd08db LP |
1707 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1708 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1709 | ||
f70a17f8 LP |
1710 | return 0; |
1711 | } | |
1712 | ||
1713 | int btrfs_subvol_snapshot_fd(int old_fd, const char *new_path, BtrfsSnapshotFlags flags) { | |
1714 | _cleanup_close_ int new_fd = -1; | |
1715 | const char *subvolume; | |
1716 | int r; | |
1717 | ||
1718 | assert(old_fd >= 0); | |
1719 | assert(new_path); | |
1720 | ||
2904e949 | 1721 | r = btrfs_is_subvol_fd(old_fd); |
f70a17f8 LP |
1722 | if (r < 0) |
1723 | return r; | |
1724 | if (r == 0) { | |
17cbb288 LP |
1725 | bool plain_directory = false; |
1726 | ||
1727 | /* If the source isn't a proper subvolume, fail unless fallback is requested */ | |
f70a17f8 LP |
1728 | if (!(flags & BTRFS_SNAPSHOT_FALLBACK_COPY)) |
1729 | return -EISDIR; | |
1730 | ||
1731 | r = btrfs_subvol_make(new_path); | |
17cbb288 LP |
1732 | if (r == -ENOTTY && (flags & BTRFS_SNAPSHOT_FALLBACK_DIRECTORY)) { |
1733 | /* If the destination doesn't support subvolumes, then use a plain directory, if that's requested. */ | |
1734 | if (mkdir(new_path, 0755) < 0) | |
1735 | return r; | |
1736 | ||
1737 | plain_directory = true; | |
1738 | } else if (r < 0) | |
f70a17f8 LP |
1739 | return r; |
1740 | ||
1c876927 | 1741 | r = copy_directory_fd(old_fd, new_path, COPY_MERGE|COPY_REFLINK); |
17cbb288 LP |
1742 | if (r < 0) |
1743 | goto fallback_fail; | |
f70a17f8 LP |
1744 | |
1745 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
17cbb288 LP |
1746 | |
1747 | if (plain_directory) { | |
1748 | /* Plain directories have no recursive read-only flag, but something pretty close to | |
1749 | * it: the IMMUTABLE bit. Let's use this here, if this is requested. */ | |
1750 | ||
1751 | if (flags & BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE) | |
1752 | (void) chattr_path(new_path, FS_IMMUTABLE_FL, FS_IMMUTABLE_FL); | |
1753 | } else { | |
1754 | r = btrfs_subvol_set_read_only(new_path, true); | |
1755 | if (r < 0) | |
1756 | goto fallback_fail; | |
f70a17f8 LP |
1757 | } |
1758 | } | |
1759 | ||
1760 | return 0; | |
17cbb288 LP |
1761 | |
1762 | fallback_fail: | |
1763 | (void) rm_rf(new_path, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME); | |
1764 | return r; | |
f70a17f8 LP |
1765 | } |
1766 | ||
1767 | r = extract_subvolume_name(new_path, &subvolume); | |
1768 | if (r < 0) | |
1769 | return r; | |
1770 | ||
1771 | new_fd = open_parent(new_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1772 | if (new_fd < 0) | |
1773 | return new_fd; | |
1774 | ||
1775 | return subvol_snapshot_children(old_fd, new_fd, subvolume, 0, flags); | |
1776 | } | |
1777 | ||
1778 | int btrfs_subvol_snapshot(const char *old_path, const char *new_path, BtrfsSnapshotFlags flags) { | |
1779 | _cleanup_close_ int old_fd = -1; | |
1780 | ||
1781 | assert(old_path); | |
1782 | assert(new_path); | |
1783 | ||
1784 | old_fd = open(old_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1785 | if (old_fd < 0) | |
1786 | return -errno; | |
1787 | ||
1788 | return btrfs_subvol_snapshot_fd(old_fd, new_path, flags); | |
1789 | } | |
5bcd08db LP |
1790 | |
1791 | int btrfs_qgroup_find_parents(int fd, uint64_t qgroupid, uint64_t **ret) { | |
1792 | ||
1793 | struct btrfs_ioctl_search_args args = { | |
1794 | /* Tree of quota items */ | |
1795 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1796 | ||
1797 | /* Look precisely for the quota relation items */ | |
1798 | .key.min_type = BTRFS_QGROUP_RELATION_KEY, | |
1799 | .key.max_type = BTRFS_QGROUP_RELATION_KEY, | |
1800 | ||
1801 | /* No restrictions on the other components */ | |
1802 | .key.min_offset = 0, | |
1803 | .key.max_offset = (uint64_t) -1, | |
1804 | ||
1805 | .key.min_transid = 0, | |
1806 | .key.max_transid = (uint64_t) -1, | |
1807 | }; | |
1808 | ||
1809 | _cleanup_free_ uint64_t *items = NULL; | |
1810 | size_t n_items = 0, n_allocated = 0; | |
1811 | int r; | |
1812 | ||
1813 | assert(fd >= 0); | |
1814 | assert(ret); | |
1815 | ||
1816 | if (qgroupid == 0) { | |
1817 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
1818 | if (r < 0) | |
1819 | return r; | |
1820 | } else { | |
1821 | r = btrfs_is_filesystem(fd); | |
1822 | if (r < 0) | |
1823 | return r; | |
1824 | if (!r) | |
1825 | return -ENOTTY; | |
1826 | } | |
1827 | ||
1828 | args.key.min_objectid = args.key.max_objectid = qgroupid; | |
1829 | ||
1830 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1831 | const struct btrfs_ioctl_search_header *sh; | |
1832 | unsigned i; | |
1833 | ||
1834 | args.key.nr_items = 256; | |
12ee6186 LP |
1835 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1836 | if (errno == ENOENT) /* quota tree missing: quota is disabled */ | |
1837 | break; | |
1838 | ||
5bcd08db | 1839 | return -errno; |
12ee6186 | 1840 | } |
5bcd08db LP |
1841 | |
1842 | if (args.key.nr_items <= 0) | |
1843 | break; | |
1844 | ||
1845 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1846 | ||
1847 | /* Make sure we start the next search at least from this entry */ | |
1848 | btrfs_ioctl_search_args_set(&args, sh); | |
1849 | ||
1850 | if (sh->type != BTRFS_QGROUP_RELATION_KEY) | |
1851 | continue; | |
1852 | if (sh->offset < sh->objectid) | |
1853 | continue; | |
1854 | if (sh->objectid != qgroupid) | |
1855 | continue; | |
1856 | ||
1857 | if (!GREEDY_REALLOC(items, n_allocated, n_items+1)) | |
1858 | return -ENOMEM; | |
1859 | ||
1860 | items[n_items++] = sh->offset; | |
1861 | } | |
1862 | ||
1863 | /* Increase search key by one, to read the next item, if we can. */ | |
1864 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1865 | break; | |
1866 | } | |
1867 | ||
1868 | if (n_items <= 0) { | |
1869 | *ret = NULL; | |
1870 | return 0; | |
1871 | } | |
1872 | ||
1873 | *ret = items; | |
1874 | items = NULL; | |
1875 | ||
1876 | return (int) n_items; | |
1877 | } | |
1878 | ||
1879 | int btrfs_subvol_auto_qgroup_fd(int fd, uint64_t subvol_id, bool insert_intermediary_qgroup) { | |
1880 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1881 | uint64_t parent_subvol; | |
1882 | bool changed = false; | |
1883 | int n = 0, r; | |
1884 | ||
1885 | assert(fd >= 0); | |
1886 | ||
1887 | /* | |
1888 | * Sets up the specified subvolume's qgroup automatically in | |
1889 | * one of two ways: | |
1890 | * | |
1891 | * If insert_intermediary_qgroup is false, the subvolume's | |
1892 | * leaf qgroup will be assigned to the same parent qgroups as | |
1893 | * the subvolume's parent subvolume. | |
1894 | * | |
1895 | * If insert_intermediary_qgroup is true a new intermediary | |
1896 | * higher-level qgroup is created, with a higher level number, | |
1897 | * but reusing the id of the subvolume. The level number is | |
1898 | * picked as one smaller than the lowest level qgroup the | |
1899 | * parent subvolume is a member of. If the parent subvolume's | |
1900 | * leaf qgroup is assigned to no higher-level qgroup a new | |
1901 | * qgroup of level 255 is created instead. Either way, the new | |
1902 | * qgroup is then assigned to the parent's higher-level | |
1903 | * qgroup, and the subvolume itself is assigned to it. | |
1904 | * | |
1905 | * If the subvolume is already assigned to a higher level | |
1906 | * qgroup, no operation is executed. | |
1907 | * | |
1908 | * Effectively this means: regardless if | |
1909 | * insert_intermediary_qgroup is true or not, after this | |
1910 | * function is invoked the subvolume will be accounted within | |
1911 | * the same qgroups as the parent. However, if it is true, it | |
1912 | * will also get its own higher-level qgroup, which may in | |
1913 | * turn be used by subvolumes created beneath this subvolume | |
1914 | * later on. | |
1915 | * | |
1916 | * This hence defines a simple default qgroup setup for | |
1917 | * subvolumes, as long as this function is invoked on each | |
1918 | * created subvolume: each subvolume is always accounting | |
1919 | * together with its immediate parents. Optionally, if | |
1920 | * insert_intermediary_qgroup is true, it will also get a | |
1921 | * qgroup that then includes all its own child subvolumes. | |
1922 | */ | |
1923 | ||
1924 | if (subvol_id == 0) { | |
2904e949 | 1925 | r = btrfs_is_subvol_fd(fd); |
5bcd08db LP |
1926 | if (r < 0) |
1927 | return r; | |
1928 | if (!r) | |
1929 | return -ENOTTY; | |
1930 | ||
1931 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
1932 | if (r < 0) | |
1933 | return r; | |
1934 | } | |
1935 | ||
1936 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
1937 | if (n < 0) | |
1938 | return n; | |
1939 | if (n > 0) /* already parent qgroups set up, let's bail */ | |
1940 | return 0; | |
1941 | ||
08c77cf3 LP |
1942 | qgroups = mfree(qgroups); |
1943 | ||
5bcd08db | 1944 | r = btrfs_subvol_get_parent(fd, subvol_id, &parent_subvol); |
08c77cf3 LP |
1945 | if (r == -ENXIO) |
1946 | /* No parent, hence no qgroup memberships */ | |
1947 | n = 0; | |
1948 | else if (r < 0) | |
5bcd08db | 1949 | return r; |
08c77cf3 LP |
1950 | else { |
1951 | n = btrfs_qgroup_find_parents(fd, parent_subvol, &qgroups); | |
1952 | if (n < 0) | |
1953 | return n; | |
1954 | } | |
5bcd08db LP |
1955 | |
1956 | if (insert_intermediary_qgroup) { | |
1957 | uint64_t lowest = 256, new_qgroupid; | |
1958 | bool created = false; | |
1959 | int i; | |
1960 | ||
1961 | /* Determine the lowest qgroup that the parent | |
1962 | * subvolume is assigned to. */ | |
1963 | ||
1964 | for (i = 0; i < n; i++) { | |
1965 | uint64_t level; | |
1966 | ||
1967 | r = btrfs_qgroupid_split(qgroups[i], &level, NULL); | |
1968 | if (r < 0) | |
1969 | return r; | |
1970 | ||
1971 | if (level < lowest) | |
1972 | lowest = level; | |
1973 | } | |
1974 | ||
1975 | if (lowest <= 1) /* There are no levels left we could use insert an intermediary qgroup at */ | |
1976 | return -EBUSY; | |
1977 | ||
1978 | r = btrfs_qgroupid_make(lowest - 1, subvol_id, &new_qgroupid); | |
1979 | if (r < 0) | |
1980 | return r; | |
1981 | ||
1982 | /* Create the new intermediary group, unless it already exists */ | |
1983 | r = btrfs_qgroup_create(fd, new_qgroupid); | |
1984 | if (r < 0 && r != -EEXIST) | |
1985 | return r; | |
1986 | if (r >= 0) | |
1987 | changed = created = true; | |
1988 | ||
1989 | for (i = 0; i < n; i++) { | |
1990 | r = btrfs_qgroup_assign(fd, new_qgroupid, qgroups[i]); | |
1991 | if (r < 0 && r != -EEXIST) { | |
1992 | if (created) | |
1993 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
1994 | ||
1995 | return r; | |
1996 | } | |
1997 | if (r >= 0) | |
1998 | changed = true; | |
1999 | } | |
2000 | ||
2001 | r = btrfs_qgroup_assign(fd, subvol_id, new_qgroupid); | |
2002 | if (r < 0 && r != -EEXIST) { | |
2003 | if (created) | |
2004 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
2005 | return r; | |
2006 | } | |
2007 | if (r >= 0) | |
2008 | changed = true; | |
2009 | ||
2010 | } else { | |
2011 | int i; | |
2012 | ||
2013 | /* Assign our subvolume to all the same qgroups as the parent */ | |
2014 | ||
2015 | for (i = 0; i < n; i++) { | |
2016 | r = btrfs_qgroup_assign(fd, subvol_id, qgroups[i]); | |
2017 | if (r < 0 && r != -EEXIST) | |
2018 | return r; | |
2019 | if (r >= 0) | |
2020 | changed = true; | |
2021 | } | |
2022 | } | |
2023 | ||
2024 | return changed; | |
2025 | } | |
2026 | ||
2027 | int btrfs_subvol_auto_qgroup(const char *path, uint64_t subvol_id, bool create_intermediary_qgroup) { | |
2028 | _cleanup_close_ int fd = -1; | |
2029 | ||
2030 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
2031 | if (fd < 0) | |
2032 | return -errno; | |
2033 | ||
2034 | return btrfs_subvol_auto_qgroup_fd(fd, subvol_id, create_intermediary_qgroup); | |
2035 | } | |
2036 | ||
2037 | int btrfs_subvol_get_parent(int fd, uint64_t subvol_id, uint64_t *ret) { | |
2038 | ||
2039 | struct btrfs_ioctl_search_args args = { | |
2040 | /* Tree of tree roots */ | |
2041 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
2042 | ||
2043 | /* Look precisely for the subvolume items */ | |
2044 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
2045 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
2046 | ||
2047 | /* No restrictions on the other components */ | |
2048 | .key.min_offset = 0, | |
2049 | .key.max_offset = (uint64_t) -1, | |
2050 | ||
2051 | .key.min_transid = 0, | |
2052 | .key.max_transid = (uint64_t) -1, | |
2053 | }; | |
2054 | int r; | |
2055 | ||
2056 | assert(fd >= 0); | |
2057 | assert(ret); | |
2058 | ||
2059 | if (subvol_id == 0) { | |
2060 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
2061 | if (r < 0) | |
2062 | return r; | |
2063 | } else { | |
2064 | r = btrfs_is_filesystem(fd); | |
2065 | if (r < 0) | |
2066 | return r; | |
2067 | if (!r) | |
2068 | return -ENOTTY; | |
2069 | } | |
2070 | ||
2071 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
2072 | ||
2073 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
2074 | const struct btrfs_ioctl_search_header *sh; | |
2075 | unsigned i; | |
2076 | ||
2077 | args.key.nr_items = 256; | |
2078 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
9c4615fb | 2079 | return negative_errno(); |
5bcd08db LP |
2080 | |
2081 | if (args.key.nr_items <= 0) | |
2082 | break; | |
2083 | ||
2084 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
2085 | ||
2086 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
2087 | continue; | |
2088 | if (sh->objectid != subvol_id) | |
2089 | continue; | |
2090 | ||
2091 | *ret = sh->offset; | |
2092 | return 0; | |
2093 | } | |
2094 | } | |
2095 | ||
2096 | return -ENXIO; | |
2097 | } |