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