]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
d7c7c334 | 2 | |
11c3a366 TA |
3 | #include <errno.h> |
4 | #include <fcntl.h> | |
5 | #include <inttypes.h> | |
17cbb288 | 6 | #include <linux/fs.h> |
11c3a366 TA |
7 | #include <linux/loop.h> |
8 | #include <stddef.h> | |
9 | #include <stdio.h> | |
d7c7c334 | 10 | #include <stdlib.h> |
11c3a366 TA |
11 | #include <string.h> |
12 | #include <sys/ioctl.h> | |
d7c7c334 | 13 | #include <sys/stat.h> |
11c3a366 TA |
14 | #include <sys/statfs.h> |
15 | #include <sys/sysmacros.h> | |
16 | #include <unistd.h> | |
17 | ||
349cc4a5 | 18 | #if HAVE_LINUX_BTRFS_H |
d7c7c334 LP |
19 | #include <linux/btrfs.h> |
20 | #endif | |
21 | ||
b5efdb8a | 22 | #include "alloc-util.h" |
18c528e9 | 23 | #include "blockdev-util.h" |
07630cea | 24 | #include "btrfs-ctree.h" |
3ffd4af2 | 25 | #include "btrfs-util.h" |
17cbb288 | 26 | #include "chattr-util.h" |
07630cea | 27 | #include "copy.h" |
553e15f2 | 28 | #include "device-nodes.h" |
3ffd4af2 | 29 | #include "fd-util.h" |
07630cea | 30 | #include "fileio.h" |
ef8becfa | 31 | #include "fs-util.h" |
a90fb858 | 32 | #include "io-util.h" |
07630cea | 33 | #include "macro.h" |
d7c7c334 | 34 | #include "missing.h" |
d7c7c334 | 35 | #include "path-util.h" |
17cbb288 | 36 | #include "rm-rf.h" |
d7b8eec7 | 37 | #include "smack-util.h" |
11c3a366 | 38 | #include "sparse-endian.h" |
872a590e | 39 | #include "stat-util.h" |
07630cea | 40 | #include "string-util.h" |
93cc7779 | 41 | #include "time-util.h" |
07630cea | 42 | #include "util.h" |
d7c7c334 | 43 | |
62572894 LP |
44 | /* WARNING: Be careful with file system ioctls! When we get an fd, we |
45 | * need to make sure it either refers to only a regular file or | |
46 | * directory, or that it is located on btrfs, before invoking any | |
47 | * btrfs ioctls. The ioctl numbers are reused by some device drivers | |
48 | * (such as DRM), and hence might have bad effects when invoked on | |
49 | * device nodes (that reference drivers) rather than fds to normal | |
50 | * files or directories. */ | |
51 | ||
d7c7c334 LP |
52 | static int validate_subvolume_name(const char *name) { |
53 | ||
54 | if (!filename_is_valid(name)) | |
55 | return -EINVAL; | |
56 | ||
57 | if (strlen(name) > BTRFS_SUBVOL_NAME_MAX) | |
58 | return -E2BIG; | |
59 | ||
60 | return 0; | |
61 | } | |
62 | ||
d7c7c334 LP |
63 | static int extract_subvolume_name(const char *path, const char **subvolume) { |
64 | const char *fn; | |
65 | int r; | |
66 | ||
67 | assert(path); | |
68 | assert(subvolume); | |
69 | ||
70 | fn = basename(path); | |
71 | ||
72 | r = validate_subvolume_name(fn); | |
73 | if (r < 0) | |
74 | return r; | |
75 | ||
76 | *subvolume = fn; | |
77 | return 0; | |
78 | } | |
79 | ||
21222ea5 | 80 | int btrfs_is_filesystem(int fd) { |
d7c7c334 LP |
81 | struct statfs sfs; |
82 | ||
21222ea5 LP |
83 | assert(fd >= 0); |
84 | ||
85 | if (fstatfs(fd, &sfs) < 0) | |
86 | return -errno; | |
87 | ||
88 | return F_TYPE_EQUAL(sfs.f_type, BTRFS_SUPER_MAGIC); | |
89 | } | |
90 | ||
2904e949 | 91 | int btrfs_is_subvol_fd(int fd) { |
21222ea5 LP |
92 | struct stat st; |
93 | ||
94 | assert(fd >= 0); | |
95 | ||
cd61c3bf LP |
96 | /* On btrfs subvolumes always have the inode 256 */ |
97 | ||
98 | if (fstat(fd, &st) < 0) | |
d7c7c334 LP |
99 | return -errno; |
100 | ||
cd61c3bf | 101 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) |
d7c7c334 LP |
102 | return 0; |
103 | ||
21222ea5 | 104 | return btrfs_is_filesystem(fd); |
d7c7c334 LP |
105 | } |
106 | ||
2904e949 LP |
107 | int btrfs_is_subvol(const char *path) { |
108 | _cleanup_close_ int fd = -1; | |
109 | ||
110 | assert(path); | |
111 | ||
112 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
113 | if (fd < 0) | |
114 | return -errno; | |
115 | ||
116 | return btrfs_is_subvol_fd(fd); | |
117 | } | |
118 | ||
d7c7c334 LP |
119 | int btrfs_subvol_make(const char *path) { |
120 | struct btrfs_ioctl_vol_args args = {}; | |
121 | _cleanup_close_ int fd = -1; | |
122 | const char *subvolume; | |
123 | int r; | |
124 | ||
125 | assert(path); | |
126 | ||
127 | r = extract_subvolume_name(path, &subvolume); | |
128 | if (r < 0) | |
129 | return r; | |
130 | ||
ef8becfa | 131 | fd = open_parent(path, O_CLOEXEC, 0); |
d7c7c334 LP |
132 | if (fd < 0) |
133 | return fd; | |
134 | ||
135 | strncpy(args.name, subvolume, sizeof(args.name)-1); | |
136 | ||
137 | if (ioctl(fd, BTRFS_IOC_SUBVOL_CREATE, &args) < 0) | |
138 | return -errno; | |
139 | ||
140 | return 0; | |
141 | } | |
142 | ||
0d6e763b | 143 | int btrfs_subvol_set_read_only_fd(int fd, bool b) { |
d7c7c334 | 144 | uint64_t flags, nflags; |
0d6e763b | 145 | struct stat st; |
d7c7c334 | 146 | |
0d6e763b LP |
147 | assert(fd >= 0); |
148 | ||
149 | if (fstat(fd, &st) < 0) | |
d7c7c334 LP |
150 | return -errno; |
151 | ||
0d6e763b LP |
152 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) |
153 | return -EINVAL; | |
154 | ||
d7c7c334 LP |
155 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) |
156 | return -errno; | |
157 | ||
158 | if (b) | |
159 | nflags = flags | BTRFS_SUBVOL_RDONLY; | |
160 | else | |
161 | nflags = flags & ~BTRFS_SUBVOL_RDONLY; | |
162 | ||
163 | if (flags == nflags) | |
164 | return 0; | |
165 | ||
166 | if (ioctl(fd, BTRFS_IOC_SUBVOL_SETFLAGS, &nflags) < 0) | |
167 | return -errno; | |
168 | ||
169 | return 0; | |
170 | } | |
171 | ||
0d6e763b LP |
172 | int btrfs_subvol_set_read_only(const char *path, bool b) { |
173 | _cleanup_close_ int fd = -1; | |
174 | ||
175 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
176 | if (fd < 0) | |
177 | return -errno; | |
178 | ||
179 | return btrfs_subvol_set_read_only_fd(fd, b); | |
180 | } | |
181 | ||
10f9c755 | 182 | int btrfs_subvol_get_read_only_fd(int fd) { |
cd61c3bf | 183 | uint64_t flags; |
62572894 LP |
184 | struct stat st; |
185 | ||
186 | assert(fd >= 0); | |
187 | ||
188 | if (fstat(fd, &st) < 0) | |
189 | return -errno; | |
190 | ||
191 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) | |
192 | return -EINVAL; | |
cd61c3bf LP |
193 | |
194 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) | |
195 | return -errno; | |
196 | ||
197 | return !!(flags & BTRFS_SUBVOL_RDONLY); | |
198 | } | |
199 | ||
d7c7c334 LP |
200 | int btrfs_reflink(int infd, int outfd) { |
201 | int r; | |
202 | ||
203 | assert(infd >= 0); | |
204 | assert(outfd >= 0); | |
205 | ||
3cc44114 | 206 | /* Make sure we invoke the ioctl on a regular file, so that no device driver accidentally gets it. */ |
62572894 | 207 | |
3cc44114 | 208 | r = fd_verify_regular(outfd); |
d7c7c334 | 209 | if (r < 0) |
3cc44114 LP |
210 | return r; |
211 | ||
212 | if (ioctl(outfd, BTRFS_IOC_CLONE, infd) < 0) | |
d7c7c334 LP |
213 | return -errno; |
214 | ||
215 | return 0; | |
216 | } | |
217 | ||
1c7dd825 LP |
218 | int btrfs_clone_range(int infd, uint64_t in_offset, int outfd, uint64_t out_offset, uint64_t sz) { |
219 | struct btrfs_ioctl_clone_range_args args = { | |
220 | .src_fd = infd, | |
221 | .src_offset = in_offset, | |
222 | .src_length = sz, | |
223 | .dest_offset = out_offset, | |
224 | }; | |
225 | int r; | |
226 | ||
227 | assert(infd >= 0); | |
228 | assert(outfd >= 0); | |
229 | assert(sz > 0); | |
230 | ||
3cc44114 | 231 | r = fd_verify_regular(outfd); |
1c7dd825 | 232 | if (r < 0) |
3cc44114 LP |
233 | return r; |
234 | ||
235 | if (ioctl(outfd, BTRFS_IOC_CLONE_RANGE, &args) < 0) | |
1c7dd825 LP |
236 | return -errno; |
237 | ||
238 | return 0; | |
239 | } | |
240 | ||
efe02862 | 241 | int btrfs_get_block_device_fd(int fd, dev_t *dev) { |
d7c7c334 | 242 | struct btrfs_ioctl_fs_info_args fsi = {}; |
d7c7c334 | 243 | uint64_t id; |
62572894 | 244 | int r; |
d7c7c334 | 245 | |
efe02862 | 246 | assert(fd >= 0); |
d7c7c334 LP |
247 | assert(dev); |
248 | ||
62572894 LP |
249 | r = btrfs_is_filesystem(fd); |
250 | if (r < 0) | |
251 | return r; | |
252 | if (!r) | |
253 | return -ENOTTY; | |
254 | ||
d7c7c334 LP |
255 | if (ioctl(fd, BTRFS_IOC_FS_INFO, &fsi) < 0) |
256 | return -errno; | |
257 | ||
258 | /* We won't do this for btrfs RAID */ | |
66ae5130 LP |
259 | if (fsi.num_devices != 1) { |
260 | *dev = 0; | |
d7c7c334 | 261 | return 0; |
66ae5130 | 262 | } |
d7c7c334 LP |
263 | |
264 | for (id = 1; id <= fsi.max_id; id++) { | |
265 | struct btrfs_ioctl_dev_info_args di = { | |
266 | .devid = id, | |
267 | }; | |
268 | struct stat st; | |
269 | ||
270 | if (ioctl(fd, BTRFS_IOC_DEV_INFO, &di) < 0) { | |
271 | if (errno == ENODEV) | |
272 | continue; | |
273 | ||
274 | return -errno; | |
275 | } | |
276 | ||
277 | if (stat((char*) di.path, &st) < 0) | |
278 | return -errno; | |
279 | ||
280 | if (!S_ISBLK(st.st_mode)) | |
281 | return -ENODEV; | |
282 | ||
283 | if (major(st.st_rdev) == 0) | |
284 | return -ENODEV; | |
285 | ||
286 | *dev = st.st_rdev; | |
287 | return 1; | |
288 | } | |
289 | ||
290 | return -ENODEV; | |
291 | } | |
10f9c755 | 292 | |
efe02862 LP |
293 | int btrfs_get_block_device(const char *path, dev_t *dev) { |
294 | _cleanup_close_ int fd = -1; | |
295 | ||
296 | assert(path); | |
297 | assert(dev); | |
298 | ||
299 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
300 | if (fd < 0) | |
301 | return -errno; | |
302 | ||
303 | return btrfs_get_block_device_fd(fd, dev); | |
304 | } | |
305 | ||
10f9c755 LP |
306 | int btrfs_subvol_get_id_fd(int fd, uint64_t *ret) { |
307 | struct btrfs_ioctl_ino_lookup_args args = { | |
308 | .objectid = BTRFS_FIRST_FREE_OBJECTID | |
309 | }; | |
62572894 | 310 | int r; |
10f9c755 LP |
311 | |
312 | assert(fd >= 0); | |
313 | assert(ret); | |
314 | ||
62572894 LP |
315 | r = btrfs_is_filesystem(fd); |
316 | if (r < 0) | |
317 | return r; | |
318 | if (!r) | |
319 | return -ENOTTY; | |
320 | ||
10f9c755 LP |
321 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args) < 0) |
322 | return -errno; | |
323 | ||
324 | *ret = args.treeid; | |
325 | return 0; | |
326 | } | |
327 | ||
90578cbd LP |
328 | int btrfs_subvol_get_id(int fd, const char *subvol, uint64_t *ret) { |
329 | _cleanup_close_ int subvol_fd = -1; | |
330 | ||
331 | assert(fd >= 0); | |
332 | assert(ret); | |
333 | ||
334 | subvol_fd = openat(fd, subvol, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
335 | if (subvol_fd < 0) | |
336 | return -errno; | |
337 | ||
338 | return btrfs_subvol_get_id_fd(subvol_fd, ret); | |
339 | } | |
340 | ||
5743a585 LP |
341 | static bool btrfs_ioctl_search_args_inc(struct btrfs_ioctl_search_args *args) { |
342 | assert(args); | |
343 | ||
344 | /* the objectid, type, offset together make up the btrfs key, | |
345 | * which is considered a single 136byte integer when | |
346 | * comparing. This call increases the counter by one, dealing | |
347 | * with the overflow between the overflows */ | |
348 | ||
349 | if (args->key.min_offset < (uint64_t) -1) { | |
350 | args->key.min_offset++; | |
351 | return true; | |
352 | } | |
353 | ||
354 | if (args->key.min_type < (uint8_t) -1) { | |
355 | args->key.min_type++; | |
356 | args->key.min_offset = 0; | |
357 | return true; | |
358 | } | |
359 | ||
360 | if (args->key.min_objectid < (uint64_t) -1) { | |
361 | args->key.min_objectid++; | |
362 | args->key.min_offset = 0; | |
363 | args->key.min_type = 0; | |
364 | return true; | |
365 | } | |
366 | ||
367 | return 0; | |
368 | } | |
369 | ||
370 | static void btrfs_ioctl_search_args_set(struct btrfs_ioctl_search_args *args, const struct btrfs_ioctl_search_header *h) { | |
371 | assert(args); | |
372 | assert(h); | |
373 | ||
374 | args->key.min_objectid = h->objectid; | |
375 | args->key.min_type = h->type; | |
376 | args->key.min_offset = h->offset; | |
377 | } | |
378 | ||
379 | static int btrfs_ioctl_search_args_compare(const struct btrfs_ioctl_search_args *args) { | |
380 | assert(args); | |
381 | ||
382 | /* Compare min and max */ | |
383 | ||
384 | if (args->key.min_objectid < args->key.max_objectid) | |
385 | return -1; | |
386 | if (args->key.min_objectid > args->key.max_objectid) | |
387 | return 1; | |
388 | ||
389 | if (args->key.min_type < args->key.max_type) | |
390 | return -1; | |
391 | if (args->key.min_type > args->key.max_type) | |
392 | return 1; | |
393 | ||
394 | if (args->key.min_offset < args->key.max_offset) | |
395 | return -1; | |
396 | if (args->key.min_offset > args->key.max_offset) | |
397 | return 1; | |
398 | ||
399 | return 0; | |
400 | } | |
401 | ||
402 | #define FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) \ | |
403 | for ((i) = 0, \ | |
404 | (sh) = (const struct btrfs_ioctl_search_header*) (args).buf; \ | |
405 | (i) < (args).key.nr_items; \ | |
406 | (i)++, \ | |
407 | (sh) = (const struct btrfs_ioctl_search_header*) ((uint8_t*) (sh) + sizeof(struct btrfs_ioctl_search_header) + (sh)->len)) | |
408 | ||
409 | #define BTRFS_IOCTL_SEARCH_HEADER_BODY(sh) \ | |
410 | ((void*) ((uint8_t*) sh + sizeof(struct btrfs_ioctl_search_header))) | |
411 | ||
5bcd08db | 412 | int btrfs_subvol_get_info_fd(int fd, uint64_t subvol_id, BtrfsSubvolInfo *ret) { |
10f9c755 LP |
413 | struct btrfs_ioctl_search_args args = { |
414 | /* Tree of tree roots */ | |
b6b18498 | 415 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, |
10f9c755 LP |
416 | |
417 | /* Look precisely for the subvolume items */ | |
418 | .key.min_type = BTRFS_ROOT_ITEM_KEY, | |
419 | .key.max_type = BTRFS_ROOT_ITEM_KEY, | |
420 | ||
10f9c755 LP |
421 | .key.min_offset = 0, |
422 | .key.max_offset = (uint64_t) -1, | |
5743a585 LP |
423 | |
424 | /* No restrictions on the other components */ | |
10f9c755 LP |
425 | .key.min_transid = 0, |
426 | .key.max_transid = (uint64_t) -1, | |
10f9c755 LP |
427 | }; |
428 | ||
b6b18498 | 429 | bool found = false; |
10f9c755 LP |
430 | int r; |
431 | ||
432 | assert(fd >= 0); | |
433 | assert(ret); | |
434 | ||
5bcd08db LP |
435 | if (subvol_id == 0) { |
436 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
437 | if (r < 0) | |
438 | return r; | |
439 | } else { | |
440 | r = btrfs_is_filesystem(fd); | |
441 | if (r < 0) | |
442 | return r; | |
443 | if (!r) | |
444 | return -ENOTTY; | |
445 | } | |
10f9c755 LP |
446 | |
447 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
10f9c755 | 448 | |
5743a585 | 449 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
450 | const struct btrfs_ioctl_search_header *sh; |
451 | unsigned i; | |
452 | ||
453 | args.key.nr_items = 256; | |
454 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
455 | return -errno; | |
456 | ||
457 | if (args.key.nr_items <= 0) | |
458 | break; | |
10f9c755 | 459 | |
5743a585 | 460 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
10f9c755 | 461 | |
b6b18498 LP |
462 | const struct btrfs_root_item *ri; |
463 | ||
5743a585 LP |
464 | /* Make sure we start the next search at least from this entry */ |
465 | btrfs_ioctl_search_args_set(&args, sh); | |
466 | ||
b6b18498 LP |
467 | if (sh->objectid != subvol_id) |
468 | continue; | |
469 | if (sh->type != BTRFS_ROOT_ITEM_KEY) | |
470 | continue; | |
5743a585 LP |
471 | |
472 | /* Older versions of the struct lacked the otime setting */ | |
b6b18498 LP |
473 | if (sh->len < offsetof(struct btrfs_root_item, otime) + sizeof(struct btrfs_timespec)) |
474 | continue; | |
10f9c755 | 475 | |
5743a585 | 476 | ri = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
10f9c755 | 477 | |
b6b18498 LP |
478 | ret->otime = (usec_t) le64toh(ri->otime.sec) * USEC_PER_SEC + |
479 | (usec_t) le32toh(ri->otime.nsec) / NSEC_PER_USEC; | |
10f9c755 | 480 | |
b6b18498 | 481 | ret->subvol_id = subvol_id; |
5d904a6a | 482 | ret->read_only = le64toh(ri->flags) & BTRFS_ROOT_SUBVOL_RDONLY; |
10f9c755 | 483 | |
b6b18498 LP |
484 | assert_cc(sizeof(ri->uuid) == sizeof(ret->uuid)); |
485 | memcpy(&ret->uuid, ri->uuid, sizeof(ret->uuid)); | |
486 | memcpy(&ret->parent_uuid, ri->parent_uuid, sizeof(ret->parent_uuid)); | |
487 | ||
488 | found = true; | |
489 | goto finish; | |
490 | } | |
491 | ||
5743a585 LP |
492 | /* Increase search key by one, to read the next item, if we can. */ |
493 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
494 | break; |
495 | } | |
496 | ||
497 | finish: | |
498 | if (!found) | |
499 | return -ENODATA; | |
500 | ||
501 | return 0; | |
502 | } | |
503 | ||
5bcd08db | 504 | int btrfs_qgroup_get_quota_fd(int fd, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
b6b18498 LP |
505 | |
506 | struct btrfs_ioctl_search_args args = { | |
507 | /* Tree of quota items */ | |
508 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
509 | ||
5743a585 LP |
510 | /* The object ID is always 0 */ |
511 | .key.min_objectid = 0, | |
512 | .key.max_objectid = 0, | |
513 | ||
b6b18498 LP |
514 | /* Look precisely for the quota items */ |
515 | .key.min_type = BTRFS_QGROUP_STATUS_KEY, | |
516 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
517 | ||
b6b18498 LP |
518 | /* No restrictions on the other components */ |
519 | .key.min_transid = 0, | |
520 | .key.max_transid = (uint64_t) -1, | |
521 | }; | |
522 | ||
b6b18498 LP |
523 | bool found_info = false, found_limit = false; |
524 | int r; | |
525 | ||
526 | assert(fd >= 0); | |
527 | assert(ret); | |
528 | ||
5bcd08db LP |
529 | if (qgroupid == 0) { |
530 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
531 | if (r < 0) | |
532 | return r; | |
533 | } else { | |
534 | r = btrfs_is_filesystem(fd); | |
535 | if (r < 0) | |
536 | return r; | |
537 | if (!r) | |
538 | return -ENOTTY; | |
539 | } | |
b6b18498 | 540 | |
5bcd08db | 541 | args.key.min_offset = args.key.max_offset = qgroupid; |
b6b18498 | 542 | |
5743a585 | 543 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
544 | const struct btrfs_ioctl_search_header *sh; |
545 | unsigned i; | |
546 | ||
547 | args.key.nr_items = 256; | |
12ee6186 LP |
548 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
549 | if (errno == ENOENT) /* quota tree is missing: quota disabled */ | |
550 | break; | |
551 | ||
b6b18498 | 552 | return -errno; |
12ee6186 | 553 | } |
b6b18498 LP |
554 | |
555 | if (args.key.nr_items <= 0) | |
556 | break; | |
557 | ||
5743a585 | 558 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
b6b18498 | 559 | |
5743a585 LP |
560 | /* Make sure we start the next search at least from this entry */ |
561 | btrfs_ioctl_search_args_set(&args, sh); | |
b6b18498 LP |
562 | |
563 | if (sh->objectid != 0) | |
564 | continue; | |
5bcd08db | 565 | if (sh->offset != qgroupid) |
b6b18498 LP |
566 | continue; |
567 | ||
b6b18498 | 568 | if (sh->type == BTRFS_QGROUP_INFO_KEY) { |
5743a585 | 569 | const struct btrfs_qgroup_info_item *qii = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 570 | |
cb81cd80 | 571 | ret->referenced = le64toh(qii->rfer); |
b6b18498 LP |
572 | ret->exclusive = le64toh(qii->excl); |
573 | ||
574 | found_info = true; | |
575 | ||
576 | } else if (sh->type == BTRFS_QGROUP_LIMIT_KEY) { | |
5743a585 | 577 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 578 | |
5bcd08db LP |
579 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_RFER) |
580 | ret->referenced_max = le64toh(qli->max_rfer); | |
581 | else | |
cb81cd80 | 582 | ret->referenced_max = (uint64_t) -1; |
5bcd08db LP |
583 | |
584 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_EXCL) | |
585 | ret->exclusive_max = le64toh(qli->max_excl); | |
586 | else | |
b6b18498 LP |
587 | ret->exclusive_max = (uint64_t) -1; |
588 | ||
589 | found_limit = true; | |
590 | } | |
591 | ||
592 | if (found_info && found_limit) | |
593 | goto finish; | |
594 | } | |
595 | ||
5743a585 LP |
596 | /* Increase search key by one, to read the next item, if we can. */ |
597 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
598 | break; |
599 | } | |
600 | ||
601 | finish: | |
602 | if (!found_limit && !found_info) | |
603 | return -ENODATA; | |
604 | ||
605 | if (!found_info) { | |
cb81cd80 | 606 | ret->referenced = (uint64_t) -1; |
b6b18498 LP |
607 | ret->exclusive = (uint64_t) -1; |
608 | } | |
609 | ||
610 | if (!found_limit) { | |
cb81cd80 | 611 | ret->referenced_max = (uint64_t) -1; |
b6b18498 LP |
612 | ret->exclusive_max = (uint64_t) -1; |
613 | } | |
10f9c755 LP |
614 | |
615 | return 0; | |
616 | } | |
f27a3864 | 617 | |
5bcd08db LP |
618 | int btrfs_qgroup_get_quota(const char *path, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
619 | _cleanup_close_ int fd = -1; | |
620 | ||
621 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
622 | if (fd < 0) | |
623 | return -errno; | |
624 | ||
625 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
626 | } | |
627 | ||
628 | int btrfs_subvol_find_subtree_qgroup(int fd, uint64_t subvol_id, uint64_t *ret) { | |
629 | uint64_t level, lowest = (uint64_t) -1, lowest_qgroupid = 0; | |
630 | _cleanup_free_ uint64_t *qgroups = NULL; | |
631 | int r, n, i; | |
632 | ||
633 | assert(fd >= 0); | |
634 | assert(ret); | |
635 | ||
636 | /* This finds the "subtree" qgroup for a specific | |
637 | * subvolume. This only works for subvolumes that have been | |
638 | * prepared with btrfs_subvol_auto_qgroup_fd() with | |
639 | * insert_intermediary_qgroup=true (or equivalent). For others | |
640 | * it will return the leaf qgroup instead. The two cases may | |
641 | * be distuingished via the return value, which is 1 in case | |
642 | * an appropriate "subtree" qgroup was found, and 0 | |
643 | * otherwise. */ | |
644 | ||
645 | if (subvol_id == 0) { | |
646 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
647 | if (r < 0) | |
648 | return r; | |
649 | } | |
650 | ||
651 | r = btrfs_qgroupid_split(subvol_id, &level, NULL); | |
652 | if (r < 0) | |
653 | return r; | |
654 | if (level != 0) /* Input must be a leaf qgroup */ | |
655 | return -EINVAL; | |
656 | ||
657 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
658 | if (n < 0) | |
659 | return n; | |
660 | ||
661 | for (i = 0; i < n; i++) { | |
662 | uint64_t id; | |
663 | ||
664 | r = btrfs_qgroupid_split(qgroups[i], &level, &id); | |
665 | if (r < 0) | |
666 | return r; | |
667 | ||
668 | if (id != subvol_id) | |
669 | continue; | |
670 | ||
671 | if (lowest == (uint64_t) -1 || level < lowest) { | |
672 | lowest_qgroupid = qgroups[i]; | |
673 | lowest = level; | |
674 | } | |
675 | } | |
676 | ||
677 | if (lowest == (uint64_t) -1) { | |
678 | /* No suitable higher-level qgroup found, let's return | |
679 | * the leaf qgroup instead, and indicate that with the | |
680 | * return value. */ | |
681 | ||
682 | *ret = subvol_id; | |
683 | return 0; | |
684 | } | |
685 | ||
686 | *ret = lowest_qgroupid; | |
687 | return 1; | |
688 | } | |
689 | ||
690 | int btrfs_subvol_get_subtree_quota_fd(int fd, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
691 | uint64_t qgroupid; | |
692 | int r; | |
693 | ||
694 | assert(fd >= 0); | |
695 | assert(ret); | |
696 | ||
697 | /* This determines the quota data of the qgroup with the | |
698 | * lowest level, that shares the id part with the specified | |
699 | * subvolume. This is useful for determining the quota data | |
700 | * for entire subvolume subtrees, as long as the subtrees have | |
701 | * been set up with btrfs_qgroup_subvol_auto_fd() or in a | |
702 | * compatible way */ | |
703 | ||
704 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
705 | if (r < 0) | |
706 | return r; | |
707 | ||
708 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
709 | } | |
710 | ||
711 | int btrfs_subvol_get_subtree_quota(const char *path, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
712 | _cleanup_close_ int fd = -1; | |
713 | ||
714 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
715 | if (fd < 0) | |
716 | return -errno; | |
717 | ||
718 | return btrfs_subvol_get_subtree_quota_fd(fd, subvol_id, ret); | |
719 | } | |
720 | ||
f27a3864 | 721 | int btrfs_defrag_fd(int fd) { |
3cc44114 | 722 | int r; |
62572894 | 723 | |
f27a3864 LP |
724 | assert(fd >= 0); |
725 | ||
3cc44114 LP |
726 | r = fd_verify_regular(fd); |
727 | if (r < 0) | |
728 | return r; | |
62572894 | 729 | |
f27a3864 LP |
730 | if (ioctl(fd, BTRFS_IOC_DEFRAG, NULL) < 0) |
731 | return -errno; | |
732 | ||
733 | return 0; | |
734 | } | |
735 | ||
736 | int btrfs_defrag(const char *p) { | |
737 | _cleanup_close_ int fd = -1; | |
738 | ||
739 | fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
740 | if (fd < 0) | |
741 | return -errno; | |
742 | ||
743 | return btrfs_defrag_fd(fd); | |
744 | } | |
754061ce LP |
745 | |
746 | int btrfs_quota_enable_fd(int fd, bool b) { | |
747 | struct btrfs_ioctl_quota_ctl_args args = { | |
748 | .cmd = b ? BTRFS_QUOTA_CTL_ENABLE : BTRFS_QUOTA_CTL_DISABLE, | |
749 | }; | |
62572894 | 750 | int r; |
754061ce LP |
751 | |
752 | assert(fd >= 0); | |
753 | ||
62572894 LP |
754 | r = btrfs_is_filesystem(fd); |
755 | if (r < 0) | |
756 | return r; | |
757 | if (!r) | |
758 | return -ENOTTY; | |
759 | ||
754061ce LP |
760 | if (ioctl(fd, BTRFS_IOC_QUOTA_CTL, &args) < 0) |
761 | return -errno; | |
762 | ||
763 | return 0; | |
764 | } | |
765 | ||
766 | int btrfs_quota_enable(const char *path, bool b) { | |
767 | _cleanup_close_ int fd = -1; | |
768 | ||
769 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
770 | if (fd < 0) | |
771 | return -errno; | |
772 | ||
773 | return btrfs_quota_enable_fd(fd, b); | |
774 | } | |
d6ce17c7 | 775 | |
5bcd08db LP |
776 | int btrfs_qgroup_set_limit_fd(int fd, uint64_t qgroupid, uint64_t referenced_max) { |
777 | ||
d6ce17c7 | 778 | struct btrfs_ioctl_qgroup_limit_args args = { |
5bcd08db | 779 | .lim.max_rfer = referenced_max, |
d6ce17c7 LP |
780 | .lim.flags = BTRFS_QGROUP_LIMIT_MAX_RFER, |
781 | }; | |
5bcd08db | 782 | unsigned c; |
62572894 | 783 | int r; |
d6ce17c7 LP |
784 | |
785 | assert(fd >= 0); | |
786 | ||
5bcd08db LP |
787 | if (qgroupid == 0) { |
788 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
789 | if (r < 0) | |
790 | return r; | |
791 | } else { | |
792 | r = btrfs_is_filesystem(fd); | |
793 | if (r < 0) | |
794 | return r; | |
795 | if (!r) | |
796 | return -ENOTTY; | |
797 | } | |
62572894 | 798 | |
5bcd08db LP |
799 | args.qgroupid = qgroupid; |
800 | ||
801 | for (c = 0;; c++) { | |
802 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &args) < 0) { | |
803 | ||
804 | if (errno == EBUSY && c < 10) { | |
805 | (void) btrfs_quota_scan_wait(fd); | |
806 | continue; | |
807 | } | |
808 | ||
809 | return -errno; | |
810 | } | |
811 | ||
812 | break; | |
813 | } | |
d6ce17c7 LP |
814 | |
815 | return 0; | |
816 | } | |
817 | ||
5bcd08db LP |
818 | int btrfs_qgroup_set_limit(const char *path, uint64_t qgroupid, uint64_t referenced_max) { |
819 | _cleanup_close_ int fd = -1; | |
820 | ||
821 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
822 | if (fd < 0) | |
823 | return -errno; | |
824 | ||
825 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
826 | } | |
827 | ||
828 | int btrfs_subvol_set_subtree_quota_limit_fd(int fd, uint64_t subvol_id, uint64_t referenced_max) { | |
829 | uint64_t qgroupid; | |
830 | int r; | |
831 | ||
832 | assert(fd >= 0); | |
833 | ||
834 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
835 | if (r < 0) | |
836 | return r; | |
837 | ||
838 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
839 | } | |
840 | ||
841 | int btrfs_subvol_set_subtree_quota_limit(const char *path, uint64_t subvol_id, uint64_t referenced_max) { | |
d6ce17c7 LP |
842 | _cleanup_close_ int fd = -1; |
843 | ||
844 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
845 | if (fd < 0) | |
846 | return -errno; | |
847 | ||
5bcd08db | 848 | return btrfs_subvol_set_subtree_quota_limit_fd(fd, subvol_id, referenced_max); |
d6ce17c7 | 849 | } |
efe02862 | 850 | |
26166c88 | 851 | int btrfs_resize_loopback_fd(int fd, uint64_t new_size, bool grow_only) { |
efe02862 | 852 | struct btrfs_ioctl_vol_args args = {}; |
553e15f2 ZJS |
853 | char p[SYS_BLOCK_PATH_MAX("/loop/backing_file")]; |
854 | _cleanup_free_ char *backing = NULL; | |
efe02862 LP |
855 | _cleanup_close_ int loop_fd = -1, backing_fd = -1; |
856 | struct stat st; | |
a7f7d1bd | 857 | dev_t dev = 0; |
efe02862 LP |
858 | int r; |
859 | ||
a90fb858 LP |
860 | /* In contrast to btrfs quota ioctls ftruncate() cannot make sense of "infinity" or file sizes > 2^31 */ |
861 | if (!FILE_SIZE_VALID(new_size)) | |
862 | return -EINVAL; | |
863 | ||
efe02862 LP |
864 | /* btrfs cannot handle file systems < 16M, hence use this as minimum */ |
865 | if (new_size < 16*1024*1024) | |
866 | new_size = 16*1024*1024; | |
867 | ||
868 | r = btrfs_get_block_device_fd(fd, &dev); | |
869 | if (r < 0) | |
870 | return r; | |
871 | if (r == 0) | |
872 | return -ENODEV; | |
873 | ||
553e15f2 | 874 | xsprintf_sys_block_path(p, "/loop/backing_file", dev); |
efe02862 | 875 | r = read_one_line_file(p, &backing); |
26166c88 LP |
876 | if (r == -ENOENT) |
877 | return -ENODEV; | |
efe02862 LP |
878 | if (r < 0) |
879 | return r; | |
880 | if (isempty(backing) || !path_is_absolute(backing)) | |
881 | return -ENODEV; | |
882 | ||
883 | backing_fd = open(backing, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
884 | if (backing_fd < 0) | |
885 | return -errno; | |
886 | ||
887 | if (fstat(backing_fd, &st) < 0) | |
888 | return -errno; | |
889 | if (!S_ISREG(st.st_mode)) | |
890 | return -ENODEV; | |
891 | ||
892 | if (new_size == (uint64_t) st.st_size) | |
893 | return 0; | |
894 | ||
26166c88 LP |
895 | if (grow_only && new_size < (uint64_t) st.st_size) |
896 | return -EINVAL; | |
897 | ||
553e15f2 ZJS |
898 | xsprintf_sys_block_path(p, NULL, dev); |
899 | loop_fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
efe02862 LP |
900 | if (loop_fd < 0) |
901 | return -errno; | |
902 | ||
903 | if (snprintf(args.name, sizeof(args.name), "%" PRIu64, new_size) >= (int) sizeof(args.name)) | |
904 | return -EINVAL; | |
905 | ||
906 | if (new_size < (uint64_t) st.st_size) { | |
907 | /* Decrease size: first decrease btrfs size, then shorten loopback */ | |
908 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
909 | return -errno; | |
910 | } | |
911 | ||
912 | if (ftruncate(backing_fd, new_size) < 0) | |
913 | return -errno; | |
914 | ||
915 | if (ioctl(loop_fd, LOOP_SET_CAPACITY, 0) < 0) | |
916 | return -errno; | |
917 | ||
918 | if (new_size > (uint64_t) st.st_size) { | |
919 | /* Increase size: first enlarge loopback, then increase btrfs size */ | |
920 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
921 | return -errno; | |
922 | } | |
923 | ||
26166c88 LP |
924 | /* Make sure the free disk space is correctly updated for both file systems */ |
925 | (void) fsync(fd); | |
926 | (void) fsync(backing_fd); | |
927 | ||
928 | return 1; | |
efe02862 LP |
929 | } |
930 | ||
26166c88 | 931 | int btrfs_resize_loopback(const char *p, uint64_t new_size, bool grow_only) { |
efe02862 LP |
932 | _cleanup_close_ int fd = -1; |
933 | ||
934 | fd = open(p, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
935 | if (fd < 0) | |
936 | return -errno; | |
937 | ||
26166c88 | 938 | return btrfs_resize_loopback_fd(fd, new_size, grow_only); |
efe02862 | 939 | } |
d9e2daaf | 940 | |
5bcd08db | 941 | int btrfs_qgroupid_make(uint64_t level, uint64_t id, uint64_t *ret) { |
3f952f92 LP |
942 | assert(ret); |
943 | ||
944 | if (level >= (UINT64_C(1) << (64 - BTRFS_QGROUP_LEVEL_SHIFT))) | |
945 | return -EINVAL; | |
946 | ||
947 | if (id >= (UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT)) | |
948 | return -EINVAL; | |
949 | ||
950 | *ret = (level << BTRFS_QGROUP_LEVEL_SHIFT) | id; | |
951 | return 0; | |
952 | } | |
953 | ||
5bcd08db LP |
954 | int btrfs_qgroupid_split(uint64_t qgroupid, uint64_t *level, uint64_t *id) { |
955 | assert(level || id); | |
956 | ||
957 | if (level) | |
958 | *level = qgroupid >> BTRFS_QGROUP_LEVEL_SHIFT; | |
959 | ||
960 | if (id) | |
961 | *id = qgroupid & ((UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT) - 1); | |
962 | ||
963 | return 0; | |
964 | } | |
965 | ||
966 | static int qgroup_create_or_destroy(int fd, bool b, uint64_t qgroupid) { | |
3f952f92 LP |
967 | |
968 | struct btrfs_ioctl_qgroup_create_args args = { | |
969 | .create = b, | |
5bcd08db | 970 | .qgroupid = qgroupid, |
3f952f92 | 971 | }; |
5bcd08db | 972 | unsigned c; |
3f952f92 LP |
973 | int r; |
974 | ||
5bcd08db | 975 | r = btrfs_is_filesystem(fd); |
3f952f92 LP |
976 | if (r < 0) |
977 | return r; | |
5bcd08db LP |
978 | if (r == 0) |
979 | return -ENOTTY; | |
980 | ||
981 | for (c = 0;; c++) { | |
982 | if (ioctl(fd, BTRFS_IOC_QGROUP_CREATE, &args) < 0) { | |
983 | ||
be6d467c LP |
984 | /* If quota is not enabled, we get EINVAL. Turn this into a recognizable error */ |
985 | if (errno == EINVAL) | |
986 | return -ENOPROTOOPT; | |
987 | ||
5bcd08db LP |
988 | if (errno == EBUSY && c < 10) { |
989 | (void) btrfs_quota_scan_wait(fd); | |
990 | continue; | |
991 | } | |
992 | ||
993 | return -errno; | |
994 | } | |
3f952f92 | 995 | |
5bcd08db LP |
996 | break; |
997 | } | |
998 | ||
999 | return 0; | |
1000 | } | |
1001 | ||
1002 | int btrfs_qgroup_create(int fd, uint64_t qgroupid) { | |
1003 | return qgroup_create_or_destroy(fd, true, qgroupid); | |
1004 | } | |
1005 | ||
1006 | int btrfs_qgroup_destroy(int fd, uint64_t qgroupid) { | |
1007 | return qgroup_create_or_destroy(fd, false, qgroupid); | |
1008 | } | |
1009 | ||
1010 | int btrfs_qgroup_destroy_recursive(int fd, uint64_t qgroupid) { | |
1011 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1012 | uint64_t subvol_id; | |
1013 | int i, n, r; | |
1014 | ||
1015 | /* Destroys the specified qgroup, but unassigns it from all | |
1016 | * its parents first. Also, it recursively destroys all | |
1017 | * qgroups it is assgined to that have the same id part of the | |
1018 | * qgroupid as the specified group. */ | |
1019 | ||
1020 | r = btrfs_qgroupid_split(qgroupid, NULL, &subvol_id); | |
1021 | if (r < 0) | |
1022 | return r; | |
1023 | ||
1024 | n = btrfs_qgroup_find_parents(fd, qgroupid, &qgroups); | |
1025 | if (n < 0) | |
1026 | return n; | |
1027 | ||
1028 | for (i = 0; i < n; i++) { | |
1029 | uint64_t id; | |
1030 | ||
1031 | r = btrfs_qgroupid_split(qgroups[i], NULL, &id); | |
1032 | if (r < 0) | |
1033 | return r; | |
1034 | ||
1035 | r = btrfs_qgroup_unassign(fd, qgroupid, qgroups[i]); | |
1036 | if (r < 0) | |
1037 | return r; | |
1038 | ||
1039 | if (id != subvol_id) | |
1040 | continue; | |
1041 | ||
1042 | /* The parent qgroupid shares the same id part with | |
1043 | * us? If so, destroy it too. */ | |
1044 | ||
1045 | (void) btrfs_qgroup_destroy_recursive(fd, qgroups[i]); | |
1046 | } | |
1047 | ||
1048 | return btrfs_qgroup_destroy(fd, qgroupid); | |
1049 | } | |
1050 | ||
1051 | int btrfs_quota_scan_start(int fd) { | |
1052 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1053 | ||
1054 | assert(fd >= 0); | |
1055 | ||
1056 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN, &args) < 0) | |
3f952f92 LP |
1057 | return -errno; |
1058 | ||
1059 | return 0; | |
1060 | } | |
1061 | ||
5bcd08db LP |
1062 | int btrfs_quota_scan_wait(int fd) { |
1063 | assert(fd >= 0); | |
1064 | ||
1065 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_WAIT) < 0) | |
1066 | return -errno; | |
1067 | ||
1068 | return 0; | |
3f952f92 LP |
1069 | } |
1070 | ||
5bcd08db LP |
1071 | int btrfs_quota_scan_ongoing(int fd) { |
1072 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1073 | ||
1074 | assert(fd >= 0); | |
1075 | ||
1076 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_STATUS, &args) < 0) | |
1077 | return -errno; | |
1078 | ||
1079 | return !!args.flags; | |
3f952f92 LP |
1080 | } |
1081 | ||
5bcd08db LP |
1082 | static int qgroup_assign_or_unassign(int fd, bool b, uint64_t child, uint64_t parent) { |
1083 | struct btrfs_ioctl_qgroup_assign_args args = { | |
1084 | .assign = b, | |
1085 | .src = child, | |
1086 | .dst = parent, | |
1087 | }; | |
1088 | unsigned c; | |
1089 | int r; | |
1090 | ||
1091 | r = btrfs_is_filesystem(fd); | |
1092 | if (r < 0) | |
1093 | return r; | |
1094 | if (r == 0) | |
1095 | return -ENOTTY; | |
1096 | ||
1097 | for (c = 0;; c++) { | |
1098 | r = ioctl(fd, BTRFS_IOC_QGROUP_ASSIGN, &args); | |
1099 | if (r < 0) { | |
1100 | if (errno == EBUSY && c < 10) { | |
1101 | (void) btrfs_quota_scan_wait(fd); | |
1102 | continue; | |
1103 | } | |
1104 | ||
1105 | return -errno; | |
1106 | } | |
1107 | ||
1108 | if (r == 0) | |
1109 | return 0; | |
1110 | ||
1111 | /* If the return value is > 0, we need to request a rescan */ | |
1112 | ||
1113 | (void) btrfs_quota_scan_start(fd); | |
1114 | return 1; | |
1115 | } | |
1116 | } | |
1117 | ||
1118 | int btrfs_qgroup_assign(int fd, uint64_t child, uint64_t parent) { | |
1119 | return qgroup_assign_or_unassign(fd, true, child, parent); | |
1120 | } | |
1121 | ||
1122 | int btrfs_qgroup_unassign(int fd, uint64_t child, uint64_t parent) { | |
1123 | return qgroup_assign_or_unassign(fd, false, child, parent); | |
1124 | } | |
1125 | ||
1126 | static int subvol_remove_children(int fd, const char *subvolume, uint64_t subvol_id, BtrfsRemoveFlags flags) { | |
d9e2daaf LP |
1127 | struct btrfs_ioctl_search_args args = { |
1128 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1129 | ||
1130 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1131 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1132 | ||
1133 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1134 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1135 | ||
1136 | .key.min_transid = 0, | |
1137 | .key.max_transid = (uint64_t) -1, | |
1138 | }; | |
1139 | ||
1140 | struct btrfs_ioctl_vol_args vol_args = {}; | |
1141 | _cleanup_close_ int subvol_fd = -1; | |
62572894 | 1142 | struct stat st; |
3986b258 | 1143 | bool made_writable = false; |
d9e2daaf LP |
1144 | int r; |
1145 | ||
1146 | assert(fd >= 0); | |
1147 | assert(subvolume); | |
1148 | ||
62572894 LP |
1149 | if (fstat(fd, &st) < 0) |
1150 | return -errno; | |
1151 | ||
1152 | if (!S_ISDIR(st.st_mode)) | |
1153 | return -EINVAL; | |
1154 | ||
f7c9f4a2 | 1155 | subvol_fd = openat(fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1156 | if (subvol_fd < 0) |
1157 | return -errno; | |
1158 | ||
1159 | if (subvol_id == 0) { | |
1160 | r = btrfs_subvol_get_id_fd(subvol_fd, &subvol_id); | |
1161 | if (r < 0) | |
1162 | return r; | |
1163 | } | |
1164 | ||
3f952f92 LP |
1165 | /* First, try to remove the subvolume. If it happens to be |
1166 | * already empty, this will just work. */ | |
1167 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
1168 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) >= 0) { | |
5bcd08db | 1169 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); /* for the leaf subvolumes, the qgroup id is identical to the subvol id */ |
3f952f92 LP |
1170 | return 0; |
1171 | } | |
5bcd08db | 1172 | if (!(flags & BTRFS_REMOVE_RECURSIVE) || errno != ENOTEMPTY) |
3f952f92 LP |
1173 | return -errno; |
1174 | ||
1175 | /* OK, the subvolume is not empty, let's look for child | |
1176 | * subvolumes, and remove them, first */ | |
1177 | ||
d9e2daaf LP |
1178 | args.key.min_offset = args.key.max_offset = subvol_id; |
1179 | ||
1180 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1181 | const struct btrfs_ioctl_search_header *sh; | |
1182 | unsigned i; | |
1183 | ||
1184 | args.key.nr_items = 256; | |
1185 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1186 | return -errno; | |
1187 | ||
1188 | if (args.key.nr_items <= 0) | |
1189 | break; | |
1190 | ||
1191 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1192 | _cleanup_free_ char *p = NULL; | |
1193 | const struct btrfs_root_ref *ref; | |
1194 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1195 | ||
1196 | btrfs_ioctl_search_args_set(&args, sh); | |
1197 | ||
1198 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1199 | continue; | |
1200 | if (sh->offset != subvol_id) | |
1201 | continue; | |
1202 | ||
1203 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1204 | ||
1205 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); | |
1206 | if (!p) | |
1207 | return -ENOMEM; | |
1208 | ||
1209 | zero(ino_args); | |
1210 | ino_args.treeid = subvol_id; | |
cbf21ecc | 1211 | ino_args.objectid = htole64(ref->dirid); |
d9e2daaf LP |
1212 | |
1213 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1214 | return -errno; | |
1215 | ||
3986b258 LP |
1216 | if (!made_writable) { |
1217 | r = btrfs_subvol_set_read_only_fd(subvol_fd, false); | |
1218 | if (r < 0) | |
1219 | return r; | |
1220 | ||
1221 | made_writable = true; | |
1222 | } | |
1223 | ||
d9e2daaf LP |
1224 | if (isempty(ino_args.name)) |
1225 | /* Subvolume is in the top-level | |
1226 | * directory of the subvolume. */ | |
5bcd08db | 1227 | r = subvol_remove_children(subvol_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1228 | else { |
1229 | _cleanup_close_ int child_fd = -1; | |
1230 | ||
1231 | /* Subvolume is somewhere further down, | |
1232 | * hence we need to open the | |
1233 | * containing directory first */ | |
1234 | ||
f7c9f4a2 | 1235 | child_fd = openat(subvol_fd, ino_args.name, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1236 | if (child_fd < 0) |
1237 | return -errno; | |
1238 | ||
5bcd08db | 1239 | r = subvol_remove_children(child_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1240 | } |
1241 | if (r < 0) | |
1242 | return r; | |
1243 | } | |
1244 | ||
1245 | /* Increase search key by one, to read the next item, if we can. */ | |
1246 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1247 | break; | |
1248 | } | |
1249 | ||
1250 | /* OK, the child subvolumes should all be gone now, let's try | |
1251 | * again to remove the subvolume */ | |
1252 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) < 0) | |
1253 | return -errno; | |
1254 | ||
5bcd08db | 1255 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); |
d9e2daaf LP |
1256 | return 0; |
1257 | } | |
1258 | ||
5bcd08db | 1259 | int btrfs_subvol_remove(const char *path, BtrfsRemoveFlags flags) { |
d9e2daaf LP |
1260 | _cleanup_close_ int fd = -1; |
1261 | const char *subvolume; | |
1262 | int r; | |
1263 | ||
1264 | assert(path); | |
1265 | ||
1266 | r = extract_subvolume_name(path, &subvolume); | |
1267 | if (r < 0) | |
1268 | return r; | |
1269 | ||
ef8becfa | 1270 | fd = open_parent(path, O_CLOEXEC, 0); |
d9e2daaf LP |
1271 | if (fd < 0) |
1272 | return fd; | |
1273 | ||
5bcd08db LP |
1274 | return subvol_remove_children(fd, subvolume, 0, flags); |
1275 | } | |
1276 | ||
1277 | int btrfs_subvol_remove_fd(int fd, const char *subvolume, BtrfsRemoveFlags flags) { | |
1278 | return subvol_remove_children(fd, subvolume, 0, flags); | |
1279 | } | |
1280 | ||
1281 | int btrfs_qgroup_copy_limits(int fd, uint64_t old_qgroupid, uint64_t new_qgroupid) { | |
1282 | ||
1283 | struct btrfs_ioctl_search_args args = { | |
1284 | /* Tree of quota items */ | |
1285 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1286 | ||
1287 | /* The object ID is always 0 */ | |
1288 | .key.min_objectid = 0, | |
1289 | .key.max_objectid = 0, | |
1290 | ||
1291 | /* Look precisely for the quota items */ | |
1292 | .key.min_type = BTRFS_QGROUP_LIMIT_KEY, | |
1293 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
1294 | ||
1295 | /* For our qgroup */ | |
1296 | .key.min_offset = old_qgroupid, | |
1297 | .key.max_offset = old_qgroupid, | |
1298 | ||
1299 | /* No restrictions on the other components */ | |
1300 | .key.min_transid = 0, | |
1301 | .key.max_transid = (uint64_t) -1, | |
1302 | }; | |
1303 | ||
1304 | int r; | |
1305 | ||
1306 | r = btrfs_is_filesystem(fd); | |
1307 | if (r < 0) | |
1308 | return r; | |
1309 | if (!r) | |
1310 | return -ENOTTY; | |
1311 | ||
1312 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1313 | const struct btrfs_ioctl_search_header *sh; | |
1314 | unsigned i; | |
1315 | ||
1316 | args.key.nr_items = 256; | |
12ee6186 LP |
1317 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1318 | if (errno == ENOENT) /* quota tree missing: quota is not enabled, hence nothing to copy */ | |
1319 | break; | |
1320 | ||
5bcd08db | 1321 | return -errno; |
12ee6186 | 1322 | } |
5bcd08db LP |
1323 | |
1324 | if (args.key.nr_items <= 0) | |
1325 | break; | |
1326 | ||
1327 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1328 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1329 | struct btrfs_ioctl_qgroup_limit_args qargs; | |
1330 | unsigned c; | |
1331 | ||
1332 | /* Make sure we start the next search at least from this entry */ | |
1333 | btrfs_ioctl_search_args_set(&args, sh); | |
1334 | ||
1335 | if (sh->objectid != 0) | |
1336 | continue; | |
1337 | if (sh->type != BTRFS_QGROUP_LIMIT_KEY) | |
1338 | continue; | |
1339 | if (sh->offset != old_qgroupid) | |
1340 | continue; | |
1341 | ||
1342 | /* We found the entry, now copy things over. */ | |
1343 | ||
1344 | qargs = (struct btrfs_ioctl_qgroup_limit_args) { | |
1345 | .qgroupid = new_qgroupid, | |
1346 | ||
1347 | .lim.max_rfer = le64toh(qli->max_rfer), | |
1348 | .lim.max_excl = le64toh(qli->max_excl), | |
1349 | .lim.rsv_rfer = le64toh(qli->rsv_rfer), | |
1350 | .lim.rsv_excl = le64toh(qli->rsv_excl), | |
1351 | ||
1352 | .lim.flags = le64toh(qli->flags) & (BTRFS_QGROUP_LIMIT_MAX_RFER| | |
1353 | BTRFS_QGROUP_LIMIT_MAX_EXCL| | |
1354 | BTRFS_QGROUP_LIMIT_RSV_RFER| | |
1355 | BTRFS_QGROUP_LIMIT_RSV_EXCL), | |
1356 | }; | |
1357 | ||
1358 | for (c = 0;; c++) { | |
1359 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &qargs) < 0) { | |
1360 | if (errno == EBUSY && c < 10) { | |
1361 | (void) btrfs_quota_scan_wait(fd); | |
1362 | continue; | |
1363 | } | |
1364 | return -errno; | |
1365 | } | |
1366 | ||
1367 | break; | |
1368 | } | |
1369 | ||
1370 | return 1; | |
1371 | } | |
1372 | ||
1373 | /* Increase search key by one, to read the next item, if we can. */ | |
1374 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1375 | break; | |
1376 | } | |
1377 | ||
1378 | return 0; | |
d9e2daaf LP |
1379 | } |
1380 | ||
5bcd08db LP |
1381 | static int copy_quota_hierarchy(int fd, uint64_t old_subvol_id, uint64_t new_subvol_id) { |
1382 | _cleanup_free_ uint64_t *old_qgroups = NULL, *old_parent_qgroups = NULL; | |
1383 | bool copy_from_parent = false, insert_intermediary_qgroup = false; | |
1384 | int n_old_qgroups, n_old_parent_qgroups, r, i; | |
1385 | uint64_t old_parent_id; | |
1386 | ||
1387 | assert(fd >= 0); | |
1388 | ||
1389 | /* Copies a reduced form of quota information from the old to | |
1390 | * the new subvolume. */ | |
1391 | ||
1392 | n_old_qgroups = btrfs_qgroup_find_parents(fd, old_subvol_id, &old_qgroups); | |
1393 | if (n_old_qgroups <= 0) /* Nothing to copy */ | |
1394 | return n_old_qgroups; | |
1395 | ||
1396 | r = btrfs_subvol_get_parent(fd, old_subvol_id, &old_parent_id); | |
08c77cf3 LP |
1397 | if (r == -ENXIO) |
1398 | /* We have no parent, hence nothing to copy. */ | |
1399 | n_old_parent_qgroups = 0; | |
1400 | else if (r < 0) | |
5bcd08db | 1401 | return r; |
08c77cf3 LP |
1402 | else { |
1403 | n_old_parent_qgroups = btrfs_qgroup_find_parents(fd, old_parent_id, &old_parent_qgroups); | |
1404 | if (n_old_parent_qgroups < 0) | |
1405 | return n_old_parent_qgroups; | |
1406 | } | |
5bcd08db LP |
1407 | |
1408 | for (i = 0; i < n_old_qgroups; i++) { | |
1409 | uint64_t id; | |
1410 | int j; | |
1411 | ||
1412 | r = btrfs_qgroupid_split(old_qgroups[i], NULL, &id); | |
1413 | if (r < 0) | |
1414 | return r; | |
1415 | ||
1416 | if (id == old_subvol_id) { | |
1417 | /* The old subvolume was member of a qgroup | |
1418 | * that had the same id, but a different level | |
1419 | * as it self. Let's set up something similar | |
1420 | * in the destination. */ | |
1421 | insert_intermediary_qgroup = true; | |
1422 | break; | |
1423 | } | |
1424 | ||
1425 | for (j = 0; j < n_old_parent_qgroups; j++) | |
1426 | if (old_parent_qgroups[j] == old_qgroups[i]) { | |
1427 | /* The old subvolume shared a common | |
1428 | * parent qgroup with its parent | |
1429 | * subvolume. Let's set up something | |
1430 | * similar in the destination. */ | |
1431 | copy_from_parent = true; | |
1432 | } | |
1433 | } | |
1434 | ||
1435 | if (!insert_intermediary_qgroup && !copy_from_parent) | |
1436 | return 0; | |
1437 | ||
1438 | return btrfs_subvol_auto_qgroup_fd(fd, new_subvol_id, insert_intermediary_qgroup); | |
1439 | } | |
1440 | ||
1441 | static int copy_subtree_quota_limits(int fd, uint64_t old_subvol, uint64_t new_subvol) { | |
1442 | uint64_t old_subtree_qgroup, new_subtree_qgroup; | |
1443 | bool changed; | |
1444 | int r; | |
1445 | ||
1446 | /* First copy the leaf limits */ | |
1447 | r = btrfs_qgroup_copy_limits(fd, old_subvol, new_subvol); | |
1448 | if (r < 0) | |
1449 | return r; | |
1450 | changed = r > 0; | |
1451 | ||
1452 | /* Then, try to copy the subtree limits, if there are any. */ | |
1453 | r = btrfs_subvol_find_subtree_qgroup(fd, old_subvol, &old_subtree_qgroup); | |
1454 | if (r < 0) | |
1455 | return r; | |
1456 | if (r == 0) | |
1457 | return changed; | |
1458 | ||
1459 | r = btrfs_subvol_find_subtree_qgroup(fd, new_subvol, &new_subtree_qgroup); | |
1460 | if (r < 0) | |
1461 | return r; | |
1462 | if (r == 0) | |
1463 | return changed; | |
1464 | ||
1465 | r = btrfs_qgroup_copy_limits(fd, old_subtree_qgroup, new_subtree_qgroup); | |
1466 | if (r != 0) | |
1467 | return r; | |
1468 | ||
1469 | return changed; | |
d9e2daaf | 1470 | } |
f70a17f8 | 1471 | |
90578cbd | 1472 | static int subvol_snapshot_children(int old_fd, int new_fd, const char *subvolume, uint64_t old_subvol_id, BtrfsSnapshotFlags flags) { |
f70a17f8 LP |
1473 | |
1474 | struct btrfs_ioctl_search_args args = { | |
1475 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1476 | ||
1477 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1478 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1479 | ||
1480 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1481 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1482 | ||
1483 | .key.min_transid = 0, | |
1484 | .key.max_transid = (uint64_t) -1, | |
1485 | }; | |
1486 | ||
1487 | struct btrfs_ioctl_vol_args_v2 vol_args = { | |
1488 | .flags = flags & BTRFS_SNAPSHOT_READ_ONLY ? BTRFS_SUBVOL_RDONLY : 0, | |
1489 | .fd = old_fd, | |
1490 | }; | |
ffb296b2 | 1491 | _cleanup_close_ int subvolume_fd = -1; |
90578cbd LP |
1492 | uint64_t new_subvol_id; |
1493 | int r; | |
f70a17f8 LP |
1494 | |
1495 | assert(old_fd >= 0); | |
1496 | assert(new_fd >= 0); | |
1497 | assert(subvolume); | |
1498 | ||
1499 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
f70a17f8 LP |
1500 | |
1501 | if (ioctl(new_fd, BTRFS_IOC_SNAP_CREATE_V2, &vol_args) < 0) | |
1502 | return -errno; | |
1503 | ||
5bcd08db LP |
1504 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE) && |
1505 | !(flags & BTRFS_SNAPSHOT_QUOTA)) | |
f70a17f8 LP |
1506 | return 0; |
1507 | ||
90578cbd LP |
1508 | if (old_subvol_id == 0) { |
1509 | r = btrfs_subvol_get_id_fd(old_fd, &old_subvol_id); | |
f70a17f8 LP |
1510 | if (r < 0) |
1511 | return r; | |
1512 | } | |
1513 | ||
90578cbd LP |
1514 | r = btrfs_subvol_get_id(new_fd, vol_args.name, &new_subvol_id); |
1515 | if (r < 0) | |
1516 | return r; | |
1517 | ||
5bcd08db LP |
1518 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1519 | (void) copy_quota_hierarchy(new_fd, old_subvol_id, new_subvol_id); | |
1520 | ||
1521 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE)) { | |
1522 | ||
1523 | if (flags & BTRFS_SNAPSHOT_QUOTA) | |
1524 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1525 | ||
1526 | return 0; | |
1527 | } | |
1528 | ||
90578cbd | 1529 | args.key.min_offset = args.key.max_offset = old_subvol_id; |
f70a17f8 LP |
1530 | |
1531 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1532 | const struct btrfs_ioctl_search_header *sh; | |
1533 | unsigned i; | |
1534 | ||
1535 | args.key.nr_items = 256; | |
1536 | if (ioctl(old_fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1537 | return -errno; | |
1538 | ||
1539 | if (args.key.nr_items <= 0) | |
1540 | break; | |
1541 | ||
1542 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1543 | _cleanup_free_ char *p = NULL, *c = NULL, *np = NULL; | |
1544 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1545 | const struct btrfs_root_ref *ref; | |
1546 | _cleanup_close_ int old_child_fd = -1, new_child_fd = -1; | |
1547 | ||
1548 | btrfs_ioctl_search_args_set(&args, sh); | |
1549 | ||
1550 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1551 | continue; | |
90578cbd LP |
1552 | |
1553 | /* Avoid finding the source subvolume a second | |
1554 | * time */ | |
1555 | if (sh->offset != old_subvol_id) | |
f70a17f8 LP |
1556 | continue; |
1557 | ||
90578cbd LP |
1558 | /* Avoid running into loops if the new |
1559 | * subvolume is below the old one. */ | |
1560 | if (sh->objectid == new_subvol_id) | |
1561 | continue; | |
f70a17f8 | 1562 | |
90578cbd | 1563 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
f70a17f8 LP |
1564 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); |
1565 | if (!p) | |
1566 | return -ENOMEM; | |
1567 | ||
1568 | zero(ino_args); | |
90578cbd | 1569 | ino_args.treeid = old_subvol_id; |
f70a17f8 LP |
1570 | ino_args.objectid = htole64(ref->dirid); |
1571 | ||
1572 | if (ioctl(old_fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1573 | return -errno; | |
1574 | ||
1575 | /* The kernel returns an empty name if the | |
1576 | * subvolume is in the top-level directory, | |
1577 | * and otherwise appends a slash, so that we | |
1578 | * can just concatenate easily here, without | |
1579 | * adding a slash. */ | |
1580 | c = strappend(ino_args.name, p); | |
1581 | if (!c) | |
1582 | return -ENOMEM; | |
1583 | ||
f7c9f4a2 | 1584 | old_child_fd = openat(old_fd, c, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1585 | if (old_child_fd < 0) |
1586 | return -errno; | |
1587 | ||
605405c6 | 1588 | np = strjoin(subvolume, "/", ino_args.name); |
f70a17f8 LP |
1589 | if (!np) |
1590 | return -ENOMEM; | |
1591 | ||
f7c9f4a2 | 1592 | new_child_fd = openat(new_fd, np, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1593 | if (new_child_fd < 0) |
1594 | return -errno; | |
1595 | ||
ffb296b2 LP |
1596 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { |
1597 | /* If the snapshot is read-only we | |
1598 | * need to mark it writable | |
1599 | * temporarily, to put the subsnapshot | |
1600 | * into place. */ | |
1601 | ||
1602 | if (subvolume_fd < 0) { | |
f7c9f4a2 | 1603 | subvolume_fd = openat(new_fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
ffb296b2 LP |
1604 | if (subvolume_fd < 0) |
1605 | return -errno; | |
1606 | } | |
1607 | ||
1608 | r = btrfs_subvol_set_read_only_fd(subvolume_fd, false); | |
1609 | if (r < 0) | |
1610 | return r; | |
1611 | } | |
1612 | ||
f70a17f8 | 1613 | /* When btrfs clones the subvolumes, child |
90578cbd | 1614 | * subvolumes appear as empty directories. Remove |
f70a17f8 LP |
1615 | * them, so that we can create a new snapshot |
1616 | * in their place */ | |
ffb296b2 LP |
1617 | if (unlinkat(new_child_fd, p, AT_REMOVEDIR) < 0) { |
1618 | int k = -errno; | |
1619 | ||
1620 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) | |
1621 | (void) btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1622 | ||
1623 | return k; | |
1624 | } | |
f70a17f8 LP |
1625 | |
1626 | r = subvol_snapshot_children(old_child_fd, new_child_fd, p, sh->objectid, flags & ~BTRFS_SNAPSHOT_FALLBACK_COPY); | |
ffb296b2 LP |
1627 | |
1628 | /* Restore the readonly flag */ | |
1629 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
1630 | int k; | |
1631 | ||
1632 | k = btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1633 | if (r >= 0 && k < 0) | |
1634 | return k; | |
1635 | } | |
1636 | ||
f70a17f8 LP |
1637 | if (r < 0) |
1638 | return r; | |
1639 | } | |
1640 | ||
1641 | /* Increase search key by one, to read the next item, if we can. */ | |
1642 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1643 | break; | |
1644 | } | |
1645 | ||
5bcd08db LP |
1646 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1647 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1648 | ||
f70a17f8 LP |
1649 | return 0; |
1650 | } | |
1651 | ||
1652 | int btrfs_subvol_snapshot_fd(int old_fd, const char *new_path, BtrfsSnapshotFlags flags) { | |
1653 | _cleanup_close_ int new_fd = -1; | |
1654 | const char *subvolume; | |
1655 | int r; | |
1656 | ||
1657 | assert(old_fd >= 0); | |
1658 | assert(new_path); | |
1659 | ||
2904e949 | 1660 | r = btrfs_is_subvol_fd(old_fd); |
f70a17f8 LP |
1661 | if (r < 0) |
1662 | return r; | |
1663 | if (r == 0) { | |
17cbb288 LP |
1664 | bool plain_directory = false; |
1665 | ||
1666 | /* If the source isn't a proper subvolume, fail unless fallback is requested */ | |
f70a17f8 LP |
1667 | if (!(flags & BTRFS_SNAPSHOT_FALLBACK_COPY)) |
1668 | return -EISDIR; | |
1669 | ||
1670 | r = btrfs_subvol_make(new_path); | |
17cbb288 LP |
1671 | if (r == -ENOTTY && (flags & BTRFS_SNAPSHOT_FALLBACK_DIRECTORY)) { |
1672 | /* If the destination doesn't support subvolumes, then use a plain directory, if that's requested. */ | |
1673 | if (mkdir(new_path, 0755) < 0) | |
679def2a | 1674 | return -errno; |
17cbb288 LP |
1675 | |
1676 | plain_directory = true; | |
1677 | } else if (r < 0) | |
f70a17f8 LP |
1678 | return r; |
1679 | ||
1c876927 | 1680 | r = copy_directory_fd(old_fd, new_path, COPY_MERGE|COPY_REFLINK); |
17cbb288 LP |
1681 | if (r < 0) |
1682 | goto fallback_fail; | |
f70a17f8 LP |
1683 | |
1684 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
17cbb288 LP |
1685 | |
1686 | if (plain_directory) { | |
1687 | /* Plain directories have no recursive read-only flag, but something pretty close to | |
1688 | * it: the IMMUTABLE bit. Let's use this here, if this is requested. */ | |
1689 | ||
1690 | if (flags & BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE) | |
1691 | (void) chattr_path(new_path, FS_IMMUTABLE_FL, FS_IMMUTABLE_FL); | |
1692 | } else { | |
1693 | r = btrfs_subvol_set_read_only(new_path, true); | |
1694 | if (r < 0) | |
1695 | goto fallback_fail; | |
f70a17f8 LP |
1696 | } |
1697 | } | |
1698 | ||
1699 | return 0; | |
17cbb288 LP |
1700 | |
1701 | fallback_fail: | |
1702 | (void) rm_rf(new_path, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME); | |
1703 | return r; | |
f70a17f8 LP |
1704 | } |
1705 | ||
1706 | r = extract_subvolume_name(new_path, &subvolume); | |
1707 | if (r < 0) | |
1708 | return r; | |
1709 | ||
ef8becfa | 1710 | new_fd = open_parent(new_path, O_CLOEXEC, 0); |
f70a17f8 LP |
1711 | if (new_fd < 0) |
1712 | return new_fd; | |
1713 | ||
1714 | return subvol_snapshot_children(old_fd, new_fd, subvolume, 0, flags); | |
1715 | } | |
1716 | ||
1717 | int btrfs_subvol_snapshot(const char *old_path, const char *new_path, BtrfsSnapshotFlags flags) { | |
1718 | _cleanup_close_ int old_fd = -1; | |
1719 | ||
1720 | assert(old_path); | |
1721 | assert(new_path); | |
1722 | ||
1723 | old_fd = open(old_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1724 | if (old_fd < 0) | |
1725 | return -errno; | |
1726 | ||
1727 | return btrfs_subvol_snapshot_fd(old_fd, new_path, flags); | |
1728 | } | |
5bcd08db LP |
1729 | |
1730 | int btrfs_qgroup_find_parents(int fd, uint64_t qgroupid, uint64_t **ret) { | |
1731 | ||
1732 | struct btrfs_ioctl_search_args args = { | |
1733 | /* Tree of quota items */ | |
1734 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1735 | ||
1736 | /* Look precisely for the quota relation items */ | |
1737 | .key.min_type = BTRFS_QGROUP_RELATION_KEY, | |
1738 | .key.max_type = BTRFS_QGROUP_RELATION_KEY, | |
1739 | ||
1740 | /* No restrictions on the other components */ | |
1741 | .key.min_offset = 0, | |
1742 | .key.max_offset = (uint64_t) -1, | |
1743 | ||
1744 | .key.min_transid = 0, | |
1745 | .key.max_transid = (uint64_t) -1, | |
1746 | }; | |
1747 | ||
1748 | _cleanup_free_ uint64_t *items = NULL; | |
1749 | size_t n_items = 0, n_allocated = 0; | |
1750 | int r; | |
1751 | ||
1752 | assert(fd >= 0); | |
1753 | assert(ret); | |
1754 | ||
1755 | if (qgroupid == 0) { | |
1756 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
1757 | if (r < 0) | |
1758 | return r; | |
1759 | } else { | |
1760 | r = btrfs_is_filesystem(fd); | |
1761 | if (r < 0) | |
1762 | return r; | |
1763 | if (!r) | |
1764 | return -ENOTTY; | |
1765 | } | |
1766 | ||
1767 | args.key.min_objectid = args.key.max_objectid = qgroupid; | |
1768 | ||
1769 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1770 | const struct btrfs_ioctl_search_header *sh; | |
1771 | unsigned i; | |
1772 | ||
1773 | args.key.nr_items = 256; | |
12ee6186 LP |
1774 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1775 | if (errno == ENOENT) /* quota tree missing: quota is disabled */ | |
1776 | break; | |
1777 | ||
5bcd08db | 1778 | return -errno; |
12ee6186 | 1779 | } |
5bcd08db LP |
1780 | |
1781 | if (args.key.nr_items <= 0) | |
1782 | break; | |
1783 | ||
1784 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1785 | ||
1786 | /* Make sure we start the next search at least from this entry */ | |
1787 | btrfs_ioctl_search_args_set(&args, sh); | |
1788 | ||
1789 | if (sh->type != BTRFS_QGROUP_RELATION_KEY) | |
1790 | continue; | |
1791 | if (sh->offset < sh->objectid) | |
1792 | continue; | |
1793 | if (sh->objectid != qgroupid) | |
1794 | continue; | |
1795 | ||
1796 | if (!GREEDY_REALLOC(items, n_allocated, n_items+1)) | |
1797 | return -ENOMEM; | |
1798 | ||
1799 | items[n_items++] = sh->offset; | |
1800 | } | |
1801 | ||
1802 | /* Increase search key by one, to read the next item, if we can. */ | |
1803 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1804 | break; | |
1805 | } | |
1806 | ||
1807 | if (n_items <= 0) { | |
1808 | *ret = NULL; | |
1809 | return 0; | |
1810 | } | |
1811 | ||
1cc6c93a | 1812 | *ret = TAKE_PTR(items); |
5bcd08db LP |
1813 | |
1814 | return (int) n_items; | |
1815 | } | |
1816 | ||
1817 | int btrfs_subvol_auto_qgroup_fd(int fd, uint64_t subvol_id, bool insert_intermediary_qgroup) { | |
1818 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1819 | uint64_t parent_subvol; | |
1820 | bool changed = false; | |
1821 | int n = 0, r; | |
1822 | ||
1823 | assert(fd >= 0); | |
1824 | ||
1825 | /* | |
1826 | * Sets up the specified subvolume's qgroup automatically in | |
1827 | * one of two ways: | |
1828 | * | |
1829 | * If insert_intermediary_qgroup is false, the subvolume's | |
1830 | * leaf qgroup will be assigned to the same parent qgroups as | |
1831 | * the subvolume's parent subvolume. | |
1832 | * | |
1833 | * If insert_intermediary_qgroup is true a new intermediary | |
1834 | * higher-level qgroup is created, with a higher level number, | |
1835 | * but reusing the id of the subvolume. The level number is | |
1836 | * picked as one smaller than the lowest level qgroup the | |
1837 | * parent subvolume is a member of. If the parent subvolume's | |
1838 | * leaf qgroup is assigned to no higher-level qgroup a new | |
1839 | * qgroup of level 255 is created instead. Either way, the new | |
1840 | * qgroup is then assigned to the parent's higher-level | |
1841 | * qgroup, and the subvolume itself is assigned to it. | |
1842 | * | |
1843 | * If the subvolume is already assigned to a higher level | |
1844 | * qgroup, no operation is executed. | |
1845 | * | |
1846 | * Effectively this means: regardless if | |
1847 | * insert_intermediary_qgroup is true or not, after this | |
1848 | * function is invoked the subvolume will be accounted within | |
1849 | * the same qgroups as the parent. However, if it is true, it | |
1850 | * will also get its own higher-level qgroup, which may in | |
1851 | * turn be used by subvolumes created beneath this subvolume | |
1852 | * later on. | |
1853 | * | |
1854 | * This hence defines a simple default qgroup setup for | |
1855 | * subvolumes, as long as this function is invoked on each | |
1856 | * created subvolume: each subvolume is always accounting | |
1857 | * together with its immediate parents. Optionally, if | |
1858 | * insert_intermediary_qgroup is true, it will also get a | |
1859 | * qgroup that then includes all its own child subvolumes. | |
1860 | */ | |
1861 | ||
1862 | if (subvol_id == 0) { | |
2904e949 | 1863 | r = btrfs_is_subvol_fd(fd); |
5bcd08db LP |
1864 | if (r < 0) |
1865 | return r; | |
1866 | if (!r) | |
1867 | return -ENOTTY; | |
1868 | ||
1869 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
1870 | if (r < 0) | |
1871 | return r; | |
1872 | } | |
1873 | ||
1874 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
1875 | if (n < 0) | |
1876 | return n; | |
1877 | if (n > 0) /* already parent qgroups set up, let's bail */ | |
1878 | return 0; | |
1879 | ||
08c77cf3 LP |
1880 | qgroups = mfree(qgroups); |
1881 | ||
5bcd08db | 1882 | r = btrfs_subvol_get_parent(fd, subvol_id, &parent_subvol); |
08c77cf3 LP |
1883 | if (r == -ENXIO) |
1884 | /* No parent, hence no qgroup memberships */ | |
1885 | n = 0; | |
1886 | else if (r < 0) | |
5bcd08db | 1887 | return r; |
08c77cf3 LP |
1888 | else { |
1889 | n = btrfs_qgroup_find_parents(fd, parent_subvol, &qgroups); | |
1890 | if (n < 0) | |
1891 | return n; | |
1892 | } | |
5bcd08db LP |
1893 | |
1894 | if (insert_intermediary_qgroup) { | |
1895 | uint64_t lowest = 256, new_qgroupid; | |
1896 | bool created = false; | |
1897 | int i; | |
1898 | ||
1899 | /* Determine the lowest qgroup that the parent | |
1900 | * subvolume is assigned to. */ | |
1901 | ||
1902 | for (i = 0; i < n; i++) { | |
1903 | uint64_t level; | |
1904 | ||
1905 | r = btrfs_qgroupid_split(qgroups[i], &level, NULL); | |
1906 | if (r < 0) | |
1907 | return r; | |
1908 | ||
1909 | if (level < lowest) | |
1910 | lowest = level; | |
1911 | } | |
1912 | ||
1913 | if (lowest <= 1) /* There are no levels left we could use insert an intermediary qgroup at */ | |
1914 | return -EBUSY; | |
1915 | ||
1916 | r = btrfs_qgroupid_make(lowest - 1, subvol_id, &new_qgroupid); | |
1917 | if (r < 0) | |
1918 | return r; | |
1919 | ||
1920 | /* Create the new intermediary group, unless it already exists */ | |
1921 | r = btrfs_qgroup_create(fd, new_qgroupid); | |
1922 | if (r < 0 && r != -EEXIST) | |
1923 | return r; | |
1924 | if (r >= 0) | |
1925 | changed = created = true; | |
1926 | ||
1927 | for (i = 0; i < n; i++) { | |
1928 | r = btrfs_qgroup_assign(fd, new_qgroupid, qgroups[i]); | |
1929 | if (r < 0 && r != -EEXIST) { | |
1930 | if (created) | |
1931 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
1932 | ||
1933 | return r; | |
1934 | } | |
1935 | if (r >= 0) | |
1936 | changed = true; | |
1937 | } | |
1938 | ||
1939 | r = btrfs_qgroup_assign(fd, subvol_id, new_qgroupid); | |
1940 | if (r < 0 && r != -EEXIST) { | |
1941 | if (created) | |
1942 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
1943 | return r; | |
1944 | } | |
1945 | if (r >= 0) | |
1946 | changed = true; | |
1947 | ||
1948 | } else { | |
1949 | int i; | |
1950 | ||
1951 | /* Assign our subvolume to all the same qgroups as the parent */ | |
1952 | ||
1953 | for (i = 0; i < n; i++) { | |
1954 | r = btrfs_qgroup_assign(fd, subvol_id, qgroups[i]); | |
1955 | if (r < 0 && r != -EEXIST) | |
1956 | return r; | |
1957 | if (r >= 0) | |
1958 | changed = true; | |
1959 | } | |
1960 | } | |
1961 | ||
1962 | return changed; | |
1963 | } | |
1964 | ||
1965 | int btrfs_subvol_auto_qgroup(const char *path, uint64_t subvol_id, bool create_intermediary_qgroup) { | |
1966 | _cleanup_close_ int fd = -1; | |
1967 | ||
1968 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1969 | if (fd < 0) | |
1970 | return -errno; | |
1971 | ||
1972 | return btrfs_subvol_auto_qgroup_fd(fd, subvol_id, create_intermediary_qgroup); | |
1973 | } | |
1974 | ||
1975 | int btrfs_subvol_get_parent(int fd, uint64_t subvol_id, uint64_t *ret) { | |
1976 | ||
1977 | struct btrfs_ioctl_search_args args = { | |
1978 | /* Tree of tree roots */ | |
1979 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1980 | ||
1981 | /* Look precisely for the subvolume items */ | |
1982 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1983 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1984 | ||
1985 | /* No restrictions on the other components */ | |
1986 | .key.min_offset = 0, | |
1987 | .key.max_offset = (uint64_t) -1, | |
1988 | ||
1989 | .key.min_transid = 0, | |
1990 | .key.max_transid = (uint64_t) -1, | |
1991 | }; | |
1992 | int r; | |
1993 | ||
1994 | assert(fd >= 0); | |
1995 | assert(ret); | |
1996 | ||
1997 | if (subvol_id == 0) { | |
1998 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
1999 | if (r < 0) | |
2000 | return r; | |
2001 | } else { | |
2002 | r = btrfs_is_filesystem(fd); | |
2003 | if (r < 0) | |
2004 | return r; | |
2005 | if (!r) | |
2006 | return -ENOTTY; | |
2007 | } | |
2008 | ||
2009 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
2010 | ||
2011 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
2012 | const struct btrfs_ioctl_search_header *sh; | |
2013 | unsigned i; | |
2014 | ||
2015 | args.key.nr_items = 256; | |
2016 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
9c4615fb | 2017 | return negative_errno(); |
5bcd08db LP |
2018 | |
2019 | if (args.key.nr_items <= 0) | |
2020 | break; | |
2021 | ||
2022 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
2023 | ||
2024 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
2025 | continue; | |
2026 | if (sh->objectid != subvol_id) | |
2027 | continue; | |
2028 | ||
2029 | *ret = sh->offset; | |
2030 | return 0; | |
2031 | } | |
2032 | } | |
2033 | ||
2034 | return -ENXIO; | |
2035 | } |