]>
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 | ||
62f9666a | 119 | int btrfs_subvol_make_fd(int fd, const char *subvolume) { |
d7c7c334 | 120 | struct btrfs_ioctl_vol_args args = {}; |
62f9666a FB |
121 | int r; |
122 | ||
123 | assert(subvolume); | |
124 | ||
125 | r = validate_subvolume_name(subvolume); | |
126 | if (r < 0) | |
127 | return r; | |
128 | ||
129 | strncpy(args.name, subvolume, sizeof(args.name)-1); | |
130 | ||
131 | if (ioctl(fd, BTRFS_IOC_SUBVOL_CREATE, &args) < 0) | |
132 | return -errno; | |
133 | ||
134 | return 0; | |
135 | } | |
136 | ||
137 | int btrfs_subvol_make(const char *path) { | |
d7c7c334 LP |
138 | _cleanup_close_ int fd = -1; |
139 | const char *subvolume; | |
140 | int r; | |
141 | ||
142 | assert(path); | |
143 | ||
144 | r = extract_subvolume_name(path, &subvolume); | |
145 | if (r < 0) | |
146 | return r; | |
147 | ||
ef8becfa | 148 | fd = open_parent(path, O_CLOEXEC, 0); |
d7c7c334 LP |
149 | if (fd < 0) |
150 | return fd; | |
151 | ||
62f9666a | 152 | return btrfs_subvol_make_fd(fd, subvolume); |
d7c7c334 LP |
153 | } |
154 | ||
0d6e763b | 155 | int btrfs_subvol_set_read_only_fd(int fd, bool b) { |
d7c7c334 | 156 | uint64_t flags, nflags; |
0d6e763b | 157 | struct stat st; |
d7c7c334 | 158 | |
0d6e763b LP |
159 | assert(fd >= 0); |
160 | ||
161 | if (fstat(fd, &st) < 0) | |
d7c7c334 LP |
162 | return -errno; |
163 | ||
0d6e763b LP |
164 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) |
165 | return -EINVAL; | |
166 | ||
d7c7c334 LP |
167 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) |
168 | return -errno; | |
169 | ||
170 | if (b) | |
171 | nflags = flags | BTRFS_SUBVOL_RDONLY; | |
172 | else | |
173 | nflags = flags & ~BTRFS_SUBVOL_RDONLY; | |
174 | ||
175 | if (flags == nflags) | |
176 | return 0; | |
177 | ||
178 | if (ioctl(fd, BTRFS_IOC_SUBVOL_SETFLAGS, &nflags) < 0) | |
179 | return -errno; | |
180 | ||
181 | return 0; | |
182 | } | |
183 | ||
0d6e763b LP |
184 | int btrfs_subvol_set_read_only(const char *path, bool b) { |
185 | _cleanup_close_ int fd = -1; | |
186 | ||
187 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
188 | if (fd < 0) | |
189 | return -errno; | |
190 | ||
191 | return btrfs_subvol_set_read_only_fd(fd, b); | |
192 | } | |
193 | ||
10f9c755 | 194 | int btrfs_subvol_get_read_only_fd(int fd) { |
cd61c3bf | 195 | uint64_t flags; |
62572894 LP |
196 | struct stat st; |
197 | ||
198 | assert(fd >= 0); | |
199 | ||
200 | if (fstat(fd, &st) < 0) | |
201 | return -errno; | |
202 | ||
203 | if (!S_ISDIR(st.st_mode) || st.st_ino != 256) | |
204 | return -EINVAL; | |
cd61c3bf LP |
205 | |
206 | if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0) | |
207 | return -errno; | |
208 | ||
209 | return !!(flags & BTRFS_SUBVOL_RDONLY); | |
210 | } | |
211 | ||
d7c7c334 LP |
212 | int btrfs_reflink(int infd, int outfd) { |
213 | int r; | |
214 | ||
215 | assert(infd >= 0); | |
216 | assert(outfd >= 0); | |
217 | ||
3cc44114 | 218 | /* Make sure we invoke the ioctl on a regular file, so that no device driver accidentally gets it. */ |
62572894 | 219 | |
3cc44114 | 220 | r = fd_verify_regular(outfd); |
d7c7c334 | 221 | if (r < 0) |
3cc44114 LP |
222 | return r; |
223 | ||
224 | if (ioctl(outfd, BTRFS_IOC_CLONE, infd) < 0) | |
d7c7c334 LP |
225 | return -errno; |
226 | ||
227 | return 0; | |
228 | } | |
229 | ||
1c7dd825 LP |
230 | int btrfs_clone_range(int infd, uint64_t in_offset, int outfd, uint64_t out_offset, uint64_t sz) { |
231 | struct btrfs_ioctl_clone_range_args args = { | |
232 | .src_fd = infd, | |
233 | .src_offset = in_offset, | |
234 | .src_length = sz, | |
235 | .dest_offset = out_offset, | |
236 | }; | |
237 | int r; | |
238 | ||
239 | assert(infd >= 0); | |
240 | assert(outfd >= 0); | |
241 | assert(sz > 0); | |
242 | ||
3cc44114 | 243 | r = fd_verify_regular(outfd); |
1c7dd825 | 244 | if (r < 0) |
3cc44114 LP |
245 | return r; |
246 | ||
247 | if (ioctl(outfd, BTRFS_IOC_CLONE_RANGE, &args) < 0) | |
1c7dd825 LP |
248 | return -errno; |
249 | ||
250 | return 0; | |
251 | } | |
252 | ||
efe02862 | 253 | int btrfs_get_block_device_fd(int fd, dev_t *dev) { |
d7c7c334 | 254 | struct btrfs_ioctl_fs_info_args fsi = {}; |
d7c7c334 | 255 | uint64_t id; |
62572894 | 256 | int r; |
d7c7c334 | 257 | |
efe02862 | 258 | assert(fd >= 0); |
d7c7c334 LP |
259 | assert(dev); |
260 | ||
62572894 LP |
261 | r = btrfs_is_filesystem(fd); |
262 | if (r < 0) | |
263 | return r; | |
264 | if (!r) | |
265 | return -ENOTTY; | |
266 | ||
d7c7c334 LP |
267 | if (ioctl(fd, BTRFS_IOC_FS_INFO, &fsi) < 0) |
268 | return -errno; | |
269 | ||
270 | /* We won't do this for btrfs RAID */ | |
66ae5130 LP |
271 | if (fsi.num_devices != 1) { |
272 | *dev = 0; | |
d7c7c334 | 273 | return 0; |
66ae5130 | 274 | } |
d7c7c334 LP |
275 | |
276 | for (id = 1; id <= fsi.max_id; id++) { | |
277 | struct btrfs_ioctl_dev_info_args di = { | |
278 | .devid = id, | |
279 | }; | |
280 | struct stat st; | |
281 | ||
282 | if (ioctl(fd, BTRFS_IOC_DEV_INFO, &di) < 0) { | |
283 | if (errno == ENODEV) | |
284 | continue; | |
285 | ||
286 | return -errno; | |
287 | } | |
288 | ||
289 | if (stat((char*) di.path, &st) < 0) | |
290 | return -errno; | |
291 | ||
292 | if (!S_ISBLK(st.st_mode)) | |
293 | return -ENODEV; | |
294 | ||
295 | if (major(st.st_rdev) == 0) | |
296 | return -ENODEV; | |
297 | ||
298 | *dev = st.st_rdev; | |
299 | return 1; | |
300 | } | |
301 | ||
302 | return -ENODEV; | |
303 | } | |
10f9c755 | 304 | |
efe02862 LP |
305 | int btrfs_get_block_device(const char *path, dev_t *dev) { |
306 | _cleanup_close_ int fd = -1; | |
307 | ||
308 | assert(path); | |
309 | assert(dev); | |
310 | ||
311 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
312 | if (fd < 0) | |
313 | return -errno; | |
314 | ||
315 | return btrfs_get_block_device_fd(fd, dev); | |
316 | } | |
317 | ||
10f9c755 LP |
318 | int btrfs_subvol_get_id_fd(int fd, uint64_t *ret) { |
319 | struct btrfs_ioctl_ino_lookup_args args = { | |
320 | .objectid = BTRFS_FIRST_FREE_OBJECTID | |
321 | }; | |
62572894 | 322 | int r; |
10f9c755 LP |
323 | |
324 | assert(fd >= 0); | |
325 | assert(ret); | |
326 | ||
62572894 LP |
327 | r = btrfs_is_filesystem(fd); |
328 | if (r < 0) | |
329 | return r; | |
330 | if (!r) | |
331 | return -ENOTTY; | |
332 | ||
10f9c755 LP |
333 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args) < 0) |
334 | return -errno; | |
335 | ||
336 | *ret = args.treeid; | |
337 | return 0; | |
338 | } | |
339 | ||
90578cbd LP |
340 | int btrfs_subvol_get_id(int fd, const char *subvol, uint64_t *ret) { |
341 | _cleanup_close_ int subvol_fd = -1; | |
342 | ||
343 | assert(fd >= 0); | |
344 | assert(ret); | |
345 | ||
346 | subvol_fd = openat(fd, subvol, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
347 | if (subvol_fd < 0) | |
348 | return -errno; | |
349 | ||
350 | return btrfs_subvol_get_id_fd(subvol_fd, ret); | |
351 | } | |
352 | ||
5743a585 LP |
353 | static bool btrfs_ioctl_search_args_inc(struct btrfs_ioctl_search_args *args) { |
354 | assert(args); | |
355 | ||
356 | /* the objectid, type, offset together make up the btrfs key, | |
357 | * which is considered a single 136byte integer when | |
358 | * comparing. This call increases the counter by one, dealing | |
359 | * with the overflow between the overflows */ | |
360 | ||
361 | if (args->key.min_offset < (uint64_t) -1) { | |
362 | args->key.min_offset++; | |
363 | return true; | |
364 | } | |
365 | ||
366 | if (args->key.min_type < (uint8_t) -1) { | |
367 | args->key.min_type++; | |
368 | args->key.min_offset = 0; | |
369 | return true; | |
370 | } | |
371 | ||
372 | if (args->key.min_objectid < (uint64_t) -1) { | |
373 | args->key.min_objectid++; | |
374 | args->key.min_offset = 0; | |
375 | args->key.min_type = 0; | |
376 | return true; | |
377 | } | |
378 | ||
379 | return 0; | |
380 | } | |
381 | ||
382 | static void btrfs_ioctl_search_args_set(struct btrfs_ioctl_search_args *args, const struct btrfs_ioctl_search_header *h) { | |
383 | assert(args); | |
384 | assert(h); | |
385 | ||
386 | args->key.min_objectid = h->objectid; | |
387 | args->key.min_type = h->type; | |
388 | args->key.min_offset = h->offset; | |
389 | } | |
390 | ||
391 | static int btrfs_ioctl_search_args_compare(const struct btrfs_ioctl_search_args *args) { | |
392 | assert(args); | |
393 | ||
394 | /* Compare min and max */ | |
395 | ||
396 | if (args->key.min_objectid < args->key.max_objectid) | |
397 | return -1; | |
398 | if (args->key.min_objectid > args->key.max_objectid) | |
399 | return 1; | |
400 | ||
401 | if (args->key.min_type < args->key.max_type) | |
402 | return -1; | |
403 | if (args->key.min_type > args->key.max_type) | |
404 | return 1; | |
405 | ||
406 | if (args->key.min_offset < args->key.max_offset) | |
407 | return -1; | |
408 | if (args->key.min_offset > args->key.max_offset) | |
409 | return 1; | |
410 | ||
411 | return 0; | |
412 | } | |
413 | ||
414 | #define FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) \ | |
415 | for ((i) = 0, \ | |
416 | (sh) = (const struct btrfs_ioctl_search_header*) (args).buf; \ | |
417 | (i) < (args).key.nr_items; \ | |
418 | (i)++, \ | |
419 | (sh) = (const struct btrfs_ioctl_search_header*) ((uint8_t*) (sh) + sizeof(struct btrfs_ioctl_search_header) + (sh)->len)) | |
420 | ||
421 | #define BTRFS_IOCTL_SEARCH_HEADER_BODY(sh) \ | |
422 | ((void*) ((uint8_t*) sh + sizeof(struct btrfs_ioctl_search_header))) | |
423 | ||
5bcd08db | 424 | int btrfs_subvol_get_info_fd(int fd, uint64_t subvol_id, BtrfsSubvolInfo *ret) { |
10f9c755 LP |
425 | struct btrfs_ioctl_search_args args = { |
426 | /* Tree of tree roots */ | |
b6b18498 | 427 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, |
10f9c755 LP |
428 | |
429 | /* Look precisely for the subvolume items */ | |
430 | .key.min_type = BTRFS_ROOT_ITEM_KEY, | |
431 | .key.max_type = BTRFS_ROOT_ITEM_KEY, | |
432 | ||
10f9c755 LP |
433 | .key.min_offset = 0, |
434 | .key.max_offset = (uint64_t) -1, | |
5743a585 LP |
435 | |
436 | /* No restrictions on the other components */ | |
10f9c755 LP |
437 | .key.min_transid = 0, |
438 | .key.max_transid = (uint64_t) -1, | |
10f9c755 LP |
439 | }; |
440 | ||
b6b18498 | 441 | bool found = false; |
10f9c755 LP |
442 | int r; |
443 | ||
444 | assert(fd >= 0); | |
445 | assert(ret); | |
446 | ||
5bcd08db LP |
447 | if (subvol_id == 0) { |
448 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
449 | if (r < 0) | |
450 | return r; | |
451 | } else { | |
452 | r = btrfs_is_filesystem(fd); | |
453 | if (r < 0) | |
454 | return r; | |
455 | if (!r) | |
456 | return -ENOTTY; | |
457 | } | |
10f9c755 LP |
458 | |
459 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
10f9c755 | 460 | |
5743a585 | 461 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
462 | const struct btrfs_ioctl_search_header *sh; |
463 | unsigned i; | |
464 | ||
465 | args.key.nr_items = 256; | |
466 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
467 | return -errno; | |
468 | ||
469 | if (args.key.nr_items <= 0) | |
470 | break; | |
10f9c755 | 471 | |
5743a585 | 472 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
10f9c755 | 473 | |
b6b18498 LP |
474 | const struct btrfs_root_item *ri; |
475 | ||
5743a585 LP |
476 | /* Make sure we start the next search at least from this entry */ |
477 | btrfs_ioctl_search_args_set(&args, sh); | |
478 | ||
b6b18498 LP |
479 | if (sh->objectid != subvol_id) |
480 | continue; | |
481 | if (sh->type != BTRFS_ROOT_ITEM_KEY) | |
482 | continue; | |
5743a585 LP |
483 | |
484 | /* Older versions of the struct lacked the otime setting */ | |
b6b18498 LP |
485 | if (sh->len < offsetof(struct btrfs_root_item, otime) + sizeof(struct btrfs_timespec)) |
486 | continue; | |
10f9c755 | 487 | |
5743a585 | 488 | ri = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
10f9c755 | 489 | |
b6b18498 LP |
490 | ret->otime = (usec_t) le64toh(ri->otime.sec) * USEC_PER_SEC + |
491 | (usec_t) le32toh(ri->otime.nsec) / NSEC_PER_USEC; | |
10f9c755 | 492 | |
b6b18498 | 493 | ret->subvol_id = subvol_id; |
5d904a6a | 494 | ret->read_only = le64toh(ri->flags) & BTRFS_ROOT_SUBVOL_RDONLY; |
10f9c755 | 495 | |
b6b18498 LP |
496 | assert_cc(sizeof(ri->uuid) == sizeof(ret->uuid)); |
497 | memcpy(&ret->uuid, ri->uuid, sizeof(ret->uuid)); | |
498 | memcpy(&ret->parent_uuid, ri->parent_uuid, sizeof(ret->parent_uuid)); | |
499 | ||
500 | found = true; | |
501 | goto finish; | |
502 | } | |
503 | ||
5743a585 LP |
504 | /* Increase search key by one, to read the next item, if we can. */ |
505 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
506 | break; |
507 | } | |
508 | ||
509 | finish: | |
510 | if (!found) | |
511 | return -ENODATA; | |
512 | ||
513 | return 0; | |
514 | } | |
515 | ||
5bcd08db | 516 | int btrfs_qgroup_get_quota_fd(int fd, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
b6b18498 LP |
517 | |
518 | struct btrfs_ioctl_search_args args = { | |
519 | /* Tree of quota items */ | |
520 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
521 | ||
5743a585 LP |
522 | /* The object ID is always 0 */ |
523 | .key.min_objectid = 0, | |
524 | .key.max_objectid = 0, | |
525 | ||
b6b18498 LP |
526 | /* Look precisely for the quota items */ |
527 | .key.min_type = BTRFS_QGROUP_STATUS_KEY, | |
528 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
529 | ||
b6b18498 LP |
530 | /* No restrictions on the other components */ |
531 | .key.min_transid = 0, | |
532 | .key.max_transid = (uint64_t) -1, | |
533 | }; | |
534 | ||
b6b18498 LP |
535 | bool found_info = false, found_limit = false; |
536 | int r; | |
537 | ||
538 | assert(fd >= 0); | |
539 | assert(ret); | |
540 | ||
5bcd08db LP |
541 | if (qgroupid == 0) { |
542 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
543 | if (r < 0) | |
544 | return r; | |
545 | } else { | |
546 | r = btrfs_is_filesystem(fd); | |
547 | if (r < 0) | |
548 | return r; | |
549 | if (!r) | |
550 | return -ENOTTY; | |
551 | } | |
b6b18498 | 552 | |
5bcd08db | 553 | args.key.min_offset = args.key.max_offset = qgroupid; |
b6b18498 | 554 | |
5743a585 | 555 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { |
b6b18498 LP |
556 | const struct btrfs_ioctl_search_header *sh; |
557 | unsigned i; | |
558 | ||
559 | args.key.nr_items = 256; | |
12ee6186 LP |
560 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
561 | if (errno == ENOENT) /* quota tree is missing: quota disabled */ | |
562 | break; | |
563 | ||
b6b18498 | 564 | return -errno; |
12ee6186 | 565 | } |
b6b18498 LP |
566 | |
567 | if (args.key.nr_items <= 0) | |
568 | break; | |
569 | ||
5743a585 | 570 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { |
b6b18498 | 571 | |
5743a585 LP |
572 | /* Make sure we start the next search at least from this entry */ |
573 | btrfs_ioctl_search_args_set(&args, sh); | |
b6b18498 LP |
574 | |
575 | if (sh->objectid != 0) | |
576 | continue; | |
5bcd08db | 577 | if (sh->offset != qgroupid) |
b6b18498 LP |
578 | continue; |
579 | ||
b6b18498 | 580 | if (sh->type == BTRFS_QGROUP_INFO_KEY) { |
5743a585 | 581 | const struct btrfs_qgroup_info_item *qii = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 582 | |
cb81cd80 | 583 | ret->referenced = le64toh(qii->rfer); |
b6b18498 LP |
584 | ret->exclusive = le64toh(qii->excl); |
585 | ||
586 | found_info = true; | |
587 | ||
588 | } else if (sh->type == BTRFS_QGROUP_LIMIT_KEY) { | |
5743a585 | 589 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
b6b18498 | 590 | |
5bcd08db LP |
591 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_RFER) |
592 | ret->referenced_max = le64toh(qli->max_rfer); | |
593 | else | |
cb81cd80 | 594 | ret->referenced_max = (uint64_t) -1; |
5bcd08db LP |
595 | |
596 | if (le64toh(qli->flags) & BTRFS_QGROUP_LIMIT_MAX_EXCL) | |
597 | ret->exclusive_max = le64toh(qli->max_excl); | |
598 | else | |
b6b18498 LP |
599 | ret->exclusive_max = (uint64_t) -1; |
600 | ||
601 | found_limit = true; | |
602 | } | |
603 | ||
604 | if (found_info && found_limit) | |
605 | goto finish; | |
606 | } | |
607 | ||
5743a585 LP |
608 | /* Increase search key by one, to read the next item, if we can. */ |
609 | if (!btrfs_ioctl_search_args_inc(&args)) | |
b6b18498 LP |
610 | break; |
611 | } | |
612 | ||
613 | finish: | |
614 | if (!found_limit && !found_info) | |
615 | return -ENODATA; | |
616 | ||
617 | if (!found_info) { | |
cb81cd80 | 618 | ret->referenced = (uint64_t) -1; |
b6b18498 LP |
619 | ret->exclusive = (uint64_t) -1; |
620 | } | |
621 | ||
622 | if (!found_limit) { | |
cb81cd80 | 623 | ret->referenced_max = (uint64_t) -1; |
b6b18498 LP |
624 | ret->exclusive_max = (uint64_t) -1; |
625 | } | |
10f9c755 LP |
626 | |
627 | return 0; | |
628 | } | |
f27a3864 | 629 | |
5bcd08db LP |
630 | int btrfs_qgroup_get_quota(const char *path, uint64_t qgroupid, BtrfsQuotaInfo *ret) { |
631 | _cleanup_close_ int fd = -1; | |
632 | ||
633 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
634 | if (fd < 0) | |
635 | return -errno; | |
636 | ||
637 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
638 | } | |
639 | ||
640 | int btrfs_subvol_find_subtree_qgroup(int fd, uint64_t subvol_id, uint64_t *ret) { | |
641 | uint64_t level, lowest = (uint64_t) -1, lowest_qgroupid = 0; | |
642 | _cleanup_free_ uint64_t *qgroups = NULL; | |
643 | int r, n, i; | |
644 | ||
645 | assert(fd >= 0); | |
646 | assert(ret); | |
647 | ||
648 | /* This finds the "subtree" qgroup for a specific | |
649 | * subvolume. This only works for subvolumes that have been | |
650 | * prepared with btrfs_subvol_auto_qgroup_fd() with | |
651 | * insert_intermediary_qgroup=true (or equivalent). For others | |
652 | * it will return the leaf qgroup instead. The two cases may | |
653 | * be distuingished via the return value, which is 1 in case | |
654 | * an appropriate "subtree" qgroup was found, and 0 | |
655 | * otherwise. */ | |
656 | ||
657 | if (subvol_id == 0) { | |
658 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
659 | if (r < 0) | |
660 | return r; | |
661 | } | |
662 | ||
663 | r = btrfs_qgroupid_split(subvol_id, &level, NULL); | |
664 | if (r < 0) | |
665 | return r; | |
666 | if (level != 0) /* Input must be a leaf qgroup */ | |
667 | return -EINVAL; | |
668 | ||
669 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
670 | if (n < 0) | |
671 | return n; | |
672 | ||
673 | for (i = 0; i < n; i++) { | |
674 | uint64_t id; | |
675 | ||
676 | r = btrfs_qgroupid_split(qgroups[i], &level, &id); | |
677 | if (r < 0) | |
678 | return r; | |
679 | ||
680 | if (id != subvol_id) | |
681 | continue; | |
682 | ||
683 | if (lowest == (uint64_t) -1 || level < lowest) { | |
684 | lowest_qgroupid = qgroups[i]; | |
685 | lowest = level; | |
686 | } | |
687 | } | |
688 | ||
689 | if (lowest == (uint64_t) -1) { | |
690 | /* No suitable higher-level qgroup found, let's return | |
691 | * the leaf qgroup instead, and indicate that with the | |
692 | * return value. */ | |
693 | ||
694 | *ret = subvol_id; | |
695 | return 0; | |
696 | } | |
697 | ||
698 | *ret = lowest_qgroupid; | |
699 | return 1; | |
700 | } | |
701 | ||
702 | int btrfs_subvol_get_subtree_quota_fd(int fd, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
703 | uint64_t qgroupid; | |
704 | int r; | |
705 | ||
706 | assert(fd >= 0); | |
707 | assert(ret); | |
708 | ||
709 | /* This determines the quota data of the qgroup with the | |
710 | * lowest level, that shares the id part with the specified | |
711 | * subvolume. This is useful for determining the quota data | |
712 | * for entire subvolume subtrees, as long as the subtrees have | |
713 | * been set up with btrfs_qgroup_subvol_auto_fd() or in a | |
714 | * compatible way */ | |
715 | ||
716 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
717 | if (r < 0) | |
718 | return r; | |
719 | ||
720 | return btrfs_qgroup_get_quota_fd(fd, qgroupid, ret); | |
721 | } | |
722 | ||
723 | int btrfs_subvol_get_subtree_quota(const char *path, uint64_t subvol_id, BtrfsQuotaInfo *ret) { | |
724 | _cleanup_close_ int fd = -1; | |
725 | ||
726 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
727 | if (fd < 0) | |
728 | return -errno; | |
729 | ||
730 | return btrfs_subvol_get_subtree_quota_fd(fd, subvol_id, ret); | |
731 | } | |
732 | ||
f27a3864 | 733 | int btrfs_defrag_fd(int fd) { |
3cc44114 | 734 | int r; |
62572894 | 735 | |
f27a3864 LP |
736 | assert(fd >= 0); |
737 | ||
3cc44114 LP |
738 | r = fd_verify_regular(fd); |
739 | if (r < 0) | |
740 | return r; | |
62572894 | 741 | |
f27a3864 LP |
742 | if (ioctl(fd, BTRFS_IOC_DEFRAG, NULL) < 0) |
743 | return -errno; | |
744 | ||
745 | return 0; | |
746 | } | |
747 | ||
748 | int btrfs_defrag(const char *p) { | |
749 | _cleanup_close_ int fd = -1; | |
750 | ||
751 | fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
752 | if (fd < 0) | |
753 | return -errno; | |
754 | ||
755 | return btrfs_defrag_fd(fd); | |
756 | } | |
754061ce LP |
757 | |
758 | int btrfs_quota_enable_fd(int fd, bool b) { | |
759 | struct btrfs_ioctl_quota_ctl_args args = { | |
760 | .cmd = b ? BTRFS_QUOTA_CTL_ENABLE : BTRFS_QUOTA_CTL_DISABLE, | |
761 | }; | |
62572894 | 762 | int r; |
754061ce LP |
763 | |
764 | assert(fd >= 0); | |
765 | ||
62572894 LP |
766 | r = btrfs_is_filesystem(fd); |
767 | if (r < 0) | |
768 | return r; | |
769 | if (!r) | |
770 | return -ENOTTY; | |
771 | ||
754061ce LP |
772 | if (ioctl(fd, BTRFS_IOC_QUOTA_CTL, &args) < 0) |
773 | return -errno; | |
774 | ||
775 | return 0; | |
776 | } | |
777 | ||
778 | int btrfs_quota_enable(const char *path, bool b) { | |
779 | _cleanup_close_ int fd = -1; | |
780 | ||
781 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
782 | if (fd < 0) | |
783 | return -errno; | |
784 | ||
785 | return btrfs_quota_enable_fd(fd, b); | |
786 | } | |
d6ce17c7 | 787 | |
5bcd08db LP |
788 | int btrfs_qgroup_set_limit_fd(int fd, uint64_t qgroupid, uint64_t referenced_max) { |
789 | ||
d6ce17c7 | 790 | struct btrfs_ioctl_qgroup_limit_args args = { |
5bcd08db | 791 | .lim.max_rfer = referenced_max, |
d6ce17c7 LP |
792 | .lim.flags = BTRFS_QGROUP_LIMIT_MAX_RFER, |
793 | }; | |
5bcd08db | 794 | unsigned c; |
62572894 | 795 | int r; |
d6ce17c7 LP |
796 | |
797 | assert(fd >= 0); | |
798 | ||
5bcd08db LP |
799 | if (qgroupid == 0) { |
800 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
801 | if (r < 0) | |
802 | return r; | |
803 | } else { | |
804 | r = btrfs_is_filesystem(fd); | |
805 | if (r < 0) | |
806 | return r; | |
807 | if (!r) | |
808 | return -ENOTTY; | |
809 | } | |
62572894 | 810 | |
5bcd08db LP |
811 | args.qgroupid = qgroupid; |
812 | ||
813 | for (c = 0;; c++) { | |
814 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &args) < 0) { | |
815 | ||
816 | if (errno == EBUSY && c < 10) { | |
817 | (void) btrfs_quota_scan_wait(fd); | |
818 | continue; | |
819 | } | |
820 | ||
821 | return -errno; | |
822 | } | |
823 | ||
824 | break; | |
825 | } | |
d6ce17c7 LP |
826 | |
827 | return 0; | |
828 | } | |
829 | ||
5bcd08db LP |
830 | int btrfs_qgroup_set_limit(const char *path, uint64_t qgroupid, uint64_t referenced_max) { |
831 | _cleanup_close_ int fd = -1; | |
832 | ||
833 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
834 | if (fd < 0) | |
835 | return -errno; | |
836 | ||
837 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
838 | } | |
839 | ||
840 | int btrfs_subvol_set_subtree_quota_limit_fd(int fd, uint64_t subvol_id, uint64_t referenced_max) { | |
841 | uint64_t qgroupid; | |
842 | int r; | |
843 | ||
844 | assert(fd >= 0); | |
845 | ||
846 | r = btrfs_subvol_find_subtree_qgroup(fd, subvol_id, &qgroupid); | |
847 | if (r < 0) | |
848 | return r; | |
849 | ||
850 | return btrfs_qgroup_set_limit_fd(fd, qgroupid, referenced_max); | |
851 | } | |
852 | ||
853 | int btrfs_subvol_set_subtree_quota_limit(const char *path, uint64_t subvol_id, uint64_t referenced_max) { | |
d6ce17c7 LP |
854 | _cleanup_close_ int fd = -1; |
855 | ||
856 | fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW); | |
857 | if (fd < 0) | |
858 | return -errno; | |
859 | ||
5bcd08db | 860 | return btrfs_subvol_set_subtree_quota_limit_fd(fd, subvol_id, referenced_max); |
d6ce17c7 | 861 | } |
efe02862 | 862 | |
26166c88 | 863 | int btrfs_resize_loopback_fd(int fd, uint64_t new_size, bool grow_only) { |
efe02862 | 864 | struct btrfs_ioctl_vol_args args = {}; |
553e15f2 ZJS |
865 | char p[SYS_BLOCK_PATH_MAX("/loop/backing_file")]; |
866 | _cleanup_free_ char *backing = NULL; | |
efe02862 LP |
867 | _cleanup_close_ int loop_fd = -1, backing_fd = -1; |
868 | struct stat st; | |
a7f7d1bd | 869 | dev_t dev = 0; |
efe02862 LP |
870 | int r; |
871 | ||
a90fb858 LP |
872 | /* In contrast to btrfs quota ioctls ftruncate() cannot make sense of "infinity" or file sizes > 2^31 */ |
873 | if (!FILE_SIZE_VALID(new_size)) | |
874 | return -EINVAL; | |
875 | ||
efe02862 LP |
876 | /* btrfs cannot handle file systems < 16M, hence use this as minimum */ |
877 | if (new_size < 16*1024*1024) | |
878 | new_size = 16*1024*1024; | |
879 | ||
880 | r = btrfs_get_block_device_fd(fd, &dev); | |
881 | if (r < 0) | |
882 | return r; | |
883 | if (r == 0) | |
884 | return -ENODEV; | |
885 | ||
553e15f2 | 886 | xsprintf_sys_block_path(p, "/loop/backing_file", dev); |
efe02862 | 887 | r = read_one_line_file(p, &backing); |
26166c88 LP |
888 | if (r == -ENOENT) |
889 | return -ENODEV; | |
efe02862 LP |
890 | if (r < 0) |
891 | return r; | |
892 | if (isempty(backing) || !path_is_absolute(backing)) | |
893 | return -ENODEV; | |
894 | ||
895 | backing_fd = open(backing, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
896 | if (backing_fd < 0) | |
897 | return -errno; | |
898 | ||
899 | if (fstat(backing_fd, &st) < 0) | |
900 | return -errno; | |
901 | if (!S_ISREG(st.st_mode)) | |
902 | return -ENODEV; | |
903 | ||
904 | if (new_size == (uint64_t) st.st_size) | |
905 | return 0; | |
906 | ||
26166c88 LP |
907 | if (grow_only && new_size < (uint64_t) st.st_size) |
908 | return -EINVAL; | |
909 | ||
553e15f2 ZJS |
910 | xsprintf_sys_block_path(p, NULL, dev); |
911 | loop_fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY); | |
efe02862 LP |
912 | if (loop_fd < 0) |
913 | return -errno; | |
914 | ||
915 | if (snprintf(args.name, sizeof(args.name), "%" PRIu64, new_size) >= (int) sizeof(args.name)) | |
916 | return -EINVAL; | |
917 | ||
918 | if (new_size < (uint64_t) st.st_size) { | |
919 | /* Decrease size: first decrease btrfs size, then shorten loopback */ | |
920 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
921 | return -errno; | |
922 | } | |
923 | ||
924 | if (ftruncate(backing_fd, new_size) < 0) | |
925 | return -errno; | |
926 | ||
927 | if (ioctl(loop_fd, LOOP_SET_CAPACITY, 0) < 0) | |
928 | return -errno; | |
929 | ||
930 | if (new_size > (uint64_t) st.st_size) { | |
931 | /* Increase size: first enlarge loopback, then increase btrfs size */ | |
932 | if (ioctl(fd, BTRFS_IOC_RESIZE, &args) < 0) | |
933 | return -errno; | |
934 | } | |
935 | ||
26166c88 LP |
936 | /* Make sure the free disk space is correctly updated for both file systems */ |
937 | (void) fsync(fd); | |
938 | (void) fsync(backing_fd); | |
939 | ||
940 | return 1; | |
efe02862 LP |
941 | } |
942 | ||
26166c88 | 943 | int btrfs_resize_loopback(const char *p, uint64_t new_size, bool grow_only) { |
efe02862 LP |
944 | _cleanup_close_ int fd = -1; |
945 | ||
946 | fd = open(p, O_RDONLY|O_NOCTTY|O_CLOEXEC); | |
947 | if (fd < 0) | |
948 | return -errno; | |
949 | ||
26166c88 | 950 | return btrfs_resize_loopback_fd(fd, new_size, grow_only); |
efe02862 | 951 | } |
d9e2daaf | 952 | |
5bcd08db | 953 | int btrfs_qgroupid_make(uint64_t level, uint64_t id, uint64_t *ret) { |
3f952f92 LP |
954 | assert(ret); |
955 | ||
956 | if (level >= (UINT64_C(1) << (64 - BTRFS_QGROUP_LEVEL_SHIFT))) | |
957 | return -EINVAL; | |
958 | ||
959 | if (id >= (UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT)) | |
960 | return -EINVAL; | |
961 | ||
962 | *ret = (level << BTRFS_QGROUP_LEVEL_SHIFT) | id; | |
963 | return 0; | |
964 | } | |
965 | ||
5bcd08db LP |
966 | int btrfs_qgroupid_split(uint64_t qgroupid, uint64_t *level, uint64_t *id) { |
967 | assert(level || id); | |
968 | ||
969 | if (level) | |
970 | *level = qgroupid >> BTRFS_QGROUP_LEVEL_SHIFT; | |
971 | ||
972 | if (id) | |
973 | *id = qgroupid & ((UINT64_C(1) << BTRFS_QGROUP_LEVEL_SHIFT) - 1); | |
974 | ||
975 | return 0; | |
976 | } | |
977 | ||
978 | static int qgroup_create_or_destroy(int fd, bool b, uint64_t qgroupid) { | |
3f952f92 LP |
979 | |
980 | struct btrfs_ioctl_qgroup_create_args args = { | |
981 | .create = b, | |
5bcd08db | 982 | .qgroupid = qgroupid, |
3f952f92 | 983 | }; |
5bcd08db | 984 | unsigned c; |
3f952f92 LP |
985 | int r; |
986 | ||
5bcd08db | 987 | r = btrfs_is_filesystem(fd); |
3f952f92 LP |
988 | if (r < 0) |
989 | return r; | |
5bcd08db LP |
990 | if (r == 0) |
991 | return -ENOTTY; | |
992 | ||
993 | for (c = 0;; c++) { | |
994 | if (ioctl(fd, BTRFS_IOC_QGROUP_CREATE, &args) < 0) { | |
995 | ||
be6d467c LP |
996 | /* If quota is not enabled, we get EINVAL. Turn this into a recognizable error */ |
997 | if (errno == EINVAL) | |
998 | return -ENOPROTOOPT; | |
999 | ||
5bcd08db LP |
1000 | if (errno == EBUSY && c < 10) { |
1001 | (void) btrfs_quota_scan_wait(fd); | |
1002 | continue; | |
1003 | } | |
1004 | ||
1005 | return -errno; | |
1006 | } | |
3f952f92 | 1007 | |
5bcd08db LP |
1008 | break; |
1009 | } | |
1010 | ||
1011 | return 0; | |
1012 | } | |
1013 | ||
1014 | int btrfs_qgroup_create(int fd, uint64_t qgroupid) { | |
1015 | return qgroup_create_or_destroy(fd, true, qgroupid); | |
1016 | } | |
1017 | ||
1018 | int btrfs_qgroup_destroy(int fd, uint64_t qgroupid) { | |
1019 | return qgroup_create_or_destroy(fd, false, qgroupid); | |
1020 | } | |
1021 | ||
1022 | int btrfs_qgroup_destroy_recursive(int fd, uint64_t qgroupid) { | |
1023 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1024 | uint64_t subvol_id; | |
1025 | int i, n, r; | |
1026 | ||
1027 | /* Destroys the specified qgroup, but unassigns it from all | |
1028 | * its parents first. Also, it recursively destroys all | |
1029 | * qgroups it is assgined to that have the same id part of the | |
1030 | * qgroupid as the specified group. */ | |
1031 | ||
1032 | r = btrfs_qgroupid_split(qgroupid, NULL, &subvol_id); | |
1033 | if (r < 0) | |
1034 | return r; | |
1035 | ||
1036 | n = btrfs_qgroup_find_parents(fd, qgroupid, &qgroups); | |
1037 | if (n < 0) | |
1038 | return n; | |
1039 | ||
1040 | for (i = 0; i < n; i++) { | |
1041 | uint64_t id; | |
1042 | ||
1043 | r = btrfs_qgroupid_split(qgroups[i], NULL, &id); | |
1044 | if (r < 0) | |
1045 | return r; | |
1046 | ||
1047 | r = btrfs_qgroup_unassign(fd, qgroupid, qgroups[i]); | |
1048 | if (r < 0) | |
1049 | return r; | |
1050 | ||
1051 | if (id != subvol_id) | |
1052 | continue; | |
1053 | ||
1054 | /* The parent qgroupid shares the same id part with | |
1055 | * us? If so, destroy it too. */ | |
1056 | ||
1057 | (void) btrfs_qgroup_destroy_recursive(fd, qgroups[i]); | |
1058 | } | |
1059 | ||
1060 | return btrfs_qgroup_destroy(fd, qgroupid); | |
1061 | } | |
1062 | ||
1063 | int btrfs_quota_scan_start(int fd) { | |
1064 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1065 | ||
1066 | assert(fd >= 0); | |
1067 | ||
1068 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN, &args) < 0) | |
3f952f92 LP |
1069 | return -errno; |
1070 | ||
1071 | return 0; | |
1072 | } | |
1073 | ||
5bcd08db LP |
1074 | int btrfs_quota_scan_wait(int fd) { |
1075 | assert(fd >= 0); | |
1076 | ||
1077 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_WAIT) < 0) | |
1078 | return -errno; | |
1079 | ||
1080 | return 0; | |
3f952f92 LP |
1081 | } |
1082 | ||
5bcd08db LP |
1083 | int btrfs_quota_scan_ongoing(int fd) { |
1084 | struct btrfs_ioctl_quota_rescan_args args = {}; | |
1085 | ||
1086 | assert(fd >= 0); | |
1087 | ||
1088 | if (ioctl(fd, BTRFS_IOC_QUOTA_RESCAN_STATUS, &args) < 0) | |
1089 | return -errno; | |
1090 | ||
1091 | return !!args.flags; | |
3f952f92 LP |
1092 | } |
1093 | ||
5bcd08db LP |
1094 | static int qgroup_assign_or_unassign(int fd, bool b, uint64_t child, uint64_t parent) { |
1095 | struct btrfs_ioctl_qgroup_assign_args args = { | |
1096 | .assign = b, | |
1097 | .src = child, | |
1098 | .dst = parent, | |
1099 | }; | |
1100 | unsigned c; | |
1101 | int r; | |
1102 | ||
1103 | r = btrfs_is_filesystem(fd); | |
1104 | if (r < 0) | |
1105 | return r; | |
1106 | if (r == 0) | |
1107 | return -ENOTTY; | |
1108 | ||
1109 | for (c = 0;; c++) { | |
1110 | r = ioctl(fd, BTRFS_IOC_QGROUP_ASSIGN, &args); | |
1111 | if (r < 0) { | |
1112 | if (errno == EBUSY && c < 10) { | |
1113 | (void) btrfs_quota_scan_wait(fd); | |
1114 | continue; | |
1115 | } | |
1116 | ||
1117 | return -errno; | |
1118 | } | |
1119 | ||
1120 | if (r == 0) | |
1121 | return 0; | |
1122 | ||
1123 | /* If the return value is > 0, we need to request a rescan */ | |
1124 | ||
1125 | (void) btrfs_quota_scan_start(fd); | |
1126 | return 1; | |
1127 | } | |
1128 | } | |
1129 | ||
1130 | int btrfs_qgroup_assign(int fd, uint64_t child, uint64_t parent) { | |
1131 | return qgroup_assign_or_unassign(fd, true, child, parent); | |
1132 | } | |
1133 | ||
1134 | int btrfs_qgroup_unassign(int fd, uint64_t child, uint64_t parent) { | |
1135 | return qgroup_assign_or_unassign(fd, false, child, parent); | |
1136 | } | |
1137 | ||
1138 | static int subvol_remove_children(int fd, const char *subvolume, uint64_t subvol_id, BtrfsRemoveFlags flags) { | |
d9e2daaf LP |
1139 | struct btrfs_ioctl_search_args args = { |
1140 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1141 | ||
1142 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1143 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1144 | ||
1145 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1146 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1147 | ||
1148 | .key.min_transid = 0, | |
1149 | .key.max_transid = (uint64_t) -1, | |
1150 | }; | |
1151 | ||
1152 | struct btrfs_ioctl_vol_args vol_args = {}; | |
1153 | _cleanup_close_ int subvol_fd = -1; | |
62572894 | 1154 | struct stat st; |
3986b258 | 1155 | bool made_writable = false; |
d9e2daaf LP |
1156 | int r; |
1157 | ||
1158 | assert(fd >= 0); | |
1159 | assert(subvolume); | |
1160 | ||
62572894 LP |
1161 | if (fstat(fd, &st) < 0) |
1162 | return -errno; | |
1163 | ||
1164 | if (!S_ISDIR(st.st_mode)) | |
1165 | return -EINVAL; | |
1166 | ||
f7c9f4a2 | 1167 | subvol_fd = openat(fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1168 | if (subvol_fd < 0) |
1169 | return -errno; | |
1170 | ||
1171 | if (subvol_id == 0) { | |
1172 | r = btrfs_subvol_get_id_fd(subvol_fd, &subvol_id); | |
1173 | if (r < 0) | |
1174 | return r; | |
1175 | } | |
1176 | ||
3f952f92 LP |
1177 | /* First, try to remove the subvolume. If it happens to be |
1178 | * already empty, this will just work. */ | |
1179 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
1180 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) >= 0) { | |
5bcd08db | 1181 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); /* for the leaf subvolumes, the qgroup id is identical to the subvol id */ |
3f952f92 LP |
1182 | return 0; |
1183 | } | |
5bcd08db | 1184 | if (!(flags & BTRFS_REMOVE_RECURSIVE) || errno != ENOTEMPTY) |
3f952f92 LP |
1185 | return -errno; |
1186 | ||
1187 | /* OK, the subvolume is not empty, let's look for child | |
1188 | * subvolumes, and remove them, first */ | |
1189 | ||
d9e2daaf LP |
1190 | args.key.min_offset = args.key.max_offset = subvol_id; |
1191 | ||
1192 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1193 | const struct btrfs_ioctl_search_header *sh; | |
1194 | unsigned i; | |
1195 | ||
1196 | args.key.nr_items = 256; | |
1197 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1198 | return -errno; | |
1199 | ||
1200 | if (args.key.nr_items <= 0) | |
1201 | break; | |
1202 | ||
1203 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1204 | _cleanup_free_ char *p = NULL; | |
1205 | const struct btrfs_root_ref *ref; | |
1206 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1207 | ||
1208 | btrfs_ioctl_search_args_set(&args, sh); | |
1209 | ||
1210 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1211 | continue; | |
1212 | if (sh->offset != subvol_id) | |
1213 | continue; | |
1214 | ||
1215 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1216 | ||
1217 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); | |
1218 | if (!p) | |
1219 | return -ENOMEM; | |
1220 | ||
1221 | zero(ino_args); | |
1222 | ino_args.treeid = subvol_id; | |
cbf21ecc | 1223 | ino_args.objectid = htole64(ref->dirid); |
d9e2daaf LP |
1224 | |
1225 | if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1226 | return -errno; | |
1227 | ||
3986b258 LP |
1228 | if (!made_writable) { |
1229 | r = btrfs_subvol_set_read_only_fd(subvol_fd, false); | |
1230 | if (r < 0) | |
1231 | return r; | |
1232 | ||
1233 | made_writable = true; | |
1234 | } | |
1235 | ||
d9e2daaf LP |
1236 | if (isempty(ino_args.name)) |
1237 | /* Subvolume is in the top-level | |
1238 | * directory of the subvolume. */ | |
5bcd08db | 1239 | r = subvol_remove_children(subvol_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1240 | else { |
1241 | _cleanup_close_ int child_fd = -1; | |
1242 | ||
1243 | /* Subvolume is somewhere further down, | |
1244 | * hence we need to open the | |
1245 | * containing directory first */ | |
1246 | ||
f7c9f4a2 | 1247 | child_fd = openat(subvol_fd, ino_args.name, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
d9e2daaf LP |
1248 | if (child_fd < 0) |
1249 | return -errno; | |
1250 | ||
5bcd08db | 1251 | r = subvol_remove_children(child_fd, p, sh->objectid, flags); |
d9e2daaf LP |
1252 | } |
1253 | if (r < 0) | |
1254 | return r; | |
1255 | } | |
1256 | ||
1257 | /* Increase search key by one, to read the next item, if we can. */ | |
1258 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1259 | break; | |
1260 | } | |
1261 | ||
1262 | /* OK, the child subvolumes should all be gone now, let's try | |
1263 | * again to remove the subvolume */ | |
1264 | if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &vol_args) < 0) | |
1265 | return -errno; | |
1266 | ||
5bcd08db | 1267 | (void) btrfs_qgroup_destroy_recursive(fd, subvol_id); |
d9e2daaf LP |
1268 | return 0; |
1269 | } | |
1270 | ||
5bcd08db | 1271 | int btrfs_subvol_remove(const char *path, BtrfsRemoveFlags flags) { |
d9e2daaf LP |
1272 | _cleanup_close_ int fd = -1; |
1273 | const char *subvolume; | |
1274 | int r; | |
1275 | ||
1276 | assert(path); | |
1277 | ||
1278 | r = extract_subvolume_name(path, &subvolume); | |
1279 | if (r < 0) | |
1280 | return r; | |
1281 | ||
ef8becfa | 1282 | fd = open_parent(path, O_CLOEXEC, 0); |
d9e2daaf LP |
1283 | if (fd < 0) |
1284 | return fd; | |
1285 | ||
5bcd08db LP |
1286 | return subvol_remove_children(fd, subvolume, 0, flags); |
1287 | } | |
1288 | ||
1289 | int btrfs_subvol_remove_fd(int fd, const char *subvolume, BtrfsRemoveFlags flags) { | |
1290 | return subvol_remove_children(fd, subvolume, 0, flags); | |
1291 | } | |
1292 | ||
1293 | int btrfs_qgroup_copy_limits(int fd, uint64_t old_qgroupid, uint64_t new_qgroupid) { | |
1294 | ||
1295 | struct btrfs_ioctl_search_args args = { | |
1296 | /* Tree of quota items */ | |
1297 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1298 | ||
1299 | /* The object ID is always 0 */ | |
1300 | .key.min_objectid = 0, | |
1301 | .key.max_objectid = 0, | |
1302 | ||
1303 | /* Look precisely for the quota items */ | |
1304 | .key.min_type = BTRFS_QGROUP_LIMIT_KEY, | |
1305 | .key.max_type = BTRFS_QGROUP_LIMIT_KEY, | |
1306 | ||
1307 | /* For our qgroup */ | |
1308 | .key.min_offset = old_qgroupid, | |
1309 | .key.max_offset = old_qgroupid, | |
1310 | ||
1311 | /* No restrictions on the other components */ | |
1312 | .key.min_transid = 0, | |
1313 | .key.max_transid = (uint64_t) -1, | |
1314 | }; | |
1315 | ||
1316 | int r; | |
1317 | ||
1318 | r = btrfs_is_filesystem(fd); | |
1319 | if (r < 0) | |
1320 | return r; | |
1321 | if (!r) | |
1322 | return -ENOTTY; | |
1323 | ||
1324 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1325 | const struct btrfs_ioctl_search_header *sh; | |
1326 | unsigned i; | |
1327 | ||
1328 | args.key.nr_items = 256; | |
12ee6186 LP |
1329 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1330 | if (errno == ENOENT) /* quota tree missing: quota is not enabled, hence nothing to copy */ | |
1331 | break; | |
1332 | ||
5bcd08db | 1333 | return -errno; |
12ee6186 | 1334 | } |
5bcd08db LP |
1335 | |
1336 | if (args.key.nr_items <= 0) | |
1337 | break; | |
1338 | ||
1339 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1340 | const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); | |
1341 | struct btrfs_ioctl_qgroup_limit_args qargs; | |
1342 | unsigned c; | |
1343 | ||
1344 | /* Make sure we start the next search at least from this entry */ | |
1345 | btrfs_ioctl_search_args_set(&args, sh); | |
1346 | ||
1347 | if (sh->objectid != 0) | |
1348 | continue; | |
1349 | if (sh->type != BTRFS_QGROUP_LIMIT_KEY) | |
1350 | continue; | |
1351 | if (sh->offset != old_qgroupid) | |
1352 | continue; | |
1353 | ||
1354 | /* We found the entry, now copy things over. */ | |
1355 | ||
1356 | qargs = (struct btrfs_ioctl_qgroup_limit_args) { | |
1357 | .qgroupid = new_qgroupid, | |
1358 | ||
1359 | .lim.max_rfer = le64toh(qli->max_rfer), | |
1360 | .lim.max_excl = le64toh(qli->max_excl), | |
1361 | .lim.rsv_rfer = le64toh(qli->rsv_rfer), | |
1362 | .lim.rsv_excl = le64toh(qli->rsv_excl), | |
1363 | ||
1364 | .lim.flags = le64toh(qli->flags) & (BTRFS_QGROUP_LIMIT_MAX_RFER| | |
1365 | BTRFS_QGROUP_LIMIT_MAX_EXCL| | |
1366 | BTRFS_QGROUP_LIMIT_RSV_RFER| | |
1367 | BTRFS_QGROUP_LIMIT_RSV_EXCL), | |
1368 | }; | |
1369 | ||
1370 | for (c = 0;; c++) { | |
1371 | if (ioctl(fd, BTRFS_IOC_QGROUP_LIMIT, &qargs) < 0) { | |
1372 | if (errno == EBUSY && c < 10) { | |
1373 | (void) btrfs_quota_scan_wait(fd); | |
1374 | continue; | |
1375 | } | |
1376 | return -errno; | |
1377 | } | |
1378 | ||
1379 | break; | |
1380 | } | |
1381 | ||
1382 | return 1; | |
1383 | } | |
1384 | ||
1385 | /* Increase search key by one, to read the next item, if we can. */ | |
1386 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1387 | break; | |
1388 | } | |
1389 | ||
1390 | return 0; | |
d9e2daaf LP |
1391 | } |
1392 | ||
5bcd08db LP |
1393 | static int copy_quota_hierarchy(int fd, uint64_t old_subvol_id, uint64_t new_subvol_id) { |
1394 | _cleanup_free_ uint64_t *old_qgroups = NULL, *old_parent_qgroups = NULL; | |
1395 | bool copy_from_parent = false, insert_intermediary_qgroup = false; | |
1396 | int n_old_qgroups, n_old_parent_qgroups, r, i; | |
1397 | uint64_t old_parent_id; | |
1398 | ||
1399 | assert(fd >= 0); | |
1400 | ||
1401 | /* Copies a reduced form of quota information from the old to | |
1402 | * the new subvolume. */ | |
1403 | ||
1404 | n_old_qgroups = btrfs_qgroup_find_parents(fd, old_subvol_id, &old_qgroups); | |
1405 | if (n_old_qgroups <= 0) /* Nothing to copy */ | |
1406 | return n_old_qgroups; | |
1407 | ||
1408 | r = btrfs_subvol_get_parent(fd, old_subvol_id, &old_parent_id); | |
08c77cf3 LP |
1409 | if (r == -ENXIO) |
1410 | /* We have no parent, hence nothing to copy. */ | |
1411 | n_old_parent_qgroups = 0; | |
1412 | else if (r < 0) | |
5bcd08db | 1413 | return r; |
08c77cf3 LP |
1414 | else { |
1415 | n_old_parent_qgroups = btrfs_qgroup_find_parents(fd, old_parent_id, &old_parent_qgroups); | |
1416 | if (n_old_parent_qgroups < 0) | |
1417 | return n_old_parent_qgroups; | |
1418 | } | |
5bcd08db LP |
1419 | |
1420 | for (i = 0; i < n_old_qgroups; i++) { | |
1421 | uint64_t id; | |
1422 | int j; | |
1423 | ||
1424 | r = btrfs_qgroupid_split(old_qgroups[i], NULL, &id); | |
1425 | if (r < 0) | |
1426 | return r; | |
1427 | ||
1428 | if (id == old_subvol_id) { | |
1429 | /* The old subvolume was member of a qgroup | |
1430 | * that had the same id, but a different level | |
1431 | * as it self. Let's set up something similar | |
1432 | * in the destination. */ | |
1433 | insert_intermediary_qgroup = true; | |
1434 | break; | |
1435 | } | |
1436 | ||
1437 | for (j = 0; j < n_old_parent_qgroups; j++) | |
1438 | if (old_parent_qgroups[j] == old_qgroups[i]) { | |
1439 | /* The old subvolume shared a common | |
1440 | * parent qgroup with its parent | |
1441 | * subvolume. Let's set up something | |
1442 | * similar in the destination. */ | |
1443 | copy_from_parent = true; | |
1444 | } | |
1445 | } | |
1446 | ||
1447 | if (!insert_intermediary_qgroup && !copy_from_parent) | |
1448 | return 0; | |
1449 | ||
1450 | return btrfs_subvol_auto_qgroup_fd(fd, new_subvol_id, insert_intermediary_qgroup); | |
1451 | } | |
1452 | ||
1453 | static int copy_subtree_quota_limits(int fd, uint64_t old_subvol, uint64_t new_subvol) { | |
1454 | uint64_t old_subtree_qgroup, new_subtree_qgroup; | |
1455 | bool changed; | |
1456 | int r; | |
1457 | ||
1458 | /* First copy the leaf limits */ | |
1459 | r = btrfs_qgroup_copy_limits(fd, old_subvol, new_subvol); | |
1460 | if (r < 0) | |
1461 | return r; | |
1462 | changed = r > 0; | |
1463 | ||
1464 | /* Then, try to copy the subtree limits, if there are any. */ | |
1465 | r = btrfs_subvol_find_subtree_qgroup(fd, old_subvol, &old_subtree_qgroup); | |
1466 | if (r < 0) | |
1467 | return r; | |
1468 | if (r == 0) | |
1469 | return changed; | |
1470 | ||
1471 | r = btrfs_subvol_find_subtree_qgroup(fd, new_subvol, &new_subtree_qgroup); | |
1472 | if (r < 0) | |
1473 | return r; | |
1474 | if (r == 0) | |
1475 | return changed; | |
1476 | ||
1477 | r = btrfs_qgroup_copy_limits(fd, old_subtree_qgroup, new_subtree_qgroup); | |
1478 | if (r != 0) | |
1479 | return r; | |
1480 | ||
1481 | return changed; | |
d9e2daaf | 1482 | } |
f70a17f8 | 1483 | |
90578cbd | 1484 | static int subvol_snapshot_children(int old_fd, int new_fd, const char *subvolume, uint64_t old_subvol_id, BtrfsSnapshotFlags flags) { |
f70a17f8 LP |
1485 | |
1486 | struct btrfs_ioctl_search_args args = { | |
1487 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1488 | ||
1489 | .key.min_objectid = BTRFS_FIRST_FREE_OBJECTID, | |
1490 | .key.max_objectid = BTRFS_LAST_FREE_OBJECTID, | |
1491 | ||
1492 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1493 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1494 | ||
1495 | .key.min_transid = 0, | |
1496 | .key.max_transid = (uint64_t) -1, | |
1497 | }; | |
1498 | ||
1499 | struct btrfs_ioctl_vol_args_v2 vol_args = { | |
1500 | .flags = flags & BTRFS_SNAPSHOT_READ_ONLY ? BTRFS_SUBVOL_RDONLY : 0, | |
1501 | .fd = old_fd, | |
1502 | }; | |
ffb296b2 | 1503 | _cleanup_close_ int subvolume_fd = -1; |
90578cbd LP |
1504 | uint64_t new_subvol_id; |
1505 | int r; | |
f70a17f8 LP |
1506 | |
1507 | assert(old_fd >= 0); | |
1508 | assert(new_fd >= 0); | |
1509 | assert(subvolume); | |
1510 | ||
1511 | strncpy(vol_args.name, subvolume, sizeof(vol_args.name)-1); | |
f70a17f8 LP |
1512 | |
1513 | if (ioctl(new_fd, BTRFS_IOC_SNAP_CREATE_V2, &vol_args) < 0) | |
1514 | return -errno; | |
1515 | ||
5bcd08db LP |
1516 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE) && |
1517 | !(flags & BTRFS_SNAPSHOT_QUOTA)) | |
f70a17f8 LP |
1518 | return 0; |
1519 | ||
90578cbd LP |
1520 | if (old_subvol_id == 0) { |
1521 | r = btrfs_subvol_get_id_fd(old_fd, &old_subvol_id); | |
f70a17f8 LP |
1522 | if (r < 0) |
1523 | return r; | |
1524 | } | |
1525 | ||
90578cbd LP |
1526 | r = btrfs_subvol_get_id(new_fd, vol_args.name, &new_subvol_id); |
1527 | if (r < 0) | |
1528 | return r; | |
1529 | ||
5bcd08db LP |
1530 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1531 | (void) copy_quota_hierarchy(new_fd, old_subvol_id, new_subvol_id); | |
1532 | ||
1533 | if (!(flags & BTRFS_SNAPSHOT_RECURSIVE)) { | |
1534 | ||
1535 | if (flags & BTRFS_SNAPSHOT_QUOTA) | |
1536 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1537 | ||
1538 | return 0; | |
1539 | } | |
1540 | ||
90578cbd | 1541 | args.key.min_offset = args.key.max_offset = old_subvol_id; |
f70a17f8 LP |
1542 | |
1543 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1544 | const struct btrfs_ioctl_search_header *sh; | |
1545 | unsigned i; | |
1546 | ||
1547 | args.key.nr_items = 256; | |
1548 | if (ioctl(old_fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
1549 | return -errno; | |
1550 | ||
1551 | if (args.key.nr_items <= 0) | |
1552 | break; | |
1553 | ||
1554 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1555 | _cleanup_free_ char *p = NULL, *c = NULL, *np = NULL; | |
1556 | struct btrfs_ioctl_ino_lookup_args ino_args; | |
1557 | const struct btrfs_root_ref *ref; | |
1558 | _cleanup_close_ int old_child_fd = -1, new_child_fd = -1; | |
1559 | ||
1560 | btrfs_ioctl_search_args_set(&args, sh); | |
1561 | ||
1562 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
1563 | continue; | |
90578cbd LP |
1564 | |
1565 | /* Avoid finding the source subvolume a second | |
1566 | * time */ | |
1567 | if (sh->offset != old_subvol_id) | |
f70a17f8 LP |
1568 | continue; |
1569 | ||
90578cbd LP |
1570 | /* Avoid running into loops if the new |
1571 | * subvolume is below the old one. */ | |
1572 | if (sh->objectid == new_subvol_id) | |
1573 | continue; | |
f70a17f8 | 1574 | |
90578cbd | 1575 | ref = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh); |
f70a17f8 LP |
1576 | p = strndup((char*) ref + sizeof(struct btrfs_root_ref), le64toh(ref->name_len)); |
1577 | if (!p) | |
1578 | return -ENOMEM; | |
1579 | ||
1580 | zero(ino_args); | |
90578cbd | 1581 | ino_args.treeid = old_subvol_id; |
f70a17f8 LP |
1582 | ino_args.objectid = htole64(ref->dirid); |
1583 | ||
1584 | if (ioctl(old_fd, BTRFS_IOC_INO_LOOKUP, &ino_args) < 0) | |
1585 | return -errno; | |
1586 | ||
1587 | /* The kernel returns an empty name if the | |
1588 | * subvolume is in the top-level directory, | |
1589 | * and otherwise appends a slash, so that we | |
1590 | * can just concatenate easily here, without | |
1591 | * adding a slash. */ | |
1592 | c = strappend(ino_args.name, p); | |
1593 | if (!c) | |
1594 | return -ENOMEM; | |
1595 | ||
f7c9f4a2 | 1596 | old_child_fd = openat(old_fd, c, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1597 | if (old_child_fd < 0) |
1598 | return -errno; | |
1599 | ||
605405c6 | 1600 | np = strjoin(subvolume, "/", ino_args.name); |
f70a17f8 LP |
1601 | if (!np) |
1602 | return -ENOMEM; | |
1603 | ||
f7c9f4a2 | 1604 | new_child_fd = openat(new_fd, np, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
f70a17f8 LP |
1605 | if (new_child_fd < 0) |
1606 | return -errno; | |
1607 | ||
ffb296b2 LP |
1608 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { |
1609 | /* If the snapshot is read-only we | |
1610 | * need to mark it writable | |
1611 | * temporarily, to put the subsnapshot | |
1612 | * into place. */ | |
1613 | ||
1614 | if (subvolume_fd < 0) { | |
f7c9f4a2 | 1615 | subvolume_fd = openat(new_fd, subvolume, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW); |
ffb296b2 LP |
1616 | if (subvolume_fd < 0) |
1617 | return -errno; | |
1618 | } | |
1619 | ||
1620 | r = btrfs_subvol_set_read_only_fd(subvolume_fd, false); | |
1621 | if (r < 0) | |
1622 | return r; | |
1623 | } | |
1624 | ||
f70a17f8 | 1625 | /* When btrfs clones the subvolumes, child |
90578cbd | 1626 | * subvolumes appear as empty directories. Remove |
f70a17f8 LP |
1627 | * them, so that we can create a new snapshot |
1628 | * in their place */ | |
ffb296b2 LP |
1629 | if (unlinkat(new_child_fd, p, AT_REMOVEDIR) < 0) { |
1630 | int k = -errno; | |
1631 | ||
1632 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) | |
1633 | (void) btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1634 | ||
1635 | return k; | |
1636 | } | |
f70a17f8 LP |
1637 | |
1638 | r = subvol_snapshot_children(old_child_fd, new_child_fd, p, sh->objectid, flags & ~BTRFS_SNAPSHOT_FALLBACK_COPY); | |
ffb296b2 LP |
1639 | |
1640 | /* Restore the readonly flag */ | |
1641 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
1642 | int k; | |
1643 | ||
1644 | k = btrfs_subvol_set_read_only_fd(subvolume_fd, true); | |
1645 | if (r >= 0 && k < 0) | |
1646 | return k; | |
1647 | } | |
1648 | ||
f70a17f8 LP |
1649 | if (r < 0) |
1650 | return r; | |
1651 | } | |
1652 | ||
1653 | /* Increase search key by one, to read the next item, if we can. */ | |
1654 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1655 | break; | |
1656 | } | |
1657 | ||
5bcd08db LP |
1658 | if (flags & BTRFS_SNAPSHOT_QUOTA) |
1659 | (void) copy_subtree_quota_limits(new_fd, old_subvol_id, new_subvol_id); | |
1660 | ||
f70a17f8 LP |
1661 | return 0; |
1662 | } | |
1663 | ||
1664 | int btrfs_subvol_snapshot_fd(int old_fd, const char *new_path, BtrfsSnapshotFlags flags) { | |
1665 | _cleanup_close_ int new_fd = -1; | |
1666 | const char *subvolume; | |
1667 | int r; | |
1668 | ||
1669 | assert(old_fd >= 0); | |
1670 | assert(new_path); | |
1671 | ||
2904e949 | 1672 | r = btrfs_is_subvol_fd(old_fd); |
f70a17f8 LP |
1673 | if (r < 0) |
1674 | return r; | |
1675 | if (r == 0) { | |
17cbb288 LP |
1676 | bool plain_directory = false; |
1677 | ||
1678 | /* If the source isn't a proper subvolume, fail unless fallback is requested */ | |
f70a17f8 LP |
1679 | if (!(flags & BTRFS_SNAPSHOT_FALLBACK_COPY)) |
1680 | return -EISDIR; | |
1681 | ||
1682 | r = btrfs_subvol_make(new_path); | |
17cbb288 LP |
1683 | if (r == -ENOTTY && (flags & BTRFS_SNAPSHOT_FALLBACK_DIRECTORY)) { |
1684 | /* If the destination doesn't support subvolumes, then use a plain directory, if that's requested. */ | |
1685 | if (mkdir(new_path, 0755) < 0) | |
679def2a | 1686 | return -errno; |
17cbb288 LP |
1687 | |
1688 | plain_directory = true; | |
1689 | } else if (r < 0) | |
f70a17f8 LP |
1690 | return r; |
1691 | ||
1c876927 | 1692 | r = copy_directory_fd(old_fd, new_path, COPY_MERGE|COPY_REFLINK); |
17cbb288 LP |
1693 | if (r < 0) |
1694 | goto fallback_fail; | |
f70a17f8 LP |
1695 | |
1696 | if (flags & BTRFS_SNAPSHOT_READ_ONLY) { | |
17cbb288 LP |
1697 | |
1698 | if (plain_directory) { | |
1699 | /* Plain directories have no recursive read-only flag, but something pretty close to | |
1700 | * it: the IMMUTABLE bit. Let's use this here, if this is requested. */ | |
1701 | ||
1702 | if (flags & BTRFS_SNAPSHOT_FALLBACK_IMMUTABLE) | |
1703 | (void) chattr_path(new_path, FS_IMMUTABLE_FL, FS_IMMUTABLE_FL); | |
1704 | } else { | |
1705 | r = btrfs_subvol_set_read_only(new_path, true); | |
1706 | if (r < 0) | |
1707 | goto fallback_fail; | |
f70a17f8 LP |
1708 | } |
1709 | } | |
1710 | ||
1711 | return 0; | |
17cbb288 LP |
1712 | |
1713 | fallback_fail: | |
1714 | (void) rm_rf(new_path, REMOVE_ROOT|REMOVE_PHYSICAL|REMOVE_SUBVOLUME); | |
1715 | return r; | |
f70a17f8 LP |
1716 | } |
1717 | ||
1718 | r = extract_subvolume_name(new_path, &subvolume); | |
1719 | if (r < 0) | |
1720 | return r; | |
1721 | ||
ef8becfa | 1722 | new_fd = open_parent(new_path, O_CLOEXEC, 0); |
f70a17f8 LP |
1723 | if (new_fd < 0) |
1724 | return new_fd; | |
1725 | ||
1726 | return subvol_snapshot_children(old_fd, new_fd, subvolume, 0, flags); | |
1727 | } | |
1728 | ||
1729 | int btrfs_subvol_snapshot(const char *old_path, const char *new_path, BtrfsSnapshotFlags flags) { | |
1730 | _cleanup_close_ int old_fd = -1; | |
1731 | ||
1732 | assert(old_path); | |
1733 | assert(new_path); | |
1734 | ||
1735 | old_fd = open(old_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1736 | if (old_fd < 0) | |
1737 | return -errno; | |
1738 | ||
1739 | return btrfs_subvol_snapshot_fd(old_fd, new_path, flags); | |
1740 | } | |
5bcd08db LP |
1741 | |
1742 | int btrfs_qgroup_find_parents(int fd, uint64_t qgroupid, uint64_t **ret) { | |
1743 | ||
1744 | struct btrfs_ioctl_search_args args = { | |
1745 | /* Tree of quota items */ | |
1746 | .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID, | |
1747 | ||
1748 | /* Look precisely for the quota relation items */ | |
1749 | .key.min_type = BTRFS_QGROUP_RELATION_KEY, | |
1750 | .key.max_type = BTRFS_QGROUP_RELATION_KEY, | |
1751 | ||
1752 | /* No restrictions on the other components */ | |
1753 | .key.min_offset = 0, | |
1754 | .key.max_offset = (uint64_t) -1, | |
1755 | ||
1756 | .key.min_transid = 0, | |
1757 | .key.max_transid = (uint64_t) -1, | |
1758 | }; | |
1759 | ||
1760 | _cleanup_free_ uint64_t *items = NULL; | |
1761 | size_t n_items = 0, n_allocated = 0; | |
1762 | int r; | |
1763 | ||
1764 | assert(fd >= 0); | |
1765 | assert(ret); | |
1766 | ||
1767 | if (qgroupid == 0) { | |
1768 | r = btrfs_subvol_get_id_fd(fd, &qgroupid); | |
1769 | if (r < 0) | |
1770 | return r; | |
1771 | } else { | |
1772 | r = btrfs_is_filesystem(fd); | |
1773 | if (r < 0) | |
1774 | return r; | |
1775 | if (!r) | |
1776 | return -ENOTTY; | |
1777 | } | |
1778 | ||
1779 | args.key.min_objectid = args.key.max_objectid = qgroupid; | |
1780 | ||
1781 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
1782 | const struct btrfs_ioctl_search_header *sh; | |
1783 | unsigned i; | |
1784 | ||
1785 | args.key.nr_items = 256; | |
12ee6186 LP |
1786 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) { |
1787 | if (errno == ENOENT) /* quota tree missing: quota is disabled */ | |
1788 | break; | |
1789 | ||
5bcd08db | 1790 | return -errno; |
12ee6186 | 1791 | } |
5bcd08db LP |
1792 | |
1793 | if (args.key.nr_items <= 0) | |
1794 | break; | |
1795 | ||
1796 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
1797 | ||
1798 | /* Make sure we start the next search at least from this entry */ | |
1799 | btrfs_ioctl_search_args_set(&args, sh); | |
1800 | ||
1801 | if (sh->type != BTRFS_QGROUP_RELATION_KEY) | |
1802 | continue; | |
1803 | if (sh->offset < sh->objectid) | |
1804 | continue; | |
1805 | if (sh->objectid != qgroupid) | |
1806 | continue; | |
1807 | ||
1808 | if (!GREEDY_REALLOC(items, n_allocated, n_items+1)) | |
1809 | return -ENOMEM; | |
1810 | ||
1811 | items[n_items++] = sh->offset; | |
1812 | } | |
1813 | ||
1814 | /* Increase search key by one, to read the next item, if we can. */ | |
1815 | if (!btrfs_ioctl_search_args_inc(&args)) | |
1816 | break; | |
1817 | } | |
1818 | ||
1819 | if (n_items <= 0) { | |
1820 | *ret = NULL; | |
1821 | return 0; | |
1822 | } | |
1823 | ||
1cc6c93a | 1824 | *ret = TAKE_PTR(items); |
5bcd08db LP |
1825 | |
1826 | return (int) n_items; | |
1827 | } | |
1828 | ||
1829 | int btrfs_subvol_auto_qgroup_fd(int fd, uint64_t subvol_id, bool insert_intermediary_qgroup) { | |
1830 | _cleanup_free_ uint64_t *qgroups = NULL; | |
1831 | uint64_t parent_subvol; | |
1832 | bool changed = false; | |
1833 | int n = 0, r; | |
1834 | ||
1835 | assert(fd >= 0); | |
1836 | ||
1837 | /* | |
1838 | * Sets up the specified subvolume's qgroup automatically in | |
1839 | * one of two ways: | |
1840 | * | |
1841 | * If insert_intermediary_qgroup is false, the subvolume's | |
1842 | * leaf qgroup will be assigned to the same parent qgroups as | |
1843 | * the subvolume's parent subvolume. | |
1844 | * | |
1845 | * If insert_intermediary_qgroup is true a new intermediary | |
1846 | * higher-level qgroup is created, with a higher level number, | |
1847 | * but reusing the id of the subvolume. The level number is | |
1848 | * picked as one smaller than the lowest level qgroup the | |
1849 | * parent subvolume is a member of. If the parent subvolume's | |
1850 | * leaf qgroup is assigned to no higher-level qgroup a new | |
1851 | * qgroup of level 255 is created instead. Either way, the new | |
1852 | * qgroup is then assigned to the parent's higher-level | |
1853 | * qgroup, and the subvolume itself is assigned to it. | |
1854 | * | |
1855 | * If the subvolume is already assigned to a higher level | |
1856 | * qgroup, no operation is executed. | |
1857 | * | |
1858 | * Effectively this means: regardless if | |
1859 | * insert_intermediary_qgroup is true or not, after this | |
1860 | * function is invoked the subvolume will be accounted within | |
1861 | * the same qgroups as the parent. However, if it is true, it | |
1862 | * will also get its own higher-level qgroup, which may in | |
1863 | * turn be used by subvolumes created beneath this subvolume | |
1864 | * later on. | |
1865 | * | |
1866 | * This hence defines a simple default qgroup setup for | |
1867 | * subvolumes, as long as this function is invoked on each | |
1868 | * created subvolume: each subvolume is always accounting | |
1869 | * together with its immediate parents. Optionally, if | |
1870 | * insert_intermediary_qgroup is true, it will also get a | |
1871 | * qgroup that then includes all its own child subvolumes. | |
1872 | */ | |
1873 | ||
1874 | if (subvol_id == 0) { | |
2904e949 | 1875 | r = btrfs_is_subvol_fd(fd); |
5bcd08db LP |
1876 | if (r < 0) |
1877 | return r; | |
1878 | if (!r) | |
1879 | return -ENOTTY; | |
1880 | ||
1881 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
1882 | if (r < 0) | |
1883 | return r; | |
1884 | } | |
1885 | ||
1886 | n = btrfs_qgroup_find_parents(fd, subvol_id, &qgroups); | |
1887 | if (n < 0) | |
1888 | return n; | |
1889 | if (n > 0) /* already parent qgroups set up, let's bail */ | |
1890 | return 0; | |
1891 | ||
08c77cf3 LP |
1892 | qgroups = mfree(qgroups); |
1893 | ||
5bcd08db | 1894 | r = btrfs_subvol_get_parent(fd, subvol_id, &parent_subvol); |
08c77cf3 LP |
1895 | if (r == -ENXIO) |
1896 | /* No parent, hence no qgroup memberships */ | |
1897 | n = 0; | |
1898 | else if (r < 0) | |
5bcd08db | 1899 | return r; |
08c77cf3 LP |
1900 | else { |
1901 | n = btrfs_qgroup_find_parents(fd, parent_subvol, &qgroups); | |
1902 | if (n < 0) | |
1903 | return n; | |
1904 | } | |
5bcd08db LP |
1905 | |
1906 | if (insert_intermediary_qgroup) { | |
1907 | uint64_t lowest = 256, new_qgroupid; | |
1908 | bool created = false; | |
1909 | int i; | |
1910 | ||
1911 | /* Determine the lowest qgroup that the parent | |
1912 | * subvolume is assigned to. */ | |
1913 | ||
1914 | for (i = 0; i < n; i++) { | |
1915 | uint64_t level; | |
1916 | ||
1917 | r = btrfs_qgroupid_split(qgroups[i], &level, NULL); | |
1918 | if (r < 0) | |
1919 | return r; | |
1920 | ||
1921 | if (level < lowest) | |
1922 | lowest = level; | |
1923 | } | |
1924 | ||
1925 | if (lowest <= 1) /* There are no levels left we could use insert an intermediary qgroup at */ | |
1926 | return -EBUSY; | |
1927 | ||
1928 | r = btrfs_qgroupid_make(lowest - 1, subvol_id, &new_qgroupid); | |
1929 | if (r < 0) | |
1930 | return r; | |
1931 | ||
1932 | /* Create the new intermediary group, unless it already exists */ | |
1933 | r = btrfs_qgroup_create(fd, new_qgroupid); | |
1934 | if (r < 0 && r != -EEXIST) | |
1935 | return r; | |
1936 | if (r >= 0) | |
1937 | changed = created = true; | |
1938 | ||
1939 | for (i = 0; i < n; i++) { | |
1940 | r = btrfs_qgroup_assign(fd, new_qgroupid, qgroups[i]); | |
1941 | if (r < 0 && r != -EEXIST) { | |
1942 | if (created) | |
1943 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
1944 | ||
1945 | return r; | |
1946 | } | |
1947 | if (r >= 0) | |
1948 | changed = true; | |
1949 | } | |
1950 | ||
1951 | r = btrfs_qgroup_assign(fd, subvol_id, new_qgroupid); | |
1952 | if (r < 0 && r != -EEXIST) { | |
1953 | if (created) | |
1954 | (void) btrfs_qgroup_destroy_recursive(fd, new_qgroupid); | |
1955 | return r; | |
1956 | } | |
1957 | if (r >= 0) | |
1958 | changed = true; | |
1959 | ||
1960 | } else { | |
1961 | int i; | |
1962 | ||
1963 | /* Assign our subvolume to all the same qgroups as the parent */ | |
1964 | ||
1965 | for (i = 0; i < n; i++) { | |
1966 | r = btrfs_qgroup_assign(fd, subvol_id, qgroups[i]); | |
1967 | if (r < 0 && r != -EEXIST) | |
1968 | return r; | |
1969 | if (r >= 0) | |
1970 | changed = true; | |
1971 | } | |
1972 | } | |
1973 | ||
1974 | return changed; | |
1975 | } | |
1976 | ||
1977 | int btrfs_subvol_auto_qgroup(const char *path, uint64_t subvol_id, bool create_intermediary_qgroup) { | |
1978 | _cleanup_close_ int fd = -1; | |
1979 | ||
1980 | fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY); | |
1981 | if (fd < 0) | |
1982 | return -errno; | |
1983 | ||
1984 | return btrfs_subvol_auto_qgroup_fd(fd, subvol_id, create_intermediary_qgroup); | |
1985 | } | |
1986 | ||
1987 | int btrfs_subvol_get_parent(int fd, uint64_t subvol_id, uint64_t *ret) { | |
1988 | ||
1989 | struct btrfs_ioctl_search_args args = { | |
1990 | /* Tree of tree roots */ | |
1991 | .key.tree_id = BTRFS_ROOT_TREE_OBJECTID, | |
1992 | ||
1993 | /* Look precisely for the subvolume items */ | |
1994 | .key.min_type = BTRFS_ROOT_BACKREF_KEY, | |
1995 | .key.max_type = BTRFS_ROOT_BACKREF_KEY, | |
1996 | ||
1997 | /* No restrictions on the other components */ | |
1998 | .key.min_offset = 0, | |
1999 | .key.max_offset = (uint64_t) -1, | |
2000 | ||
2001 | .key.min_transid = 0, | |
2002 | .key.max_transid = (uint64_t) -1, | |
2003 | }; | |
2004 | int r; | |
2005 | ||
2006 | assert(fd >= 0); | |
2007 | assert(ret); | |
2008 | ||
2009 | if (subvol_id == 0) { | |
2010 | r = btrfs_subvol_get_id_fd(fd, &subvol_id); | |
2011 | if (r < 0) | |
2012 | return r; | |
2013 | } else { | |
2014 | r = btrfs_is_filesystem(fd); | |
2015 | if (r < 0) | |
2016 | return r; | |
2017 | if (!r) | |
2018 | return -ENOTTY; | |
2019 | } | |
2020 | ||
2021 | args.key.min_objectid = args.key.max_objectid = subvol_id; | |
2022 | ||
2023 | while (btrfs_ioctl_search_args_compare(&args) <= 0) { | |
2024 | const struct btrfs_ioctl_search_header *sh; | |
2025 | unsigned i; | |
2026 | ||
2027 | args.key.nr_items = 256; | |
2028 | if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0) | |
9c4615fb | 2029 | return negative_errno(); |
5bcd08db LP |
2030 | |
2031 | if (args.key.nr_items <= 0) | |
2032 | break; | |
2033 | ||
2034 | FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) { | |
2035 | ||
2036 | if (sh->type != BTRFS_ROOT_BACKREF_KEY) | |
2037 | continue; | |
2038 | if (sh->objectid != subvol_id) | |
2039 | continue; | |
2040 | ||
2041 | *ret = sh->offset; | |
2042 | return 0; | |
2043 | } | |
2044 | } | |
2045 | ||
2046 | return -ENXIO; | |
2047 | } |