Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
e20d96d6 | 6 | #include <linux/fs.h> |
d98237b3 | 7 | #include <linux/blkdev.h> |
fc7cbcd4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
0b86a832 | 26 | #include "volumes.h" |
db94535d | 27 | #include "print-tree.h" |
925baedd | 28 | #include "locking.h" |
e02119d5 | 29 | #include "tree-log.h" |
fa9c0d79 | 30 | #include "free-space-cache.h" |
70f6d82e | 31 | #include "free-space-tree.h" |
21adbd5c | 32 | #include "check-integrity.h" |
606686ee | 33 | #include "rcu-string.h" |
8dabb742 | 34 | #include "dev-replace.h" |
53b381b3 | 35 | #include "raid56.h" |
5ac1d209 | 36 | #include "sysfs.h" |
fcebe456 | 37 | #include "qgroup.h" |
ebb8765b | 38 | #include "compression.h" |
557ea5dd | 39 | #include "tree-checker.h" |
fd708b81 | 40 | #include "ref-verify.h" |
aac0023c | 41 | #include "block-group.h" |
b0643e59 | 42 | #include "discard.h" |
f603bb94 | 43 | #include "space-info.h" |
b70f5097 | 44 | #include "zoned.h" |
139e8cd3 | 45 | #include "subpage.h" |
eb60ceac | 46 | |
319e4d06 QW |
47 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
48 | BTRFS_HEADER_FLAG_RELOC |\ | |
49 | BTRFS_SUPER_FLAG_ERROR |\ | |
50 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
51 | BTRFS_SUPER_FLAG_METADUMP |\ |
52 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 53 | |
143bede5 | 54 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 56 | struct btrfs_fs_info *fs_info); |
143bede5 | 57 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 58 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 59 | struct extent_io_tree *dirty_pages, |
60 | int mark); | |
2ff7e61e | 61 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 62 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
63 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
64 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 65 | |
141386e1 JB |
66 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
67 | { | |
68 | if (fs_info->csum_shash) | |
69 | crypto_free_shash(fs_info->csum_shash); | |
70 | } | |
71 | ||
d352ac68 CM |
72 | /* |
73 | * async submit bios are used to offload expensive checksumming | |
74 | * onto the worker threads. They checksum file and metadata bios | |
75 | * just before they are sent down the IO stack. | |
76 | */ | |
44b8bd7e | 77 | struct async_submit_bio { |
8896a08d | 78 | struct inode *inode; |
44b8bd7e | 79 | struct bio *bio; |
a758781d | 80 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 81 | int mirror_num; |
1941b64b QW |
82 | |
83 | /* Optional parameter for submit_bio_start used by direct io */ | |
84 | u64 dio_file_offset; | |
8b712842 | 85 | struct btrfs_work work; |
4e4cbee9 | 86 | blk_status_t status; |
44b8bd7e CM |
87 | }; |
88 | ||
d352ac68 | 89 | /* |
2996e1f8 | 90 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 91 | */ |
c67b3892 | 92 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 93 | { |
d5178578 | 94 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 95 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 96 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 97 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 98 | char *kaddr; |
e9be5a30 | 99 | int i; |
d5178578 JT |
100 | |
101 | shash->tfm = fs_info->csum_shash; | |
102 | crypto_shash_init(shash); | |
a26663e7 | 103 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 104 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 105 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 106 | |
e9be5a30 DS |
107 | for (i = 1; i < num_pages; i++) { |
108 | kaddr = page_address(buf->pages[i]); | |
109 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 110 | } |
71a63551 | 111 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 112 | crypto_shash_final(shash, result); |
19c00ddc CM |
113 | } |
114 | ||
d352ac68 CM |
115 | /* |
116 | * we can't consider a given block up to date unless the transid of the | |
117 | * block matches the transid in the parent node's pointer. This is how we | |
118 | * detect blocks that either didn't get written at all or got written | |
119 | * in the wrong place. | |
120 | */ | |
1259ab75 | 121 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
122 | struct extent_buffer *eb, u64 parent_transid, |
123 | int atomic) | |
1259ab75 | 124 | { |
2ac55d41 | 125 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
126 | int ret; |
127 | ||
128 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
129 | return 0; | |
130 | ||
b9fab919 CM |
131 | if (atomic) |
132 | return -EAGAIN; | |
133 | ||
570eb97b | 134 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); |
0b32f4bb | 135 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
136 | btrfs_header_generation(eb) == parent_transid) { |
137 | ret = 0; | |
138 | goto out; | |
139 | } | |
94647322 | 140 | btrfs_err_rl(eb->fs_info, |
8f0ed7d4 QW |
141 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
142 | eb->start, eb->read_mirror, | |
29549aec | 143 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 144 | ret = 1; |
35b22c19 | 145 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 146 | out: |
570eb97b JB |
147 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
148 | &cached_state); | |
1259ab75 | 149 | return ret; |
1259ab75 CM |
150 | } |
151 | ||
e7e16f48 JT |
152 | static bool btrfs_supported_super_csum(u16 csum_type) |
153 | { | |
154 | switch (csum_type) { | |
155 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 156 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 157 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 158 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
159 | return true; |
160 | default: | |
161 | return false; | |
162 | } | |
163 | } | |
164 | ||
1104a885 DS |
165 | /* |
166 | * Return 0 if the superblock checksum type matches the checksum value of that | |
167 | * algorithm. Pass the raw disk superblock data. | |
168 | */ | |
3d17adea QW |
169 | int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
170 | const struct btrfs_super_block *disk_sb) | |
1104a885 | 171 | { |
51bce6c9 | 172 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
173 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
174 | ||
175 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 176 | |
51bce6c9 JT |
177 | /* |
178 | * The super_block structure does not span the whole | |
179 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
180 | * filled with zeros and is included in the checksum. | |
181 | */ | |
3d17adea | 182 | crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, |
fd08001f | 183 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); |
1104a885 | 184 | |
55fc29be | 185 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 186 | return 1; |
1104a885 | 187 | |
e7e16f48 | 188 | return 0; |
1104a885 DS |
189 | } |
190 | ||
e064d5e9 | 191 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 192 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 193 | { |
e064d5e9 | 194 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
195 | int found_level; |
196 | struct btrfs_key found_key; | |
197 | int ret; | |
198 | ||
199 | found_level = btrfs_header_level(eb); | |
200 | if (found_level != level) { | |
63489055 QW |
201 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
202 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
203 | btrfs_err(fs_info, |
204 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
205 | eb->start, level, found_level); | |
581c1760 QW |
206 | return -EIO; |
207 | } | |
208 | ||
209 | if (!first_key) | |
210 | return 0; | |
211 | ||
5d41be6f QW |
212 | /* |
213 | * For live tree block (new tree blocks in current transaction), | |
214 | * we need proper lock context to avoid race, which is impossible here. | |
215 | * So we only checks tree blocks which is read from disk, whose | |
216 | * generation <= fs_info->last_trans_committed. | |
217 | */ | |
218 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
219 | return 0; | |
62fdaa52 QW |
220 | |
221 | /* We have @first_key, so this @eb must have at least one item */ | |
222 | if (btrfs_header_nritems(eb) == 0) { | |
223 | btrfs_err(fs_info, | |
224 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
225 | eb->start); | |
226 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
227 | return -EUCLEAN; | |
228 | } | |
229 | ||
581c1760 QW |
230 | if (found_level) |
231 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
232 | else | |
233 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
234 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
235 | ||
581c1760 | 236 | if (ret) { |
63489055 QW |
237 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
238 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 239 | btrfs_err(fs_info, |
ff76a864 LB |
240 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
241 | eb->start, parent_transid, first_key->objectid, | |
242 | first_key->type, first_key->offset, | |
243 | found_key.objectid, found_key.type, | |
244 | found_key.offset); | |
581c1760 | 245 | } |
581c1760 QW |
246 | return ret; |
247 | } | |
248 | ||
d352ac68 CM |
249 | /* |
250 | * helper to read a given tree block, doing retries as required when | |
251 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
252 | * |
253 | * @parent_transid: expected transid, skip check if 0 | |
254 | * @level: expected level, mandatory check | |
255 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 256 | */ |
6a2e9dc4 FM |
257 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
258 | u64 parent_transid, int level, | |
259 | struct btrfs_key *first_key) | |
f188591e | 260 | { |
5ab12d1f | 261 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 262 | struct extent_io_tree *io_tree; |
ea466794 | 263 | int failed = 0; |
f188591e CM |
264 | int ret; |
265 | int num_copies = 0; | |
266 | int mirror_num = 0; | |
ea466794 | 267 | int failed_mirror = 0; |
f188591e | 268 | |
0b246afa | 269 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 270 | while (1) { |
f8397d69 | 271 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 272 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 273 | if (!ret) { |
581c1760 | 274 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 275 | parent_transid, 0)) |
256dd1bb | 276 | ret = -EIO; |
e064d5e9 | 277 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 278 | first_key, parent_transid)) |
581c1760 QW |
279 | ret = -EUCLEAN; |
280 | else | |
281 | break; | |
256dd1bb | 282 | } |
d397712b | 283 | |
0b246afa | 284 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 285 | eb->start, eb->len); |
4235298e | 286 | if (num_copies == 1) |
ea466794 | 287 | break; |
4235298e | 288 | |
5cf1ab56 JB |
289 | if (!failed_mirror) { |
290 | failed = 1; | |
291 | failed_mirror = eb->read_mirror; | |
292 | } | |
293 | ||
f188591e | 294 | mirror_num++; |
ea466794 JB |
295 | if (mirror_num == failed_mirror) |
296 | mirror_num++; | |
297 | ||
4235298e | 298 | if (mirror_num > num_copies) |
ea466794 | 299 | break; |
f188591e | 300 | } |
ea466794 | 301 | |
c0901581 | 302 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 303 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
304 | |
305 | return ret; | |
f188591e | 306 | } |
19c00ddc | 307 | |
eca0f6f6 QW |
308 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
309 | { | |
310 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
311 | u8 result[BTRFS_CSUM_SIZE]; | |
312 | int ret; | |
313 | ||
314 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
315 | offsetof(struct btrfs_header, fsid), | |
316 | BTRFS_FSID_SIZE) == 0); | |
317 | csum_tree_block(eb, result); | |
318 | ||
319 | if (btrfs_header_level(eb)) | |
320 | ret = btrfs_check_node(eb); | |
321 | else | |
322 | ret = btrfs_check_leaf_full(eb); | |
323 | ||
3777369f QW |
324 | if (ret < 0) |
325 | goto error; | |
326 | ||
327 | /* | |
328 | * Also check the generation, the eb reached here must be newer than | |
329 | * last committed. Or something seriously wrong happened. | |
330 | */ | |
331 | if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { | |
332 | ret = -EUCLEAN; | |
eca0f6f6 | 333 | btrfs_err(fs_info, |
3777369f QW |
334 | "block=%llu bad generation, have %llu expect > %llu", |
335 | eb->start, btrfs_header_generation(eb), | |
336 | fs_info->last_trans_committed); | |
337 | goto error; | |
eca0f6f6 QW |
338 | } |
339 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
340 | ||
341 | return 0; | |
3777369f QW |
342 | |
343 | error: | |
344 | btrfs_print_tree(eb, 0); | |
345 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
346 | eb->start); | |
347 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
348 | return ret; | |
eca0f6f6 QW |
349 | } |
350 | ||
351 | /* Checksum all dirty extent buffers in one bio_vec */ | |
352 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
353 | struct bio_vec *bvec) | |
354 | { | |
355 | struct page *page = bvec->bv_page; | |
356 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
357 | u64 cur; | |
358 | int ret = 0; | |
359 | ||
360 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
361 | cur += fs_info->nodesize) { | |
362 | struct extent_buffer *eb; | |
363 | bool uptodate; | |
364 | ||
365 | eb = find_extent_buffer(fs_info, cur); | |
366 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
367 | fs_info->nodesize); | |
368 | ||
01cd3909 | 369 | /* A dirty eb shouldn't disappear from buffer_radix */ |
eca0f6f6 QW |
370 | if (WARN_ON(!eb)) |
371 | return -EUCLEAN; | |
372 | ||
373 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
374 | free_extent_buffer(eb); | |
375 | return -EUCLEAN; | |
376 | } | |
377 | if (WARN_ON(!uptodate)) { | |
378 | free_extent_buffer(eb); | |
379 | return -EUCLEAN; | |
380 | } | |
381 | ||
382 | ret = csum_one_extent_buffer(eb); | |
383 | free_extent_buffer(eb); | |
384 | if (ret < 0) | |
385 | return ret; | |
386 | } | |
387 | return ret; | |
388 | } | |
389 | ||
d352ac68 | 390 | /* |
ac303b69 QW |
391 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
392 | * we only fill in the checksum field in the first page of a multi-page block. | |
393 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 394 | */ |
ac303b69 | 395 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 396 | { |
ac303b69 | 397 | struct page *page = bvec->bv_page; |
4eee4fa4 | 398 | u64 start = page_offset(page); |
19c00ddc | 399 | u64 found_start; |
19c00ddc | 400 | struct extent_buffer *eb; |
eca0f6f6 | 401 | |
fbca46eb | 402 | if (fs_info->nodesize < PAGE_SIZE) |
eca0f6f6 | 403 | return csum_dirty_subpage_buffers(fs_info, bvec); |
f188591e | 404 | |
4f2de97a JB |
405 | eb = (struct extent_buffer *)page->private; |
406 | if (page != eb->pages[0]) | |
407 | return 0; | |
0f805531 | 408 | |
19c00ddc | 409 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
410 | |
411 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
412 | WARN_ON(found_start != 0); | |
413 | return 0; | |
414 | } | |
415 | ||
0f805531 AL |
416 | /* |
417 | * Please do not consolidate these warnings into a single if. | |
418 | * It is useful to know what went wrong. | |
419 | */ | |
420 | if (WARN_ON(found_start != start)) | |
421 | return -EUCLEAN; | |
422 | if (WARN_ON(!PageUptodate(page))) | |
423 | return -EUCLEAN; | |
424 | ||
eca0f6f6 | 425 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
426 | } |
427 | ||
b0c9b3b0 | 428 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 429 | { |
b0c9b3b0 | 430 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 431 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 432 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 433 | u8 *metadata_uuid; |
2b82032c | 434 | |
9a8658e3 DS |
435 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
436 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
437 | /* |
438 | * Checking the incompat flag is only valid for the current fs. For | |
439 | * seed devices it's forbidden to have their uuid changed so reading | |
440 | * ->fsid in this case is fine | |
441 | */ | |
442 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
443 | metadata_uuid = fs_devices->metadata_uuid; | |
444 | else | |
445 | metadata_uuid = fs_devices->fsid; | |
446 | ||
447 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
448 | return 0; | |
449 | ||
450 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
451 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
452 | return 0; | |
453 | ||
454 | return 1; | |
2b82032c YZ |
455 | } |
456 | ||
77bf40a2 QW |
457 | /* Do basic extent buffer checks at read time */ |
458 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 459 | { |
77bf40a2 | 460 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 461 | u64 found_start; |
77bf40a2 QW |
462 | const u32 csum_size = fs_info->csum_size; |
463 | u8 found_level; | |
2996e1f8 | 464 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 465 | const u8 *header_csum; |
77bf40a2 | 466 | int ret = 0; |
ea466794 | 467 | |
ce9adaa5 | 468 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 469 | if (found_start != eb->start) { |
8f0ed7d4 QW |
470 | btrfs_err_rl(fs_info, |
471 | "bad tree block start, mirror %u want %llu have %llu", | |
472 | eb->read_mirror, eb->start, found_start); | |
f188591e | 473 | ret = -EIO; |
77bf40a2 | 474 | goto out; |
ce9adaa5 | 475 | } |
b0c9b3b0 | 476 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
477 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
478 | eb->start, eb->read_mirror); | |
1259ab75 | 479 | ret = -EIO; |
77bf40a2 | 480 | goto out; |
1259ab75 | 481 | } |
ce9adaa5 | 482 | found_level = btrfs_header_level(eb); |
1c24c3ce | 483 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
484 | btrfs_err(fs_info, |
485 | "bad tree block level, mirror %u level %d on logical %llu", | |
486 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 487 | ret = -EIO; |
77bf40a2 | 488 | goto out; |
1c24c3ce | 489 | } |
ce9adaa5 | 490 | |
c67b3892 | 491 | csum_tree_block(eb, result); |
dfd29eed DS |
492 | header_csum = page_address(eb->pages[0]) + |
493 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 494 | |
dfd29eed | 495 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 496 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
497 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
498 | eb->start, eb->read_mirror, | |
dfd29eed | 499 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
500 | CSUM_FMT_VALUE(csum_size, result), |
501 | btrfs_header_level(eb)); | |
2996e1f8 | 502 | ret = -EUCLEAN; |
77bf40a2 | 503 | goto out; |
2996e1f8 JT |
504 | } |
505 | ||
a826d6dc JB |
506 | /* |
507 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
508 | * that we don't try and read the other copies of this block, just | |
509 | * return -EIO. | |
510 | */ | |
1c4360ee | 511 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
512 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
513 | ret = -EIO; | |
514 | } | |
ce9adaa5 | 515 | |
813fd1dc | 516 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
517 | ret = -EIO; |
518 | ||
0b32f4bb JB |
519 | if (!ret) |
520 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
521 | else |
522 | btrfs_err(fs_info, | |
8f0ed7d4 QW |
523 | "read time tree block corruption detected on logical %llu mirror %u", |
524 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
525 | out: |
526 | return ret; | |
527 | } | |
528 | ||
371cdc07 QW |
529 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
530 | int mirror) | |
531 | { | |
532 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
533 | struct extent_buffer *eb; | |
534 | bool reads_done; | |
535 | int ret = 0; | |
536 | ||
537 | /* | |
538 | * We don't allow bio merge for subpage metadata read, so we should | |
539 | * only get one eb for each endio hook. | |
540 | */ | |
541 | ASSERT(end == start + fs_info->nodesize - 1); | |
542 | ASSERT(PagePrivate(page)); | |
543 | ||
544 | eb = find_extent_buffer(fs_info, start); | |
545 | /* | |
546 | * When we are reading one tree block, eb must have been inserted into | |
547 | * the radix tree. If not, something is wrong. | |
548 | */ | |
549 | ASSERT(eb); | |
550 | ||
551 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
552 | /* Subpage read must finish in page read */ | |
553 | ASSERT(reads_done); | |
554 | ||
555 | eb->read_mirror = mirror; | |
556 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
557 | ret = -EIO; | |
558 | goto err; | |
559 | } | |
560 | ret = validate_extent_buffer(eb); | |
561 | if (ret < 0) | |
562 | goto err; | |
563 | ||
371cdc07 QW |
564 | set_extent_buffer_uptodate(eb); |
565 | ||
566 | free_extent_buffer(eb); | |
567 | return ret; | |
568 | err: | |
569 | /* | |
570 | * end_bio_extent_readpage decrements io_pages in case of error, | |
571 | * make sure it has something to decrement. | |
572 | */ | |
573 | atomic_inc(&eb->io_pages); | |
574 | clear_extent_buffer_uptodate(eb); | |
575 | free_extent_buffer(eb); | |
576 | return ret; | |
577 | } | |
578 | ||
c3a3b19b | 579 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
580 | struct page *page, u64 start, u64 end, |
581 | int mirror) | |
582 | { | |
583 | struct extent_buffer *eb; | |
584 | int ret = 0; | |
585 | int reads_done; | |
586 | ||
587 | ASSERT(page->private); | |
371cdc07 | 588 | |
fbca46eb | 589 | if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) |
371cdc07 QW |
590 | return validate_subpage_buffer(page, start, end, mirror); |
591 | ||
77bf40a2 QW |
592 | eb = (struct extent_buffer *)page->private; |
593 | ||
594 | /* | |
595 | * The pending IO might have been the only thing that kept this buffer | |
596 | * in memory. Make sure we have a ref for all this other checks | |
597 | */ | |
598 | atomic_inc(&eb->refs); | |
599 | ||
600 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
601 | if (!reads_done) | |
602 | goto err; | |
603 | ||
604 | eb->read_mirror = mirror; | |
605 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
606 | ret = -EIO; | |
607 | goto err; | |
608 | } | |
609 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 610 | err: |
53b381b3 DW |
611 | if (ret) { |
612 | /* | |
613 | * our io error hook is going to dec the io pages | |
614 | * again, we have to make sure it has something | |
615 | * to decrement | |
616 | */ | |
617 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 618 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 619 | } |
0b32f4bb | 620 | free_extent_buffer(eb); |
77bf40a2 | 621 | |
f188591e | 622 | return ret; |
ce9adaa5 CM |
623 | } |
624 | ||
4a69a410 CM |
625 | static void run_one_async_start(struct btrfs_work *work) |
626 | { | |
4a69a410 | 627 | struct async_submit_bio *async; |
4e4cbee9 | 628 | blk_status_t ret; |
4a69a410 CM |
629 | |
630 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
631 | ret = async->submit_bio_start(async->inode, async->bio, |
632 | async->dio_file_offset); | |
79787eaa | 633 | if (ret) |
4e4cbee9 | 634 | async->status = ret; |
4a69a410 CM |
635 | } |
636 | ||
06ea01b1 DS |
637 | /* |
638 | * In order to insert checksums into the metadata in large chunks, we wait | |
639 | * until bio submission time. All the pages in the bio are checksummed and | |
640 | * sums are attached onto the ordered extent record. | |
641 | * | |
642 | * At IO completion time the csums attached on the ordered extent record are | |
643 | * inserted into the tree. | |
644 | */ | |
4a69a410 | 645 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 646 | { |
917f32a2 CH |
647 | struct async_submit_bio *async = |
648 | container_of(work, struct async_submit_bio, work); | |
649 | struct inode *inode = async->inode; | |
650 | struct btrfs_bio *bbio = btrfs_bio(async->bio); | |
4854ddd0 | 651 | |
bb7ab3b9 | 652 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 | 653 | if (async->status) { |
917f32a2 | 654 | btrfs_bio_end_io(bbio, async->status); |
79787eaa JM |
655 | return; |
656 | } | |
657 | ||
ec39f769 CM |
658 | /* |
659 | * All of the bios that pass through here are from async helpers. | |
660 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
661 | * This changes nothing when cgroups aren't in use. | |
662 | */ | |
663 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
1a722d8f | 664 | btrfs_submit_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
4a69a410 CM |
665 | } |
666 | ||
667 | static void run_one_async_free(struct btrfs_work *work) | |
668 | { | |
669 | struct async_submit_bio *async; | |
670 | ||
671 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
672 | kfree(async); |
673 | } | |
674 | ||
ea1f0ced CH |
675 | /* |
676 | * Submit bio to an async queue. | |
677 | * | |
678 | * Retrun: | |
679 | * - true if the work has been succesfuly submitted | |
680 | * - false in case of error | |
681 | */ | |
682 | bool btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, int mirror_num, | |
683 | u64 dio_file_offset, | |
684 | extent_submit_bio_start_t *submit_bio_start) | |
44b8bd7e | 685 | { |
8896a08d | 686 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
687 | struct async_submit_bio *async; |
688 | ||
689 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
690 | if (!async) | |
ea1f0ced | 691 | return false; |
44b8bd7e | 692 | |
8896a08d | 693 | async->inode = inode; |
44b8bd7e CM |
694 | async->bio = bio; |
695 | async->mirror_num = mirror_num; | |
4a69a410 | 696 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 697 | |
a0cac0ec OS |
698 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
699 | run_one_async_free); | |
4a69a410 | 700 | |
1941b64b | 701 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 702 | |
4e4cbee9 | 703 | async->status = 0; |
79787eaa | 704 | |
67f055c7 | 705 | if (op_is_sync(bio->bi_opf)) |
a31b4a43 CH |
706 | btrfs_queue_work(fs_info->hipri_workers, &async->work); |
707 | else | |
708 | btrfs_queue_work(fs_info->workers, &async->work); | |
ea1f0ced | 709 | return true; |
44b8bd7e CM |
710 | } |
711 | ||
4e4cbee9 | 712 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 713 | { |
2c30c71b | 714 | struct bio_vec *bvec; |
ce3ed71a | 715 | struct btrfs_root *root; |
2b070cfe | 716 | int ret = 0; |
6dc4f100 | 717 | struct bvec_iter_all iter_all; |
ce3ed71a | 718 | |
c09abff8 | 719 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 720 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 721 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 722 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
723 | if (ret) |
724 | break; | |
ce3ed71a | 725 | } |
2c30c71b | 726 | |
4e4cbee9 | 727 | return errno_to_blk_status(ret); |
ce3ed71a CM |
728 | } |
729 | ||
8896a08d | 730 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 731 | u64 dio_file_offset) |
22c59948 | 732 | { |
8b712842 CM |
733 | /* |
734 | * when we're called for a write, we're already in the async | |
1a722d8f | 735 | * submission context. Just jump into btrfs_submit_bio. |
8b712842 | 736 | */ |
79787eaa | 737 | return btree_csum_one_bio(bio); |
4a69a410 | 738 | } |
22c59948 | 739 | |
f4dcfb30 | 740 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 741 | struct btrfs_inode *bi) |
de0022b9 | 742 | { |
4eef29ef | 743 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 744 | return false; |
6300463b | 745 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 746 | return false; |
9b4e675a | 747 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
748 | return false; |
749 | return true; | |
de0022b9 JB |
750 | } |
751 | ||
94d9e11b | 752 | void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) |
44b8bd7e | 753 | { |
0b246afa | 754 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
917f32a2 | 755 | struct btrfs_bio *bbio = btrfs_bio(bio); |
4e4cbee9 | 756 | blk_status_t ret; |
cad321ad | 757 | |
08a6f464 CH |
758 | bio->bi_opf |= REQ_META; |
759 | ||
cfe94440 | 760 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
1a722d8f CH |
761 | btrfs_submit_bio(fs_info, bio, mirror_num); |
762 | return; | |
44b8bd7e | 763 | } |
d313d7a3 | 764 | |
ea1f0ced CH |
765 | /* |
766 | * Kthread helpers are used to submit writes so that checksumming can | |
767 | * happen in parallel across all CPUs. | |
768 | */ | |
769 | if (should_async_write(fs_info, BTRFS_I(inode)) && | |
770 | btrfs_wq_submit_bio(inode, bio, mirror_num, 0, btree_submit_bio_start)) | |
771 | return; | |
772 | ||
773 | ret = btree_csum_one_bio(bio); | |
94d9e11b | 774 | if (ret) { |
917f32a2 | 775 | btrfs_bio_end_io(bbio, ret); |
ea1f0ced | 776 | return; |
94d9e11b | 777 | } |
ea1f0ced CH |
778 | |
779 | btrfs_submit_bio(fs_info, bio, mirror_num); | |
44b8bd7e CM |
780 | } |
781 | ||
3dd1462e | 782 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
783 | static int btree_migrate_folio(struct address_space *mapping, |
784 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
785 | { |
786 | /* | |
787 | * we can't safely write a btree page from here, | |
788 | * we haven't done the locking hook | |
789 | */ | |
8958b551 | 790 | if (folio_test_dirty(src)) |
784b4e29 CM |
791 | return -EAGAIN; |
792 | /* | |
793 | * Buffers may be managed in a filesystem specific way. | |
794 | * We must have no buffers or drop them. | |
795 | */ | |
8958b551 MWO |
796 | if (folio_get_private(src) && |
797 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 798 | return -EAGAIN; |
54184650 | 799 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 800 | } |
8958b551 MWO |
801 | #else |
802 | #define btree_migrate_folio NULL | |
3dd1462e | 803 | #endif |
784b4e29 | 804 | |
0da5468f CM |
805 | static int btree_writepages(struct address_space *mapping, |
806 | struct writeback_control *wbc) | |
807 | { | |
e2d84521 MX |
808 | struct btrfs_fs_info *fs_info; |
809 | int ret; | |
810 | ||
d8d5f3e1 | 811 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
812 | |
813 | if (wbc->for_kupdate) | |
814 | return 0; | |
815 | ||
e2d84521 | 816 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 817 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
818 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
819 | BTRFS_DIRTY_METADATA_THRESH, | |
820 | fs_info->dirty_metadata_batch); | |
e2d84521 | 821 | if (ret < 0) |
793955bc | 822 | return 0; |
793955bc | 823 | } |
0b32f4bb | 824 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
825 | } |
826 | ||
f913cff3 | 827 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 828 | { |
f913cff3 MWO |
829 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
830 | return false; | |
0c4e538b | 831 | |
f913cff3 | 832 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
833 | } |
834 | ||
895586eb MWO |
835 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
836 | size_t length) | |
d98237b3 | 837 | { |
d1310b2e | 838 | struct extent_io_tree *tree; |
895586eb MWO |
839 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
840 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 841 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
842 | if (folio_get_private(folio)) { |
843 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
844 | "folio private not zero on folio %llu", | |
845 | (unsigned long long)folio_pos(folio)); | |
846 | folio_detach_private(folio); | |
9ad6b7bc | 847 | } |
d98237b3 CM |
848 | } |
849 | ||
bb146eb2 | 850 | #ifdef DEBUG |
0079c3b1 MWO |
851 | static bool btree_dirty_folio(struct address_space *mapping, |
852 | struct folio *folio) | |
853 | { | |
854 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
139e8cd3 | 855 | struct btrfs_subpage *subpage; |
0b32f4bb | 856 | struct extent_buffer *eb; |
139e8cd3 | 857 | int cur_bit = 0; |
0079c3b1 | 858 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
859 | |
860 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 861 | eb = folio_get_private(folio); |
139e8cd3 QW |
862 | BUG_ON(!eb); |
863 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
864 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 865 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 866 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 867 | } |
0079c3b1 | 868 | subpage = folio_get_private(folio); |
139e8cd3 QW |
869 | |
870 | ASSERT(subpage->dirty_bitmap); | |
871 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
872 | unsigned long flags; | |
873 | u64 cur; | |
874 | u16 tmp = (1 << cur_bit); | |
875 | ||
876 | spin_lock_irqsave(&subpage->lock, flags); | |
877 | if (!(tmp & subpage->dirty_bitmap)) { | |
878 | spin_unlock_irqrestore(&subpage->lock, flags); | |
879 | cur_bit++; | |
880 | continue; | |
881 | } | |
882 | spin_unlock_irqrestore(&subpage->lock, flags); | |
883 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 884 | |
139e8cd3 QW |
885 | eb = find_extent_buffer(fs_info, cur); |
886 | ASSERT(eb); | |
887 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
888 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 889 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
890 | free_extent_buffer(eb); |
891 | ||
892 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
893 | } | |
0079c3b1 | 894 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 895 | } |
0079c3b1 MWO |
896 | #else |
897 | #define btree_dirty_folio filemap_dirty_folio | |
898 | #endif | |
0b32f4bb | 899 | |
7f09410b | 900 | static const struct address_space_operations btree_aops = { |
0da5468f | 901 | .writepages = btree_writepages, |
f913cff3 | 902 | .release_folio = btree_release_folio, |
895586eb | 903 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
904 | .migrate_folio = btree_migrate_folio, |
905 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
906 | }; |
907 | ||
2ff7e61e JM |
908 | struct extent_buffer *btrfs_find_create_tree_block( |
909 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
910 | u64 bytenr, u64 owner_root, |
911 | int level) | |
0999df54 | 912 | { |
0b246afa JM |
913 | if (btrfs_is_testing(fs_info)) |
914 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 915 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
916 | } |
917 | ||
581c1760 QW |
918 | /* |
919 | * Read tree block at logical address @bytenr and do variant basic but critical | |
920 | * verification. | |
921 | * | |
1b7ec85e | 922 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
923 | * @parent_transid: expected transid of this tree block, skip check if 0 |
924 | * @level: expected level, mandatory check | |
925 | * @first_key: expected key in slot 0, skip check if NULL | |
926 | */ | |
2ff7e61e | 927 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
928 | u64 owner_root, u64 parent_transid, |
929 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
930 | { |
931 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
932 | int ret; |
933 | ||
3fbaf258 | 934 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
935 | if (IS_ERR(buf)) |
936 | return buf; | |
0999df54 | 937 | |
6a2e9dc4 | 938 | ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); |
0f0fe8f7 | 939 | if (ret) { |
537f38f0 | 940 | free_extent_buffer_stale(buf); |
64c043de | 941 | return ERR_PTR(ret); |
0f0fe8f7 | 942 | } |
88c602ab QW |
943 | if (btrfs_check_eb_owner(buf, owner_root)) { |
944 | free_extent_buffer_stale(buf); | |
945 | return ERR_PTR(-EUCLEAN); | |
946 | } | |
5f39d397 | 947 | return buf; |
ce9adaa5 | 948 | |
eb60ceac CM |
949 | } |
950 | ||
6a884d7d | 951 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 952 | { |
6a884d7d | 953 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 954 | if (btrfs_header_generation(buf) == |
e2d84521 | 955 | fs_info->running_transaction->transid) { |
49d0c642 | 956 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 957 | |
b9473439 | 958 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
959 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
960 | -buf->len, | |
961 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
962 | clear_extent_buffer_dirty(buf); |
963 | } | |
925baedd | 964 | } |
5f39d397 CM |
965 | } |
966 | ||
da17066c | 967 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 968 | u64 objectid) |
d97e63b6 | 969 | { |
7c0260ee | 970 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
971 | |
972 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
973 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
974 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 975 | root->fs_info = fs_info; |
2e608bd1 | 976 | root->root_key.objectid = objectid; |
cfaa7295 | 977 | root->node = NULL; |
a28ec197 | 978 | root->commit_root = NULL; |
27cdeb70 | 979 | root->state = 0; |
abed4aaa | 980 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 981 | |
0f7d52f4 | 982 | root->last_trans = 0; |
6b8fad57 | 983 | root->free_objectid = 0; |
eb73c1b7 | 984 | root->nr_delalloc_inodes = 0; |
199c2a9c | 985 | root->nr_ordered_extents = 0; |
6bef4d31 | 986 | root->inode_tree = RB_ROOT; |
088aea3b | 987 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
2e608bd1 JB |
988 | |
989 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
990 | |
991 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 992 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
993 | INIT_LIST_HEAD(&root->delalloc_inodes); |
994 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
995 | INIT_LIST_HEAD(&root->ordered_extents); |
996 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 997 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
998 | INIT_LIST_HEAD(&root->logged_list[0]); |
999 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1000 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1001 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1002 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1003 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1004 | spin_lock_init(&root->log_extents_lock[0]); |
1005 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1006 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1007 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1008 | mutex_init(&root->log_mutex); |
31f3d255 | 1009 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1010 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1011 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1012 | init_waitqueue_head(&root->log_writer_wait); |
1013 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1014 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1015 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1016 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1017 | atomic_set(&root->log_commit[0], 0); |
1018 | atomic_set(&root->log_commit[1], 0); | |
1019 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1020 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1021 | refcount_set(&root->refs, 1); |
8ecebf4d | 1022 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1023 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1024 | root->log_transid = 0; |
d1433deb | 1025 | root->log_transid_committed = -1; |
257c62e1 | 1026 | root->last_log_commit = 0; |
2e608bd1 | 1027 | root->anon_dev = 0; |
e289f03e | 1028 | if (!dummy) { |
43eb5f29 QW |
1029 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1030 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1031 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1032 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1033 | } | |
017e5369 | 1034 | |
5f3ab90a | 1035 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1036 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1037 | #ifdef CONFIG_BTRFS_DEBUG |
1038 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 1039 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1040 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 1041 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1042 | #endif |
3768f368 CM |
1043 | } |
1044 | ||
74e4d827 | 1045 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1046 | u64 objectid, gfp_t flags) |
6f07e42e | 1047 | { |
74e4d827 | 1048 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1049 | if (root) |
96dfcb46 | 1050 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1051 | return root; |
1052 | } | |
1053 | ||
06ea65a3 JB |
1054 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1055 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1056 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1057 | { |
1058 | struct btrfs_root *root; | |
1059 | ||
7c0260ee JM |
1060 | if (!fs_info) |
1061 | return ERR_PTR(-EINVAL); | |
1062 | ||
96dfcb46 | 1063 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1064 | if (!root) |
1065 | return ERR_PTR(-ENOMEM); | |
da17066c | 1066 | |
b9ef22de | 1067 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1068 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1069 | |
1070 | return root; | |
1071 | } | |
1072 | #endif | |
1073 | ||
abed4aaa JB |
1074 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1075 | { | |
1076 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1077 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1078 | ||
1079 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1080 | } | |
1081 | ||
1082 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1083 | { | |
1084 | const struct btrfs_key *key = k; | |
1085 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1086 | ||
1087 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1088 | } | |
1089 | ||
1090 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1091 | { | |
1092 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1093 | struct rb_node *tmp; | |
1094 | ||
1095 | write_lock(&fs_info->global_root_lock); | |
1096 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1097 | write_unlock(&fs_info->global_root_lock); | |
1098 | ASSERT(!tmp); | |
1099 | ||
1100 | return tmp ? -EEXIST : 0; | |
1101 | } | |
1102 | ||
1103 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1104 | { | |
1105 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1106 | ||
1107 | write_lock(&fs_info->global_root_lock); | |
1108 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1109 | write_unlock(&fs_info->global_root_lock); | |
1110 | } | |
1111 | ||
1112 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1113 | struct btrfs_key *key) | |
1114 | { | |
1115 | struct rb_node *node; | |
1116 | struct btrfs_root *root = NULL; | |
1117 | ||
1118 | read_lock(&fs_info->global_root_lock); | |
1119 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1120 | if (node) | |
1121 | root = container_of(node, struct btrfs_root, rb_node); | |
1122 | read_unlock(&fs_info->global_root_lock); | |
1123 | ||
1124 | return root; | |
1125 | } | |
1126 | ||
f7238e50 JB |
1127 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
1128 | { | |
1129 | struct btrfs_block_group *block_group; | |
1130 | u64 ret; | |
1131 | ||
1132 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1133 | return 0; | |
1134 | ||
1135 | if (bytenr) | |
1136 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
1137 | else | |
1138 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
1139 | ASSERT(block_group); | |
1140 | if (!block_group) | |
1141 | return 0; | |
1142 | ret = block_group->global_root_id; | |
1143 | btrfs_put_block_group(block_group); | |
1144 | ||
1145 | return ret; | |
1146 | } | |
1147 | ||
abed4aaa JB |
1148 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
1149 | { | |
1150 | struct btrfs_key key = { | |
1151 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1152 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1153 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1154 | }; |
1155 | ||
1156 | return btrfs_global_root(fs_info, &key); | |
1157 | } | |
1158 | ||
1159 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1160 | { | |
1161 | struct btrfs_key key = { | |
1162 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1163 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1164 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1165 | }; |
1166 | ||
1167 | return btrfs_global_root(fs_info, &key); | |
1168 | } | |
1169 | ||
20897f5c | 1170 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1171 | u64 objectid) |
1172 | { | |
9b7a2440 | 1173 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1174 | struct extent_buffer *leaf; |
1175 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1176 | struct btrfs_root *root; | |
1177 | struct btrfs_key key; | |
b89f6d1f | 1178 | unsigned int nofs_flag; |
20897f5c | 1179 | int ret = 0; |
20897f5c | 1180 | |
b89f6d1f FM |
1181 | /* |
1182 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1183 | * context to avoid deadlock if reclaim happens. | |
1184 | */ | |
1185 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1186 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1187 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1188 | if (!root) |
1189 | return ERR_PTR(-ENOMEM); | |
1190 | ||
20897f5c AJ |
1191 | root->root_key.objectid = objectid; |
1192 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1193 | root->root_key.offset = 0; | |
1194 | ||
9631e4cc JB |
1195 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1196 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1197 | if (IS_ERR(leaf)) { |
1198 | ret = PTR_ERR(leaf); | |
1dd05682 | 1199 | leaf = NULL; |
8a6a87cd | 1200 | goto fail_unlock; |
20897f5c AJ |
1201 | } |
1202 | ||
20897f5c | 1203 | root->node = leaf; |
20897f5c AJ |
1204 | btrfs_mark_buffer_dirty(leaf); |
1205 | ||
1206 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1207 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1208 | |
f944d2cb DS |
1209 | btrfs_set_root_flags(&root->root_item, 0); |
1210 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1211 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1212 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1213 | btrfs_set_root_level(&root->root_item, 0); | |
1214 | btrfs_set_root_refs(&root->root_item, 1); | |
1215 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1216 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1217 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1218 | if (is_fstree(objectid)) |
807fc790 AS |
1219 | generate_random_guid(root->root_item.uuid); |
1220 | else | |
1221 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1222 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1223 | |
8a6a87cd BB |
1224 | btrfs_tree_unlock(leaf); |
1225 | ||
20897f5c AJ |
1226 | key.objectid = objectid; |
1227 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1228 | key.offset = 0; | |
1229 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1230 | if (ret) | |
1231 | goto fail; | |
1232 | ||
1dd05682 TI |
1233 | return root; |
1234 | ||
8a6a87cd | 1235 | fail_unlock: |
8c38938c | 1236 | if (leaf) |
1dd05682 | 1237 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1238 | fail: |
00246528 | 1239 | btrfs_put_root(root); |
20897f5c | 1240 | |
1dd05682 | 1241 | return ERR_PTR(ret); |
20897f5c AJ |
1242 | } |
1243 | ||
7237f183 YZ |
1244 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1245 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1246 | { |
1247 | struct btrfs_root *root; | |
e02119d5 | 1248 | |
96dfcb46 | 1249 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1250 | if (!root) |
7237f183 | 1251 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1252 | |
e02119d5 CM |
1253 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1254 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1255 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1256 | |
6ab6ebb7 NA |
1257 | return root; |
1258 | } | |
1259 | ||
1260 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1261 | struct btrfs_root *root) | |
1262 | { | |
1263 | struct extent_buffer *leaf; | |
1264 | ||
7237f183 | 1265 | /* |
92a7cc42 | 1266 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1267 | * |
92a7cc42 QW |
1268 | * Log trees are not exposed to user space thus can't be snapshotted, |
1269 | * and they go away before a real commit is actually done. | |
1270 | * | |
1271 | * They do store pointers to file data extents, and those reference | |
1272 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1273 | */ |
e02119d5 | 1274 | |
4d75f8a9 | 1275 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1276 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1277 | if (IS_ERR(leaf)) |
1278 | return PTR_ERR(leaf); | |
e02119d5 | 1279 | |
7237f183 | 1280 | root->node = leaf; |
e02119d5 | 1281 | |
e02119d5 CM |
1282 | btrfs_mark_buffer_dirty(root->node); |
1283 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1284 | |
1285 | return 0; | |
7237f183 YZ |
1286 | } |
1287 | ||
1288 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1289 | struct btrfs_fs_info *fs_info) | |
1290 | { | |
1291 | struct btrfs_root *log_root; | |
1292 | ||
1293 | log_root = alloc_log_tree(trans, fs_info); | |
1294 | if (IS_ERR(log_root)) | |
1295 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1296 | |
3ddebf27 NA |
1297 | if (!btrfs_is_zoned(fs_info)) { |
1298 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1299 | ||
1300 | if (ret) { | |
1301 | btrfs_put_root(log_root); | |
1302 | return ret; | |
1303 | } | |
6ab6ebb7 NA |
1304 | } |
1305 | ||
7237f183 YZ |
1306 | WARN_ON(fs_info->log_root_tree); |
1307 | fs_info->log_root_tree = log_root; | |
1308 | return 0; | |
1309 | } | |
1310 | ||
1311 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1312 | struct btrfs_root *root) | |
1313 | { | |
0b246afa | 1314 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1315 | struct btrfs_root *log_root; |
1316 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1317 | int ret; |
7237f183 | 1318 | |
0b246afa | 1319 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1320 | if (IS_ERR(log_root)) |
1321 | return PTR_ERR(log_root); | |
1322 | ||
6ab6ebb7 NA |
1323 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1324 | if (ret) { | |
1325 | btrfs_put_root(log_root); | |
1326 | return ret; | |
1327 | } | |
1328 | ||
7237f183 YZ |
1329 | log_root->last_trans = trans->transid; |
1330 | log_root->root_key.offset = root->root_key.objectid; | |
1331 | ||
1332 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1333 | btrfs_set_stack_inode_generation(inode_item, 1); |
1334 | btrfs_set_stack_inode_size(inode_item, 3); | |
1335 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1336 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1337 | fs_info->nodesize); |
3cae210f | 1338 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1339 | |
5d4f98a2 | 1340 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1341 | |
1342 | WARN_ON(root->log_root); | |
1343 | root->log_root = log_root; | |
1344 | root->log_transid = 0; | |
d1433deb | 1345 | root->log_transid_committed = -1; |
257c62e1 | 1346 | root->last_log_commit = 0; |
e02119d5 CM |
1347 | return 0; |
1348 | } | |
1349 | ||
49d11bea JB |
1350 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1351 | struct btrfs_path *path, | |
1352 | struct btrfs_key *key) | |
e02119d5 CM |
1353 | { |
1354 | struct btrfs_root *root; | |
1355 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1356 | u64 generation; |
cb517eab | 1357 | int ret; |
581c1760 | 1358 | int level; |
0f7d52f4 | 1359 | |
96dfcb46 | 1360 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1361 | if (!root) |
1362 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1363 | |
cb517eab MX |
1364 | ret = btrfs_find_root(tree_root, key, path, |
1365 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1366 | if (ret) { |
13a8a7c8 YZ |
1367 | if (ret > 0) |
1368 | ret = -ENOENT; | |
49d11bea | 1369 | goto fail; |
0f7d52f4 | 1370 | } |
13a8a7c8 | 1371 | |
84234f3a | 1372 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1373 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1374 | root->node = read_tree_block(fs_info, |
1375 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1376 | key->objectid, generation, level, NULL); |
64c043de LB |
1377 | if (IS_ERR(root->node)) { |
1378 | ret = PTR_ERR(root->node); | |
8c38938c | 1379 | root->node = NULL; |
49d11bea | 1380 | goto fail; |
4eb150d6 QW |
1381 | } |
1382 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1383 | ret = -EIO; |
49d11bea | 1384 | goto fail; |
416bc658 | 1385 | } |
88c602ab QW |
1386 | |
1387 | /* | |
1388 | * For real fs, and not log/reloc trees, root owner must | |
1389 | * match its root node owner | |
1390 | */ | |
1391 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1392 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1393 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1394 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1395 | btrfs_crit(fs_info, | |
1396 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1397 | root->root_key.objectid, root->node->start, | |
1398 | btrfs_header_owner(root->node), | |
1399 | root->root_key.objectid); | |
1400 | ret = -EUCLEAN; | |
1401 | goto fail; | |
1402 | } | |
5d4f98a2 | 1403 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1404 | return root; |
49d11bea | 1405 | fail: |
00246528 | 1406 | btrfs_put_root(root); |
49d11bea JB |
1407 | return ERR_PTR(ret); |
1408 | } | |
1409 | ||
1410 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1411 | struct btrfs_key *key) | |
1412 | { | |
1413 | struct btrfs_root *root; | |
1414 | struct btrfs_path *path; | |
1415 | ||
1416 | path = btrfs_alloc_path(); | |
1417 | if (!path) | |
1418 | return ERR_PTR(-ENOMEM); | |
1419 | root = read_tree_root_path(tree_root, path, key); | |
1420 | btrfs_free_path(path); | |
1421 | ||
1422 | return root; | |
cb517eab MX |
1423 | } |
1424 | ||
2dfb1e43 QW |
1425 | /* |
1426 | * Initialize subvolume root in-memory structure | |
1427 | * | |
1428 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1429 | */ | |
1430 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1431 | { |
1432 | int ret; | |
dcc3eb96 | 1433 | unsigned int nofs_flag; |
cb517eab | 1434 | |
dcc3eb96 NB |
1435 | /* |
1436 | * We might be called under a transaction (e.g. indirect backref | |
1437 | * resolution) which could deadlock if it triggers memory reclaim | |
1438 | */ | |
1439 | nofs_flag = memalloc_nofs_save(); | |
1440 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1441 | memalloc_nofs_restore(nofs_flag); | |
1442 | if (ret) | |
8257b2dc | 1443 | goto fail; |
8257b2dc | 1444 | |
aeb935a4 | 1445 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1446 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1447 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1448 | btrfs_check_and_init_root_item(&root->root_item); |
1449 | } | |
1450 | ||
851fd730 QW |
1451 | /* |
1452 | * Don't assign anonymous block device to roots that are not exposed to | |
1453 | * userspace, the id pool is limited to 1M | |
1454 | */ | |
1455 | if (is_fstree(root->root_key.objectid) && | |
1456 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1457 | if (!anon_dev) { |
1458 | ret = get_anon_bdev(&root->anon_dev); | |
1459 | if (ret) | |
1460 | goto fail; | |
1461 | } else { | |
1462 | root->anon_dev = anon_dev; | |
1463 | } | |
851fd730 | 1464 | } |
f32e48e9 CR |
1465 | |
1466 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1467 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1468 | if (ret) { |
1469 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1470 | goto fail; |
f32e48e9 CR |
1471 | } |
1472 | ||
6b8fad57 | 1473 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1474 | |
1475 | mutex_unlock(&root->objectid_mutex); | |
1476 | ||
cb517eab MX |
1477 | return 0; |
1478 | fail: | |
84db5ccf | 1479 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1480 | return ret; |
1481 | } | |
1482 | ||
a98db0f3 JB |
1483 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1484 | u64 root_id) | |
cb517eab MX |
1485 | { |
1486 | struct btrfs_root *root; | |
1487 | ||
fc7cbcd4 DS |
1488 | spin_lock(&fs_info->fs_roots_radix_lock); |
1489 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1490 | (unsigned long)root_id); | |
bc44d7c4 | 1491 | if (root) |
00246528 | 1492 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1493 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1494 | return root; |
1495 | } | |
1496 | ||
49d11bea JB |
1497 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1498 | u64 objectid) | |
1499 | { | |
abed4aaa JB |
1500 | struct btrfs_key key = { |
1501 | .objectid = objectid, | |
1502 | .type = BTRFS_ROOT_ITEM_KEY, | |
1503 | .offset = 0, | |
1504 | }; | |
1505 | ||
49d11bea JB |
1506 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1507 | return btrfs_grab_root(fs_info->tree_root); | |
1508 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1509 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1510 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1511 | return btrfs_grab_root(fs_info->chunk_root); | |
1512 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1513 | return btrfs_grab_root(fs_info->dev_root); | |
1514 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1515 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1516 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1517 | return btrfs_grab_root(fs_info->quota_root) ? | |
1518 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1519 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1520 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1521 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
14033b08 QW |
1522 | if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) |
1523 | return btrfs_grab_root(fs_info->block_group_root) ? | |
1524 | fs_info->block_group_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1525 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1526 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1527 | ||
1528 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1529 | } | |
49d11bea JB |
1530 | return NULL; |
1531 | } | |
1532 | ||
cb517eab MX |
1533 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1534 | struct btrfs_root *root) | |
1535 | { | |
1536 | int ret; | |
1537 | ||
fc7cbcd4 DS |
1538 | ret = radix_tree_preload(GFP_NOFS); |
1539 | if (ret) | |
1540 | return ret; | |
1541 | ||
1542 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1543 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1544 | (unsigned long)root->root_key.objectid, | |
1545 | root); | |
af01d2e5 | 1546 | if (ret == 0) { |
00246528 | 1547 | btrfs_grab_root(root); |
fc7cbcd4 | 1548 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1549 | } |
fc7cbcd4 DS |
1550 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1551 | radix_tree_preload_end(); | |
cb517eab MX |
1552 | |
1553 | return ret; | |
1554 | } | |
1555 | ||
bd647ce3 JB |
1556 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1557 | { | |
1558 | #ifdef CONFIG_BTRFS_DEBUG | |
1559 | struct btrfs_root *root; | |
1560 | ||
1561 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1562 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1563 | ||
bd647ce3 JB |
1564 | root = list_first_entry(&fs_info->allocated_roots, |
1565 | struct btrfs_root, leak_list); | |
457f1864 | 1566 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1567 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1568 | refcount_read(&root->refs)); |
1569 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1570 | btrfs_put_root(root); |
1571 | btrfs_put_root(root); | |
bd647ce3 JB |
1572 | } |
1573 | #endif | |
1574 | } | |
1575 | ||
abed4aaa JB |
1576 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1577 | { | |
1578 | struct btrfs_root *root; | |
1579 | struct rb_node *node; | |
1580 | ||
1581 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1582 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1583 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1584 | btrfs_put_root(root); | |
1585 | } | |
1586 | } | |
1587 | ||
0d4b0463 JB |
1588 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1589 | { | |
141386e1 JB |
1590 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1591 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1592 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1593 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1594 | btrfs_free_csum_hash(fs_info); | |
1595 | btrfs_free_stripe_hash_table(fs_info); | |
1596 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1597 | kfree(fs_info->balance_ctl); |
1598 | kfree(fs_info->delayed_root); | |
abed4aaa | 1599 | free_global_roots(fs_info); |
00246528 JB |
1600 | btrfs_put_root(fs_info->tree_root); |
1601 | btrfs_put_root(fs_info->chunk_root); | |
1602 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1603 | btrfs_put_root(fs_info->quota_root); |
1604 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1605 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1606 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1607 | btrfs_put_root(fs_info->block_group_root); |
bd647ce3 | 1608 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1609 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1610 | kfree(fs_info->super_copy); |
1611 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1612 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1613 | kvfree(fs_info); |
1614 | } | |
1615 | ||
1616 | ||
2dfb1e43 QW |
1617 | /* |
1618 | * Get an in-memory reference of a root structure. | |
1619 | * | |
1620 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1621 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1622 | * found, then read it from disk and add it to cached fs roots. | |
1623 | * | |
1624 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1625 | * | |
1626 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1627 | * same root objectid. | |
1628 | * | |
1629 | * @objectid: root id | |
1630 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1631 | * pass 0 for new allocation. | |
1632 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1633 | * for orphan roots | |
1634 | */ | |
1635 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1636 | u64 objectid, dev_t anon_dev, | |
1637 | bool check_ref) | |
5eda7b5e CM |
1638 | { |
1639 | struct btrfs_root *root; | |
381cf658 | 1640 | struct btrfs_path *path; |
1d4c08e0 | 1641 | struct btrfs_key key; |
5eda7b5e CM |
1642 | int ret; |
1643 | ||
49d11bea JB |
1644 | root = btrfs_get_global_root(fs_info, objectid); |
1645 | if (root) | |
1646 | return root; | |
4df27c4d | 1647 | again: |
56e9357a | 1648 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1649 | if (root) { |
2dfb1e43 QW |
1650 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1651 | ASSERT(!anon_dev); | |
bc44d7c4 | 1652 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1653 | btrfs_put_root(root); |
48475471 | 1654 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1655 | } |
5eda7b5e | 1656 | return root; |
48475471 | 1657 | } |
5eda7b5e | 1658 | |
56e9357a DS |
1659 | key.objectid = objectid; |
1660 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1661 | key.offset = (u64)-1; | |
1662 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1663 | if (IS_ERR(root)) |
1664 | return root; | |
3394e160 | 1665 | |
c00869f1 | 1666 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1667 | ret = -ENOENT; |
581bb050 | 1668 | goto fail; |
35a30d7c | 1669 | } |
581bb050 | 1670 | |
2dfb1e43 | 1671 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1672 | if (ret) |
1673 | goto fail; | |
3394e160 | 1674 | |
381cf658 DS |
1675 | path = btrfs_alloc_path(); |
1676 | if (!path) { | |
1677 | ret = -ENOMEM; | |
1678 | goto fail; | |
1679 | } | |
1d4c08e0 DS |
1680 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1681 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1682 | key.offset = objectid; |
1d4c08e0 DS |
1683 | |
1684 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1685 | btrfs_free_path(path); |
d68fc57b YZ |
1686 | if (ret < 0) |
1687 | goto fail; | |
1688 | if (ret == 0) | |
27cdeb70 | 1689 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1690 | |
cb517eab | 1691 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1692 | if (ret) { |
168a2f77 JJB |
1693 | if (ret == -EEXIST) { |
1694 | btrfs_put_root(root); | |
4df27c4d | 1695 | goto again; |
168a2f77 | 1696 | } |
4df27c4d | 1697 | goto fail; |
0f7d52f4 | 1698 | } |
edbd8d4e | 1699 | return root; |
4df27c4d | 1700 | fail: |
33fab972 FM |
1701 | /* |
1702 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1703 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1704 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1705 | * and once again by our caller. | |
1706 | */ | |
1707 | if (anon_dev) | |
1708 | root->anon_dev = 0; | |
8c38938c | 1709 | btrfs_put_root(root); |
4df27c4d | 1710 | return ERR_PTR(ret); |
edbd8d4e CM |
1711 | } |
1712 | ||
2dfb1e43 QW |
1713 | /* |
1714 | * Get in-memory reference of a root structure | |
1715 | * | |
1716 | * @objectid: tree objectid | |
1717 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1718 | * one reference | |
1719 | */ | |
1720 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1721 | u64 objectid, bool check_ref) | |
1722 | { | |
1723 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1724 | } | |
1725 | ||
1726 | /* | |
1727 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1728 | * the anonymous block device id | |
1729 | * | |
1730 | * @objectid: tree objectid | |
1731 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1732 | * parameter value | |
1733 | */ | |
1734 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1735 | u64 objectid, dev_t anon_dev) | |
1736 | { | |
1737 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1738 | } | |
1739 | ||
49d11bea JB |
1740 | /* |
1741 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1742 | * @fs_info: the fs_info | |
1743 | * @objectid: the objectid we need to lookup | |
1744 | * | |
1745 | * This is exclusively used for backref walking, and exists specifically because | |
1746 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1747 | * creation time, which means we may have to read the tree_root in order to look | |
1748 | * up a fs root that is not in memory. If the root is not in memory we will | |
1749 | * read the tree root commit root and look up the fs root from there. This is a | |
1750 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1751 | * have the most uptodate information, it'll simply be discarded once the | |
1752 | * backref code is finished using the root. | |
1753 | */ | |
1754 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1755 | struct btrfs_path *path, | |
1756 | u64 objectid) | |
1757 | { | |
1758 | struct btrfs_root *root; | |
1759 | struct btrfs_key key; | |
1760 | ||
1761 | ASSERT(path->search_commit_root && path->skip_locking); | |
1762 | ||
1763 | /* | |
1764 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1765 | * since this is called via the backref walking code we won't be looking | |
1766 | * up a root that doesn't exist, unless there's corruption. So if root | |
1767 | * != NULL just return it. | |
1768 | */ | |
1769 | root = btrfs_get_global_root(fs_info, objectid); | |
1770 | if (root) | |
1771 | return root; | |
1772 | ||
1773 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1774 | if (root) | |
1775 | return root; | |
1776 | ||
1777 | key.objectid = objectid; | |
1778 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1779 | key.offset = (u64)-1; | |
1780 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1781 | btrfs_release_path(path); | |
1782 | ||
1783 | return root; | |
1784 | } | |
1785 | ||
a74a4b97 CM |
1786 | static int cleaner_kthread(void *arg) |
1787 | { | |
0d031dc4 | 1788 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1789 | int again; |
a74a4b97 | 1790 | |
d6fd0ae2 | 1791 | while (1) { |
d0278245 | 1792 | again = 0; |
a74a4b97 | 1793 | |
fd340d0f JB |
1794 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1795 | ||
d0278245 | 1796 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1797 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1798 | goto sleep; |
1799 | ||
90c711ab ZB |
1800 | /* |
1801 | * Do not do anything if we might cause open_ctree() to block | |
1802 | * before we have finished mounting the filesystem. | |
1803 | */ | |
0b246afa | 1804 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1805 | goto sleep; |
1806 | ||
0b246afa | 1807 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1808 | goto sleep; |
1809 | ||
dc7f370c MX |
1810 | /* |
1811 | * Avoid the problem that we change the status of the fs | |
1812 | * during the above check and trylock. | |
1813 | */ | |
2ff7e61e | 1814 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1815 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1816 | goto sleep; |
76dda93c | 1817 | } |
a74a4b97 | 1818 | |
2ff7e61e | 1819 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1820 | |
33c44184 | 1821 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1822 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1823 | |
1824 | /* | |
05323cd1 MX |
1825 | * The defragger has dealt with the R/O remount and umount, |
1826 | * needn't do anything special here. | |
d0278245 | 1827 | */ |
0b246afa | 1828 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1829 | |
1830 | /* | |
f3372065 | 1831 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1832 | * with relocation (btrfs_relocate_chunk) and relocation |
1833 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1834 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1835 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1836 | * unused block groups. | |
1837 | */ | |
0b246afa | 1838 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1839 | |
1840 | /* | |
1841 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1842 | * all unused block_groups. This possibly gives us some more free | |
1843 | * space. | |
1844 | */ | |
1845 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1846 | sleep: |
a0a1db70 | 1847 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1848 | if (kthread_should_park()) |
1849 | kthread_parkme(); | |
1850 | if (kthread_should_stop()) | |
1851 | return 0; | |
838fe188 | 1852 | if (!again) { |
a74a4b97 | 1853 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1854 | schedule(); |
a74a4b97 CM |
1855 | __set_current_state(TASK_RUNNING); |
1856 | } | |
da288d28 | 1857 | } |
a74a4b97 CM |
1858 | } |
1859 | ||
1860 | static int transaction_kthread(void *arg) | |
1861 | { | |
1862 | struct btrfs_root *root = arg; | |
0b246afa | 1863 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1864 | struct btrfs_trans_handle *trans; |
1865 | struct btrfs_transaction *cur; | |
8929ecfa | 1866 | u64 transid; |
643900be | 1867 | time64_t delta; |
a74a4b97 | 1868 | unsigned long delay; |
914b2007 | 1869 | bool cannot_commit; |
a74a4b97 CM |
1870 | |
1871 | do { | |
914b2007 | 1872 | cannot_commit = false; |
ba1bc00f | 1873 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1874 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1875 | |
0b246afa JM |
1876 | spin_lock(&fs_info->trans_lock); |
1877 | cur = fs_info->running_transaction; | |
a74a4b97 | 1878 | if (!cur) { |
0b246afa | 1879 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1880 | goto sleep; |
1881 | } | |
31153d81 | 1882 | |
643900be | 1883 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
1884 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
1885 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 1886 | delta < fs_info->commit_interval) { |
0b246afa | 1887 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1888 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1889 | delay = min(delay, | |
1890 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1891 | goto sleep; |
1892 | } | |
8929ecfa | 1893 | transid = cur->transid; |
0b246afa | 1894 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1895 | |
79787eaa | 1896 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1897 | trans = btrfs_attach_transaction(root); |
914b2007 | 1898 | if (IS_ERR(trans)) { |
354aa0fb MX |
1899 | if (PTR_ERR(trans) != -ENOENT) |
1900 | cannot_commit = true; | |
79787eaa | 1901 | goto sleep; |
914b2007 | 1902 | } |
8929ecfa | 1903 | if (transid == trans->transid) { |
3a45bb20 | 1904 | btrfs_commit_transaction(trans); |
8929ecfa | 1905 | } else { |
3a45bb20 | 1906 | btrfs_end_transaction(trans); |
8929ecfa | 1907 | } |
a74a4b97 | 1908 | sleep: |
0b246afa JM |
1909 | wake_up_process(fs_info->cleaner_kthread); |
1910 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1911 | |
84961539 | 1912 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1913 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1914 | if (!kthread_should_stop() && |
0b246afa | 1915 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1916 | cannot_commit)) |
bc5511d0 | 1917 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1918 | } while (!kthread_should_stop()); |
1919 | return 0; | |
1920 | } | |
1921 | ||
af31f5e5 | 1922 | /* |
01f0f9da NB |
1923 | * This will find the highest generation in the array of root backups. The |
1924 | * index of the highest array is returned, or -EINVAL if we can't find | |
1925 | * anything. | |
af31f5e5 CM |
1926 | * |
1927 | * We check to make sure the array is valid by comparing the | |
1928 | * generation of the latest root in the array with the generation | |
1929 | * in the super block. If they don't match we pitch it. | |
1930 | */ | |
01f0f9da | 1931 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1932 | { |
01f0f9da | 1933 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1934 | u64 cur; |
af31f5e5 CM |
1935 | struct btrfs_root_backup *root_backup; |
1936 | int i; | |
1937 | ||
1938 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1939 | root_backup = info->super_copy->super_roots + i; | |
1940 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1941 | if (cur == newest_gen) | |
01f0f9da | 1942 | return i; |
af31f5e5 CM |
1943 | } |
1944 | ||
01f0f9da | 1945 | return -EINVAL; |
af31f5e5 CM |
1946 | } |
1947 | ||
af31f5e5 CM |
1948 | /* |
1949 | * copy all the root pointers into the super backup array. | |
1950 | * this will bump the backup pointer by one when it is | |
1951 | * done | |
1952 | */ | |
1953 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1954 | { | |
6ef108dd | 1955 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1956 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1957 | |
1958 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1959 | ||
1960 | /* | |
1961 | * make sure all of our padding and empty slots get zero filled | |
1962 | * regardless of which ones we use today | |
1963 | */ | |
1964 | memset(root_backup, 0, sizeof(*root_backup)); | |
1965 | ||
1966 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1967 | ||
1968 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1969 | btrfs_set_backup_tree_root_gen(root_backup, | |
1970 | btrfs_header_generation(info->tree_root->node)); | |
1971 | ||
1972 | btrfs_set_backup_tree_root_level(root_backup, | |
1973 | btrfs_header_level(info->tree_root->node)); | |
1974 | ||
1975 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1976 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1977 | btrfs_header_generation(info->chunk_root->node)); | |
1978 | btrfs_set_backup_chunk_root_level(root_backup, | |
1979 | btrfs_header_level(info->chunk_root->node)); | |
1980 | ||
1c56ab99 | 1981 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1982 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1983 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1984 | |
1985 | btrfs_set_backup_extent_root(root_backup, | |
1986 | extent_root->node->start); | |
1987 | btrfs_set_backup_extent_root_gen(root_backup, | |
1988 | btrfs_header_generation(extent_root->node)); | |
1989 | btrfs_set_backup_extent_root_level(root_backup, | |
1990 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1991 | |
1992 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1993 | btrfs_set_backup_csum_root_gen(root_backup, | |
1994 | btrfs_header_generation(csum_root->node)); | |
1995 | btrfs_set_backup_csum_root_level(root_backup, | |
1996 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1997 | } |
af31f5e5 | 1998 | |
7c7e82a7 CM |
1999 | /* |
2000 | * we might commit during log recovery, which happens before we set | |
2001 | * the fs_root. Make sure it is valid before we fill it in. | |
2002 | */ | |
2003 | if (info->fs_root && info->fs_root->node) { | |
2004 | btrfs_set_backup_fs_root(root_backup, | |
2005 | info->fs_root->node->start); | |
2006 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2007 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2008 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2009 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2010 | } |
af31f5e5 CM |
2011 | |
2012 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2013 | btrfs_set_backup_dev_root_gen(root_backup, | |
2014 | btrfs_header_generation(info->dev_root->node)); | |
2015 | btrfs_set_backup_dev_root_level(root_backup, | |
2016 | btrfs_header_level(info->dev_root->node)); | |
2017 | ||
af31f5e5 CM |
2018 | btrfs_set_backup_total_bytes(root_backup, |
2019 | btrfs_super_total_bytes(info->super_copy)); | |
2020 | btrfs_set_backup_bytes_used(root_backup, | |
2021 | btrfs_super_bytes_used(info->super_copy)); | |
2022 | btrfs_set_backup_num_devices(root_backup, | |
2023 | btrfs_super_num_devices(info->super_copy)); | |
2024 | ||
2025 | /* | |
2026 | * if we don't copy this out to the super_copy, it won't get remembered | |
2027 | * for the next commit | |
2028 | */ | |
2029 | memcpy(&info->super_copy->super_roots, | |
2030 | &info->super_for_commit->super_roots, | |
2031 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2032 | } | |
2033 | ||
bd2336b2 NB |
2034 | /* |
2035 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2036 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2037 | * | |
2038 | * fs_info - filesystem whose backup roots need to be read | |
2039 | * priority - priority of backup root required | |
2040 | * | |
2041 | * Returns backup root index on success and -EINVAL otherwise. | |
2042 | */ | |
2043 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2044 | { | |
2045 | int backup_index = find_newest_super_backup(fs_info); | |
2046 | struct btrfs_super_block *super = fs_info->super_copy; | |
2047 | struct btrfs_root_backup *root_backup; | |
2048 | ||
2049 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2050 | if (priority == 0) | |
2051 | return backup_index; | |
2052 | ||
2053 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2054 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2055 | } else { | |
2056 | return -EINVAL; | |
2057 | } | |
2058 | ||
2059 | root_backup = super->super_roots + backup_index; | |
2060 | ||
2061 | btrfs_set_super_generation(super, | |
2062 | btrfs_backup_tree_root_gen(root_backup)); | |
2063 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2064 | btrfs_set_super_root_level(super, | |
2065 | btrfs_backup_tree_root_level(root_backup)); | |
2066 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2067 | ||
2068 | /* | |
2069 | * Fixme: the total bytes and num_devices need to match or we should | |
2070 | * need a fsck | |
2071 | */ | |
2072 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2073 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2074 | ||
2075 | return backup_index; | |
2076 | } | |
2077 | ||
7abadb64 LB |
2078 | /* helper to cleanup workers */ |
2079 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2080 | { | |
dc6e3209 | 2081 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2082 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
a31b4a43 | 2083 | btrfs_destroy_workqueue(fs_info->hipri_workers); |
5cdc7ad3 | 2084 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
2085 | if (fs_info->endio_workers) |
2086 | destroy_workqueue(fs_info->endio_workers); | |
d34e123d CH |
2087 | if (fs_info->endio_raid56_workers) |
2088 | destroy_workqueue(fs_info->endio_raid56_workers); | |
385de0ef CH |
2089 | if (fs_info->rmw_workers) |
2090 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
2091 | if (fs_info->compressed_write_workers) |
2092 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
2093 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2094 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2095 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2096 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2097 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2098 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2099 | if (fs_info->discard_ctl.discard_workers) |
2100 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2101 | /* |
2102 | * Now that all other work queues are destroyed, we can safely destroy | |
2103 | * the queues used for metadata I/O, since tasks from those other work | |
2104 | * queues can do metadata I/O operations. | |
2105 | */ | |
d7b9416f CH |
2106 | if (fs_info->endio_meta_workers) |
2107 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
2108 | } |
2109 | ||
2e9f5954 R |
2110 | static void free_root_extent_buffers(struct btrfs_root *root) |
2111 | { | |
2112 | if (root) { | |
2113 | free_extent_buffer(root->node); | |
2114 | free_extent_buffer(root->commit_root); | |
2115 | root->node = NULL; | |
2116 | root->commit_root = NULL; | |
2117 | } | |
2118 | } | |
2119 | ||
abed4aaa JB |
2120 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2121 | { | |
2122 | struct btrfs_root *root, *tmp; | |
2123 | ||
2124 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2125 | &fs_info->global_root_tree, | |
2126 | rb_node) | |
2127 | free_root_extent_buffers(root); | |
2128 | } | |
2129 | ||
af31f5e5 | 2130 | /* helper to cleanup tree roots */ |
4273eaff | 2131 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2132 | { |
2e9f5954 | 2133 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2134 | |
abed4aaa | 2135 | free_global_root_pointers(info); |
2e9f5954 | 2136 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2137 | free_root_extent_buffers(info->quota_root); |
2138 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2139 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2140 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 2141 | free_root_extent_buffers(info->block_group_root); |
4273eaff | 2142 | if (free_chunk_root) |
2e9f5954 | 2143 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2144 | } |
2145 | ||
8c38938c JB |
2146 | void btrfs_put_root(struct btrfs_root *root) |
2147 | { | |
2148 | if (!root) | |
2149 | return; | |
2150 | ||
2151 | if (refcount_dec_and_test(&root->refs)) { | |
2152 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2153 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2154 | if (root->anon_dev) |
2155 | free_anon_bdev(root->anon_dev); | |
2156 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2157 | free_root_extent_buffers(root); |
8c38938c | 2158 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 2159 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 2160 | list_del_init(&root->leak_list); |
fc7cbcd4 | 2161 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
2162 | #endif |
2163 | kfree(root); | |
2164 | } | |
2165 | } | |
2166 | ||
faa2dbf0 | 2167 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 2168 | { |
fc7cbcd4 DS |
2169 | int ret; |
2170 | struct btrfs_root *gang[8]; | |
2171 | int i; | |
171f6537 JB |
2172 | |
2173 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
2174 | gang[0] = list_entry(fs_info->dead_roots.next, |
2175 | struct btrfs_root, root_list); | |
2176 | list_del(&gang[0]->root_list); | |
171f6537 | 2177 | |
fc7cbcd4 DS |
2178 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
2179 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
2180 | btrfs_put_root(gang[0]); | |
171f6537 JB |
2181 | } |
2182 | ||
fc7cbcd4 DS |
2183 | while (1) { |
2184 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2185 | (void **)gang, 0, | |
2186 | ARRAY_SIZE(gang)); | |
2187 | if (!ret) | |
2188 | break; | |
2189 | for (i = 0; i < ret; i++) | |
2190 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
2191 | } |
2192 | } | |
af31f5e5 | 2193 | |
638aa7ed ES |
2194 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2195 | { | |
2196 | mutex_init(&fs_info->scrub_lock); | |
2197 | atomic_set(&fs_info->scrubs_running, 0); | |
2198 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2199 | atomic_set(&fs_info->scrubs_paused, 0); | |
2200 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2201 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2202 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2203 | } |
2204 | ||
779a65a4 ES |
2205 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2206 | { | |
2207 | spin_lock_init(&fs_info->balance_lock); | |
2208 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2209 | atomic_set(&fs_info->balance_pause_req, 0); |
2210 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2211 | fs_info->balance_ctl = NULL; | |
2212 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2213 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2214 | } |
2215 | ||
6bccf3ab | 2216 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2217 | { |
2ff7e61e | 2218 | struct inode *inode = fs_info->btree_inode; |
e256927b JB |
2219 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
2220 | fs_info->tree_root); | |
2ff7e61e JM |
2221 | |
2222 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2223 | set_nlink(inode, 1); | |
f37938e0 ES |
2224 | /* |
2225 | * we set the i_size on the btree inode to the max possible int. | |
2226 | * the real end of the address space is determined by all of | |
2227 | * the devices in the system | |
2228 | */ | |
2ff7e61e JM |
2229 | inode->i_size = OFFSET_MAX; |
2230 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2231 | |
2ff7e61e | 2232 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2233 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
efb0645b | 2234 | IO_TREE_BTREE_INODE_IO, NULL); |
2ff7e61e | 2235 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2236 | |
5c8fd99f | 2237 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
2238 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
2239 | BTRFS_I(inode)->location.type = 0; | |
2240 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 2241 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 2242 | __insert_inode_hash(inode, hash); |
f37938e0 ES |
2243 | } |
2244 | ||
ad618368 ES |
2245 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2246 | { | |
ad618368 | 2247 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2248 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2249 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2250 | } |
2251 | ||
f9e92e40 ES |
2252 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2253 | { | |
2254 | spin_lock_init(&fs_info->qgroup_lock); | |
2255 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2256 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2257 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2258 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2259 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2260 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 2261 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
2262 | mutex_init(&fs_info->qgroup_rescan_lock); |
2263 | } | |
2264 | ||
d21deec5 | 2265 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2266 | { |
f7b885be | 2267 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2268 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2269 | |
2270 | fs_info->workers = | |
a31b4a43 CH |
2271 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2272 | fs_info->hipri_workers = | |
2273 | btrfs_alloc_workqueue(fs_info, "worker-high", | |
cb001095 | 2274 | flags | WQ_HIGHPRI, max_active, 16); |
2a458198 ES |
2275 | |
2276 | fs_info->delalloc_workers = | |
cb001095 JM |
2277 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2278 | flags, max_active, 2); | |
2a458198 ES |
2279 | |
2280 | fs_info->flush_workers = | |
cb001095 JM |
2281 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2282 | flags, max_active, 0); | |
2a458198 ES |
2283 | |
2284 | fs_info->caching_workers = | |
cb001095 | 2285 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2286 | |
2a458198 | 2287 | fs_info->fixup_workers = |
cb001095 | 2288 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 | 2289 | |
2a458198 | 2290 | fs_info->endio_workers = |
d7b9416f | 2291 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 2292 | fs_info->endio_meta_workers = |
d7b9416f | 2293 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
2a458198 | 2294 | fs_info->endio_raid56_workers = |
d34e123d | 2295 | alloc_workqueue("btrfs-endio-raid56", flags, max_active); |
385de0ef | 2296 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2297 | fs_info->endio_write_workers = |
cb001095 JM |
2298 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2299 | max_active, 2); | |
fed8a72d CH |
2300 | fs_info->compressed_write_workers = |
2301 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2302 | fs_info->endio_freespace_worker = |
cb001095 JM |
2303 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2304 | max_active, 0); | |
2a458198 | 2305 | fs_info->delayed_workers = |
cb001095 JM |
2306 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2307 | max_active, 0); | |
2a458198 | 2308 | fs_info->qgroup_rescan_workers = |
cb001095 | 2309 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2310 | fs_info->discard_ctl.discard_workers = |
2311 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 | 2312 | |
a31b4a43 CH |
2313 | if (!(fs_info->workers && fs_info->hipri_workers && |
2314 | fs_info->delalloc_workers && fs_info->flush_workers && | |
2a458198 | 2315 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2316 | fs_info->compressed_write_workers && |
2a458198 ES |
2317 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2318 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
f26c9238 QW |
2319 | fs_info->caching_workers && fs_info->fixup_workers && |
2320 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2321 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2322 | return -ENOMEM; |
2323 | } | |
2324 | ||
2325 | return 0; | |
2326 | } | |
2327 | ||
6d97c6e3 JT |
2328 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2329 | { | |
2330 | struct crypto_shash *csum_shash; | |
b4e967be | 2331 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2332 | |
b4e967be | 2333 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2334 | |
2335 | if (IS_ERR(csum_shash)) { | |
2336 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2337 | csum_driver); |
6d97c6e3 JT |
2338 | return PTR_ERR(csum_shash); |
2339 | } | |
2340 | ||
2341 | fs_info->csum_shash = csum_shash; | |
2342 | ||
c8a5f8ca DS |
2343 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2344 | btrfs_super_csum_name(csum_type), | |
2345 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2346 | return 0; |
2347 | } | |
2348 | ||
63443bf5 ES |
2349 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2350 | struct btrfs_fs_devices *fs_devices) | |
2351 | { | |
2352 | int ret; | |
63443bf5 ES |
2353 | struct btrfs_root *log_tree_root; |
2354 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2355 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2356 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2357 | |
2358 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2359 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2360 | return -EIO; |
2361 | } | |
2362 | ||
96dfcb46 JB |
2363 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2364 | GFP_KERNEL); | |
63443bf5 ES |
2365 | if (!log_tree_root) |
2366 | return -ENOMEM; | |
2367 | ||
2ff7e61e | 2368 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2369 | BTRFS_TREE_LOG_OBJECTID, |
2370 | fs_info->generation + 1, level, | |
2371 | NULL); | |
64c043de | 2372 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2373 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2374 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2375 | log_tree_root->node = NULL; |
00246528 | 2376 | btrfs_put_root(log_tree_root); |
0eeff236 | 2377 | return ret; |
4eb150d6 QW |
2378 | } |
2379 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2380 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2381 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2382 | return -EIO; |
2383 | } | |
4eb150d6 | 2384 | |
63443bf5 ES |
2385 | /* returns with log_tree_root freed on success */ |
2386 | ret = btrfs_recover_log_trees(log_tree_root); | |
2387 | if (ret) { | |
0b246afa JM |
2388 | btrfs_handle_fs_error(fs_info, ret, |
2389 | "Failed to recover log tree"); | |
00246528 | 2390 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2391 | return ret; |
2392 | } | |
2393 | ||
bc98a42c | 2394 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2395 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2396 | if (ret) |
2397 | return ret; | |
2398 | } | |
2399 | ||
2400 | return 0; | |
2401 | } | |
2402 | ||
abed4aaa JB |
2403 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2404 | struct btrfs_path *path, u64 objectid, | |
2405 | const char *name) | |
2406 | { | |
2407 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2408 | struct btrfs_root *root; | |
f7238e50 | 2409 | u64 max_global_id = 0; |
abed4aaa JB |
2410 | int ret; |
2411 | struct btrfs_key key = { | |
2412 | .objectid = objectid, | |
2413 | .type = BTRFS_ROOT_ITEM_KEY, | |
2414 | .offset = 0, | |
2415 | }; | |
2416 | bool found = false; | |
2417 | ||
2418 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2419 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2420 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2421 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2422 | return 0; | |
2423 | } | |
2424 | ||
2425 | while (1) { | |
2426 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2427 | if (ret < 0) | |
2428 | break; | |
2429 | ||
2430 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2431 | ret = btrfs_next_leaf(tree_root, path); | |
2432 | if (ret) { | |
2433 | if (ret > 0) | |
2434 | ret = 0; | |
2435 | break; | |
2436 | } | |
2437 | } | |
2438 | ret = 0; | |
2439 | ||
2440 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2441 | if (key.objectid != objectid) | |
2442 | break; | |
2443 | btrfs_release_path(path); | |
2444 | ||
f7238e50 JB |
2445 | /* |
2446 | * Just worry about this for extent tree, it'll be the same for | |
2447 | * everybody. | |
2448 | */ | |
2449 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2450 | max_global_id = max(max_global_id, key.offset); | |
2451 | ||
abed4aaa JB |
2452 | found = true; |
2453 | root = read_tree_root_path(tree_root, path, &key); | |
2454 | if (IS_ERR(root)) { | |
2455 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2456 | ret = PTR_ERR(root); | |
2457 | break; | |
2458 | } | |
2459 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2460 | ret = btrfs_global_root_insert(root); | |
2461 | if (ret) { | |
2462 | btrfs_put_root(root); | |
2463 | break; | |
2464 | } | |
2465 | key.offset++; | |
2466 | } | |
2467 | btrfs_release_path(path); | |
2468 | ||
f7238e50 JB |
2469 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2470 | fs_info->nr_global_roots = max_global_id + 1; | |
2471 | ||
abed4aaa JB |
2472 | if (!found || ret) { |
2473 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2474 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2475 | ||
2476 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2477 | ret = ret ? ret : -ENOENT; | |
2478 | else | |
2479 | ret = 0; | |
2480 | btrfs_err(fs_info, "failed to load root %s", name); | |
2481 | } | |
2482 | return ret; | |
2483 | } | |
2484 | ||
2485 | static int load_global_roots(struct btrfs_root *tree_root) | |
2486 | { | |
2487 | struct btrfs_path *path; | |
2488 | int ret = 0; | |
2489 | ||
2490 | path = btrfs_alloc_path(); | |
2491 | if (!path) | |
2492 | return -ENOMEM; | |
2493 | ||
2494 | ret = load_global_roots_objectid(tree_root, path, | |
2495 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2496 | if (ret) | |
2497 | goto out; | |
2498 | ret = load_global_roots_objectid(tree_root, path, | |
2499 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2500 | if (ret) | |
2501 | goto out; | |
2502 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2503 | goto out; | |
2504 | ret = load_global_roots_objectid(tree_root, path, | |
2505 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2506 | "free space"); | |
2507 | out: | |
2508 | btrfs_free_path(path); | |
2509 | return ret; | |
2510 | } | |
2511 | ||
6bccf3ab | 2512 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2513 | { |
6bccf3ab | 2514 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2515 | struct btrfs_root *root; |
4bbcaa64 ES |
2516 | struct btrfs_key location; |
2517 | int ret; | |
2518 | ||
6bccf3ab JM |
2519 | BUG_ON(!fs_info->tree_root); |
2520 | ||
abed4aaa JB |
2521 | ret = load_global_roots(tree_root); |
2522 | if (ret) | |
2523 | return ret; | |
2524 | ||
4bbcaa64 ES |
2525 | location.type = BTRFS_ROOT_ITEM_KEY; |
2526 | location.offset = 0; | |
2527 | ||
1c56ab99 | 2528 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2529 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2530 | root = btrfs_read_tree_root(tree_root, &location); | |
2531 | if (IS_ERR(root)) { | |
2532 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2533 | ret = PTR_ERR(root); | |
2534 | goto out; | |
2535 | } | |
2536 | } else { | |
2537 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2538 | fs_info->block_group_root = root; | |
2539 | } | |
2540 | } | |
2541 | ||
2542 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2543 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2544 | if (IS_ERR(root)) { |
42437a63 JB |
2545 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2546 | ret = PTR_ERR(root); | |
2547 | goto out; | |
2548 | } | |
2549 | } else { | |
2550 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2551 | fs_info->dev_root = root; | |
f50f4353 | 2552 | } |
820a49da JB |
2553 | /* Initialize fs_info for all devices in any case */ |
2554 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2555 | |
aeb935a4 QW |
2556 | /* |
2557 | * This tree can share blocks with some other fs tree during relocation | |
2558 | * and we need a proper setup by btrfs_get_fs_root | |
2559 | */ | |
56e9357a DS |
2560 | root = btrfs_get_fs_root(tree_root->fs_info, |
2561 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2562 | if (IS_ERR(root)) { |
42437a63 JB |
2563 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2564 | ret = PTR_ERR(root); | |
2565 | goto out; | |
2566 | } | |
2567 | } else { | |
2568 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2569 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2570 | } |
aeb935a4 | 2571 | |
4bbcaa64 | 2572 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2573 | root = btrfs_read_tree_root(tree_root, &location); |
2574 | if (!IS_ERR(root)) { | |
2575 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2576 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2577 | fs_info->quota_root = root; |
4bbcaa64 ES |
2578 | } |
2579 | ||
2580 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2581 | root = btrfs_read_tree_root(tree_root, &location); |
2582 | if (IS_ERR(root)) { | |
42437a63 JB |
2583 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2584 | ret = PTR_ERR(root); | |
2585 | if (ret != -ENOENT) | |
2586 | goto out; | |
2587 | } | |
4bbcaa64 | 2588 | } else { |
a4f3d2c4 DS |
2589 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2590 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2591 | } |
2592 | ||
2593 | return 0; | |
f50f4353 LB |
2594 | out: |
2595 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2596 | location.objectid, ret); | |
2597 | return ret; | |
4bbcaa64 ES |
2598 | } |
2599 | ||
069ec957 QW |
2600 | /* |
2601 | * Real super block validation | |
2602 | * NOTE: super csum type and incompat features will not be checked here. | |
2603 | * | |
2604 | * @sb: super block to check | |
2605 | * @mirror_num: the super block number to check its bytenr: | |
2606 | * 0 the primary (1st) sb | |
2607 | * 1, 2 2nd and 3rd backup copy | |
2608 | * -1 skip bytenr check | |
2609 | */ | |
a05d3c91 QW |
2610 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2611 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2612 | { |
21a852b0 QW |
2613 | u64 nodesize = btrfs_super_nodesize(sb); |
2614 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2615 | int ret = 0; | |
2616 | ||
2617 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2618 | btrfs_err(fs_info, "no valid FS found"); | |
2619 | ret = -EINVAL; | |
2620 | } | |
2621 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2622 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2623 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2624 | ret = -EINVAL; | |
2625 | } | |
2626 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2627 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2628 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2629 | ret = -EINVAL; | |
2630 | } | |
2631 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2632 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2633 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2634 | ret = -EINVAL; | |
2635 | } | |
2636 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2637 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2638 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2639 | ret = -EINVAL; | |
2640 | } | |
2641 | ||
2642 | /* | |
2643 | * Check sectorsize and nodesize first, other check will need it. | |
2644 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2645 | */ | |
2646 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2647 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2648 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2649 | ret = -EINVAL; | |
2650 | } | |
0bb3eb3e QW |
2651 | |
2652 | /* | |
1a42daab QW |
2653 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2654 | * | |
2655 | * We can support 16K sectorsize with 64K page size without problem, | |
2656 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2657 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2658 | * beginning. | |
0bb3eb3e | 2659 | */ |
1a42daab | 2660 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2661 | btrfs_err(fs_info, |
0bb3eb3e | 2662 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2663 | sectorsize, PAGE_SIZE); |
2664 | ret = -EINVAL; | |
2665 | } | |
0bb3eb3e | 2666 | |
21a852b0 QW |
2667 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2668 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2669 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2670 | ret = -EINVAL; | |
2671 | } | |
2672 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2673 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2674 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2675 | ret = -EINVAL; | |
2676 | } | |
2677 | ||
2678 | /* Root alignment check */ | |
2679 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2680 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2681 | btrfs_super_root(sb)); | |
2682 | ret = -EINVAL; | |
2683 | } | |
2684 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2685 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2686 | btrfs_super_chunk_root(sb)); | |
2687 | ret = -EINVAL; | |
2688 | } | |
2689 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2690 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2691 | btrfs_super_log_root(sb)); | |
2692 | ret = -EINVAL; | |
2693 | } | |
2694 | ||
aefd7f70 NB |
2695 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2696 | BTRFS_FSID_SIZE)) { | |
2697 | btrfs_err(fs_info, | |
2698 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2699 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2700 | ret = -EINVAL; | |
2701 | } | |
2702 | ||
2703 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2704 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2705 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2706 | btrfs_err(fs_info, | |
2707 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2708 | fs_info->super_copy->metadata_uuid, | |
2709 | fs_info->fs_devices->metadata_uuid); | |
2710 | ret = -EINVAL; | |
2711 | } | |
2712 | ||
1c56ab99 QW |
2713 | /* |
2714 | * Artificial requirement for block-group-tree to force newer features | |
2715 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2716 | */ | |
2717 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2718 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2719 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2720 | btrfs_err(fs_info, | |
2721 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2722 | ret = -EINVAL; | |
2723 | } | |
2724 | ||
de37aa51 | 2725 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2726 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2727 | btrfs_err(fs_info, |
7239ff4b | 2728 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2729 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2730 | ret = -EINVAL; |
2731 | } | |
2732 | ||
2733 | /* | |
2734 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2735 | * done later | |
2736 | */ | |
2737 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2738 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2739 | btrfs_super_bytes_used(sb)); | |
2740 | ret = -EINVAL; | |
2741 | } | |
2742 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2743 | btrfs_err(fs_info, "invalid stripesize %u", | |
2744 | btrfs_super_stripesize(sb)); | |
2745 | ret = -EINVAL; | |
2746 | } | |
2747 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2748 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2749 | btrfs_super_num_devices(sb)); | |
2750 | if (btrfs_super_num_devices(sb) == 0) { | |
2751 | btrfs_err(fs_info, "number of devices is 0"); | |
2752 | ret = -EINVAL; | |
2753 | } | |
2754 | ||
069ec957 QW |
2755 | if (mirror_num >= 0 && |
2756 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2757 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2758 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2759 | ret = -EINVAL; | |
2760 | } | |
2761 | ||
2762 | /* | |
2763 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2764 | * and one chunk | |
2765 | */ | |
2766 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2767 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2768 | btrfs_super_sys_array_size(sb), | |
2769 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2770 | ret = -EINVAL; | |
2771 | } | |
2772 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2773 | + sizeof(struct btrfs_chunk)) { | |
2774 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2775 | btrfs_super_sys_array_size(sb), | |
2776 | sizeof(struct btrfs_disk_key) | |
2777 | + sizeof(struct btrfs_chunk)); | |
2778 | ret = -EINVAL; | |
2779 | } | |
2780 | ||
2781 | /* | |
2782 | * The generation is a global counter, we'll trust it more than the others | |
2783 | * but it's still possible that it's the one that's wrong. | |
2784 | */ | |
2785 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2786 | btrfs_warn(fs_info, | |
2787 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2788 | btrfs_super_generation(sb), | |
2789 | btrfs_super_chunk_root_generation(sb)); | |
2790 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2791 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2792 | btrfs_warn(fs_info, | |
2793 | "suspicious: generation < cache_generation: %llu < %llu", | |
2794 | btrfs_super_generation(sb), | |
2795 | btrfs_super_cache_generation(sb)); | |
2796 | ||
2797 | return ret; | |
2798 | } | |
2799 | ||
069ec957 QW |
2800 | /* |
2801 | * Validation of super block at mount time. | |
2802 | * Some checks already done early at mount time, like csum type and incompat | |
2803 | * flags will be skipped. | |
2804 | */ | |
2805 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2806 | { | |
a05d3c91 | 2807 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2808 | } |
2809 | ||
75cb857d QW |
2810 | /* |
2811 | * Validation of super block at write time. | |
2812 | * Some checks like bytenr check will be skipped as their values will be | |
2813 | * overwritten soon. | |
2814 | * Extra checks like csum type and incompat flags will be done here. | |
2815 | */ | |
2816 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2817 | struct btrfs_super_block *sb) | |
2818 | { | |
2819 | int ret; | |
2820 | ||
a05d3c91 | 2821 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2822 | if (ret < 0) |
2823 | goto out; | |
e7e16f48 | 2824 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2825 | ret = -EUCLEAN; |
2826 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2827 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2828 | goto out; | |
2829 | } | |
2830 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2831 | ret = -EUCLEAN; | |
2832 | btrfs_err(fs_info, | |
2833 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2834 | btrfs_super_incompat_flags(sb), | |
2835 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2836 | goto out; | |
2837 | } | |
2838 | out: | |
2839 | if (ret < 0) | |
2840 | btrfs_err(fs_info, | |
2841 | "super block corruption detected before writing it to disk"); | |
2842 | return ret; | |
2843 | } | |
2844 | ||
bd676446 JB |
2845 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2846 | { | |
2847 | int ret = 0; | |
2848 | ||
2849 | root->node = read_tree_block(root->fs_info, bytenr, | |
2850 | root->root_key.objectid, gen, level, NULL); | |
2851 | if (IS_ERR(root->node)) { | |
2852 | ret = PTR_ERR(root->node); | |
2853 | root->node = NULL; | |
4eb150d6 QW |
2854 | return ret; |
2855 | } | |
2856 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2857 | free_extent_buffer(root->node); |
2858 | root->node = NULL; | |
4eb150d6 | 2859 | return -EIO; |
bd676446 JB |
2860 | } |
2861 | ||
bd676446 JB |
2862 | btrfs_set_root_node(&root->root_item, root->node); |
2863 | root->commit_root = btrfs_root_node(root); | |
2864 | btrfs_set_root_refs(&root->root_item, 1); | |
2865 | return ret; | |
2866 | } | |
2867 | ||
2868 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2869 | { | |
2870 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2871 | u64 gen, bytenr; | |
2872 | int level, ret; | |
2873 | ||
2874 | bytenr = btrfs_super_root(sb); | |
2875 | gen = btrfs_super_generation(sb); | |
2876 | level = btrfs_super_root_level(sb); | |
2877 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2878 | if (ret) { |
bd676446 | 2879 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2880 | return ret; |
2881 | } | |
14033b08 | 2882 | return 0; |
bd676446 JB |
2883 | } |
2884 | ||
6ef108dd | 2885 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2886 | { |
6ef108dd | 2887 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2888 | struct btrfs_super_block *sb = fs_info->super_copy; |
2889 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2890 | bool handle_error = false; | |
2891 | int ret = 0; | |
2892 | int i; | |
2893 | ||
2894 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2895 | if (handle_error) { |
2896 | if (!IS_ERR(tree_root->node)) | |
2897 | free_extent_buffer(tree_root->node); | |
2898 | tree_root->node = NULL; | |
2899 | ||
2900 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2901 | break; | |
2902 | ||
2903 | free_root_pointers(fs_info, 0); | |
2904 | ||
2905 | /* | |
2906 | * Don't use the log in recovery mode, it won't be | |
2907 | * valid | |
2908 | */ | |
2909 | btrfs_set_super_log_root(sb, 0); | |
2910 | ||
2911 | /* We can't trust the free space cache either */ | |
2912 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2913 | ||
2914 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2915 | backup_index = ret; |
b8522a1e NB |
2916 | if (ret < 0) |
2917 | return ret; | |
2918 | } | |
b8522a1e | 2919 | |
bd676446 JB |
2920 | ret = load_important_roots(fs_info); |
2921 | if (ret) { | |
217f5004 | 2922 | handle_error = true; |
b8522a1e NB |
2923 | continue; |
2924 | } | |
2925 | ||
336a0d8d NB |
2926 | /* |
2927 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2928 | * hasn't been fully initialised and we are the only user | |
2929 | */ | |
453e4873 | 2930 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2931 | if (ret < 0) { |
b8522a1e NB |
2932 | handle_error = true; |
2933 | continue; | |
2934 | } | |
2935 | ||
6b8fad57 | 2936 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2937 | |
2938 | ret = btrfs_read_roots(fs_info); | |
2939 | if (ret < 0) { | |
2940 | handle_error = true; | |
2941 | continue; | |
2942 | } | |
2943 | ||
2944 | /* All successful */ | |
bd676446 JB |
2945 | fs_info->generation = btrfs_header_generation(tree_root->node); |
2946 | fs_info->last_trans_committed = fs_info->generation; | |
d96b3424 | 2947 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2948 | |
2949 | /* Always begin writing backup roots after the one being used */ | |
2950 | if (backup_index < 0) { | |
2951 | fs_info->backup_root_index = 0; | |
2952 | } else { | |
2953 | fs_info->backup_root_index = backup_index + 1; | |
2954 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2955 | } | |
b8522a1e NB |
2956 | break; |
2957 | } | |
2958 | ||
2959 | return ret; | |
2960 | } | |
2961 | ||
8260edba | 2962 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2963 | { |
fc7cbcd4 | 2964 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2965 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2966 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2967 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2968 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2969 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2970 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2971 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2972 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2973 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2974 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2975 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2976 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2977 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2978 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2979 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2980 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2981 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2982 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2983 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2984 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2985 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2986 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2987 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2988 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2989 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2990 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2991 | |
e1489b4f | 2992 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2993 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2994 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2995 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
3e738c53 IA |
2996 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, |
2997 | BTRFS_LOCKDEP_TRANS_COMMIT_START); | |
2998 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, | |
2999 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
3000 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
3001 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
3002 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
3003 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 3004 | |
0b86a832 | 3005 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3006 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3007 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3008 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3009 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3010 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3011 | #ifdef CONFIG_BTRFS_DEBUG |
3012 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3013 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3014 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3015 | #endif |
c8bf1b67 | 3016 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3017 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3018 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3019 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3020 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3021 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3022 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3023 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3024 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3025 | BTRFS_BLOCK_RSV_DELREFS); | |
3026 | ||
771ed689 | 3027 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3028 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3029 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3030 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3031 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3032 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3033 | fs_info->metadata_ratio = 0; |
4cb5300b | 3034 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3035 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3036 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3037 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3038 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3039 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3040 | |
b34b086c CM |
3041 | fs_info->thread_pool_size = min_t(unsigned long, |
3042 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3043 | |
199c2a9c MX |
3044 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3045 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3046 | |
638aa7ed | 3047 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3048 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3049 | fs_info->check_integrity_print_mask = 0; | |
3050 | #endif | |
779a65a4 | 3051 | btrfs_init_balance(fs_info); |
57056740 | 3052 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3053 | |
16b0c258 | 3054 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 3055 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 3056 | |
fe119a6e NB |
3057 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
3058 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
39279cc3 | 3059 | |
5a3f23d5 | 3060 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3061 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3062 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3063 | mutex_init(&fs_info->transaction_kthread_mutex); |
3064 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3065 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3066 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3067 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3068 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3069 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3070 | |
ad618368 | 3071 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3072 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3073 | btrfs_discard_init(fs_info); |
416ac51d | 3074 | |
fa9c0d79 CM |
3075 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3076 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3077 | ||
e6dcd2dc | 3078 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3079 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3080 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3081 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3082 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3083 | |
da17066c JM |
3084 | /* Usable values until the real ones are cached from the superblock */ |
3085 | fs_info->nodesize = 4096; | |
3086 | fs_info->sectorsize = 4096; | |
ab108d99 | 3087 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3088 | fs_info->stripesize = 4096; |
3089 | ||
f7b12a62 NA |
3090 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
3091 | ||
eede2bf3 OS |
3092 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3093 | fs_info->swapfile_pins = RB_ROOT; | |
3094 | ||
18bb8bbf JT |
3095 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3096 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3097 | } |
3098 | ||
3099 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3100 | { | |
3101 | int ret; | |
3102 | ||
3103 | fs_info->sb = sb; | |
3104 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3105 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3106 | |
5deb17e1 | 3107 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3108 | if (ret) |
c75e8394 | 3109 | return ret; |
ae18c37a JB |
3110 | |
3111 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3112 | if (ret) | |
c75e8394 | 3113 | return ret; |
ae18c37a JB |
3114 | |
3115 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3116 | (1 + ilog2(nr_cpu_ids)); | |
3117 | ||
3118 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3119 | if (ret) | |
c75e8394 | 3120 | return ret; |
ae18c37a JB |
3121 | |
3122 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3123 | GFP_KERNEL); | |
3124 | if (ret) | |
c75e8394 | 3125 | return ret; |
ae18c37a JB |
3126 | |
3127 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3128 | GFP_KERNEL); | |
c75e8394 JB |
3129 | if (!fs_info->delayed_root) |
3130 | return -ENOMEM; | |
ae18c37a JB |
3131 | btrfs_init_delayed_root(fs_info->delayed_root); |
3132 | ||
a0a1db70 FM |
3133 | if (sb_rdonly(sb)) |
3134 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3135 | ||
c75e8394 | 3136 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3137 | } |
3138 | ||
97f4dd09 NB |
3139 | static int btrfs_uuid_rescan_kthread(void *data) |
3140 | { | |
0d031dc4 | 3141 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
3142 | int ret; |
3143 | ||
3144 | /* | |
3145 | * 1st step is to iterate through the existing UUID tree and | |
3146 | * to delete all entries that contain outdated data. | |
3147 | * 2nd step is to add all missing entries to the UUID tree. | |
3148 | */ | |
3149 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3150 | if (ret < 0) { | |
c94bec2c JB |
3151 | if (ret != -EINTR) |
3152 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3153 | ret); | |
97f4dd09 NB |
3154 | up(&fs_info->uuid_tree_rescan_sem); |
3155 | return ret; | |
3156 | } | |
3157 | return btrfs_uuid_scan_kthread(data); | |
3158 | } | |
3159 | ||
3160 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3161 | { | |
3162 | struct task_struct *task; | |
3163 | ||
3164 | down(&fs_info->uuid_tree_rescan_sem); | |
3165 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3166 | if (IS_ERR(task)) { | |
3167 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3168 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3169 | up(&fs_info->uuid_tree_rescan_sem); | |
3170 | return PTR_ERR(task); | |
3171 | } | |
3172 | ||
3173 | return 0; | |
3174 | } | |
3175 | ||
8cd29088 BB |
3176 | /* |
3177 | * Some options only have meaning at mount time and shouldn't persist across | |
3178 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3179 | * code paths. | |
3180 | */ | |
3181 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3182 | { | |
3183 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3184 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3185 | } |
3186 | ||
44c0ca21 BB |
3187 | /* |
3188 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3189 | * and btrfs_remount when remounting from read-only to read-write. | |
3190 | */ | |
3191 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3192 | { | |
3193 | int ret; | |
94846229 | 3194 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3195 | bool clear_free_space_tree = false; |
3196 | ||
3197 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3198 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3199 | clear_free_space_tree = true; | |
3200 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3201 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3202 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3203 | clear_free_space_tree = true; | |
3204 | } | |
3205 | ||
3206 | if (clear_free_space_tree) { | |
3207 | btrfs_info(fs_info, "clearing free space tree"); | |
3208 | ret = btrfs_clear_free_space_tree(fs_info); | |
3209 | if (ret) { | |
3210 | btrfs_warn(fs_info, | |
3211 | "failed to clear free space tree: %d", ret); | |
3212 | goto out; | |
3213 | } | |
3214 | } | |
44c0ca21 | 3215 | |
8d488a8c FM |
3216 | /* |
3217 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3218 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3219 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3220 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3221 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3222 | * item before the root's tree is deleted - this means that if we unmount | |
3223 | * or crash before the deletion completes, on the next mount we will not | |
3224 | * delete what remains of the tree because the orphan item does not | |
3225 | * exists anymore, which is what tells us we have a pending deletion. | |
3226 | */ | |
3227 | ret = btrfs_find_orphan_roots(fs_info); | |
3228 | if (ret) | |
3229 | goto out; | |
3230 | ||
44c0ca21 BB |
3231 | ret = btrfs_cleanup_fs_roots(fs_info); |
3232 | if (ret) | |
3233 | goto out; | |
3234 | ||
8f1c21d7 BB |
3235 | down_read(&fs_info->cleanup_work_sem); |
3236 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3237 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3238 | up_read(&fs_info->cleanup_work_sem); | |
3239 | goto out; | |
3240 | } | |
3241 | up_read(&fs_info->cleanup_work_sem); | |
3242 | ||
44c0ca21 | 3243 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3244 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3245 | mutex_unlock(&fs_info->cleaner_mutex); |
3246 | if (ret < 0) { | |
3247 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3248 | goto out; | |
3249 | } | |
3250 | ||
5011139a BB |
3251 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3252 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3253 | btrfs_info(fs_info, "creating free space tree"); | |
3254 | ret = btrfs_create_free_space_tree(fs_info); | |
3255 | if (ret) { | |
3256 | btrfs_warn(fs_info, | |
3257 | "failed to create free space tree: %d", ret); | |
3258 | goto out; | |
3259 | } | |
3260 | } | |
3261 | ||
94846229 BB |
3262 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3263 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3264 | if (ret) | |
3265 | goto out; | |
3266 | } | |
3267 | ||
44c0ca21 BB |
3268 | ret = btrfs_resume_balance_async(fs_info); |
3269 | if (ret) | |
3270 | goto out; | |
3271 | ||
3272 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3273 | if (ret) { | |
3274 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3275 | goto out; | |
3276 | } | |
3277 | ||
3278 | btrfs_qgroup_rescan_resume(fs_info); | |
3279 | ||
3280 | if (!fs_info->uuid_root) { | |
3281 | btrfs_info(fs_info, "creating UUID tree"); | |
3282 | ret = btrfs_create_uuid_tree(fs_info); | |
3283 | if (ret) { | |
3284 | btrfs_warn(fs_info, | |
3285 | "failed to create the UUID tree %d", ret); | |
3286 | goto out; | |
3287 | } | |
3288 | } | |
3289 | ||
3290 | out: | |
3291 | return ret; | |
3292 | } | |
3293 | ||
d7f67ac9 QW |
3294 | /* |
3295 | * Do various sanity and dependency checks of different features. | |
3296 | * | |
3297 | * This is the place for less strict checks (like for subpage or artificial | |
3298 | * feature dependencies). | |
3299 | * | |
3300 | * For strict checks or possible corruption detection, see | |
3301 | * btrfs_validate_super(). | |
3302 | * | |
3303 | * This should be called after btrfs_parse_options(), as some mount options | |
3304 | * (space cache related) can modify on-disk format like free space tree and | |
3305 | * screw up certain feature dependencies. | |
3306 | */ | |
3307 | int btrfs_check_features(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3308 | { | |
3309 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3310 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3311 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3312 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3313 | ||
3314 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3315 | btrfs_err(fs_info, | |
3316 | "cannot mount because of unknown incompat features (0x%llx)", | |
3317 | incompat); | |
3318 | return -EINVAL; | |
3319 | } | |
3320 | ||
3321 | /* Runtime limitation for mixed block groups. */ | |
3322 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3323 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3324 | btrfs_err(fs_info, | |
3325 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3326 | fs_info->nodesize, fs_info->sectorsize); | |
3327 | return -EINVAL; | |
3328 | } | |
3329 | ||
3330 | /* Mixed backref is an always-enabled feature. */ | |
3331 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3332 | ||
3333 | /* Set compression related flags just in case. */ | |
3334 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3335 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3336 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3337 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3338 | ||
3339 | /* | |
3340 | * An ancient flag, which should really be marked deprecated. | |
3341 | * Such runtime limitation doesn't really need a incompat flag. | |
3342 | */ | |
3343 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3344 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3345 | ||
3346 | if (compat_ro_unsupp && !sb_rdonly(sb)) { | |
3347 | btrfs_err(fs_info, | |
3348 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3349 | compat_ro); | |
3350 | return -EINVAL; | |
3351 | } | |
3352 | ||
3353 | /* | |
3354 | * We have unsupported RO compat features, although RO mounted, we | |
3355 | * should not cause any metadata writes, including log replay. | |
3356 | * Or we could screw up whatever the new feature requires. | |
3357 | */ | |
3358 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3359 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3360 | btrfs_err(fs_info, | |
3361 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3362 | compat_ro); | |
3363 | return -EINVAL; | |
3364 | } | |
3365 | ||
3366 | /* | |
3367 | * Artificial limitations for block group tree, to force | |
3368 | * block-group-tree to rely on no-holes and free-space-tree. | |
3369 | */ | |
3370 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3371 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3372 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3373 | btrfs_err(fs_info, | |
3374 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3375 | return -EINVAL; | |
3376 | } | |
3377 | ||
3378 | /* | |
3379 | * Subpage runtime limitation on v1 cache. | |
3380 | * | |
3381 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3382 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3383 | * going to be deprecated anyway. | |
3384 | */ | |
3385 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3386 | btrfs_warn(fs_info, | |
3387 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3388 | PAGE_SIZE, fs_info->sectorsize); | |
3389 | return -EINVAL; | |
3390 | } | |
3391 | ||
3392 | /* This can be called by remount, we need to protect the super block. */ | |
3393 | spin_lock(&fs_info->super_lock); | |
3394 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3395 | spin_unlock(&fs_info->super_lock); | |
3396 | ||
3397 | return 0; | |
3398 | } | |
3399 | ||
ae18c37a JB |
3400 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3401 | char *options) | |
3402 | { | |
3403 | u32 sectorsize; | |
3404 | u32 nodesize; | |
3405 | u32 stripesize; | |
3406 | u64 generation; | |
3407 | u64 features; | |
3408 | u16 csum_type; | |
ae18c37a JB |
3409 | struct btrfs_super_block *disk_super; |
3410 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3411 | struct btrfs_root *tree_root; | |
3412 | struct btrfs_root *chunk_root; | |
3413 | int ret; | |
3414 | int err = -EINVAL; | |
ae18c37a JB |
3415 | int level; |
3416 | ||
8260edba | 3417 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3418 | if (ret) { |
83c8266a | 3419 | err = ret; |
ae18c37a | 3420 | goto fail; |
53b381b3 DW |
3421 | } |
3422 | ||
ae18c37a JB |
3423 | /* These need to be init'ed before we start creating inodes and such. */ |
3424 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3425 | GFP_KERNEL); | |
3426 | fs_info->tree_root = tree_root; | |
3427 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3428 | GFP_KERNEL); | |
3429 | fs_info->chunk_root = chunk_root; | |
3430 | if (!tree_root || !chunk_root) { | |
3431 | err = -ENOMEM; | |
c75e8394 | 3432 | goto fail; |
ae18c37a JB |
3433 | } |
3434 | ||
3435 | fs_info->btree_inode = new_inode(sb); | |
3436 | if (!fs_info->btree_inode) { | |
3437 | err = -ENOMEM; | |
c75e8394 | 3438 | goto fail; |
ae18c37a JB |
3439 | } |
3440 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3441 | btrfs_init_btree_inode(fs_info); | |
3442 | ||
d24fa5c1 | 3443 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3444 | |
3445 | /* | |
3446 | * Read super block and check the signature bytes only | |
3447 | */ | |
d24fa5c1 | 3448 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3449 | if (IS_ERR(disk_super)) { |
3450 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3451 | goto fail_alloc; |
20b45077 | 3452 | } |
39279cc3 | 3453 | |
8dc3f22c | 3454 | /* |
260db43c | 3455 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3456 | * corrupted, we'll find out |
3457 | */ | |
8f32380d | 3458 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3459 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3460 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3461 | csum_type); |
8dc3f22c | 3462 | err = -EINVAL; |
8f32380d | 3463 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3464 | goto fail_alloc; |
3465 | } | |
3466 | ||
83c68bbc SY |
3467 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3468 | ||
6d97c6e3 JT |
3469 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3470 | if (ret) { | |
3471 | err = ret; | |
8f32380d | 3472 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3473 | goto fail_alloc; |
3474 | } | |
3475 | ||
1104a885 DS |
3476 | /* |
3477 | * We want to check superblock checksum, the type is stored inside. | |
3478 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3479 | */ | |
3d17adea | 3480 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3481 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3482 | err = -EINVAL; |
8f32380d | 3483 | btrfs_release_disk_super(disk_super); |
141386e1 | 3484 | goto fail_alloc; |
1104a885 DS |
3485 | } |
3486 | ||
3487 | /* | |
3488 | * super_copy is zeroed at allocation time and we never touch the | |
3489 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3490 | * the whole block of INFO_SIZE | |
3491 | */ | |
8f32380d JT |
3492 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3493 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3494 | |
fbc6feae NB |
3495 | disk_super = fs_info->super_copy; |
3496 | ||
0b86a832 | 3497 | |
fbc6feae NB |
3498 | features = btrfs_super_flags(disk_super); |
3499 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3500 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3501 | btrfs_set_super_flags(disk_super, features); | |
3502 | btrfs_info(fs_info, | |
3503 | "found metadata UUID change in progress flag, clearing"); | |
3504 | } | |
3505 | ||
3506 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3507 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3508 | |
069ec957 | 3509 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3510 | if (ret) { |
05135f59 | 3511 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3512 | err = -EINVAL; |
141386e1 | 3513 | goto fail_alloc; |
1104a885 DS |
3514 | } |
3515 | ||
0f7d52f4 | 3516 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3517 | goto fail_alloc; |
0f7d52f4 | 3518 | |
acce952b | 3519 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3520 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3521 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3522 | |
75e7cb7f LB |
3523 | /* |
3524 | * In the long term, we'll store the compression type in the super | |
3525 | * block, and it'll be used for per file compression control. | |
3526 | */ | |
3527 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3528 | ||
6f93e834 AJ |
3529 | |
3530 | /* Set up fs_info before parsing mount options */ | |
3531 | nodesize = btrfs_super_nodesize(disk_super); | |
3532 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3533 | stripesize = sectorsize; | |
3534 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3535 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3536 | ||
3537 | fs_info->nodesize = nodesize; | |
3538 | fs_info->sectorsize = sectorsize; | |
3539 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3540 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3541 | fs_info->stripesize = stripesize; | |
3542 | ||
2ff7e61e | 3543 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3544 | if (ret) { |
3545 | err = ret; | |
141386e1 | 3546 | goto fail_alloc; |
2b82032c | 3547 | } |
dfe25020 | 3548 | |
d7f67ac9 QW |
3549 | ret = btrfs_check_features(fs_info, sb); |
3550 | if (ret < 0) { | |
3551 | err = ret; | |
dc4d3168 QW |
3552 | goto fail_alloc; |
3553 | } | |
3554 | ||
8481dd80 QW |
3555 | if (sectorsize < PAGE_SIZE) { |
3556 | struct btrfs_subpage_info *subpage_info; | |
3557 | ||
9f73f1ae QW |
3558 | /* |
3559 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3560 | * going to be deprecated. | |
3561 | * | |
3562 | * Force to use v2 cache for subpage case. | |
3563 | */ | |
3564 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3565 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3566 | "forcing free space tree for sector size %u with page size %lu", | |
3567 | sectorsize, PAGE_SIZE); | |
3568 | ||
95ea0486 QW |
3569 | btrfs_warn(fs_info, |
3570 | "read-write for sector size %u with page size %lu is experimental", | |
3571 | sectorsize, PAGE_SIZE); | |
8481dd80 QW |
3572 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3573 | if (!subpage_info) | |
3574 | goto fail_alloc; | |
3575 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3576 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3577 | } |
0bb3eb3e | 3578 | |
d21deec5 | 3579 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3580 | if (ret) { |
3581 | err = ret; | |
0dc3b84a JB |
3582 | goto fail_sb_buffer; |
3583 | } | |
4543df7e | 3584 | |
9e11ceee JK |
3585 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3586 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3587 | |
a061fc8d CM |
3588 | sb->s_blocksize = sectorsize; |
3589 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3590 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3591 | |
925baedd | 3592 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3593 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3594 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3595 | if (ret) { |
05135f59 | 3596 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3597 | goto fail_sb_buffer; |
84eed90f | 3598 | } |
0b86a832 | 3599 | |
84234f3a | 3600 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3601 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3602 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3603 | generation, level); | |
3604 | if (ret) { | |
05135f59 | 3605 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3606 | goto fail_tree_roots; |
83121942 | 3607 | } |
0b86a832 | 3608 | |
e17cade2 | 3609 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3610 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3611 | BTRFS_UUID_SIZE); | |
e17cade2 | 3612 | |
5b4aacef | 3613 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3614 | if (ret) { |
05135f59 | 3615 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3616 | goto fail_tree_roots; |
2b82032c | 3617 | } |
0b86a832 | 3618 | |
8dabb742 | 3619 | /* |
bacce86a AJ |
3620 | * At this point we know all the devices that make this filesystem, |
3621 | * including the seed devices but we don't know yet if the replace | |
3622 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3623 | * filesystem but skip the replace target device which is checked |
bacce86a | 3624 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3625 | */ |
bacce86a | 3626 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3627 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3628 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3629 | goto fail_tree_roots; |
3630 | } | |
3631 | ||
b8522a1e | 3632 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3633 | if (ret) |
b8522a1e | 3634 | goto fail_tree_roots; |
8929ecfa | 3635 | |
73651042 NA |
3636 | /* |
3637 | * Get zone type information of zoned block devices. This will also | |
3638 | * handle emulation of a zoned filesystem if a regular device has the | |
3639 | * zoned incompat feature flag set. | |
3640 | */ | |
3641 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3642 | if (ret) { | |
3643 | btrfs_err(fs_info, | |
3644 | "zoned: failed to read device zone info: %d", | |
3645 | ret); | |
3646 | goto fail_block_groups; | |
3647 | } | |
3648 | ||
75ec1db8 JB |
3649 | /* |
3650 | * If we have a uuid root and we're not being told to rescan we need to | |
3651 | * check the generation here so we can set the | |
3652 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3653 | * transaction during a balance or the log replay without updating the | |
3654 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3655 | * even though it was perfectly fine. | |
3656 | */ | |
3657 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3658 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3659 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3660 | ||
cf90d884 QW |
3661 | ret = btrfs_verify_dev_extents(fs_info); |
3662 | if (ret) { | |
3663 | btrfs_err(fs_info, | |
3664 | "failed to verify dev extents against chunks: %d", | |
3665 | ret); | |
3666 | goto fail_block_groups; | |
3667 | } | |
68310a5e ID |
3668 | ret = btrfs_recover_balance(fs_info); |
3669 | if (ret) { | |
05135f59 | 3670 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3671 | goto fail_block_groups; |
3672 | } | |
3673 | ||
733f4fbb SB |
3674 | ret = btrfs_init_dev_stats(fs_info); |
3675 | if (ret) { | |
05135f59 | 3676 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3677 | goto fail_block_groups; |
3678 | } | |
3679 | ||
8dabb742 SB |
3680 | ret = btrfs_init_dev_replace(fs_info); |
3681 | if (ret) { | |
05135f59 | 3682 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3683 | goto fail_block_groups; |
3684 | } | |
3685 | ||
b70f5097 NA |
3686 | ret = btrfs_check_zoned_mode(fs_info); |
3687 | if (ret) { | |
3688 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3689 | ret); | |
3690 | goto fail_block_groups; | |
3691 | } | |
3692 | ||
c6761a9e | 3693 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3694 | if (ret) { |
05135f59 DS |
3695 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3696 | ret); | |
b7c35e81 AJ |
3697 | goto fail_block_groups; |
3698 | } | |
3699 | ||
96f3136e | 3700 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3701 | if (ret) { |
05135f59 | 3702 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3703 | goto fail_fsdev_sysfs; |
c59021f8 | 3704 | } |
3705 | ||
c59021f8 | 3706 | ret = btrfs_init_space_info(fs_info); |
3707 | if (ret) { | |
05135f59 | 3708 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3709 | goto fail_sysfs; |
c59021f8 | 3710 | } |
3711 | ||
5b4aacef | 3712 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3713 | if (ret) { |
05135f59 | 3714 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3715 | goto fail_sysfs; |
1b1d1f66 | 3716 | } |
4330e183 | 3717 | |
16beac87 NA |
3718 | btrfs_free_zone_cache(fs_info); |
3719 | ||
5c78a5e7 AJ |
3720 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3721 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3722 | btrfs_warn(fs_info, |
52042d8e | 3723 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3724 | goto fail_sysfs; |
292fd7fc | 3725 | } |
9078a3e1 | 3726 | |
33c44184 | 3727 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3728 | "btrfs-cleaner"); |
57506d50 | 3729 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3730 | goto fail_sysfs; |
a74a4b97 CM |
3731 | |
3732 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3733 | tree_root, | |
3734 | "btrfs-transaction"); | |
57506d50 | 3735 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3736 | goto fail_cleaner; |
a74a4b97 | 3737 | |
583b7231 | 3738 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3739 | !fs_info->fs_devices->rotating) { |
583b7231 | 3740 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3741 | } |
3742 | ||
572d9ab7 | 3743 | /* |
01327610 | 3744 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3745 | * not have to wait for transaction commit |
3746 | */ | |
3747 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3748 | |
21adbd5c | 3749 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3750 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3751 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3752 | btrfs_test_opt(fs_info, |
cbeaae4f | 3753 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3754 | fs_info->check_integrity_print_mask); |
3755 | if (ret) | |
05135f59 DS |
3756 | btrfs_warn(fs_info, |
3757 | "failed to initialize integrity check module: %d", | |
3758 | ret); | |
21adbd5c SB |
3759 | } |
3760 | #endif | |
bcef60f2 AJ |
3761 | ret = btrfs_read_qgroup_config(fs_info); |
3762 | if (ret) | |
3763 | goto fail_trans_kthread; | |
21adbd5c | 3764 | |
fd708b81 JB |
3765 | if (btrfs_build_ref_tree(fs_info)) |
3766 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3767 | ||
96da0919 QW |
3768 | /* do not make disk changes in broken FS or nologreplay is given */ |
3769 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3770 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3771 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3772 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3773 | if (ret) { |
63443bf5 | 3774 | err = ret; |
28c16cbb | 3775 | goto fail_qgroup; |
79787eaa | 3776 | } |
e02119d5 | 3777 | } |
1a40e23b | 3778 | |
56e9357a | 3779 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3780 | if (IS_ERR(fs_info->fs_root)) { |
3781 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3782 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3783 | fs_info->fs_root = NULL; |
bcef60f2 | 3784 | goto fail_qgroup; |
3140c9a3 | 3785 | } |
c289811c | 3786 | |
bc98a42c | 3787 | if (sb_rdonly(sb)) |
8cd29088 | 3788 | goto clear_oneshot; |
59641015 | 3789 | |
44c0ca21 | 3790 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3791 | if (ret) { |
6bccf3ab | 3792 | close_ctree(fs_info); |
2b6ba629 | 3793 | return ret; |
e3acc2a6 | 3794 | } |
b0643e59 | 3795 | btrfs_discard_resume(fs_info); |
b382a324 | 3796 | |
44c0ca21 BB |
3797 | if (fs_info->uuid_root && |
3798 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3799 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3800 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3801 | ret = btrfs_check_uuid_tree(fs_info); |
3802 | if (ret) { | |
05135f59 DS |
3803 | btrfs_warn(fs_info, |
3804 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3805 | close_ctree(fs_info); |
70f80175 SB |
3806 | return ret; |
3807 | } | |
f7a81ea4 | 3808 | } |
94846229 | 3809 | |
afcdd129 | 3810 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3811 | |
b4be6aef JB |
3812 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3813 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3814 | wake_up_process(fs_info->cleaner_kthread); | |
3815 | ||
8cd29088 BB |
3816 | clear_oneshot: |
3817 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3818 | return 0; |
39279cc3 | 3819 | |
bcef60f2 AJ |
3820 | fail_qgroup: |
3821 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3822 | fail_trans_kthread: |
3823 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3824 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3825 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3826 | fail_cleaner: |
a74a4b97 | 3827 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3828 | |
3829 | /* | |
3830 | * make sure we're done with the btree inode before we stop our | |
3831 | * kthreads | |
3832 | */ | |
3833 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3834 | |
2365dd3c | 3835 | fail_sysfs: |
6618a59b | 3836 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3837 | |
b7c35e81 AJ |
3838 | fail_fsdev_sysfs: |
3839 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3840 | ||
1b1d1f66 | 3841 | fail_block_groups: |
54067ae9 | 3842 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3843 | |
3844 | fail_tree_roots: | |
9e3aa805 JB |
3845 | if (fs_info->data_reloc_root) |
3846 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3847 | free_root_pointers(fs_info, true); |
2b8195bb | 3848 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3849 | |
39279cc3 | 3850 | fail_sb_buffer: |
7abadb64 | 3851 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3852 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3853 | fail_alloc: |
586e46e2 ID |
3854 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3855 | ||
4543df7e | 3856 | iput(fs_info->btree_inode); |
7e662854 | 3857 | fail: |
586e46e2 | 3858 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3859 | return err; |
eb60ceac | 3860 | } |
663faf9f | 3861 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3862 | |
314b6dd0 | 3863 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3864 | { |
314b6dd0 JT |
3865 | struct btrfs_device *device = bio->bi_private; |
3866 | struct bio_vec *bvec; | |
3867 | struct bvec_iter_all iter_all; | |
3868 | struct page *page; | |
3869 | ||
3870 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3871 | page = bvec->bv_page; | |
3872 | ||
3873 | if (bio->bi_status) { | |
3874 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3875 | "lost page write due to IO error on %s (%d)", | |
3876 | rcu_str_deref(device->name), | |
3877 | blk_status_to_errno(bio->bi_status)); | |
3878 | ClearPageUptodate(page); | |
3879 | SetPageError(page); | |
3880 | btrfs_dev_stat_inc_and_print(device, | |
3881 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3882 | } else { | |
3883 | SetPageUptodate(page); | |
3884 | } | |
3885 | ||
3886 | put_page(page); | |
3887 | unlock_page(page); | |
f2984462 | 3888 | } |
314b6dd0 JT |
3889 | |
3890 | bio_put(bio); | |
f2984462 CM |
3891 | } |
3892 | ||
8f32380d | 3893 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3894 | int copy_num, bool drop_cache) |
29c36d72 | 3895 | { |
29c36d72 | 3896 | struct btrfs_super_block *super; |
8f32380d | 3897 | struct page *page; |
12659251 | 3898 | u64 bytenr, bytenr_orig; |
8f32380d | 3899 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3900 | int ret; |
3901 | ||
3902 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3903 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3904 | if (ret == -ENOENT) | |
3905 | return ERR_PTR(-EINVAL); | |
3906 | else if (ret) | |
3907 | return ERR_PTR(ret); | |
29c36d72 | 3908 | |
cda00eba | 3909 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3910 | return ERR_PTR(-EINVAL); |
29c36d72 | 3911 | |
a05d3c91 QW |
3912 | if (drop_cache) { |
3913 | /* This should only be called with the primary sb. */ | |
3914 | ASSERT(copy_num == 0); | |
3915 | ||
3916 | /* | |
3917 | * Drop the page of the primary superblock, so later read will | |
3918 | * always read from the device. | |
3919 | */ | |
3920 | invalidate_inode_pages2_range(mapping, | |
3921 | bytenr >> PAGE_SHIFT, | |
3922 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3923 | } | |
3924 | ||
8f32380d JT |
3925 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3926 | if (IS_ERR(page)) | |
3927 | return ERR_CAST(page); | |
29c36d72 | 3928 | |
8f32380d | 3929 | super = page_address(page); |
96c2e067 AJ |
3930 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3931 | btrfs_release_disk_super(super); | |
3932 | return ERR_PTR(-ENODATA); | |
3933 | } | |
3934 | ||
12659251 | 3935 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3936 | btrfs_release_disk_super(super); |
3937 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3938 | } |
3939 | ||
8f32380d | 3940 | return super; |
29c36d72 AJ |
3941 | } |
3942 | ||
3943 | ||
8f32380d | 3944 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3945 | { |
8f32380d | 3946 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3947 | int i; |
3948 | u64 transid = 0; | |
a512bbf8 YZ |
3949 | |
3950 | /* we would like to check all the supers, but that would make | |
3951 | * a btrfs mount succeed after a mkfs from a different FS. | |
3952 | * So, we need to add a special mount option to scan for | |
3953 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3954 | */ | |
3955 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3956 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3957 | if (IS_ERR(super)) |
a512bbf8 YZ |
3958 | continue; |
3959 | ||
a512bbf8 | 3960 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3961 | if (latest) |
3962 | btrfs_release_disk_super(super); | |
3963 | ||
3964 | latest = super; | |
a512bbf8 | 3965 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3966 | } |
3967 | } | |
92fc03fb | 3968 | |
8f32380d | 3969 | return super; |
a512bbf8 YZ |
3970 | } |
3971 | ||
4eedeb75 | 3972 | /* |
abbb3b8e | 3973 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3974 | * pages we use for writing are locked. |
4eedeb75 | 3975 | * |
abbb3b8e DS |
3976 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3977 | * the expected device size at commit time. Note that max_mirrors must be | |
3978 | * same for write and wait phases. | |
4eedeb75 | 3979 | * |
314b6dd0 | 3980 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3981 | */ |
a512bbf8 | 3982 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3983 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3984 | { |
d5178578 | 3985 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3986 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3987 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3988 | int i; |
a512bbf8 | 3989 | int errors = 0; |
12659251 NA |
3990 | int ret; |
3991 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3992 | |
3993 | if (max_mirrors == 0) | |
3994 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3995 | ||
d5178578 JT |
3996 | shash->tfm = fs_info->csum_shash; |
3997 | ||
a512bbf8 | 3998 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3999 | struct page *page; |
4000 | struct bio *bio; | |
4001 | struct btrfs_super_block *disk_super; | |
4002 | ||
12659251 NA |
4003 | bytenr_orig = btrfs_sb_offset(i); |
4004 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
4005 | if (ret == -ENOENT) { | |
4006 | continue; | |
4007 | } else if (ret < 0) { | |
4008 | btrfs_err(device->fs_info, | |
4009 | "couldn't get super block location for mirror %d", | |
4010 | i); | |
4011 | errors++; | |
4012 | continue; | |
4013 | } | |
935e5cc9 MX |
4014 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4015 | device->commit_total_bytes) | |
a512bbf8 YZ |
4016 | break; |
4017 | ||
12659251 | 4018 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 4019 | |
fd08001f EB |
4020 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
4021 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
4022 | sb->csum); | |
4eedeb75 | 4023 | |
314b6dd0 JT |
4024 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
4025 | GFP_NOFS); | |
4026 | if (!page) { | |
abbb3b8e | 4027 | btrfs_err(device->fs_info, |
314b6dd0 | 4028 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
4029 | bytenr); |
4030 | errors++; | |
4eedeb75 | 4031 | continue; |
abbb3b8e | 4032 | } |
634554dc | 4033 | |
314b6dd0 JT |
4034 | /* Bump the refcount for wait_dev_supers() */ |
4035 | get_page(page); | |
a512bbf8 | 4036 | |
314b6dd0 JT |
4037 | disk_super = page_address(page); |
4038 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 4039 | |
314b6dd0 JT |
4040 | /* |
4041 | * Directly use bios here instead of relying on the page cache | |
4042 | * to do I/O, so we don't lose the ability to do integrity | |
4043 | * checking. | |
4044 | */ | |
07888c66 CH |
4045 | bio = bio_alloc(device->bdev, 1, |
4046 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
4047 | GFP_NOFS); | |
314b6dd0 JT |
4048 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
4049 | bio->bi_private = device; | |
4050 | bio->bi_end_io = btrfs_end_super_write; | |
4051 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4052 | offset_in_page(bytenr)); | |
a512bbf8 | 4053 | |
387125fc | 4054 | /* |
314b6dd0 JT |
4055 | * We FUA only the first super block. The others we allow to |
4056 | * go down lazy and there's a short window where the on-disk | |
4057 | * copies might still contain the older version. | |
387125fc | 4058 | */ |
1b9e619c | 4059 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4060 | bio->bi_opf |= REQ_FUA; |
4061 | ||
58ff51f1 CH |
4062 | btrfsic_check_bio(bio); |
4063 | submit_bio(bio); | |
8376d9e1 NA |
4064 | |
4065 | if (btrfs_advance_sb_log(device, i)) | |
4066 | errors++; | |
a512bbf8 YZ |
4067 | } |
4068 | return errors < i ? 0 : -1; | |
4069 | } | |
4070 | ||
abbb3b8e DS |
4071 | /* |
4072 | * Wait for write completion of superblocks done by write_dev_supers, | |
4073 | * @max_mirrors same for write and wait phases. | |
4074 | * | |
314b6dd0 | 4075 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4076 | * date. |
4077 | */ | |
4078 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4079 | { | |
abbb3b8e DS |
4080 | int i; |
4081 | int errors = 0; | |
b6a535fa | 4082 | bool primary_failed = false; |
12659251 | 4083 | int ret; |
abbb3b8e DS |
4084 | u64 bytenr; |
4085 | ||
4086 | if (max_mirrors == 0) | |
4087 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4088 | ||
4089 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4090 | struct page *page; |
4091 | ||
12659251 NA |
4092 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4093 | if (ret == -ENOENT) { | |
4094 | break; | |
4095 | } else if (ret < 0) { | |
4096 | errors++; | |
4097 | if (i == 0) | |
4098 | primary_failed = true; | |
4099 | continue; | |
4100 | } | |
abbb3b8e DS |
4101 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4102 | device->commit_total_bytes) | |
4103 | break; | |
4104 | ||
314b6dd0 JT |
4105 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4106 | bytenr >> PAGE_SHIFT); | |
4107 | if (!page) { | |
abbb3b8e | 4108 | errors++; |
b6a535fa HM |
4109 | if (i == 0) |
4110 | primary_failed = true; | |
abbb3b8e DS |
4111 | continue; |
4112 | } | |
314b6dd0 JT |
4113 | /* Page is submitted locked and unlocked once the IO completes */ |
4114 | wait_on_page_locked(page); | |
4115 | if (PageError(page)) { | |
abbb3b8e | 4116 | errors++; |
b6a535fa HM |
4117 | if (i == 0) |
4118 | primary_failed = true; | |
4119 | } | |
abbb3b8e | 4120 | |
314b6dd0 JT |
4121 | /* Drop our reference */ |
4122 | put_page(page); | |
abbb3b8e | 4123 | |
314b6dd0 JT |
4124 | /* Drop the reference from the writing run */ |
4125 | put_page(page); | |
abbb3b8e DS |
4126 | } |
4127 | ||
b6a535fa HM |
4128 | /* log error, force error return */ |
4129 | if (primary_failed) { | |
4130 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4131 | device->devid); | |
4132 | return -1; | |
4133 | } | |
4134 | ||
abbb3b8e DS |
4135 | return errors < i ? 0 : -1; |
4136 | } | |
4137 | ||
387125fc CM |
4138 | /* |
4139 | * endio for the write_dev_flush, this will wake anyone waiting | |
4140 | * for the barrier when it is done | |
4141 | */ | |
4246a0b6 | 4142 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4143 | { |
f9e69aa9 | 4144 | bio_uninit(bio); |
e0ae9994 | 4145 | complete(bio->bi_private); |
387125fc CM |
4146 | } |
4147 | ||
4148 | /* | |
4fc6441a AJ |
4149 | * Submit a flush request to the device if it supports it. Error handling is |
4150 | * done in the waiting counterpart. | |
387125fc | 4151 | */ |
4fc6441a | 4152 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4153 | { |
f9e69aa9 | 4154 | struct bio *bio = &device->flush_bio; |
387125fc | 4155 | |
a91cf0ff WY |
4156 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4157 | /* | |
4158 | * When a disk has write caching disabled, we skip submission of a bio | |
4159 | * with flush and sync requests before writing the superblock, since | |
4160 | * it's not needed. However when the integrity checker is enabled, this | |
4161 | * results in reports that there are metadata blocks referred by a | |
4162 | * superblock that were not properly flushed. So don't skip the bio | |
4163 | * submission only when the integrity checker is enabled for the sake | |
4164 | * of simplicity, since this is a debug tool and not meant for use in | |
4165 | * non-debug builds. | |
4166 | */ | |
08e688fd | 4167 | if (!bdev_write_cache(device->bdev)) |
4fc6441a | 4168 | return; |
a91cf0ff | 4169 | #endif |
387125fc | 4170 | |
f9e69aa9 CH |
4171 | bio_init(bio, device->bdev, NULL, 0, |
4172 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 4173 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
4174 | init_completion(&device->flush_wait); |
4175 | bio->bi_private = &device->flush_wait; | |
387125fc | 4176 | |
58ff51f1 CH |
4177 | btrfsic_check_bio(bio); |
4178 | submit_bio(bio); | |
1c3063b6 | 4179 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4180 | } |
387125fc | 4181 | |
4fc6441a AJ |
4182 | /* |
4183 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4184 | */ | |
8c27cb35 | 4185 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4186 | { |
f9e69aa9 | 4187 | struct bio *bio = &device->flush_bio; |
387125fc | 4188 | |
1c3063b6 | 4189 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4190 | return BLK_STS_OK; |
387125fc | 4191 | |
1c3063b6 | 4192 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4193 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4194 | |
8c27cb35 | 4195 | return bio->bi_status; |
387125fc | 4196 | } |
387125fc | 4197 | |
d10b82fe | 4198 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4199 | { |
6528b99d | 4200 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4201 | return -EIO; |
387125fc CM |
4202 | return 0; |
4203 | } | |
4204 | ||
4205 | /* | |
4206 | * send an empty flush down to each device in parallel, | |
4207 | * then wait for them | |
4208 | */ | |
4209 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4210 | { | |
4211 | struct list_head *head; | |
4212 | struct btrfs_device *dev; | |
5af3e8cc | 4213 | int errors_wait = 0; |
4e4cbee9 | 4214 | blk_status_t ret; |
387125fc | 4215 | |
1538e6c5 | 4216 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4217 | /* send down all the barriers */ |
4218 | head = &info->fs_devices->devices; | |
1538e6c5 | 4219 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4220 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4221 | continue; |
cea7c8bf | 4222 | if (!dev->bdev) |
387125fc | 4223 | continue; |
e12c9621 | 4224 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4225 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4226 | continue; |
4227 | ||
4fc6441a | 4228 | write_dev_flush(dev); |
58efbc9f | 4229 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4230 | } |
4231 | ||
4232 | /* wait for all the barriers */ | |
1538e6c5 | 4233 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4234 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4235 | continue; |
387125fc | 4236 | if (!dev->bdev) { |
5af3e8cc | 4237 | errors_wait++; |
387125fc CM |
4238 | continue; |
4239 | } | |
e12c9621 | 4240 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4241 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4242 | continue; |
4243 | ||
4fc6441a | 4244 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4245 | if (ret) { |
4246 | dev->last_flush_error = ret; | |
66b4993e DS |
4247 | btrfs_dev_stat_inc_and_print(dev, |
4248 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4249 | errors_wait++; |
401b41e5 AJ |
4250 | } |
4251 | } | |
4252 | ||
cea7c8bf | 4253 | if (errors_wait) { |
401b41e5 AJ |
4254 | /* |
4255 | * At some point we need the status of all disks | |
4256 | * to arrive at the volume status. So error checking | |
4257 | * is being pushed to a separate loop. | |
4258 | */ | |
d10b82fe | 4259 | return check_barrier_error(info); |
387125fc | 4260 | } |
387125fc CM |
4261 | return 0; |
4262 | } | |
4263 | ||
943c6e99 ZL |
4264 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4265 | { | |
8789f4fe ZL |
4266 | int raid_type; |
4267 | int min_tolerated = INT_MAX; | |
943c6e99 | 4268 | |
8789f4fe ZL |
4269 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4270 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4271 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4272 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4273 | tolerated_failures); | |
943c6e99 | 4274 | |
8789f4fe ZL |
4275 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4276 | if (raid_type == BTRFS_RAID_SINGLE) | |
4277 | continue; | |
41a6e891 | 4278 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4279 | continue; |
8c3e3582 | 4280 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4281 | btrfs_raid_array[raid_type]. |
4282 | tolerated_failures); | |
4283 | } | |
943c6e99 | 4284 | |
8789f4fe | 4285 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4286 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4287 | min_tolerated = 0; |
4288 | } | |
4289 | ||
4290 | return min_tolerated; | |
943c6e99 ZL |
4291 | } |
4292 | ||
eece6a9c | 4293 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4294 | { |
e5e9a520 | 4295 | struct list_head *head; |
f2984462 | 4296 | struct btrfs_device *dev; |
a061fc8d | 4297 | struct btrfs_super_block *sb; |
f2984462 | 4298 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4299 | int ret; |
4300 | int do_barriers; | |
a236aed1 CM |
4301 | int max_errors; |
4302 | int total_errors = 0; | |
a061fc8d | 4303 | u64 flags; |
f2984462 | 4304 | |
0b246afa | 4305 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4306 | |
4307 | /* | |
4308 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4309 | * not from fsync where the tree roots in fs_info have not | |
4310 | * been consistent on disk. | |
4311 | */ | |
4312 | if (max_mirrors == 0) | |
4313 | backup_super_roots(fs_info); | |
f2984462 | 4314 | |
0b246afa | 4315 | sb = fs_info->super_for_commit; |
a061fc8d | 4316 | dev_item = &sb->dev_item; |
e5e9a520 | 4317 | |
0b246afa JM |
4318 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4319 | head = &fs_info->fs_devices->devices; | |
4320 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4321 | |
5af3e8cc | 4322 | if (do_barriers) { |
0b246afa | 4323 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4324 | if (ret) { |
4325 | mutex_unlock( | |
0b246afa JM |
4326 | &fs_info->fs_devices->device_list_mutex); |
4327 | btrfs_handle_fs_error(fs_info, ret, | |
4328 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4329 | return ret; |
4330 | } | |
4331 | } | |
387125fc | 4332 | |
1538e6c5 | 4333 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4334 | if (!dev->bdev) { |
4335 | total_errors++; | |
4336 | continue; | |
4337 | } | |
e12c9621 | 4338 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4339 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4340 | continue; |
4341 | ||
2b82032c | 4342 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4343 | btrfs_set_stack_device_type(dev_item, dev->type); |
4344 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4345 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4346 | dev->commit_total_bytes); |
ce7213c7 MX |
4347 | btrfs_set_stack_device_bytes_used(dev_item, |
4348 | dev->commit_bytes_used); | |
a061fc8d CM |
4349 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4350 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4351 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4352 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4353 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4354 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4355 | |
a061fc8d CM |
4356 | flags = btrfs_super_flags(sb); |
4357 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4358 | ||
75cb857d QW |
4359 | ret = btrfs_validate_write_super(fs_info, sb); |
4360 | if (ret < 0) { | |
4361 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4362 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4363 | "unexpected superblock corruption detected"); | |
4364 | return -EUCLEAN; | |
4365 | } | |
4366 | ||
abbb3b8e | 4367 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4368 | if (ret) |
4369 | total_errors++; | |
f2984462 | 4370 | } |
a236aed1 | 4371 | if (total_errors > max_errors) { |
0b246afa JM |
4372 | btrfs_err(fs_info, "%d errors while writing supers", |
4373 | total_errors); | |
4374 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4375 | |
9d565ba4 | 4376 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4377 | btrfs_handle_fs_error(fs_info, -EIO, |
4378 | "%d errors while writing supers", | |
4379 | total_errors); | |
9d565ba4 | 4380 | return -EIO; |
a236aed1 | 4381 | } |
f2984462 | 4382 | |
a512bbf8 | 4383 | total_errors = 0; |
1538e6c5 | 4384 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4385 | if (!dev->bdev) |
4386 | continue; | |
e12c9621 | 4387 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4388 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4389 | continue; |
4390 | ||
abbb3b8e | 4391 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4392 | if (ret) |
4393 | total_errors++; | |
f2984462 | 4394 | } |
0b246afa | 4395 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4396 | if (total_errors > max_errors) { |
0b246afa JM |
4397 | btrfs_handle_fs_error(fs_info, -EIO, |
4398 | "%d errors while writing supers", | |
4399 | total_errors); | |
79787eaa | 4400 | return -EIO; |
a236aed1 | 4401 | } |
f2984462 CM |
4402 | return 0; |
4403 | } | |
4404 | ||
cb517eab MX |
4405 | /* Drop a fs root from the radix tree and free it. */ |
4406 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4407 | struct btrfs_root *root) | |
2619ba1f | 4408 | { |
4785e24f JB |
4409 | bool drop_ref = false; |
4410 | ||
fc7cbcd4 DS |
4411 | spin_lock(&fs_info->fs_roots_radix_lock); |
4412 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4413 | (unsigned long)root->root_key.objectid); | |
4414 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4415 | drop_ref = true; |
fc7cbcd4 | 4416 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4417 | |
84961539 | 4418 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4419 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4420 | if (root->reloc_root) { |
00246528 | 4421 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4422 | root->reloc_root = NULL; |
4423 | } | |
4424 | } | |
3321719e | 4425 | |
4785e24f JB |
4426 | if (drop_ref) |
4427 | btrfs_put_root(root); | |
2619ba1f CM |
4428 | } |
4429 | ||
c146afad | 4430 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4431 | { |
fc7cbcd4 DS |
4432 | u64 root_objectid = 0; |
4433 | struct btrfs_root *gang[8]; | |
4434 | int i = 0; | |
65d33fd7 | 4435 | int err = 0; |
fc7cbcd4 | 4436 | unsigned int ret = 0; |
e089f05c | 4437 | |
c146afad | 4438 | while (1) { |
fc7cbcd4 DS |
4439 | spin_lock(&fs_info->fs_roots_radix_lock); |
4440 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4441 | (void **)gang, root_objectid, | |
4442 | ARRAY_SIZE(gang)); | |
4443 | if (!ret) { | |
4444 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4445 | break; | |
65d33fd7 | 4446 | } |
fc7cbcd4 | 4447 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4448 | |
fc7cbcd4 DS |
4449 | for (i = 0; i < ret; i++) { |
4450 | /* Avoid to grab roots in dead_roots */ | |
4451 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4452 | gang[i] = NULL; | |
4453 | continue; | |
4454 | } | |
4455 | /* grab all the search result for later use */ | |
4456 | gang[i] = btrfs_grab_root(gang[i]); | |
65d33fd7 | 4457 | } |
fc7cbcd4 | 4458 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4459 | |
fc7cbcd4 DS |
4460 | for (i = 0; i < ret; i++) { |
4461 | if (!gang[i]) | |
65d33fd7 | 4462 | continue; |
fc7cbcd4 DS |
4463 | root_objectid = gang[i]->root_key.objectid; |
4464 | err = btrfs_orphan_cleanup(gang[i]); | |
66b4ffd1 | 4465 | if (err) |
fc7cbcd4 DS |
4466 | break; |
4467 | btrfs_put_root(gang[i]); | |
c146afad | 4468 | } |
fc7cbcd4 | 4469 | root_objectid++; |
c146afad | 4470 | } |
65d33fd7 | 4471 | |
fc7cbcd4 DS |
4472 | /* release the uncleaned roots due to error */ |
4473 | for (; i < ret; i++) { | |
4474 | if (gang[i]) | |
4475 | btrfs_put_root(gang[i]); | |
65d33fd7 QW |
4476 | } |
4477 | return err; | |
c146afad | 4478 | } |
a2135011 | 4479 | |
6bccf3ab | 4480 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4481 | { |
6bccf3ab | 4482 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4483 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4484 | |
0b246afa | 4485 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4486 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4487 | mutex_unlock(&fs_info->cleaner_mutex); |
4488 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4489 | |
4490 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4491 | down_write(&fs_info->cleanup_work_sem); |
4492 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4493 | |
7a7eaa40 | 4494 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4495 | if (IS_ERR(trans)) |
4496 | return PTR_ERR(trans); | |
3a45bb20 | 4497 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4498 | } |
4499 | ||
36c86a9e QW |
4500 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4501 | { | |
4502 | struct btrfs_transaction *trans; | |
4503 | struct btrfs_transaction *tmp; | |
4504 | bool found = false; | |
4505 | ||
4506 | if (list_empty(&fs_info->trans_list)) | |
4507 | return; | |
4508 | ||
4509 | /* | |
4510 | * This function is only called at the very end of close_ctree(), | |
4511 | * thus no other running transaction, no need to take trans_lock. | |
4512 | */ | |
4513 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4514 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4515 | struct extent_state *cached = NULL; | |
4516 | u64 dirty_bytes = 0; | |
4517 | u64 cur = 0; | |
4518 | u64 found_start; | |
4519 | u64 found_end; | |
4520 | ||
4521 | found = true; | |
4522 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4523 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4524 | dirty_bytes += found_end + 1 - found_start; | |
4525 | cur = found_end + 1; | |
4526 | } | |
4527 | btrfs_warn(fs_info, | |
4528 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4529 | trans->transid, dirty_bytes); | |
4530 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4531 | ||
4532 | if (trans == fs_info->running_transaction) | |
4533 | fs_info->running_transaction = NULL; | |
4534 | list_del_init(&trans->list); | |
4535 | ||
4536 | btrfs_put_transaction(trans); | |
4537 | trace_btrfs_transaction_commit(fs_info); | |
4538 | } | |
4539 | ASSERT(!found); | |
4540 | } | |
4541 | ||
b105e927 | 4542 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4543 | { |
c146afad YZ |
4544 | int ret; |
4545 | ||
afcdd129 | 4546 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4547 | |
8a1f1e3d FM |
4548 | /* |
4549 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4550 | * clear that bit and wake up relocation so it can stop. | |
4551 | * We must do this before stopping the block group reclaim task, because | |
4552 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4553 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4554 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4555 | * return 1. | |
4556 | */ | |
4557 | btrfs_wake_unfinished_drop(fs_info); | |
4558 | ||
31e70e52 FM |
4559 | /* |
4560 | * We may have the reclaim task running and relocating a data block group, | |
4561 | * in which case it may create delayed iputs. So stop it before we park | |
4562 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4563 | * parking the cleaner, and that can make the async reclaim task to hang | |
4564 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4565 | * parked and can not run delayed iputs - this will make us hang when | |
4566 | * trying to stop the async reclaim task. | |
4567 | */ | |
4568 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4569 | /* |
4570 | * We don't want the cleaner to start new transactions, add more delayed | |
4571 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4572 | * because that frees the task_struct, and the transaction kthread might | |
4573 | * still try to wake up the cleaner. | |
4574 | */ | |
4575 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4576 | |
7343dd61 | 4577 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4578 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4579 | |
803b2f54 SB |
4580 | /* wait for the uuid_scan task to finish */ |
4581 | down(&fs_info->uuid_tree_rescan_sem); | |
4582 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4583 | up(&fs_info->uuid_tree_rescan_sem); | |
4584 | ||
837d5b6e | 4585 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4586 | btrfs_pause_balance(fs_info); |
837d5b6e | 4587 | |
8dabb742 SB |
4588 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4589 | ||
aa1b8cd4 | 4590 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4591 | |
4592 | /* wait for any defraggers to finish */ | |
4593 | wait_event(fs_info->transaction_wait, | |
4594 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4595 | ||
4596 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4597 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4598 | |
a362bb86 FM |
4599 | /* |
4600 | * After we parked the cleaner kthread, ordered extents may have | |
4601 | * completed and created new delayed iputs. If one of the async reclaim | |
4602 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4603 | * can hang forever trying to stop it, because if a delayed iput is | |
4604 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4605 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4606 | * no one else to run iputs. | |
4607 | * | |
4608 | * So wait for all ongoing ordered extents to complete and then run | |
4609 | * delayed iputs. This works because once we reach this point no one | |
4610 | * can either create new ordered extents nor create delayed iputs | |
4611 | * through some other means. | |
4612 | * | |
4613 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4614 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4615 | * but the delayed iput for the respective inode is made only when doing | |
4616 | * the final btrfs_put_ordered_extent() (which must happen at | |
4617 | * btrfs_finish_ordered_io() when we are unmounting). | |
4618 | */ | |
4619 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4620 | /* Ordered extents for free space inodes. */ | |
4621 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4622 | btrfs_run_delayed_iputs(fs_info); | |
4623 | ||
21c7e756 | 4624 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4625 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4626 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4627 | |
b0643e59 DZ |
4628 | /* Cancel or finish ongoing discard work */ |
4629 | btrfs_discard_cleanup(fs_info); | |
4630 | ||
bc98a42c | 4631 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4632 | /* |
d6fd0ae2 OS |
4633 | * The cleaner kthread is stopped, so do one final pass over |
4634 | * unused block groups. | |
e44163e1 | 4635 | */ |
0b246afa | 4636 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4637 | |
f0cc2cd7 FM |
4638 | /* |
4639 | * There might be existing delayed inode workers still running | |
4640 | * and holding an empty delayed inode item. We must wait for | |
4641 | * them to complete first because they can create a transaction. | |
4642 | * This happens when someone calls btrfs_balance_delayed_items() | |
4643 | * and then a transaction commit runs the same delayed nodes | |
4644 | * before any delayed worker has done something with the nodes. | |
4645 | * We must wait for any worker here and not at transaction | |
4646 | * commit time since that could cause a deadlock. | |
4647 | * This is a very rare case. | |
4648 | */ | |
4649 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4650 | ||
6bccf3ab | 4651 | ret = btrfs_commit_super(fs_info); |
acce952b | 4652 | if (ret) |
04892340 | 4653 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4654 | } |
4655 | ||
84961539 | 4656 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4657 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4658 | |
e3029d9f AV |
4659 | kthread_stop(fs_info->transaction_kthread); |
4660 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4661 | |
e187831e | 4662 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4663 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4664 | |
5958253c QW |
4665 | if (btrfs_check_quota_leak(fs_info)) { |
4666 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4667 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4668 | } | |
4669 | ||
04892340 | 4670 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4671 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4672 | |
963d678b | 4673 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4674 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4675 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4676 | } |
bcc63abb | 4677 | |
5deb17e1 | 4678 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4679 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4680 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4681 | |
6618a59b | 4682 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4683 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4684 | |
1a4319cc LB |
4685 | btrfs_put_block_group_cache(fs_info); |
4686 | ||
de348ee0 WS |
4687 | /* |
4688 | * we must make sure there is not any read request to | |
4689 | * submit after we stopping all workers. | |
4690 | */ | |
4691 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4692 | btrfs_stop_all_workers(fs_info); |
4693 | ||
0a31daa4 | 4694 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4695 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4696 | |
afcdd129 | 4697 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4698 | free_root_pointers(fs_info, true); |
8c38938c | 4699 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4700 | |
4e19443d JB |
4701 | /* |
4702 | * We must free the block groups after dropping the fs_roots as we could | |
4703 | * have had an IO error and have left over tree log blocks that aren't | |
4704 | * cleaned up until the fs roots are freed. This makes the block group | |
4705 | * accounting appear to be wrong because there's pending reserved bytes, | |
4706 | * so make sure we do the block group cleanup afterwards. | |
4707 | */ | |
4708 | btrfs_free_block_groups(fs_info); | |
4709 | ||
13e6c37b | 4710 | iput(fs_info->btree_inode); |
d6bfde87 | 4711 | |
21adbd5c | 4712 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4713 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4714 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4715 | #endif |
4716 | ||
0b86a832 | 4717 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4718 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4719 | } |
4720 | ||
b9fab919 CM |
4721 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4722 | int atomic) | |
5f39d397 | 4723 | { |
1259ab75 | 4724 | int ret; |
727011e0 | 4725 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4726 | |
0b32f4bb | 4727 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4728 | if (!ret) |
4729 | return ret; | |
4730 | ||
4731 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4732 | parent_transid, atomic); |
4733 | if (ret == -EAGAIN) | |
4734 | return ret; | |
1259ab75 | 4735 | return !ret; |
5f39d397 CM |
4736 | } |
4737 | ||
5f39d397 CM |
4738 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4739 | { | |
2f4d60df | 4740 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4741 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4742 | int was_dirty; |
b4ce94de | 4743 | |
06ea65a3 JB |
4744 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4745 | /* | |
4746 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4747 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4748 | * outside of the sanity tests. |
4749 | */ | |
b0132a3b | 4750 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4751 | return; |
4752 | #endif | |
49d0c642 | 4753 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4754 | if (transid != fs_info->generation) |
5d163e0e | 4755 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4756 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4757 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4758 | if (!was_dirty) |
104b4e51 NB |
4759 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4760 | buf->len, | |
4761 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4762 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4763 | /* |
4764 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4765 | * but item data not updated. | |
4766 | * So here we should only check item pointers, not item data. | |
4767 | */ | |
4768 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4769 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4770 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4771 | ASSERT(0); |
4772 | } | |
4773 | #endif | |
eb60ceac CM |
4774 | } |
4775 | ||
2ff7e61e | 4776 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4777 | int flush_delayed) |
16cdcec7 MX |
4778 | { |
4779 | /* | |
4780 | * looks as though older kernels can get into trouble with | |
4781 | * this code, they end up stuck in balance_dirty_pages forever | |
4782 | */ | |
e2d84521 | 4783 | int ret; |
16cdcec7 MX |
4784 | |
4785 | if (current->flags & PF_MEMALLOC) | |
4786 | return; | |
4787 | ||
b53d3f5d | 4788 | if (flush_delayed) |
2ff7e61e | 4789 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4790 | |
d814a491 EL |
4791 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4792 | BTRFS_DIRTY_METADATA_THRESH, | |
4793 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4794 | if (ret > 0) { |
0b246afa | 4795 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4796 | } |
16cdcec7 MX |
4797 | } |
4798 | ||
2ff7e61e | 4799 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4800 | { |
2ff7e61e | 4801 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4802 | } |
585ad2c3 | 4803 | |
2ff7e61e | 4804 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4805 | { |
2ff7e61e | 4806 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4807 | } |
6b80053d | 4808 | |
2ff7e61e | 4809 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4810 | { |
fe816d0f NB |
4811 | /* cleanup FS via transaction */ |
4812 | btrfs_cleanup_transaction(fs_info); | |
4813 | ||
0b246afa | 4814 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4815 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4816 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4817 | |
0b246afa JM |
4818 | down_write(&fs_info->cleanup_work_sem); |
4819 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4820 | } |
4821 | ||
ef67963d JB |
4822 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4823 | { | |
fc7cbcd4 DS |
4824 | struct btrfs_root *gang[8]; |
4825 | u64 root_objectid = 0; | |
4826 | int ret; | |
4827 | ||
4828 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4829 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4830 | (void **)gang, root_objectid, | |
4831 | ARRAY_SIZE(gang))) != 0) { | |
4832 | int i; | |
4833 | ||
4834 | for (i = 0; i < ret; i++) | |
4835 | gang[i] = btrfs_grab_root(gang[i]); | |
4836 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4837 | ||
4838 | for (i = 0; i < ret; i++) { | |
4839 | if (!gang[i]) | |
ef67963d | 4840 | continue; |
fc7cbcd4 DS |
4841 | root_objectid = gang[i]->root_key.objectid; |
4842 | btrfs_free_log(NULL, gang[i]); | |
4843 | btrfs_put_root(gang[i]); | |
ef67963d | 4844 | } |
fc7cbcd4 DS |
4845 | root_objectid++; |
4846 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4847 | } |
fc7cbcd4 | 4848 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4849 | btrfs_free_log_root_tree(NULL, fs_info); |
4850 | } | |
4851 | ||
143bede5 | 4852 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4853 | { |
acce952b | 4854 | struct btrfs_ordered_extent *ordered; |
acce952b | 4855 | |
199c2a9c | 4856 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4857 | /* |
4858 | * This will just short circuit the ordered completion stuff which will | |
4859 | * make sure the ordered extent gets properly cleaned up. | |
4860 | */ | |
199c2a9c | 4861 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4862 | root_extent_list) |
4863 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4864 | spin_unlock(&root->ordered_extent_lock); |
4865 | } | |
4866 | ||
4867 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4868 | { | |
4869 | struct btrfs_root *root; | |
4870 | struct list_head splice; | |
4871 | ||
4872 | INIT_LIST_HEAD(&splice); | |
4873 | ||
4874 | spin_lock(&fs_info->ordered_root_lock); | |
4875 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4876 | while (!list_empty(&splice)) { | |
4877 | root = list_first_entry(&splice, struct btrfs_root, | |
4878 | ordered_root); | |
1de2cfde JB |
4879 | list_move_tail(&root->ordered_root, |
4880 | &fs_info->ordered_roots); | |
199c2a9c | 4881 | |
2a85d9ca | 4882 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4883 | btrfs_destroy_ordered_extents(root); |
4884 | ||
2a85d9ca LB |
4885 | cond_resched(); |
4886 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4887 | } |
4888 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4889 | |
4890 | /* | |
4891 | * We need this here because if we've been flipped read-only we won't | |
4892 | * get sync() from the umount, so we need to make sure any ordered | |
4893 | * extents that haven't had their dirty pages IO start writeout yet | |
4894 | * actually get run and error out properly. | |
4895 | */ | |
4896 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4897 | } |
4898 | ||
35a3621b | 4899 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4900 | struct btrfs_fs_info *fs_info) |
acce952b | 4901 | { |
4902 | struct rb_node *node; | |
4903 | struct btrfs_delayed_ref_root *delayed_refs; | |
4904 | struct btrfs_delayed_ref_node *ref; | |
4905 | int ret = 0; | |
4906 | ||
4907 | delayed_refs = &trans->delayed_refs; | |
4908 | ||
4909 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4910 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4911 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4912 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4913 | return ret; |
4914 | } | |
4915 | ||
5c9d028b | 4916 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4917 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4918 | struct rb_node *n; |
e78417d1 | 4919 | bool pin_bytes = false; |
acce952b | 4920 | |
d7df2c79 JB |
4921 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4922 | href_node); | |
3069bd26 | 4923 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4924 | continue; |
3069bd26 | 4925 | |
d7df2c79 | 4926 | spin_lock(&head->lock); |
e3d03965 | 4927 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4928 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4929 | ref_node); | |
d7df2c79 | 4930 | ref->in_tree = 0; |
e3d03965 | 4931 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4932 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4933 | if (!list_empty(&ref->add_list)) |
4934 | list_del(&ref->add_list); | |
d7df2c79 JB |
4935 | atomic_dec(&delayed_refs->num_entries); |
4936 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4937 | } |
d7df2c79 JB |
4938 | if (head->must_insert_reserved) |
4939 | pin_bytes = true; | |
4940 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4941 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4942 | spin_unlock(&head->lock); |
4943 | spin_unlock(&delayed_refs->lock); | |
4944 | mutex_unlock(&head->mutex); | |
acce952b | 4945 | |
f603bb94 NB |
4946 | if (pin_bytes) { |
4947 | struct btrfs_block_group *cache; | |
4948 | ||
4949 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4950 | BUG_ON(!cache); | |
4951 | ||
4952 | spin_lock(&cache->space_info->lock); | |
4953 | spin_lock(&cache->lock); | |
4954 | cache->pinned += head->num_bytes; | |
4955 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4956 | cache->space_info, head->num_bytes); | |
4957 | cache->reserved -= head->num_bytes; | |
4958 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4959 | spin_unlock(&cache->lock); | |
4960 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4961 | |
4962 | btrfs_put_block_group(cache); | |
4963 | ||
4964 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4965 | head->bytenr + head->num_bytes - 1); | |
4966 | } | |
31890da0 | 4967 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4968 | btrfs_put_delayed_ref_head(head); |
acce952b | 4969 | cond_resched(); |
4970 | spin_lock(&delayed_refs->lock); | |
4971 | } | |
81f7eb00 | 4972 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4973 | |
4974 | spin_unlock(&delayed_refs->lock); | |
4975 | ||
4976 | return ret; | |
4977 | } | |
4978 | ||
143bede5 | 4979 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4980 | { |
4981 | struct btrfs_inode *btrfs_inode; | |
4982 | struct list_head splice; | |
4983 | ||
4984 | INIT_LIST_HEAD(&splice); | |
4985 | ||
eb73c1b7 MX |
4986 | spin_lock(&root->delalloc_lock); |
4987 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4988 | |
4989 | while (!list_empty(&splice)) { | |
fe816d0f | 4990 | struct inode *inode = NULL; |
eb73c1b7 MX |
4991 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4992 | delalloc_inodes); | |
fe816d0f | 4993 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4994 | spin_unlock(&root->delalloc_lock); |
acce952b | 4995 | |
fe816d0f NB |
4996 | /* |
4997 | * Make sure we get a live inode and that it'll not disappear | |
4998 | * meanwhile. | |
4999 | */ | |
5000 | inode = igrab(&btrfs_inode->vfs_inode); | |
5001 | if (inode) { | |
5002 | invalidate_inode_pages2(inode->i_mapping); | |
5003 | iput(inode); | |
5004 | } | |
eb73c1b7 | 5005 | spin_lock(&root->delalloc_lock); |
acce952b | 5006 | } |
eb73c1b7 MX |
5007 | spin_unlock(&root->delalloc_lock); |
5008 | } | |
5009 | ||
5010 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
5011 | { | |
5012 | struct btrfs_root *root; | |
5013 | struct list_head splice; | |
5014 | ||
5015 | INIT_LIST_HEAD(&splice); | |
5016 | ||
5017 | spin_lock(&fs_info->delalloc_root_lock); | |
5018 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
5019 | while (!list_empty(&splice)) { | |
5020 | root = list_first_entry(&splice, struct btrfs_root, | |
5021 | delalloc_root); | |
00246528 | 5022 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
5023 | BUG_ON(!root); |
5024 | spin_unlock(&fs_info->delalloc_root_lock); | |
5025 | ||
5026 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 5027 | btrfs_put_root(root); |
eb73c1b7 MX |
5028 | |
5029 | spin_lock(&fs_info->delalloc_root_lock); | |
5030 | } | |
5031 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 5032 | } |
5033 | ||
2ff7e61e | 5034 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 5035 | struct extent_io_tree *dirty_pages, |
5036 | int mark) | |
5037 | { | |
5038 | int ret; | |
acce952b | 5039 | struct extent_buffer *eb; |
5040 | u64 start = 0; | |
5041 | u64 end; | |
acce952b | 5042 | |
5043 | while (1) { | |
5044 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5045 | mark, NULL); |
acce952b | 5046 | if (ret) |
5047 | break; | |
5048 | ||
91166212 | 5049 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5050 | while (start <= end) { |
0b246afa JM |
5051 | eb = find_extent_buffer(fs_info, start); |
5052 | start += fs_info->nodesize; | |
fd8b2b61 | 5053 | if (!eb) |
acce952b | 5054 | continue; |
fd8b2b61 | 5055 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5056 | |
fd8b2b61 JB |
5057 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5058 | &eb->bflags)) | |
5059 | clear_extent_buffer_dirty(eb); | |
5060 | free_extent_buffer_stale(eb); | |
acce952b | 5061 | } |
5062 | } | |
5063 | ||
5064 | return ret; | |
5065 | } | |
5066 | ||
2ff7e61e | 5067 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5068 | struct extent_io_tree *unpin) |
acce952b | 5069 | { |
acce952b | 5070 | u64 start; |
5071 | u64 end; | |
5072 | int ret; | |
5073 | ||
acce952b | 5074 | while (1) { |
0e6ec385 FM |
5075 | struct extent_state *cached_state = NULL; |
5076 | ||
fcd5e742 LF |
5077 | /* |
5078 | * The btrfs_finish_extent_commit() may get the same range as | |
5079 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5080 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5081 | * the same extent range. | |
5082 | */ | |
5083 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5084 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5085 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5086 | if (ret) { |
5087 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5088 | break; |
fcd5e742 | 5089 | } |
acce952b | 5090 | |
0e6ec385 FM |
5091 | clear_extent_dirty(unpin, start, end, &cached_state); |
5092 | free_extent_state(cached_state); | |
2ff7e61e | 5093 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5094 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5095 | cond_resched(); |
5096 | } | |
5097 | ||
5098 | return 0; | |
5099 | } | |
5100 | ||
32da5386 | 5101 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5102 | { |
5103 | struct inode *inode; | |
5104 | ||
5105 | inode = cache->io_ctl.inode; | |
5106 | if (inode) { | |
5107 | invalidate_inode_pages2(inode->i_mapping); | |
5108 | BTRFS_I(inode)->generation = 0; | |
5109 | cache->io_ctl.inode = NULL; | |
5110 | iput(inode); | |
5111 | } | |
bbc37d6e | 5112 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5113 | btrfs_put_block_group(cache); |
5114 | } | |
5115 | ||
5116 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5117 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5118 | { |
32da5386 | 5119 | struct btrfs_block_group *cache; |
c79a1751 LB |
5120 | |
5121 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5122 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5123 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5124 | struct btrfs_block_group, |
c79a1751 | 5125 | dirty_list); |
c79a1751 LB |
5126 | |
5127 | if (!list_empty(&cache->io_list)) { | |
5128 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5129 | list_del_init(&cache->io_list); | |
5130 | btrfs_cleanup_bg_io(cache); | |
5131 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5132 | } | |
5133 | ||
5134 | list_del_init(&cache->dirty_list); | |
5135 | spin_lock(&cache->lock); | |
5136 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5137 | spin_unlock(&cache->lock); | |
5138 | ||
5139 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5140 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5141 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5142 | spin_lock(&cur_trans->dirty_bgs_lock); |
5143 | } | |
5144 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5145 | ||
45ae2c18 NB |
5146 | /* |
5147 | * Refer to the definition of io_bgs member for details why it's safe | |
5148 | * to use it without any locking | |
5149 | */ | |
c79a1751 LB |
5150 | while (!list_empty(&cur_trans->io_bgs)) { |
5151 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5152 | struct btrfs_block_group, |
c79a1751 | 5153 | io_list); |
c79a1751 LB |
5154 | |
5155 | list_del_init(&cache->io_list); | |
5156 | spin_lock(&cache->lock); | |
5157 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5158 | spin_unlock(&cache->lock); | |
5159 | btrfs_cleanup_bg_io(cache); | |
5160 | } | |
5161 | } | |
5162 | ||
49b25e05 | 5163 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5164 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5165 | { |
bbbf7243 NB |
5166 | struct btrfs_device *dev, *tmp; |
5167 | ||
2ff7e61e | 5168 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5169 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5170 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5171 | ||
bbbf7243 NB |
5172 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5173 | post_commit_list) { | |
5174 | list_del_init(&dev->post_commit_list); | |
5175 | } | |
5176 | ||
2ff7e61e | 5177 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5178 | |
4a9d8bde | 5179 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5180 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5181 | |
4a9d8bde | 5182 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5183 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5184 | |
ccdf9b30 | 5185 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5186 | |
2ff7e61e | 5187 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5188 | EXTENT_DIRTY); |
fe119a6e | 5189 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5190 | |
d3575156 NA |
5191 | btrfs_free_redirty_list(cur_trans); |
5192 | ||
4a9d8bde MX |
5193 | cur_trans->state =TRANS_STATE_COMPLETED; |
5194 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5195 | } |
5196 | ||
2ff7e61e | 5197 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5198 | { |
5199 | struct btrfs_transaction *t; | |
acce952b | 5200 | |
0b246afa | 5201 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5202 | |
0b246afa JM |
5203 | spin_lock(&fs_info->trans_lock); |
5204 | while (!list_empty(&fs_info->trans_list)) { | |
5205 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5206 | struct btrfs_transaction, list); |
5207 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5208 | refcount_inc(&t->use_count); |
0b246afa | 5209 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5210 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5211 | btrfs_put_transaction(t); |
0b246afa | 5212 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5213 | continue; |
5214 | } | |
0b246afa | 5215 | if (t == fs_info->running_transaction) { |
724e2315 | 5216 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5217 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5218 | /* |
5219 | * We wait for 0 num_writers since we don't hold a trans | |
5220 | * handle open currently for this transaction. | |
5221 | */ | |
5222 | wait_event(t->writer_wait, | |
5223 | atomic_read(&t->num_writers) == 0); | |
5224 | } else { | |
0b246afa | 5225 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5226 | } |
2ff7e61e | 5227 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5228 | |
0b246afa JM |
5229 | spin_lock(&fs_info->trans_lock); |
5230 | if (t == fs_info->running_transaction) | |
5231 | fs_info->running_transaction = NULL; | |
acce952b | 5232 | list_del_init(&t->list); |
0b246afa | 5233 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5234 | |
724e2315 | 5235 | btrfs_put_transaction(t); |
2e4e97ab | 5236 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5237 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5238 | } |
0b246afa JM |
5239 | spin_unlock(&fs_info->trans_lock); |
5240 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5241 | btrfs_destroy_delayed_inodes(fs_info); |
5242 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5243 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5244 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5245 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5246 | |
5247 | return 0; | |
5248 | } | |
ec7d6dfd | 5249 | |
453e4873 | 5250 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5251 | { |
5252 | struct btrfs_path *path; | |
5253 | int ret; | |
5254 | struct extent_buffer *l; | |
5255 | struct btrfs_key search_key; | |
5256 | struct btrfs_key found_key; | |
5257 | int slot; | |
5258 | ||
5259 | path = btrfs_alloc_path(); | |
5260 | if (!path) | |
5261 | return -ENOMEM; | |
5262 | ||
5263 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5264 | search_key.type = -1; | |
5265 | search_key.offset = (u64)-1; | |
5266 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5267 | if (ret < 0) | |
5268 | goto error; | |
5269 | BUG_ON(ret == 0); /* Corruption */ | |
5270 | if (path->slots[0] > 0) { | |
5271 | slot = path->slots[0] - 1; | |
5272 | l = path->nodes[0]; | |
5273 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5274 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5275 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5276 | } else { |
23125104 | 5277 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5278 | } |
5279 | ret = 0; | |
5280 | error: | |
5281 | btrfs_free_path(path); | |
5282 | return ret; | |
5283 | } | |
5284 | ||
543068a2 | 5285 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5286 | { |
5287 | int ret; | |
5288 | mutex_lock(&root->objectid_mutex); | |
5289 | ||
6b8fad57 | 5290 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5291 | btrfs_warn(root->fs_info, |
5292 | "the objectid of root %llu reaches its highest value", | |
5293 | root->root_key.objectid); | |
5294 | ret = -ENOSPC; | |
5295 | goto out; | |
5296 | } | |
5297 | ||
23125104 | 5298 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5299 | ret = 0; |
5300 | out: | |
5301 | mutex_unlock(&root->objectid_mutex); | |
5302 | return ret; | |
5303 | } |