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" |
103c1972 | 26 | #include "bio.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" |
606686ee | 32 | #include "rcu-string.h" |
8dabb742 | 33 | #include "dev-replace.h" |
53b381b3 | 34 | #include "raid56.h" |
5ac1d209 | 35 | #include "sysfs.h" |
fcebe456 | 36 | #include "qgroup.h" |
ebb8765b | 37 | #include "compression.h" |
557ea5dd | 38 | #include "tree-checker.h" |
fd708b81 | 39 | #include "ref-verify.h" |
aac0023c | 40 | #include "block-group.h" |
b0643e59 | 41 | #include "discard.h" |
f603bb94 | 42 | #include "space-info.h" |
b70f5097 | 43 | #include "zoned.h" |
139e8cd3 | 44 | #include "subpage.h" |
c7f13d42 | 45 | #include "fs.h" |
07e81dc9 | 46 | #include "accessors.h" |
a0231804 | 47 | #include "extent-tree.h" |
45c40c8f | 48 | #include "root-tree.h" |
59b818e0 | 49 | #include "defrag.h" |
c7a03b52 | 50 | #include "uuid-tree.h" |
67707479 | 51 | #include "relocation.h" |
2fc6822c | 52 | #include "scrub.h" |
c03b2207 | 53 | #include "super.h" |
eb60ceac | 54 | |
319e4d06 QW |
55 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
56 | BTRFS_HEADER_FLAG_RELOC |\ | |
57 | BTRFS_SUPER_FLAG_ERROR |\ | |
58 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
59 | BTRFS_SUPER_FLAG_METADUMP |\ |
60 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 61 | |
2ff7e61e JM |
62 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
63 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 64 | |
141386e1 JB |
65 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
66 | { | |
67 | if (fs_info->csum_shash) | |
68 | crypto_free_shash(fs_info->csum_shash); | |
69 | } | |
70 | ||
d352ac68 | 71 | /* |
2996e1f8 | 72 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 73 | */ |
c67b3892 | 74 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 75 | { |
d5178578 | 76 | struct btrfs_fs_info *fs_info = buf->fs_info; |
397239ed QW |
77 | int num_pages; |
78 | u32 first_page_part; | |
d5178578 | 79 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 80 | char *kaddr; |
e9be5a30 | 81 | int i; |
d5178578 JT |
82 | |
83 | shash->tfm = fs_info->csum_shash; | |
84 | crypto_shash_init(shash); | |
397239ed QW |
85 | |
86 | if (buf->addr) { | |
87 | /* Pages are contiguous, handle them as a big one. */ | |
88 | kaddr = buf->addr; | |
89 | first_page_part = fs_info->nodesize; | |
90 | num_pages = 1; | |
91 | } else { | |
082d5bb9 | 92 | kaddr = folio_address(buf->folios[0]); |
397239ed QW |
93 | first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
94 | num_pages = num_extent_pages(buf); | |
95 | } | |
96 | ||
e9be5a30 | 97 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 98 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 99 | |
13df3775 QW |
100 | /* |
101 | * Multiple single-page folios case would reach here. | |
102 | * | |
103 | * nodesize <= PAGE_SIZE and large folio all handled by above | |
104 | * crypto_shash_update() already. | |
105 | */ | |
5ad9b471 | 106 | for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) { |
082d5bb9 | 107 | kaddr = folio_address(buf->folios[i]); |
e9be5a30 | 108 | crypto_shash_update(shash, kaddr, PAGE_SIZE); |
19c00ddc | 109 | } |
71a63551 | 110 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 111 | crypto_shash_final(shash, result); |
19c00ddc CM |
112 | } |
113 | ||
d352ac68 CM |
114 | /* |
115 | * we can't consider a given block up to date unless the transid of the | |
116 | * block matches the transid in the parent node's pointer. This is how we | |
117 | * detect blocks that either didn't get written at all or got written | |
118 | * in the wrong place. | |
119 | */ | |
d87e6575 | 120 | int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) |
1259ab75 | 121 | { |
d87e6575 | 122 | if (!extent_buffer_uptodate(eb)) |
1259ab75 CM |
123 | return 0; |
124 | ||
d87e6575 CH |
125 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) |
126 | return 1; | |
127 | ||
b9fab919 CM |
128 | if (atomic) |
129 | return -EAGAIN; | |
130 | ||
d87e6575 CH |
131 | if (!extent_buffer_uptodate(eb) || |
132 | btrfs_header_generation(eb) != parent_transid) { | |
133 | btrfs_err_rl(eb->fs_info, | |
8f0ed7d4 QW |
134 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
135 | eb->start, eb->read_mirror, | |
29549aec | 136 | parent_transid, btrfs_header_generation(eb)); |
d87e6575 | 137 | clear_extent_buffer_uptodate(eb); |
9e2aff90 | 138 | return 0; |
d87e6575 | 139 | } |
9e2aff90 | 140 | return 1; |
1259ab75 CM |
141 | } |
142 | ||
e7e16f48 JT |
143 | static bool btrfs_supported_super_csum(u16 csum_type) |
144 | { | |
145 | switch (csum_type) { | |
146 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 147 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 148 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 149 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
150 | return true; |
151 | default: | |
152 | return false; | |
153 | } | |
154 | } | |
155 | ||
1104a885 DS |
156 | /* |
157 | * Return 0 if the superblock checksum type matches the checksum value of that | |
158 | * algorithm. Pass the raw disk superblock data. | |
159 | */ | |
3d17adea QW |
160 | int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
161 | const struct btrfs_super_block *disk_sb) | |
1104a885 | 162 | { |
51bce6c9 | 163 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
164 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
165 | ||
166 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 167 | |
51bce6c9 JT |
168 | /* |
169 | * The super_block structure does not span the whole | |
170 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
171 | * filled with zeros and is included in the checksum. | |
172 | */ | |
3d17adea | 173 | crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, |
fd08001f | 174 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); |
1104a885 | 175 | |
55fc29be | 176 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 177 | return 1; |
1104a885 | 178 | |
e7e16f48 | 179 | return 0; |
1104a885 DS |
180 | } |
181 | ||
bacf60e5 CH |
182 | static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, |
183 | int mirror_num) | |
184 | { | |
185 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
96c36eaa | 186 | int num_folios = num_extent_folios(eb); |
bacf60e5 CH |
187 | int ret = 0; |
188 | ||
189 | if (sb_rdonly(fs_info->sb)) | |
190 | return -EROFS; | |
191 | ||
96c36eaa QW |
192 | for (int i = 0; i < num_folios; i++) { |
193 | struct folio *folio = eb->folios[i]; | |
194 | u64 start = max_t(u64, eb->start, folio_pos(folio)); | |
082d5bb9 | 195 | u64 end = min_t(u64, eb->start + eb->len, |
84cda1a6 | 196 | folio_pos(folio) + eb->folio_size); |
917ac778 | 197 | u32 len = end - start; |
bacf60e5 | 198 | |
917ac778 | 199 | ret = btrfs_repair_io_failure(fs_info, 0, start, len, |
96c36eaa QW |
200 | start, folio, offset_in_folio(folio, start), |
201 | mirror_num); | |
bacf60e5 CH |
202 | if (ret) |
203 | break; | |
bacf60e5 CH |
204 | } |
205 | ||
206 | return ret; | |
207 | } | |
208 | ||
d352ac68 CM |
209 | /* |
210 | * helper to read a given tree block, doing retries as required when | |
211 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 | 212 | * |
789d6a3a QW |
213 | * @check: expected tree parentness check, see the comments of the |
214 | * structure for details. | |
d352ac68 | 215 | */ |
6a2e9dc4 | 216 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
789d6a3a | 217 | struct btrfs_tree_parent_check *check) |
f188591e | 218 | { |
5ab12d1f | 219 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ea466794 | 220 | int failed = 0; |
f188591e CM |
221 | int ret; |
222 | int num_copies = 0; | |
223 | int mirror_num = 0; | |
ea466794 | 224 | int failed_mirror = 0; |
f188591e | 225 | |
789d6a3a QW |
226 | ASSERT(check); |
227 | ||
f188591e | 228 | while (1) { |
f8397d69 | 229 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
947a6299 QW |
230 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); |
231 | if (!ret) | |
232 | break; | |
d397712b | 233 | |
0b246afa | 234 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 235 | eb->start, eb->len); |
4235298e | 236 | if (num_copies == 1) |
ea466794 | 237 | break; |
4235298e | 238 | |
5cf1ab56 JB |
239 | if (!failed_mirror) { |
240 | failed = 1; | |
241 | failed_mirror = eb->read_mirror; | |
242 | } | |
243 | ||
f188591e | 244 | mirror_num++; |
ea466794 JB |
245 | if (mirror_num == failed_mirror) |
246 | mirror_num++; | |
247 | ||
4235298e | 248 | if (mirror_num > num_copies) |
ea466794 | 249 | break; |
f188591e | 250 | } |
ea466794 | 251 | |
c0901581 | 252 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 253 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
254 | |
255 | return ret; | |
f188591e | 256 | } |
19c00ddc | 257 | |
31d89399 CH |
258 | /* |
259 | * Checksum a dirty tree block before IO. | |
260 | */ | |
261 | blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) | |
eca0f6f6 | 262 | { |
31d89399 | 263 | struct extent_buffer *eb = bbio->private; |
eca0f6f6 | 264 | struct btrfs_fs_info *fs_info = eb->fs_info; |
31d89399 | 265 | u64 found_start = btrfs_header_bytenr(eb); |
0124855f | 266 | u64 last_trans; |
eca0f6f6 QW |
267 | u8 result[BTRFS_CSUM_SIZE]; |
268 | int ret; | |
269 | ||
31d89399 CH |
270 | /* Btree blocks are always contiguous on disk. */ |
271 | if (WARN_ON_ONCE(bbio->file_offset != eb->start)) | |
272 | return BLK_STS_IOERR; | |
273 | if (WARN_ON_ONCE(bbio->bio.bi_iter.bi_size != eb->len)) | |
274 | return BLK_STS_IOERR; | |
275 | ||
aa6313e6 JT |
276 | /* |
277 | * If an extent_buffer is marked as EXTENT_BUFFER_ZONED_ZEROOUT, don't | |
278 | * checksum it but zero-out its content. This is done to preserve | |
279 | * ordering of I/O without unnecessarily writing out data. | |
280 | */ | |
cbf44cd9 | 281 | if (test_bit(EXTENT_BUFFER_ZONED_ZEROOUT, &eb->bflags)) { |
aa6313e6 | 282 | memzero_extent_buffer(eb, 0, eb->len); |
31d89399 CH |
283 | return BLK_STS_OK; |
284 | } | |
285 | ||
286 | if (WARN_ON_ONCE(found_start != eb->start)) | |
287 | return BLK_STS_IOERR; | |
55151ea9 QW |
288 | if (WARN_ON(!btrfs_folio_test_uptodate(fs_info, eb->folios[0], |
289 | eb->start, eb->len))) | |
31d89399 CH |
290 | return BLK_STS_IOERR; |
291 | ||
eca0f6f6 QW |
292 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, |
293 | offsetof(struct btrfs_header, fsid), | |
294 | BTRFS_FSID_SIZE) == 0); | |
295 | csum_tree_block(eb, result); | |
296 | ||
297 | if (btrfs_header_level(eb)) | |
298 | ret = btrfs_check_node(eb); | |
299 | else | |
85d8a826 | 300 | ret = btrfs_check_leaf(eb); |
eca0f6f6 | 301 | |
3777369f QW |
302 | if (ret < 0) |
303 | goto error; | |
304 | ||
305 | /* | |
306 | * Also check the generation, the eb reached here must be newer than | |
307 | * last committed. Or something seriously wrong happened. | |
308 | */ | |
0124855f FM |
309 | last_trans = btrfs_get_last_trans_committed(fs_info); |
310 | if (unlikely(btrfs_header_generation(eb) <= last_trans)) { | |
3777369f | 311 | ret = -EUCLEAN; |
eca0f6f6 | 312 | btrfs_err(fs_info, |
3777369f | 313 | "block=%llu bad generation, have %llu expect > %llu", |
0124855f | 314 | eb->start, btrfs_header_generation(eb), last_trans); |
3777369f | 315 | goto error; |
eca0f6f6 QW |
316 | } |
317 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
31d89399 | 318 | return BLK_STS_OK; |
3777369f QW |
319 | |
320 | error: | |
321 | btrfs_print_tree(eb, 0); | |
322 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
323 | eb->start); | |
16199ad9 FM |
324 | /* |
325 | * Be noisy if this is an extent buffer from a log tree. We don't abort | |
326 | * a transaction in case there's a bad log tree extent buffer, we just | |
327 | * fallback to a transaction commit. Still we want to know when there is | |
328 | * a bad log tree extent buffer, as that may signal a bug somewhere. | |
329 | */ | |
330 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || | |
331 | btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); | |
deb6216f CH |
332 | return errno_to_blk_status(ret); |
333 | } | |
334 | ||
413fb1bc | 335 | static bool check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 336 | { |
b0c9b3b0 | 337 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 338 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 339 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c | 340 | |
9a8658e3 DS |
341 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
342 | BTRFS_FSID_SIZE); | |
67bc5ad0 | 343 | |
944d3f9f | 344 | /* |
f7361d8c AJ |
345 | * alloc_fsid_devices() copies the fsid into fs_devices::metadata_uuid. |
346 | * This is then overwritten by metadata_uuid if it is present in the | |
347 | * device_list_add(). The same true for a seed device as well. So use of | |
348 | * fs_devices::metadata_uuid is appropriate here. | |
944d3f9f | 349 | */ |
67bc5ad0 | 350 | if (memcmp(fsid, fs_info->fs_devices->metadata_uuid, BTRFS_FSID_SIZE) == 0) |
413fb1bc | 351 | return false; |
944d3f9f NB |
352 | |
353 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
354 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
413fb1bc | 355 | return false; |
944d3f9f | 356 | |
413fb1bc | 357 | return true; |
2b82032c YZ |
358 | } |
359 | ||
77bf40a2 | 360 | /* Do basic extent buffer checks at read time */ |
046b562b CH |
361 | int btrfs_validate_extent_buffer(struct extent_buffer *eb, |
362 | struct btrfs_tree_parent_check *check) | |
ce9adaa5 | 363 | { |
77bf40a2 | 364 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 365 | u64 found_start; |
77bf40a2 QW |
366 | const u32 csum_size = fs_info->csum_size; |
367 | u8 found_level; | |
2996e1f8 | 368 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 369 | const u8 *header_csum; |
77bf40a2 | 370 | int ret = 0; |
ea466794 | 371 | |
947a6299 QW |
372 | ASSERT(check); |
373 | ||
ce9adaa5 | 374 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 375 | if (found_start != eb->start) { |
8f0ed7d4 QW |
376 | btrfs_err_rl(fs_info, |
377 | "bad tree block start, mirror %u want %llu have %llu", | |
378 | eb->read_mirror, eb->start, found_start); | |
f188591e | 379 | ret = -EIO; |
77bf40a2 | 380 | goto out; |
ce9adaa5 | 381 | } |
b0c9b3b0 | 382 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
383 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
384 | eb->start, eb->read_mirror); | |
1259ab75 | 385 | ret = -EIO; |
77bf40a2 | 386 | goto out; |
1259ab75 | 387 | } |
ce9adaa5 | 388 | found_level = btrfs_header_level(eb); |
1c24c3ce | 389 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
390 | btrfs_err(fs_info, |
391 | "bad tree block level, mirror %u level %d on logical %llu", | |
392 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 393 | ret = -EIO; |
77bf40a2 | 394 | goto out; |
1c24c3ce | 395 | } |
ce9adaa5 | 396 | |
c67b3892 | 397 | csum_tree_block(eb, result); |
082d5bb9 | 398 | header_csum = folio_address(eb->folios[0]) + |
8d993618 | 399 | get_eb_offset_in_folio(eb, offsetof(struct btrfs_header, csum)); |
a826d6dc | 400 | |
dfd29eed | 401 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 402 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
403 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
404 | eb->start, eb->read_mirror, | |
dfd29eed | 405 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
406 | CSUM_FMT_VALUE(csum_size, result), |
407 | btrfs_header_level(eb)); | |
2996e1f8 | 408 | ret = -EUCLEAN; |
77bf40a2 | 409 | goto out; |
2996e1f8 JT |
410 | } |
411 | ||
947a6299 | 412 | if (found_level != check->level) { |
77177ed1 QW |
413 | btrfs_err(fs_info, |
414 | "level verify failed on logical %llu mirror %u wanted %u found %u", | |
415 | eb->start, eb->read_mirror, check->level, found_level); | |
947a6299 QW |
416 | ret = -EIO; |
417 | goto out; | |
418 | } | |
419 | if (unlikely(check->transid && | |
420 | btrfs_header_generation(eb) != check->transid)) { | |
421 | btrfs_err_rl(eb->fs_info, | |
422 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", | |
423 | eb->start, eb->read_mirror, check->transid, | |
424 | btrfs_header_generation(eb)); | |
425 | ret = -EIO; | |
426 | goto out; | |
427 | } | |
428 | if (check->has_first_key) { | |
429 | struct btrfs_key *expect_key = &check->first_key; | |
430 | struct btrfs_key found_key; | |
431 | ||
432 | if (found_level) | |
433 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
434 | else | |
435 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
436 | if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { | |
437 | btrfs_err(fs_info, | |
438 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", | |
439 | eb->start, check->transid, | |
440 | expect_key->objectid, | |
441 | expect_key->type, expect_key->offset, | |
442 | found_key.objectid, found_key.type, | |
443 | found_key.offset); | |
444 | ret = -EUCLEAN; | |
445 | goto out; | |
446 | } | |
447 | } | |
448 | if (check->owner_root) { | |
449 | ret = btrfs_check_eb_owner(eb, check->owner_root); | |
450 | if (ret < 0) | |
451 | goto out; | |
452 | } | |
453 | ||
a826d6dc JB |
454 | /* |
455 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
456 | * that we don't try and read the other copies of this block, just | |
457 | * return -EIO. | |
458 | */ | |
85d8a826 | 459 | if (found_level == 0 && btrfs_check_leaf(eb)) { |
a826d6dc JB |
460 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
461 | ret = -EIO; | |
462 | } | |
ce9adaa5 | 463 | |
813fd1dc | 464 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
465 | ret = -EIO; |
466 | ||
aebcc159 | 467 | if (ret) |
75391f0d | 468 | btrfs_err(fs_info, |
8f0ed7d4 QW |
469 | "read time tree block corruption detected on logical %llu mirror %u", |
470 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
471 | out: |
472 | return ret; | |
473 | } | |
474 | ||
3dd1462e | 475 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
476 | static int btree_migrate_folio(struct address_space *mapping, |
477 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
478 | { |
479 | /* | |
480 | * we can't safely write a btree page from here, | |
481 | * we haven't done the locking hook | |
482 | */ | |
8958b551 | 483 | if (folio_test_dirty(src)) |
784b4e29 CM |
484 | return -EAGAIN; |
485 | /* | |
486 | * Buffers may be managed in a filesystem specific way. | |
487 | * We must have no buffers or drop them. | |
488 | */ | |
8958b551 MWO |
489 | if (folio_get_private(src) && |
490 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 491 | return -EAGAIN; |
54184650 | 492 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 493 | } |
8958b551 MWO |
494 | #else |
495 | #define btree_migrate_folio NULL | |
3dd1462e | 496 | #endif |
784b4e29 | 497 | |
0da5468f CM |
498 | static int btree_writepages(struct address_space *mapping, |
499 | struct writeback_control *wbc) | |
500 | { | |
e2d84521 MX |
501 | struct btrfs_fs_info *fs_info; |
502 | int ret; | |
503 | ||
d8d5f3e1 | 504 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
505 | |
506 | if (wbc->for_kupdate) | |
507 | return 0; | |
508 | ||
e2d84521 | 509 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 510 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
511 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
512 | BTRFS_DIRTY_METADATA_THRESH, | |
513 | fs_info->dirty_metadata_batch); | |
e2d84521 | 514 | if (ret < 0) |
793955bc | 515 | return 0; |
793955bc | 516 | } |
0b32f4bb | 517 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
518 | } |
519 | ||
f913cff3 | 520 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 521 | { |
f913cff3 MWO |
522 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
523 | return false; | |
0c4e538b | 524 | |
f913cff3 | 525 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
526 | } |
527 | ||
895586eb MWO |
528 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
529 | size_t length) | |
d98237b3 | 530 | { |
d1310b2e | 531 | struct extent_io_tree *tree; |
895586eb MWO |
532 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
533 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 534 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
535 | if (folio_get_private(folio)) { |
536 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
537 | "folio private not zero on folio %llu", | |
538 | (unsigned long long)folio_pos(folio)); | |
539 | folio_detach_private(folio); | |
9ad6b7bc | 540 | } |
d98237b3 CM |
541 | } |
542 | ||
bb146eb2 | 543 | #ifdef DEBUG |
0079c3b1 MWO |
544 | static bool btree_dirty_folio(struct address_space *mapping, |
545 | struct folio *folio) | |
546 | { | |
547 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
5e0e8799 | 548 | struct btrfs_subpage_info *spi = fs_info->subpage_info; |
139e8cd3 | 549 | struct btrfs_subpage *subpage; |
0b32f4bb | 550 | struct extent_buffer *eb; |
139e8cd3 | 551 | int cur_bit = 0; |
0079c3b1 | 552 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
553 | |
554 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 555 | eb = folio_get_private(folio); |
139e8cd3 QW |
556 | BUG_ON(!eb); |
557 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
558 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 559 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 560 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 561 | } |
5e0e8799 QW |
562 | |
563 | ASSERT(spi); | |
0079c3b1 | 564 | subpage = folio_get_private(folio); |
139e8cd3 | 565 | |
5e0e8799 QW |
566 | for (cur_bit = spi->dirty_offset; |
567 | cur_bit < spi->dirty_offset + spi->bitmap_nr_bits; | |
568 | cur_bit++) { | |
139e8cd3 QW |
569 | unsigned long flags; |
570 | u64 cur; | |
139e8cd3 QW |
571 | |
572 | spin_lock_irqsave(&subpage->lock, flags); | |
5e0e8799 | 573 | if (!test_bit(cur_bit, subpage->bitmaps)) { |
139e8cd3 | 574 | spin_unlock_irqrestore(&subpage->lock, flags); |
139e8cd3 QW |
575 | continue; |
576 | } | |
577 | spin_unlock_irqrestore(&subpage->lock, flags); | |
578 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 579 | |
139e8cd3 QW |
580 | eb = find_extent_buffer(fs_info, cur); |
581 | ASSERT(eb); | |
582 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
583 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 584 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
585 | free_extent_buffer(eb); |
586 | ||
5e0e8799 | 587 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits) - 1; |
139e8cd3 | 588 | } |
0079c3b1 | 589 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 590 | } |
0079c3b1 MWO |
591 | #else |
592 | #define btree_dirty_folio filemap_dirty_folio | |
593 | #endif | |
0b32f4bb | 594 | |
7f09410b | 595 | static const struct address_space_operations btree_aops = { |
0da5468f | 596 | .writepages = btree_writepages, |
f913cff3 | 597 | .release_folio = btree_release_folio, |
895586eb | 598 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
599 | .migrate_folio = btree_migrate_folio, |
600 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
601 | }; |
602 | ||
2ff7e61e JM |
603 | struct extent_buffer *btrfs_find_create_tree_block( |
604 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
605 | u64 bytenr, u64 owner_root, |
606 | int level) | |
0999df54 | 607 | { |
0b246afa JM |
608 | if (btrfs_is_testing(fs_info)) |
609 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 610 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
611 | } |
612 | ||
581c1760 QW |
613 | /* |
614 | * Read tree block at logical address @bytenr and do variant basic but critical | |
615 | * verification. | |
616 | * | |
789d6a3a QW |
617 | * @check: expected tree parentness check, see comments of the |
618 | * structure for details. | |
581c1760 | 619 | */ |
2ff7e61e | 620 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
789d6a3a | 621 | struct btrfs_tree_parent_check *check) |
0999df54 CM |
622 | { |
623 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
624 | int ret; |
625 | ||
789d6a3a QW |
626 | ASSERT(check); |
627 | ||
628 | buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, | |
629 | check->level); | |
c871b0f2 LB |
630 | if (IS_ERR(buf)) |
631 | return buf; | |
0999df54 | 632 | |
789d6a3a | 633 | ret = btrfs_read_extent_buffer(buf, check); |
0f0fe8f7 | 634 | if (ret) { |
537f38f0 | 635 | free_extent_buffer_stale(buf); |
64c043de | 636 | return ERR_PTR(ret); |
0f0fe8f7 | 637 | } |
789d6a3a | 638 | if (btrfs_check_eb_owner(buf, check->owner_root)) { |
88c602ab QW |
639 | free_extent_buffer_stale(buf); |
640 | return ERR_PTR(-EUCLEAN); | |
641 | } | |
5f39d397 | 642 | return buf; |
ce9adaa5 | 643 | |
eb60ceac CM |
644 | } |
645 | ||
da17066c | 646 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 647 | u64 objectid) |
d97e63b6 | 648 | { |
7c0260ee | 649 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
650 | |
651 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
652 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
653 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 654 | root->fs_info = fs_info; |
2e608bd1 | 655 | root->root_key.objectid = objectid; |
cfaa7295 | 656 | root->node = NULL; |
a28ec197 | 657 | root->commit_root = NULL; |
27cdeb70 | 658 | root->state = 0; |
abed4aaa | 659 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 660 | |
0f7d52f4 | 661 | root->last_trans = 0; |
6b8fad57 | 662 | root->free_objectid = 0; |
eb73c1b7 | 663 | root->nr_delalloc_inodes = 0; |
199c2a9c | 664 | root->nr_ordered_extents = 0; |
6bef4d31 | 665 | root->inode_tree = RB_ROOT; |
6140ba8a DS |
666 | /* GFP flags are compatible with XA_FLAGS_*. */ |
667 | xa_init_flags(&root->delayed_nodes, GFP_ATOMIC); | |
2e608bd1 JB |
668 | |
669 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
670 | |
671 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 672 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
673 | INIT_LIST_HEAD(&root->delalloc_inodes); |
674 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
675 | INIT_LIST_HEAD(&root->ordered_extents); |
676 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 677 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
5d4f98a2 | 678 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 679 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 680 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 681 | spin_lock_init(&root->accounting_lock); |
8287475a | 682 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 683 | mutex_init(&root->objectid_mutex); |
e02119d5 | 684 | mutex_init(&root->log_mutex); |
31f3d255 | 685 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 686 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 687 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
688 | init_waitqueue_head(&root->log_writer_wait); |
689 | init_waitqueue_head(&root->log_commit_wait[0]); | |
690 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
691 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
692 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
693 | atomic_set(&root->log_commit[0], 0); |
694 | atomic_set(&root->log_commit[1], 0); | |
695 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 696 | atomic_set(&root->log_batch, 0); |
0700cea7 | 697 | refcount_set(&root->refs, 1); |
8ecebf4d | 698 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 699 | atomic_set(&root->nr_swapfiles, 0); |
6008859b | 700 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 701 | root->log_transid_committed = -1; |
f9850787 | 702 | btrfs_set_root_last_log_commit(root, 0); |
2e608bd1 | 703 | root->anon_dev = 0; |
e289f03e | 704 | if (!dummy) { |
43eb5f29 | 705 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
35da5a7e | 706 | IO_TREE_ROOT_DIRTY_LOG_PAGES); |
e289f03e | 707 | extent_io_tree_init(fs_info, &root->log_csum_range, |
35da5a7e | 708 | IO_TREE_LOG_CSUM_RANGE); |
e289f03e | 709 | } |
017e5369 | 710 | |
5f3ab90a | 711 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 712 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
713 | #ifdef CONFIG_BTRFS_DEBUG |
714 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 715 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 716 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 717 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 718 | #endif |
3768f368 CM |
719 | } |
720 | ||
74e4d827 | 721 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 722 | u64 objectid, gfp_t flags) |
6f07e42e | 723 | { |
74e4d827 | 724 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 725 | if (root) |
96dfcb46 | 726 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
727 | return root; |
728 | } | |
729 | ||
06ea65a3 JB |
730 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
731 | /* Should only be used by the testing infrastructure */ | |
da17066c | 732 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
733 | { |
734 | struct btrfs_root *root; | |
735 | ||
7c0260ee JM |
736 | if (!fs_info) |
737 | return ERR_PTR(-EINVAL); | |
738 | ||
96dfcb46 | 739 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
740 | if (!root) |
741 | return ERR_PTR(-ENOMEM); | |
da17066c | 742 | |
b9ef22de | 743 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 744 | root->alloc_bytenr = 0; |
06ea65a3 JB |
745 | |
746 | return root; | |
747 | } | |
748 | #endif | |
749 | ||
abed4aaa JB |
750 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
751 | { | |
752 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
753 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
754 | ||
755 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
756 | } | |
757 | ||
758 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
759 | { | |
760 | const struct btrfs_key *key = k; | |
761 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
762 | ||
763 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
764 | } | |
765 | ||
766 | int btrfs_global_root_insert(struct btrfs_root *root) | |
767 | { | |
768 | struct btrfs_fs_info *fs_info = root->fs_info; | |
769 | struct rb_node *tmp; | |
745806fb | 770 | int ret = 0; |
abed4aaa JB |
771 | |
772 | write_lock(&fs_info->global_root_lock); | |
773 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
774 | write_unlock(&fs_info->global_root_lock); | |
abed4aaa | 775 | |
745806fb QW |
776 | if (tmp) { |
777 | ret = -EEXIST; | |
778 | btrfs_warn(fs_info, "global root %llu %llu already exists", | |
779 | root->root_key.objectid, root->root_key.offset); | |
780 | } | |
781 | return ret; | |
abed4aaa JB |
782 | } |
783 | ||
784 | void btrfs_global_root_delete(struct btrfs_root *root) | |
785 | { | |
786 | struct btrfs_fs_info *fs_info = root->fs_info; | |
787 | ||
788 | write_lock(&fs_info->global_root_lock); | |
789 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
790 | write_unlock(&fs_info->global_root_lock); | |
791 | } | |
792 | ||
793 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
794 | struct btrfs_key *key) | |
795 | { | |
796 | struct rb_node *node; | |
797 | struct btrfs_root *root = NULL; | |
798 | ||
799 | read_lock(&fs_info->global_root_lock); | |
800 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
801 | if (node) | |
802 | root = container_of(node, struct btrfs_root, rb_node); | |
803 | read_unlock(&fs_info->global_root_lock); | |
804 | ||
805 | return root; | |
806 | } | |
807 | ||
f7238e50 JB |
808 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
809 | { | |
810 | struct btrfs_block_group *block_group; | |
811 | u64 ret; | |
812 | ||
813 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
814 | return 0; | |
815 | ||
816 | if (bytenr) | |
817 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
818 | else | |
819 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
820 | ASSERT(block_group); | |
821 | if (!block_group) | |
822 | return 0; | |
823 | ret = block_group->global_root_id; | |
824 | btrfs_put_block_group(block_group); | |
825 | ||
826 | return ret; | |
827 | } | |
828 | ||
abed4aaa JB |
829 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
830 | { | |
831 | struct btrfs_key key = { | |
832 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
833 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 834 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
835 | }; |
836 | ||
837 | return btrfs_global_root(fs_info, &key); | |
838 | } | |
839 | ||
840 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
841 | { | |
842 | struct btrfs_key key = { | |
843 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
844 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 845 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
846 | }; |
847 | ||
848 | return btrfs_global_root(fs_info, &key); | |
849 | } | |
850 | ||
51129b33 JB |
851 | struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) |
852 | { | |
853 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) | |
854 | return fs_info->block_group_root; | |
855 | return btrfs_extent_root(fs_info, 0); | |
856 | } | |
857 | ||
20897f5c | 858 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
859 | u64 objectid) |
860 | { | |
9b7a2440 | 861 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
862 | struct extent_buffer *leaf; |
863 | struct btrfs_root *tree_root = fs_info->tree_root; | |
864 | struct btrfs_root *root; | |
865 | struct btrfs_key key; | |
b89f6d1f | 866 | unsigned int nofs_flag; |
20897f5c | 867 | int ret = 0; |
20897f5c | 868 | |
b89f6d1f FM |
869 | /* |
870 | * We're holding a transaction handle, so use a NOFS memory allocation | |
871 | * context to avoid deadlock if reclaim happens. | |
872 | */ | |
873 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 874 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 875 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
876 | if (!root) |
877 | return ERR_PTR(-ENOMEM); | |
878 | ||
20897f5c AJ |
879 | root->root_key.objectid = objectid; |
880 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
881 | root->root_key.offset = 0; | |
882 | ||
9631e4cc | 883 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
60ea105a | 884 | 0, BTRFS_NESTING_NORMAL); |
20897f5c AJ |
885 | if (IS_ERR(leaf)) { |
886 | ret = PTR_ERR(leaf); | |
1dd05682 | 887 | leaf = NULL; |
c1b07854 | 888 | goto fail; |
20897f5c AJ |
889 | } |
890 | ||
20897f5c | 891 | root->node = leaf; |
50564b65 | 892 | btrfs_mark_buffer_dirty(trans, leaf); |
20897f5c AJ |
893 | |
894 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 895 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 896 | |
f944d2cb DS |
897 | btrfs_set_root_flags(&root->root_item, 0); |
898 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
899 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
900 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
901 | btrfs_set_root_level(&root->root_item, 0); | |
902 | btrfs_set_root_refs(&root->root_item, 1); | |
903 | btrfs_set_root_used(&root->root_item, leaf->len); | |
904 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
905 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 906 | if (is_fstree(objectid)) |
807fc790 AS |
907 | generate_random_guid(root->root_item.uuid); |
908 | else | |
909 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 910 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 911 | |
8a6a87cd BB |
912 | btrfs_tree_unlock(leaf); |
913 | ||
20897f5c AJ |
914 | key.objectid = objectid; |
915 | key.type = BTRFS_ROOT_ITEM_KEY; | |
916 | key.offset = 0; | |
917 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
918 | if (ret) | |
919 | goto fail; | |
920 | ||
1dd05682 TI |
921 | return root; |
922 | ||
8a6a87cd | 923 | fail: |
00246528 | 924 | btrfs_put_root(root); |
20897f5c | 925 | |
1dd05682 | 926 | return ERR_PTR(ret); |
20897f5c AJ |
927 | } |
928 | ||
7237f183 YZ |
929 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
930 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
931 | { |
932 | struct btrfs_root *root; | |
e02119d5 | 933 | |
96dfcb46 | 934 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 935 | if (!root) |
7237f183 | 936 | return ERR_PTR(-ENOMEM); |
e02119d5 | 937 | |
e02119d5 CM |
938 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
939 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
940 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 941 | |
6ab6ebb7 NA |
942 | return root; |
943 | } | |
944 | ||
945 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
946 | struct btrfs_root *root) | |
947 | { | |
948 | struct extent_buffer *leaf; | |
949 | ||
7237f183 | 950 | /* |
92a7cc42 | 951 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 952 | * |
92a7cc42 QW |
953 | * Log trees are not exposed to user space thus can't be snapshotted, |
954 | * and they go away before a real commit is actually done. | |
955 | * | |
956 | * They do store pointers to file data extents, and those reference | |
957 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 958 | */ |
e02119d5 | 959 | |
4d75f8a9 | 960 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
60ea105a | 961 | NULL, 0, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
962 | if (IS_ERR(leaf)) |
963 | return PTR_ERR(leaf); | |
e02119d5 | 964 | |
7237f183 | 965 | root->node = leaf; |
e02119d5 | 966 | |
50564b65 | 967 | btrfs_mark_buffer_dirty(trans, root->node); |
e02119d5 | 968 | btrfs_tree_unlock(root->node); |
6ab6ebb7 NA |
969 | |
970 | return 0; | |
7237f183 YZ |
971 | } |
972 | ||
973 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
974 | struct btrfs_fs_info *fs_info) | |
975 | { | |
976 | struct btrfs_root *log_root; | |
977 | ||
978 | log_root = alloc_log_tree(trans, fs_info); | |
979 | if (IS_ERR(log_root)) | |
980 | return PTR_ERR(log_root); | |
6ab6ebb7 | 981 | |
3ddebf27 NA |
982 | if (!btrfs_is_zoned(fs_info)) { |
983 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
984 | ||
985 | if (ret) { | |
986 | btrfs_put_root(log_root); | |
987 | return ret; | |
988 | } | |
6ab6ebb7 NA |
989 | } |
990 | ||
7237f183 YZ |
991 | WARN_ON(fs_info->log_root_tree); |
992 | fs_info->log_root_tree = log_root; | |
993 | return 0; | |
994 | } | |
995 | ||
996 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
997 | struct btrfs_root *root) | |
998 | { | |
0b246afa | 999 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1000 | struct btrfs_root *log_root; |
1001 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1002 | int ret; |
7237f183 | 1003 | |
0b246afa | 1004 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1005 | if (IS_ERR(log_root)) |
1006 | return PTR_ERR(log_root); | |
1007 | ||
6ab6ebb7 NA |
1008 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1009 | if (ret) { | |
1010 | btrfs_put_root(log_root); | |
1011 | return ret; | |
1012 | } | |
1013 | ||
7237f183 YZ |
1014 | log_root->last_trans = trans->transid; |
1015 | log_root->root_key.offset = root->root_key.objectid; | |
1016 | ||
1017 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1018 | btrfs_set_stack_inode_generation(inode_item, 1); |
1019 | btrfs_set_stack_inode_size(inode_item, 3); | |
1020 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1021 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1022 | fs_info->nodesize); |
3cae210f | 1023 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1024 | |
5d4f98a2 | 1025 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1026 | |
1027 | WARN_ON(root->log_root); | |
1028 | root->log_root = log_root; | |
6008859b | 1029 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 1030 | root->log_transid_committed = -1; |
f9850787 | 1031 | btrfs_set_root_last_log_commit(root, 0); |
e02119d5 CM |
1032 | return 0; |
1033 | } | |
1034 | ||
49d11bea JB |
1035 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1036 | struct btrfs_path *path, | |
1037 | struct btrfs_key *key) | |
e02119d5 CM |
1038 | { |
1039 | struct btrfs_root *root; | |
789d6a3a | 1040 | struct btrfs_tree_parent_check check = { 0 }; |
e02119d5 | 1041 | struct btrfs_fs_info *fs_info = tree_root->fs_info; |
84234f3a | 1042 | u64 generation; |
cb517eab | 1043 | int ret; |
581c1760 | 1044 | int level; |
0f7d52f4 | 1045 | |
96dfcb46 | 1046 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1047 | if (!root) |
1048 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1049 | |
cb517eab MX |
1050 | ret = btrfs_find_root(tree_root, key, path, |
1051 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1052 | if (ret) { |
13a8a7c8 YZ |
1053 | if (ret > 0) |
1054 | ret = -ENOENT; | |
49d11bea | 1055 | goto fail; |
0f7d52f4 | 1056 | } |
13a8a7c8 | 1057 | |
84234f3a | 1058 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1059 | level = btrfs_root_level(&root->root_item); |
789d6a3a QW |
1060 | check.level = level; |
1061 | check.transid = generation; | |
1062 | check.owner_root = key->objectid; | |
1063 | root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), | |
1064 | &check); | |
64c043de LB |
1065 | if (IS_ERR(root->node)) { |
1066 | ret = PTR_ERR(root->node); | |
8c38938c | 1067 | root->node = NULL; |
49d11bea | 1068 | goto fail; |
4eb150d6 QW |
1069 | } |
1070 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1071 | ret = -EIO; |
49d11bea | 1072 | goto fail; |
416bc658 | 1073 | } |
88c602ab QW |
1074 | |
1075 | /* | |
1076 | * For real fs, and not log/reloc trees, root owner must | |
1077 | * match its root node owner | |
1078 | */ | |
1079 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1080 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1081 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1082 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1083 | btrfs_crit(fs_info, | |
1084 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1085 | root->root_key.objectid, root->node->start, | |
1086 | btrfs_header_owner(root->node), | |
1087 | root->root_key.objectid); | |
1088 | ret = -EUCLEAN; | |
1089 | goto fail; | |
1090 | } | |
5d4f98a2 | 1091 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1092 | return root; |
49d11bea | 1093 | fail: |
00246528 | 1094 | btrfs_put_root(root); |
49d11bea JB |
1095 | return ERR_PTR(ret); |
1096 | } | |
1097 | ||
1098 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1099 | struct btrfs_key *key) | |
1100 | { | |
1101 | struct btrfs_root *root; | |
1102 | struct btrfs_path *path; | |
1103 | ||
1104 | path = btrfs_alloc_path(); | |
1105 | if (!path) | |
1106 | return ERR_PTR(-ENOMEM); | |
1107 | root = read_tree_root_path(tree_root, path, key); | |
1108 | btrfs_free_path(path); | |
1109 | ||
1110 | return root; | |
cb517eab MX |
1111 | } |
1112 | ||
2dfb1e43 QW |
1113 | /* |
1114 | * Initialize subvolume root in-memory structure | |
1115 | * | |
1116 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1117 | */ | |
1118 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1119 | { |
1120 | int ret; | |
1121 | ||
0b548539 | 1122 | btrfs_drew_lock_init(&root->snapshot_lock); |
8257b2dc | 1123 | |
aeb935a4 | 1124 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
6ebcd021 QW |
1125 | !btrfs_is_data_reloc_root(root) && |
1126 | is_fstree(root->root_key.objectid)) { | |
92a7cc42 | 1127 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1128 | btrfs_check_and_init_root_item(&root->root_item); |
1129 | } | |
1130 | ||
851fd730 QW |
1131 | /* |
1132 | * Don't assign anonymous block device to roots that are not exposed to | |
1133 | * userspace, the id pool is limited to 1M | |
1134 | */ | |
1135 | if (is_fstree(root->root_key.objectid) && | |
1136 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1137 | if (!anon_dev) { |
1138 | ret = get_anon_bdev(&root->anon_dev); | |
1139 | if (ret) | |
1140 | goto fail; | |
1141 | } else { | |
1142 | root->anon_dev = anon_dev; | |
1143 | } | |
851fd730 | 1144 | } |
f32e48e9 CR |
1145 | |
1146 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1147 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1148 | if (ret) { |
1149 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1150 | goto fail; |
f32e48e9 CR |
1151 | } |
1152 | ||
6b8fad57 | 1153 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1154 | |
1155 | mutex_unlock(&root->objectid_mutex); | |
1156 | ||
cb517eab MX |
1157 | return 0; |
1158 | fail: | |
84db5ccf | 1159 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1160 | return ret; |
1161 | } | |
1162 | ||
a98db0f3 JB |
1163 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1164 | u64 root_id) | |
cb517eab MX |
1165 | { |
1166 | struct btrfs_root *root; | |
1167 | ||
fc7cbcd4 DS |
1168 | spin_lock(&fs_info->fs_roots_radix_lock); |
1169 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1170 | (unsigned long)root_id); | |
25ac047c | 1171 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1172 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1173 | return root; |
1174 | } | |
1175 | ||
49d11bea JB |
1176 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1177 | u64 objectid) | |
1178 | { | |
abed4aaa JB |
1179 | struct btrfs_key key = { |
1180 | .objectid = objectid, | |
1181 | .type = BTRFS_ROOT_ITEM_KEY, | |
1182 | .offset = 0, | |
1183 | }; | |
1184 | ||
e91909aa CH |
1185 | switch (objectid) { |
1186 | case BTRFS_ROOT_TREE_OBJECTID: | |
49d11bea | 1187 | return btrfs_grab_root(fs_info->tree_root); |
e91909aa | 1188 | case BTRFS_EXTENT_TREE_OBJECTID: |
abed4aaa | 1189 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1190 | case BTRFS_CHUNK_TREE_OBJECTID: |
49d11bea | 1191 | return btrfs_grab_root(fs_info->chunk_root); |
e91909aa | 1192 | case BTRFS_DEV_TREE_OBJECTID: |
49d11bea | 1193 | return btrfs_grab_root(fs_info->dev_root); |
e91909aa | 1194 | case BTRFS_CSUM_TREE_OBJECTID: |
abed4aaa | 1195 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1196 | case BTRFS_QUOTA_TREE_OBJECTID: |
85724171 | 1197 | return btrfs_grab_root(fs_info->quota_root); |
e91909aa | 1198 | case BTRFS_UUID_TREE_OBJECTID: |
85724171 | 1199 | return btrfs_grab_root(fs_info->uuid_root); |
e91909aa | 1200 | case BTRFS_BLOCK_GROUP_TREE_OBJECTID: |
85724171 | 1201 | return btrfs_grab_root(fs_info->block_group_root); |
e91909aa | 1202 | case BTRFS_FREE_SPACE_TREE_OBJECTID: |
85724171 | 1203 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
51502090 JT |
1204 | case BTRFS_RAID_STRIPE_TREE_OBJECTID: |
1205 | return btrfs_grab_root(fs_info->stripe_root); | |
e91909aa CH |
1206 | default: |
1207 | return NULL; | |
1208 | } | |
49d11bea JB |
1209 | } |
1210 | ||
cb517eab MX |
1211 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1212 | struct btrfs_root *root) | |
1213 | { | |
1214 | int ret; | |
1215 | ||
fc7cbcd4 DS |
1216 | ret = radix_tree_preload(GFP_NOFS); |
1217 | if (ret) | |
1218 | return ret; | |
1219 | ||
1220 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1221 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1222 | (unsigned long)root->root_key.objectid, | |
1223 | root); | |
af01d2e5 | 1224 | if (ret == 0) { |
00246528 | 1225 | btrfs_grab_root(root); |
fc7cbcd4 | 1226 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1227 | } |
fc7cbcd4 DS |
1228 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1229 | radix_tree_preload_end(); | |
cb517eab MX |
1230 | |
1231 | return ret; | |
1232 | } | |
1233 | ||
bd647ce3 JB |
1234 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1235 | { | |
1236 | #ifdef CONFIG_BTRFS_DEBUG | |
1237 | struct btrfs_root *root; | |
1238 | ||
1239 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1240 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1241 | ||
bd647ce3 JB |
1242 | root = list_first_entry(&fs_info->allocated_roots, |
1243 | struct btrfs_root, leak_list); | |
457f1864 | 1244 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1245 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 | 1246 | refcount_read(&root->refs)); |
8fd2b12e | 1247 | WARN_ON_ONCE(1); |
bd647ce3 | 1248 | while (refcount_read(&root->refs) > 1) |
00246528 JB |
1249 | btrfs_put_root(root); |
1250 | btrfs_put_root(root); | |
bd647ce3 JB |
1251 | } |
1252 | #endif | |
1253 | } | |
1254 | ||
abed4aaa JB |
1255 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1256 | { | |
1257 | struct btrfs_root *root; | |
1258 | struct rb_node *node; | |
1259 | ||
1260 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1261 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1262 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1263 | btrfs_put_root(root); | |
1264 | } | |
1265 | } | |
1266 | ||
0d4b0463 JB |
1267 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1268 | { | |
141386e1 JB |
1269 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1270 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1271 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1272 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1273 | btrfs_free_csum_hash(fs_info); | |
1274 | btrfs_free_stripe_hash_table(fs_info); | |
1275 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1276 | kfree(fs_info->balance_ctl); |
1277 | kfree(fs_info->delayed_root); | |
abed4aaa | 1278 | free_global_roots(fs_info); |
00246528 JB |
1279 | btrfs_put_root(fs_info->tree_root); |
1280 | btrfs_put_root(fs_info->chunk_root); | |
1281 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1282 | btrfs_put_root(fs_info->quota_root); |
1283 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1284 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1285 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1286 | btrfs_put_root(fs_info->block_group_root); |
51502090 | 1287 | btrfs_put_root(fs_info->stripe_root); |
bd647ce3 | 1288 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1289 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1290 | kfree(fs_info->super_copy); |
1291 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1292 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1293 | kvfree(fs_info); |
1294 | } | |
1295 | ||
1296 | ||
2dfb1e43 QW |
1297 | /* |
1298 | * Get an in-memory reference of a root structure. | |
1299 | * | |
1300 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1301 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1302 | * found, then read it from disk and add it to cached fs roots. | |
1303 | * | |
1304 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1305 | * | |
1306 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1307 | * same root objectid. | |
1308 | * | |
1309 | * @objectid: root id | |
1310 | * @anon_dev: preallocated anonymous block device number for new roots, | |
e2b54eaf | 1311 | * pass NULL for a new allocation. |
2dfb1e43 QW |
1312 | * @check_ref: whether to check root item references, If true, return -ENOENT |
1313 | * for orphan roots | |
1314 | */ | |
1315 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1316 | u64 objectid, dev_t *anon_dev, |
2dfb1e43 | 1317 | bool check_ref) |
5eda7b5e CM |
1318 | { |
1319 | struct btrfs_root *root; | |
381cf658 | 1320 | struct btrfs_path *path; |
1d4c08e0 | 1321 | struct btrfs_key key; |
5eda7b5e CM |
1322 | int ret; |
1323 | ||
49d11bea JB |
1324 | root = btrfs_get_global_root(fs_info, objectid); |
1325 | if (root) | |
1326 | return root; | |
773e722a QW |
1327 | |
1328 | /* | |
1329 | * If we're called for non-subvolume trees, and above function didn't | |
1330 | * find one, do not try to read it from disk. | |
1331 | * | |
1332 | * This is namely for free-space-tree and quota tree, which can change | |
1333 | * at runtime and should only be grabbed from fs_info. | |
1334 | */ | |
1335 | if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) | |
1336 | return ERR_PTR(-ENOENT); | |
4df27c4d | 1337 | again: |
56e9357a | 1338 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1339 | if (root) { |
e03ee2fe QW |
1340 | /* |
1341 | * Some other caller may have read out the newly inserted | |
1342 | * subvolume already (for things like backref walk etc). Not | |
1343 | * that common but still possible. In that case, we just need | |
1344 | * to free the anon_dev. | |
1345 | */ | |
e2b54eaf FM |
1346 | if (unlikely(anon_dev && *anon_dev)) { |
1347 | free_anon_bdev(*anon_dev); | |
1348 | *anon_dev = 0; | |
e03ee2fe QW |
1349 | } |
1350 | ||
bc44d7c4 | 1351 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1352 | btrfs_put_root(root); |
48475471 | 1353 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1354 | } |
5eda7b5e | 1355 | return root; |
48475471 | 1356 | } |
5eda7b5e | 1357 | |
56e9357a DS |
1358 | key.objectid = objectid; |
1359 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1360 | key.offset = (u64)-1; | |
1361 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1362 | if (IS_ERR(root)) |
1363 | return root; | |
3394e160 | 1364 | |
c00869f1 | 1365 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1366 | ret = -ENOENT; |
581bb050 | 1367 | goto fail; |
35a30d7c | 1368 | } |
581bb050 | 1369 | |
e2b54eaf | 1370 | ret = btrfs_init_fs_root(root, anon_dev ? *anon_dev : 0); |
ac08aedf CM |
1371 | if (ret) |
1372 | goto fail; | |
3394e160 | 1373 | |
381cf658 DS |
1374 | path = btrfs_alloc_path(); |
1375 | if (!path) { | |
1376 | ret = -ENOMEM; | |
1377 | goto fail; | |
1378 | } | |
1d4c08e0 DS |
1379 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1380 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1381 | key.offset = objectid; |
1d4c08e0 DS |
1382 | |
1383 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1384 | btrfs_free_path(path); |
d68fc57b YZ |
1385 | if (ret < 0) |
1386 | goto fail; | |
1387 | if (ret == 0) | |
27cdeb70 | 1388 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1389 | |
cb517eab | 1390 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1391 | if (ret) { |
168a2f77 JJB |
1392 | if (ret == -EEXIST) { |
1393 | btrfs_put_root(root); | |
4df27c4d | 1394 | goto again; |
168a2f77 | 1395 | } |
4df27c4d | 1396 | goto fail; |
0f7d52f4 | 1397 | } |
edbd8d4e | 1398 | return root; |
4df27c4d | 1399 | fail: |
33fab972 FM |
1400 | /* |
1401 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1402 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1403 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1404 | * and once again by our caller. | |
1405 | */ | |
e2b54eaf | 1406 | if (anon_dev && *anon_dev) |
33fab972 | 1407 | root->anon_dev = 0; |
8c38938c | 1408 | btrfs_put_root(root); |
4df27c4d | 1409 | return ERR_PTR(ret); |
edbd8d4e CM |
1410 | } |
1411 | ||
2dfb1e43 QW |
1412 | /* |
1413 | * Get in-memory reference of a root structure | |
1414 | * | |
1415 | * @objectid: tree objectid | |
1416 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1417 | * one reference | |
1418 | */ | |
1419 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1420 | u64 objectid, bool check_ref) | |
1421 | { | |
e2b54eaf | 1422 | return btrfs_get_root_ref(fs_info, objectid, NULL, check_ref); |
2dfb1e43 QW |
1423 | } |
1424 | ||
1425 | /* | |
1426 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1427 | * the anonymous block device id | |
1428 | * | |
1429 | * @objectid: tree objectid | |
e2b54eaf FM |
1430 | * @anon_dev: if NULL, allocate a new anonymous block device or use the |
1431 | * parameter value if not NULL | |
2dfb1e43 QW |
1432 | */ |
1433 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1434 | u64 objectid, dev_t *anon_dev) |
2dfb1e43 QW |
1435 | { |
1436 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1437 | } | |
1438 | ||
49d11bea | 1439 | /* |
9580503b DS |
1440 | * Return a root for the given objectid. |
1441 | * | |
49d11bea JB |
1442 | * @fs_info: the fs_info |
1443 | * @objectid: the objectid we need to lookup | |
1444 | * | |
1445 | * This is exclusively used for backref walking, and exists specifically because | |
1446 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1447 | * creation time, which means we may have to read the tree_root in order to look | |
1448 | * up a fs root that is not in memory. If the root is not in memory we will | |
1449 | * read the tree root commit root and look up the fs root from there. This is a | |
1450 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1451 | * have the most uptodate information, it'll simply be discarded once the | |
1452 | * backref code is finished using the root. | |
1453 | */ | |
1454 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1455 | struct btrfs_path *path, | |
1456 | u64 objectid) | |
1457 | { | |
1458 | struct btrfs_root *root; | |
1459 | struct btrfs_key key; | |
1460 | ||
1461 | ASSERT(path->search_commit_root && path->skip_locking); | |
1462 | ||
1463 | /* | |
1464 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1465 | * since this is called via the backref walking code we won't be looking | |
1466 | * up a root that doesn't exist, unless there's corruption. So if root | |
1467 | * != NULL just return it. | |
1468 | */ | |
1469 | root = btrfs_get_global_root(fs_info, objectid); | |
1470 | if (root) | |
1471 | return root; | |
1472 | ||
1473 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1474 | if (root) | |
1475 | return root; | |
1476 | ||
1477 | key.objectid = objectid; | |
1478 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1479 | key.offset = (u64)-1; | |
1480 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1481 | btrfs_release_path(path); | |
1482 | ||
1483 | return root; | |
1484 | } | |
1485 | ||
a74a4b97 CM |
1486 | static int cleaner_kthread(void *arg) |
1487 | { | |
0d031dc4 | 1488 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1489 | int again; |
a74a4b97 | 1490 | |
d6fd0ae2 | 1491 | while (1) { |
d0278245 | 1492 | again = 0; |
a74a4b97 | 1493 | |
fd340d0f JB |
1494 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1495 | ||
d0278245 | 1496 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1497 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1498 | goto sleep; |
1499 | ||
90c711ab ZB |
1500 | /* |
1501 | * Do not do anything if we might cause open_ctree() to block | |
1502 | * before we have finished mounting the filesystem. | |
1503 | */ | |
0b246afa | 1504 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1505 | goto sleep; |
1506 | ||
0b246afa | 1507 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1508 | goto sleep; |
1509 | ||
dc7f370c MX |
1510 | /* |
1511 | * Avoid the problem that we change the status of the fs | |
1512 | * during the above check and trylock. | |
1513 | */ | |
2ff7e61e | 1514 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1515 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1516 | goto sleep; |
76dda93c | 1517 | } |
a74a4b97 | 1518 | |
b7625f46 QW |
1519 | if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) |
1520 | btrfs_sysfs_feature_update(fs_info); | |
1521 | ||
2ff7e61e | 1522 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1523 | |
33c44184 | 1524 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1525 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1526 | |
1527 | /* | |
05323cd1 MX |
1528 | * The defragger has dealt with the R/O remount and umount, |
1529 | * needn't do anything special here. | |
d0278245 | 1530 | */ |
0b246afa | 1531 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1532 | |
1533 | /* | |
f3372065 | 1534 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1535 | * with relocation (btrfs_relocate_chunk) and relocation |
1536 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1537 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1538 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1539 | * unused block groups. | |
1540 | */ | |
0b246afa | 1541 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1542 | |
1543 | /* | |
1544 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1545 | * all unused block_groups. This possibly gives us some more free | |
1546 | * space. | |
1547 | */ | |
1548 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1549 | sleep: |
a0a1db70 | 1550 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1551 | if (kthread_should_park()) |
1552 | kthread_parkme(); | |
1553 | if (kthread_should_stop()) | |
1554 | return 0; | |
838fe188 | 1555 | if (!again) { |
a74a4b97 | 1556 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1557 | schedule(); |
a74a4b97 CM |
1558 | __set_current_state(TASK_RUNNING); |
1559 | } | |
da288d28 | 1560 | } |
a74a4b97 CM |
1561 | } |
1562 | ||
1563 | static int transaction_kthread(void *arg) | |
1564 | { | |
1565 | struct btrfs_root *root = arg; | |
0b246afa | 1566 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1567 | struct btrfs_trans_handle *trans; |
1568 | struct btrfs_transaction *cur; | |
8929ecfa | 1569 | u64 transid; |
643900be | 1570 | time64_t delta; |
a74a4b97 | 1571 | unsigned long delay; |
914b2007 | 1572 | bool cannot_commit; |
a74a4b97 CM |
1573 | |
1574 | do { | |
914b2007 | 1575 | cannot_commit = false; |
ba1bc00f | 1576 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1577 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1578 | |
0b246afa JM |
1579 | spin_lock(&fs_info->trans_lock); |
1580 | cur = fs_info->running_transaction; | |
a74a4b97 | 1581 | if (!cur) { |
0b246afa | 1582 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1583 | goto sleep; |
1584 | } | |
31153d81 | 1585 | |
643900be | 1586 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 | 1587 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
77d20c68 | 1588 | cur->state < TRANS_STATE_COMMIT_PREP && |
643900be | 1589 | delta < fs_info->commit_interval) { |
0b246afa | 1590 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1591 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1592 | delay = min(delay, | |
1593 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1594 | goto sleep; |
1595 | } | |
8929ecfa | 1596 | transid = cur->transid; |
0b246afa | 1597 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1598 | |
79787eaa | 1599 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1600 | trans = btrfs_attach_transaction(root); |
914b2007 | 1601 | if (IS_ERR(trans)) { |
354aa0fb MX |
1602 | if (PTR_ERR(trans) != -ENOENT) |
1603 | cannot_commit = true; | |
79787eaa | 1604 | goto sleep; |
914b2007 | 1605 | } |
8929ecfa | 1606 | if (transid == trans->transid) { |
3a45bb20 | 1607 | btrfs_commit_transaction(trans); |
8929ecfa | 1608 | } else { |
3a45bb20 | 1609 | btrfs_end_transaction(trans); |
8929ecfa | 1610 | } |
a74a4b97 | 1611 | sleep: |
0b246afa JM |
1612 | wake_up_process(fs_info->cleaner_kthread); |
1613 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1614 | |
84961539 | 1615 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1616 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1617 | if (!kthread_should_stop() && |
0b246afa | 1618 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1619 | cannot_commit)) |
bc5511d0 | 1620 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1621 | } while (!kthread_should_stop()); |
1622 | return 0; | |
1623 | } | |
1624 | ||
af31f5e5 | 1625 | /* |
01f0f9da NB |
1626 | * This will find the highest generation in the array of root backups. The |
1627 | * index of the highest array is returned, or -EINVAL if we can't find | |
1628 | * anything. | |
af31f5e5 CM |
1629 | * |
1630 | * We check to make sure the array is valid by comparing the | |
1631 | * generation of the latest root in the array with the generation | |
1632 | * in the super block. If they don't match we pitch it. | |
1633 | */ | |
01f0f9da | 1634 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1635 | { |
01f0f9da | 1636 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1637 | u64 cur; |
af31f5e5 CM |
1638 | struct btrfs_root_backup *root_backup; |
1639 | int i; | |
1640 | ||
1641 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1642 | root_backup = info->super_copy->super_roots + i; | |
1643 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1644 | if (cur == newest_gen) | |
01f0f9da | 1645 | return i; |
af31f5e5 CM |
1646 | } |
1647 | ||
01f0f9da | 1648 | return -EINVAL; |
af31f5e5 CM |
1649 | } |
1650 | ||
af31f5e5 CM |
1651 | /* |
1652 | * copy all the root pointers into the super backup array. | |
1653 | * this will bump the backup pointer by one when it is | |
1654 | * done | |
1655 | */ | |
1656 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1657 | { | |
6ef108dd | 1658 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1659 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1660 | |
1661 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1662 | ||
1663 | /* | |
1664 | * make sure all of our padding and empty slots get zero filled | |
1665 | * regardless of which ones we use today | |
1666 | */ | |
1667 | memset(root_backup, 0, sizeof(*root_backup)); | |
1668 | ||
1669 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1670 | ||
1671 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1672 | btrfs_set_backup_tree_root_gen(root_backup, | |
1673 | btrfs_header_generation(info->tree_root->node)); | |
1674 | ||
1675 | btrfs_set_backup_tree_root_level(root_backup, | |
1676 | btrfs_header_level(info->tree_root->node)); | |
1677 | ||
1678 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1679 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1680 | btrfs_header_generation(info->chunk_root->node)); | |
1681 | btrfs_set_backup_chunk_root_level(root_backup, | |
1682 | btrfs_header_level(info->chunk_root->node)); | |
1683 | ||
1c56ab99 | 1684 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1685 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1686 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1687 | |
1688 | btrfs_set_backup_extent_root(root_backup, | |
1689 | extent_root->node->start); | |
1690 | btrfs_set_backup_extent_root_gen(root_backup, | |
1691 | btrfs_header_generation(extent_root->node)); | |
1692 | btrfs_set_backup_extent_root_level(root_backup, | |
1693 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1694 | |
1695 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1696 | btrfs_set_backup_csum_root_gen(root_backup, | |
1697 | btrfs_header_generation(csum_root->node)); | |
1698 | btrfs_set_backup_csum_root_level(root_backup, | |
1699 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1700 | } |
af31f5e5 | 1701 | |
7c7e82a7 CM |
1702 | /* |
1703 | * we might commit during log recovery, which happens before we set | |
1704 | * the fs_root. Make sure it is valid before we fill it in. | |
1705 | */ | |
1706 | if (info->fs_root && info->fs_root->node) { | |
1707 | btrfs_set_backup_fs_root(root_backup, | |
1708 | info->fs_root->node->start); | |
1709 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1710 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1711 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1712 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1713 | } |
af31f5e5 CM |
1714 | |
1715 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1716 | btrfs_set_backup_dev_root_gen(root_backup, | |
1717 | btrfs_header_generation(info->dev_root->node)); | |
1718 | btrfs_set_backup_dev_root_level(root_backup, | |
1719 | btrfs_header_level(info->dev_root->node)); | |
1720 | ||
af31f5e5 CM |
1721 | btrfs_set_backup_total_bytes(root_backup, |
1722 | btrfs_super_total_bytes(info->super_copy)); | |
1723 | btrfs_set_backup_bytes_used(root_backup, | |
1724 | btrfs_super_bytes_used(info->super_copy)); | |
1725 | btrfs_set_backup_num_devices(root_backup, | |
1726 | btrfs_super_num_devices(info->super_copy)); | |
1727 | ||
1728 | /* | |
1729 | * if we don't copy this out to the super_copy, it won't get remembered | |
1730 | * for the next commit | |
1731 | */ | |
1732 | memcpy(&info->super_copy->super_roots, | |
1733 | &info->super_for_commit->super_roots, | |
1734 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1735 | } | |
1736 | ||
bd2336b2 | 1737 | /* |
9580503b DS |
1738 | * Reads a backup root based on the passed priority. Prio 0 is the newest, prio |
1739 | * 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
bd2336b2 | 1740 | * |
9580503b DS |
1741 | * @fs_info: filesystem whose backup roots need to be read |
1742 | * @priority: priority of backup root required | |
bd2336b2 NB |
1743 | * |
1744 | * Returns backup root index on success and -EINVAL otherwise. | |
1745 | */ | |
1746 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1747 | { | |
1748 | int backup_index = find_newest_super_backup(fs_info); | |
1749 | struct btrfs_super_block *super = fs_info->super_copy; | |
1750 | struct btrfs_root_backup *root_backup; | |
1751 | ||
1752 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1753 | if (priority == 0) | |
1754 | return backup_index; | |
1755 | ||
1756 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1757 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1758 | } else { | |
1759 | return -EINVAL; | |
1760 | } | |
1761 | ||
1762 | root_backup = super->super_roots + backup_index; | |
1763 | ||
1764 | btrfs_set_super_generation(super, | |
1765 | btrfs_backup_tree_root_gen(root_backup)); | |
1766 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1767 | btrfs_set_super_root_level(super, | |
1768 | btrfs_backup_tree_root_level(root_backup)); | |
1769 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1770 | ||
1771 | /* | |
1772 | * Fixme: the total bytes and num_devices need to match or we should | |
1773 | * need a fsck | |
1774 | */ | |
1775 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1776 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1777 | ||
1778 | return backup_index; | |
1779 | } | |
1780 | ||
7abadb64 LB |
1781 | /* helper to cleanup workers */ |
1782 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1783 | { | |
dc6e3209 | 1784 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 1785 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 1786 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
1787 | if (fs_info->endio_workers) |
1788 | destroy_workqueue(fs_info->endio_workers); | |
385de0ef CH |
1789 | if (fs_info->rmw_workers) |
1790 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
1791 | if (fs_info->compressed_write_workers) |
1792 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
1793 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
1794 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 1795 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 1796 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 1797 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 1798 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
1799 | if (fs_info->discard_ctl.discard_workers) |
1800 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
1801 | /* |
1802 | * Now that all other work queues are destroyed, we can safely destroy | |
1803 | * the queues used for metadata I/O, since tasks from those other work | |
1804 | * queues can do metadata I/O operations. | |
1805 | */ | |
d7b9416f CH |
1806 | if (fs_info->endio_meta_workers) |
1807 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
1808 | } |
1809 | ||
2e9f5954 R |
1810 | static void free_root_extent_buffers(struct btrfs_root *root) |
1811 | { | |
1812 | if (root) { | |
1813 | free_extent_buffer(root->node); | |
1814 | free_extent_buffer(root->commit_root); | |
1815 | root->node = NULL; | |
1816 | root->commit_root = NULL; | |
1817 | } | |
1818 | } | |
1819 | ||
abed4aaa JB |
1820 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
1821 | { | |
1822 | struct btrfs_root *root, *tmp; | |
1823 | ||
1824 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
1825 | &fs_info->global_root_tree, | |
1826 | rb_node) | |
1827 | free_root_extent_buffers(root); | |
1828 | } | |
1829 | ||
af31f5e5 | 1830 | /* helper to cleanup tree roots */ |
4273eaff | 1831 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 1832 | { |
2e9f5954 | 1833 | free_root_extent_buffers(info->tree_root); |
655b09fe | 1834 | |
abed4aaa | 1835 | free_global_root_pointers(info); |
2e9f5954 | 1836 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
1837 | free_root_extent_buffers(info->quota_root); |
1838 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 1839 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 1840 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 1841 | free_root_extent_buffers(info->block_group_root); |
51502090 | 1842 | free_root_extent_buffers(info->stripe_root); |
4273eaff | 1843 | if (free_chunk_root) |
2e9f5954 | 1844 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
1845 | } |
1846 | ||
8c38938c JB |
1847 | void btrfs_put_root(struct btrfs_root *root) |
1848 | { | |
1849 | if (!root) | |
1850 | return; | |
1851 | ||
1852 | if (refcount_dec_and_test(&root->refs)) { | |
1853 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 1854 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
1855 | if (root->anon_dev) |
1856 | free_anon_bdev(root->anon_dev); | |
923eb523 | 1857 | free_root_extent_buffers(root); |
8c38938c | 1858 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 1859 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 1860 | list_del_init(&root->leak_list); |
fc7cbcd4 | 1861 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
1862 | #endif |
1863 | kfree(root); | |
1864 | } | |
1865 | } | |
1866 | ||
faa2dbf0 | 1867 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 1868 | { |
fc7cbcd4 DS |
1869 | int ret; |
1870 | struct btrfs_root *gang[8]; | |
1871 | int i; | |
171f6537 JB |
1872 | |
1873 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
1874 | gang[0] = list_entry(fs_info->dead_roots.next, |
1875 | struct btrfs_root, root_list); | |
1876 | list_del(&gang[0]->root_list); | |
171f6537 | 1877 | |
fc7cbcd4 DS |
1878 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
1879 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
1880 | btrfs_put_root(gang[0]); | |
171f6537 JB |
1881 | } |
1882 | ||
fc7cbcd4 DS |
1883 | while (1) { |
1884 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1885 | (void **)gang, 0, | |
1886 | ARRAY_SIZE(gang)); | |
1887 | if (!ret) | |
1888 | break; | |
1889 | for (i = 0; i < ret; i++) | |
1890 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
1891 | } |
1892 | } | |
af31f5e5 | 1893 | |
638aa7ed ES |
1894 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
1895 | { | |
1896 | mutex_init(&fs_info->scrub_lock); | |
1897 | atomic_set(&fs_info->scrubs_running, 0); | |
1898 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1899 | atomic_set(&fs_info->scrubs_paused, 0); | |
1900 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1901 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 1902 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
1903 | } |
1904 | ||
779a65a4 ES |
1905 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
1906 | { | |
1907 | spin_lock_init(&fs_info->balance_lock); | |
1908 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
1909 | atomic_set(&fs_info->balance_pause_req, 0); |
1910 | atomic_set(&fs_info->balance_cancel_req, 0); | |
1911 | fs_info->balance_ctl = NULL; | |
1912 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 1913 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
1914 | } |
1915 | ||
dcb2137c | 1916 | static int btrfs_init_btree_inode(struct super_block *sb) |
f37938e0 | 1917 | { |
dcb2137c | 1918 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
e256927b JB |
1919 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
1920 | fs_info->tree_root); | |
dcb2137c CH |
1921 | struct inode *inode; |
1922 | ||
1923 | inode = new_inode(sb); | |
1924 | if (!inode) | |
1925 | return -ENOMEM; | |
2ff7e61e JM |
1926 | |
1927 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
1928 | set_nlink(inode, 1); | |
f37938e0 ES |
1929 | /* |
1930 | * we set the i_size on the btree inode to the max possible int. | |
1931 | * the real end of the address space is determined by all of | |
1932 | * the devices in the system | |
1933 | */ | |
2ff7e61e JM |
1934 | inode->i_size = OFFSET_MAX; |
1935 | inode->i_mapping->a_ops = &btree_aops; | |
dcb2137c | 1936 | mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); |
f37938e0 | 1937 | |
2ff7e61e | 1938 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 1939 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
35da5a7e | 1940 | IO_TREE_BTREE_INODE_IO); |
2ff7e61e | 1941 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 1942 | |
5c8fd99f | 1943 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
1944 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
1945 | BTRFS_I(inode)->location.type = 0; | |
1946 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 1947 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 1948 | __insert_inode_hash(inode, hash); |
dcb2137c CH |
1949 | fs_info->btree_inode = inode; |
1950 | ||
1951 | return 0; | |
f37938e0 ES |
1952 | } |
1953 | ||
ad618368 ES |
1954 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
1955 | { | |
ad618368 | 1956 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 1957 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 1958 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
1959 | } |
1960 | ||
f9e92e40 ES |
1961 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
1962 | { | |
1963 | spin_lock_init(&fs_info->qgroup_lock); | |
1964 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
1965 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
1966 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
1967 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 1968 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 1969 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 1970 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
1971 | mutex_init(&fs_info->qgroup_rescan_lock); |
1972 | } | |
1973 | ||
d21deec5 | 1974 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 1975 | { |
f7b885be | 1976 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 1977 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
58e814fc | 1978 | unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; |
2a458198 ES |
1979 | |
1980 | fs_info->workers = | |
a31b4a43 | 1981 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2a458198 ES |
1982 | |
1983 | fs_info->delalloc_workers = | |
cb001095 JM |
1984 | btrfs_alloc_workqueue(fs_info, "delalloc", |
1985 | flags, max_active, 2); | |
2a458198 ES |
1986 | |
1987 | fs_info->flush_workers = | |
cb001095 JM |
1988 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
1989 | flags, max_active, 0); | |
2a458198 ES |
1990 | |
1991 | fs_info->caching_workers = | |
cb001095 | 1992 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 1993 | |
2a458198 | 1994 | fs_info->fixup_workers = |
58e814fc | 1995 | btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); |
2a458198 | 1996 | |
2a458198 | 1997 | fs_info->endio_workers = |
d7b9416f | 1998 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 1999 | fs_info->endio_meta_workers = |
d7b9416f | 2000 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
385de0ef | 2001 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2002 | fs_info->endio_write_workers = |
cb001095 JM |
2003 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2004 | max_active, 2); | |
fed8a72d CH |
2005 | fs_info->compressed_write_workers = |
2006 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2007 | fs_info->endio_freespace_worker = |
cb001095 JM |
2008 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2009 | max_active, 0); | |
2a458198 | 2010 | fs_info->delayed_workers = |
cb001095 JM |
2011 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2012 | max_active, 0); | |
2a458198 | 2013 | fs_info->qgroup_rescan_workers = |
58e814fc TH |
2014 | btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", |
2015 | ordered_flags); | |
b0643e59 | 2016 | fs_info->discard_ctl.discard_workers = |
58e814fc | 2017 | alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); |
2a458198 | 2018 | |
8bfec2e4 | 2019 | if (!(fs_info->workers && |
a31b4a43 | 2020 | fs_info->delalloc_workers && fs_info->flush_workers && |
2a458198 | 2021 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2022 | fs_info->compressed_write_workers && |
1a1a2851 | 2023 | fs_info->endio_write_workers && |
2a458198 | 2024 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
f26c9238 QW |
2025 | fs_info->caching_workers && fs_info->fixup_workers && |
2026 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2027 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2028 | return -ENOMEM; |
2029 | } | |
2030 | ||
2031 | return 0; | |
2032 | } | |
2033 | ||
6d97c6e3 JT |
2034 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2035 | { | |
2036 | struct crypto_shash *csum_shash; | |
b4e967be | 2037 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2038 | |
b4e967be | 2039 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2040 | |
2041 | if (IS_ERR(csum_shash)) { | |
2042 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2043 | csum_driver); |
6d97c6e3 JT |
2044 | return PTR_ERR(csum_shash); |
2045 | } | |
2046 | ||
2047 | fs_info->csum_shash = csum_shash; | |
2048 | ||
68d99ab0 CH |
2049 | /* |
2050 | * Check if the checksum implementation is a fast accelerated one. | |
2051 | * As-is this is a bit of a hack and should be replaced once the csum | |
2052 | * implementations provide that information themselves. | |
2053 | */ | |
2054 | switch (csum_type) { | |
2055 | case BTRFS_CSUM_TYPE_CRC32: | |
2056 | if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) | |
2057 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2058 | break; | |
efcfcbc6 DS |
2059 | case BTRFS_CSUM_TYPE_XXHASH: |
2060 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2061 | break; | |
68d99ab0 CH |
2062 | default: |
2063 | break; | |
2064 | } | |
2065 | ||
c8a5f8ca DS |
2066 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2067 | btrfs_super_csum_name(csum_type), | |
2068 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2069 | return 0; |
2070 | } | |
2071 | ||
63443bf5 ES |
2072 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2073 | struct btrfs_fs_devices *fs_devices) | |
2074 | { | |
2075 | int ret; | |
789d6a3a | 2076 | struct btrfs_tree_parent_check check = { 0 }; |
63443bf5 ES |
2077 | struct btrfs_root *log_tree_root; |
2078 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2079 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2080 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2081 | |
2082 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2083 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2084 | return -EIO; |
2085 | } | |
2086 | ||
96dfcb46 JB |
2087 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2088 | GFP_KERNEL); | |
63443bf5 ES |
2089 | if (!log_tree_root) |
2090 | return -ENOMEM; | |
2091 | ||
789d6a3a QW |
2092 | check.level = level; |
2093 | check.transid = fs_info->generation + 1; | |
2094 | check.owner_root = BTRFS_TREE_LOG_OBJECTID; | |
2095 | log_tree_root->node = read_tree_block(fs_info, bytenr, &check); | |
64c043de | 2096 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2097 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2098 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2099 | log_tree_root->node = NULL; |
00246528 | 2100 | btrfs_put_root(log_tree_root); |
0eeff236 | 2101 | return ret; |
4eb150d6 QW |
2102 | } |
2103 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2104 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2105 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2106 | return -EIO; |
2107 | } | |
4eb150d6 | 2108 | |
63443bf5 ES |
2109 | /* returns with log_tree_root freed on success */ |
2110 | ret = btrfs_recover_log_trees(log_tree_root); | |
2111 | if (ret) { | |
0b246afa JM |
2112 | btrfs_handle_fs_error(fs_info, ret, |
2113 | "Failed to recover log tree"); | |
00246528 | 2114 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2115 | return ret; |
2116 | } | |
2117 | ||
bc98a42c | 2118 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2119 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2120 | if (ret) |
2121 | return ret; | |
2122 | } | |
2123 | ||
2124 | return 0; | |
2125 | } | |
2126 | ||
abed4aaa JB |
2127 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2128 | struct btrfs_path *path, u64 objectid, | |
2129 | const char *name) | |
2130 | { | |
2131 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2132 | struct btrfs_root *root; | |
f7238e50 | 2133 | u64 max_global_id = 0; |
abed4aaa JB |
2134 | int ret; |
2135 | struct btrfs_key key = { | |
2136 | .objectid = objectid, | |
2137 | .type = BTRFS_ROOT_ITEM_KEY, | |
2138 | .offset = 0, | |
2139 | }; | |
2140 | bool found = false; | |
2141 | ||
2142 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2143 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2144 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2145 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2146 | return 0; | |
2147 | } | |
2148 | ||
2149 | while (1) { | |
2150 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2151 | if (ret < 0) | |
2152 | break; | |
2153 | ||
2154 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2155 | ret = btrfs_next_leaf(tree_root, path); | |
2156 | if (ret) { | |
2157 | if (ret > 0) | |
2158 | ret = 0; | |
2159 | break; | |
2160 | } | |
2161 | } | |
2162 | ret = 0; | |
2163 | ||
2164 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2165 | if (key.objectid != objectid) | |
2166 | break; | |
2167 | btrfs_release_path(path); | |
2168 | ||
f7238e50 JB |
2169 | /* |
2170 | * Just worry about this for extent tree, it'll be the same for | |
2171 | * everybody. | |
2172 | */ | |
2173 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2174 | max_global_id = max(max_global_id, key.offset); | |
2175 | ||
abed4aaa JB |
2176 | found = true; |
2177 | root = read_tree_root_path(tree_root, path, &key); | |
2178 | if (IS_ERR(root)) { | |
2179 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2180 | ret = PTR_ERR(root); | |
2181 | break; | |
2182 | } | |
2183 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2184 | ret = btrfs_global_root_insert(root); | |
2185 | if (ret) { | |
2186 | btrfs_put_root(root); | |
2187 | break; | |
2188 | } | |
2189 | key.offset++; | |
2190 | } | |
2191 | btrfs_release_path(path); | |
2192 | ||
f7238e50 JB |
2193 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2194 | fs_info->nr_global_roots = max_global_id + 1; | |
2195 | ||
abed4aaa JB |
2196 | if (!found || ret) { |
2197 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2198 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2199 | ||
2200 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2201 | ret = ret ? ret : -ENOENT; | |
2202 | else | |
2203 | ret = 0; | |
2204 | btrfs_err(fs_info, "failed to load root %s", name); | |
2205 | } | |
2206 | return ret; | |
2207 | } | |
2208 | ||
2209 | static int load_global_roots(struct btrfs_root *tree_root) | |
2210 | { | |
2211 | struct btrfs_path *path; | |
2212 | int ret = 0; | |
2213 | ||
2214 | path = btrfs_alloc_path(); | |
2215 | if (!path) | |
2216 | return -ENOMEM; | |
2217 | ||
2218 | ret = load_global_roots_objectid(tree_root, path, | |
2219 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2220 | if (ret) | |
2221 | goto out; | |
2222 | ret = load_global_roots_objectid(tree_root, path, | |
2223 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2224 | if (ret) | |
2225 | goto out; | |
2226 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2227 | goto out; | |
2228 | ret = load_global_roots_objectid(tree_root, path, | |
2229 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2230 | "free space"); | |
2231 | out: | |
2232 | btrfs_free_path(path); | |
2233 | return ret; | |
2234 | } | |
2235 | ||
6bccf3ab | 2236 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2237 | { |
6bccf3ab | 2238 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2239 | struct btrfs_root *root; |
4bbcaa64 ES |
2240 | struct btrfs_key location; |
2241 | int ret; | |
2242 | ||
6bccf3ab JM |
2243 | BUG_ON(!fs_info->tree_root); |
2244 | ||
abed4aaa JB |
2245 | ret = load_global_roots(tree_root); |
2246 | if (ret) | |
2247 | return ret; | |
2248 | ||
4bbcaa64 ES |
2249 | location.type = BTRFS_ROOT_ITEM_KEY; |
2250 | location.offset = 0; | |
2251 | ||
1c56ab99 | 2252 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2253 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2254 | root = btrfs_read_tree_root(tree_root, &location); | |
2255 | if (IS_ERR(root)) { | |
2256 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2257 | ret = PTR_ERR(root); | |
2258 | goto out; | |
2259 | } | |
2260 | } else { | |
2261 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2262 | fs_info->block_group_root = root; | |
2263 | } | |
2264 | } | |
2265 | ||
2266 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2267 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2268 | if (IS_ERR(root)) { |
42437a63 JB |
2269 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2270 | ret = PTR_ERR(root); | |
2271 | goto out; | |
2272 | } | |
2273 | } else { | |
2274 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2275 | fs_info->dev_root = root; | |
f50f4353 | 2276 | } |
820a49da | 2277 | /* Initialize fs_info for all devices in any case */ |
a8d1b164 JT |
2278 | ret = btrfs_init_devices_late(fs_info); |
2279 | if (ret) | |
2280 | goto out; | |
4bbcaa64 | 2281 | |
aeb935a4 QW |
2282 | /* |
2283 | * This tree can share blocks with some other fs tree during relocation | |
2284 | * and we need a proper setup by btrfs_get_fs_root | |
2285 | */ | |
56e9357a DS |
2286 | root = btrfs_get_fs_root(tree_root->fs_info, |
2287 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2288 | if (IS_ERR(root)) { |
42437a63 JB |
2289 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2290 | ret = PTR_ERR(root); | |
2291 | goto out; | |
2292 | } | |
2293 | } else { | |
2294 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2295 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2296 | } |
aeb935a4 | 2297 | |
4bbcaa64 | 2298 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2299 | root = btrfs_read_tree_root(tree_root, &location); |
2300 | if (!IS_ERR(root)) { | |
2301 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
a4f3d2c4 | 2302 | fs_info->quota_root = root; |
4bbcaa64 ES |
2303 | } |
2304 | ||
2305 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2306 | root = btrfs_read_tree_root(tree_root, &location); |
2307 | if (IS_ERR(root)) { | |
42437a63 JB |
2308 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2309 | ret = PTR_ERR(root); | |
2310 | if (ret != -ENOENT) | |
2311 | goto out; | |
2312 | } | |
4bbcaa64 | 2313 | } else { |
a4f3d2c4 DS |
2314 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2315 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2316 | } |
2317 | ||
51502090 JT |
2318 | if (btrfs_fs_incompat(fs_info, RAID_STRIPE_TREE)) { |
2319 | location.objectid = BTRFS_RAID_STRIPE_TREE_OBJECTID; | |
2320 | root = btrfs_read_tree_root(tree_root, &location); | |
2321 | if (IS_ERR(root)) { | |
2322 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2323 | ret = PTR_ERR(root); | |
2324 | goto out; | |
2325 | } | |
2326 | } else { | |
2327 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2328 | fs_info->stripe_root = root; | |
2329 | } | |
2330 | } | |
2331 | ||
4bbcaa64 | 2332 | return 0; |
f50f4353 LB |
2333 | out: |
2334 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2335 | location.objectid, ret); | |
2336 | return ret; | |
4bbcaa64 ES |
2337 | } |
2338 | ||
069ec957 QW |
2339 | /* |
2340 | * Real super block validation | |
2341 | * NOTE: super csum type and incompat features will not be checked here. | |
2342 | * | |
2343 | * @sb: super block to check | |
2344 | * @mirror_num: the super block number to check its bytenr: | |
2345 | * 0 the primary (1st) sb | |
2346 | * 1, 2 2nd and 3rd backup copy | |
2347 | * -1 skip bytenr check | |
2348 | */ | |
a05d3c91 QW |
2349 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2350 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2351 | { |
21a852b0 QW |
2352 | u64 nodesize = btrfs_super_nodesize(sb); |
2353 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2354 | int ret = 0; | |
2355 | ||
2356 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2357 | btrfs_err(fs_info, "no valid FS found"); | |
2358 | ret = -EINVAL; | |
2359 | } | |
2360 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2361 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2362 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2363 | ret = -EINVAL; | |
2364 | } | |
2365 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2366 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2367 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2368 | ret = -EINVAL; | |
2369 | } | |
2370 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2371 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2372 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2373 | ret = -EINVAL; | |
2374 | } | |
2375 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2376 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2377 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2378 | ret = -EINVAL; | |
2379 | } | |
2380 | ||
2381 | /* | |
2382 | * Check sectorsize and nodesize first, other check will need it. | |
2383 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2384 | */ | |
2385 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2386 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2387 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2388 | ret = -EINVAL; | |
2389 | } | |
0bb3eb3e QW |
2390 | |
2391 | /* | |
1a42daab QW |
2392 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2393 | * | |
2394 | * We can support 16K sectorsize with 64K page size without problem, | |
2395 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2396 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2397 | * beginning. | |
0bb3eb3e | 2398 | */ |
1a42daab | 2399 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2400 | btrfs_err(fs_info, |
0bb3eb3e | 2401 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2402 | sectorsize, PAGE_SIZE); |
2403 | ret = -EINVAL; | |
2404 | } | |
0bb3eb3e | 2405 | |
21a852b0 QW |
2406 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2407 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2408 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2409 | ret = -EINVAL; | |
2410 | } | |
2411 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2412 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2413 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2414 | ret = -EINVAL; | |
2415 | } | |
2416 | ||
2417 | /* Root alignment check */ | |
2418 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2419 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2420 | btrfs_super_root(sb)); | |
2421 | ret = -EINVAL; | |
2422 | } | |
2423 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2424 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2425 | btrfs_super_chunk_root(sb)); | |
2426 | ret = -EINVAL; | |
2427 | } | |
2428 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2429 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2430 | btrfs_super_log_root(sb)); | |
2431 | ret = -EINVAL; | |
2432 | } | |
2433 | ||
a5b8a5f9 AJ |
2434 | if (!fs_info->fs_devices->temp_fsid && |
2435 | memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2436 | btrfs_err(fs_info, |
2437 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
d167aa76 | 2438 | sb->fsid, fs_info->fs_devices->fsid); |
aefd7f70 NB |
2439 | ret = -EINVAL; |
2440 | } | |
2441 | ||
6bfe3959 AJ |
2442 | if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), |
2443 | BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2444 | btrfs_err(fs_info, |
2445 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
6bfe3959 | 2446 | btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); |
aefd7f70 NB |
2447 | ret = -EINVAL; |
2448 | } | |
2449 | ||
25984a5a AJ |
2450 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
2451 | BTRFS_FSID_SIZE) != 0) { | |
2452 | btrfs_err(fs_info, | |
2453 | "dev_item UUID does not match metadata fsid: %pU != %pU", | |
2454 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); | |
2455 | ret = -EINVAL; | |
2456 | } | |
2457 | ||
1c56ab99 QW |
2458 | /* |
2459 | * Artificial requirement for block-group-tree to force newer features | |
2460 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2461 | */ | |
2462 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2463 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2464 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2465 | btrfs_err(fs_info, | |
2466 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2467 | ret = -EINVAL; | |
2468 | } | |
2469 | ||
21a852b0 QW |
2470 | /* |
2471 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2472 | * done later | |
2473 | */ | |
2474 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2475 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2476 | btrfs_super_bytes_used(sb)); | |
2477 | ret = -EINVAL; | |
2478 | } | |
2479 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2480 | btrfs_err(fs_info, "invalid stripesize %u", | |
2481 | btrfs_super_stripesize(sb)); | |
2482 | ret = -EINVAL; | |
2483 | } | |
2484 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2485 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2486 | btrfs_super_num_devices(sb)); | |
2487 | if (btrfs_super_num_devices(sb) == 0) { | |
2488 | btrfs_err(fs_info, "number of devices is 0"); | |
2489 | ret = -EINVAL; | |
2490 | } | |
2491 | ||
069ec957 QW |
2492 | if (mirror_num >= 0 && |
2493 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2494 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2495 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2496 | ret = -EINVAL; | |
2497 | } | |
2498 | ||
2499 | /* | |
2500 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2501 | * and one chunk | |
2502 | */ | |
2503 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2504 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2505 | btrfs_super_sys_array_size(sb), | |
2506 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2507 | ret = -EINVAL; | |
2508 | } | |
2509 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2510 | + sizeof(struct btrfs_chunk)) { | |
2511 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2512 | btrfs_super_sys_array_size(sb), | |
2513 | sizeof(struct btrfs_disk_key) | |
2514 | + sizeof(struct btrfs_chunk)); | |
2515 | ret = -EINVAL; | |
2516 | } | |
2517 | ||
2518 | /* | |
2519 | * The generation is a global counter, we'll trust it more than the others | |
2520 | * but it's still possible that it's the one that's wrong. | |
2521 | */ | |
2522 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2523 | btrfs_warn(fs_info, | |
2524 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2525 | btrfs_super_generation(sb), | |
2526 | btrfs_super_chunk_root_generation(sb)); | |
2527 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2528 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2529 | btrfs_warn(fs_info, | |
2530 | "suspicious: generation < cache_generation: %llu < %llu", | |
2531 | btrfs_super_generation(sb), | |
2532 | btrfs_super_cache_generation(sb)); | |
2533 | ||
2534 | return ret; | |
2535 | } | |
2536 | ||
069ec957 QW |
2537 | /* |
2538 | * Validation of super block at mount time. | |
2539 | * Some checks already done early at mount time, like csum type and incompat | |
2540 | * flags will be skipped. | |
2541 | */ | |
2542 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2543 | { | |
a05d3c91 | 2544 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2545 | } |
2546 | ||
75cb857d QW |
2547 | /* |
2548 | * Validation of super block at write time. | |
2549 | * Some checks like bytenr check will be skipped as their values will be | |
2550 | * overwritten soon. | |
2551 | * Extra checks like csum type and incompat flags will be done here. | |
2552 | */ | |
2553 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2554 | struct btrfs_super_block *sb) | |
2555 | { | |
2556 | int ret; | |
2557 | ||
a05d3c91 | 2558 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2559 | if (ret < 0) |
2560 | goto out; | |
e7e16f48 | 2561 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2562 | ret = -EUCLEAN; |
2563 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2564 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2565 | goto out; | |
2566 | } | |
2567 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2568 | ret = -EUCLEAN; | |
2569 | btrfs_err(fs_info, | |
2570 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2571 | btrfs_super_incompat_flags(sb), | |
2572 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2573 | goto out; | |
2574 | } | |
2575 | out: | |
2576 | if (ret < 0) | |
2577 | btrfs_err(fs_info, | |
2578 | "super block corruption detected before writing it to disk"); | |
2579 | return ret; | |
2580 | } | |
2581 | ||
bd676446 JB |
2582 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2583 | { | |
789d6a3a QW |
2584 | struct btrfs_tree_parent_check check = { |
2585 | .level = level, | |
2586 | .transid = gen, | |
2587 | .owner_root = root->root_key.objectid | |
2588 | }; | |
bd676446 JB |
2589 | int ret = 0; |
2590 | ||
789d6a3a | 2591 | root->node = read_tree_block(root->fs_info, bytenr, &check); |
bd676446 JB |
2592 | if (IS_ERR(root->node)) { |
2593 | ret = PTR_ERR(root->node); | |
2594 | root->node = NULL; | |
4eb150d6 QW |
2595 | return ret; |
2596 | } | |
2597 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2598 | free_extent_buffer(root->node); |
2599 | root->node = NULL; | |
4eb150d6 | 2600 | return -EIO; |
bd676446 JB |
2601 | } |
2602 | ||
bd676446 JB |
2603 | btrfs_set_root_node(&root->root_item, root->node); |
2604 | root->commit_root = btrfs_root_node(root); | |
2605 | btrfs_set_root_refs(&root->root_item, 1); | |
2606 | return ret; | |
2607 | } | |
2608 | ||
2609 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2610 | { | |
2611 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2612 | u64 gen, bytenr; | |
2613 | int level, ret; | |
2614 | ||
2615 | bytenr = btrfs_super_root(sb); | |
2616 | gen = btrfs_super_generation(sb); | |
2617 | level = btrfs_super_root_level(sb); | |
2618 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2619 | if (ret) { |
bd676446 | 2620 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2621 | return ret; |
2622 | } | |
14033b08 | 2623 | return 0; |
bd676446 JB |
2624 | } |
2625 | ||
6ef108dd | 2626 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2627 | { |
6ef108dd | 2628 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2629 | struct btrfs_super_block *sb = fs_info->super_copy; |
2630 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2631 | bool handle_error = false; | |
2632 | int ret = 0; | |
2633 | int i; | |
2634 | ||
2635 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2636 | if (handle_error) { |
2637 | if (!IS_ERR(tree_root->node)) | |
2638 | free_extent_buffer(tree_root->node); | |
2639 | tree_root->node = NULL; | |
2640 | ||
2641 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2642 | break; | |
2643 | ||
2644 | free_root_pointers(fs_info, 0); | |
2645 | ||
2646 | /* | |
2647 | * Don't use the log in recovery mode, it won't be | |
2648 | * valid | |
2649 | */ | |
2650 | btrfs_set_super_log_root(sb, 0); | |
2651 | ||
745806fb | 2652 | btrfs_warn(fs_info, "try to load backup roots slot %d", i); |
b8522a1e | 2653 | ret = read_backup_root(fs_info, i); |
6ef108dd | 2654 | backup_index = ret; |
b8522a1e NB |
2655 | if (ret < 0) |
2656 | return ret; | |
2657 | } | |
b8522a1e | 2658 | |
bd676446 JB |
2659 | ret = load_important_roots(fs_info); |
2660 | if (ret) { | |
217f5004 | 2661 | handle_error = true; |
b8522a1e NB |
2662 | continue; |
2663 | } | |
2664 | ||
336a0d8d NB |
2665 | /* |
2666 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2667 | * hasn't been fully initialised and we are the only user | |
2668 | */ | |
453e4873 | 2669 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2670 | if (ret < 0) { |
b8522a1e NB |
2671 | handle_error = true; |
2672 | continue; | |
2673 | } | |
2674 | ||
6b8fad57 | 2675 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2676 | |
2677 | ret = btrfs_read_roots(fs_info); | |
2678 | if (ret < 0) { | |
2679 | handle_error = true; | |
2680 | continue; | |
2681 | } | |
2682 | ||
2683 | /* All successful */ | |
bd676446 | 2684 | fs_info->generation = btrfs_header_generation(tree_root->node); |
0124855f | 2685 | btrfs_set_last_trans_committed(fs_info, fs_info->generation); |
d96b3424 | 2686 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2687 | |
2688 | /* Always begin writing backup roots after the one being used */ | |
2689 | if (backup_index < 0) { | |
2690 | fs_info->backup_root_index = 0; | |
2691 | } else { | |
2692 | fs_info->backup_root_index = backup_index + 1; | |
2693 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2694 | } | |
b8522a1e NB |
2695 | break; |
2696 | } | |
2697 | ||
2698 | return ret; | |
2699 | } | |
2700 | ||
8260edba | 2701 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2702 | { |
fc7cbcd4 | 2703 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2704 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2705 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2706 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2707 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2708 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2709 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2710 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2711 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2712 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2713 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2714 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2715 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2716 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2717 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2718 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2719 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2720 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2721 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2722 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2723 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2724 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2725 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2726 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2727 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2728 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2729 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2730 | |
e1489b4f | 2731 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2732 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2733 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2734 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
77d20c68 JB |
2735 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_prep, |
2736 | BTRFS_LOCKDEP_TRANS_COMMIT_PREP); | |
3e738c53 IA |
2737 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, |
2738 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
2739 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
2740 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
2741 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
2742 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 2743 | |
0b86a832 | 2744 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2745 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2746 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2747 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2748 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 2749 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
2750 | #ifdef CONFIG_BTRFS_DEBUG |
2751 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2752 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2753 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2754 | #endif |
7dc66abb FM |
2755 | fs_info->mapping_tree = RB_ROOT_CACHED; |
2756 | rwlock_init(&fs_info->mapping_tree_lock); | |
66d8f3dd MX |
2757 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2758 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2759 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2760 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2761 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2762 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2763 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2764 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2765 | BTRFS_BLOCK_RSV_DELREFS); | |
2766 | ||
771ed689 | 2767 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2768 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 2769 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2770 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 2771 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 2772 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2773 | fs_info->metadata_ratio = 0; |
4cb5300b | 2774 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2775 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2776 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2777 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
fd708b81 | 2778 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2779 | |
b34b086c CM |
2780 | fs_info->thread_pool_size = min_t(unsigned long, |
2781 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2782 | |
199c2a9c MX |
2783 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2784 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2785 | |
638aa7ed | 2786 | btrfs_init_scrub(fs_info); |
779a65a4 | 2787 | btrfs_init_balance(fs_info); |
57056740 | 2788 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2789 | |
16b0c258 | 2790 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 2791 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 2792 | |
fe119a6e | 2793 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
35da5a7e | 2794 | IO_TREE_FS_EXCLUDED_EXTENTS); |
39279cc3 | 2795 | |
5a3f23d5 | 2796 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2797 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2798 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2799 | mutex_init(&fs_info->transaction_kthread_mutex); |
2800 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2801 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2802 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2803 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2804 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2805 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2806 | |
ad618368 | 2807 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2808 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2809 | btrfs_discard_init(fs_info); |
416ac51d | 2810 | |
fa9c0d79 CM |
2811 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2812 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2813 | ||
e6dcd2dc | 2814 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2815 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2816 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2817 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2818 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2819 | |
da17066c JM |
2820 | /* Usable values until the real ones are cached from the superblock */ |
2821 | fs_info->nodesize = 4096; | |
2822 | fs_info->sectorsize = 4096; | |
ab108d99 | 2823 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
2824 | fs_info->stripesize = 4096; |
2825 | ||
6207c9e3 JB |
2826 | /* Default compress algorithm when user does -o compress */ |
2827 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2828 | ||
f7b12a62 NA |
2829 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
2830 | ||
eede2bf3 OS |
2831 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2832 | fs_info->swapfile_pins = RB_ROOT; | |
2833 | ||
18bb8bbf JT |
2834 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
2835 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
2836 | } |
2837 | ||
2838 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2839 | { | |
2840 | int ret; | |
2841 | ||
2842 | fs_info->sb = sb; | |
4e00422e | 2843 | /* Temporary fixed values for block size until we read the superblock. */ |
8260edba JB |
2844 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2845 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 2846 | |
5deb17e1 | 2847 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 2848 | if (ret) |
c75e8394 | 2849 | return ret; |
ae18c37a JB |
2850 | |
2851 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
2852 | if (ret) | |
c75e8394 | 2853 | return ret; |
ae18c37a JB |
2854 | |
2855 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
2856 | (1 + ilog2(nr_cpu_ids)); | |
2857 | ||
2858 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
2859 | if (ret) | |
c75e8394 | 2860 | return ret; |
ae18c37a JB |
2861 | |
2862 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
2863 | GFP_KERNEL); | |
2864 | if (ret) | |
c75e8394 | 2865 | return ret; |
ae18c37a JB |
2866 | |
2867 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
2868 | GFP_KERNEL); | |
c75e8394 JB |
2869 | if (!fs_info->delayed_root) |
2870 | return -ENOMEM; | |
ae18c37a JB |
2871 | btrfs_init_delayed_root(fs_info->delayed_root); |
2872 | ||
a0a1db70 FM |
2873 | if (sb_rdonly(sb)) |
2874 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
2875 | ||
c75e8394 | 2876 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
2877 | } |
2878 | ||
97f4dd09 NB |
2879 | static int btrfs_uuid_rescan_kthread(void *data) |
2880 | { | |
0d031dc4 | 2881 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
2882 | int ret; |
2883 | ||
2884 | /* | |
2885 | * 1st step is to iterate through the existing UUID tree and | |
2886 | * to delete all entries that contain outdated data. | |
2887 | * 2nd step is to add all missing entries to the UUID tree. | |
2888 | */ | |
2889 | ret = btrfs_uuid_tree_iterate(fs_info); | |
2890 | if (ret < 0) { | |
c94bec2c JB |
2891 | if (ret != -EINTR) |
2892 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
2893 | ret); | |
97f4dd09 NB |
2894 | up(&fs_info->uuid_tree_rescan_sem); |
2895 | return ret; | |
2896 | } | |
2897 | return btrfs_uuid_scan_kthread(data); | |
2898 | } | |
2899 | ||
2900 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
2901 | { | |
2902 | struct task_struct *task; | |
2903 | ||
2904 | down(&fs_info->uuid_tree_rescan_sem); | |
2905 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
2906 | if (IS_ERR(task)) { | |
2907 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
2908 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
2909 | up(&fs_info->uuid_tree_rescan_sem); | |
2910 | return PTR_ERR(task); | |
2911 | } | |
2912 | ||
2913 | return 0; | |
2914 | } | |
2915 | ||
504b1596 FM |
2916 | static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
2917 | { | |
2918 | u64 root_objectid = 0; | |
2919 | struct btrfs_root *gang[8]; | |
2920 | int i = 0; | |
2921 | int err = 0; | |
2922 | unsigned int ret = 0; | |
2923 | ||
2924 | while (1) { | |
2925 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2926 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2927 | (void **)gang, root_objectid, | |
2928 | ARRAY_SIZE(gang)); | |
2929 | if (!ret) { | |
2930 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2931 | break; | |
2932 | } | |
2933 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
2934 | ||
2935 | for (i = 0; i < ret; i++) { | |
2936 | /* Avoid to grab roots in dead_roots. */ | |
2937 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
2938 | gang[i] = NULL; | |
2939 | continue; | |
2940 | } | |
2941 | /* Grab all the search result for later use. */ | |
2942 | gang[i] = btrfs_grab_root(gang[i]); | |
2943 | } | |
2944 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2945 | ||
2946 | for (i = 0; i < ret; i++) { | |
2947 | if (!gang[i]) | |
2948 | continue; | |
2949 | root_objectid = gang[i]->root_key.objectid; | |
2950 | err = btrfs_orphan_cleanup(gang[i]); | |
2951 | if (err) | |
2952 | goto out; | |
2953 | btrfs_put_root(gang[i]); | |
2954 | } | |
2955 | root_objectid++; | |
2956 | } | |
2957 | out: | |
2958 | /* Release the uncleaned roots due to error. */ | |
2959 | for (; i < ret; i++) { | |
2960 | if (gang[i]) | |
2961 | btrfs_put_root(gang[i]); | |
2962 | } | |
2963 | return err; | |
2964 | } | |
2965 | ||
44c0ca21 BB |
2966 | /* |
2967 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
2968 | * and btrfs_remount when remounting from read-only to read-write. | |
2969 | */ | |
2970 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
2971 | { | |
2972 | int ret; | |
94846229 | 2973 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
1d6a4fc8 | 2974 | bool rebuild_free_space_tree = false; |
8b228324 BB |
2975 | |
2976 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
2977 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
272efa30 JB |
2978 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) |
2979 | btrfs_warn(fs_info, | |
2980 | "'clear_cache' option is ignored with extent tree v2"); | |
2981 | else | |
2982 | rebuild_free_space_tree = true; | |
8b228324 BB |
2983 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && |
2984 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
2985 | btrfs_warn(fs_info, "free space tree is invalid"); | |
1d6a4fc8 | 2986 | rebuild_free_space_tree = true; |
8b228324 BB |
2987 | } |
2988 | ||
1d6a4fc8 QW |
2989 | if (rebuild_free_space_tree) { |
2990 | btrfs_info(fs_info, "rebuilding free space tree"); | |
2991 | ret = btrfs_rebuild_free_space_tree(fs_info); | |
8b228324 BB |
2992 | if (ret) { |
2993 | btrfs_warn(fs_info, | |
1d6a4fc8 QW |
2994 | "failed to rebuild free space tree: %d", ret); |
2995 | goto out; | |
2996 | } | |
2997 | } | |
2998 | ||
2999 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3000 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { | |
3001 | btrfs_info(fs_info, "disabling free space tree"); | |
3002 | ret = btrfs_delete_free_space_tree(fs_info); | |
3003 | if (ret) { | |
3004 | btrfs_warn(fs_info, | |
3005 | "failed to disable free space tree: %d", ret); | |
8b228324 BB |
3006 | goto out; |
3007 | } | |
3008 | } | |
44c0ca21 | 3009 | |
8d488a8c FM |
3010 | /* |
3011 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3012 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3013 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3014 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3015 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3016 | * item before the root's tree is deleted - this means that if we unmount | |
3017 | * or crash before the deletion completes, on the next mount we will not | |
3018 | * delete what remains of the tree because the orphan item does not | |
3019 | * exists anymore, which is what tells us we have a pending deletion. | |
3020 | */ | |
3021 | ret = btrfs_find_orphan_roots(fs_info); | |
3022 | if (ret) | |
3023 | goto out; | |
3024 | ||
44c0ca21 BB |
3025 | ret = btrfs_cleanup_fs_roots(fs_info); |
3026 | if (ret) | |
3027 | goto out; | |
3028 | ||
8f1c21d7 BB |
3029 | down_read(&fs_info->cleanup_work_sem); |
3030 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3031 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3032 | up_read(&fs_info->cleanup_work_sem); | |
3033 | goto out; | |
3034 | } | |
3035 | up_read(&fs_info->cleanup_work_sem); | |
3036 | ||
44c0ca21 | 3037 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3038 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3039 | mutex_unlock(&fs_info->cleaner_mutex); |
3040 | if (ret < 0) { | |
3041 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3042 | goto out; | |
3043 | } | |
3044 | ||
5011139a BB |
3045 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3046 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3047 | btrfs_info(fs_info, "creating free space tree"); | |
3048 | ret = btrfs_create_free_space_tree(fs_info); | |
3049 | if (ret) { | |
3050 | btrfs_warn(fs_info, | |
3051 | "failed to create free space tree: %d", ret); | |
3052 | goto out; | |
3053 | } | |
3054 | } | |
3055 | ||
94846229 BB |
3056 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3057 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3058 | if (ret) | |
3059 | goto out; | |
3060 | } | |
3061 | ||
44c0ca21 BB |
3062 | ret = btrfs_resume_balance_async(fs_info); |
3063 | if (ret) | |
3064 | goto out; | |
3065 | ||
3066 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3067 | if (ret) { | |
3068 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3069 | goto out; | |
3070 | } | |
3071 | ||
3072 | btrfs_qgroup_rescan_resume(fs_info); | |
3073 | ||
3074 | if (!fs_info->uuid_root) { | |
3075 | btrfs_info(fs_info, "creating UUID tree"); | |
3076 | ret = btrfs_create_uuid_tree(fs_info); | |
3077 | if (ret) { | |
3078 | btrfs_warn(fs_info, | |
3079 | "failed to create the UUID tree %d", ret); | |
3080 | goto out; | |
3081 | } | |
3082 | } | |
3083 | ||
3084 | out: | |
3085 | return ret; | |
3086 | } | |
3087 | ||
d7f67ac9 QW |
3088 | /* |
3089 | * Do various sanity and dependency checks of different features. | |
3090 | * | |
2ba48b20 QW |
3091 | * @is_rw_mount: If the mount is read-write. |
3092 | * | |
d7f67ac9 QW |
3093 | * This is the place for less strict checks (like for subpage or artificial |
3094 | * feature dependencies). | |
3095 | * | |
3096 | * For strict checks or possible corruption detection, see | |
3097 | * btrfs_validate_super(). | |
3098 | * | |
3099 | * This should be called after btrfs_parse_options(), as some mount options | |
3100 | * (space cache related) can modify on-disk format like free space tree and | |
3101 | * screw up certain feature dependencies. | |
3102 | */ | |
2ba48b20 | 3103 | int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) |
d7f67ac9 QW |
3104 | { |
3105 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3106 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3107 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3108 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3109 | ||
3110 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3111 | btrfs_err(fs_info, | |
3112 | "cannot mount because of unknown incompat features (0x%llx)", | |
3113 | incompat); | |
3114 | return -EINVAL; | |
3115 | } | |
3116 | ||
3117 | /* Runtime limitation for mixed block groups. */ | |
3118 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3119 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3120 | btrfs_err(fs_info, | |
3121 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3122 | fs_info->nodesize, fs_info->sectorsize); | |
3123 | return -EINVAL; | |
3124 | } | |
3125 | ||
3126 | /* Mixed backref is an always-enabled feature. */ | |
3127 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3128 | ||
3129 | /* Set compression related flags just in case. */ | |
3130 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3131 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3132 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3133 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3134 | ||
3135 | /* | |
3136 | * An ancient flag, which should really be marked deprecated. | |
3137 | * Such runtime limitation doesn't really need a incompat flag. | |
3138 | */ | |
3139 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3140 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3141 | ||
2ba48b20 | 3142 | if (compat_ro_unsupp && is_rw_mount) { |
d7f67ac9 QW |
3143 | btrfs_err(fs_info, |
3144 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3145 | compat_ro); | |
3146 | return -EINVAL; | |
3147 | } | |
3148 | ||
3149 | /* | |
3150 | * We have unsupported RO compat features, although RO mounted, we | |
3151 | * should not cause any metadata writes, including log replay. | |
3152 | * Or we could screw up whatever the new feature requires. | |
3153 | */ | |
3154 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3155 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3156 | btrfs_err(fs_info, | |
3157 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3158 | compat_ro); | |
3159 | return -EINVAL; | |
3160 | } | |
3161 | ||
3162 | /* | |
3163 | * Artificial limitations for block group tree, to force | |
3164 | * block-group-tree to rely on no-holes and free-space-tree. | |
3165 | */ | |
3166 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3167 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3168 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3169 | btrfs_err(fs_info, | |
3170 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3171 | return -EINVAL; | |
3172 | } | |
3173 | ||
3174 | /* | |
3175 | * Subpage runtime limitation on v1 cache. | |
3176 | * | |
3177 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3178 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3179 | * going to be deprecated anyway. | |
3180 | */ | |
3181 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3182 | btrfs_warn(fs_info, | |
3183 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3184 | PAGE_SIZE, fs_info->sectorsize); | |
3185 | return -EINVAL; | |
3186 | } | |
3187 | ||
3188 | /* This can be called by remount, we need to protect the super block. */ | |
3189 | spin_lock(&fs_info->super_lock); | |
3190 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3191 | spin_unlock(&fs_info->super_lock); | |
3192 | ||
3193 | return 0; | |
3194 | } | |
3195 | ||
ae18c37a JB |
3196 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3197 | char *options) | |
3198 | { | |
3199 | u32 sectorsize; | |
3200 | u32 nodesize; | |
3201 | u32 stripesize; | |
3202 | u64 generation; | |
ae18c37a | 3203 | u16 csum_type; |
ae18c37a JB |
3204 | struct btrfs_super_block *disk_super; |
3205 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3206 | struct btrfs_root *tree_root; | |
3207 | struct btrfs_root *chunk_root; | |
3208 | int ret; | |
ae18c37a JB |
3209 | int level; |
3210 | ||
8260edba | 3211 | ret = init_mount_fs_info(fs_info, sb); |
4871c33b | 3212 | if (ret) |
ae18c37a | 3213 | goto fail; |
53b381b3 | 3214 | |
ae18c37a JB |
3215 | /* These need to be init'ed before we start creating inodes and such. */ |
3216 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3217 | GFP_KERNEL); | |
3218 | fs_info->tree_root = tree_root; | |
3219 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3220 | GFP_KERNEL); | |
3221 | fs_info->chunk_root = chunk_root; | |
3222 | if (!tree_root || !chunk_root) { | |
4871c33b | 3223 | ret = -ENOMEM; |
c75e8394 | 3224 | goto fail; |
ae18c37a JB |
3225 | } |
3226 | ||
dcb2137c | 3227 | ret = btrfs_init_btree_inode(sb); |
4871c33b | 3228 | if (ret) |
c75e8394 | 3229 | goto fail; |
ae18c37a | 3230 | |
d24fa5c1 | 3231 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3232 | |
3233 | /* | |
3234 | * Read super block and check the signature bytes only | |
3235 | */ | |
d24fa5c1 | 3236 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d | 3237 | if (IS_ERR(disk_super)) { |
4871c33b | 3238 | ret = PTR_ERR(disk_super); |
16cdcec7 | 3239 | goto fail_alloc; |
20b45077 | 3240 | } |
39279cc3 | 3241 | |
2db31320 | 3242 | btrfs_info(fs_info, "first mount of filesystem %pU", disk_super->fsid); |
8dc3f22c | 3243 | /* |
260db43c | 3244 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3245 | * corrupted, we'll find out |
3246 | */ | |
8f32380d | 3247 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3248 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3249 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3250 | csum_type); |
4871c33b | 3251 | ret = -EINVAL; |
8f32380d | 3252 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3253 | goto fail_alloc; |
3254 | } | |
3255 | ||
83c68bbc SY |
3256 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3257 | ||
6d97c6e3 JT |
3258 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3259 | if (ret) { | |
8f32380d | 3260 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3261 | goto fail_alloc; |
3262 | } | |
3263 | ||
1104a885 DS |
3264 | /* |
3265 | * We want to check superblock checksum, the type is stored inside. | |
3266 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3267 | */ | |
3d17adea | 3268 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3269 | btrfs_err(fs_info, "superblock checksum mismatch"); |
4871c33b | 3270 | ret = -EINVAL; |
8f32380d | 3271 | btrfs_release_disk_super(disk_super); |
141386e1 | 3272 | goto fail_alloc; |
1104a885 DS |
3273 | } |
3274 | ||
3275 | /* | |
3276 | * super_copy is zeroed at allocation time and we never touch the | |
3277 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3278 | * the whole block of INFO_SIZE | |
3279 | */ | |
8f32380d JT |
3280 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3281 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3282 | |
fbc6feae NB |
3283 | disk_super = fs_info->super_copy; |
3284 | ||
fbc6feae NB |
3285 | memcpy(fs_info->super_for_commit, fs_info->super_copy, |
3286 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3287 | |
069ec957 | 3288 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3289 | if (ret) { |
05135f59 | 3290 | btrfs_err(fs_info, "superblock contains fatal errors"); |
4871c33b | 3291 | ret = -EINVAL; |
141386e1 | 3292 | goto fail_alloc; |
1104a885 DS |
3293 | } |
3294 | ||
4871c33b QW |
3295 | if (!btrfs_super_root(disk_super)) { |
3296 | btrfs_err(fs_info, "invalid superblock tree root bytenr"); | |
3297 | ret = -EINVAL; | |
141386e1 | 3298 | goto fail_alloc; |
4871c33b | 3299 | } |
0f7d52f4 | 3300 | |
acce952b | 3301 | /* check FS state, whether FS is broken. */ |
87533c47 | 3302 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
ae3364e5 | 3303 | WRITE_ONCE(fs_info->fs_error, -EUCLEAN); |
acce952b | 3304 | |
6f93e834 AJ |
3305 | /* Set up fs_info before parsing mount options */ |
3306 | nodesize = btrfs_super_nodesize(disk_super); | |
3307 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3308 | stripesize = sectorsize; | |
3309 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3310 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3311 | ||
3312 | fs_info->nodesize = nodesize; | |
3313 | fs_info->sectorsize = sectorsize; | |
3314 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3315 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3316 | fs_info->stripesize = stripesize; | |
3317 | ||
a6a8f22a JB |
3318 | /* |
3319 | * Handle the space caching options appropriately now that we have the | |
3320 | * super block loaded and validated. | |
3321 | */ | |
3322 | btrfs_set_free_space_cache_settings(fs_info); | |
3323 | ||
ad21f15b JB |
3324 | if (!btrfs_check_options(fs_info, &fs_info->mount_opt, sb->s_flags)) { |
3325 | ret = -EINVAL; | |
141386e1 | 3326 | goto fail_alloc; |
ad21f15b | 3327 | } |
dfe25020 | 3328 | |
2ba48b20 | 3329 | ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); |
4871c33b | 3330 | if (ret < 0) |
dc4d3168 | 3331 | goto fail_alloc; |
dc4d3168 | 3332 | |
ad21f15b JB |
3333 | /* |
3334 | * At this point our mount options are validated, if we set ->max_inline | |
3335 | * to something non-standard make sure we truncate it to sectorsize. | |
3336 | */ | |
3337 | fs_info->max_inline = min_t(u64, fs_info->max_inline, fs_info->sectorsize); | |
3338 | ||
8481dd80 QW |
3339 | if (sectorsize < PAGE_SIZE) { |
3340 | struct btrfs_subpage_info *subpage_info; | |
3341 | ||
95ea0486 QW |
3342 | btrfs_warn(fs_info, |
3343 | "read-write for sector size %u with page size %lu is experimental", | |
3344 | sectorsize, PAGE_SIZE); | |
8481dd80 | 3345 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
4871c33b QW |
3346 | if (!subpage_info) { |
3347 | ret = -ENOMEM; | |
8481dd80 | 3348 | goto fail_alloc; |
4871c33b | 3349 | } |
8481dd80 QW |
3350 | btrfs_init_subpage_info(subpage_info, sectorsize); |
3351 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3352 | } |
0bb3eb3e | 3353 | |
d21deec5 | 3354 | ret = btrfs_init_workqueues(fs_info); |
4871c33b | 3355 | if (ret) |
0dc3b84a | 3356 | goto fail_sb_buffer; |
4543df7e | 3357 | |
9e11ceee JK |
3358 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3359 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3360 | |
4e00422e | 3361 | /* Update the values for the current filesystem. */ |
a061fc8d CM |
3362 | sb->s_blocksize = sectorsize; |
3363 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3364 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3365 | |
925baedd | 3366 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3367 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3368 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3369 | if (ret) { |
05135f59 | 3370 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3371 | goto fail_sb_buffer; |
84eed90f | 3372 | } |
0b86a832 | 3373 | |
84234f3a | 3374 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3375 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3376 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3377 | generation, level); | |
3378 | if (ret) { | |
05135f59 | 3379 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3380 | goto fail_tree_roots; |
83121942 | 3381 | } |
0b86a832 | 3382 | |
e17cade2 | 3383 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3384 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3385 | BTRFS_UUID_SIZE); | |
e17cade2 | 3386 | |
5b4aacef | 3387 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3388 | if (ret) { |
05135f59 | 3389 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3390 | goto fail_tree_roots; |
2b82032c | 3391 | } |
0b86a832 | 3392 | |
8dabb742 | 3393 | /* |
bacce86a AJ |
3394 | * At this point we know all the devices that make this filesystem, |
3395 | * including the seed devices but we don't know yet if the replace | |
3396 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3397 | * filesystem but skip the replace target device which is checked |
bacce86a | 3398 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3399 | */ |
bacce86a | 3400 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3401 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3402 | btrfs_err(fs_info, "failed to read devices"); |
4871c33b | 3403 | ret = -EIO; |
a6b0d5c8 CM |
3404 | goto fail_tree_roots; |
3405 | } | |
3406 | ||
b8522a1e | 3407 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3408 | if (ret) |
b8522a1e | 3409 | goto fail_tree_roots; |
8929ecfa | 3410 | |
73651042 NA |
3411 | /* |
3412 | * Get zone type information of zoned block devices. This will also | |
3413 | * handle emulation of a zoned filesystem if a regular device has the | |
3414 | * zoned incompat feature flag set. | |
3415 | */ | |
3416 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3417 | if (ret) { | |
3418 | btrfs_err(fs_info, | |
4871c33b | 3419 | "zoned: failed to read device zone info: %d", ret); |
73651042 NA |
3420 | goto fail_block_groups; |
3421 | } | |
3422 | ||
75ec1db8 JB |
3423 | /* |
3424 | * If we have a uuid root and we're not being told to rescan we need to | |
3425 | * check the generation here so we can set the | |
3426 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3427 | * transaction during a balance or the log replay without updating the | |
3428 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3429 | * even though it was perfectly fine. | |
3430 | */ | |
3431 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3432 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3433 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3434 | ||
cf90d884 QW |
3435 | ret = btrfs_verify_dev_extents(fs_info); |
3436 | if (ret) { | |
3437 | btrfs_err(fs_info, | |
3438 | "failed to verify dev extents against chunks: %d", | |
3439 | ret); | |
3440 | goto fail_block_groups; | |
3441 | } | |
68310a5e ID |
3442 | ret = btrfs_recover_balance(fs_info); |
3443 | if (ret) { | |
05135f59 | 3444 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3445 | goto fail_block_groups; |
3446 | } | |
3447 | ||
733f4fbb SB |
3448 | ret = btrfs_init_dev_stats(fs_info); |
3449 | if (ret) { | |
05135f59 | 3450 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3451 | goto fail_block_groups; |
3452 | } | |
3453 | ||
8dabb742 SB |
3454 | ret = btrfs_init_dev_replace(fs_info); |
3455 | if (ret) { | |
05135f59 | 3456 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3457 | goto fail_block_groups; |
3458 | } | |
3459 | ||
b70f5097 NA |
3460 | ret = btrfs_check_zoned_mode(fs_info); |
3461 | if (ret) { | |
3462 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3463 | ret); | |
3464 | goto fail_block_groups; | |
3465 | } | |
3466 | ||
c6761a9e | 3467 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3468 | if (ret) { |
05135f59 DS |
3469 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3470 | ret); | |
b7c35e81 AJ |
3471 | goto fail_block_groups; |
3472 | } | |
3473 | ||
96f3136e | 3474 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3475 | if (ret) { |
05135f59 | 3476 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3477 | goto fail_fsdev_sysfs; |
c59021f8 | 3478 | } |
3479 | ||
c59021f8 | 3480 | ret = btrfs_init_space_info(fs_info); |
3481 | if (ret) { | |
05135f59 | 3482 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3483 | goto fail_sysfs; |
c59021f8 | 3484 | } |
3485 | ||
5b4aacef | 3486 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3487 | if (ret) { |
05135f59 | 3488 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3489 | goto fail_sysfs; |
1b1d1f66 | 3490 | } |
4330e183 | 3491 | |
16beac87 NA |
3492 | btrfs_free_zone_cache(fs_info); |
3493 | ||
a7e1ac7b NA |
3494 | btrfs_check_active_zone_reservation(fs_info); |
3495 | ||
5c78a5e7 AJ |
3496 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3497 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3498 | btrfs_warn(fs_info, |
52042d8e | 3499 | "writable mount is not allowed due to too many missing devices"); |
4871c33b | 3500 | ret = -EINVAL; |
2365dd3c | 3501 | goto fail_sysfs; |
292fd7fc | 3502 | } |
9078a3e1 | 3503 | |
33c44184 | 3504 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3505 | "btrfs-cleaner"); |
4871c33b QW |
3506 | if (IS_ERR(fs_info->cleaner_kthread)) { |
3507 | ret = PTR_ERR(fs_info->cleaner_kthread); | |
2365dd3c | 3508 | goto fail_sysfs; |
4871c33b | 3509 | } |
a74a4b97 CM |
3510 | |
3511 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3512 | tree_root, | |
3513 | "btrfs-transaction"); | |
4871c33b QW |
3514 | if (IS_ERR(fs_info->transaction_kthread)) { |
3515 | ret = PTR_ERR(fs_info->transaction_kthread); | |
3f157a2f | 3516 | goto fail_cleaner; |
4871c33b | 3517 | } |
a74a4b97 | 3518 | |
bcef60f2 AJ |
3519 | ret = btrfs_read_qgroup_config(fs_info); |
3520 | if (ret) | |
3521 | goto fail_trans_kthread; | |
21adbd5c | 3522 | |
fd708b81 JB |
3523 | if (btrfs_build_ref_tree(fs_info)) |
3524 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3525 | ||
96da0919 QW |
3526 | /* do not make disk changes in broken FS or nologreplay is given */ |
3527 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3528 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3529 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3530 | ret = btrfs_replay_log(fs_info, fs_devices); |
4871c33b | 3531 | if (ret) |
28c16cbb | 3532 | goto fail_qgroup; |
e02119d5 | 3533 | } |
1a40e23b | 3534 | |
56e9357a | 3535 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 | 3536 | if (IS_ERR(fs_info->fs_root)) { |
4871c33b QW |
3537 | ret = PTR_ERR(fs_info->fs_root); |
3538 | btrfs_warn(fs_info, "failed to read fs tree: %d", ret); | |
315bf8ef | 3539 | fs_info->fs_root = NULL; |
bcef60f2 | 3540 | goto fail_qgroup; |
3140c9a3 | 3541 | } |
c289811c | 3542 | |
bc98a42c | 3543 | if (sb_rdonly(sb)) |
83e3a40a | 3544 | return 0; |
59641015 | 3545 | |
44c0ca21 | 3546 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3547 | if (ret) { |
6bccf3ab | 3548 | close_ctree(fs_info); |
2b6ba629 | 3549 | return ret; |
e3acc2a6 | 3550 | } |
b0643e59 | 3551 | btrfs_discard_resume(fs_info); |
b382a324 | 3552 | |
44c0ca21 BB |
3553 | if (fs_info->uuid_root && |
3554 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3555 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3556 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3557 | ret = btrfs_check_uuid_tree(fs_info); |
3558 | if (ret) { | |
05135f59 DS |
3559 | btrfs_warn(fs_info, |
3560 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3561 | close_ctree(fs_info); |
70f80175 SB |
3562 | return ret; |
3563 | } | |
f7a81ea4 | 3564 | } |
94846229 | 3565 | |
afcdd129 | 3566 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3567 | |
b4be6aef JB |
3568 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3569 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3570 | wake_up_process(fs_info->cleaner_kthread); | |
3571 | ||
ad2b2c80 | 3572 | return 0; |
39279cc3 | 3573 | |
bcef60f2 AJ |
3574 | fail_qgroup: |
3575 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3576 | fail_trans_kthread: |
3577 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3578 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3579 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3580 | fail_cleaner: |
a74a4b97 | 3581 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3582 | |
3583 | /* | |
3584 | * make sure we're done with the btree inode before we stop our | |
3585 | * kthreads | |
3586 | */ | |
3587 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3588 | |
2365dd3c | 3589 | fail_sysfs: |
6618a59b | 3590 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3591 | |
b7c35e81 AJ |
3592 | fail_fsdev_sysfs: |
3593 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3594 | ||
1b1d1f66 | 3595 | fail_block_groups: |
54067ae9 | 3596 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3597 | |
3598 | fail_tree_roots: | |
9e3aa805 JB |
3599 | if (fs_info->data_reloc_root) |
3600 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3601 | free_root_pointers(fs_info, true); |
2b8195bb | 3602 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3603 | |
39279cc3 | 3604 | fail_sb_buffer: |
7abadb64 | 3605 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3606 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3607 | fail_alloc: |
7dc66abb | 3608 | btrfs_mapping_tree_free(fs_info); |
586e46e2 | 3609 | |
4543df7e | 3610 | iput(fs_info->btree_inode); |
7e662854 | 3611 | fail: |
586e46e2 | 3612 | btrfs_close_devices(fs_info->fs_devices); |
4871c33b QW |
3613 | ASSERT(ret < 0); |
3614 | return ret; | |
eb60ceac | 3615 | } |
663faf9f | 3616 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3617 | |
314b6dd0 | 3618 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3619 | { |
314b6dd0 JT |
3620 | struct btrfs_device *device = bio->bi_private; |
3621 | struct bio_vec *bvec; | |
3622 | struct bvec_iter_all iter_all; | |
3623 | struct page *page; | |
3624 | ||
3625 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3626 | page = bvec->bv_page; | |
3627 | ||
3628 | if (bio->bi_status) { | |
3629 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3630 | "lost page write due to IO error on %s (%d)", | |
cb3e217b | 3631 | btrfs_dev_name(device), |
314b6dd0 JT |
3632 | blk_status_to_errno(bio->bi_status)); |
3633 | ClearPageUptodate(page); | |
3634 | SetPageError(page); | |
3635 | btrfs_dev_stat_inc_and_print(device, | |
3636 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3637 | } else { | |
3638 | SetPageUptodate(page); | |
3639 | } | |
3640 | ||
3641 | put_page(page); | |
3642 | unlock_page(page); | |
f2984462 | 3643 | } |
314b6dd0 JT |
3644 | |
3645 | bio_put(bio); | |
f2984462 CM |
3646 | } |
3647 | ||
8f32380d | 3648 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3649 | int copy_num, bool drop_cache) |
29c36d72 | 3650 | { |
29c36d72 | 3651 | struct btrfs_super_block *super; |
8f32380d | 3652 | struct page *page; |
12659251 | 3653 | u64 bytenr, bytenr_orig; |
8f32380d | 3654 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3655 | int ret; |
3656 | ||
3657 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3658 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3659 | if (ret == -ENOENT) | |
3660 | return ERR_PTR(-EINVAL); | |
3661 | else if (ret) | |
3662 | return ERR_PTR(ret); | |
29c36d72 | 3663 | |
cda00eba | 3664 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3665 | return ERR_PTR(-EINVAL); |
29c36d72 | 3666 | |
a05d3c91 QW |
3667 | if (drop_cache) { |
3668 | /* This should only be called with the primary sb. */ | |
3669 | ASSERT(copy_num == 0); | |
3670 | ||
3671 | /* | |
3672 | * Drop the page of the primary superblock, so later read will | |
3673 | * always read from the device. | |
3674 | */ | |
3675 | invalidate_inode_pages2_range(mapping, | |
3676 | bytenr >> PAGE_SHIFT, | |
3677 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3678 | } | |
3679 | ||
8f32380d JT |
3680 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3681 | if (IS_ERR(page)) | |
3682 | return ERR_CAST(page); | |
29c36d72 | 3683 | |
8f32380d | 3684 | super = page_address(page); |
96c2e067 AJ |
3685 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3686 | btrfs_release_disk_super(super); | |
3687 | return ERR_PTR(-ENODATA); | |
3688 | } | |
3689 | ||
12659251 | 3690 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3691 | btrfs_release_disk_super(super); |
3692 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3693 | } |
3694 | ||
8f32380d | 3695 | return super; |
29c36d72 AJ |
3696 | } |
3697 | ||
3698 | ||
8f32380d | 3699 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3700 | { |
8f32380d | 3701 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3702 | int i; |
3703 | u64 transid = 0; | |
a512bbf8 YZ |
3704 | |
3705 | /* we would like to check all the supers, but that would make | |
3706 | * a btrfs mount succeed after a mkfs from a different FS. | |
3707 | * So, we need to add a special mount option to scan for | |
3708 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3709 | */ | |
3710 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3711 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3712 | if (IS_ERR(super)) |
a512bbf8 YZ |
3713 | continue; |
3714 | ||
a512bbf8 | 3715 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3716 | if (latest) |
3717 | btrfs_release_disk_super(super); | |
3718 | ||
3719 | latest = super; | |
a512bbf8 | 3720 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3721 | } |
3722 | } | |
92fc03fb | 3723 | |
8f32380d | 3724 | return super; |
a512bbf8 YZ |
3725 | } |
3726 | ||
4eedeb75 | 3727 | /* |
abbb3b8e | 3728 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3729 | * pages we use for writing are locked. |
4eedeb75 | 3730 | * |
abbb3b8e DS |
3731 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3732 | * the expected device size at commit time. Note that max_mirrors must be | |
3733 | * same for write and wait phases. | |
4eedeb75 | 3734 | * |
314b6dd0 | 3735 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3736 | */ |
a512bbf8 | 3737 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3738 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3739 | { |
d5178578 | 3740 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3741 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3742 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3743 | int i; |
a512bbf8 | 3744 | int errors = 0; |
12659251 NA |
3745 | int ret; |
3746 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3747 | |
3748 | if (max_mirrors == 0) | |
3749 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3750 | ||
d5178578 JT |
3751 | shash->tfm = fs_info->csum_shash; |
3752 | ||
a512bbf8 | 3753 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3754 | struct page *page; |
3755 | struct bio *bio; | |
3756 | struct btrfs_super_block *disk_super; | |
3757 | ||
12659251 NA |
3758 | bytenr_orig = btrfs_sb_offset(i); |
3759 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3760 | if (ret == -ENOENT) { | |
3761 | continue; | |
3762 | } else if (ret < 0) { | |
3763 | btrfs_err(device->fs_info, | |
3764 | "couldn't get super block location for mirror %d", | |
3765 | i); | |
3766 | errors++; | |
3767 | continue; | |
3768 | } | |
935e5cc9 MX |
3769 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3770 | device->commit_total_bytes) | |
a512bbf8 YZ |
3771 | break; |
3772 | ||
12659251 | 3773 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3774 | |
fd08001f EB |
3775 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3776 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3777 | sb->csum); | |
4eedeb75 | 3778 | |
314b6dd0 JT |
3779 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3780 | GFP_NOFS); | |
3781 | if (!page) { | |
abbb3b8e | 3782 | btrfs_err(device->fs_info, |
314b6dd0 | 3783 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3784 | bytenr); |
3785 | errors++; | |
4eedeb75 | 3786 | continue; |
abbb3b8e | 3787 | } |
634554dc | 3788 | |
314b6dd0 JT |
3789 | /* Bump the refcount for wait_dev_supers() */ |
3790 | get_page(page); | |
a512bbf8 | 3791 | |
314b6dd0 JT |
3792 | disk_super = page_address(page); |
3793 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3794 | |
314b6dd0 JT |
3795 | /* |
3796 | * Directly use bios here instead of relying on the page cache | |
3797 | * to do I/O, so we don't lose the ability to do integrity | |
3798 | * checking. | |
3799 | */ | |
07888c66 CH |
3800 | bio = bio_alloc(device->bdev, 1, |
3801 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
3802 | GFP_NOFS); | |
314b6dd0 JT |
3803 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
3804 | bio->bi_private = device; | |
3805 | bio->bi_end_io = btrfs_end_super_write; | |
3806 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3807 | offset_in_page(bytenr)); | |
a512bbf8 | 3808 | |
387125fc | 3809 | /* |
314b6dd0 JT |
3810 | * We FUA only the first super block. The others we allow to |
3811 | * go down lazy and there's a short window where the on-disk | |
3812 | * copies might still contain the older version. | |
387125fc | 3813 | */ |
1b9e619c | 3814 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 | 3815 | bio->bi_opf |= REQ_FUA; |
58ff51f1 | 3816 | submit_bio(bio); |
8376d9e1 NA |
3817 | |
3818 | if (btrfs_advance_sb_log(device, i)) | |
3819 | errors++; | |
a512bbf8 YZ |
3820 | } |
3821 | return errors < i ? 0 : -1; | |
3822 | } | |
3823 | ||
abbb3b8e DS |
3824 | /* |
3825 | * Wait for write completion of superblocks done by write_dev_supers, | |
3826 | * @max_mirrors same for write and wait phases. | |
3827 | * | |
314b6dd0 | 3828 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3829 | * date. |
3830 | */ | |
3831 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3832 | { | |
abbb3b8e DS |
3833 | int i; |
3834 | int errors = 0; | |
b6a535fa | 3835 | bool primary_failed = false; |
12659251 | 3836 | int ret; |
abbb3b8e DS |
3837 | u64 bytenr; |
3838 | ||
3839 | if (max_mirrors == 0) | |
3840 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3841 | ||
3842 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3843 | struct page *page; |
3844 | ||
12659251 NA |
3845 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3846 | if (ret == -ENOENT) { | |
3847 | break; | |
3848 | } else if (ret < 0) { | |
3849 | errors++; | |
3850 | if (i == 0) | |
3851 | primary_failed = true; | |
3852 | continue; | |
3853 | } | |
abbb3b8e DS |
3854 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3855 | device->commit_total_bytes) | |
3856 | break; | |
3857 | ||
314b6dd0 JT |
3858 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3859 | bytenr >> PAGE_SHIFT); | |
3860 | if (!page) { | |
abbb3b8e | 3861 | errors++; |
b6a535fa HM |
3862 | if (i == 0) |
3863 | primary_failed = true; | |
abbb3b8e DS |
3864 | continue; |
3865 | } | |
314b6dd0 JT |
3866 | /* Page is submitted locked and unlocked once the IO completes */ |
3867 | wait_on_page_locked(page); | |
3868 | if (PageError(page)) { | |
abbb3b8e | 3869 | errors++; |
b6a535fa HM |
3870 | if (i == 0) |
3871 | primary_failed = true; | |
3872 | } | |
abbb3b8e | 3873 | |
314b6dd0 JT |
3874 | /* Drop our reference */ |
3875 | put_page(page); | |
abbb3b8e | 3876 | |
314b6dd0 JT |
3877 | /* Drop the reference from the writing run */ |
3878 | put_page(page); | |
abbb3b8e DS |
3879 | } |
3880 | ||
b6a535fa HM |
3881 | /* log error, force error return */ |
3882 | if (primary_failed) { | |
3883 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3884 | device->devid); | |
3885 | return -1; | |
3886 | } | |
3887 | ||
abbb3b8e DS |
3888 | return errors < i ? 0 : -1; |
3889 | } | |
3890 | ||
387125fc CM |
3891 | /* |
3892 | * endio for the write_dev_flush, this will wake anyone waiting | |
3893 | * for the barrier when it is done | |
3894 | */ | |
4246a0b6 | 3895 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3896 | { |
f9e69aa9 | 3897 | bio_uninit(bio); |
e0ae9994 | 3898 | complete(bio->bi_private); |
387125fc CM |
3899 | } |
3900 | ||
3901 | /* | |
4fc6441a AJ |
3902 | * Submit a flush request to the device if it supports it. Error handling is |
3903 | * done in the waiting counterpart. | |
387125fc | 3904 | */ |
4fc6441a | 3905 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3906 | { |
f9e69aa9 | 3907 | struct bio *bio = &device->flush_bio; |
387125fc | 3908 | |
bfd3ea94 AJ |
3909 | device->last_flush_error = BLK_STS_OK; |
3910 | ||
f9e69aa9 CH |
3911 | bio_init(bio, device->bdev, NULL, 0, |
3912 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 3913 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
3914 | init_completion(&device->flush_wait); |
3915 | bio->bi_private = &device->flush_wait; | |
58ff51f1 | 3916 | submit_bio(bio); |
1c3063b6 | 3917 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3918 | } |
387125fc | 3919 | |
4fc6441a AJ |
3920 | /* |
3921 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
1b465784 | 3922 | * Return true for any error, and false otherwise. |
4fc6441a | 3923 | */ |
1b465784 | 3924 | static bool wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3925 | { |
f9e69aa9 | 3926 | struct bio *bio = &device->flush_bio; |
387125fc | 3927 | |
7e812f20 | 3928 | if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
1b465784 | 3929 | return false; |
387125fc | 3930 | |
2980d574 | 3931 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3932 | |
bfd3ea94 AJ |
3933 | if (bio->bi_status) { |
3934 | device->last_flush_error = bio->bi_status; | |
3935 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); | |
1b465784 | 3936 | return true; |
bfd3ea94 AJ |
3937 | } |
3938 | ||
1b465784 | 3939 | return false; |
387125fc | 3940 | } |
387125fc | 3941 | |
387125fc CM |
3942 | /* |
3943 | * send an empty flush down to each device in parallel, | |
3944 | * then wait for them | |
3945 | */ | |
3946 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3947 | { | |
3948 | struct list_head *head; | |
3949 | struct btrfs_device *dev; | |
5af3e8cc | 3950 | int errors_wait = 0; |
387125fc | 3951 | |
1538e6c5 | 3952 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3953 | /* send down all the barriers */ |
3954 | head = &info->fs_devices->devices; | |
1538e6c5 | 3955 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3956 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3957 | continue; |
cea7c8bf | 3958 | if (!dev->bdev) |
387125fc | 3959 | continue; |
e12c9621 | 3960 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3961 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3962 | continue; |
3963 | ||
4fc6441a | 3964 | write_dev_flush(dev); |
387125fc CM |
3965 | } |
3966 | ||
3967 | /* wait for all the barriers */ | |
1538e6c5 | 3968 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3969 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3970 | continue; |
387125fc | 3971 | if (!dev->bdev) { |
5af3e8cc | 3972 | errors_wait++; |
387125fc CM |
3973 | continue; |
3974 | } | |
e12c9621 | 3975 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3976 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3977 | continue; |
3978 | ||
1b465784 | 3979 | if (wait_dev_flush(dev)) |
5af3e8cc | 3980 | errors_wait++; |
401b41e5 AJ |
3981 | } |
3982 | ||
de38a206 AJ |
3983 | /* |
3984 | * Checks last_flush_error of disks in order to determine the device | |
3985 | * state. | |
3986 | */ | |
3987 | if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) | |
3988 | return -EIO; | |
3989 | ||
387125fc CM |
3990 | return 0; |
3991 | } | |
3992 | ||
943c6e99 ZL |
3993 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3994 | { | |
8789f4fe ZL |
3995 | int raid_type; |
3996 | int min_tolerated = INT_MAX; | |
943c6e99 | 3997 | |
8789f4fe ZL |
3998 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3999 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4000 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4001 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4002 | tolerated_failures); | |
943c6e99 | 4003 | |
8789f4fe ZL |
4004 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4005 | if (raid_type == BTRFS_RAID_SINGLE) | |
4006 | continue; | |
41a6e891 | 4007 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4008 | continue; |
8c3e3582 | 4009 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4010 | btrfs_raid_array[raid_type]. |
4011 | tolerated_failures); | |
4012 | } | |
943c6e99 | 4013 | |
8789f4fe | 4014 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4015 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4016 | min_tolerated = 0; |
4017 | } | |
4018 | ||
4019 | return min_tolerated; | |
943c6e99 ZL |
4020 | } |
4021 | ||
eece6a9c | 4022 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4023 | { |
e5e9a520 | 4024 | struct list_head *head; |
f2984462 | 4025 | struct btrfs_device *dev; |
a061fc8d | 4026 | struct btrfs_super_block *sb; |
f2984462 | 4027 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4028 | int ret; |
4029 | int do_barriers; | |
a236aed1 CM |
4030 | int max_errors; |
4031 | int total_errors = 0; | |
a061fc8d | 4032 | u64 flags; |
f2984462 | 4033 | |
0b246afa | 4034 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4035 | |
4036 | /* | |
4037 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4038 | * not from fsync where the tree roots in fs_info have not | |
4039 | * been consistent on disk. | |
4040 | */ | |
4041 | if (max_mirrors == 0) | |
4042 | backup_super_roots(fs_info); | |
f2984462 | 4043 | |
0b246afa | 4044 | sb = fs_info->super_for_commit; |
a061fc8d | 4045 | dev_item = &sb->dev_item; |
e5e9a520 | 4046 | |
0b246afa JM |
4047 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4048 | head = &fs_info->fs_devices->devices; | |
4049 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4050 | |
5af3e8cc | 4051 | if (do_barriers) { |
0b246afa | 4052 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4053 | if (ret) { |
4054 | mutex_unlock( | |
0b246afa JM |
4055 | &fs_info->fs_devices->device_list_mutex); |
4056 | btrfs_handle_fs_error(fs_info, ret, | |
4057 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4058 | return ret; |
4059 | } | |
4060 | } | |
387125fc | 4061 | |
1538e6c5 | 4062 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4063 | if (!dev->bdev) { |
4064 | total_errors++; | |
4065 | continue; | |
4066 | } | |
e12c9621 | 4067 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4068 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4069 | continue; |
4070 | ||
2b82032c | 4071 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4072 | btrfs_set_stack_device_type(dev_item, dev->type); |
4073 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4074 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4075 | dev->commit_total_bytes); |
ce7213c7 MX |
4076 | btrfs_set_stack_device_bytes_used(dev_item, |
4077 | dev->commit_bytes_used); | |
a061fc8d CM |
4078 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4079 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4080 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4081 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4082 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4083 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4084 | |
a061fc8d CM |
4085 | flags = btrfs_super_flags(sb); |
4086 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4087 | ||
75cb857d QW |
4088 | ret = btrfs_validate_write_super(fs_info, sb); |
4089 | if (ret < 0) { | |
4090 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4091 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4092 | "unexpected superblock corruption detected"); | |
4093 | return -EUCLEAN; | |
4094 | } | |
4095 | ||
abbb3b8e | 4096 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4097 | if (ret) |
4098 | total_errors++; | |
f2984462 | 4099 | } |
a236aed1 | 4100 | if (total_errors > max_errors) { |
0b246afa JM |
4101 | btrfs_err(fs_info, "%d errors while writing supers", |
4102 | total_errors); | |
4103 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4104 | |
9d565ba4 | 4105 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4106 | btrfs_handle_fs_error(fs_info, -EIO, |
4107 | "%d errors while writing supers", | |
4108 | total_errors); | |
9d565ba4 | 4109 | return -EIO; |
a236aed1 | 4110 | } |
f2984462 | 4111 | |
a512bbf8 | 4112 | total_errors = 0; |
1538e6c5 | 4113 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4114 | if (!dev->bdev) |
4115 | continue; | |
e12c9621 | 4116 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4117 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4118 | continue; |
4119 | ||
abbb3b8e | 4120 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4121 | if (ret) |
4122 | total_errors++; | |
f2984462 | 4123 | } |
0b246afa | 4124 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4125 | if (total_errors > max_errors) { |
0b246afa JM |
4126 | btrfs_handle_fs_error(fs_info, -EIO, |
4127 | "%d errors while writing supers", | |
4128 | total_errors); | |
79787eaa | 4129 | return -EIO; |
a236aed1 | 4130 | } |
f2984462 CM |
4131 | return 0; |
4132 | } | |
4133 | ||
cb517eab MX |
4134 | /* Drop a fs root from the radix tree and free it. */ |
4135 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4136 | struct btrfs_root *root) | |
2619ba1f | 4137 | { |
4785e24f JB |
4138 | bool drop_ref = false; |
4139 | ||
fc7cbcd4 DS |
4140 | spin_lock(&fs_info->fs_roots_radix_lock); |
4141 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4142 | (unsigned long)root->root_key.objectid); | |
4143 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4144 | drop_ref = true; |
fc7cbcd4 | 4145 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4146 | |
84961539 | 4147 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4148 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4149 | if (root->reloc_root) { |
00246528 | 4150 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4151 | root->reloc_root = NULL; |
4152 | } | |
4153 | } | |
3321719e | 4154 | |
4785e24f JB |
4155 | if (drop_ref) |
4156 | btrfs_put_root(root); | |
2619ba1f CM |
4157 | } |
4158 | ||
6bccf3ab | 4159 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4160 | { |
6bccf3ab | 4161 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4162 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4163 | |
0b246afa | 4164 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4165 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4166 | mutex_unlock(&fs_info->cleaner_mutex); |
4167 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4168 | |
4169 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4170 | down_write(&fs_info->cleanup_work_sem); |
4171 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4172 | |
7a7eaa40 | 4173 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4174 | if (IS_ERR(trans)) |
4175 | return PTR_ERR(trans); | |
3a45bb20 | 4176 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4177 | } |
4178 | ||
36c86a9e QW |
4179 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4180 | { | |
4181 | struct btrfs_transaction *trans; | |
4182 | struct btrfs_transaction *tmp; | |
4183 | bool found = false; | |
4184 | ||
4185 | if (list_empty(&fs_info->trans_list)) | |
4186 | return; | |
4187 | ||
4188 | /* | |
4189 | * This function is only called at the very end of close_ctree(), | |
4190 | * thus no other running transaction, no need to take trans_lock. | |
4191 | */ | |
4192 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4193 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4194 | struct extent_state *cached = NULL; | |
4195 | u64 dirty_bytes = 0; | |
4196 | u64 cur = 0; | |
4197 | u64 found_start; | |
4198 | u64 found_end; | |
4199 | ||
4200 | found = true; | |
e5860f82 | 4201 | while (find_first_extent_bit(&trans->dirty_pages, cur, |
36c86a9e QW |
4202 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { |
4203 | dirty_bytes += found_end + 1 - found_start; | |
4204 | cur = found_end + 1; | |
4205 | } | |
4206 | btrfs_warn(fs_info, | |
4207 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4208 | trans->transid, dirty_bytes); | |
4209 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4210 | ||
4211 | if (trans == fs_info->running_transaction) | |
4212 | fs_info->running_transaction = NULL; | |
4213 | list_del_init(&trans->list); | |
4214 | ||
4215 | btrfs_put_transaction(trans); | |
4216 | trace_btrfs_transaction_commit(fs_info); | |
4217 | } | |
4218 | ASSERT(!found); | |
4219 | } | |
4220 | ||
b105e927 | 4221 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4222 | { |
c146afad YZ |
4223 | int ret; |
4224 | ||
afcdd129 | 4225 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4226 | |
8a1f1e3d FM |
4227 | /* |
4228 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4229 | * clear that bit and wake up relocation so it can stop. | |
4230 | * We must do this before stopping the block group reclaim task, because | |
4231 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4232 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4233 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4234 | * return 1. | |
4235 | */ | |
4236 | btrfs_wake_unfinished_drop(fs_info); | |
4237 | ||
31e70e52 FM |
4238 | /* |
4239 | * We may have the reclaim task running and relocating a data block group, | |
4240 | * in which case it may create delayed iputs. So stop it before we park | |
4241 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4242 | * parking the cleaner, and that can make the async reclaim task to hang | |
4243 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4244 | * parked and can not run delayed iputs - this will make us hang when | |
4245 | * trying to stop the async reclaim task. | |
4246 | */ | |
4247 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4248 | /* |
4249 | * We don't want the cleaner to start new transactions, add more delayed | |
4250 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4251 | * because that frees the task_struct, and the transaction kthread might | |
4252 | * still try to wake up the cleaner. | |
4253 | */ | |
4254 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4255 | |
7343dd61 | 4256 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4257 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4258 | |
803b2f54 SB |
4259 | /* wait for the uuid_scan task to finish */ |
4260 | down(&fs_info->uuid_tree_rescan_sem); | |
4261 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4262 | up(&fs_info->uuid_tree_rescan_sem); | |
4263 | ||
837d5b6e | 4264 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4265 | btrfs_pause_balance(fs_info); |
837d5b6e | 4266 | |
8dabb742 SB |
4267 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4268 | ||
aa1b8cd4 | 4269 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4270 | |
4271 | /* wait for any defraggers to finish */ | |
4272 | wait_event(fs_info->transaction_wait, | |
4273 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4274 | ||
4275 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4276 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4277 | |
a362bb86 FM |
4278 | /* |
4279 | * After we parked the cleaner kthread, ordered extents may have | |
4280 | * completed and created new delayed iputs. If one of the async reclaim | |
4281 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4282 | * can hang forever trying to stop it, because if a delayed iput is | |
4283 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4284 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4285 | * no one else to run iputs. | |
4286 | * | |
4287 | * So wait for all ongoing ordered extents to complete and then run | |
4288 | * delayed iputs. This works because once we reach this point no one | |
4289 | * can either create new ordered extents nor create delayed iputs | |
4290 | * through some other means. | |
4291 | * | |
4292 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4293 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4294 | * but the delayed iput for the respective inode is made only when doing | |
4295 | * the final btrfs_put_ordered_extent() (which must happen at | |
4296 | * btrfs_finish_ordered_io() when we are unmounting). | |
4297 | */ | |
4298 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4299 | /* Ordered extents for free space inodes. */ | |
4300 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4301 | btrfs_run_delayed_iputs(fs_info); | |
4302 | ||
21c7e756 | 4303 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4304 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4305 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4306 | |
b0643e59 DZ |
4307 | /* Cancel or finish ongoing discard work */ |
4308 | btrfs_discard_cleanup(fs_info); | |
4309 | ||
bc98a42c | 4310 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4311 | /* |
d6fd0ae2 OS |
4312 | * The cleaner kthread is stopped, so do one final pass over |
4313 | * unused block groups. | |
e44163e1 | 4314 | */ |
0b246afa | 4315 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4316 | |
f0cc2cd7 FM |
4317 | /* |
4318 | * There might be existing delayed inode workers still running | |
4319 | * and holding an empty delayed inode item. We must wait for | |
4320 | * them to complete first because they can create a transaction. | |
4321 | * This happens when someone calls btrfs_balance_delayed_items() | |
4322 | * and then a transaction commit runs the same delayed nodes | |
4323 | * before any delayed worker has done something with the nodes. | |
4324 | * We must wait for any worker here and not at transaction | |
4325 | * commit time since that could cause a deadlock. | |
4326 | * This is a very rare case. | |
4327 | */ | |
4328 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4329 | ||
6bccf3ab | 4330 | ret = btrfs_commit_super(fs_info); |
acce952b | 4331 | if (ret) |
04892340 | 4332 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4333 | } |
4334 | ||
84961539 | 4335 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4336 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4337 | |
e3029d9f AV |
4338 | kthread_stop(fs_info->transaction_kthread); |
4339 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4340 | |
e187831e | 4341 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4342 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4343 | |
5958253c QW |
4344 | if (btrfs_check_quota_leak(fs_info)) { |
4345 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4346 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4347 | } | |
4348 | ||
04892340 | 4349 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4350 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4351 | |
963d678b | 4352 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4353 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4354 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4355 | } |
bcc63abb | 4356 | |
5deb17e1 | 4357 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4358 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4359 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4360 | |
6618a59b | 4361 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4362 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4363 | |
1a4319cc LB |
4364 | btrfs_put_block_group_cache(fs_info); |
4365 | ||
de348ee0 WS |
4366 | /* |
4367 | * we must make sure there is not any read request to | |
4368 | * submit after we stopping all workers. | |
4369 | */ | |
4370 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4371 | btrfs_stop_all_workers(fs_info); |
4372 | ||
0a31daa4 | 4373 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4374 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4375 | |
afcdd129 | 4376 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4377 | free_root_pointers(fs_info, true); |
8c38938c | 4378 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4379 | |
4e19443d JB |
4380 | /* |
4381 | * We must free the block groups after dropping the fs_roots as we could | |
4382 | * have had an IO error and have left over tree log blocks that aren't | |
4383 | * cleaned up until the fs roots are freed. This makes the block group | |
4384 | * accounting appear to be wrong because there's pending reserved bytes, | |
4385 | * so make sure we do the block group cleanup afterwards. | |
4386 | */ | |
4387 | btrfs_free_block_groups(fs_info); | |
4388 | ||
13e6c37b | 4389 | iput(fs_info->btree_inode); |
d6bfde87 | 4390 | |
7dc66abb | 4391 | btrfs_mapping_tree_free(fs_info); |
68c94e55 | 4392 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4393 | } |
4394 | ||
50564b65 FM |
4395 | void btrfs_mark_buffer_dirty(struct btrfs_trans_handle *trans, |
4396 | struct extent_buffer *buf) | |
5f39d397 | 4397 | { |
2f4d60df | 4398 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4399 | u64 transid = btrfs_header_generation(buf); |
b4ce94de | 4400 | |
06ea65a3 JB |
4401 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4402 | /* | |
4403 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4404 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4405 | * outside of the sanity tests. |
4406 | */ | |
b0132a3b | 4407 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4408 | return; |
4409 | #endif | |
50564b65 FM |
4410 | /* This is an active transaction (its state < TRANS_STATE_UNBLOCKED). */ |
4411 | ASSERT(trans->transid == fs_info->generation); | |
49d0c642 | 4412 | btrfs_assert_tree_write_locked(buf); |
4ebe8d47 | 4413 | if (unlikely(transid != fs_info->generation)) { |
50564b65 | 4414 | btrfs_abort_transaction(trans, -EUCLEAN); |
20cbe460 FM |
4415 | btrfs_crit(fs_info, |
4416 | "dirty buffer transid mismatch, logical %llu found transid %llu running transid %llu", | |
4417 | buf->start, transid, fs_info->generation); | |
50564b65 | 4418 | } |
f18cc978 | 4419 | set_extent_buffer_dirty(buf); |
eb60ceac CM |
4420 | } |
4421 | ||
2ff7e61e | 4422 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4423 | int flush_delayed) |
16cdcec7 MX |
4424 | { |
4425 | /* | |
4426 | * looks as though older kernels can get into trouble with | |
4427 | * this code, they end up stuck in balance_dirty_pages forever | |
4428 | */ | |
e2d84521 | 4429 | int ret; |
16cdcec7 MX |
4430 | |
4431 | if (current->flags & PF_MEMALLOC) | |
4432 | return; | |
4433 | ||
b53d3f5d | 4434 | if (flush_delayed) |
2ff7e61e | 4435 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4436 | |
d814a491 EL |
4437 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4438 | BTRFS_DIRTY_METADATA_THRESH, | |
4439 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4440 | if (ret > 0) { |
0b246afa | 4441 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4442 | } |
16cdcec7 MX |
4443 | } |
4444 | ||
2ff7e61e | 4445 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4446 | { |
2ff7e61e | 4447 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4448 | } |
585ad2c3 | 4449 | |
2ff7e61e | 4450 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4451 | { |
2ff7e61e | 4452 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4453 | } |
6b80053d | 4454 | |
2ff7e61e | 4455 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4456 | { |
fe816d0f NB |
4457 | /* cleanup FS via transaction */ |
4458 | btrfs_cleanup_transaction(fs_info); | |
4459 | ||
0b246afa | 4460 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4461 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4462 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4463 | |
0b246afa JM |
4464 | down_write(&fs_info->cleanup_work_sem); |
4465 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4466 | } |
4467 | ||
ef67963d JB |
4468 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4469 | { | |
fc7cbcd4 DS |
4470 | struct btrfs_root *gang[8]; |
4471 | u64 root_objectid = 0; | |
4472 | int ret; | |
4473 | ||
4474 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4475 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4476 | (void **)gang, root_objectid, | |
4477 | ARRAY_SIZE(gang))) != 0) { | |
4478 | int i; | |
4479 | ||
4480 | for (i = 0; i < ret; i++) | |
4481 | gang[i] = btrfs_grab_root(gang[i]); | |
4482 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4483 | ||
4484 | for (i = 0; i < ret; i++) { | |
4485 | if (!gang[i]) | |
ef67963d | 4486 | continue; |
fc7cbcd4 DS |
4487 | root_objectid = gang[i]->root_key.objectid; |
4488 | btrfs_free_log(NULL, gang[i]); | |
4489 | btrfs_put_root(gang[i]); | |
ef67963d | 4490 | } |
fc7cbcd4 DS |
4491 | root_objectid++; |
4492 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4493 | } |
fc7cbcd4 | 4494 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4495 | btrfs_free_log_root_tree(NULL, fs_info); |
4496 | } | |
4497 | ||
143bede5 | 4498 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4499 | { |
acce952b | 4500 | struct btrfs_ordered_extent *ordered; |
acce952b | 4501 | |
199c2a9c | 4502 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4503 | /* |
4504 | * This will just short circuit the ordered completion stuff which will | |
4505 | * make sure the ordered extent gets properly cleaned up. | |
4506 | */ | |
199c2a9c | 4507 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4508 | root_extent_list) |
4509 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4510 | spin_unlock(&root->ordered_extent_lock); |
4511 | } | |
4512 | ||
4513 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4514 | { | |
4515 | struct btrfs_root *root; | |
84af994b | 4516 | LIST_HEAD(splice); |
199c2a9c MX |
4517 | |
4518 | spin_lock(&fs_info->ordered_root_lock); | |
4519 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4520 | while (!list_empty(&splice)) { | |
4521 | root = list_first_entry(&splice, struct btrfs_root, | |
4522 | ordered_root); | |
1de2cfde JB |
4523 | list_move_tail(&root->ordered_root, |
4524 | &fs_info->ordered_roots); | |
199c2a9c | 4525 | |
2a85d9ca | 4526 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4527 | btrfs_destroy_ordered_extents(root); |
4528 | ||
2a85d9ca LB |
4529 | cond_resched(); |
4530 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4531 | } |
4532 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4533 | |
4534 | /* | |
4535 | * We need this here because if we've been flipped read-only we won't | |
4536 | * get sync() from the umount, so we need to make sure any ordered | |
4537 | * extents that haven't had their dirty pages IO start writeout yet | |
4538 | * actually get run and error out properly. | |
4539 | */ | |
4540 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4541 | } |
4542 | ||
99f09ce3 FM |
4543 | static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4544 | struct btrfs_fs_info *fs_info) | |
acce952b | 4545 | { |
4546 | struct rb_node *node; | |
4547 | struct btrfs_delayed_ref_root *delayed_refs; | |
4548 | struct btrfs_delayed_ref_node *ref; | |
acce952b | 4549 | |
4550 | delayed_refs = &trans->delayed_refs; | |
4551 | ||
4552 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4553 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4554 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4555 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
99f09ce3 | 4556 | return; |
acce952b | 4557 | } |
4558 | ||
5c9d028b | 4559 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4560 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4561 | struct rb_node *n; |
e78417d1 | 4562 | bool pin_bytes = false; |
acce952b | 4563 | |
d7df2c79 JB |
4564 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4565 | href_node); | |
3069bd26 | 4566 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4567 | continue; |
3069bd26 | 4568 | |
d7df2c79 | 4569 | spin_lock(&head->lock); |
e3d03965 | 4570 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4571 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4572 | ref_node); | |
e3d03965 | 4573 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4574 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4575 | if (!list_empty(&ref->add_list)) |
4576 | list_del(&ref->add_list); | |
d7df2c79 JB |
4577 | atomic_dec(&delayed_refs->num_entries); |
4578 | btrfs_put_delayed_ref(ref); | |
adb86dbe | 4579 | btrfs_delayed_refs_rsv_release(fs_info, 1, 0); |
e78417d1 | 4580 | } |
d7df2c79 JB |
4581 | if (head->must_insert_reserved) |
4582 | pin_bytes = true; | |
4583 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4584 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4585 | spin_unlock(&head->lock); |
4586 | spin_unlock(&delayed_refs->lock); | |
4587 | mutex_unlock(&head->mutex); | |
acce952b | 4588 | |
f603bb94 NB |
4589 | if (pin_bytes) { |
4590 | struct btrfs_block_group *cache; | |
4591 | ||
4592 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4593 | BUG_ON(!cache); | |
4594 | ||
4595 | spin_lock(&cache->space_info->lock); | |
4596 | spin_lock(&cache->lock); | |
4597 | cache->pinned += head->num_bytes; | |
4598 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4599 | cache->space_info, head->num_bytes); | |
4600 | cache->reserved -= head->num_bytes; | |
4601 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4602 | spin_unlock(&cache->lock); | |
4603 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4604 | |
4605 | btrfs_put_block_group(cache); | |
4606 | ||
4607 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4608 | head->bytenr + head->num_bytes - 1); | |
4609 | } | |
31890da0 | 4610 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4611 | btrfs_put_delayed_ref_head(head); |
acce952b | 4612 | cond_resched(); |
4613 | spin_lock(&delayed_refs->lock); | |
4614 | } | |
81f7eb00 | 4615 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4616 | |
4617 | spin_unlock(&delayed_refs->lock); | |
acce952b | 4618 | } |
4619 | ||
143bede5 | 4620 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4621 | { |
4622 | struct btrfs_inode *btrfs_inode; | |
84af994b | 4623 | LIST_HEAD(splice); |
acce952b | 4624 | |
eb73c1b7 MX |
4625 | spin_lock(&root->delalloc_lock); |
4626 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4627 | |
4628 | while (!list_empty(&splice)) { | |
fe816d0f | 4629 | struct inode *inode = NULL; |
eb73c1b7 MX |
4630 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4631 | delalloc_inodes); | |
fe816d0f | 4632 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4633 | spin_unlock(&root->delalloc_lock); |
acce952b | 4634 | |
fe816d0f NB |
4635 | /* |
4636 | * Make sure we get a live inode and that it'll not disappear | |
4637 | * meanwhile. | |
4638 | */ | |
4639 | inode = igrab(&btrfs_inode->vfs_inode); | |
4640 | if (inode) { | |
597441b3 JB |
4641 | unsigned int nofs_flag; |
4642 | ||
4643 | nofs_flag = memalloc_nofs_save(); | |
fe816d0f | 4644 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 | 4645 | memalloc_nofs_restore(nofs_flag); |
fe816d0f NB |
4646 | iput(inode); |
4647 | } | |
eb73c1b7 | 4648 | spin_lock(&root->delalloc_lock); |
acce952b | 4649 | } |
eb73c1b7 MX |
4650 | spin_unlock(&root->delalloc_lock); |
4651 | } | |
4652 | ||
4653 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4654 | { | |
4655 | struct btrfs_root *root; | |
84af994b | 4656 | LIST_HEAD(splice); |
eb73c1b7 MX |
4657 | |
4658 | spin_lock(&fs_info->delalloc_root_lock); | |
4659 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4660 | while (!list_empty(&splice)) { | |
4661 | root = list_first_entry(&splice, struct btrfs_root, | |
4662 | delalloc_root); | |
00246528 | 4663 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4664 | BUG_ON(!root); |
4665 | spin_unlock(&fs_info->delalloc_root_lock); | |
4666 | ||
4667 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4668 | btrfs_put_root(root); |
eb73c1b7 MX |
4669 | |
4670 | spin_lock(&fs_info->delalloc_root_lock); | |
4671 | } | |
4672 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4673 | } |
4674 | ||
aec5716c FM |
4675 | static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
4676 | struct extent_io_tree *dirty_pages, | |
4677 | int mark) | |
acce952b | 4678 | { |
acce952b | 4679 | struct extent_buffer *eb; |
4680 | u64 start = 0; | |
4681 | u64 end; | |
acce952b | 4682 | |
e5860f82 FM |
4683 | while (find_first_extent_bit(dirty_pages, start, &start, &end, |
4684 | mark, NULL)) { | |
91166212 | 4685 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4686 | while (start <= end) { |
0b246afa JM |
4687 | eb = find_extent_buffer(fs_info, start); |
4688 | start += fs_info->nodesize; | |
fd8b2b61 | 4689 | if (!eb) |
acce952b | 4690 | continue; |
c4e54a65 JB |
4691 | |
4692 | btrfs_tree_lock(eb); | |
fd8b2b61 | 4693 | wait_on_extent_buffer_writeback(eb); |
190a8339 | 4694 | btrfs_clear_buffer_dirty(NULL, eb); |
c4e54a65 | 4695 | btrfs_tree_unlock(eb); |
acce952b | 4696 | |
fd8b2b61 | 4697 | free_extent_buffer_stale(eb); |
acce952b | 4698 | } |
4699 | } | |
acce952b | 4700 | } |
4701 | ||
46d81ebd FM |
4702 | static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
4703 | struct extent_io_tree *unpin) | |
acce952b | 4704 | { |
acce952b | 4705 | u64 start; |
4706 | u64 end; | |
acce952b | 4707 | |
acce952b | 4708 | while (1) { |
0e6ec385 FM |
4709 | struct extent_state *cached_state = NULL; |
4710 | ||
fcd5e742 LF |
4711 | /* |
4712 | * The btrfs_finish_extent_commit() may get the same range as | |
4713 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4714 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4715 | * the same extent range. | |
4716 | */ | |
4717 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
e5860f82 FM |
4718 | if (!find_first_extent_bit(unpin, 0, &start, &end, |
4719 | EXTENT_DIRTY, &cached_state)) { | |
fcd5e742 | 4720 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4721 | break; |
fcd5e742 | 4722 | } |
acce952b | 4723 | |
0e6ec385 FM |
4724 | clear_extent_dirty(unpin, start, end, &cached_state); |
4725 | free_extent_state(cached_state); | |
2ff7e61e | 4726 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4727 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4728 | cond_resched(); |
4729 | } | |
acce952b | 4730 | } |
4731 | ||
32da5386 | 4732 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4733 | { |
4734 | struct inode *inode; | |
4735 | ||
4736 | inode = cache->io_ctl.inode; | |
4737 | if (inode) { | |
597441b3 JB |
4738 | unsigned int nofs_flag; |
4739 | ||
4740 | nofs_flag = memalloc_nofs_save(); | |
c79a1751 | 4741 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 JB |
4742 | memalloc_nofs_restore(nofs_flag); |
4743 | ||
c79a1751 LB |
4744 | BTRFS_I(inode)->generation = 0; |
4745 | cache->io_ctl.inode = NULL; | |
4746 | iput(inode); | |
4747 | } | |
bbc37d6e | 4748 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4749 | btrfs_put_block_group(cache); |
4750 | } | |
4751 | ||
4752 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4753 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4754 | { |
32da5386 | 4755 | struct btrfs_block_group *cache; |
c79a1751 LB |
4756 | |
4757 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4758 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4759 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4760 | struct btrfs_block_group, |
c79a1751 | 4761 | dirty_list); |
c79a1751 LB |
4762 | |
4763 | if (!list_empty(&cache->io_list)) { | |
4764 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4765 | list_del_init(&cache->io_list); | |
4766 | btrfs_cleanup_bg_io(cache); | |
4767 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4768 | } | |
4769 | ||
4770 | list_del_init(&cache->dirty_list); | |
4771 | spin_lock(&cache->lock); | |
4772 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4773 | spin_unlock(&cache->lock); | |
4774 | ||
4775 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4776 | btrfs_put_block_group(cache); | |
f66e0209 | 4777 | btrfs_dec_delayed_refs_rsv_bg_updates(fs_info); |
c79a1751 LB |
4778 | spin_lock(&cur_trans->dirty_bgs_lock); |
4779 | } | |
4780 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4781 | ||
45ae2c18 NB |
4782 | /* |
4783 | * Refer to the definition of io_bgs member for details why it's safe | |
4784 | * to use it without any locking | |
4785 | */ | |
c79a1751 LB |
4786 | while (!list_empty(&cur_trans->io_bgs)) { |
4787 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4788 | struct btrfs_block_group, |
c79a1751 | 4789 | io_list); |
c79a1751 LB |
4790 | |
4791 | list_del_init(&cache->io_list); | |
4792 | spin_lock(&cache->lock); | |
4793 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4794 | spin_unlock(&cache->lock); | |
4795 | btrfs_cleanup_bg_io(cache); | |
4796 | } | |
4797 | } | |
4798 | ||
b321a52c BB |
4799 | static void btrfs_free_all_qgroup_pertrans(struct btrfs_fs_info *fs_info) |
4800 | { | |
4801 | struct btrfs_root *gang[8]; | |
4802 | int i; | |
4803 | int ret; | |
4804 | ||
4805 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4806 | while (1) { | |
4807 | ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, | |
4808 | (void **)gang, 0, | |
4809 | ARRAY_SIZE(gang), | |
4810 | BTRFS_ROOT_TRANS_TAG); | |
4811 | if (ret == 0) | |
4812 | break; | |
4813 | for (i = 0; i < ret; i++) { | |
4814 | struct btrfs_root *root = gang[i]; | |
4815 | ||
4816 | btrfs_qgroup_free_meta_all_pertrans(root); | |
4817 | radix_tree_tag_clear(&fs_info->fs_roots_radix, | |
4818 | (unsigned long)root->root_key.objectid, | |
4819 | BTRFS_ROOT_TRANS_TAG); | |
4820 | } | |
4821 | } | |
4822 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4823 | } | |
4824 | ||
49b25e05 | 4825 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4826 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4827 | { |
bbbf7243 NB |
4828 | struct btrfs_device *dev, *tmp; |
4829 | ||
2ff7e61e | 4830 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4831 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4832 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4833 | ||
bbbf7243 NB |
4834 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4835 | post_commit_list) { | |
4836 | list_del_init(&dev->post_commit_list); | |
4837 | } | |
4838 | ||
2ff7e61e | 4839 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4840 | |
4a9d8bde | 4841 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4842 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4843 | |
4a9d8bde | 4844 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4845 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4846 | |
ccdf9b30 | 4847 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4848 | |
2ff7e61e | 4849 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4850 | EXTENT_DIRTY); |
fe119a6e | 4851 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4852 | |
b321a52c BB |
4853 | btrfs_free_all_qgroup_pertrans(fs_info); |
4854 | ||
4a9d8bde MX |
4855 | cur_trans->state =TRANS_STATE_COMPLETED; |
4856 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4857 | } |
4858 | ||
2ff7e61e | 4859 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4860 | { |
4861 | struct btrfs_transaction *t; | |
acce952b | 4862 | |
0b246afa | 4863 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4864 | |
0b246afa JM |
4865 | spin_lock(&fs_info->trans_lock); |
4866 | while (!list_empty(&fs_info->trans_list)) { | |
4867 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 | 4868 | struct btrfs_transaction, list); |
77d20c68 | 4869 | if (t->state >= TRANS_STATE_COMMIT_PREP) { |
9b64f57d | 4870 | refcount_inc(&t->use_count); |
0b246afa | 4871 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4872 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4873 | btrfs_put_transaction(t); |
0b246afa | 4874 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4875 | continue; |
4876 | } | |
0b246afa | 4877 | if (t == fs_info->running_transaction) { |
724e2315 | 4878 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4879 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4880 | /* |
4881 | * We wait for 0 num_writers since we don't hold a trans | |
4882 | * handle open currently for this transaction. | |
4883 | */ | |
4884 | wait_event(t->writer_wait, | |
4885 | atomic_read(&t->num_writers) == 0); | |
4886 | } else { | |
0b246afa | 4887 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4888 | } |
2ff7e61e | 4889 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4890 | |
0b246afa JM |
4891 | spin_lock(&fs_info->trans_lock); |
4892 | if (t == fs_info->running_transaction) | |
4893 | fs_info->running_transaction = NULL; | |
acce952b | 4894 | list_del_init(&t->list); |
0b246afa | 4895 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4896 | |
724e2315 | 4897 | btrfs_put_transaction(t); |
2e4e97ab | 4898 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 4899 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4900 | } |
0b246afa JM |
4901 | spin_unlock(&fs_info->trans_lock); |
4902 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4903 | btrfs_destroy_delayed_inodes(fs_info); |
4904 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4905 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4906 | btrfs_drop_all_logs(fs_info); |
0b246afa | 4907 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4908 | |
4909 | return 0; | |
4910 | } | |
ec7d6dfd | 4911 | |
453e4873 | 4912 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
4913 | { |
4914 | struct btrfs_path *path; | |
4915 | int ret; | |
4916 | struct extent_buffer *l; | |
4917 | struct btrfs_key search_key; | |
4918 | struct btrfs_key found_key; | |
4919 | int slot; | |
4920 | ||
4921 | path = btrfs_alloc_path(); | |
4922 | if (!path) | |
4923 | return -ENOMEM; | |
4924 | ||
4925 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
4926 | search_key.type = -1; | |
4927 | search_key.offset = (u64)-1; | |
4928 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
4929 | if (ret < 0) | |
4930 | goto error; | |
4931 | BUG_ON(ret == 0); /* Corruption */ | |
4932 | if (path->slots[0] > 0) { | |
4933 | slot = path->slots[0] - 1; | |
4934 | l = path->nodes[0]; | |
4935 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
4936 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
4937 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 4938 | } else { |
23125104 | 4939 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
4940 | } |
4941 | ret = 0; | |
4942 | error: | |
4943 | btrfs_free_path(path); | |
4944 | return ret; | |
4945 | } | |
4946 | ||
543068a2 | 4947 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
4948 | { |
4949 | int ret; | |
4950 | mutex_lock(&root->objectid_mutex); | |
4951 | ||
6b8fad57 | 4952 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
4953 | btrfs_warn(root->fs_info, |
4954 | "the objectid of root %llu reaches its highest value", | |
4955 | root->root_key.objectid); | |
4956 | ret = -ENOSPC; | |
4957 | goto out; | |
4958 | } | |
4959 | ||
23125104 | 4960 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
4961 | ret = 0; |
4962 | out: | |
4963 | mutex_unlock(&root->objectid_mutex); | |
4964 | return ret; | |
4965 | } |