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" |
8dabb742 | 32 | #include "dev-replace.h" |
53b381b3 | 33 | #include "raid56.h" |
5ac1d209 | 34 | #include "sysfs.h" |
fcebe456 | 35 | #include "qgroup.h" |
ebb8765b | 36 | #include "compression.h" |
557ea5dd | 37 | #include "tree-checker.h" |
fd708b81 | 38 | #include "ref-verify.h" |
aac0023c | 39 | #include "block-group.h" |
b0643e59 | 40 | #include "discard.h" |
f603bb94 | 41 | #include "space-info.h" |
b70f5097 | 42 | #include "zoned.h" |
139e8cd3 | 43 | #include "subpage.h" |
c7f13d42 | 44 | #include "fs.h" |
07e81dc9 | 45 | #include "accessors.h" |
a0231804 | 46 | #include "extent-tree.h" |
45c40c8f | 47 | #include "root-tree.h" |
59b818e0 | 48 | #include "defrag.h" |
c7a03b52 | 49 | #include "uuid-tree.h" |
67707479 | 50 | #include "relocation.h" |
2fc6822c | 51 | #include "scrub.h" |
c03b2207 | 52 | #include "super.h" |
eb60ceac | 53 | |
319e4d06 QW |
54 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
55 | BTRFS_HEADER_FLAG_RELOC |\ | |
56 | BTRFS_SUPER_FLAG_ERROR |\ | |
57 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
58 | BTRFS_SUPER_FLAG_METADUMP |\ |
59 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 60 | |
2ff7e61e JM |
61 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
62 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 63 | |
141386e1 JB |
64 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
65 | { | |
66 | if (fs_info->csum_shash) | |
67 | crypto_free_shash(fs_info->csum_shash); | |
68 | } | |
69 | ||
d352ac68 | 70 | /* |
2996e1f8 | 71 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 72 | */ |
c67b3892 | 73 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 74 | { |
d5178578 | 75 | struct btrfs_fs_info *fs_info = buf->fs_info; |
397239ed QW |
76 | int num_pages; |
77 | u32 first_page_part; | |
d5178578 | 78 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 79 | char *kaddr; |
e9be5a30 | 80 | int i; |
d5178578 JT |
81 | |
82 | shash->tfm = fs_info->csum_shash; | |
83 | crypto_shash_init(shash); | |
397239ed QW |
84 | |
85 | if (buf->addr) { | |
86 | /* Pages are contiguous, handle them as a big one. */ | |
87 | kaddr = buf->addr; | |
88 | first_page_part = fs_info->nodesize; | |
89 | num_pages = 1; | |
90 | } else { | |
082d5bb9 | 91 | kaddr = folio_address(buf->folios[0]); |
397239ed QW |
92 | first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
93 | num_pages = num_extent_pages(buf); | |
94 | } | |
95 | ||
e9be5a30 | 96 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 97 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 98 | |
13df3775 QW |
99 | /* |
100 | * Multiple single-page folios case would reach here. | |
101 | * | |
102 | * nodesize <= PAGE_SIZE and large folio all handled by above | |
103 | * crypto_shash_update() already. | |
104 | */ | |
5ad9b471 | 105 | for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) { |
082d5bb9 | 106 | kaddr = folio_address(buf->folios[i]); |
e9be5a30 | 107 | crypto_shash_update(shash, kaddr, PAGE_SIZE); |
19c00ddc | 108 | } |
71a63551 | 109 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 110 | crypto_shash_final(shash, result); |
19c00ddc CM |
111 | } |
112 | ||
d352ac68 CM |
113 | /* |
114 | * we can't consider a given block up to date unless the transid of the | |
115 | * block matches the transid in the parent node's pointer. This is how we | |
116 | * detect blocks that either didn't get written at all or got written | |
117 | * in the wrong place. | |
118 | */ | |
d87e6575 | 119 | int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) |
1259ab75 | 120 | { |
d87e6575 | 121 | if (!extent_buffer_uptodate(eb)) |
1259ab75 CM |
122 | return 0; |
123 | ||
d87e6575 CH |
124 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) |
125 | return 1; | |
126 | ||
b9fab919 CM |
127 | if (atomic) |
128 | return -EAGAIN; | |
129 | ||
d87e6575 CH |
130 | if (!extent_buffer_uptodate(eb) || |
131 | btrfs_header_generation(eb) != parent_transid) { | |
132 | btrfs_err_rl(eb->fs_info, | |
8f0ed7d4 QW |
133 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
134 | eb->start, eb->read_mirror, | |
29549aec | 135 | parent_transid, btrfs_header_generation(eb)); |
d87e6575 | 136 | clear_extent_buffer_uptodate(eb); |
9e2aff90 | 137 | return 0; |
d87e6575 | 138 | } |
9e2aff90 | 139 | return 1; |
1259ab75 CM |
140 | } |
141 | ||
e7e16f48 JT |
142 | static bool btrfs_supported_super_csum(u16 csum_type) |
143 | { | |
144 | switch (csum_type) { | |
145 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 146 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 147 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 148 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
149 | return true; |
150 | default: | |
151 | return false; | |
152 | } | |
153 | } | |
154 | ||
1104a885 DS |
155 | /* |
156 | * Return 0 if the superblock checksum type matches the checksum value of that | |
157 | * algorithm. Pass the raw disk superblock data. | |
158 | */ | |
3d17adea QW |
159 | int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
160 | const struct btrfs_super_block *disk_sb) | |
1104a885 | 161 | { |
51bce6c9 | 162 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
163 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
164 | ||
165 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 166 | |
51bce6c9 JT |
167 | /* |
168 | * The super_block structure does not span the whole | |
169 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
170 | * filled with zeros and is included in the checksum. | |
171 | */ | |
3d17adea | 172 | crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, |
fd08001f | 173 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); |
1104a885 | 174 | |
55fc29be | 175 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 176 | return 1; |
1104a885 | 177 | |
e7e16f48 | 178 | return 0; |
1104a885 DS |
179 | } |
180 | ||
bacf60e5 CH |
181 | static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, |
182 | int mirror_num) | |
183 | { | |
184 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
96c36eaa | 185 | int num_folios = num_extent_folios(eb); |
bacf60e5 CH |
186 | int ret = 0; |
187 | ||
188 | if (sb_rdonly(fs_info->sb)) | |
189 | return -EROFS; | |
190 | ||
96c36eaa QW |
191 | for (int i = 0; i < num_folios; i++) { |
192 | struct folio *folio = eb->folios[i]; | |
193 | u64 start = max_t(u64, eb->start, folio_pos(folio)); | |
082d5bb9 | 194 | u64 end = min_t(u64, eb->start + eb->len, |
84cda1a6 | 195 | folio_pos(folio) + eb->folio_size); |
917ac778 | 196 | u32 len = end - start; |
bacf60e5 | 197 | |
917ac778 | 198 | ret = btrfs_repair_io_failure(fs_info, 0, start, len, |
96c36eaa QW |
199 | start, folio, offset_in_folio(folio, start), |
200 | mirror_num); | |
bacf60e5 CH |
201 | if (ret) |
202 | break; | |
bacf60e5 CH |
203 | } |
204 | ||
205 | return ret; | |
206 | } | |
207 | ||
d352ac68 CM |
208 | /* |
209 | * helper to read a given tree block, doing retries as required when | |
210 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 | 211 | * |
789d6a3a QW |
212 | * @check: expected tree parentness check, see the comments of the |
213 | * structure for details. | |
d352ac68 | 214 | */ |
6a2e9dc4 | 215 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
789d6a3a | 216 | struct btrfs_tree_parent_check *check) |
f188591e | 217 | { |
5ab12d1f | 218 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ea466794 | 219 | int failed = 0; |
f188591e CM |
220 | int ret; |
221 | int num_copies = 0; | |
222 | int mirror_num = 0; | |
ea466794 | 223 | int failed_mirror = 0; |
f188591e | 224 | |
789d6a3a QW |
225 | ASSERT(check); |
226 | ||
f188591e | 227 | while (1) { |
f8397d69 | 228 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
947a6299 QW |
229 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); |
230 | if (!ret) | |
231 | break; | |
d397712b | 232 | |
0b246afa | 233 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 234 | eb->start, eb->len); |
4235298e | 235 | if (num_copies == 1) |
ea466794 | 236 | break; |
4235298e | 237 | |
5cf1ab56 JB |
238 | if (!failed_mirror) { |
239 | failed = 1; | |
240 | failed_mirror = eb->read_mirror; | |
241 | } | |
242 | ||
f188591e | 243 | mirror_num++; |
ea466794 JB |
244 | if (mirror_num == failed_mirror) |
245 | mirror_num++; | |
246 | ||
4235298e | 247 | if (mirror_num > num_copies) |
ea466794 | 248 | break; |
f188591e | 249 | } |
ea466794 | 250 | |
c0901581 | 251 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 252 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
253 | |
254 | return ret; | |
f188591e | 255 | } |
19c00ddc | 256 | |
31d89399 CH |
257 | /* |
258 | * Checksum a dirty tree block before IO. | |
259 | */ | |
260 | blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) | |
eca0f6f6 | 261 | { |
31d89399 | 262 | struct extent_buffer *eb = bbio->private; |
eca0f6f6 | 263 | struct btrfs_fs_info *fs_info = eb->fs_info; |
31d89399 | 264 | u64 found_start = btrfs_header_bytenr(eb); |
0124855f | 265 | u64 last_trans; |
eca0f6f6 QW |
266 | u8 result[BTRFS_CSUM_SIZE]; |
267 | int ret; | |
268 | ||
31d89399 CH |
269 | /* Btree blocks are always contiguous on disk. */ |
270 | if (WARN_ON_ONCE(bbio->file_offset != eb->start)) | |
271 | return BLK_STS_IOERR; | |
272 | if (WARN_ON_ONCE(bbio->bio.bi_iter.bi_size != eb->len)) | |
273 | return BLK_STS_IOERR; | |
274 | ||
aa6313e6 JT |
275 | /* |
276 | * If an extent_buffer is marked as EXTENT_BUFFER_ZONED_ZEROOUT, don't | |
277 | * checksum it but zero-out its content. This is done to preserve | |
278 | * ordering of I/O without unnecessarily writing out data. | |
279 | */ | |
cbf44cd9 | 280 | if (test_bit(EXTENT_BUFFER_ZONED_ZEROOUT, &eb->bflags)) { |
aa6313e6 | 281 | memzero_extent_buffer(eb, 0, eb->len); |
31d89399 CH |
282 | return BLK_STS_OK; |
283 | } | |
284 | ||
285 | if (WARN_ON_ONCE(found_start != eb->start)) | |
286 | return BLK_STS_IOERR; | |
55151ea9 QW |
287 | if (WARN_ON(!btrfs_folio_test_uptodate(fs_info, eb->folios[0], |
288 | eb->start, eb->len))) | |
31d89399 CH |
289 | return BLK_STS_IOERR; |
290 | ||
eca0f6f6 QW |
291 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, |
292 | offsetof(struct btrfs_header, fsid), | |
293 | BTRFS_FSID_SIZE) == 0); | |
294 | csum_tree_block(eb, result); | |
295 | ||
296 | if (btrfs_header_level(eb)) | |
297 | ret = btrfs_check_node(eb); | |
298 | else | |
85d8a826 | 299 | ret = btrfs_check_leaf(eb); |
eca0f6f6 | 300 | |
3777369f QW |
301 | if (ret < 0) |
302 | goto error; | |
303 | ||
304 | /* | |
305 | * Also check the generation, the eb reached here must be newer than | |
306 | * last committed. Or something seriously wrong happened. | |
307 | */ | |
0124855f FM |
308 | last_trans = btrfs_get_last_trans_committed(fs_info); |
309 | if (unlikely(btrfs_header_generation(eb) <= last_trans)) { | |
3777369f | 310 | ret = -EUCLEAN; |
eca0f6f6 | 311 | btrfs_err(fs_info, |
3777369f | 312 | "block=%llu bad generation, have %llu expect > %llu", |
0124855f | 313 | eb->start, btrfs_header_generation(eb), last_trans); |
3777369f | 314 | goto error; |
eca0f6f6 QW |
315 | } |
316 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
31d89399 | 317 | return BLK_STS_OK; |
3777369f QW |
318 | |
319 | error: | |
320 | btrfs_print_tree(eb, 0); | |
321 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
322 | eb->start); | |
16199ad9 FM |
323 | /* |
324 | * Be noisy if this is an extent buffer from a log tree. We don't abort | |
325 | * a transaction in case there's a bad log tree extent buffer, we just | |
326 | * fallback to a transaction commit. Still we want to know when there is | |
327 | * a bad log tree extent buffer, as that may signal a bug somewhere. | |
328 | */ | |
329 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || | |
330 | btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); | |
deb6216f CH |
331 | return errno_to_blk_status(ret); |
332 | } | |
333 | ||
413fb1bc | 334 | static bool check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 335 | { |
b0c9b3b0 | 336 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 337 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 338 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c | 339 | |
9a8658e3 DS |
340 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
341 | BTRFS_FSID_SIZE); | |
67bc5ad0 | 342 | |
944d3f9f | 343 | /* |
f7361d8c AJ |
344 | * alloc_fsid_devices() copies the fsid into fs_devices::metadata_uuid. |
345 | * This is then overwritten by metadata_uuid if it is present in the | |
346 | * device_list_add(). The same true for a seed device as well. So use of | |
347 | * fs_devices::metadata_uuid is appropriate here. | |
944d3f9f | 348 | */ |
67bc5ad0 | 349 | if (memcmp(fsid, fs_info->fs_devices->metadata_uuid, BTRFS_FSID_SIZE) == 0) |
413fb1bc | 350 | return false; |
944d3f9f NB |
351 | |
352 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
353 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
413fb1bc | 354 | return false; |
944d3f9f | 355 | |
413fb1bc | 356 | return true; |
2b82032c YZ |
357 | } |
358 | ||
77bf40a2 | 359 | /* Do basic extent buffer checks at read time */ |
046b562b CH |
360 | int btrfs_validate_extent_buffer(struct extent_buffer *eb, |
361 | struct btrfs_tree_parent_check *check) | |
ce9adaa5 | 362 | { |
77bf40a2 | 363 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 364 | u64 found_start; |
77bf40a2 QW |
365 | const u32 csum_size = fs_info->csum_size; |
366 | u8 found_level; | |
2996e1f8 | 367 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 368 | const u8 *header_csum; |
77bf40a2 | 369 | int ret = 0; |
ea466794 | 370 | |
947a6299 QW |
371 | ASSERT(check); |
372 | ||
ce9adaa5 | 373 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 374 | if (found_start != eb->start) { |
8f0ed7d4 QW |
375 | btrfs_err_rl(fs_info, |
376 | "bad tree block start, mirror %u want %llu have %llu", | |
377 | eb->read_mirror, eb->start, found_start); | |
f188591e | 378 | ret = -EIO; |
77bf40a2 | 379 | goto out; |
ce9adaa5 | 380 | } |
b0c9b3b0 | 381 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
382 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
383 | eb->start, eb->read_mirror); | |
1259ab75 | 384 | ret = -EIO; |
77bf40a2 | 385 | goto out; |
1259ab75 | 386 | } |
ce9adaa5 | 387 | found_level = btrfs_header_level(eb); |
1c24c3ce | 388 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
389 | btrfs_err(fs_info, |
390 | "bad tree block level, mirror %u level %d on logical %llu", | |
391 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 392 | ret = -EIO; |
77bf40a2 | 393 | goto out; |
1c24c3ce | 394 | } |
ce9adaa5 | 395 | |
c67b3892 | 396 | csum_tree_block(eb, result); |
082d5bb9 | 397 | header_csum = folio_address(eb->folios[0]) + |
8d993618 | 398 | get_eb_offset_in_folio(eb, offsetof(struct btrfs_header, csum)); |
a826d6dc | 399 | |
dfd29eed | 400 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 401 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
402 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
403 | eb->start, eb->read_mirror, | |
dfd29eed | 404 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
405 | CSUM_FMT_VALUE(csum_size, result), |
406 | btrfs_header_level(eb)); | |
2996e1f8 | 407 | ret = -EUCLEAN; |
77bf40a2 | 408 | goto out; |
2996e1f8 JT |
409 | } |
410 | ||
947a6299 | 411 | if (found_level != check->level) { |
77177ed1 QW |
412 | btrfs_err(fs_info, |
413 | "level verify failed on logical %llu mirror %u wanted %u found %u", | |
414 | eb->start, eb->read_mirror, check->level, found_level); | |
947a6299 QW |
415 | ret = -EIO; |
416 | goto out; | |
417 | } | |
418 | if (unlikely(check->transid && | |
419 | btrfs_header_generation(eb) != check->transid)) { | |
420 | btrfs_err_rl(eb->fs_info, | |
421 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", | |
422 | eb->start, eb->read_mirror, check->transid, | |
423 | btrfs_header_generation(eb)); | |
424 | ret = -EIO; | |
425 | goto out; | |
426 | } | |
427 | if (check->has_first_key) { | |
428 | struct btrfs_key *expect_key = &check->first_key; | |
429 | struct btrfs_key found_key; | |
430 | ||
431 | if (found_level) | |
432 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
433 | else | |
434 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
435 | if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { | |
436 | btrfs_err(fs_info, | |
437 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", | |
438 | eb->start, check->transid, | |
439 | expect_key->objectid, | |
440 | expect_key->type, expect_key->offset, | |
441 | found_key.objectid, found_key.type, | |
442 | found_key.offset); | |
443 | ret = -EUCLEAN; | |
444 | goto out; | |
445 | } | |
446 | } | |
447 | if (check->owner_root) { | |
448 | ret = btrfs_check_eb_owner(eb, check->owner_root); | |
449 | if (ret < 0) | |
450 | goto out; | |
451 | } | |
452 | ||
a826d6dc JB |
453 | /* |
454 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
455 | * that we don't try and read the other copies of this block, just | |
456 | * return -EIO. | |
457 | */ | |
85d8a826 | 458 | if (found_level == 0 && btrfs_check_leaf(eb)) { |
a826d6dc JB |
459 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
460 | ret = -EIO; | |
461 | } | |
ce9adaa5 | 462 | |
813fd1dc | 463 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
464 | ret = -EIO; |
465 | ||
aebcc159 | 466 | if (ret) |
75391f0d | 467 | btrfs_err(fs_info, |
8f0ed7d4 QW |
468 | "read time tree block corruption detected on logical %llu mirror %u", |
469 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
470 | out: |
471 | return ret; | |
472 | } | |
473 | ||
3dd1462e | 474 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
475 | static int btree_migrate_folio(struct address_space *mapping, |
476 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
477 | { |
478 | /* | |
479 | * we can't safely write a btree page from here, | |
480 | * we haven't done the locking hook | |
481 | */ | |
8958b551 | 482 | if (folio_test_dirty(src)) |
784b4e29 CM |
483 | return -EAGAIN; |
484 | /* | |
485 | * Buffers may be managed in a filesystem specific way. | |
486 | * We must have no buffers or drop them. | |
487 | */ | |
8958b551 MWO |
488 | if (folio_get_private(src) && |
489 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 490 | return -EAGAIN; |
54184650 | 491 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 492 | } |
8958b551 MWO |
493 | #else |
494 | #define btree_migrate_folio NULL | |
3dd1462e | 495 | #endif |
784b4e29 | 496 | |
0da5468f CM |
497 | static int btree_writepages(struct address_space *mapping, |
498 | struct writeback_control *wbc) | |
499 | { | |
e2d84521 MX |
500 | int ret; |
501 | ||
d8d5f3e1 | 502 | if (wbc->sync_mode == WB_SYNC_NONE) { |
41044b41 | 503 | struct btrfs_fs_info *fs_info; |
448d640b CM |
504 | |
505 | if (wbc->for_kupdate) | |
506 | return 0; | |
507 | ||
41044b41 | 508 | fs_info = inode_to_fs_info(mapping->host); |
b9473439 | 509 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
510 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
511 | BTRFS_DIRTY_METADATA_THRESH, | |
512 | fs_info->dirty_metadata_batch); | |
e2d84521 | 513 | if (ret < 0) |
793955bc | 514 | return 0; |
793955bc | 515 | } |
0b32f4bb | 516 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
517 | } |
518 | ||
f913cff3 | 519 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 520 | { |
f913cff3 MWO |
521 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
522 | return false; | |
0c4e538b | 523 | |
f913cff3 | 524 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
525 | } |
526 | ||
895586eb MWO |
527 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
528 | size_t length) | |
d98237b3 | 529 | { |
d1310b2e | 530 | struct extent_io_tree *tree; |
c8293894 DS |
531 | |
532 | tree = &folio_to_inode(folio)->io_tree; | |
895586eb | 533 | extent_invalidate_folio(tree, folio, offset); |
f913cff3 | 534 | btree_release_folio(folio, GFP_NOFS); |
895586eb | 535 | if (folio_get_private(folio)) { |
b33d2e53 | 536 | btrfs_warn(folio_to_fs_info(folio), |
895586eb MWO |
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 | { | |
41044b41 | 547 | struct btrfs_fs_info *fs_info = inode_to_fs_info(mapping->host); |
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 | { |
fef998d1 | 649 | bool dummy = btrfs_is_testing(fs_info); |
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; |
905a95f3 | 666 | xa_init(&root->delayed_nodes); |
2e608bd1 JB |
667 | |
668 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
669 | |
670 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 671 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
672 | INIT_LIST_HEAD(&root->delalloc_inodes); |
673 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
674 | INIT_LIST_HEAD(&root->ordered_extents); |
675 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 676 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
5d4f98a2 | 677 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 678 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 679 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 680 | spin_lock_init(&root->accounting_lock); |
8287475a | 681 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 682 | mutex_init(&root->objectid_mutex); |
e02119d5 | 683 | mutex_init(&root->log_mutex); |
31f3d255 | 684 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 685 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 686 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
687 | init_waitqueue_head(&root->log_writer_wait); |
688 | init_waitqueue_head(&root->log_commit_wait[0]); | |
689 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
690 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
691 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
692 | atomic_set(&root->log_commit[0], 0); |
693 | atomic_set(&root->log_commit[1], 0); | |
694 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 695 | atomic_set(&root->log_batch, 0); |
0700cea7 | 696 | refcount_set(&root->refs, 1); |
8ecebf4d | 697 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 698 | atomic_set(&root->nr_swapfiles, 0); |
6008859b | 699 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 700 | root->log_transid_committed = -1; |
f9850787 | 701 | btrfs_set_root_last_log_commit(root, 0); |
2e608bd1 | 702 | root->anon_dev = 0; |
e289f03e | 703 | if (!dummy) { |
43eb5f29 | 704 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
35da5a7e | 705 | IO_TREE_ROOT_DIRTY_LOG_PAGES); |
e289f03e | 706 | extent_io_tree_init(fs_info, &root->log_csum_range, |
35da5a7e | 707 | IO_TREE_LOG_CSUM_RANGE); |
e289f03e | 708 | } |
017e5369 | 709 | |
5f3ab90a | 710 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 711 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
712 | #ifdef CONFIG_BTRFS_DEBUG |
713 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 714 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 715 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 716 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 717 | #endif |
3768f368 CM |
718 | } |
719 | ||
74e4d827 | 720 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 721 | u64 objectid, gfp_t flags) |
6f07e42e | 722 | { |
74e4d827 | 723 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 724 | if (root) |
96dfcb46 | 725 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
726 | return root; |
727 | } | |
728 | ||
06ea65a3 JB |
729 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
730 | /* Should only be used by the testing infrastructure */ | |
da17066c | 731 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
732 | { |
733 | struct btrfs_root *root; | |
734 | ||
7c0260ee JM |
735 | if (!fs_info) |
736 | return ERR_PTR(-EINVAL); | |
737 | ||
96dfcb46 | 738 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
739 | if (!root) |
740 | return ERR_PTR(-ENOMEM); | |
da17066c | 741 | |
b9ef22de | 742 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 743 | root->alloc_bytenr = 0; |
06ea65a3 JB |
744 | |
745 | return root; | |
746 | } | |
747 | #endif | |
748 | ||
abed4aaa JB |
749 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
750 | { | |
751 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
752 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
753 | ||
754 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
755 | } | |
756 | ||
757 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
758 | { | |
759 | const struct btrfs_key *key = k; | |
760 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
761 | ||
762 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
763 | } | |
764 | ||
765 | int btrfs_global_root_insert(struct btrfs_root *root) | |
766 | { | |
767 | struct btrfs_fs_info *fs_info = root->fs_info; | |
768 | struct rb_node *tmp; | |
745806fb | 769 | int ret = 0; |
abed4aaa JB |
770 | |
771 | write_lock(&fs_info->global_root_lock); | |
772 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
773 | write_unlock(&fs_info->global_root_lock); | |
abed4aaa | 774 | |
745806fb QW |
775 | if (tmp) { |
776 | ret = -EEXIST; | |
777 | btrfs_warn(fs_info, "global root %llu %llu already exists", | |
e094f480 | 778 | btrfs_root_id(root), root->root_key.offset); |
745806fb QW |
779 | } |
780 | return ret; | |
abed4aaa JB |
781 | } |
782 | ||
783 | void btrfs_global_root_delete(struct btrfs_root *root) | |
784 | { | |
785 | struct btrfs_fs_info *fs_info = root->fs_info; | |
786 | ||
787 | write_lock(&fs_info->global_root_lock); | |
788 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
789 | write_unlock(&fs_info->global_root_lock); | |
790 | } | |
791 | ||
792 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
793 | struct btrfs_key *key) | |
794 | { | |
795 | struct rb_node *node; | |
796 | struct btrfs_root *root = NULL; | |
797 | ||
798 | read_lock(&fs_info->global_root_lock); | |
799 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
800 | if (node) | |
801 | root = container_of(node, struct btrfs_root, rb_node); | |
802 | read_unlock(&fs_info->global_root_lock); | |
803 | ||
804 | return root; | |
805 | } | |
806 | ||
f7238e50 JB |
807 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
808 | { | |
809 | struct btrfs_block_group *block_group; | |
810 | u64 ret; | |
811 | ||
812 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
813 | return 0; | |
814 | ||
815 | if (bytenr) | |
816 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
817 | else | |
818 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
819 | ASSERT(block_group); | |
820 | if (!block_group) | |
821 | return 0; | |
822 | ret = block_group->global_root_id; | |
823 | btrfs_put_block_group(block_group); | |
824 | ||
825 | return ret; | |
826 | } | |
827 | ||
abed4aaa JB |
828 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
829 | { | |
830 | struct btrfs_key key = { | |
831 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
832 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 833 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
834 | }; |
835 | ||
836 | return btrfs_global_root(fs_info, &key); | |
837 | } | |
838 | ||
839 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
840 | { | |
841 | struct btrfs_key key = { | |
842 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
843 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 844 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
845 | }; |
846 | ||
847 | return btrfs_global_root(fs_info, &key); | |
848 | } | |
849 | ||
51129b33 JB |
850 | struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) |
851 | { | |
852 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) | |
853 | return fs_info->block_group_root; | |
854 | return btrfs_extent_root(fs_info, 0); | |
855 | } | |
856 | ||
20897f5c | 857 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
858 | u64 objectid) |
859 | { | |
9b7a2440 | 860 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
861 | struct extent_buffer *leaf; |
862 | struct btrfs_root *tree_root = fs_info->tree_root; | |
863 | struct btrfs_root *root; | |
864 | struct btrfs_key key; | |
b89f6d1f | 865 | unsigned int nofs_flag; |
20897f5c | 866 | int ret = 0; |
20897f5c | 867 | |
b89f6d1f FM |
868 | /* |
869 | * We're holding a transaction handle, so use a NOFS memory allocation | |
870 | * context to avoid deadlock if reclaim happens. | |
871 | */ | |
872 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 873 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 874 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
875 | if (!root) |
876 | return ERR_PTR(-ENOMEM); | |
877 | ||
20897f5c AJ |
878 | root->root_key.objectid = objectid; |
879 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
880 | root->root_key.offset = 0; | |
881 | ||
9631e4cc | 882 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
60ea105a | 883 | 0, BTRFS_NESTING_NORMAL); |
20897f5c AJ |
884 | if (IS_ERR(leaf)) { |
885 | ret = PTR_ERR(leaf); | |
1dd05682 | 886 | leaf = NULL; |
c1b07854 | 887 | goto fail; |
20897f5c AJ |
888 | } |
889 | ||
20897f5c | 890 | root->node = leaf; |
50564b65 | 891 | btrfs_mark_buffer_dirty(trans, leaf); |
20897f5c AJ |
892 | |
893 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 894 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 895 | |
f944d2cb DS |
896 | btrfs_set_root_flags(&root->root_item, 0); |
897 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
898 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
899 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
900 | btrfs_set_root_level(&root->root_item, 0); | |
901 | btrfs_set_root_refs(&root->root_item, 1); | |
902 | btrfs_set_root_used(&root->root_item, leaf->len); | |
903 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
904 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 905 | if (is_fstree(objectid)) |
807fc790 AS |
906 | generate_random_guid(root->root_item.uuid); |
907 | else | |
908 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 909 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 910 | |
8a6a87cd BB |
911 | btrfs_tree_unlock(leaf); |
912 | ||
20897f5c AJ |
913 | key.objectid = objectid; |
914 | key.type = BTRFS_ROOT_ITEM_KEY; | |
915 | key.offset = 0; | |
916 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
917 | if (ret) | |
918 | goto fail; | |
919 | ||
1dd05682 TI |
920 | return root; |
921 | ||
8a6a87cd | 922 | fail: |
00246528 | 923 | btrfs_put_root(root); |
20897f5c | 924 | |
1dd05682 | 925 | return ERR_PTR(ret); |
20897f5c AJ |
926 | } |
927 | ||
7237f183 YZ |
928 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
929 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
930 | { |
931 | struct btrfs_root *root; | |
e02119d5 | 932 | |
96dfcb46 | 933 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 934 | if (!root) |
7237f183 | 935 | return ERR_PTR(-ENOMEM); |
e02119d5 | 936 | |
e02119d5 CM |
937 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
938 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
939 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 940 | |
6ab6ebb7 NA |
941 | return root; |
942 | } | |
943 | ||
944 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
945 | struct btrfs_root *root) | |
946 | { | |
947 | struct extent_buffer *leaf; | |
948 | ||
7237f183 | 949 | /* |
92a7cc42 | 950 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 951 | * |
92a7cc42 QW |
952 | * Log trees are not exposed to user space thus can't be snapshotted, |
953 | * and they go away before a real commit is actually done. | |
954 | * | |
955 | * They do store pointers to file data extents, and those reference | |
956 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 957 | */ |
e02119d5 | 958 | |
4d75f8a9 | 959 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
60ea105a | 960 | NULL, 0, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
961 | if (IS_ERR(leaf)) |
962 | return PTR_ERR(leaf); | |
e02119d5 | 963 | |
7237f183 | 964 | root->node = leaf; |
e02119d5 | 965 | |
50564b65 | 966 | btrfs_mark_buffer_dirty(trans, root->node); |
e02119d5 | 967 | btrfs_tree_unlock(root->node); |
6ab6ebb7 NA |
968 | |
969 | return 0; | |
7237f183 YZ |
970 | } |
971 | ||
972 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
973 | struct btrfs_fs_info *fs_info) | |
974 | { | |
975 | struct btrfs_root *log_root; | |
976 | ||
977 | log_root = alloc_log_tree(trans, fs_info); | |
978 | if (IS_ERR(log_root)) | |
979 | return PTR_ERR(log_root); | |
6ab6ebb7 | 980 | |
3ddebf27 NA |
981 | if (!btrfs_is_zoned(fs_info)) { |
982 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
983 | ||
984 | if (ret) { | |
985 | btrfs_put_root(log_root); | |
986 | return ret; | |
987 | } | |
6ab6ebb7 NA |
988 | } |
989 | ||
7237f183 YZ |
990 | WARN_ON(fs_info->log_root_tree); |
991 | fs_info->log_root_tree = log_root; | |
992 | return 0; | |
993 | } | |
994 | ||
995 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
996 | struct btrfs_root *root) | |
997 | { | |
0b246afa | 998 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
999 | struct btrfs_root *log_root; |
1000 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1001 | int ret; |
7237f183 | 1002 | |
0b246afa | 1003 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1004 | if (IS_ERR(log_root)) |
1005 | return PTR_ERR(log_root); | |
1006 | ||
6ab6ebb7 NA |
1007 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1008 | if (ret) { | |
1009 | btrfs_put_root(log_root); | |
1010 | return ret; | |
1011 | } | |
1012 | ||
7237f183 | 1013 | log_root->last_trans = trans->transid; |
e094f480 | 1014 | log_root->root_key.offset = btrfs_root_id(root); |
7237f183 YZ |
1015 | |
1016 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1017 | btrfs_set_stack_inode_generation(inode_item, 1); |
1018 | btrfs_set_stack_inode_size(inode_item, 3); | |
1019 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1020 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1021 | fs_info->nodesize); |
3cae210f | 1022 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1023 | |
5d4f98a2 | 1024 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1025 | |
1026 | WARN_ON(root->log_root); | |
1027 | root->log_root = log_root; | |
6008859b | 1028 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 1029 | root->log_transid_committed = -1; |
f9850787 | 1030 | btrfs_set_root_last_log_commit(root, 0); |
e02119d5 CM |
1031 | return 0; |
1032 | } | |
1033 | ||
49d11bea JB |
1034 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1035 | struct btrfs_path *path, | |
1036 | struct btrfs_key *key) | |
e02119d5 CM |
1037 | { |
1038 | struct btrfs_root *root; | |
789d6a3a | 1039 | struct btrfs_tree_parent_check check = { 0 }; |
e02119d5 | 1040 | struct btrfs_fs_info *fs_info = tree_root->fs_info; |
84234f3a | 1041 | u64 generation; |
cb517eab | 1042 | int ret; |
581c1760 | 1043 | int level; |
0f7d52f4 | 1044 | |
96dfcb46 | 1045 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1046 | if (!root) |
1047 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1048 | |
cb517eab MX |
1049 | ret = btrfs_find_root(tree_root, key, path, |
1050 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1051 | if (ret) { |
13a8a7c8 YZ |
1052 | if (ret > 0) |
1053 | ret = -ENOENT; | |
49d11bea | 1054 | goto fail; |
0f7d52f4 | 1055 | } |
13a8a7c8 | 1056 | |
84234f3a | 1057 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1058 | level = btrfs_root_level(&root->root_item); |
789d6a3a QW |
1059 | check.level = level; |
1060 | check.transid = generation; | |
1061 | check.owner_root = key->objectid; | |
1062 | root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), | |
1063 | &check); | |
64c043de LB |
1064 | if (IS_ERR(root->node)) { |
1065 | ret = PTR_ERR(root->node); | |
8c38938c | 1066 | root->node = NULL; |
49d11bea | 1067 | goto fail; |
4eb150d6 QW |
1068 | } |
1069 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1070 | ret = -EIO; |
49d11bea | 1071 | goto fail; |
416bc658 | 1072 | } |
88c602ab QW |
1073 | |
1074 | /* | |
1075 | * For real fs, and not log/reloc trees, root owner must | |
1076 | * match its root node owner | |
1077 | */ | |
fef998d1 | 1078 | if (!btrfs_is_testing(fs_info) && |
e094f480 JB |
1079 | btrfs_root_id(root) != BTRFS_TREE_LOG_OBJECTID && |
1080 | btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID && | |
1081 | btrfs_root_id(root) != btrfs_header_owner(root->node)) { | |
88c602ab QW |
1082 | btrfs_crit(fs_info, |
1083 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
e094f480 | 1084 | btrfs_root_id(root), root->node->start, |
88c602ab | 1085 | btrfs_header_owner(root->node), |
e094f480 | 1086 | btrfs_root_id(root)); |
88c602ab QW |
1087 | ret = -EUCLEAN; |
1088 | goto fail; | |
1089 | } | |
5d4f98a2 | 1090 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1091 | return root; |
49d11bea | 1092 | fail: |
00246528 | 1093 | btrfs_put_root(root); |
49d11bea JB |
1094 | return ERR_PTR(ret); |
1095 | } | |
1096 | ||
1097 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1098 | struct btrfs_key *key) | |
1099 | { | |
1100 | struct btrfs_root *root; | |
1101 | struct btrfs_path *path; | |
1102 | ||
1103 | path = btrfs_alloc_path(); | |
1104 | if (!path) | |
1105 | return ERR_PTR(-ENOMEM); | |
1106 | root = read_tree_root_path(tree_root, path, key); | |
1107 | btrfs_free_path(path); | |
1108 | ||
1109 | return root; | |
cb517eab MX |
1110 | } |
1111 | ||
2dfb1e43 QW |
1112 | /* |
1113 | * Initialize subvolume root in-memory structure | |
1114 | * | |
1115 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1116 | */ | |
1117 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1118 | { |
1119 | int ret; | |
1120 | ||
0b548539 | 1121 | btrfs_drew_lock_init(&root->snapshot_lock); |
8257b2dc | 1122 | |
e094f480 | 1123 | if (btrfs_root_id(root) != BTRFS_TREE_LOG_OBJECTID && |
6ebcd021 | 1124 | !btrfs_is_data_reloc_root(root) && |
e094f480 | 1125 | is_fstree(btrfs_root_id(root))) { |
92a7cc42 | 1126 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1127 | btrfs_check_and_init_root_item(&root->root_item); |
1128 | } | |
1129 | ||
851fd730 QW |
1130 | /* |
1131 | * Don't assign anonymous block device to roots that are not exposed to | |
1132 | * userspace, the id pool is limited to 1M | |
1133 | */ | |
e094f480 | 1134 | if (is_fstree(btrfs_root_id(root)) && |
851fd730 | 1135 | btrfs_root_refs(&root->root_item) > 0) { |
2dfb1e43 QW |
1136 | if (!anon_dev) { |
1137 | ret = get_anon_bdev(&root->anon_dev); | |
1138 | if (ret) | |
1139 | goto fail; | |
1140 | } else { | |
1141 | root->anon_dev = anon_dev; | |
1142 | } | |
851fd730 | 1143 | } |
f32e48e9 CR |
1144 | |
1145 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1146 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1147 | if (ret) { |
1148 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1149 | goto fail; |
f32e48e9 CR |
1150 | } |
1151 | ||
6b8fad57 | 1152 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1153 | |
1154 | mutex_unlock(&root->objectid_mutex); | |
1155 | ||
cb517eab MX |
1156 | return 0; |
1157 | fail: | |
84db5ccf | 1158 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1159 | return ret; |
1160 | } | |
1161 | ||
a98db0f3 JB |
1162 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1163 | u64 root_id) | |
cb517eab MX |
1164 | { |
1165 | struct btrfs_root *root; | |
1166 | ||
fc7cbcd4 DS |
1167 | spin_lock(&fs_info->fs_roots_radix_lock); |
1168 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1169 | (unsigned long)root_id); | |
25ac047c | 1170 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1171 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1172 | return root; |
1173 | } | |
1174 | ||
49d11bea JB |
1175 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1176 | u64 objectid) | |
1177 | { | |
abed4aaa JB |
1178 | struct btrfs_key key = { |
1179 | .objectid = objectid, | |
1180 | .type = BTRFS_ROOT_ITEM_KEY, | |
1181 | .offset = 0, | |
1182 | }; | |
1183 | ||
e91909aa CH |
1184 | switch (objectid) { |
1185 | case BTRFS_ROOT_TREE_OBJECTID: | |
49d11bea | 1186 | return btrfs_grab_root(fs_info->tree_root); |
e91909aa | 1187 | case BTRFS_EXTENT_TREE_OBJECTID: |
abed4aaa | 1188 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1189 | case BTRFS_CHUNK_TREE_OBJECTID: |
49d11bea | 1190 | return btrfs_grab_root(fs_info->chunk_root); |
e91909aa | 1191 | case BTRFS_DEV_TREE_OBJECTID: |
49d11bea | 1192 | return btrfs_grab_root(fs_info->dev_root); |
e91909aa | 1193 | case BTRFS_CSUM_TREE_OBJECTID: |
abed4aaa | 1194 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1195 | case BTRFS_QUOTA_TREE_OBJECTID: |
85724171 | 1196 | return btrfs_grab_root(fs_info->quota_root); |
e91909aa | 1197 | case BTRFS_UUID_TREE_OBJECTID: |
85724171 | 1198 | return btrfs_grab_root(fs_info->uuid_root); |
e91909aa | 1199 | case BTRFS_BLOCK_GROUP_TREE_OBJECTID: |
85724171 | 1200 | return btrfs_grab_root(fs_info->block_group_root); |
e91909aa | 1201 | case BTRFS_FREE_SPACE_TREE_OBJECTID: |
85724171 | 1202 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
51502090 JT |
1203 | case BTRFS_RAID_STRIPE_TREE_OBJECTID: |
1204 | return btrfs_grab_root(fs_info->stripe_root); | |
e91909aa CH |
1205 | default: |
1206 | return NULL; | |
1207 | } | |
49d11bea JB |
1208 | } |
1209 | ||
cb517eab MX |
1210 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1211 | struct btrfs_root *root) | |
1212 | { | |
1213 | int ret; | |
1214 | ||
fc7cbcd4 DS |
1215 | ret = radix_tree_preload(GFP_NOFS); |
1216 | if (ret) | |
1217 | return ret; | |
1218 | ||
1219 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1220 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
e094f480 | 1221 | (unsigned long)btrfs_root_id(root), |
fc7cbcd4 | 1222 | root); |
af01d2e5 | 1223 | if (ret == 0) { |
00246528 | 1224 | btrfs_grab_root(root); |
fc7cbcd4 | 1225 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1226 | } |
fc7cbcd4 DS |
1227 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1228 | radix_tree_preload_end(); | |
cb517eab MX |
1229 | |
1230 | return ret; | |
1231 | } | |
1232 | ||
bd647ce3 JB |
1233 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1234 | { | |
1235 | #ifdef CONFIG_BTRFS_DEBUG | |
1236 | struct btrfs_root *root; | |
1237 | ||
1238 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1239 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1240 | ||
bd647ce3 JB |
1241 | root = list_first_entry(&fs_info->allocated_roots, |
1242 | struct btrfs_root, leak_list); | |
457f1864 | 1243 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1244 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 | 1245 | refcount_read(&root->refs)); |
8fd2b12e | 1246 | WARN_ON_ONCE(1); |
bd647ce3 | 1247 | while (refcount_read(&root->refs) > 1) |
00246528 JB |
1248 | btrfs_put_root(root); |
1249 | btrfs_put_root(root); | |
bd647ce3 JB |
1250 | } |
1251 | #endif | |
1252 | } | |
1253 | ||
abed4aaa JB |
1254 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1255 | { | |
1256 | struct btrfs_root *root; | |
1257 | struct rb_node *node; | |
1258 | ||
1259 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1260 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1261 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1262 | btrfs_put_root(root); | |
1263 | } | |
1264 | } | |
1265 | ||
0d4b0463 JB |
1266 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1267 | { | |
f1d97e76 FM |
1268 | struct percpu_counter *em_counter = &fs_info->evictable_extent_maps; |
1269 | ||
141386e1 JB |
1270 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1271 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1272 | percpu_counter_destroy(&fs_info->ordered_bytes); |
f1d97e76 FM |
1273 | if (percpu_counter_initialized(em_counter)) |
1274 | ASSERT(percpu_counter_sum_positive(em_counter) == 0); | |
1275 | percpu_counter_destroy(em_counter); | |
141386e1 JB |
1276 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1277 | btrfs_free_csum_hash(fs_info); | |
1278 | btrfs_free_stripe_hash_table(fs_info); | |
1279 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1280 | kfree(fs_info->balance_ctl); |
1281 | kfree(fs_info->delayed_root); | |
abed4aaa | 1282 | free_global_roots(fs_info); |
00246528 JB |
1283 | btrfs_put_root(fs_info->tree_root); |
1284 | btrfs_put_root(fs_info->chunk_root); | |
1285 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1286 | btrfs_put_root(fs_info->quota_root); |
1287 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1288 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1289 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1290 | btrfs_put_root(fs_info->block_group_root); |
51502090 | 1291 | btrfs_put_root(fs_info->stripe_root); |
bd647ce3 | 1292 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1293 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1294 | kfree(fs_info->super_copy); |
1295 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1296 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1297 | kvfree(fs_info); |
1298 | } | |
1299 | ||
1300 | ||
2dfb1e43 QW |
1301 | /* |
1302 | * Get an in-memory reference of a root structure. | |
1303 | * | |
1304 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1305 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1306 | * found, then read it from disk and add it to cached fs roots. | |
1307 | * | |
1308 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1309 | * | |
1310 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1311 | * same root objectid. | |
1312 | * | |
1313 | * @objectid: root id | |
1314 | * @anon_dev: preallocated anonymous block device number for new roots, | |
e2b54eaf | 1315 | * pass NULL for a new allocation. |
2dfb1e43 QW |
1316 | * @check_ref: whether to check root item references, If true, return -ENOENT |
1317 | * for orphan roots | |
1318 | */ | |
1319 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1320 | u64 objectid, dev_t *anon_dev, |
2dfb1e43 | 1321 | bool check_ref) |
5eda7b5e CM |
1322 | { |
1323 | struct btrfs_root *root; | |
381cf658 | 1324 | struct btrfs_path *path; |
1d4c08e0 | 1325 | struct btrfs_key key; |
5eda7b5e CM |
1326 | int ret; |
1327 | ||
49d11bea JB |
1328 | root = btrfs_get_global_root(fs_info, objectid); |
1329 | if (root) | |
1330 | return root; | |
773e722a QW |
1331 | |
1332 | /* | |
1333 | * If we're called for non-subvolume trees, and above function didn't | |
1334 | * find one, do not try to read it from disk. | |
1335 | * | |
1336 | * This is namely for free-space-tree and quota tree, which can change | |
1337 | * at runtime and should only be grabbed from fs_info. | |
1338 | */ | |
1339 | if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) | |
1340 | return ERR_PTR(-ENOENT); | |
4df27c4d | 1341 | again: |
56e9357a | 1342 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1343 | if (root) { |
e03ee2fe QW |
1344 | /* |
1345 | * Some other caller may have read out the newly inserted | |
1346 | * subvolume already (for things like backref walk etc). Not | |
1347 | * that common but still possible. In that case, we just need | |
1348 | * to free the anon_dev. | |
1349 | */ | |
e2b54eaf FM |
1350 | if (unlikely(anon_dev && *anon_dev)) { |
1351 | free_anon_bdev(*anon_dev); | |
1352 | *anon_dev = 0; | |
e03ee2fe QW |
1353 | } |
1354 | ||
bc44d7c4 | 1355 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1356 | btrfs_put_root(root); |
48475471 | 1357 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1358 | } |
5eda7b5e | 1359 | return root; |
48475471 | 1360 | } |
5eda7b5e | 1361 | |
56e9357a DS |
1362 | key.objectid = objectid; |
1363 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1364 | key.offset = (u64)-1; | |
1365 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1366 | if (IS_ERR(root)) |
1367 | return root; | |
3394e160 | 1368 | |
c00869f1 | 1369 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1370 | ret = -ENOENT; |
581bb050 | 1371 | goto fail; |
35a30d7c | 1372 | } |
581bb050 | 1373 | |
e2b54eaf | 1374 | ret = btrfs_init_fs_root(root, anon_dev ? *anon_dev : 0); |
ac08aedf CM |
1375 | if (ret) |
1376 | goto fail; | |
3394e160 | 1377 | |
381cf658 DS |
1378 | path = btrfs_alloc_path(); |
1379 | if (!path) { | |
1380 | ret = -ENOMEM; | |
1381 | goto fail; | |
1382 | } | |
1d4c08e0 DS |
1383 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1384 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1385 | key.offset = objectid; |
1d4c08e0 DS |
1386 | |
1387 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1388 | btrfs_free_path(path); |
d68fc57b YZ |
1389 | if (ret < 0) |
1390 | goto fail; | |
1391 | if (ret == 0) | |
27cdeb70 | 1392 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1393 | |
cb517eab | 1394 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1395 | if (ret) { |
168a2f77 JJB |
1396 | if (ret == -EEXIST) { |
1397 | btrfs_put_root(root); | |
4df27c4d | 1398 | goto again; |
168a2f77 | 1399 | } |
4df27c4d | 1400 | goto fail; |
0f7d52f4 | 1401 | } |
edbd8d4e | 1402 | return root; |
4df27c4d | 1403 | fail: |
33fab972 FM |
1404 | /* |
1405 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1406 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1407 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1408 | * and once again by our caller. | |
1409 | */ | |
e2b54eaf | 1410 | if (anon_dev && *anon_dev) |
33fab972 | 1411 | root->anon_dev = 0; |
8c38938c | 1412 | btrfs_put_root(root); |
4df27c4d | 1413 | return ERR_PTR(ret); |
edbd8d4e CM |
1414 | } |
1415 | ||
2dfb1e43 QW |
1416 | /* |
1417 | * Get in-memory reference of a root structure | |
1418 | * | |
1419 | * @objectid: tree objectid | |
1420 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1421 | * one reference | |
1422 | */ | |
1423 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1424 | u64 objectid, bool check_ref) | |
1425 | { | |
e2b54eaf | 1426 | return btrfs_get_root_ref(fs_info, objectid, NULL, check_ref); |
2dfb1e43 QW |
1427 | } |
1428 | ||
1429 | /* | |
1430 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1431 | * the anonymous block device id | |
1432 | * | |
1433 | * @objectid: tree objectid | |
e2b54eaf FM |
1434 | * @anon_dev: if NULL, allocate a new anonymous block device or use the |
1435 | * parameter value if not NULL | |
2dfb1e43 QW |
1436 | */ |
1437 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1438 | u64 objectid, dev_t *anon_dev) |
2dfb1e43 QW |
1439 | { |
1440 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1441 | } | |
1442 | ||
49d11bea | 1443 | /* |
9580503b DS |
1444 | * Return a root for the given objectid. |
1445 | * | |
49d11bea JB |
1446 | * @fs_info: the fs_info |
1447 | * @objectid: the objectid we need to lookup | |
1448 | * | |
1449 | * This is exclusively used for backref walking, and exists specifically because | |
1450 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1451 | * creation time, which means we may have to read the tree_root in order to look | |
1452 | * up a fs root that is not in memory. If the root is not in memory we will | |
1453 | * read the tree root commit root and look up the fs root from there. This is a | |
1454 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1455 | * have the most uptodate information, it'll simply be discarded once the | |
1456 | * backref code is finished using the root. | |
1457 | */ | |
1458 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1459 | struct btrfs_path *path, | |
1460 | u64 objectid) | |
1461 | { | |
1462 | struct btrfs_root *root; | |
1463 | struct btrfs_key key; | |
1464 | ||
1465 | ASSERT(path->search_commit_root && path->skip_locking); | |
1466 | ||
1467 | /* | |
1468 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1469 | * since this is called via the backref walking code we won't be looking | |
1470 | * up a root that doesn't exist, unless there's corruption. So if root | |
1471 | * != NULL just return it. | |
1472 | */ | |
1473 | root = btrfs_get_global_root(fs_info, objectid); | |
1474 | if (root) | |
1475 | return root; | |
1476 | ||
1477 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1478 | if (root) | |
1479 | return root; | |
1480 | ||
1481 | key.objectid = objectid; | |
1482 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1483 | key.offset = (u64)-1; | |
1484 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1485 | btrfs_release_path(path); | |
1486 | ||
1487 | return root; | |
1488 | } | |
1489 | ||
a74a4b97 CM |
1490 | static int cleaner_kthread(void *arg) |
1491 | { | |
0d031dc4 | 1492 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1493 | int again; |
a74a4b97 | 1494 | |
d6fd0ae2 | 1495 | while (1) { |
d0278245 | 1496 | again = 0; |
a74a4b97 | 1497 | |
fd340d0f JB |
1498 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1499 | ||
d0278245 | 1500 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1501 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1502 | goto sleep; |
1503 | ||
90c711ab ZB |
1504 | /* |
1505 | * Do not do anything if we might cause open_ctree() to block | |
1506 | * before we have finished mounting the filesystem. | |
1507 | */ | |
0b246afa | 1508 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1509 | goto sleep; |
1510 | ||
0b246afa | 1511 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1512 | goto sleep; |
1513 | ||
dc7f370c MX |
1514 | /* |
1515 | * Avoid the problem that we change the status of the fs | |
1516 | * during the above check and trylock. | |
1517 | */ | |
2ff7e61e | 1518 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1519 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1520 | goto sleep; |
76dda93c | 1521 | } |
a74a4b97 | 1522 | |
b7625f46 QW |
1523 | if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) |
1524 | btrfs_sysfs_feature_update(fs_info); | |
1525 | ||
2ff7e61e | 1526 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1527 | |
33c44184 | 1528 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1529 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1530 | |
1531 | /* | |
05323cd1 MX |
1532 | * The defragger has dealt with the R/O remount and umount, |
1533 | * needn't do anything special here. | |
d0278245 | 1534 | */ |
0b246afa | 1535 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1536 | |
1537 | /* | |
f3372065 | 1538 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1539 | * with relocation (btrfs_relocate_chunk) and relocation |
1540 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1541 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1542 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1543 | * unused block groups. | |
1544 | */ | |
0b246afa | 1545 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1546 | |
1547 | /* | |
1548 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1549 | * all unused block_groups. This possibly gives us some more free | |
1550 | * space. | |
1551 | */ | |
1552 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1553 | sleep: |
a0a1db70 | 1554 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1555 | if (kthread_should_park()) |
1556 | kthread_parkme(); | |
1557 | if (kthread_should_stop()) | |
1558 | return 0; | |
838fe188 | 1559 | if (!again) { |
a74a4b97 | 1560 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1561 | schedule(); |
a74a4b97 CM |
1562 | __set_current_state(TASK_RUNNING); |
1563 | } | |
da288d28 | 1564 | } |
a74a4b97 CM |
1565 | } |
1566 | ||
1567 | static int transaction_kthread(void *arg) | |
1568 | { | |
1569 | struct btrfs_root *root = arg; | |
0b246afa | 1570 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1571 | struct btrfs_trans_handle *trans; |
1572 | struct btrfs_transaction *cur; | |
8929ecfa | 1573 | u64 transid; |
643900be | 1574 | time64_t delta; |
a74a4b97 | 1575 | unsigned long delay; |
914b2007 | 1576 | bool cannot_commit; |
a74a4b97 CM |
1577 | |
1578 | do { | |
914b2007 | 1579 | cannot_commit = false; |
ba1bc00f | 1580 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1581 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1582 | |
0b246afa JM |
1583 | spin_lock(&fs_info->trans_lock); |
1584 | cur = fs_info->running_transaction; | |
a74a4b97 | 1585 | if (!cur) { |
0b246afa | 1586 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1587 | goto sleep; |
1588 | } | |
31153d81 | 1589 | |
643900be | 1590 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 | 1591 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
77d20c68 | 1592 | cur->state < TRANS_STATE_COMMIT_PREP && |
643900be | 1593 | delta < fs_info->commit_interval) { |
0b246afa | 1594 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1595 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1596 | delay = min(delay, | |
1597 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1598 | goto sleep; |
1599 | } | |
8929ecfa | 1600 | transid = cur->transid; |
0b246afa | 1601 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1602 | |
79787eaa | 1603 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1604 | trans = btrfs_attach_transaction(root); |
914b2007 | 1605 | if (IS_ERR(trans)) { |
354aa0fb MX |
1606 | if (PTR_ERR(trans) != -ENOENT) |
1607 | cannot_commit = true; | |
79787eaa | 1608 | goto sleep; |
914b2007 | 1609 | } |
8929ecfa | 1610 | if (transid == trans->transid) { |
3a45bb20 | 1611 | btrfs_commit_transaction(trans); |
8929ecfa | 1612 | } else { |
3a45bb20 | 1613 | btrfs_end_transaction(trans); |
8929ecfa | 1614 | } |
a74a4b97 | 1615 | sleep: |
0b246afa JM |
1616 | wake_up_process(fs_info->cleaner_kthread); |
1617 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1618 | |
84961539 | 1619 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1620 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1621 | if (!kthread_should_stop() && |
0b246afa | 1622 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1623 | cannot_commit)) |
bc5511d0 | 1624 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1625 | } while (!kthread_should_stop()); |
1626 | return 0; | |
1627 | } | |
1628 | ||
af31f5e5 | 1629 | /* |
01f0f9da NB |
1630 | * This will find the highest generation in the array of root backups. The |
1631 | * index of the highest array is returned, or -EINVAL if we can't find | |
1632 | * anything. | |
af31f5e5 CM |
1633 | * |
1634 | * We check to make sure the array is valid by comparing the | |
1635 | * generation of the latest root in the array with the generation | |
1636 | * in the super block. If they don't match we pitch it. | |
1637 | */ | |
01f0f9da | 1638 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1639 | { |
01f0f9da | 1640 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1641 | u64 cur; |
af31f5e5 CM |
1642 | struct btrfs_root_backup *root_backup; |
1643 | int i; | |
1644 | ||
1645 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1646 | root_backup = info->super_copy->super_roots + i; | |
1647 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1648 | if (cur == newest_gen) | |
01f0f9da | 1649 | return i; |
af31f5e5 CM |
1650 | } |
1651 | ||
01f0f9da | 1652 | return -EINVAL; |
af31f5e5 CM |
1653 | } |
1654 | ||
af31f5e5 CM |
1655 | /* |
1656 | * copy all the root pointers into the super backup array. | |
1657 | * this will bump the backup pointer by one when it is | |
1658 | * done | |
1659 | */ | |
1660 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1661 | { | |
6ef108dd | 1662 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1663 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1664 | |
1665 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1666 | ||
1667 | /* | |
1668 | * make sure all of our padding and empty slots get zero filled | |
1669 | * regardless of which ones we use today | |
1670 | */ | |
1671 | memset(root_backup, 0, sizeof(*root_backup)); | |
1672 | ||
1673 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1674 | ||
1675 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1676 | btrfs_set_backup_tree_root_gen(root_backup, | |
1677 | btrfs_header_generation(info->tree_root->node)); | |
1678 | ||
1679 | btrfs_set_backup_tree_root_level(root_backup, | |
1680 | btrfs_header_level(info->tree_root->node)); | |
1681 | ||
1682 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1683 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1684 | btrfs_header_generation(info->chunk_root->node)); | |
1685 | btrfs_set_backup_chunk_root_level(root_backup, | |
1686 | btrfs_header_level(info->chunk_root->node)); | |
1687 | ||
1c56ab99 | 1688 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1689 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1690 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1691 | |
1692 | btrfs_set_backup_extent_root(root_backup, | |
1693 | extent_root->node->start); | |
1694 | btrfs_set_backup_extent_root_gen(root_backup, | |
1695 | btrfs_header_generation(extent_root->node)); | |
1696 | btrfs_set_backup_extent_root_level(root_backup, | |
1697 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1698 | |
1699 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1700 | btrfs_set_backup_csum_root_gen(root_backup, | |
1701 | btrfs_header_generation(csum_root->node)); | |
1702 | btrfs_set_backup_csum_root_level(root_backup, | |
1703 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1704 | } |
af31f5e5 | 1705 | |
7c7e82a7 CM |
1706 | /* |
1707 | * we might commit during log recovery, which happens before we set | |
1708 | * the fs_root. Make sure it is valid before we fill it in. | |
1709 | */ | |
1710 | if (info->fs_root && info->fs_root->node) { | |
1711 | btrfs_set_backup_fs_root(root_backup, | |
1712 | info->fs_root->node->start); | |
1713 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1714 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1715 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1716 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1717 | } |
af31f5e5 CM |
1718 | |
1719 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1720 | btrfs_set_backup_dev_root_gen(root_backup, | |
1721 | btrfs_header_generation(info->dev_root->node)); | |
1722 | btrfs_set_backup_dev_root_level(root_backup, | |
1723 | btrfs_header_level(info->dev_root->node)); | |
1724 | ||
af31f5e5 CM |
1725 | btrfs_set_backup_total_bytes(root_backup, |
1726 | btrfs_super_total_bytes(info->super_copy)); | |
1727 | btrfs_set_backup_bytes_used(root_backup, | |
1728 | btrfs_super_bytes_used(info->super_copy)); | |
1729 | btrfs_set_backup_num_devices(root_backup, | |
1730 | btrfs_super_num_devices(info->super_copy)); | |
1731 | ||
1732 | /* | |
1733 | * if we don't copy this out to the super_copy, it won't get remembered | |
1734 | * for the next commit | |
1735 | */ | |
1736 | memcpy(&info->super_copy->super_roots, | |
1737 | &info->super_for_commit->super_roots, | |
1738 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1739 | } | |
1740 | ||
bd2336b2 | 1741 | /* |
9580503b DS |
1742 | * Reads a backup root based on the passed priority. Prio 0 is the newest, prio |
1743 | * 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
bd2336b2 | 1744 | * |
9580503b DS |
1745 | * @fs_info: filesystem whose backup roots need to be read |
1746 | * @priority: priority of backup root required | |
bd2336b2 NB |
1747 | * |
1748 | * Returns backup root index on success and -EINVAL otherwise. | |
1749 | */ | |
1750 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1751 | { | |
1752 | int backup_index = find_newest_super_backup(fs_info); | |
1753 | struct btrfs_super_block *super = fs_info->super_copy; | |
1754 | struct btrfs_root_backup *root_backup; | |
1755 | ||
1756 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1757 | if (priority == 0) | |
1758 | return backup_index; | |
1759 | ||
1760 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1761 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1762 | } else { | |
1763 | return -EINVAL; | |
1764 | } | |
1765 | ||
1766 | root_backup = super->super_roots + backup_index; | |
1767 | ||
1768 | btrfs_set_super_generation(super, | |
1769 | btrfs_backup_tree_root_gen(root_backup)); | |
1770 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1771 | btrfs_set_super_root_level(super, | |
1772 | btrfs_backup_tree_root_level(root_backup)); | |
1773 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1774 | ||
1775 | /* | |
1776 | * Fixme: the total bytes and num_devices need to match or we should | |
1777 | * need a fsck | |
1778 | */ | |
1779 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1780 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1781 | ||
1782 | return backup_index; | |
1783 | } | |
1784 | ||
7abadb64 LB |
1785 | /* helper to cleanup workers */ |
1786 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1787 | { | |
dc6e3209 | 1788 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 1789 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 1790 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
1791 | if (fs_info->endio_workers) |
1792 | destroy_workqueue(fs_info->endio_workers); | |
385de0ef CH |
1793 | if (fs_info->rmw_workers) |
1794 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
1795 | if (fs_info->compressed_write_workers) |
1796 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
1797 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
1798 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 1799 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 1800 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 1801 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 1802 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
1803 | if (fs_info->discard_ctl.discard_workers) |
1804 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
1805 | /* |
1806 | * Now that all other work queues are destroyed, we can safely destroy | |
1807 | * the queues used for metadata I/O, since tasks from those other work | |
1808 | * queues can do metadata I/O operations. | |
1809 | */ | |
d7b9416f CH |
1810 | if (fs_info->endio_meta_workers) |
1811 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
1812 | } |
1813 | ||
2e9f5954 R |
1814 | static void free_root_extent_buffers(struct btrfs_root *root) |
1815 | { | |
1816 | if (root) { | |
1817 | free_extent_buffer(root->node); | |
1818 | free_extent_buffer(root->commit_root); | |
1819 | root->node = NULL; | |
1820 | root->commit_root = NULL; | |
1821 | } | |
1822 | } | |
1823 | ||
abed4aaa JB |
1824 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
1825 | { | |
1826 | struct btrfs_root *root, *tmp; | |
1827 | ||
1828 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
1829 | &fs_info->global_root_tree, | |
1830 | rb_node) | |
1831 | free_root_extent_buffers(root); | |
1832 | } | |
1833 | ||
af31f5e5 | 1834 | /* helper to cleanup tree roots */ |
4273eaff | 1835 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 1836 | { |
2e9f5954 | 1837 | free_root_extent_buffers(info->tree_root); |
655b09fe | 1838 | |
abed4aaa | 1839 | free_global_root_pointers(info); |
2e9f5954 | 1840 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
1841 | free_root_extent_buffers(info->quota_root); |
1842 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 1843 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 1844 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 1845 | free_root_extent_buffers(info->block_group_root); |
51502090 | 1846 | free_root_extent_buffers(info->stripe_root); |
4273eaff | 1847 | if (free_chunk_root) |
2e9f5954 | 1848 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
1849 | } |
1850 | ||
8c38938c JB |
1851 | void btrfs_put_root(struct btrfs_root *root) |
1852 | { | |
1853 | if (!root) | |
1854 | return; | |
1855 | ||
1856 | if (refcount_dec_and_test(&root->refs)) { | |
1857 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 1858 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
1859 | if (root->anon_dev) |
1860 | free_anon_bdev(root->anon_dev); | |
923eb523 | 1861 | free_root_extent_buffers(root); |
8c38938c | 1862 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 1863 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 1864 | list_del_init(&root->leak_list); |
fc7cbcd4 | 1865 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
1866 | #endif |
1867 | kfree(root); | |
1868 | } | |
1869 | } | |
1870 | ||
faa2dbf0 | 1871 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 1872 | { |
fc7cbcd4 DS |
1873 | int ret; |
1874 | struct btrfs_root *gang[8]; | |
1875 | int i; | |
171f6537 JB |
1876 | |
1877 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
1878 | gang[0] = list_entry(fs_info->dead_roots.next, |
1879 | struct btrfs_root, root_list); | |
1880 | list_del(&gang[0]->root_list); | |
171f6537 | 1881 | |
fc7cbcd4 DS |
1882 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
1883 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
1884 | btrfs_put_root(gang[0]); | |
171f6537 JB |
1885 | } |
1886 | ||
fc7cbcd4 DS |
1887 | while (1) { |
1888 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1889 | (void **)gang, 0, | |
1890 | ARRAY_SIZE(gang)); | |
1891 | if (!ret) | |
1892 | break; | |
1893 | for (i = 0; i < ret; i++) | |
1894 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
1895 | } |
1896 | } | |
af31f5e5 | 1897 | |
638aa7ed ES |
1898 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
1899 | { | |
1900 | mutex_init(&fs_info->scrub_lock); | |
1901 | atomic_set(&fs_info->scrubs_running, 0); | |
1902 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1903 | atomic_set(&fs_info->scrubs_paused, 0); | |
1904 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1905 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 1906 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
1907 | } |
1908 | ||
779a65a4 ES |
1909 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
1910 | { | |
1911 | spin_lock_init(&fs_info->balance_lock); | |
1912 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
1913 | atomic_set(&fs_info->balance_pause_req, 0); |
1914 | atomic_set(&fs_info->balance_cancel_req, 0); | |
1915 | fs_info->balance_ctl = NULL; | |
1916 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 1917 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
1918 | } |
1919 | ||
dcb2137c | 1920 | static int btrfs_init_btree_inode(struct super_block *sb) |
f37938e0 | 1921 | { |
dcb2137c | 1922 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
e256927b JB |
1923 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
1924 | fs_info->tree_root); | |
dcb2137c CH |
1925 | struct inode *inode; |
1926 | ||
1927 | inode = new_inode(sb); | |
1928 | if (!inode) | |
1929 | return -ENOMEM; | |
2ff7e61e JM |
1930 | |
1931 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
1932 | set_nlink(inode, 1); | |
f37938e0 ES |
1933 | /* |
1934 | * we set the i_size on the btree inode to the max possible int. | |
1935 | * the real end of the address space is determined by all of | |
1936 | * the devices in the system | |
1937 | */ | |
2ff7e61e JM |
1938 | inode->i_size = OFFSET_MAX; |
1939 | inode->i_mapping->a_ops = &btree_aops; | |
dcb2137c | 1940 | mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); |
f37938e0 | 1941 | |
2ff7e61e | 1942 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 1943 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
35da5a7e | 1944 | IO_TREE_BTREE_INODE_IO); |
2ff7e61e | 1945 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 1946 | |
5c8fd99f | 1947 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
1948 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
1949 | BTRFS_I(inode)->location.type = 0; | |
1950 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 1951 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 1952 | __insert_inode_hash(inode, hash); |
dcb2137c CH |
1953 | fs_info->btree_inode = inode; |
1954 | ||
1955 | return 0; | |
f37938e0 ES |
1956 | } |
1957 | ||
ad618368 ES |
1958 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
1959 | { | |
ad618368 | 1960 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 1961 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 1962 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
1963 | } |
1964 | ||
f9e92e40 ES |
1965 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
1966 | { | |
1967 | spin_lock_init(&fs_info->qgroup_lock); | |
1968 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
1969 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
1970 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
1971 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 1972 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 1973 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 1974 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
1975 | mutex_init(&fs_info->qgroup_rescan_lock); |
1976 | } | |
1977 | ||
d21deec5 | 1978 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 1979 | { |
f7b885be | 1980 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 1981 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
58e814fc | 1982 | unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; |
2a458198 ES |
1983 | |
1984 | fs_info->workers = | |
a31b4a43 | 1985 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2a458198 ES |
1986 | |
1987 | fs_info->delalloc_workers = | |
cb001095 JM |
1988 | btrfs_alloc_workqueue(fs_info, "delalloc", |
1989 | flags, max_active, 2); | |
2a458198 ES |
1990 | |
1991 | fs_info->flush_workers = | |
cb001095 JM |
1992 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
1993 | flags, max_active, 0); | |
2a458198 ES |
1994 | |
1995 | fs_info->caching_workers = | |
cb001095 | 1996 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 1997 | |
2a458198 | 1998 | fs_info->fixup_workers = |
58e814fc | 1999 | btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); |
2a458198 | 2000 | |
2a458198 | 2001 | fs_info->endio_workers = |
d7b9416f | 2002 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 2003 | fs_info->endio_meta_workers = |
d7b9416f | 2004 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
385de0ef | 2005 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2006 | fs_info->endio_write_workers = |
cb001095 JM |
2007 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2008 | max_active, 2); | |
fed8a72d CH |
2009 | fs_info->compressed_write_workers = |
2010 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2011 | fs_info->endio_freespace_worker = |
cb001095 JM |
2012 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2013 | max_active, 0); | |
2a458198 | 2014 | fs_info->delayed_workers = |
cb001095 JM |
2015 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2016 | max_active, 0); | |
2a458198 | 2017 | fs_info->qgroup_rescan_workers = |
58e814fc TH |
2018 | btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", |
2019 | ordered_flags); | |
b0643e59 | 2020 | fs_info->discard_ctl.discard_workers = |
58e814fc | 2021 | alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); |
2a458198 | 2022 | |
8bfec2e4 | 2023 | if (!(fs_info->workers && |
a31b4a43 | 2024 | fs_info->delalloc_workers && fs_info->flush_workers && |
2a458198 | 2025 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2026 | fs_info->compressed_write_workers && |
1a1a2851 | 2027 | fs_info->endio_write_workers && |
2a458198 | 2028 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
f26c9238 QW |
2029 | fs_info->caching_workers && fs_info->fixup_workers && |
2030 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2031 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2032 | return -ENOMEM; |
2033 | } | |
2034 | ||
2035 | return 0; | |
2036 | } | |
2037 | ||
6d97c6e3 JT |
2038 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2039 | { | |
2040 | struct crypto_shash *csum_shash; | |
b4e967be | 2041 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2042 | |
b4e967be | 2043 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2044 | |
2045 | if (IS_ERR(csum_shash)) { | |
2046 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2047 | csum_driver); |
6d97c6e3 JT |
2048 | return PTR_ERR(csum_shash); |
2049 | } | |
2050 | ||
2051 | fs_info->csum_shash = csum_shash; | |
2052 | ||
68d99ab0 CH |
2053 | /* |
2054 | * Check if the checksum implementation is a fast accelerated one. | |
2055 | * As-is this is a bit of a hack and should be replaced once the csum | |
2056 | * implementations provide that information themselves. | |
2057 | */ | |
2058 | switch (csum_type) { | |
2059 | case BTRFS_CSUM_TYPE_CRC32: | |
2060 | if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) | |
2061 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2062 | break; | |
efcfcbc6 DS |
2063 | case BTRFS_CSUM_TYPE_XXHASH: |
2064 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2065 | break; | |
68d99ab0 CH |
2066 | default: |
2067 | break; | |
2068 | } | |
2069 | ||
c8a5f8ca DS |
2070 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2071 | btrfs_super_csum_name(csum_type), | |
2072 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2073 | return 0; |
2074 | } | |
2075 | ||
63443bf5 ES |
2076 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2077 | struct btrfs_fs_devices *fs_devices) | |
2078 | { | |
2079 | int ret; | |
789d6a3a | 2080 | struct btrfs_tree_parent_check check = { 0 }; |
63443bf5 ES |
2081 | struct btrfs_root *log_tree_root; |
2082 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2083 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2084 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2085 | |
2086 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2087 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2088 | return -EIO; |
2089 | } | |
2090 | ||
96dfcb46 JB |
2091 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2092 | GFP_KERNEL); | |
63443bf5 ES |
2093 | if (!log_tree_root) |
2094 | return -ENOMEM; | |
2095 | ||
789d6a3a QW |
2096 | check.level = level; |
2097 | check.transid = fs_info->generation + 1; | |
2098 | check.owner_root = BTRFS_TREE_LOG_OBJECTID; | |
2099 | log_tree_root->node = read_tree_block(fs_info, bytenr, &check); | |
64c043de | 2100 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2101 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2102 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2103 | log_tree_root->node = NULL; |
00246528 | 2104 | btrfs_put_root(log_tree_root); |
0eeff236 | 2105 | return ret; |
4eb150d6 QW |
2106 | } |
2107 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2108 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2109 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2110 | return -EIO; |
2111 | } | |
4eb150d6 | 2112 | |
63443bf5 ES |
2113 | /* returns with log_tree_root freed on success */ |
2114 | ret = btrfs_recover_log_trees(log_tree_root); | |
2115 | if (ret) { | |
0b246afa JM |
2116 | btrfs_handle_fs_error(fs_info, ret, |
2117 | "Failed to recover log tree"); | |
00246528 | 2118 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2119 | return ret; |
2120 | } | |
2121 | ||
bc98a42c | 2122 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2123 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2124 | if (ret) |
2125 | return ret; | |
2126 | } | |
2127 | ||
2128 | return 0; | |
2129 | } | |
2130 | ||
abed4aaa JB |
2131 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2132 | struct btrfs_path *path, u64 objectid, | |
2133 | const char *name) | |
2134 | { | |
2135 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2136 | struct btrfs_root *root; | |
f7238e50 | 2137 | u64 max_global_id = 0; |
abed4aaa JB |
2138 | int ret; |
2139 | struct btrfs_key key = { | |
2140 | .objectid = objectid, | |
2141 | .type = BTRFS_ROOT_ITEM_KEY, | |
2142 | .offset = 0, | |
2143 | }; | |
2144 | bool found = false; | |
2145 | ||
2146 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2147 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2148 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2149 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2150 | return 0; | |
2151 | } | |
2152 | ||
2153 | while (1) { | |
2154 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2155 | if (ret < 0) | |
2156 | break; | |
2157 | ||
2158 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2159 | ret = btrfs_next_leaf(tree_root, path); | |
2160 | if (ret) { | |
2161 | if (ret > 0) | |
2162 | ret = 0; | |
2163 | break; | |
2164 | } | |
2165 | } | |
2166 | ret = 0; | |
2167 | ||
2168 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2169 | if (key.objectid != objectid) | |
2170 | break; | |
2171 | btrfs_release_path(path); | |
2172 | ||
f7238e50 JB |
2173 | /* |
2174 | * Just worry about this for extent tree, it'll be the same for | |
2175 | * everybody. | |
2176 | */ | |
2177 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2178 | max_global_id = max(max_global_id, key.offset); | |
2179 | ||
abed4aaa JB |
2180 | found = true; |
2181 | root = read_tree_root_path(tree_root, path, &key); | |
2182 | if (IS_ERR(root)) { | |
2183 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2184 | ret = PTR_ERR(root); | |
2185 | break; | |
2186 | } | |
2187 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2188 | ret = btrfs_global_root_insert(root); | |
2189 | if (ret) { | |
2190 | btrfs_put_root(root); | |
2191 | break; | |
2192 | } | |
2193 | key.offset++; | |
2194 | } | |
2195 | btrfs_release_path(path); | |
2196 | ||
f7238e50 JB |
2197 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2198 | fs_info->nr_global_roots = max_global_id + 1; | |
2199 | ||
abed4aaa JB |
2200 | if (!found || ret) { |
2201 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2202 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2203 | ||
2204 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2205 | ret = ret ? ret : -ENOENT; | |
2206 | else | |
2207 | ret = 0; | |
2208 | btrfs_err(fs_info, "failed to load root %s", name); | |
2209 | } | |
2210 | return ret; | |
2211 | } | |
2212 | ||
2213 | static int load_global_roots(struct btrfs_root *tree_root) | |
2214 | { | |
2215 | struct btrfs_path *path; | |
2216 | int ret = 0; | |
2217 | ||
2218 | path = btrfs_alloc_path(); | |
2219 | if (!path) | |
2220 | return -ENOMEM; | |
2221 | ||
2222 | ret = load_global_roots_objectid(tree_root, path, | |
2223 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2224 | if (ret) | |
2225 | goto out; | |
2226 | ret = load_global_roots_objectid(tree_root, path, | |
2227 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2228 | if (ret) | |
2229 | goto out; | |
2230 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2231 | goto out; | |
2232 | ret = load_global_roots_objectid(tree_root, path, | |
2233 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2234 | "free space"); | |
2235 | out: | |
2236 | btrfs_free_path(path); | |
2237 | return ret; | |
2238 | } | |
2239 | ||
6bccf3ab | 2240 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2241 | { |
6bccf3ab | 2242 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2243 | struct btrfs_root *root; |
4bbcaa64 ES |
2244 | struct btrfs_key location; |
2245 | int ret; | |
2246 | ||
2467d0fe | 2247 | ASSERT(fs_info->tree_root); |
6bccf3ab | 2248 | |
abed4aaa JB |
2249 | ret = load_global_roots(tree_root); |
2250 | if (ret) | |
2251 | return ret; | |
2252 | ||
4bbcaa64 ES |
2253 | location.type = BTRFS_ROOT_ITEM_KEY; |
2254 | location.offset = 0; | |
2255 | ||
1c56ab99 | 2256 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2257 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2258 | root = btrfs_read_tree_root(tree_root, &location); | |
2259 | if (IS_ERR(root)) { | |
2260 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2261 | ret = PTR_ERR(root); | |
2262 | goto out; | |
2263 | } | |
2264 | } else { | |
2265 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2266 | fs_info->block_group_root = root; | |
2267 | } | |
2268 | } | |
2269 | ||
2270 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2271 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2272 | if (IS_ERR(root)) { |
42437a63 JB |
2273 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2274 | ret = PTR_ERR(root); | |
2275 | goto out; | |
2276 | } | |
2277 | } else { | |
2278 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2279 | fs_info->dev_root = root; | |
f50f4353 | 2280 | } |
820a49da | 2281 | /* Initialize fs_info for all devices in any case */ |
a8d1b164 JT |
2282 | ret = btrfs_init_devices_late(fs_info); |
2283 | if (ret) | |
2284 | goto out; | |
4bbcaa64 | 2285 | |
aeb935a4 QW |
2286 | /* |
2287 | * This tree can share blocks with some other fs tree during relocation | |
2288 | * and we need a proper setup by btrfs_get_fs_root | |
2289 | */ | |
56e9357a DS |
2290 | root = btrfs_get_fs_root(tree_root->fs_info, |
2291 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2292 | if (IS_ERR(root)) { |
42437a63 JB |
2293 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2294 | ret = PTR_ERR(root); | |
2295 | goto out; | |
2296 | } | |
2297 | } else { | |
2298 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2299 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2300 | } |
aeb935a4 | 2301 | |
4bbcaa64 | 2302 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2303 | root = btrfs_read_tree_root(tree_root, &location); |
2304 | if (!IS_ERR(root)) { | |
2305 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
a4f3d2c4 | 2306 | fs_info->quota_root = root; |
4bbcaa64 ES |
2307 | } |
2308 | ||
2309 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2310 | root = btrfs_read_tree_root(tree_root, &location); |
2311 | if (IS_ERR(root)) { | |
42437a63 JB |
2312 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2313 | ret = PTR_ERR(root); | |
2314 | if (ret != -ENOENT) | |
2315 | goto out; | |
2316 | } | |
4bbcaa64 | 2317 | } else { |
a4f3d2c4 DS |
2318 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2319 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2320 | } |
2321 | ||
51502090 JT |
2322 | if (btrfs_fs_incompat(fs_info, RAID_STRIPE_TREE)) { |
2323 | location.objectid = BTRFS_RAID_STRIPE_TREE_OBJECTID; | |
2324 | root = btrfs_read_tree_root(tree_root, &location); | |
2325 | if (IS_ERR(root)) { | |
2326 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2327 | ret = PTR_ERR(root); | |
2328 | goto out; | |
2329 | } | |
2330 | } else { | |
2331 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2332 | fs_info->stripe_root = root; | |
2333 | } | |
2334 | } | |
2335 | ||
4bbcaa64 | 2336 | return 0; |
f50f4353 LB |
2337 | out: |
2338 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2339 | location.objectid, ret); | |
2340 | return ret; | |
4bbcaa64 ES |
2341 | } |
2342 | ||
069ec957 QW |
2343 | /* |
2344 | * Real super block validation | |
2345 | * NOTE: super csum type and incompat features will not be checked here. | |
2346 | * | |
2347 | * @sb: super block to check | |
2348 | * @mirror_num: the super block number to check its bytenr: | |
2349 | * 0 the primary (1st) sb | |
2350 | * 1, 2 2nd and 3rd backup copy | |
2351 | * -1 skip bytenr check | |
2352 | */ | |
a05d3c91 QW |
2353 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2354 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2355 | { |
21a852b0 QW |
2356 | u64 nodesize = btrfs_super_nodesize(sb); |
2357 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2358 | int ret = 0; | |
2359 | ||
2360 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2361 | btrfs_err(fs_info, "no valid FS found"); | |
2362 | ret = -EINVAL; | |
2363 | } | |
2364 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2365 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2366 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2367 | ret = -EINVAL; | |
2368 | } | |
2369 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2370 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2371 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2372 | ret = -EINVAL; | |
2373 | } | |
2374 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2375 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2376 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2377 | ret = -EINVAL; | |
2378 | } | |
2379 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2380 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2381 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2382 | ret = -EINVAL; | |
2383 | } | |
2384 | ||
2385 | /* | |
2386 | * Check sectorsize and nodesize first, other check will need it. | |
2387 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2388 | */ | |
2389 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2390 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2391 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2392 | ret = -EINVAL; | |
2393 | } | |
0bb3eb3e QW |
2394 | |
2395 | /* | |
1a42daab QW |
2396 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2397 | * | |
2398 | * We can support 16K sectorsize with 64K page size without problem, | |
2399 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2400 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2401 | * beginning. | |
0bb3eb3e | 2402 | */ |
1a42daab | 2403 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2404 | btrfs_err(fs_info, |
0bb3eb3e | 2405 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2406 | sectorsize, PAGE_SIZE); |
2407 | ret = -EINVAL; | |
2408 | } | |
0bb3eb3e | 2409 | |
21a852b0 QW |
2410 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2411 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2412 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2413 | ret = -EINVAL; | |
2414 | } | |
2415 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2416 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2417 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2418 | ret = -EINVAL; | |
2419 | } | |
2420 | ||
2421 | /* Root alignment check */ | |
2422 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2423 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2424 | btrfs_super_root(sb)); | |
2425 | ret = -EINVAL; | |
2426 | } | |
2427 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2428 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2429 | btrfs_super_chunk_root(sb)); | |
2430 | ret = -EINVAL; | |
2431 | } | |
2432 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2433 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2434 | btrfs_super_log_root(sb)); | |
2435 | ret = -EINVAL; | |
2436 | } | |
2437 | ||
a5b8a5f9 AJ |
2438 | if (!fs_info->fs_devices->temp_fsid && |
2439 | memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2440 | btrfs_err(fs_info, |
2441 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
d167aa76 | 2442 | sb->fsid, fs_info->fs_devices->fsid); |
aefd7f70 NB |
2443 | ret = -EINVAL; |
2444 | } | |
2445 | ||
6bfe3959 AJ |
2446 | if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), |
2447 | BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2448 | btrfs_err(fs_info, |
2449 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
6bfe3959 | 2450 | btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); |
aefd7f70 NB |
2451 | ret = -EINVAL; |
2452 | } | |
2453 | ||
25984a5a AJ |
2454 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
2455 | BTRFS_FSID_SIZE) != 0) { | |
2456 | btrfs_err(fs_info, | |
2457 | "dev_item UUID does not match metadata fsid: %pU != %pU", | |
2458 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); | |
2459 | ret = -EINVAL; | |
2460 | } | |
2461 | ||
1c56ab99 QW |
2462 | /* |
2463 | * Artificial requirement for block-group-tree to force newer features | |
2464 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2465 | */ | |
2466 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2467 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2468 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2469 | btrfs_err(fs_info, | |
2470 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2471 | ret = -EINVAL; | |
2472 | } | |
2473 | ||
21a852b0 QW |
2474 | /* |
2475 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2476 | * done later | |
2477 | */ | |
2478 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2479 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2480 | btrfs_super_bytes_used(sb)); | |
2481 | ret = -EINVAL; | |
2482 | } | |
2483 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2484 | btrfs_err(fs_info, "invalid stripesize %u", | |
2485 | btrfs_super_stripesize(sb)); | |
2486 | ret = -EINVAL; | |
2487 | } | |
2488 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2489 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2490 | btrfs_super_num_devices(sb)); | |
2491 | if (btrfs_super_num_devices(sb) == 0) { | |
2492 | btrfs_err(fs_info, "number of devices is 0"); | |
2493 | ret = -EINVAL; | |
2494 | } | |
2495 | ||
069ec957 QW |
2496 | if (mirror_num >= 0 && |
2497 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2498 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2499 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2500 | ret = -EINVAL; | |
2501 | } | |
2502 | ||
2503 | /* | |
2504 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2505 | * and one chunk | |
2506 | */ | |
2507 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2508 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2509 | btrfs_super_sys_array_size(sb), | |
2510 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2511 | ret = -EINVAL; | |
2512 | } | |
2513 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2514 | + sizeof(struct btrfs_chunk)) { | |
2515 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2516 | btrfs_super_sys_array_size(sb), | |
2517 | sizeof(struct btrfs_disk_key) | |
2518 | + sizeof(struct btrfs_chunk)); | |
2519 | ret = -EINVAL; | |
2520 | } | |
2521 | ||
2522 | /* | |
2523 | * The generation is a global counter, we'll trust it more than the others | |
2524 | * but it's still possible that it's the one that's wrong. | |
2525 | */ | |
2526 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2527 | btrfs_warn(fs_info, | |
2528 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2529 | btrfs_super_generation(sb), | |
2530 | btrfs_super_chunk_root_generation(sb)); | |
2531 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2532 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2533 | btrfs_warn(fs_info, | |
2534 | "suspicious: generation < cache_generation: %llu < %llu", | |
2535 | btrfs_super_generation(sb), | |
2536 | btrfs_super_cache_generation(sb)); | |
2537 | ||
2538 | return ret; | |
2539 | } | |
2540 | ||
069ec957 QW |
2541 | /* |
2542 | * Validation of super block at mount time. | |
2543 | * Some checks already done early at mount time, like csum type and incompat | |
2544 | * flags will be skipped. | |
2545 | */ | |
2546 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2547 | { | |
a05d3c91 | 2548 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2549 | } |
2550 | ||
75cb857d QW |
2551 | /* |
2552 | * Validation of super block at write time. | |
2553 | * Some checks like bytenr check will be skipped as their values will be | |
2554 | * overwritten soon. | |
2555 | * Extra checks like csum type and incompat flags will be done here. | |
2556 | */ | |
2557 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2558 | struct btrfs_super_block *sb) | |
2559 | { | |
2560 | int ret; | |
2561 | ||
a05d3c91 | 2562 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2563 | if (ret < 0) |
2564 | goto out; | |
e7e16f48 | 2565 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2566 | ret = -EUCLEAN; |
2567 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2568 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2569 | goto out; | |
2570 | } | |
2571 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2572 | ret = -EUCLEAN; | |
2573 | btrfs_err(fs_info, | |
2574 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2575 | btrfs_super_incompat_flags(sb), | |
2576 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2577 | goto out; | |
2578 | } | |
2579 | out: | |
2580 | if (ret < 0) | |
2581 | btrfs_err(fs_info, | |
2582 | "super block corruption detected before writing it to disk"); | |
2583 | return ret; | |
2584 | } | |
2585 | ||
bd676446 JB |
2586 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2587 | { | |
789d6a3a QW |
2588 | struct btrfs_tree_parent_check check = { |
2589 | .level = level, | |
2590 | .transid = gen, | |
e094f480 | 2591 | .owner_root = btrfs_root_id(root) |
789d6a3a | 2592 | }; |
bd676446 JB |
2593 | int ret = 0; |
2594 | ||
789d6a3a | 2595 | root->node = read_tree_block(root->fs_info, bytenr, &check); |
bd676446 JB |
2596 | if (IS_ERR(root->node)) { |
2597 | ret = PTR_ERR(root->node); | |
2598 | root->node = NULL; | |
4eb150d6 QW |
2599 | return ret; |
2600 | } | |
2601 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2602 | free_extent_buffer(root->node); |
2603 | root->node = NULL; | |
4eb150d6 | 2604 | return -EIO; |
bd676446 JB |
2605 | } |
2606 | ||
bd676446 JB |
2607 | btrfs_set_root_node(&root->root_item, root->node); |
2608 | root->commit_root = btrfs_root_node(root); | |
2609 | btrfs_set_root_refs(&root->root_item, 1); | |
2610 | return ret; | |
2611 | } | |
2612 | ||
2613 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2614 | { | |
2615 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2616 | u64 gen, bytenr; | |
2617 | int level, ret; | |
2618 | ||
2619 | bytenr = btrfs_super_root(sb); | |
2620 | gen = btrfs_super_generation(sb); | |
2621 | level = btrfs_super_root_level(sb); | |
2622 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2623 | if (ret) { |
bd676446 | 2624 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2625 | return ret; |
2626 | } | |
14033b08 | 2627 | return 0; |
bd676446 JB |
2628 | } |
2629 | ||
6ef108dd | 2630 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2631 | { |
6ef108dd | 2632 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2633 | struct btrfs_super_block *sb = fs_info->super_copy; |
2634 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2635 | bool handle_error = false; | |
2636 | int ret = 0; | |
2637 | int i; | |
2638 | ||
2639 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2640 | if (handle_error) { |
2641 | if (!IS_ERR(tree_root->node)) | |
2642 | free_extent_buffer(tree_root->node); | |
2643 | tree_root->node = NULL; | |
2644 | ||
2645 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2646 | break; | |
2647 | ||
2648 | free_root_pointers(fs_info, 0); | |
2649 | ||
2650 | /* | |
2651 | * Don't use the log in recovery mode, it won't be | |
2652 | * valid | |
2653 | */ | |
2654 | btrfs_set_super_log_root(sb, 0); | |
2655 | ||
745806fb | 2656 | btrfs_warn(fs_info, "try to load backup roots slot %d", i); |
b8522a1e | 2657 | ret = read_backup_root(fs_info, i); |
6ef108dd | 2658 | backup_index = ret; |
b8522a1e NB |
2659 | if (ret < 0) |
2660 | return ret; | |
2661 | } | |
b8522a1e | 2662 | |
bd676446 JB |
2663 | ret = load_important_roots(fs_info); |
2664 | if (ret) { | |
217f5004 | 2665 | handle_error = true; |
b8522a1e NB |
2666 | continue; |
2667 | } | |
2668 | ||
336a0d8d NB |
2669 | /* |
2670 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2671 | * hasn't been fully initialised and we are the only user | |
2672 | */ | |
453e4873 | 2673 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2674 | if (ret < 0) { |
b8522a1e NB |
2675 | handle_error = true; |
2676 | continue; | |
2677 | } | |
2678 | ||
6b8fad57 | 2679 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2680 | |
2681 | ret = btrfs_read_roots(fs_info); | |
2682 | if (ret < 0) { | |
2683 | handle_error = true; | |
2684 | continue; | |
2685 | } | |
2686 | ||
2687 | /* All successful */ | |
bd676446 | 2688 | fs_info->generation = btrfs_header_generation(tree_root->node); |
0124855f | 2689 | btrfs_set_last_trans_committed(fs_info, fs_info->generation); |
d96b3424 | 2690 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2691 | |
2692 | /* Always begin writing backup roots after the one being used */ | |
2693 | if (backup_index < 0) { | |
2694 | fs_info->backup_root_index = 0; | |
2695 | } else { | |
2696 | fs_info->backup_root_index = backup_index + 1; | |
2697 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2698 | } | |
b8522a1e NB |
2699 | break; |
2700 | } | |
2701 | ||
2702 | return ret; | |
2703 | } | |
2704 | ||
8260edba | 2705 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2706 | { |
fc7cbcd4 | 2707 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2708 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2709 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2710 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2711 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2712 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2713 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2714 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2715 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2716 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2717 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2718 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2719 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2720 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2721 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2722 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2723 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2724 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2725 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2726 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2727 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2728 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2729 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2730 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2731 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2732 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2733 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2734 | |
e1489b4f | 2735 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2736 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2737 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2738 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
77d20c68 JB |
2739 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_prep, |
2740 | BTRFS_LOCKDEP_TRANS_COMMIT_PREP); | |
3e738c53 IA |
2741 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, |
2742 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
2743 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
2744 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
2745 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
2746 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 2747 | |
0b86a832 | 2748 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2749 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2750 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2751 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2752 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 2753 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
2754 | #ifdef CONFIG_BTRFS_DEBUG |
2755 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2756 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2757 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2758 | #endif |
7dc66abb FM |
2759 | fs_info->mapping_tree = RB_ROOT_CACHED; |
2760 | rwlock_init(&fs_info->mapping_tree_lock); | |
66d8f3dd MX |
2761 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2762 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2763 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2764 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2765 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2766 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2767 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2768 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2769 | BTRFS_BLOCK_RSV_DELREFS); | |
2770 | ||
771ed689 | 2771 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2772 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 2773 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2774 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 2775 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 2776 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2777 | fs_info->metadata_ratio = 0; |
4cb5300b | 2778 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2779 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2780 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2781 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
fd708b81 | 2782 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2783 | |
b34b086c CM |
2784 | fs_info->thread_pool_size = min_t(unsigned long, |
2785 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2786 | |
199c2a9c MX |
2787 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2788 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2789 | |
638aa7ed | 2790 | btrfs_init_scrub(fs_info); |
779a65a4 | 2791 | btrfs_init_balance(fs_info); |
57056740 | 2792 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2793 | |
16b0c258 | 2794 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 2795 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 2796 | |
fe119a6e | 2797 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
35da5a7e | 2798 | IO_TREE_FS_EXCLUDED_EXTENTS); |
39279cc3 | 2799 | |
5a3f23d5 | 2800 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2801 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2802 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2803 | mutex_init(&fs_info->transaction_kthread_mutex); |
2804 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2805 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2806 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2807 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2808 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2809 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2810 | |
ad618368 | 2811 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2812 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2813 | btrfs_discard_init(fs_info); |
416ac51d | 2814 | |
fa9c0d79 CM |
2815 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2816 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2817 | ||
e6dcd2dc | 2818 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2819 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2820 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2821 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2822 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2823 | |
da17066c JM |
2824 | /* Usable values until the real ones are cached from the superblock */ |
2825 | fs_info->nodesize = 4096; | |
2826 | fs_info->sectorsize = 4096; | |
ab108d99 | 2827 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
2828 | fs_info->stripesize = 4096; |
2829 | ||
6207c9e3 JB |
2830 | /* Default compress algorithm when user does -o compress */ |
2831 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2832 | ||
f7b12a62 NA |
2833 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
2834 | ||
eede2bf3 OS |
2835 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2836 | fs_info->swapfile_pins = RB_ROOT; | |
2837 | ||
18bb8bbf JT |
2838 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
2839 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
2840 | } |
2841 | ||
2842 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2843 | { | |
2844 | int ret; | |
2845 | ||
2846 | fs_info->sb = sb; | |
4e00422e | 2847 | /* Temporary fixed values for block size until we read the superblock. */ |
8260edba JB |
2848 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2849 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 2850 | |
5deb17e1 | 2851 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 2852 | if (ret) |
c75e8394 | 2853 | return ret; |
ae18c37a | 2854 | |
f1d97e76 FM |
2855 | ret = percpu_counter_init(&fs_info->evictable_extent_maps, 0, GFP_KERNEL); |
2856 | if (ret) | |
2857 | return ret; | |
2858 | ||
ae18c37a JB |
2859 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
2860 | if (ret) | |
c75e8394 | 2861 | return ret; |
ae18c37a JB |
2862 | |
2863 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
2864 | (1 + ilog2(nr_cpu_ids)); | |
2865 | ||
2866 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
2867 | if (ret) | |
c75e8394 | 2868 | return ret; |
ae18c37a JB |
2869 | |
2870 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
2871 | GFP_KERNEL); | |
2872 | if (ret) | |
c75e8394 | 2873 | return ret; |
ae18c37a JB |
2874 | |
2875 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
2876 | GFP_KERNEL); | |
c75e8394 JB |
2877 | if (!fs_info->delayed_root) |
2878 | return -ENOMEM; | |
ae18c37a JB |
2879 | btrfs_init_delayed_root(fs_info->delayed_root); |
2880 | ||
a0a1db70 FM |
2881 | if (sb_rdonly(sb)) |
2882 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
2883 | ||
c75e8394 | 2884 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
2885 | } |
2886 | ||
97f4dd09 NB |
2887 | static int btrfs_uuid_rescan_kthread(void *data) |
2888 | { | |
0d031dc4 | 2889 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
2890 | int ret; |
2891 | ||
2892 | /* | |
2893 | * 1st step is to iterate through the existing UUID tree and | |
2894 | * to delete all entries that contain outdated data. | |
2895 | * 2nd step is to add all missing entries to the UUID tree. | |
2896 | */ | |
2897 | ret = btrfs_uuid_tree_iterate(fs_info); | |
2898 | if (ret < 0) { | |
c94bec2c JB |
2899 | if (ret != -EINTR) |
2900 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
2901 | ret); | |
97f4dd09 NB |
2902 | up(&fs_info->uuid_tree_rescan_sem); |
2903 | return ret; | |
2904 | } | |
2905 | return btrfs_uuid_scan_kthread(data); | |
2906 | } | |
2907 | ||
2908 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
2909 | { | |
2910 | struct task_struct *task; | |
2911 | ||
2912 | down(&fs_info->uuid_tree_rescan_sem); | |
2913 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
2914 | if (IS_ERR(task)) { | |
2915 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
2916 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
2917 | up(&fs_info->uuid_tree_rescan_sem); | |
2918 | return PTR_ERR(task); | |
2919 | } | |
2920 | ||
2921 | return 0; | |
2922 | } | |
2923 | ||
504b1596 FM |
2924 | static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
2925 | { | |
2926 | u64 root_objectid = 0; | |
2927 | struct btrfs_root *gang[8]; | |
2928 | int i = 0; | |
2929 | int err = 0; | |
2930 | unsigned int ret = 0; | |
2931 | ||
2932 | while (1) { | |
2933 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2934 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2935 | (void **)gang, root_objectid, | |
2936 | ARRAY_SIZE(gang)); | |
2937 | if (!ret) { | |
2938 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2939 | break; | |
2940 | } | |
e094f480 | 2941 | root_objectid = btrfs_root_id(gang[ret - 1]) + 1; |
504b1596 FM |
2942 | |
2943 | for (i = 0; i < ret; i++) { | |
2944 | /* Avoid to grab roots in dead_roots. */ | |
2945 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
2946 | gang[i] = NULL; | |
2947 | continue; | |
2948 | } | |
2949 | /* Grab all the search result for later use. */ | |
2950 | gang[i] = btrfs_grab_root(gang[i]); | |
2951 | } | |
2952 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2953 | ||
2954 | for (i = 0; i < ret; i++) { | |
2955 | if (!gang[i]) | |
2956 | continue; | |
e094f480 | 2957 | root_objectid = btrfs_root_id(gang[i]); |
504b1596 FM |
2958 | err = btrfs_orphan_cleanup(gang[i]); |
2959 | if (err) | |
2960 | goto out; | |
2961 | btrfs_put_root(gang[i]); | |
2962 | } | |
2963 | root_objectid++; | |
2964 | } | |
2965 | out: | |
2966 | /* Release the uncleaned roots due to error. */ | |
2967 | for (; i < ret; i++) { | |
2968 | if (gang[i]) | |
2969 | btrfs_put_root(gang[i]); | |
2970 | } | |
2971 | return err; | |
2972 | } | |
2973 | ||
44c0ca21 BB |
2974 | /* |
2975 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
2976 | * and btrfs_remount when remounting from read-only to read-write. | |
2977 | */ | |
2978 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
2979 | { | |
2980 | int ret; | |
94846229 | 2981 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
1d6a4fc8 | 2982 | bool rebuild_free_space_tree = false; |
8b228324 BB |
2983 | |
2984 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
2985 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
272efa30 JB |
2986 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) |
2987 | btrfs_warn(fs_info, | |
2988 | "'clear_cache' option is ignored with extent tree v2"); | |
2989 | else | |
2990 | rebuild_free_space_tree = true; | |
8b228324 BB |
2991 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && |
2992 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
2993 | btrfs_warn(fs_info, "free space tree is invalid"); | |
1d6a4fc8 | 2994 | rebuild_free_space_tree = true; |
8b228324 BB |
2995 | } |
2996 | ||
1d6a4fc8 QW |
2997 | if (rebuild_free_space_tree) { |
2998 | btrfs_info(fs_info, "rebuilding free space tree"); | |
2999 | ret = btrfs_rebuild_free_space_tree(fs_info); | |
8b228324 BB |
3000 | if (ret) { |
3001 | btrfs_warn(fs_info, | |
1d6a4fc8 QW |
3002 | "failed to rebuild free space tree: %d", ret); |
3003 | goto out; | |
3004 | } | |
3005 | } | |
3006 | ||
3007 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3008 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { | |
3009 | btrfs_info(fs_info, "disabling free space tree"); | |
3010 | ret = btrfs_delete_free_space_tree(fs_info); | |
3011 | if (ret) { | |
3012 | btrfs_warn(fs_info, | |
3013 | "failed to disable free space tree: %d", ret); | |
8b228324 BB |
3014 | goto out; |
3015 | } | |
3016 | } | |
44c0ca21 | 3017 | |
8d488a8c FM |
3018 | /* |
3019 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3020 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3021 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3022 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3023 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3024 | * item before the root's tree is deleted - this means that if we unmount | |
3025 | * or crash before the deletion completes, on the next mount we will not | |
3026 | * delete what remains of the tree because the orphan item does not | |
3027 | * exists anymore, which is what tells us we have a pending deletion. | |
3028 | */ | |
3029 | ret = btrfs_find_orphan_roots(fs_info); | |
3030 | if (ret) | |
3031 | goto out; | |
3032 | ||
44c0ca21 BB |
3033 | ret = btrfs_cleanup_fs_roots(fs_info); |
3034 | if (ret) | |
3035 | goto out; | |
3036 | ||
8f1c21d7 BB |
3037 | down_read(&fs_info->cleanup_work_sem); |
3038 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3039 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3040 | up_read(&fs_info->cleanup_work_sem); | |
3041 | goto out; | |
3042 | } | |
3043 | up_read(&fs_info->cleanup_work_sem); | |
3044 | ||
44c0ca21 | 3045 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3046 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3047 | mutex_unlock(&fs_info->cleaner_mutex); |
3048 | if (ret < 0) { | |
3049 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3050 | goto out; | |
3051 | } | |
3052 | ||
5011139a BB |
3053 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3054 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3055 | btrfs_info(fs_info, "creating free space tree"); | |
3056 | ret = btrfs_create_free_space_tree(fs_info); | |
3057 | if (ret) { | |
3058 | btrfs_warn(fs_info, | |
3059 | "failed to create free space tree: %d", ret); | |
3060 | goto out; | |
3061 | } | |
3062 | } | |
3063 | ||
94846229 BB |
3064 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3065 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3066 | if (ret) | |
3067 | goto out; | |
3068 | } | |
3069 | ||
44c0ca21 BB |
3070 | ret = btrfs_resume_balance_async(fs_info); |
3071 | if (ret) | |
3072 | goto out; | |
3073 | ||
3074 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3075 | if (ret) { | |
3076 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3077 | goto out; | |
3078 | } | |
3079 | ||
3080 | btrfs_qgroup_rescan_resume(fs_info); | |
3081 | ||
3082 | if (!fs_info->uuid_root) { | |
3083 | btrfs_info(fs_info, "creating UUID tree"); | |
3084 | ret = btrfs_create_uuid_tree(fs_info); | |
3085 | if (ret) { | |
3086 | btrfs_warn(fs_info, | |
3087 | "failed to create the UUID tree %d", ret); | |
3088 | goto out; | |
3089 | } | |
3090 | } | |
3091 | ||
3092 | out: | |
3093 | return ret; | |
3094 | } | |
3095 | ||
d7f67ac9 QW |
3096 | /* |
3097 | * Do various sanity and dependency checks of different features. | |
3098 | * | |
2ba48b20 QW |
3099 | * @is_rw_mount: If the mount is read-write. |
3100 | * | |
d7f67ac9 QW |
3101 | * This is the place for less strict checks (like for subpage or artificial |
3102 | * feature dependencies). | |
3103 | * | |
3104 | * For strict checks or possible corruption detection, see | |
3105 | * btrfs_validate_super(). | |
3106 | * | |
3107 | * This should be called after btrfs_parse_options(), as some mount options | |
3108 | * (space cache related) can modify on-disk format like free space tree and | |
3109 | * screw up certain feature dependencies. | |
3110 | */ | |
2ba48b20 | 3111 | int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) |
d7f67ac9 QW |
3112 | { |
3113 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3114 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3115 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3116 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3117 | ||
3118 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3119 | btrfs_err(fs_info, | |
3120 | "cannot mount because of unknown incompat features (0x%llx)", | |
3121 | incompat); | |
3122 | return -EINVAL; | |
3123 | } | |
3124 | ||
3125 | /* Runtime limitation for mixed block groups. */ | |
3126 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3127 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3128 | btrfs_err(fs_info, | |
3129 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3130 | fs_info->nodesize, fs_info->sectorsize); | |
3131 | return -EINVAL; | |
3132 | } | |
3133 | ||
3134 | /* Mixed backref is an always-enabled feature. */ | |
3135 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3136 | ||
3137 | /* Set compression related flags just in case. */ | |
3138 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3139 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3140 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3141 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3142 | ||
3143 | /* | |
3144 | * An ancient flag, which should really be marked deprecated. | |
3145 | * Such runtime limitation doesn't really need a incompat flag. | |
3146 | */ | |
3147 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3148 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3149 | ||
2ba48b20 | 3150 | if (compat_ro_unsupp && is_rw_mount) { |
d7f67ac9 QW |
3151 | btrfs_err(fs_info, |
3152 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3153 | compat_ro); | |
3154 | return -EINVAL; | |
3155 | } | |
3156 | ||
3157 | /* | |
3158 | * We have unsupported RO compat features, although RO mounted, we | |
3159 | * should not cause any metadata writes, including log replay. | |
3160 | * Or we could screw up whatever the new feature requires. | |
3161 | */ | |
3162 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3163 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3164 | btrfs_err(fs_info, | |
3165 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3166 | compat_ro); | |
3167 | return -EINVAL; | |
3168 | } | |
3169 | ||
3170 | /* | |
3171 | * Artificial limitations for block group tree, to force | |
3172 | * block-group-tree to rely on no-holes and free-space-tree. | |
3173 | */ | |
3174 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3175 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3176 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3177 | btrfs_err(fs_info, | |
3178 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3179 | return -EINVAL; | |
3180 | } | |
3181 | ||
3182 | /* | |
3183 | * Subpage runtime limitation on v1 cache. | |
3184 | * | |
3185 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3186 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3187 | * going to be deprecated anyway. | |
3188 | */ | |
3189 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3190 | btrfs_warn(fs_info, | |
3191 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3192 | PAGE_SIZE, fs_info->sectorsize); | |
3193 | return -EINVAL; | |
3194 | } | |
3195 | ||
3196 | /* This can be called by remount, we need to protect the super block. */ | |
3197 | spin_lock(&fs_info->super_lock); | |
3198 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3199 | spin_unlock(&fs_info->super_lock); | |
3200 | ||
3201 | return 0; | |
3202 | } | |
3203 | ||
ae18c37a JB |
3204 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3205 | char *options) | |
3206 | { | |
3207 | u32 sectorsize; | |
3208 | u32 nodesize; | |
3209 | u32 stripesize; | |
3210 | u64 generation; | |
ae18c37a | 3211 | u16 csum_type; |
ae18c37a JB |
3212 | struct btrfs_super_block *disk_super; |
3213 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3214 | struct btrfs_root *tree_root; | |
3215 | struct btrfs_root *chunk_root; | |
3216 | int ret; | |
ae18c37a JB |
3217 | int level; |
3218 | ||
8260edba | 3219 | ret = init_mount_fs_info(fs_info, sb); |
4871c33b | 3220 | if (ret) |
ae18c37a | 3221 | goto fail; |
53b381b3 | 3222 | |
ae18c37a JB |
3223 | /* These need to be init'ed before we start creating inodes and such. */ |
3224 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3225 | GFP_KERNEL); | |
3226 | fs_info->tree_root = tree_root; | |
3227 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3228 | GFP_KERNEL); | |
3229 | fs_info->chunk_root = chunk_root; | |
3230 | if (!tree_root || !chunk_root) { | |
4871c33b | 3231 | ret = -ENOMEM; |
c75e8394 | 3232 | goto fail; |
ae18c37a JB |
3233 | } |
3234 | ||
dcb2137c | 3235 | ret = btrfs_init_btree_inode(sb); |
4871c33b | 3236 | if (ret) |
c75e8394 | 3237 | goto fail; |
ae18c37a | 3238 | |
d24fa5c1 | 3239 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3240 | |
3241 | /* | |
3242 | * Read super block and check the signature bytes only | |
3243 | */ | |
d24fa5c1 | 3244 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d | 3245 | if (IS_ERR(disk_super)) { |
4871c33b | 3246 | ret = PTR_ERR(disk_super); |
16cdcec7 | 3247 | goto fail_alloc; |
20b45077 | 3248 | } |
39279cc3 | 3249 | |
2db31320 | 3250 | btrfs_info(fs_info, "first mount of filesystem %pU", disk_super->fsid); |
8dc3f22c | 3251 | /* |
260db43c | 3252 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3253 | * corrupted, we'll find out |
3254 | */ | |
8f32380d | 3255 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3256 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3257 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3258 | csum_type); |
4871c33b | 3259 | ret = -EINVAL; |
8f32380d | 3260 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3261 | goto fail_alloc; |
3262 | } | |
3263 | ||
83c68bbc SY |
3264 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3265 | ||
6d97c6e3 JT |
3266 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3267 | if (ret) { | |
8f32380d | 3268 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3269 | goto fail_alloc; |
3270 | } | |
3271 | ||
1104a885 DS |
3272 | /* |
3273 | * We want to check superblock checksum, the type is stored inside. | |
3274 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3275 | */ | |
3d17adea | 3276 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3277 | btrfs_err(fs_info, "superblock checksum mismatch"); |
4871c33b | 3278 | ret = -EINVAL; |
8f32380d | 3279 | btrfs_release_disk_super(disk_super); |
141386e1 | 3280 | goto fail_alloc; |
1104a885 DS |
3281 | } |
3282 | ||
3283 | /* | |
3284 | * super_copy is zeroed at allocation time and we never touch the | |
3285 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3286 | * the whole block of INFO_SIZE | |
3287 | */ | |
8f32380d JT |
3288 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3289 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3290 | |
fbc6feae NB |
3291 | disk_super = fs_info->super_copy; |
3292 | ||
fbc6feae NB |
3293 | memcpy(fs_info->super_for_commit, fs_info->super_copy, |
3294 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3295 | |
069ec957 | 3296 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3297 | if (ret) { |
05135f59 | 3298 | btrfs_err(fs_info, "superblock contains fatal errors"); |
4871c33b | 3299 | ret = -EINVAL; |
141386e1 | 3300 | goto fail_alloc; |
1104a885 DS |
3301 | } |
3302 | ||
4871c33b QW |
3303 | if (!btrfs_super_root(disk_super)) { |
3304 | btrfs_err(fs_info, "invalid superblock tree root bytenr"); | |
3305 | ret = -EINVAL; | |
141386e1 | 3306 | goto fail_alloc; |
4871c33b | 3307 | } |
0f7d52f4 | 3308 | |
acce952b | 3309 | /* check FS state, whether FS is broken. */ |
87533c47 | 3310 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
ae3364e5 | 3311 | WRITE_ONCE(fs_info->fs_error, -EUCLEAN); |
acce952b | 3312 | |
6f93e834 AJ |
3313 | /* Set up fs_info before parsing mount options */ |
3314 | nodesize = btrfs_super_nodesize(disk_super); | |
3315 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3316 | stripesize = sectorsize; | |
3317 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3318 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3319 | ||
3320 | fs_info->nodesize = nodesize; | |
3321 | fs_info->sectorsize = sectorsize; | |
3322 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3323 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3324 | fs_info->stripesize = stripesize; | |
3325 | ||
a6a8f22a JB |
3326 | /* |
3327 | * Handle the space caching options appropriately now that we have the | |
3328 | * super block loaded and validated. | |
3329 | */ | |
3330 | btrfs_set_free_space_cache_settings(fs_info); | |
3331 | ||
ad21f15b JB |
3332 | if (!btrfs_check_options(fs_info, &fs_info->mount_opt, sb->s_flags)) { |
3333 | ret = -EINVAL; | |
141386e1 | 3334 | goto fail_alloc; |
ad21f15b | 3335 | } |
dfe25020 | 3336 | |
2ba48b20 | 3337 | ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); |
4871c33b | 3338 | if (ret < 0) |
dc4d3168 | 3339 | goto fail_alloc; |
dc4d3168 | 3340 | |
ad21f15b JB |
3341 | /* |
3342 | * At this point our mount options are validated, if we set ->max_inline | |
3343 | * to something non-standard make sure we truncate it to sectorsize. | |
3344 | */ | |
3345 | fs_info->max_inline = min_t(u64, fs_info->max_inline, fs_info->sectorsize); | |
3346 | ||
8481dd80 QW |
3347 | if (sectorsize < PAGE_SIZE) { |
3348 | struct btrfs_subpage_info *subpage_info; | |
3349 | ||
95ea0486 QW |
3350 | btrfs_warn(fs_info, |
3351 | "read-write for sector size %u with page size %lu is experimental", | |
3352 | sectorsize, PAGE_SIZE); | |
8481dd80 | 3353 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
4871c33b QW |
3354 | if (!subpage_info) { |
3355 | ret = -ENOMEM; | |
8481dd80 | 3356 | goto fail_alloc; |
4871c33b | 3357 | } |
8481dd80 QW |
3358 | btrfs_init_subpage_info(subpage_info, sectorsize); |
3359 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3360 | } |
0bb3eb3e | 3361 | |
d21deec5 | 3362 | ret = btrfs_init_workqueues(fs_info); |
4871c33b | 3363 | if (ret) |
0dc3b84a | 3364 | goto fail_sb_buffer; |
4543df7e | 3365 | |
9e11ceee JK |
3366 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3367 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3368 | |
4e00422e | 3369 | /* Update the values for the current filesystem. */ |
a061fc8d CM |
3370 | sb->s_blocksize = sectorsize; |
3371 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3372 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3373 | |
925baedd | 3374 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3375 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3376 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3377 | if (ret) { |
05135f59 | 3378 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3379 | goto fail_sb_buffer; |
84eed90f | 3380 | } |
0b86a832 | 3381 | |
84234f3a | 3382 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3383 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3384 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3385 | generation, level); | |
3386 | if (ret) { | |
05135f59 | 3387 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3388 | goto fail_tree_roots; |
83121942 | 3389 | } |
0b86a832 | 3390 | |
e17cade2 | 3391 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3392 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3393 | BTRFS_UUID_SIZE); | |
e17cade2 | 3394 | |
5b4aacef | 3395 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3396 | if (ret) { |
05135f59 | 3397 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3398 | goto fail_tree_roots; |
2b82032c | 3399 | } |
0b86a832 | 3400 | |
8dabb742 | 3401 | /* |
bacce86a AJ |
3402 | * At this point we know all the devices that make this filesystem, |
3403 | * including the seed devices but we don't know yet if the replace | |
3404 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3405 | * filesystem but skip the replace target device which is checked |
bacce86a | 3406 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3407 | */ |
bacce86a | 3408 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3409 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3410 | btrfs_err(fs_info, "failed to read devices"); |
4871c33b | 3411 | ret = -EIO; |
a6b0d5c8 CM |
3412 | goto fail_tree_roots; |
3413 | } | |
3414 | ||
b8522a1e | 3415 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3416 | if (ret) |
b8522a1e | 3417 | goto fail_tree_roots; |
8929ecfa | 3418 | |
73651042 NA |
3419 | /* |
3420 | * Get zone type information of zoned block devices. This will also | |
3421 | * handle emulation of a zoned filesystem if a regular device has the | |
3422 | * zoned incompat feature flag set. | |
3423 | */ | |
3424 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3425 | if (ret) { | |
3426 | btrfs_err(fs_info, | |
4871c33b | 3427 | "zoned: failed to read device zone info: %d", ret); |
73651042 NA |
3428 | goto fail_block_groups; |
3429 | } | |
3430 | ||
75ec1db8 JB |
3431 | /* |
3432 | * If we have a uuid root and we're not being told to rescan we need to | |
3433 | * check the generation here so we can set the | |
3434 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3435 | * transaction during a balance or the log replay without updating the | |
3436 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3437 | * even though it was perfectly fine. | |
3438 | */ | |
3439 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3440 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3441 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3442 | ||
cf90d884 QW |
3443 | ret = btrfs_verify_dev_extents(fs_info); |
3444 | if (ret) { | |
3445 | btrfs_err(fs_info, | |
3446 | "failed to verify dev extents against chunks: %d", | |
3447 | ret); | |
3448 | goto fail_block_groups; | |
3449 | } | |
68310a5e ID |
3450 | ret = btrfs_recover_balance(fs_info); |
3451 | if (ret) { | |
05135f59 | 3452 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3453 | goto fail_block_groups; |
3454 | } | |
3455 | ||
733f4fbb SB |
3456 | ret = btrfs_init_dev_stats(fs_info); |
3457 | if (ret) { | |
05135f59 | 3458 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3459 | goto fail_block_groups; |
3460 | } | |
3461 | ||
8dabb742 SB |
3462 | ret = btrfs_init_dev_replace(fs_info); |
3463 | if (ret) { | |
05135f59 | 3464 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3465 | goto fail_block_groups; |
3466 | } | |
3467 | ||
b70f5097 NA |
3468 | ret = btrfs_check_zoned_mode(fs_info); |
3469 | if (ret) { | |
3470 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3471 | ret); | |
3472 | goto fail_block_groups; | |
3473 | } | |
3474 | ||
c6761a9e | 3475 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3476 | if (ret) { |
05135f59 DS |
3477 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3478 | ret); | |
b7c35e81 AJ |
3479 | goto fail_block_groups; |
3480 | } | |
3481 | ||
96f3136e | 3482 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3483 | if (ret) { |
05135f59 | 3484 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3485 | goto fail_fsdev_sysfs; |
c59021f8 | 3486 | } |
3487 | ||
c59021f8 | 3488 | ret = btrfs_init_space_info(fs_info); |
3489 | if (ret) { | |
05135f59 | 3490 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3491 | goto fail_sysfs; |
c59021f8 | 3492 | } |
3493 | ||
5b4aacef | 3494 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3495 | if (ret) { |
05135f59 | 3496 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3497 | goto fail_sysfs; |
1b1d1f66 | 3498 | } |
4330e183 | 3499 | |
16beac87 NA |
3500 | btrfs_free_zone_cache(fs_info); |
3501 | ||
a7e1ac7b NA |
3502 | btrfs_check_active_zone_reservation(fs_info); |
3503 | ||
5c78a5e7 AJ |
3504 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3505 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3506 | btrfs_warn(fs_info, |
52042d8e | 3507 | "writable mount is not allowed due to too many missing devices"); |
4871c33b | 3508 | ret = -EINVAL; |
2365dd3c | 3509 | goto fail_sysfs; |
292fd7fc | 3510 | } |
9078a3e1 | 3511 | |
33c44184 | 3512 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3513 | "btrfs-cleaner"); |
4871c33b QW |
3514 | if (IS_ERR(fs_info->cleaner_kthread)) { |
3515 | ret = PTR_ERR(fs_info->cleaner_kthread); | |
2365dd3c | 3516 | goto fail_sysfs; |
4871c33b | 3517 | } |
a74a4b97 CM |
3518 | |
3519 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3520 | tree_root, | |
3521 | "btrfs-transaction"); | |
4871c33b QW |
3522 | if (IS_ERR(fs_info->transaction_kthread)) { |
3523 | ret = PTR_ERR(fs_info->transaction_kthread); | |
3f157a2f | 3524 | goto fail_cleaner; |
4871c33b | 3525 | } |
a74a4b97 | 3526 | |
bcef60f2 AJ |
3527 | ret = btrfs_read_qgroup_config(fs_info); |
3528 | if (ret) | |
3529 | goto fail_trans_kthread; | |
21adbd5c | 3530 | |
fd708b81 JB |
3531 | if (btrfs_build_ref_tree(fs_info)) |
3532 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3533 | ||
96da0919 QW |
3534 | /* do not make disk changes in broken FS or nologreplay is given */ |
3535 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3536 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3537 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3538 | ret = btrfs_replay_log(fs_info, fs_devices); |
4871c33b | 3539 | if (ret) |
28c16cbb | 3540 | goto fail_qgroup; |
e02119d5 | 3541 | } |
1a40e23b | 3542 | |
56e9357a | 3543 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 | 3544 | if (IS_ERR(fs_info->fs_root)) { |
4871c33b QW |
3545 | ret = PTR_ERR(fs_info->fs_root); |
3546 | btrfs_warn(fs_info, "failed to read fs tree: %d", ret); | |
315bf8ef | 3547 | fs_info->fs_root = NULL; |
bcef60f2 | 3548 | goto fail_qgroup; |
3140c9a3 | 3549 | } |
c289811c | 3550 | |
bc98a42c | 3551 | if (sb_rdonly(sb)) |
83e3a40a | 3552 | return 0; |
59641015 | 3553 | |
44c0ca21 | 3554 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3555 | if (ret) { |
6bccf3ab | 3556 | close_ctree(fs_info); |
2b6ba629 | 3557 | return ret; |
e3acc2a6 | 3558 | } |
b0643e59 | 3559 | btrfs_discard_resume(fs_info); |
b382a324 | 3560 | |
44c0ca21 BB |
3561 | if (fs_info->uuid_root && |
3562 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3563 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3564 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3565 | ret = btrfs_check_uuid_tree(fs_info); |
3566 | if (ret) { | |
05135f59 DS |
3567 | btrfs_warn(fs_info, |
3568 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3569 | close_ctree(fs_info); |
70f80175 SB |
3570 | return ret; |
3571 | } | |
f7a81ea4 | 3572 | } |
94846229 | 3573 | |
afcdd129 | 3574 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3575 | |
b4be6aef JB |
3576 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3577 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3578 | wake_up_process(fs_info->cleaner_kthread); | |
3579 | ||
ad2b2c80 | 3580 | return 0; |
39279cc3 | 3581 | |
bcef60f2 AJ |
3582 | fail_qgroup: |
3583 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3584 | fail_trans_kthread: |
3585 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3586 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3587 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3588 | fail_cleaner: |
a74a4b97 | 3589 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3590 | |
3591 | /* | |
3592 | * make sure we're done with the btree inode before we stop our | |
3593 | * kthreads | |
3594 | */ | |
3595 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3596 | |
2365dd3c | 3597 | fail_sysfs: |
6618a59b | 3598 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3599 | |
b7c35e81 AJ |
3600 | fail_fsdev_sysfs: |
3601 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3602 | ||
1b1d1f66 | 3603 | fail_block_groups: |
54067ae9 | 3604 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3605 | |
3606 | fail_tree_roots: | |
9e3aa805 JB |
3607 | if (fs_info->data_reloc_root) |
3608 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3609 | free_root_pointers(fs_info, true); |
2b8195bb | 3610 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3611 | |
39279cc3 | 3612 | fail_sb_buffer: |
7abadb64 | 3613 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3614 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3615 | fail_alloc: |
7dc66abb | 3616 | btrfs_mapping_tree_free(fs_info); |
586e46e2 | 3617 | |
4543df7e | 3618 | iput(fs_info->btree_inode); |
7e662854 | 3619 | fail: |
586e46e2 | 3620 | btrfs_close_devices(fs_info->fs_devices); |
4871c33b QW |
3621 | ASSERT(ret < 0); |
3622 | return ret; | |
eb60ceac | 3623 | } |
663faf9f | 3624 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3625 | |
314b6dd0 | 3626 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3627 | { |
314b6dd0 | 3628 | struct btrfs_device *device = bio->bi_private; |
617fb10e | 3629 | struct folio_iter fi; |
314b6dd0 | 3630 | |
617fb10e | 3631 | bio_for_each_folio_all(fi, bio) { |
314b6dd0 JT |
3632 | if (bio->bi_status) { |
3633 | btrfs_warn_rl_in_rcu(device->fs_info, | |
617fb10e | 3634 | "lost super block write due to IO error on %s (%d)", |
cb3e217b | 3635 | btrfs_dev_name(device), |
314b6dd0 | 3636 | blk_status_to_errno(bio->bi_status)); |
314b6dd0 JT |
3637 | btrfs_dev_stat_inc_and_print(device, |
3638 | BTRFS_DEV_STAT_WRITE_ERRS); | |
bc00965d MWO |
3639 | /* Ensure failure if the primary sb fails. */ |
3640 | if (bio->bi_opf & REQ_FUA) | |
3641 | atomic_add(BTRFS_SUPER_PRIMARY_WRITE_ERROR, | |
3642 | &device->sb_write_errors); | |
3643 | else | |
3644 | atomic_inc(&device->sb_write_errors); | |
314b6dd0 | 3645 | } |
617fb10e MWO |
3646 | folio_unlock(fi.folio); |
3647 | folio_put(fi.folio); | |
f2984462 | 3648 | } |
314b6dd0 JT |
3649 | |
3650 | bio_put(bio); | |
f2984462 CM |
3651 | } |
3652 | ||
8f32380d | 3653 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3654 | int copy_num, bool drop_cache) |
29c36d72 | 3655 | { |
29c36d72 | 3656 | struct btrfs_super_block *super; |
8f32380d | 3657 | struct page *page; |
12659251 | 3658 | u64 bytenr, bytenr_orig; |
224941e8 | 3659 | struct address_space *mapping = bdev->bd_mapping; |
12659251 NA |
3660 | int ret; |
3661 | ||
3662 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3663 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3664 | if (ret == -ENOENT) | |
3665 | return ERR_PTR(-EINVAL); | |
3666 | else if (ret) | |
3667 | return ERR_PTR(ret); | |
29c36d72 | 3668 | |
cda00eba | 3669 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3670 | return ERR_PTR(-EINVAL); |
29c36d72 | 3671 | |
a05d3c91 QW |
3672 | if (drop_cache) { |
3673 | /* This should only be called with the primary sb. */ | |
3674 | ASSERT(copy_num == 0); | |
3675 | ||
3676 | /* | |
3677 | * Drop the page of the primary superblock, so later read will | |
3678 | * always read from the device. | |
3679 | */ | |
3680 | invalidate_inode_pages2_range(mapping, | |
3681 | bytenr >> PAGE_SHIFT, | |
3682 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3683 | } | |
3684 | ||
8f32380d JT |
3685 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3686 | if (IS_ERR(page)) | |
3687 | return ERR_CAST(page); | |
29c36d72 | 3688 | |
8f32380d | 3689 | super = page_address(page); |
96c2e067 AJ |
3690 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3691 | btrfs_release_disk_super(super); | |
3692 | return ERR_PTR(-ENODATA); | |
3693 | } | |
3694 | ||
12659251 | 3695 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3696 | btrfs_release_disk_super(super); |
3697 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3698 | } |
3699 | ||
8f32380d | 3700 | return super; |
29c36d72 AJ |
3701 | } |
3702 | ||
3703 | ||
8f32380d | 3704 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3705 | { |
8f32380d | 3706 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3707 | int i; |
3708 | u64 transid = 0; | |
a512bbf8 YZ |
3709 | |
3710 | /* we would like to check all the supers, but that would make | |
3711 | * a btrfs mount succeed after a mkfs from a different FS. | |
3712 | * So, we need to add a special mount option to scan for | |
3713 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3714 | */ | |
3715 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3716 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3717 | if (IS_ERR(super)) |
a512bbf8 YZ |
3718 | continue; |
3719 | ||
a512bbf8 | 3720 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3721 | if (latest) |
3722 | btrfs_release_disk_super(super); | |
3723 | ||
3724 | latest = super; | |
a512bbf8 | 3725 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3726 | } |
3727 | } | |
92fc03fb | 3728 | |
8f32380d | 3729 | return super; |
a512bbf8 YZ |
3730 | } |
3731 | ||
4eedeb75 | 3732 | /* |
abbb3b8e | 3733 | * Write superblock @sb to the @device. Do not wait for completion, all the |
f93ee0df | 3734 | * folios we use for writing are locked. |
4eedeb75 | 3735 | * |
abbb3b8e DS |
3736 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3737 | * the expected device size at commit time. Note that max_mirrors must be | |
3738 | * same for write and wait phases. | |
4eedeb75 | 3739 | * |
f93ee0df | 3740 | * Return number of errors when folio is not found or submission fails. |
4eedeb75 | 3741 | */ |
a512bbf8 | 3742 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3743 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3744 | { |
d5178578 | 3745 | struct btrfs_fs_info *fs_info = device->fs_info; |
224941e8 | 3746 | struct address_space *mapping = device->bdev->bd_mapping; |
d5178578 | 3747 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3748 | int i; |
12659251 NA |
3749 | int ret; |
3750 | u64 bytenr, bytenr_orig; | |
a512bbf8 | 3751 | |
bc00965d MWO |
3752 | atomic_set(&device->sb_write_errors, 0); |
3753 | ||
a512bbf8 YZ |
3754 | if (max_mirrors == 0) |
3755 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3756 | ||
d5178578 JT |
3757 | shash->tfm = fs_info->csum_shash; |
3758 | ||
a512bbf8 | 3759 | for (i = 0; i < max_mirrors; i++) { |
f93ee0df | 3760 | struct folio *folio; |
314b6dd0 JT |
3761 | struct bio *bio; |
3762 | struct btrfs_super_block *disk_super; | |
f93ee0df | 3763 | size_t offset; |
314b6dd0 | 3764 | |
12659251 NA |
3765 | bytenr_orig = btrfs_sb_offset(i); |
3766 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3767 | if (ret == -ENOENT) { | |
3768 | continue; | |
3769 | } else if (ret < 0) { | |
3770 | btrfs_err(device->fs_info, | |
3771 | "couldn't get super block location for mirror %d", | |
3772 | i); | |
bc00965d | 3773 | atomic_inc(&device->sb_write_errors); |
12659251 NA |
3774 | continue; |
3775 | } | |
935e5cc9 MX |
3776 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3777 | device->commit_total_bytes) | |
a512bbf8 YZ |
3778 | break; |
3779 | ||
12659251 | 3780 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3781 | |
fd08001f EB |
3782 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3783 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3784 | sb->csum); | |
4eedeb75 | 3785 | |
f93ee0df MWO |
3786 | folio = __filemap_get_folio(mapping, bytenr >> PAGE_SHIFT, |
3787 | FGP_LOCK | FGP_ACCESSED | FGP_CREAT, | |
3788 | GFP_NOFS); | |
3789 | if (IS_ERR(folio)) { | |
abbb3b8e | 3790 | btrfs_err(device->fs_info, |
314b6dd0 | 3791 | "couldn't get super block page for bytenr %llu", |
abbb3b8e | 3792 | bytenr); |
bc00965d | 3793 | atomic_inc(&device->sb_write_errors); |
4eedeb75 | 3794 | continue; |
abbb3b8e | 3795 | } |
f93ee0df | 3796 | ASSERT(folio_order(folio) == 0); |
634554dc | 3797 | |
f93ee0df MWO |
3798 | offset = offset_in_folio(folio, bytenr); |
3799 | disk_super = folio_address(folio) + offset; | |
314b6dd0 | 3800 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); |
4eedeb75 | 3801 | |
314b6dd0 JT |
3802 | /* |
3803 | * Directly use bios here instead of relying on the page cache | |
3804 | * to do I/O, so we don't lose the ability to do integrity | |
3805 | * checking. | |
3806 | */ | |
07888c66 CH |
3807 | bio = bio_alloc(device->bdev, 1, |
3808 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
3809 | GFP_NOFS); | |
314b6dd0 JT |
3810 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
3811 | bio->bi_private = device; | |
3812 | bio->bi_end_io = btrfs_end_super_write; | |
f93ee0df | 3813 | bio_add_folio_nofail(bio, folio, BTRFS_SUPER_INFO_SIZE, offset); |
a512bbf8 | 3814 | |
387125fc | 3815 | /* |
314b6dd0 JT |
3816 | * We FUA only the first super block. The others we allow to |
3817 | * go down lazy and there's a short window where the on-disk | |
3818 | * copies might still contain the older version. | |
387125fc | 3819 | */ |
1b9e619c | 3820 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 | 3821 | bio->bi_opf |= REQ_FUA; |
58ff51f1 | 3822 | submit_bio(bio); |
8376d9e1 NA |
3823 | |
3824 | if (btrfs_advance_sb_log(device, i)) | |
bc00965d | 3825 | atomic_inc(&device->sb_write_errors); |
a512bbf8 | 3826 | } |
bc00965d | 3827 | return atomic_read(&device->sb_write_errors) < i ? 0 : -1; |
a512bbf8 YZ |
3828 | } |
3829 | ||
abbb3b8e DS |
3830 | /* |
3831 | * Wait for write completion of superblocks done by write_dev_supers, | |
3832 | * @max_mirrors same for write and wait phases. | |
3833 | * | |
bc00965d MWO |
3834 | * Return -1 if primary super block write failed or when there were no super block |
3835 | * copies written. Otherwise 0. | |
abbb3b8e DS |
3836 | */ |
3837 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3838 | { | |
abbb3b8e DS |
3839 | int i; |
3840 | int errors = 0; | |
b6a535fa | 3841 | bool primary_failed = false; |
12659251 | 3842 | int ret; |
abbb3b8e DS |
3843 | u64 bytenr; |
3844 | ||
3845 | if (max_mirrors == 0) | |
3846 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3847 | ||
3848 | for (i = 0; i < max_mirrors; i++) { | |
c94b7349 | 3849 | struct folio *folio; |
314b6dd0 | 3850 | |
12659251 NA |
3851 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3852 | if (ret == -ENOENT) { | |
3853 | break; | |
3854 | } else if (ret < 0) { | |
3855 | errors++; | |
3856 | if (i == 0) | |
3857 | primary_failed = true; | |
3858 | continue; | |
3859 | } | |
abbb3b8e DS |
3860 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3861 | device->commit_total_bytes) | |
3862 | break; | |
3863 | ||
38da32ee | 3864 | folio = filemap_get_folio(device->bdev->bd_mapping, |
c94b7349 | 3865 | bytenr >> PAGE_SHIFT); |
bc00965d MWO |
3866 | /* If the folio has been removed, then we know it completed. */ |
3867 | if (IS_ERR(folio)) | |
abbb3b8e | 3868 | continue; |
c94b7349 MWO |
3869 | ASSERT(folio_order(folio) == 0); |
3870 | ||
3871 | /* Folio will be unlocked once the write completes. */ | |
3872 | folio_wait_locked(folio); | |
c94b7349 | 3873 | folio_put(folio); |
abbb3b8e DS |
3874 | } |
3875 | ||
bc00965d MWO |
3876 | errors += atomic_read(&device->sb_write_errors); |
3877 | if (errors >= BTRFS_SUPER_PRIMARY_WRITE_ERROR) | |
3878 | primary_failed = true; | |
b6a535fa HM |
3879 | if (primary_failed) { |
3880 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3881 | device->devid); | |
3882 | return -1; | |
3883 | } | |
3884 | ||
abbb3b8e DS |
3885 | return errors < i ? 0 : -1; |
3886 | } | |
3887 | ||
387125fc CM |
3888 | /* |
3889 | * endio for the write_dev_flush, this will wake anyone waiting | |
3890 | * for the barrier when it is done | |
3891 | */ | |
4246a0b6 | 3892 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3893 | { |
f9e69aa9 | 3894 | bio_uninit(bio); |
e0ae9994 | 3895 | complete(bio->bi_private); |
387125fc CM |
3896 | } |
3897 | ||
3898 | /* | |
4fc6441a AJ |
3899 | * Submit a flush request to the device if it supports it. Error handling is |
3900 | * done in the waiting counterpart. | |
387125fc | 3901 | */ |
4fc6441a | 3902 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3903 | { |
f9e69aa9 | 3904 | struct bio *bio = &device->flush_bio; |
387125fc | 3905 | |
bfd3ea94 AJ |
3906 | device->last_flush_error = BLK_STS_OK; |
3907 | ||
f9e69aa9 CH |
3908 | bio_init(bio, device->bdev, NULL, 0, |
3909 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 3910 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
3911 | init_completion(&device->flush_wait); |
3912 | bio->bi_private = &device->flush_wait; | |
58ff51f1 | 3913 | submit_bio(bio); |
1c3063b6 | 3914 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3915 | } |
387125fc | 3916 | |
4fc6441a AJ |
3917 | /* |
3918 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
1b465784 | 3919 | * Return true for any error, and false otherwise. |
4fc6441a | 3920 | */ |
1b465784 | 3921 | static bool wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3922 | { |
f9e69aa9 | 3923 | struct bio *bio = &device->flush_bio; |
387125fc | 3924 | |
7e812f20 | 3925 | if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
1b465784 | 3926 | return false; |
387125fc | 3927 | |
2980d574 | 3928 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3929 | |
bfd3ea94 AJ |
3930 | if (bio->bi_status) { |
3931 | device->last_flush_error = bio->bi_status; | |
3932 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); | |
1b465784 | 3933 | return true; |
bfd3ea94 AJ |
3934 | } |
3935 | ||
1b465784 | 3936 | return false; |
387125fc | 3937 | } |
387125fc | 3938 | |
387125fc CM |
3939 | /* |
3940 | * send an empty flush down to each device in parallel, | |
3941 | * then wait for them | |
3942 | */ | |
3943 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3944 | { | |
3945 | struct list_head *head; | |
3946 | struct btrfs_device *dev; | |
5af3e8cc | 3947 | int errors_wait = 0; |
387125fc | 3948 | |
1538e6c5 | 3949 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3950 | /* send down all the barriers */ |
3951 | head = &info->fs_devices->devices; | |
1538e6c5 | 3952 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3953 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3954 | continue; |
cea7c8bf | 3955 | if (!dev->bdev) |
387125fc | 3956 | continue; |
e12c9621 | 3957 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3958 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3959 | continue; |
3960 | ||
4fc6441a | 3961 | write_dev_flush(dev); |
387125fc CM |
3962 | } |
3963 | ||
3964 | /* wait for all the barriers */ | |
1538e6c5 | 3965 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3966 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3967 | continue; |
387125fc | 3968 | if (!dev->bdev) { |
5af3e8cc | 3969 | errors_wait++; |
387125fc CM |
3970 | continue; |
3971 | } | |
e12c9621 | 3972 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3973 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3974 | continue; |
3975 | ||
1b465784 | 3976 | if (wait_dev_flush(dev)) |
5af3e8cc | 3977 | errors_wait++; |
401b41e5 AJ |
3978 | } |
3979 | ||
de38a206 AJ |
3980 | /* |
3981 | * Checks last_flush_error of disks in order to determine the device | |
3982 | * state. | |
3983 | */ | |
3984 | if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) | |
3985 | return -EIO; | |
3986 | ||
387125fc CM |
3987 | return 0; |
3988 | } | |
3989 | ||
943c6e99 ZL |
3990 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3991 | { | |
8789f4fe ZL |
3992 | int raid_type; |
3993 | int min_tolerated = INT_MAX; | |
943c6e99 | 3994 | |
8789f4fe ZL |
3995 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3996 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 3997 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
3998 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
3999 | tolerated_failures); | |
943c6e99 | 4000 | |
8789f4fe ZL |
4001 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4002 | if (raid_type == BTRFS_RAID_SINGLE) | |
4003 | continue; | |
41a6e891 | 4004 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4005 | continue; |
8c3e3582 | 4006 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4007 | btrfs_raid_array[raid_type]. |
4008 | tolerated_failures); | |
4009 | } | |
943c6e99 | 4010 | |
8789f4fe | 4011 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4012 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4013 | min_tolerated = 0; |
4014 | } | |
4015 | ||
4016 | return min_tolerated; | |
943c6e99 ZL |
4017 | } |
4018 | ||
eece6a9c | 4019 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4020 | { |
e5e9a520 | 4021 | struct list_head *head; |
f2984462 | 4022 | struct btrfs_device *dev; |
a061fc8d | 4023 | struct btrfs_super_block *sb; |
f2984462 | 4024 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4025 | int ret; |
4026 | int do_barriers; | |
a236aed1 CM |
4027 | int max_errors; |
4028 | int total_errors = 0; | |
a061fc8d | 4029 | u64 flags; |
f2984462 | 4030 | |
0b246afa | 4031 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4032 | |
4033 | /* | |
4034 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4035 | * not from fsync where the tree roots in fs_info have not | |
4036 | * been consistent on disk. | |
4037 | */ | |
4038 | if (max_mirrors == 0) | |
4039 | backup_super_roots(fs_info); | |
f2984462 | 4040 | |
0b246afa | 4041 | sb = fs_info->super_for_commit; |
a061fc8d | 4042 | dev_item = &sb->dev_item; |
e5e9a520 | 4043 | |
0b246afa JM |
4044 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4045 | head = &fs_info->fs_devices->devices; | |
4046 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4047 | |
5af3e8cc | 4048 | if (do_barriers) { |
0b246afa | 4049 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4050 | if (ret) { |
4051 | mutex_unlock( | |
0b246afa JM |
4052 | &fs_info->fs_devices->device_list_mutex); |
4053 | btrfs_handle_fs_error(fs_info, ret, | |
4054 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4055 | return ret; |
4056 | } | |
4057 | } | |
387125fc | 4058 | |
1538e6c5 | 4059 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4060 | if (!dev->bdev) { |
4061 | total_errors++; | |
4062 | continue; | |
4063 | } | |
e12c9621 | 4064 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4065 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4066 | continue; |
4067 | ||
2b82032c | 4068 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4069 | btrfs_set_stack_device_type(dev_item, dev->type); |
4070 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4071 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4072 | dev->commit_total_bytes); |
ce7213c7 MX |
4073 | btrfs_set_stack_device_bytes_used(dev_item, |
4074 | dev->commit_bytes_used); | |
a061fc8d CM |
4075 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4076 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4077 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4078 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4079 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4080 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4081 | |
a061fc8d CM |
4082 | flags = btrfs_super_flags(sb); |
4083 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4084 | ||
75cb857d QW |
4085 | ret = btrfs_validate_write_super(fs_info, sb); |
4086 | if (ret < 0) { | |
4087 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4088 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4089 | "unexpected superblock corruption detected"); | |
4090 | return -EUCLEAN; | |
4091 | } | |
4092 | ||
abbb3b8e | 4093 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4094 | if (ret) |
4095 | total_errors++; | |
f2984462 | 4096 | } |
a236aed1 | 4097 | if (total_errors > max_errors) { |
0b246afa JM |
4098 | btrfs_err(fs_info, "%d errors while writing supers", |
4099 | total_errors); | |
4100 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4101 | |
9d565ba4 | 4102 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4103 | btrfs_handle_fs_error(fs_info, -EIO, |
4104 | "%d errors while writing supers", | |
4105 | total_errors); | |
9d565ba4 | 4106 | return -EIO; |
a236aed1 | 4107 | } |
f2984462 | 4108 | |
a512bbf8 | 4109 | total_errors = 0; |
1538e6c5 | 4110 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4111 | if (!dev->bdev) |
4112 | continue; | |
e12c9621 | 4113 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4114 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4115 | continue; |
4116 | ||
abbb3b8e | 4117 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4118 | if (ret) |
4119 | total_errors++; | |
f2984462 | 4120 | } |
0b246afa | 4121 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4122 | if (total_errors > max_errors) { |
0b246afa JM |
4123 | btrfs_handle_fs_error(fs_info, -EIO, |
4124 | "%d errors while writing supers", | |
4125 | total_errors); | |
79787eaa | 4126 | return -EIO; |
a236aed1 | 4127 | } |
f2984462 CM |
4128 | return 0; |
4129 | } | |
4130 | ||
cb517eab MX |
4131 | /* Drop a fs root from the radix tree and free it. */ |
4132 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4133 | struct btrfs_root *root) | |
2619ba1f | 4134 | { |
4785e24f JB |
4135 | bool drop_ref = false; |
4136 | ||
fc7cbcd4 DS |
4137 | spin_lock(&fs_info->fs_roots_radix_lock); |
4138 | radix_tree_delete(&fs_info->fs_roots_radix, | |
e094f480 | 4139 | (unsigned long)btrfs_root_id(root)); |
fc7cbcd4 | 4140 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
4785e24f | 4141 | drop_ref = true; |
fc7cbcd4 | 4142 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4143 | |
84961539 | 4144 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4145 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4146 | if (root->reloc_root) { |
00246528 | 4147 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4148 | root->reloc_root = NULL; |
4149 | } | |
4150 | } | |
3321719e | 4151 | |
4785e24f JB |
4152 | if (drop_ref) |
4153 | btrfs_put_root(root); | |
2619ba1f CM |
4154 | } |
4155 | ||
6bccf3ab | 4156 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4157 | { |
6bccf3ab | 4158 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4159 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4160 | |
0b246afa | 4161 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4162 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4163 | mutex_unlock(&fs_info->cleaner_mutex); |
4164 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4165 | |
4166 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4167 | down_write(&fs_info->cleanup_work_sem); |
4168 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4169 | |
7a7eaa40 | 4170 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4171 | if (IS_ERR(trans)) |
4172 | return PTR_ERR(trans); | |
3a45bb20 | 4173 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4174 | } |
4175 | ||
36c86a9e QW |
4176 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4177 | { | |
4178 | struct btrfs_transaction *trans; | |
4179 | struct btrfs_transaction *tmp; | |
4180 | bool found = false; | |
4181 | ||
36c86a9e QW |
4182 | /* |
4183 | * This function is only called at the very end of close_ctree(), | |
4184 | * thus no other running transaction, no need to take trans_lock. | |
4185 | */ | |
4186 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4187 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4188 | struct extent_state *cached = NULL; | |
4189 | u64 dirty_bytes = 0; | |
4190 | u64 cur = 0; | |
4191 | u64 found_start; | |
4192 | u64 found_end; | |
4193 | ||
4194 | found = true; | |
e5860f82 | 4195 | while (find_first_extent_bit(&trans->dirty_pages, cur, |
36c86a9e QW |
4196 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { |
4197 | dirty_bytes += found_end + 1 - found_start; | |
4198 | cur = found_end + 1; | |
4199 | } | |
4200 | btrfs_warn(fs_info, | |
4201 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4202 | trans->transid, dirty_bytes); | |
4203 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4204 | ||
4205 | if (trans == fs_info->running_transaction) | |
4206 | fs_info->running_transaction = NULL; | |
4207 | list_del_init(&trans->list); | |
4208 | ||
4209 | btrfs_put_transaction(trans); | |
4210 | trace_btrfs_transaction_commit(fs_info); | |
4211 | } | |
4212 | ASSERT(!found); | |
4213 | } | |
4214 | ||
b105e927 | 4215 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4216 | { |
c146afad YZ |
4217 | int ret; |
4218 | ||
afcdd129 | 4219 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4220 | |
8a1f1e3d FM |
4221 | /* |
4222 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4223 | * clear that bit and wake up relocation so it can stop. | |
4224 | * We must do this before stopping the block group reclaim task, because | |
4225 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4226 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4227 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4228 | * return 1. | |
4229 | */ | |
4230 | btrfs_wake_unfinished_drop(fs_info); | |
4231 | ||
31e70e52 FM |
4232 | /* |
4233 | * We may have the reclaim task running and relocating a data block group, | |
4234 | * in which case it may create delayed iputs. So stop it before we park | |
4235 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4236 | * parking the cleaner, and that can make the async reclaim task to hang | |
4237 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4238 | * parked and can not run delayed iputs - this will make us hang when | |
4239 | * trying to stop the async reclaim task. | |
4240 | */ | |
4241 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4242 | /* |
4243 | * We don't want the cleaner to start new transactions, add more delayed | |
4244 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4245 | * because that frees the task_struct, and the transaction kthread might | |
4246 | * still try to wake up the cleaner. | |
4247 | */ | |
4248 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4249 | |
7343dd61 | 4250 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4251 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4252 | |
803b2f54 SB |
4253 | /* wait for the uuid_scan task to finish */ |
4254 | down(&fs_info->uuid_tree_rescan_sem); | |
4255 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4256 | up(&fs_info->uuid_tree_rescan_sem); | |
4257 | ||
837d5b6e | 4258 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4259 | btrfs_pause_balance(fs_info); |
837d5b6e | 4260 | |
8dabb742 SB |
4261 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4262 | ||
aa1b8cd4 | 4263 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4264 | |
4265 | /* wait for any defraggers to finish */ | |
4266 | wait_event(fs_info->transaction_wait, | |
4267 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4268 | ||
4269 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4270 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4271 | |
a362bb86 FM |
4272 | /* |
4273 | * After we parked the cleaner kthread, ordered extents may have | |
4274 | * completed and created new delayed iputs. If one of the async reclaim | |
4275 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4276 | * can hang forever trying to stop it, because if a delayed iput is | |
4277 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4278 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4279 | * no one else to run iputs. | |
4280 | * | |
4281 | * So wait for all ongoing ordered extents to complete and then run | |
4282 | * delayed iputs. This works because once we reach this point no one | |
4283 | * can either create new ordered extents nor create delayed iputs | |
4284 | * through some other means. | |
4285 | * | |
4286 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4287 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4288 | * but the delayed iput for the respective inode is made only when doing | |
4289 | * the final btrfs_put_ordered_extent() (which must happen at | |
4290 | * btrfs_finish_ordered_io() when we are unmounting). | |
4291 | */ | |
4292 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4293 | /* Ordered extents for free space inodes. */ | |
4294 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4295 | btrfs_run_delayed_iputs(fs_info); | |
4296 | ||
21c7e756 | 4297 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4298 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4299 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4300 | |
b0643e59 DZ |
4301 | /* Cancel or finish ongoing discard work */ |
4302 | btrfs_discard_cleanup(fs_info); | |
4303 | ||
bc98a42c | 4304 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4305 | /* |
d6fd0ae2 OS |
4306 | * The cleaner kthread is stopped, so do one final pass over |
4307 | * unused block groups. | |
e44163e1 | 4308 | */ |
0b246afa | 4309 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4310 | |
f0cc2cd7 FM |
4311 | /* |
4312 | * There might be existing delayed inode workers still running | |
4313 | * and holding an empty delayed inode item. We must wait for | |
4314 | * them to complete first because they can create a transaction. | |
4315 | * This happens when someone calls btrfs_balance_delayed_items() | |
4316 | * and then a transaction commit runs the same delayed nodes | |
4317 | * before any delayed worker has done something with the nodes. | |
4318 | * We must wait for any worker here and not at transaction | |
4319 | * commit time since that could cause a deadlock. | |
4320 | * This is a very rare case. | |
4321 | */ | |
4322 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4323 | ||
6bccf3ab | 4324 | ret = btrfs_commit_super(fs_info); |
acce952b | 4325 | if (ret) |
04892340 | 4326 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4327 | } |
4328 | ||
84961539 | 4329 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4330 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4331 | |
e3029d9f AV |
4332 | kthread_stop(fs_info->transaction_kthread); |
4333 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4334 | |
e187831e | 4335 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4336 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4337 | |
5958253c QW |
4338 | if (btrfs_check_quota_leak(fs_info)) { |
4339 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4340 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4341 | } | |
4342 | ||
04892340 | 4343 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4344 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4345 | |
963d678b | 4346 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4347 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4348 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4349 | } |
bcc63abb | 4350 | |
5deb17e1 | 4351 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4352 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4353 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4354 | |
6618a59b | 4355 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4356 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4357 | |
1a4319cc LB |
4358 | btrfs_put_block_group_cache(fs_info); |
4359 | ||
de348ee0 WS |
4360 | /* |
4361 | * we must make sure there is not any read request to | |
4362 | * submit after we stopping all workers. | |
4363 | */ | |
4364 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4365 | btrfs_stop_all_workers(fs_info); |
4366 | ||
0a31daa4 | 4367 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4368 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4369 | |
afcdd129 | 4370 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4371 | free_root_pointers(fs_info, true); |
8c38938c | 4372 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4373 | |
4e19443d JB |
4374 | /* |
4375 | * We must free the block groups after dropping the fs_roots as we could | |
4376 | * have had an IO error and have left over tree log blocks that aren't | |
4377 | * cleaned up until the fs roots are freed. This makes the block group | |
4378 | * accounting appear to be wrong because there's pending reserved bytes, | |
4379 | * so make sure we do the block group cleanup afterwards. | |
4380 | */ | |
4381 | btrfs_free_block_groups(fs_info); | |
4382 | ||
13e6c37b | 4383 | iput(fs_info->btree_inode); |
d6bfde87 | 4384 | |
7dc66abb | 4385 | btrfs_mapping_tree_free(fs_info); |
68c94e55 | 4386 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4387 | } |
4388 | ||
50564b65 FM |
4389 | void btrfs_mark_buffer_dirty(struct btrfs_trans_handle *trans, |
4390 | struct extent_buffer *buf) | |
5f39d397 | 4391 | { |
2f4d60df | 4392 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4393 | u64 transid = btrfs_header_generation(buf); |
b4ce94de | 4394 | |
06ea65a3 JB |
4395 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4396 | /* | |
4397 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4398 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4399 | * outside of the sanity tests. |
4400 | */ | |
b0132a3b | 4401 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4402 | return; |
4403 | #endif | |
50564b65 FM |
4404 | /* This is an active transaction (its state < TRANS_STATE_UNBLOCKED). */ |
4405 | ASSERT(trans->transid == fs_info->generation); | |
49d0c642 | 4406 | btrfs_assert_tree_write_locked(buf); |
4ebe8d47 | 4407 | if (unlikely(transid != fs_info->generation)) { |
50564b65 | 4408 | btrfs_abort_transaction(trans, -EUCLEAN); |
20cbe460 FM |
4409 | btrfs_crit(fs_info, |
4410 | "dirty buffer transid mismatch, logical %llu found transid %llu running transid %llu", | |
4411 | buf->start, transid, fs_info->generation); | |
50564b65 | 4412 | } |
f18cc978 | 4413 | set_extent_buffer_dirty(buf); |
eb60ceac CM |
4414 | } |
4415 | ||
2ff7e61e | 4416 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4417 | int flush_delayed) |
16cdcec7 MX |
4418 | { |
4419 | /* | |
4420 | * looks as though older kernels can get into trouble with | |
4421 | * this code, they end up stuck in balance_dirty_pages forever | |
4422 | */ | |
e2d84521 | 4423 | int ret; |
16cdcec7 MX |
4424 | |
4425 | if (current->flags & PF_MEMALLOC) | |
4426 | return; | |
4427 | ||
b53d3f5d | 4428 | if (flush_delayed) |
2ff7e61e | 4429 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4430 | |
d814a491 EL |
4431 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4432 | BTRFS_DIRTY_METADATA_THRESH, | |
4433 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4434 | if (ret > 0) { |
0b246afa | 4435 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4436 | } |
16cdcec7 MX |
4437 | } |
4438 | ||
2ff7e61e | 4439 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4440 | { |
2ff7e61e | 4441 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4442 | } |
585ad2c3 | 4443 | |
2ff7e61e | 4444 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4445 | { |
2ff7e61e | 4446 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4447 | } |
6b80053d | 4448 | |
2ff7e61e | 4449 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4450 | { |
fe816d0f NB |
4451 | /* cleanup FS via transaction */ |
4452 | btrfs_cleanup_transaction(fs_info); | |
4453 | ||
0b246afa | 4454 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4455 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4456 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4457 | |
0b246afa JM |
4458 | down_write(&fs_info->cleanup_work_sem); |
4459 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4460 | } |
4461 | ||
ef67963d JB |
4462 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4463 | { | |
fc7cbcd4 DS |
4464 | struct btrfs_root *gang[8]; |
4465 | u64 root_objectid = 0; | |
4466 | int ret; | |
4467 | ||
4468 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4469 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4470 | (void **)gang, root_objectid, | |
4471 | ARRAY_SIZE(gang))) != 0) { | |
4472 | int i; | |
4473 | ||
4474 | for (i = 0; i < ret; i++) | |
4475 | gang[i] = btrfs_grab_root(gang[i]); | |
4476 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4477 | ||
4478 | for (i = 0; i < ret; i++) { | |
4479 | if (!gang[i]) | |
ef67963d | 4480 | continue; |
e094f480 | 4481 | root_objectid = btrfs_root_id(gang[i]); |
fc7cbcd4 DS |
4482 | btrfs_free_log(NULL, gang[i]); |
4483 | btrfs_put_root(gang[i]); | |
ef67963d | 4484 | } |
fc7cbcd4 DS |
4485 | root_objectid++; |
4486 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4487 | } |
fc7cbcd4 | 4488 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4489 | btrfs_free_log_root_tree(NULL, fs_info); |
4490 | } | |
4491 | ||
143bede5 | 4492 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4493 | { |
acce952b | 4494 | struct btrfs_ordered_extent *ordered; |
acce952b | 4495 | |
199c2a9c | 4496 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4497 | /* |
4498 | * This will just short circuit the ordered completion stuff which will | |
4499 | * make sure the ordered extent gets properly cleaned up. | |
4500 | */ | |
199c2a9c | 4501 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4502 | root_extent_list) |
4503 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4504 | spin_unlock(&root->ordered_extent_lock); |
4505 | } | |
4506 | ||
4507 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4508 | { | |
4509 | struct btrfs_root *root; | |
84af994b | 4510 | LIST_HEAD(splice); |
199c2a9c MX |
4511 | |
4512 | spin_lock(&fs_info->ordered_root_lock); | |
4513 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4514 | while (!list_empty(&splice)) { | |
4515 | root = list_first_entry(&splice, struct btrfs_root, | |
4516 | ordered_root); | |
1de2cfde JB |
4517 | list_move_tail(&root->ordered_root, |
4518 | &fs_info->ordered_roots); | |
199c2a9c | 4519 | |
2a85d9ca | 4520 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4521 | btrfs_destroy_ordered_extents(root); |
4522 | ||
2a85d9ca LB |
4523 | cond_resched(); |
4524 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4525 | } |
4526 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4527 | |
4528 | /* | |
4529 | * We need this here because if we've been flipped read-only we won't | |
4530 | * get sync() from the umount, so we need to make sure any ordered | |
4531 | * extents that haven't had their dirty pages IO start writeout yet | |
4532 | * actually get run and error out properly. | |
4533 | */ | |
4534 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4535 | } |
4536 | ||
99f09ce3 FM |
4537 | static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4538 | struct btrfs_fs_info *fs_info) | |
acce952b | 4539 | { |
4540 | struct rb_node *node; | |
4541 | struct btrfs_delayed_ref_root *delayed_refs; | |
4542 | struct btrfs_delayed_ref_node *ref; | |
acce952b | 4543 | |
4544 | delayed_refs = &trans->delayed_refs; | |
4545 | ||
4546 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4547 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4548 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4549 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
99f09ce3 | 4550 | return; |
acce952b | 4551 | } |
4552 | ||
5c9d028b | 4553 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4554 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4555 | struct rb_node *n; |
e78417d1 | 4556 | bool pin_bytes = false; |
acce952b | 4557 | |
d7df2c79 JB |
4558 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4559 | href_node); | |
3069bd26 | 4560 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4561 | continue; |
3069bd26 | 4562 | |
d7df2c79 | 4563 | spin_lock(&head->lock); |
e3d03965 | 4564 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4565 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4566 | ref_node); | |
e3d03965 | 4567 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4568 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4569 | if (!list_empty(&ref->add_list)) |
4570 | list_del(&ref->add_list); | |
d7df2c79 JB |
4571 | atomic_dec(&delayed_refs->num_entries); |
4572 | btrfs_put_delayed_ref(ref); | |
adb86dbe | 4573 | btrfs_delayed_refs_rsv_release(fs_info, 1, 0); |
e78417d1 | 4574 | } |
d7df2c79 JB |
4575 | if (head->must_insert_reserved) |
4576 | pin_bytes = true; | |
4577 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4578 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4579 | spin_unlock(&head->lock); |
4580 | spin_unlock(&delayed_refs->lock); | |
4581 | mutex_unlock(&head->mutex); | |
acce952b | 4582 | |
f603bb94 NB |
4583 | if (pin_bytes) { |
4584 | struct btrfs_block_group *cache; | |
4585 | ||
4586 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4587 | BUG_ON(!cache); | |
4588 | ||
4589 | spin_lock(&cache->space_info->lock); | |
4590 | spin_lock(&cache->lock); | |
4591 | cache->pinned += head->num_bytes; | |
4592 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4593 | cache->space_info, head->num_bytes); | |
4594 | cache->reserved -= head->num_bytes; | |
4595 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4596 | spin_unlock(&cache->lock); | |
4597 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4598 | |
4599 | btrfs_put_block_group(cache); | |
4600 | ||
4601 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4602 | head->bytenr + head->num_bytes - 1); | |
4603 | } | |
31890da0 | 4604 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4605 | btrfs_put_delayed_ref_head(head); |
acce952b | 4606 | cond_resched(); |
4607 | spin_lock(&delayed_refs->lock); | |
4608 | } | |
81f7eb00 | 4609 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4610 | |
4611 | spin_unlock(&delayed_refs->lock); | |
acce952b | 4612 | } |
4613 | ||
143bede5 | 4614 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4615 | { |
4616 | struct btrfs_inode *btrfs_inode; | |
84af994b | 4617 | LIST_HEAD(splice); |
acce952b | 4618 | |
eb73c1b7 MX |
4619 | spin_lock(&root->delalloc_lock); |
4620 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4621 | |
4622 | while (!list_empty(&splice)) { | |
fe816d0f | 4623 | struct inode *inode = NULL; |
eb73c1b7 MX |
4624 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4625 | delalloc_inodes); | |
5a8a57f9 | 4626 | btrfs_del_delalloc_inode(btrfs_inode); |
eb73c1b7 | 4627 | spin_unlock(&root->delalloc_lock); |
acce952b | 4628 | |
fe816d0f NB |
4629 | /* |
4630 | * Make sure we get a live inode and that it'll not disappear | |
4631 | * meanwhile. | |
4632 | */ | |
4633 | inode = igrab(&btrfs_inode->vfs_inode); | |
4634 | if (inode) { | |
597441b3 JB |
4635 | unsigned int nofs_flag; |
4636 | ||
4637 | nofs_flag = memalloc_nofs_save(); | |
fe816d0f | 4638 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 | 4639 | memalloc_nofs_restore(nofs_flag); |
fe816d0f NB |
4640 | iput(inode); |
4641 | } | |
eb73c1b7 | 4642 | spin_lock(&root->delalloc_lock); |
acce952b | 4643 | } |
eb73c1b7 MX |
4644 | spin_unlock(&root->delalloc_lock); |
4645 | } | |
4646 | ||
4647 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4648 | { | |
4649 | struct btrfs_root *root; | |
84af994b | 4650 | LIST_HEAD(splice); |
eb73c1b7 MX |
4651 | |
4652 | spin_lock(&fs_info->delalloc_root_lock); | |
4653 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4654 | while (!list_empty(&splice)) { | |
4655 | root = list_first_entry(&splice, struct btrfs_root, | |
4656 | delalloc_root); | |
00246528 | 4657 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4658 | BUG_ON(!root); |
4659 | spin_unlock(&fs_info->delalloc_root_lock); | |
4660 | ||
4661 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4662 | btrfs_put_root(root); |
eb73c1b7 MX |
4663 | |
4664 | spin_lock(&fs_info->delalloc_root_lock); | |
4665 | } | |
4666 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4667 | } |
4668 | ||
aec5716c FM |
4669 | static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
4670 | struct extent_io_tree *dirty_pages, | |
4671 | int mark) | |
acce952b | 4672 | { |
acce952b | 4673 | struct extent_buffer *eb; |
4674 | u64 start = 0; | |
4675 | u64 end; | |
acce952b | 4676 | |
e5860f82 FM |
4677 | while (find_first_extent_bit(dirty_pages, start, &start, &end, |
4678 | mark, NULL)) { | |
91166212 | 4679 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4680 | while (start <= end) { |
0b246afa JM |
4681 | eb = find_extent_buffer(fs_info, start); |
4682 | start += fs_info->nodesize; | |
fd8b2b61 | 4683 | if (!eb) |
acce952b | 4684 | continue; |
c4e54a65 JB |
4685 | |
4686 | btrfs_tree_lock(eb); | |
fd8b2b61 | 4687 | wait_on_extent_buffer_writeback(eb); |
190a8339 | 4688 | btrfs_clear_buffer_dirty(NULL, eb); |
c4e54a65 | 4689 | btrfs_tree_unlock(eb); |
acce952b | 4690 | |
fd8b2b61 | 4691 | free_extent_buffer_stale(eb); |
acce952b | 4692 | } |
4693 | } | |
acce952b | 4694 | } |
4695 | ||
46d81ebd FM |
4696 | static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
4697 | struct extent_io_tree *unpin) | |
acce952b | 4698 | { |
acce952b | 4699 | u64 start; |
4700 | u64 end; | |
acce952b | 4701 | |
acce952b | 4702 | while (1) { |
0e6ec385 FM |
4703 | struct extent_state *cached_state = NULL; |
4704 | ||
fcd5e742 LF |
4705 | /* |
4706 | * The btrfs_finish_extent_commit() may get the same range as | |
4707 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4708 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4709 | * the same extent range. | |
4710 | */ | |
4711 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
e5860f82 FM |
4712 | if (!find_first_extent_bit(unpin, 0, &start, &end, |
4713 | EXTENT_DIRTY, &cached_state)) { | |
fcd5e742 | 4714 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4715 | break; |
fcd5e742 | 4716 | } |
acce952b | 4717 | |
0e6ec385 FM |
4718 | clear_extent_dirty(unpin, start, end, &cached_state); |
4719 | free_extent_state(cached_state); | |
2ff7e61e | 4720 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4721 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4722 | cond_resched(); |
4723 | } | |
acce952b | 4724 | } |
4725 | ||
32da5386 | 4726 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4727 | { |
4728 | struct inode *inode; | |
4729 | ||
4730 | inode = cache->io_ctl.inode; | |
4731 | if (inode) { | |
597441b3 JB |
4732 | unsigned int nofs_flag; |
4733 | ||
4734 | nofs_flag = memalloc_nofs_save(); | |
c79a1751 | 4735 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 JB |
4736 | memalloc_nofs_restore(nofs_flag); |
4737 | ||
c79a1751 LB |
4738 | BTRFS_I(inode)->generation = 0; |
4739 | cache->io_ctl.inode = NULL; | |
4740 | iput(inode); | |
4741 | } | |
bbc37d6e | 4742 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4743 | btrfs_put_block_group(cache); |
4744 | } | |
4745 | ||
4746 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4747 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4748 | { |
32da5386 | 4749 | struct btrfs_block_group *cache; |
c79a1751 LB |
4750 | |
4751 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4752 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4753 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4754 | struct btrfs_block_group, |
c79a1751 | 4755 | dirty_list); |
c79a1751 LB |
4756 | |
4757 | if (!list_empty(&cache->io_list)) { | |
4758 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4759 | list_del_init(&cache->io_list); | |
4760 | btrfs_cleanup_bg_io(cache); | |
4761 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4762 | } | |
4763 | ||
4764 | list_del_init(&cache->dirty_list); | |
4765 | spin_lock(&cache->lock); | |
4766 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4767 | spin_unlock(&cache->lock); | |
4768 | ||
4769 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4770 | btrfs_put_block_group(cache); | |
f66e0209 | 4771 | btrfs_dec_delayed_refs_rsv_bg_updates(fs_info); |
c79a1751 LB |
4772 | spin_lock(&cur_trans->dirty_bgs_lock); |
4773 | } | |
4774 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4775 | ||
45ae2c18 NB |
4776 | /* |
4777 | * Refer to the definition of io_bgs member for details why it's safe | |
4778 | * to use it without any locking | |
4779 | */ | |
c79a1751 LB |
4780 | while (!list_empty(&cur_trans->io_bgs)) { |
4781 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4782 | struct btrfs_block_group, |
c79a1751 | 4783 | io_list); |
c79a1751 LB |
4784 | |
4785 | list_del_init(&cache->io_list); | |
4786 | spin_lock(&cache->lock); | |
4787 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4788 | spin_unlock(&cache->lock); | |
4789 | btrfs_cleanup_bg_io(cache); | |
4790 | } | |
4791 | } | |
4792 | ||
b321a52c BB |
4793 | static void btrfs_free_all_qgroup_pertrans(struct btrfs_fs_info *fs_info) |
4794 | { | |
4795 | struct btrfs_root *gang[8]; | |
4796 | int i; | |
4797 | int ret; | |
4798 | ||
4799 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4800 | while (1) { | |
4801 | ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, | |
4802 | (void **)gang, 0, | |
4803 | ARRAY_SIZE(gang), | |
4804 | BTRFS_ROOT_TRANS_TAG); | |
4805 | if (ret == 0) | |
4806 | break; | |
4807 | for (i = 0; i < ret; i++) { | |
4808 | struct btrfs_root *root = gang[i]; | |
4809 | ||
4810 | btrfs_qgroup_free_meta_all_pertrans(root); | |
4811 | radix_tree_tag_clear(&fs_info->fs_roots_radix, | |
e094f480 | 4812 | (unsigned long)btrfs_root_id(root), |
b321a52c BB |
4813 | BTRFS_ROOT_TRANS_TAG); |
4814 | } | |
4815 | } | |
4816 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4817 | } | |
4818 | ||
49b25e05 | 4819 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4820 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4821 | { |
bbbf7243 NB |
4822 | struct btrfs_device *dev, *tmp; |
4823 | ||
2ff7e61e | 4824 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4825 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4826 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4827 | ||
bbbf7243 NB |
4828 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4829 | post_commit_list) { | |
4830 | list_del_init(&dev->post_commit_list); | |
4831 | } | |
4832 | ||
2ff7e61e | 4833 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4834 | |
4a9d8bde | 4835 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4836 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4837 | |
4a9d8bde | 4838 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4839 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4840 | |
2ff7e61e | 4841 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4842 | EXTENT_DIRTY); |
fe119a6e | 4843 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4844 | |
4a9d8bde MX |
4845 | cur_trans->state =TRANS_STATE_COMPLETED; |
4846 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4847 | } |
4848 | ||
2ff7e61e | 4849 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4850 | { |
4851 | struct btrfs_transaction *t; | |
acce952b | 4852 | |
0b246afa | 4853 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4854 | |
0b246afa JM |
4855 | spin_lock(&fs_info->trans_lock); |
4856 | while (!list_empty(&fs_info->trans_list)) { | |
4857 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 | 4858 | struct btrfs_transaction, list); |
77d20c68 | 4859 | if (t->state >= TRANS_STATE_COMMIT_PREP) { |
9b64f57d | 4860 | refcount_inc(&t->use_count); |
0b246afa | 4861 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4862 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4863 | btrfs_put_transaction(t); |
0b246afa | 4864 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4865 | continue; |
4866 | } | |
0b246afa | 4867 | if (t == fs_info->running_transaction) { |
724e2315 | 4868 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4869 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4870 | /* |
4871 | * We wait for 0 num_writers since we don't hold a trans | |
4872 | * handle open currently for this transaction. | |
4873 | */ | |
4874 | wait_event(t->writer_wait, | |
4875 | atomic_read(&t->num_writers) == 0); | |
4876 | } else { | |
0b246afa | 4877 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4878 | } |
2ff7e61e | 4879 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4880 | |
0b246afa JM |
4881 | spin_lock(&fs_info->trans_lock); |
4882 | if (t == fs_info->running_transaction) | |
4883 | fs_info->running_transaction = NULL; | |
acce952b | 4884 | list_del_init(&t->list); |
0b246afa | 4885 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4886 | |
724e2315 | 4887 | btrfs_put_transaction(t); |
2e4e97ab | 4888 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 4889 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4890 | } |
0b246afa JM |
4891 | spin_unlock(&fs_info->trans_lock); |
4892 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4893 | btrfs_destroy_delayed_inodes(fs_info); |
4894 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4895 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4896 | btrfs_drop_all_logs(fs_info); |
5f2fb819 | 4897 | btrfs_free_all_qgroup_pertrans(fs_info); |
0b246afa | 4898 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4899 | |
4900 | return 0; | |
4901 | } | |
ec7d6dfd | 4902 | |
453e4873 | 4903 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
4904 | { |
4905 | struct btrfs_path *path; | |
4906 | int ret; | |
4907 | struct extent_buffer *l; | |
4908 | struct btrfs_key search_key; | |
4909 | struct btrfs_key found_key; | |
4910 | int slot; | |
4911 | ||
4912 | path = btrfs_alloc_path(); | |
4913 | if (!path) | |
4914 | return -ENOMEM; | |
4915 | ||
4916 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
4917 | search_key.type = -1; | |
4918 | search_key.offset = (u64)-1; | |
4919 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
4920 | if (ret < 0) | |
4921 | goto error; | |
a6724290 DS |
4922 | if (ret == 0) { |
4923 | /* | |
4924 | * Key with offset -1 found, there would have to exist a root | |
4925 | * with such id, but this is out of valid range. | |
4926 | */ | |
4927 | ret = -EUCLEAN; | |
4928 | goto error; | |
4929 | } | |
ec7d6dfd NB |
4930 | if (path->slots[0] > 0) { |
4931 | slot = path->slots[0] - 1; | |
4932 | l = path->nodes[0]; | |
4933 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
4934 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
4935 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 4936 | } else { |
23125104 | 4937 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
4938 | } |
4939 | ret = 0; | |
4940 | error: | |
4941 | btrfs_free_path(path); | |
4942 | return ret; | |
4943 | } | |
4944 | ||
543068a2 | 4945 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
4946 | { |
4947 | int ret; | |
4948 | mutex_lock(&root->objectid_mutex); | |
4949 | ||
6b8fad57 | 4950 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
4951 | btrfs_warn(root->fs_info, |
4952 | "the objectid of root %llu reaches its highest value", | |
e094f480 | 4953 | btrfs_root_id(root)); |
ec7d6dfd NB |
4954 | ret = -ENOSPC; |
4955 | goto out; | |
4956 | } | |
4957 | ||
23125104 | 4958 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
4959 | ret = 0; |
4960 | out: | |
4961 | mutex_unlock(&root->objectid_mutex); | |
4962 | return ret; | |
4963 | } |