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 | struct btrfs_fs_info *fs_info; |
501 | int ret; | |
502 | ||
d8d5f3e1 | 503 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
504 | |
505 | if (wbc->for_kupdate) | |
506 | return 0; | |
507 | ||
e2d84521 | 508 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
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; |
895586eb MWO |
531 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
532 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 533 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
534 | if (folio_get_private(folio)) { |
535 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
536 | "folio private not zero on folio %llu", | |
537 | (unsigned long long)folio_pos(folio)); | |
538 | folio_detach_private(folio); | |
9ad6b7bc | 539 | } |
d98237b3 CM |
540 | } |
541 | ||
bb146eb2 | 542 | #ifdef DEBUG |
0079c3b1 MWO |
543 | static bool btree_dirty_folio(struct address_space *mapping, |
544 | struct folio *folio) | |
545 | { | |
546 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
5e0e8799 | 547 | struct btrfs_subpage_info *spi = fs_info->subpage_info; |
139e8cd3 | 548 | struct btrfs_subpage *subpage; |
0b32f4bb | 549 | struct extent_buffer *eb; |
139e8cd3 | 550 | int cur_bit = 0; |
0079c3b1 | 551 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
552 | |
553 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 554 | eb = folio_get_private(folio); |
139e8cd3 QW |
555 | BUG_ON(!eb); |
556 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
557 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 558 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 559 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 560 | } |
5e0e8799 QW |
561 | |
562 | ASSERT(spi); | |
0079c3b1 | 563 | subpage = folio_get_private(folio); |
139e8cd3 | 564 | |
5e0e8799 QW |
565 | for (cur_bit = spi->dirty_offset; |
566 | cur_bit < spi->dirty_offset + spi->bitmap_nr_bits; | |
567 | cur_bit++) { | |
139e8cd3 QW |
568 | unsigned long flags; |
569 | u64 cur; | |
139e8cd3 QW |
570 | |
571 | spin_lock_irqsave(&subpage->lock, flags); | |
5e0e8799 | 572 | if (!test_bit(cur_bit, subpage->bitmaps)) { |
139e8cd3 | 573 | spin_unlock_irqrestore(&subpage->lock, flags); |
139e8cd3 QW |
574 | continue; |
575 | } | |
576 | spin_unlock_irqrestore(&subpage->lock, flags); | |
577 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 578 | |
139e8cd3 QW |
579 | eb = find_extent_buffer(fs_info, cur); |
580 | ASSERT(eb); | |
581 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
582 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 583 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
584 | free_extent_buffer(eb); |
585 | ||
5e0e8799 | 586 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits) - 1; |
139e8cd3 | 587 | } |
0079c3b1 | 588 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 589 | } |
0079c3b1 MWO |
590 | #else |
591 | #define btree_dirty_folio filemap_dirty_folio | |
592 | #endif | |
0b32f4bb | 593 | |
7f09410b | 594 | static const struct address_space_operations btree_aops = { |
0da5468f | 595 | .writepages = btree_writepages, |
f913cff3 | 596 | .release_folio = btree_release_folio, |
895586eb | 597 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
598 | .migrate_folio = btree_migrate_folio, |
599 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
600 | }; |
601 | ||
2ff7e61e JM |
602 | struct extent_buffer *btrfs_find_create_tree_block( |
603 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
604 | u64 bytenr, u64 owner_root, |
605 | int level) | |
0999df54 | 606 | { |
0b246afa JM |
607 | if (btrfs_is_testing(fs_info)) |
608 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 609 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
610 | } |
611 | ||
581c1760 QW |
612 | /* |
613 | * Read tree block at logical address @bytenr and do variant basic but critical | |
614 | * verification. | |
615 | * | |
789d6a3a QW |
616 | * @check: expected tree parentness check, see comments of the |
617 | * structure for details. | |
581c1760 | 618 | */ |
2ff7e61e | 619 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
789d6a3a | 620 | struct btrfs_tree_parent_check *check) |
0999df54 CM |
621 | { |
622 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
623 | int ret; |
624 | ||
789d6a3a QW |
625 | ASSERT(check); |
626 | ||
627 | buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, | |
628 | check->level); | |
c871b0f2 LB |
629 | if (IS_ERR(buf)) |
630 | return buf; | |
0999df54 | 631 | |
789d6a3a | 632 | ret = btrfs_read_extent_buffer(buf, check); |
0f0fe8f7 | 633 | if (ret) { |
537f38f0 | 634 | free_extent_buffer_stale(buf); |
64c043de | 635 | return ERR_PTR(ret); |
0f0fe8f7 | 636 | } |
789d6a3a | 637 | if (btrfs_check_eb_owner(buf, check->owner_root)) { |
88c602ab QW |
638 | free_extent_buffer_stale(buf); |
639 | return ERR_PTR(-EUCLEAN); | |
640 | } | |
5f39d397 | 641 | return buf; |
ce9adaa5 | 642 | |
eb60ceac CM |
643 | } |
644 | ||
da17066c | 645 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 646 | u64 objectid) |
d97e63b6 | 647 | { |
7c0260ee | 648 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
649 | |
650 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
651 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
652 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 653 | root->fs_info = fs_info; |
2e608bd1 | 654 | root->root_key.objectid = objectid; |
cfaa7295 | 655 | root->node = NULL; |
a28ec197 | 656 | root->commit_root = NULL; |
27cdeb70 | 657 | root->state = 0; |
abed4aaa | 658 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 659 | |
0f7d52f4 | 660 | root->last_trans = 0; |
6b8fad57 | 661 | root->free_objectid = 0; |
eb73c1b7 | 662 | root->nr_delalloc_inodes = 0; |
199c2a9c | 663 | root->nr_ordered_extents = 0; |
6bef4d31 | 664 | root->inode_tree = RB_ROOT; |
6140ba8a DS |
665 | /* GFP flags are compatible with XA_FLAGS_*. */ |
666 | xa_init_flags(&root->delayed_nodes, GFP_ATOMIC); | |
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", | |
778 | root->root_key.objectid, root->root_key.offset); | |
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 YZ |
1013 | log_root->last_trans = trans->transid; |
1014 | log_root->root_key.offset = root->root_key.objectid; | |
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 | */ | |
1078 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1079 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1080 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1081 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1082 | btrfs_crit(fs_info, | |
1083 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1084 | root->root_key.objectid, root->node->start, | |
1085 | btrfs_header_owner(root->node), | |
1086 | root->root_key.objectid); | |
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 | |
aeb935a4 | 1123 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
6ebcd021 QW |
1124 | !btrfs_is_data_reloc_root(root) && |
1125 | is_fstree(root->root_key.objectid)) { | |
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 | */ | |
1134 | if (is_fstree(root->root_key.objectid) && | |
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, | |
1221 | (unsigned long)root->root_key.objectid, | |
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 | { | |
141386e1 JB |
1268 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1269 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1270 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1271 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1272 | btrfs_free_csum_hash(fs_info); | |
1273 | btrfs_free_stripe_hash_table(fs_info); | |
1274 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1275 | kfree(fs_info->balance_ctl); |
1276 | kfree(fs_info->delayed_root); | |
abed4aaa | 1277 | free_global_roots(fs_info); |
00246528 JB |
1278 | btrfs_put_root(fs_info->tree_root); |
1279 | btrfs_put_root(fs_info->chunk_root); | |
1280 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1281 | btrfs_put_root(fs_info->quota_root); |
1282 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1283 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1284 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1285 | btrfs_put_root(fs_info->block_group_root); |
51502090 | 1286 | btrfs_put_root(fs_info->stripe_root); |
bd647ce3 | 1287 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1288 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1289 | kfree(fs_info->super_copy); |
1290 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1291 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1292 | kvfree(fs_info); |
1293 | } | |
1294 | ||
1295 | ||
2dfb1e43 QW |
1296 | /* |
1297 | * Get an in-memory reference of a root structure. | |
1298 | * | |
1299 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1300 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1301 | * found, then read it from disk and add it to cached fs roots. | |
1302 | * | |
1303 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1304 | * | |
1305 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1306 | * same root objectid. | |
1307 | * | |
1308 | * @objectid: root id | |
1309 | * @anon_dev: preallocated anonymous block device number for new roots, | |
e2b54eaf | 1310 | * pass NULL for a new allocation. |
2dfb1e43 QW |
1311 | * @check_ref: whether to check root item references, If true, return -ENOENT |
1312 | * for orphan roots | |
1313 | */ | |
1314 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1315 | u64 objectid, dev_t *anon_dev, |
2dfb1e43 | 1316 | bool check_ref) |
5eda7b5e CM |
1317 | { |
1318 | struct btrfs_root *root; | |
381cf658 | 1319 | struct btrfs_path *path; |
1d4c08e0 | 1320 | struct btrfs_key key; |
5eda7b5e CM |
1321 | int ret; |
1322 | ||
49d11bea JB |
1323 | root = btrfs_get_global_root(fs_info, objectid); |
1324 | if (root) | |
1325 | return root; | |
773e722a QW |
1326 | |
1327 | /* | |
1328 | * If we're called for non-subvolume trees, and above function didn't | |
1329 | * find one, do not try to read it from disk. | |
1330 | * | |
1331 | * This is namely for free-space-tree and quota tree, which can change | |
1332 | * at runtime and should only be grabbed from fs_info. | |
1333 | */ | |
1334 | if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) | |
1335 | return ERR_PTR(-ENOENT); | |
4df27c4d | 1336 | again: |
56e9357a | 1337 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1338 | if (root) { |
e03ee2fe QW |
1339 | /* |
1340 | * Some other caller may have read out the newly inserted | |
1341 | * subvolume already (for things like backref walk etc). Not | |
1342 | * that common but still possible. In that case, we just need | |
1343 | * to free the anon_dev. | |
1344 | */ | |
e2b54eaf FM |
1345 | if (unlikely(anon_dev && *anon_dev)) { |
1346 | free_anon_bdev(*anon_dev); | |
1347 | *anon_dev = 0; | |
e03ee2fe QW |
1348 | } |
1349 | ||
bc44d7c4 | 1350 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1351 | btrfs_put_root(root); |
48475471 | 1352 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1353 | } |
5eda7b5e | 1354 | return root; |
48475471 | 1355 | } |
5eda7b5e | 1356 | |
56e9357a DS |
1357 | key.objectid = objectid; |
1358 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1359 | key.offset = (u64)-1; | |
1360 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1361 | if (IS_ERR(root)) |
1362 | return root; | |
3394e160 | 1363 | |
c00869f1 | 1364 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1365 | ret = -ENOENT; |
581bb050 | 1366 | goto fail; |
35a30d7c | 1367 | } |
581bb050 | 1368 | |
e2b54eaf | 1369 | ret = btrfs_init_fs_root(root, anon_dev ? *anon_dev : 0); |
ac08aedf CM |
1370 | if (ret) |
1371 | goto fail; | |
3394e160 | 1372 | |
381cf658 DS |
1373 | path = btrfs_alloc_path(); |
1374 | if (!path) { | |
1375 | ret = -ENOMEM; | |
1376 | goto fail; | |
1377 | } | |
1d4c08e0 DS |
1378 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1379 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1380 | key.offset = objectid; |
1d4c08e0 DS |
1381 | |
1382 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1383 | btrfs_free_path(path); |
d68fc57b YZ |
1384 | if (ret < 0) |
1385 | goto fail; | |
1386 | if (ret == 0) | |
27cdeb70 | 1387 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1388 | |
cb517eab | 1389 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1390 | if (ret) { |
168a2f77 JJB |
1391 | if (ret == -EEXIST) { |
1392 | btrfs_put_root(root); | |
4df27c4d | 1393 | goto again; |
168a2f77 | 1394 | } |
4df27c4d | 1395 | goto fail; |
0f7d52f4 | 1396 | } |
edbd8d4e | 1397 | return root; |
4df27c4d | 1398 | fail: |
33fab972 FM |
1399 | /* |
1400 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1401 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1402 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1403 | * and once again by our caller. | |
1404 | */ | |
e2b54eaf | 1405 | if (anon_dev && *anon_dev) |
33fab972 | 1406 | root->anon_dev = 0; |
8c38938c | 1407 | btrfs_put_root(root); |
4df27c4d | 1408 | return ERR_PTR(ret); |
edbd8d4e CM |
1409 | } |
1410 | ||
2dfb1e43 QW |
1411 | /* |
1412 | * Get in-memory reference of a root structure | |
1413 | * | |
1414 | * @objectid: tree objectid | |
1415 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1416 | * one reference | |
1417 | */ | |
1418 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1419 | u64 objectid, bool check_ref) | |
1420 | { | |
e2b54eaf | 1421 | return btrfs_get_root_ref(fs_info, objectid, NULL, check_ref); |
2dfb1e43 QW |
1422 | } |
1423 | ||
1424 | /* | |
1425 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1426 | * the anonymous block device id | |
1427 | * | |
1428 | * @objectid: tree objectid | |
e2b54eaf FM |
1429 | * @anon_dev: if NULL, allocate a new anonymous block device or use the |
1430 | * parameter value if not NULL | |
2dfb1e43 QW |
1431 | */ |
1432 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
e2b54eaf | 1433 | u64 objectid, dev_t *anon_dev) |
2dfb1e43 QW |
1434 | { |
1435 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1436 | } | |
1437 | ||
49d11bea | 1438 | /* |
9580503b DS |
1439 | * Return a root for the given objectid. |
1440 | * | |
49d11bea JB |
1441 | * @fs_info: the fs_info |
1442 | * @objectid: the objectid we need to lookup | |
1443 | * | |
1444 | * This is exclusively used for backref walking, and exists specifically because | |
1445 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1446 | * creation time, which means we may have to read the tree_root in order to look | |
1447 | * up a fs root that is not in memory. If the root is not in memory we will | |
1448 | * read the tree root commit root and look up the fs root from there. This is a | |
1449 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1450 | * have the most uptodate information, it'll simply be discarded once the | |
1451 | * backref code is finished using the root. | |
1452 | */ | |
1453 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1454 | struct btrfs_path *path, | |
1455 | u64 objectid) | |
1456 | { | |
1457 | struct btrfs_root *root; | |
1458 | struct btrfs_key key; | |
1459 | ||
1460 | ASSERT(path->search_commit_root && path->skip_locking); | |
1461 | ||
1462 | /* | |
1463 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1464 | * since this is called via the backref walking code we won't be looking | |
1465 | * up a root that doesn't exist, unless there's corruption. So if root | |
1466 | * != NULL just return it. | |
1467 | */ | |
1468 | root = btrfs_get_global_root(fs_info, objectid); | |
1469 | if (root) | |
1470 | return root; | |
1471 | ||
1472 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1473 | if (root) | |
1474 | return root; | |
1475 | ||
1476 | key.objectid = objectid; | |
1477 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1478 | key.offset = (u64)-1; | |
1479 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1480 | btrfs_release_path(path); | |
1481 | ||
1482 | return root; | |
1483 | } | |
1484 | ||
a74a4b97 CM |
1485 | static int cleaner_kthread(void *arg) |
1486 | { | |
0d031dc4 | 1487 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1488 | int again; |
a74a4b97 | 1489 | |
d6fd0ae2 | 1490 | while (1) { |
d0278245 | 1491 | again = 0; |
a74a4b97 | 1492 | |
fd340d0f JB |
1493 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1494 | ||
d0278245 | 1495 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1496 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1497 | goto sleep; |
1498 | ||
90c711ab ZB |
1499 | /* |
1500 | * Do not do anything if we might cause open_ctree() to block | |
1501 | * before we have finished mounting the filesystem. | |
1502 | */ | |
0b246afa | 1503 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1504 | goto sleep; |
1505 | ||
0b246afa | 1506 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1507 | goto sleep; |
1508 | ||
dc7f370c MX |
1509 | /* |
1510 | * Avoid the problem that we change the status of the fs | |
1511 | * during the above check and trylock. | |
1512 | */ | |
2ff7e61e | 1513 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1514 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1515 | goto sleep; |
76dda93c | 1516 | } |
a74a4b97 | 1517 | |
b7625f46 QW |
1518 | if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) |
1519 | btrfs_sysfs_feature_update(fs_info); | |
1520 | ||
2ff7e61e | 1521 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1522 | |
33c44184 | 1523 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1524 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1525 | |
1526 | /* | |
05323cd1 MX |
1527 | * The defragger has dealt with the R/O remount and umount, |
1528 | * needn't do anything special here. | |
d0278245 | 1529 | */ |
0b246afa | 1530 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1531 | |
1532 | /* | |
f3372065 | 1533 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1534 | * with relocation (btrfs_relocate_chunk) and relocation |
1535 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1536 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1537 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1538 | * unused block groups. | |
1539 | */ | |
0b246afa | 1540 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1541 | |
1542 | /* | |
1543 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1544 | * all unused block_groups. This possibly gives us some more free | |
1545 | * space. | |
1546 | */ | |
1547 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1548 | sleep: |
a0a1db70 | 1549 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1550 | if (kthread_should_park()) |
1551 | kthread_parkme(); | |
1552 | if (kthread_should_stop()) | |
1553 | return 0; | |
838fe188 | 1554 | if (!again) { |
a74a4b97 | 1555 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1556 | schedule(); |
a74a4b97 CM |
1557 | __set_current_state(TASK_RUNNING); |
1558 | } | |
da288d28 | 1559 | } |
a74a4b97 CM |
1560 | } |
1561 | ||
1562 | static int transaction_kthread(void *arg) | |
1563 | { | |
1564 | struct btrfs_root *root = arg; | |
0b246afa | 1565 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1566 | struct btrfs_trans_handle *trans; |
1567 | struct btrfs_transaction *cur; | |
8929ecfa | 1568 | u64 transid; |
643900be | 1569 | time64_t delta; |
a74a4b97 | 1570 | unsigned long delay; |
914b2007 | 1571 | bool cannot_commit; |
a74a4b97 CM |
1572 | |
1573 | do { | |
914b2007 | 1574 | cannot_commit = false; |
ba1bc00f | 1575 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1576 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1577 | |
0b246afa JM |
1578 | spin_lock(&fs_info->trans_lock); |
1579 | cur = fs_info->running_transaction; | |
a74a4b97 | 1580 | if (!cur) { |
0b246afa | 1581 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1582 | goto sleep; |
1583 | } | |
31153d81 | 1584 | |
643900be | 1585 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 | 1586 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
77d20c68 | 1587 | cur->state < TRANS_STATE_COMMIT_PREP && |
643900be | 1588 | delta < fs_info->commit_interval) { |
0b246afa | 1589 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1590 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1591 | delay = min(delay, | |
1592 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1593 | goto sleep; |
1594 | } | |
8929ecfa | 1595 | transid = cur->transid; |
0b246afa | 1596 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1597 | |
79787eaa | 1598 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1599 | trans = btrfs_attach_transaction(root); |
914b2007 | 1600 | if (IS_ERR(trans)) { |
354aa0fb MX |
1601 | if (PTR_ERR(trans) != -ENOENT) |
1602 | cannot_commit = true; | |
79787eaa | 1603 | goto sleep; |
914b2007 | 1604 | } |
8929ecfa | 1605 | if (transid == trans->transid) { |
3a45bb20 | 1606 | btrfs_commit_transaction(trans); |
8929ecfa | 1607 | } else { |
3a45bb20 | 1608 | btrfs_end_transaction(trans); |
8929ecfa | 1609 | } |
a74a4b97 | 1610 | sleep: |
0b246afa JM |
1611 | wake_up_process(fs_info->cleaner_kthread); |
1612 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1613 | |
84961539 | 1614 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1615 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1616 | if (!kthread_should_stop() && |
0b246afa | 1617 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1618 | cannot_commit)) |
bc5511d0 | 1619 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1620 | } while (!kthread_should_stop()); |
1621 | return 0; | |
1622 | } | |
1623 | ||
af31f5e5 | 1624 | /* |
01f0f9da NB |
1625 | * This will find the highest generation in the array of root backups. The |
1626 | * index of the highest array is returned, or -EINVAL if we can't find | |
1627 | * anything. | |
af31f5e5 CM |
1628 | * |
1629 | * We check to make sure the array is valid by comparing the | |
1630 | * generation of the latest root in the array with the generation | |
1631 | * in the super block. If they don't match we pitch it. | |
1632 | */ | |
01f0f9da | 1633 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1634 | { |
01f0f9da | 1635 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1636 | u64 cur; |
af31f5e5 CM |
1637 | struct btrfs_root_backup *root_backup; |
1638 | int i; | |
1639 | ||
1640 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1641 | root_backup = info->super_copy->super_roots + i; | |
1642 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1643 | if (cur == newest_gen) | |
01f0f9da | 1644 | return i; |
af31f5e5 CM |
1645 | } |
1646 | ||
01f0f9da | 1647 | return -EINVAL; |
af31f5e5 CM |
1648 | } |
1649 | ||
af31f5e5 CM |
1650 | /* |
1651 | * copy all the root pointers into the super backup array. | |
1652 | * this will bump the backup pointer by one when it is | |
1653 | * done | |
1654 | */ | |
1655 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1656 | { | |
6ef108dd | 1657 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1658 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1659 | |
1660 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1661 | ||
1662 | /* | |
1663 | * make sure all of our padding and empty slots get zero filled | |
1664 | * regardless of which ones we use today | |
1665 | */ | |
1666 | memset(root_backup, 0, sizeof(*root_backup)); | |
1667 | ||
1668 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1669 | ||
1670 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1671 | btrfs_set_backup_tree_root_gen(root_backup, | |
1672 | btrfs_header_generation(info->tree_root->node)); | |
1673 | ||
1674 | btrfs_set_backup_tree_root_level(root_backup, | |
1675 | btrfs_header_level(info->tree_root->node)); | |
1676 | ||
1677 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1678 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1679 | btrfs_header_generation(info->chunk_root->node)); | |
1680 | btrfs_set_backup_chunk_root_level(root_backup, | |
1681 | btrfs_header_level(info->chunk_root->node)); | |
1682 | ||
1c56ab99 | 1683 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1684 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1685 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1686 | |
1687 | btrfs_set_backup_extent_root(root_backup, | |
1688 | extent_root->node->start); | |
1689 | btrfs_set_backup_extent_root_gen(root_backup, | |
1690 | btrfs_header_generation(extent_root->node)); | |
1691 | btrfs_set_backup_extent_root_level(root_backup, | |
1692 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1693 | |
1694 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1695 | btrfs_set_backup_csum_root_gen(root_backup, | |
1696 | btrfs_header_generation(csum_root->node)); | |
1697 | btrfs_set_backup_csum_root_level(root_backup, | |
1698 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1699 | } |
af31f5e5 | 1700 | |
7c7e82a7 CM |
1701 | /* |
1702 | * we might commit during log recovery, which happens before we set | |
1703 | * the fs_root. Make sure it is valid before we fill it in. | |
1704 | */ | |
1705 | if (info->fs_root && info->fs_root->node) { | |
1706 | btrfs_set_backup_fs_root(root_backup, | |
1707 | info->fs_root->node->start); | |
1708 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1709 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1710 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1711 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1712 | } |
af31f5e5 CM |
1713 | |
1714 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1715 | btrfs_set_backup_dev_root_gen(root_backup, | |
1716 | btrfs_header_generation(info->dev_root->node)); | |
1717 | btrfs_set_backup_dev_root_level(root_backup, | |
1718 | btrfs_header_level(info->dev_root->node)); | |
1719 | ||
af31f5e5 CM |
1720 | btrfs_set_backup_total_bytes(root_backup, |
1721 | btrfs_super_total_bytes(info->super_copy)); | |
1722 | btrfs_set_backup_bytes_used(root_backup, | |
1723 | btrfs_super_bytes_used(info->super_copy)); | |
1724 | btrfs_set_backup_num_devices(root_backup, | |
1725 | btrfs_super_num_devices(info->super_copy)); | |
1726 | ||
1727 | /* | |
1728 | * if we don't copy this out to the super_copy, it won't get remembered | |
1729 | * for the next commit | |
1730 | */ | |
1731 | memcpy(&info->super_copy->super_roots, | |
1732 | &info->super_for_commit->super_roots, | |
1733 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1734 | } | |
1735 | ||
bd2336b2 | 1736 | /* |
9580503b DS |
1737 | * Reads a backup root based on the passed priority. Prio 0 is the newest, prio |
1738 | * 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
bd2336b2 | 1739 | * |
9580503b DS |
1740 | * @fs_info: filesystem whose backup roots need to be read |
1741 | * @priority: priority of backup root required | |
bd2336b2 NB |
1742 | * |
1743 | * Returns backup root index on success and -EINVAL otherwise. | |
1744 | */ | |
1745 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1746 | { | |
1747 | int backup_index = find_newest_super_backup(fs_info); | |
1748 | struct btrfs_super_block *super = fs_info->super_copy; | |
1749 | struct btrfs_root_backup *root_backup; | |
1750 | ||
1751 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1752 | if (priority == 0) | |
1753 | return backup_index; | |
1754 | ||
1755 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1756 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1757 | } else { | |
1758 | return -EINVAL; | |
1759 | } | |
1760 | ||
1761 | root_backup = super->super_roots + backup_index; | |
1762 | ||
1763 | btrfs_set_super_generation(super, | |
1764 | btrfs_backup_tree_root_gen(root_backup)); | |
1765 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1766 | btrfs_set_super_root_level(super, | |
1767 | btrfs_backup_tree_root_level(root_backup)); | |
1768 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1769 | ||
1770 | /* | |
1771 | * Fixme: the total bytes and num_devices need to match or we should | |
1772 | * need a fsck | |
1773 | */ | |
1774 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1775 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1776 | ||
1777 | return backup_index; | |
1778 | } | |
1779 | ||
7abadb64 LB |
1780 | /* helper to cleanup workers */ |
1781 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1782 | { | |
dc6e3209 | 1783 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 1784 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 1785 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
1786 | if (fs_info->endio_workers) |
1787 | destroy_workqueue(fs_info->endio_workers); | |
385de0ef CH |
1788 | if (fs_info->rmw_workers) |
1789 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
1790 | if (fs_info->compressed_write_workers) |
1791 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
1792 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
1793 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 1794 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 1795 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 1796 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 1797 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
1798 | if (fs_info->discard_ctl.discard_workers) |
1799 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
1800 | /* |
1801 | * Now that all other work queues are destroyed, we can safely destroy | |
1802 | * the queues used for metadata I/O, since tasks from those other work | |
1803 | * queues can do metadata I/O operations. | |
1804 | */ | |
d7b9416f CH |
1805 | if (fs_info->endio_meta_workers) |
1806 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
1807 | } |
1808 | ||
2e9f5954 R |
1809 | static void free_root_extent_buffers(struct btrfs_root *root) |
1810 | { | |
1811 | if (root) { | |
1812 | free_extent_buffer(root->node); | |
1813 | free_extent_buffer(root->commit_root); | |
1814 | root->node = NULL; | |
1815 | root->commit_root = NULL; | |
1816 | } | |
1817 | } | |
1818 | ||
abed4aaa JB |
1819 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
1820 | { | |
1821 | struct btrfs_root *root, *tmp; | |
1822 | ||
1823 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
1824 | &fs_info->global_root_tree, | |
1825 | rb_node) | |
1826 | free_root_extent_buffers(root); | |
1827 | } | |
1828 | ||
af31f5e5 | 1829 | /* helper to cleanup tree roots */ |
4273eaff | 1830 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 1831 | { |
2e9f5954 | 1832 | free_root_extent_buffers(info->tree_root); |
655b09fe | 1833 | |
abed4aaa | 1834 | free_global_root_pointers(info); |
2e9f5954 | 1835 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
1836 | free_root_extent_buffers(info->quota_root); |
1837 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 1838 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 1839 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 1840 | free_root_extent_buffers(info->block_group_root); |
51502090 | 1841 | free_root_extent_buffers(info->stripe_root); |
4273eaff | 1842 | if (free_chunk_root) |
2e9f5954 | 1843 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
1844 | } |
1845 | ||
8c38938c JB |
1846 | void btrfs_put_root(struct btrfs_root *root) |
1847 | { | |
1848 | if (!root) | |
1849 | return; | |
1850 | ||
1851 | if (refcount_dec_and_test(&root->refs)) { | |
1852 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 1853 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
1854 | if (root->anon_dev) |
1855 | free_anon_bdev(root->anon_dev); | |
923eb523 | 1856 | free_root_extent_buffers(root); |
8c38938c | 1857 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 1858 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 1859 | list_del_init(&root->leak_list); |
fc7cbcd4 | 1860 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
1861 | #endif |
1862 | kfree(root); | |
1863 | } | |
1864 | } | |
1865 | ||
faa2dbf0 | 1866 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 1867 | { |
fc7cbcd4 DS |
1868 | int ret; |
1869 | struct btrfs_root *gang[8]; | |
1870 | int i; | |
171f6537 JB |
1871 | |
1872 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
1873 | gang[0] = list_entry(fs_info->dead_roots.next, |
1874 | struct btrfs_root, root_list); | |
1875 | list_del(&gang[0]->root_list); | |
171f6537 | 1876 | |
fc7cbcd4 DS |
1877 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
1878 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
1879 | btrfs_put_root(gang[0]); | |
171f6537 JB |
1880 | } |
1881 | ||
fc7cbcd4 DS |
1882 | while (1) { |
1883 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1884 | (void **)gang, 0, | |
1885 | ARRAY_SIZE(gang)); | |
1886 | if (!ret) | |
1887 | break; | |
1888 | for (i = 0; i < ret; i++) | |
1889 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
1890 | } |
1891 | } | |
af31f5e5 | 1892 | |
638aa7ed ES |
1893 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
1894 | { | |
1895 | mutex_init(&fs_info->scrub_lock); | |
1896 | atomic_set(&fs_info->scrubs_running, 0); | |
1897 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1898 | atomic_set(&fs_info->scrubs_paused, 0); | |
1899 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1900 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 1901 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
1902 | } |
1903 | ||
779a65a4 ES |
1904 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
1905 | { | |
1906 | spin_lock_init(&fs_info->balance_lock); | |
1907 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
1908 | atomic_set(&fs_info->balance_pause_req, 0); |
1909 | atomic_set(&fs_info->balance_cancel_req, 0); | |
1910 | fs_info->balance_ctl = NULL; | |
1911 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 1912 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
1913 | } |
1914 | ||
dcb2137c | 1915 | static int btrfs_init_btree_inode(struct super_block *sb) |
f37938e0 | 1916 | { |
dcb2137c | 1917 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
e256927b JB |
1918 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
1919 | fs_info->tree_root); | |
dcb2137c CH |
1920 | struct inode *inode; |
1921 | ||
1922 | inode = new_inode(sb); | |
1923 | if (!inode) | |
1924 | return -ENOMEM; | |
2ff7e61e JM |
1925 | |
1926 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
1927 | set_nlink(inode, 1); | |
f37938e0 ES |
1928 | /* |
1929 | * we set the i_size on the btree inode to the max possible int. | |
1930 | * the real end of the address space is determined by all of | |
1931 | * the devices in the system | |
1932 | */ | |
2ff7e61e JM |
1933 | inode->i_size = OFFSET_MAX; |
1934 | inode->i_mapping->a_ops = &btree_aops; | |
dcb2137c | 1935 | mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); |
f37938e0 | 1936 | |
2ff7e61e | 1937 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 1938 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
35da5a7e | 1939 | IO_TREE_BTREE_INODE_IO); |
2ff7e61e | 1940 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 1941 | |
5c8fd99f | 1942 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
1943 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
1944 | BTRFS_I(inode)->location.type = 0; | |
1945 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 1946 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 1947 | __insert_inode_hash(inode, hash); |
dcb2137c CH |
1948 | fs_info->btree_inode = inode; |
1949 | ||
1950 | return 0; | |
f37938e0 ES |
1951 | } |
1952 | ||
ad618368 ES |
1953 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
1954 | { | |
ad618368 | 1955 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 1956 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 1957 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
1958 | } |
1959 | ||
f9e92e40 ES |
1960 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
1961 | { | |
1962 | spin_lock_init(&fs_info->qgroup_lock); | |
1963 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
1964 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
1965 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
1966 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 1967 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 1968 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 1969 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
1970 | mutex_init(&fs_info->qgroup_rescan_lock); |
1971 | } | |
1972 | ||
d21deec5 | 1973 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 1974 | { |
f7b885be | 1975 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 1976 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
58e814fc | 1977 | unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; |
2a458198 ES |
1978 | |
1979 | fs_info->workers = | |
a31b4a43 | 1980 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2a458198 ES |
1981 | |
1982 | fs_info->delalloc_workers = | |
cb001095 JM |
1983 | btrfs_alloc_workqueue(fs_info, "delalloc", |
1984 | flags, max_active, 2); | |
2a458198 ES |
1985 | |
1986 | fs_info->flush_workers = | |
cb001095 JM |
1987 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
1988 | flags, max_active, 0); | |
2a458198 ES |
1989 | |
1990 | fs_info->caching_workers = | |
cb001095 | 1991 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 1992 | |
2a458198 | 1993 | fs_info->fixup_workers = |
58e814fc | 1994 | btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); |
2a458198 | 1995 | |
2a458198 | 1996 | fs_info->endio_workers = |
d7b9416f | 1997 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 1998 | fs_info->endio_meta_workers = |
d7b9416f | 1999 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
385de0ef | 2000 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2001 | fs_info->endio_write_workers = |
cb001095 JM |
2002 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2003 | max_active, 2); | |
fed8a72d CH |
2004 | fs_info->compressed_write_workers = |
2005 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2006 | fs_info->endio_freespace_worker = |
cb001095 JM |
2007 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2008 | max_active, 0); | |
2a458198 | 2009 | fs_info->delayed_workers = |
cb001095 JM |
2010 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2011 | max_active, 0); | |
2a458198 | 2012 | fs_info->qgroup_rescan_workers = |
58e814fc TH |
2013 | btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", |
2014 | ordered_flags); | |
b0643e59 | 2015 | fs_info->discard_ctl.discard_workers = |
58e814fc | 2016 | alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); |
2a458198 | 2017 | |
8bfec2e4 | 2018 | if (!(fs_info->workers && |
a31b4a43 | 2019 | fs_info->delalloc_workers && fs_info->flush_workers && |
2a458198 | 2020 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2021 | fs_info->compressed_write_workers && |
1a1a2851 | 2022 | fs_info->endio_write_workers && |
2a458198 | 2023 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
f26c9238 QW |
2024 | fs_info->caching_workers && fs_info->fixup_workers && |
2025 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2026 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2027 | return -ENOMEM; |
2028 | } | |
2029 | ||
2030 | return 0; | |
2031 | } | |
2032 | ||
6d97c6e3 JT |
2033 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2034 | { | |
2035 | struct crypto_shash *csum_shash; | |
b4e967be | 2036 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2037 | |
b4e967be | 2038 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2039 | |
2040 | if (IS_ERR(csum_shash)) { | |
2041 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2042 | csum_driver); |
6d97c6e3 JT |
2043 | return PTR_ERR(csum_shash); |
2044 | } | |
2045 | ||
2046 | fs_info->csum_shash = csum_shash; | |
2047 | ||
68d99ab0 CH |
2048 | /* |
2049 | * Check if the checksum implementation is a fast accelerated one. | |
2050 | * As-is this is a bit of a hack and should be replaced once the csum | |
2051 | * implementations provide that information themselves. | |
2052 | */ | |
2053 | switch (csum_type) { | |
2054 | case BTRFS_CSUM_TYPE_CRC32: | |
2055 | if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) | |
2056 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2057 | break; | |
efcfcbc6 DS |
2058 | case BTRFS_CSUM_TYPE_XXHASH: |
2059 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2060 | break; | |
68d99ab0 CH |
2061 | default: |
2062 | break; | |
2063 | } | |
2064 | ||
c8a5f8ca DS |
2065 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2066 | btrfs_super_csum_name(csum_type), | |
2067 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2068 | return 0; |
2069 | } | |
2070 | ||
63443bf5 ES |
2071 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2072 | struct btrfs_fs_devices *fs_devices) | |
2073 | { | |
2074 | int ret; | |
789d6a3a | 2075 | struct btrfs_tree_parent_check check = { 0 }; |
63443bf5 ES |
2076 | struct btrfs_root *log_tree_root; |
2077 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2078 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2079 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2080 | |
2081 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2082 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2083 | return -EIO; |
2084 | } | |
2085 | ||
96dfcb46 JB |
2086 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2087 | GFP_KERNEL); | |
63443bf5 ES |
2088 | if (!log_tree_root) |
2089 | return -ENOMEM; | |
2090 | ||
789d6a3a QW |
2091 | check.level = level; |
2092 | check.transid = fs_info->generation + 1; | |
2093 | check.owner_root = BTRFS_TREE_LOG_OBJECTID; | |
2094 | log_tree_root->node = read_tree_block(fs_info, bytenr, &check); | |
64c043de | 2095 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2096 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2097 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2098 | log_tree_root->node = NULL; |
00246528 | 2099 | btrfs_put_root(log_tree_root); |
0eeff236 | 2100 | return ret; |
4eb150d6 QW |
2101 | } |
2102 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2103 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2104 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2105 | return -EIO; |
2106 | } | |
4eb150d6 | 2107 | |
63443bf5 ES |
2108 | /* returns with log_tree_root freed on success */ |
2109 | ret = btrfs_recover_log_trees(log_tree_root); | |
2110 | if (ret) { | |
0b246afa JM |
2111 | btrfs_handle_fs_error(fs_info, ret, |
2112 | "Failed to recover log tree"); | |
00246528 | 2113 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2114 | return ret; |
2115 | } | |
2116 | ||
bc98a42c | 2117 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2118 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2119 | if (ret) |
2120 | return ret; | |
2121 | } | |
2122 | ||
2123 | return 0; | |
2124 | } | |
2125 | ||
abed4aaa JB |
2126 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2127 | struct btrfs_path *path, u64 objectid, | |
2128 | const char *name) | |
2129 | { | |
2130 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2131 | struct btrfs_root *root; | |
f7238e50 | 2132 | u64 max_global_id = 0; |
abed4aaa JB |
2133 | int ret; |
2134 | struct btrfs_key key = { | |
2135 | .objectid = objectid, | |
2136 | .type = BTRFS_ROOT_ITEM_KEY, | |
2137 | .offset = 0, | |
2138 | }; | |
2139 | bool found = false; | |
2140 | ||
2141 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2142 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2143 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2144 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2145 | return 0; | |
2146 | } | |
2147 | ||
2148 | while (1) { | |
2149 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2150 | if (ret < 0) | |
2151 | break; | |
2152 | ||
2153 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2154 | ret = btrfs_next_leaf(tree_root, path); | |
2155 | if (ret) { | |
2156 | if (ret > 0) | |
2157 | ret = 0; | |
2158 | break; | |
2159 | } | |
2160 | } | |
2161 | ret = 0; | |
2162 | ||
2163 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2164 | if (key.objectid != objectid) | |
2165 | break; | |
2166 | btrfs_release_path(path); | |
2167 | ||
f7238e50 JB |
2168 | /* |
2169 | * Just worry about this for extent tree, it'll be the same for | |
2170 | * everybody. | |
2171 | */ | |
2172 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2173 | max_global_id = max(max_global_id, key.offset); | |
2174 | ||
abed4aaa JB |
2175 | found = true; |
2176 | root = read_tree_root_path(tree_root, path, &key); | |
2177 | if (IS_ERR(root)) { | |
2178 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2179 | ret = PTR_ERR(root); | |
2180 | break; | |
2181 | } | |
2182 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2183 | ret = btrfs_global_root_insert(root); | |
2184 | if (ret) { | |
2185 | btrfs_put_root(root); | |
2186 | break; | |
2187 | } | |
2188 | key.offset++; | |
2189 | } | |
2190 | btrfs_release_path(path); | |
2191 | ||
f7238e50 JB |
2192 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2193 | fs_info->nr_global_roots = max_global_id + 1; | |
2194 | ||
abed4aaa JB |
2195 | if (!found || ret) { |
2196 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2197 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2198 | ||
2199 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2200 | ret = ret ? ret : -ENOENT; | |
2201 | else | |
2202 | ret = 0; | |
2203 | btrfs_err(fs_info, "failed to load root %s", name); | |
2204 | } | |
2205 | return ret; | |
2206 | } | |
2207 | ||
2208 | static int load_global_roots(struct btrfs_root *tree_root) | |
2209 | { | |
2210 | struct btrfs_path *path; | |
2211 | int ret = 0; | |
2212 | ||
2213 | path = btrfs_alloc_path(); | |
2214 | if (!path) | |
2215 | return -ENOMEM; | |
2216 | ||
2217 | ret = load_global_roots_objectid(tree_root, path, | |
2218 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2219 | if (ret) | |
2220 | goto out; | |
2221 | ret = load_global_roots_objectid(tree_root, path, | |
2222 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2223 | if (ret) | |
2224 | goto out; | |
2225 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2226 | goto out; | |
2227 | ret = load_global_roots_objectid(tree_root, path, | |
2228 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2229 | "free space"); | |
2230 | out: | |
2231 | btrfs_free_path(path); | |
2232 | return ret; | |
2233 | } | |
2234 | ||
6bccf3ab | 2235 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2236 | { |
6bccf3ab | 2237 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2238 | struct btrfs_root *root; |
4bbcaa64 ES |
2239 | struct btrfs_key location; |
2240 | int ret; | |
2241 | ||
2467d0fe | 2242 | ASSERT(fs_info->tree_root); |
6bccf3ab | 2243 | |
abed4aaa JB |
2244 | ret = load_global_roots(tree_root); |
2245 | if (ret) | |
2246 | return ret; | |
2247 | ||
4bbcaa64 ES |
2248 | location.type = BTRFS_ROOT_ITEM_KEY; |
2249 | location.offset = 0; | |
2250 | ||
1c56ab99 | 2251 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2252 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2253 | root = btrfs_read_tree_root(tree_root, &location); | |
2254 | if (IS_ERR(root)) { | |
2255 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2256 | ret = PTR_ERR(root); | |
2257 | goto out; | |
2258 | } | |
2259 | } else { | |
2260 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2261 | fs_info->block_group_root = root; | |
2262 | } | |
2263 | } | |
2264 | ||
2265 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2266 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2267 | if (IS_ERR(root)) { |
42437a63 JB |
2268 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2269 | ret = PTR_ERR(root); | |
2270 | goto out; | |
2271 | } | |
2272 | } else { | |
2273 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2274 | fs_info->dev_root = root; | |
f50f4353 | 2275 | } |
820a49da | 2276 | /* Initialize fs_info for all devices in any case */ |
a8d1b164 JT |
2277 | ret = btrfs_init_devices_late(fs_info); |
2278 | if (ret) | |
2279 | goto out; | |
4bbcaa64 | 2280 | |
aeb935a4 QW |
2281 | /* |
2282 | * This tree can share blocks with some other fs tree during relocation | |
2283 | * and we need a proper setup by btrfs_get_fs_root | |
2284 | */ | |
56e9357a DS |
2285 | root = btrfs_get_fs_root(tree_root->fs_info, |
2286 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2287 | if (IS_ERR(root)) { |
42437a63 JB |
2288 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2289 | ret = PTR_ERR(root); | |
2290 | goto out; | |
2291 | } | |
2292 | } else { | |
2293 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2294 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2295 | } |
aeb935a4 | 2296 | |
4bbcaa64 | 2297 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2298 | root = btrfs_read_tree_root(tree_root, &location); |
2299 | if (!IS_ERR(root)) { | |
2300 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
a4f3d2c4 | 2301 | fs_info->quota_root = root; |
4bbcaa64 ES |
2302 | } |
2303 | ||
2304 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2305 | root = btrfs_read_tree_root(tree_root, &location); |
2306 | if (IS_ERR(root)) { | |
42437a63 JB |
2307 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2308 | ret = PTR_ERR(root); | |
2309 | if (ret != -ENOENT) | |
2310 | goto out; | |
2311 | } | |
4bbcaa64 | 2312 | } else { |
a4f3d2c4 DS |
2313 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2314 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2315 | } |
2316 | ||
51502090 JT |
2317 | if (btrfs_fs_incompat(fs_info, RAID_STRIPE_TREE)) { |
2318 | location.objectid = BTRFS_RAID_STRIPE_TREE_OBJECTID; | |
2319 | root = btrfs_read_tree_root(tree_root, &location); | |
2320 | if (IS_ERR(root)) { | |
2321 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2322 | ret = PTR_ERR(root); | |
2323 | goto out; | |
2324 | } | |
2325 | } else { | |
2326 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2327 | fs_info->stripe_root = root; | |
2328 | } | |
2329 | } | |
2330 | ||
4bbcaa64 | 2331 | return 0; |
f50f4353 LB |
2332 | out: |
2333 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2334 | location.objectid, ret); | |
2335 | return ret; | |
4bbcaa64 ES |
2336 | } |
2337 | ||
069ec957 QW |
2338 | /* |
2339 | * Real super block validation | |
2340 | * NOTE: super csum type and incompat features will not be checked here. | |
2341 | * | |
2342 | * @sb: super block to check | |
2343 | * @mirror_num: the super block number to check its bytenr: | |
2344 | * 0 the primary (1st) sb | |
2345 | * 1, 2 2nd and 3rd backup copy | |
2346 | * -1 skip bytenr check | |
2347 | */ | |
a05d3c91 QW |
2348 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2349 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2350 | { |
21a852b0 QW |
2351 | u64 nodesize = btrfs_super_nodesize(sb); |
2352 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2353 | int ret = 0; | |
2354 | ||
2355 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2356 | btrfs_err(fs_info, "no valid FS found"); | |
2357 | ret = -EINVAL; | |
2358 | } | |
2359 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2360 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2361 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2362 | ret = -EINVAL; | |
2363 | } | |
2364 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2365 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2366 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2367 | ret = -EINVAL; | |
2368 | } | |
2369 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2370 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2371 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2372 | ret = -EINVAL; | |
2373 | } | |
2374 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2375 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2376 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2377 | ret = -EINVAL; | |
2378 | } | |
2379 | ||
2380 | /* | |
2381 | * Check sectorsize and nodesize first, other check will need it. | |
2382 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2383 | */ | |
2384 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2385 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2386 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2387 | ret = -EINVAL; | |
2388 | } | |
0bb3eb3e QW |
2389 | |
2390 | /* | |
1a42daab QW |
2391 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2392 | * | |
2393 | * We can support 16K sectorsize with 64K page size without problem, | |
2394 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2395 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2396 | * beginning. | |
0bb3eb3e | 2397 | */ |
1a42daab | 2398 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2399 | btrfs_err(fs_info, |
0bb3eb3e | 2400 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2401 | sectorsize, PAGE_SIZE); |
2402 | ret = -EINVAL; | |
2403 | } | |
0bb3eb3e | 2404 | |
21a852b0 QW |
2405 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2406 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2407 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2408 | ret = -EINVAL; | |
2409 | } | |
2410 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2411 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2412 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2413 | ret = -EINVAL; | |
2414 | } | |
2415 | ||
2416 | /* Root alignment check */ | |
2417 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2418 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2419 | btrfs_super_root(sb)); | |
2420 | ret = -EINVAL; | |
2421 | } | |
2422 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2423 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2424 | btrfs_super_chunk_root(sb)); | |
2425 | ret = -EINVAL; | |
2426 | } | |
2427 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2428 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2429 | btrfs_super_log_root(sb)); | |
2430 | ret = -EINVAL; | |
2431 | } | |
2432 | ||
a5b8a5f9 AJ |
2433 | if (!fs_info->fs_devices->temp_fsid && |
2434 | memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2435 | btrfs_err(fs_info, |
2436 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
d167aa76 | 2437 | sb->fsid, fs_info->fs_devices->fsid); |
aefd7f70 NB |
2438 | ret = -EINVAL; |
2439 | } | |
2440 | ||
6bfe3959 AJ |
2441 | if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), |
2442 | BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2443 | btrfs_err(fs_info, |
2444 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
6bfe3959 | 2445 | btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); |
aefd7f70 NB |
2446 | ret = -EINVAL; |
2447 | } | |
2448 | ||
25984a5a AJ |
2449 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
2450 | BTRFS_FSID_SIZE) != 0) { | |
2451 | btrfs_err(fs_info, | |
2452 | "dev_item UUID does not match metadata fsid: %pU != %pU", | |
2453 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); | |
2454 | ret = -EINVAL; | |
2455 | } | |
2456 | ||
1c56ab99 QW |
2457 | /* |
2458 | * Artificial requirement for block-group-tree to force newer features | |
2459 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2460 | */ | |
2461 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2462 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2463 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2464 | btrfs_err(fs_info, | |
2465 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2466 | ret = -EINVAL; | |
2467 | } | |
2468 | ||
21a852b0 QW |
2469 | /* |
2470 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2471 | * done later | |
2472 | */ | |
2473 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2474 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2475 | btrfs_super_bytes_used(sb)); | |
2476 | ret = -EINVAL; | |
2477 | } | |
2478 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2479 | btrfs_err(fs_info, "invalid stripesize %u", | |
2480 | btrfs_super_stripesize(sb)); | |
2481 | ret = -EINVAL; | |
2482 | } | |
2483 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2484 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2485 | btrfs_super_num_devices(sb)); | |
2486 | if (btrfs_super_num_devices(sb) == 0) { | |
2487 | btrfs_err(fs_info, "number of devices is 0"); | |
2488 | ret = -EINVAL; | |
2489 | } | |
2490 | ||
069ec957 QW |
2491 | if (mirror_num >= 0 && |
2492 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2493 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2494 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2495 | ret = -EINVAL; | |
2496 | } | |
2497 | ||
2498 | /* | |
2499 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2500 | * and one chunk | |
2501 | */ | |
2502 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2503 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2504 | btrfs_super_sys_array_size(sb), | |
2505 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2506 | ret = -EINVAL; | |
2507 | } | |
2508 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2509 | + sizeof(struct btrfs_chunk)) { | |
2510 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2511 | btrfs_super_sys_array_size(sb), | |
2512 | sizeof(struct btrfs_disk_key) | |
2513 | + sizeof(struct btrfs_chunk)); | |
2514 | ret = -EINVAL; | |
2515 | } | |
2516 | ||
2517 | /* | |
2518 | * The generation is a global counter, we'll trust it more than the others | |
2519 | * but it's still possible that it's the one that's wrong. | |
2520 | */ | |
2521 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2522 | btrfs_warn(fs_info, | |
2523 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2524 | btrfs_super_generation(sb), | |
2525 | btrfs_super_chunk_root_generation(sb)); | |
2526 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2527 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2528 | btrfs_warn(fs_info, | |
2529 | "suspicious: generation < cache_generation: %llu < %llu", | |
2530 | btrfs_super_generation(sb), | |
2531 | btrfs_super_cache_generation(sb)); | |
2532 | ||
2533 | return ret; | |
2534 | } | |
2535 | ||
069ec957 QW |
2536 | /* |
2537 | * Validation of super block at mount time. | |
2538 | * Some checks already done early at mount time, like csum type and incompat | |
2539 | * flags will be skipped. | |
2540 | */ | |
2541 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2542 | { | |
a05d3c91 | 2543 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2544 | } |
2545 | ||
75cb857d QW |
2546 | /* |
2547 | * Validation of super block at write time. | |
2548 | * Some checks like bytenr check will be skipped as their values will be | |
2549 | * overwritten soon. | |
2550 | * Extra checks like csum type and incompat flags will be done here. | |
2551 | */ | |
2552 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2553 | struct btrfs_super_block *sb) | |
2554 | { | |
2555 | int ret; | |
2556 | ||
a05d3c91 | 2557 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2558 | if (ret < 0) |
2559 | goto out; | |
e7e16f48 | 2560 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2561 | ret = -EUCLEAN; |
2562 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2563 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2564 | goto out; | |
2565 | } | |
2566 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2567 | ret = -EUCLEAN; | |
2568 | btrfs_err(fs_info, | |
2569 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2570 | btrfs_super_incompat_flags(sb), | |
2571 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2572 | goto out; | |
2573 | } | |
2574 | out: | |
2575 | if (ret < 0) | |
2576 | btrfs_err(fs_info, | |
2577 | "super block corruption detected before writing it to disk"); | |
2578 | return ret; | |
2579 | } | |
2580 | ||
bd676446 JB |
2581 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2582 | { | |
789d6a3a QW |
2583 | struct btrfs_tree_parent_check check = { |
2584 | .level = level, | |
2585 | .transid = gen, | |
2586 | .owner_root = root->root_key.objectid | |
2587 | }; | |
bd676446 JB |
2588 | int ret = 0; |
2589 | ||
789d6a3a | 2590 | root->node = read_tree_block(root->fs_info, bytenr, &check); |
bd676446 JB |
2591 | if (IS_ERR(root->node)) { |
2592 | ret = PTR_ERR(root->node); | |
2593 | root->node = NULL; | |
4eb150d6 QW |
2594 | return ret; |
2595 | } | |
2596 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2597 | free_extent_buffer(root->node); |
2598 | root->node = NULL; | |
4eb150d6 | 2599 | return -EIO; |
bd676446 JB |
2600 | } |
2601 | ||
bd676446 JB |
2602 | btrfs_set_root_node(&root->root_item, root->node); |
2603 | root->commit_root = btrfs_root_node(root); | |
2604 | btrfs_set_root_refs(&root->root_item, 1); | |
2605 | return ret; | |
2606 | } | |
2607 | ||
2608 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2609 | { | |
2610 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2611 | u64 gen, bytenr; | |
2612 | int level, ret; | |
2613 | ||
2614 | bytenr = btrfs_super_root(sb); | |
2615 | gen = btrfs_super_generation(sb); | |
2616 | level = btrfs_super_root_level(sb); | |
2617 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2618 | if (ret) { |
bd676446 | 2619 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2620 | return ret; |
2621 | } | |
14033b08 | 2622 | return 0; |
bd676446 JB |
2623 | } |
2624 | ||
6ef108dd | 2625 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2626 | { |
6ef108dd | 2627 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2628 | struct btrfs_super_block *sb = fs_info->super_copy; |
2629 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2630 | bool handle_error = false; | |
2631 | int ret = 0; | |
2632 | int i; | |
2633 | ||
2634 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2635 | if (handle_error) { |
2636 | if (!IS_ERR(tree_root->node)) | |
2637 | free_extent_buffer(tree_root->node); | |
2638 | tree_root->node = NULL; | |
2639 | ||
2640 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2641 | break; | |
2642 | ||
2643 | free_root_pointers(fs_info, 0); | |
2644 | ||
2645 | /* | |
2646 | * Don't use the log in recovery mode, it won't be | |
2647 | * valid | |
2648 | */ | |
2649 | btrfs_set_super_log_root(sb, 0); | |
2650 | ||
745806fb | 2651 | btrfs_warn(fs_info, "try to load backup roots slot %d", i); |
b8522a1e | 2652 | ret = read_backup_root(fs_info, i); |
6ef108dd | 2653 | backup_index = ret; |
b8522a1e NB |
2654 | if (ret < 0) |
2655 | return ret; | |
2656 | } | |
b8522a1e | 2657 | |
bd676446 JB |
2658 | ret = load_important_roots(fs_info); |
2659 | if (ret) { | |
217f5004 | 2660 | handle_error = true; |
b8522a1e NB |
2661 | continue; |
2662 | } | |
2663 | ||
336a0d8d NB |
2664 | /* |
2665 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2666 | * hasn't been fully initialised and we are the only user | |
2667 | */ | |
453e4873 | 2668 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2669 | if (ret < 0) { |
b8522a1e NB |
2670 | handle_error = true; |
2671 | continue; | |
2672 | } | |
2673 | ||
6b8fad57 | 2674 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2675 | |
2676 | ret = btrfs_read_roots(fs_info); | |
2677 | if (ret < 0) { | |
2678 | handle_error = true; | |
2679 | continue; | |
2680 | } | |
2681 | ||
2682 | /* All successful */ | |
bd676446 | 2683 | fs_info->generation = btrfs_header_generation(tree_root->node); |
0124855f | 2684 | btrfs_set_last_trans_committed(fs_info, fs_info->generation); |
d96b3424 | 2685 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2686 | |
2687 | /* Always begin writing backup roots after the one being used */ | |
2688 | if (backup_index < 0) { | |
2689 | fs_info->backup_root_index = 0; | |
2690 | } else { | |
2691 | fs_info->backup_root_index = backup_index + 1; | |
2692 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2693 | } | |
b8522a1e NB |
2694 | break; |
2695 | } | |
2696 | ||
2697 | return ret; | |
2698 | } | |
2699 | ||
8260edba | 2700 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2701 | { |
fc7cbcd4 | 2702 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2703 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2704 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2705 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2706 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2707 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2708 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2709 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2710 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2711 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2712 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2713 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2714 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2715 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2716 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2717 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2718 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2719 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2720 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2721 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2722 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2723 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2724 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2725 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2726 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2727 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2728 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2729 | |
e1489b4f | 2730 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2731 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2732 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2733 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
77d20c68 JB |
2734 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_prep, |
2735 | BTRFS_LOCKDEP_TRANS_COMMIT_PREP); | |
3e738c53 IA |
2736 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, |
2737 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
2738 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
2739 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
2740 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
2741 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 2742 | |
0b86a832 | 2743 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2744 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2745 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2746 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2747 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 2748 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
2749 | #ifdef CONFIG_BTRFS_DEBUG |
2750 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2751 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2752 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2753 | #endif |
7dc66abb FM |
2754 | fs_info->mapping_tree = RB_ROOT_CACHED; |
2755 | rwlock_init(&fs_info->mapping_tree_lock); | |
66d8f3dd MX |
2756 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2757 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2758 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2759 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2760 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2761 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2762 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2763 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2764 | BTRFS_BLOCK_RSV_DELREFS); | |
2765 | ||
771ed689 | 2766 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2767 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 2768 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2769 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 2770 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 2771 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2772 | fs_info->metadata_ratio = 0; |
4cb5300b | 2773 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2774 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2775 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2776 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
fd708b81 | 2777 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2778 | |
b34b086c CM |
2779 | fs_info->thread_pool_size = min_t(unsigned long, |
2780 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2781 | |
199c2a9c MX |
2782 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2783 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2784 | |
638aa7ed | 2785 | btrfs_init_scrub(fs_info); |
779a65a4 | 2786 | btrfs_init_balance(fs_info); |
57056740 | 2787 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2788 | |
16b0c258 | 2789 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 2790 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 2791 | |
fe119a6e | 2792 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
35da5a7e | 2793 | IO_TREE_FS_EXCLUDED_EXTENTS); |
39279cc3 | 2794 | |
5a3f23d5 | 2795 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2796 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2797 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2798 | mutex_init(&fs_info->transaction_kthread_mutex); |
2799 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2800 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2801 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2802 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2803 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2804 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2805 | |
ad618368 | 2806 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2807 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2808 | btrfs_discard_init(fs_info); |
416ac51d | 2809 | |
fa9c0d79 CM |
2810 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2811 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2812 | ||
e6dcd2dc | 2813 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2814 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2815 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2816 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2817 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2818 | |
da17066c JM |
2819 | /* Usable values until the real ones are cached from the superblock */ |
2820 | fs_info->nodesize = 4096; | |
2821 | fs_info->sectorsize = 4096; | |
ab108d99 | 2822 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
2823 | fs_info->stripesize = 4096; |
2824 | ||
6207c9e3 JB |
2825 | /* Default compress algorithm when user does -o compress */ |
2826 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2827 | ||
f7b12a62 NA |
2828 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
2829 | ||
eede2bf3 OS |
2830 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2831 | fs_info->swapfile_pins = RB_ROOT; | |
2832 | ||
18bb8bbf JT |
2833 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
2834 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
2835 | } |
2836 | ||
2837 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2838 | { | |
2839 | int ret; | |
2840 | ||
2841 | fs_info->sb = sb; | |
4e00422e | 2842 | /* Temporary fixed values for block size until we read the superblock. */ |
8260edba JB |
2843 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2844 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 2845 | |
5deb17e1 | 2846 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 2847 | if (ret) |
c75e8394 | 2848 | return ret; |
ae18c37a JB |
2849 | |
2850 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
2851 | if (ret) | |
c75e8394 | 2852 | return ret; |
ae18c37a JB |
2853 | |
2854 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
2855 | (1 + ilog2(nr_cpu_ids)); | |
2856 | ||
2857 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
2858 | if (ret) | |
c75e8394 | 2859 | return ret; |
ae18c37a JB |
2860 | |
2861 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
2862 | GFP_KERNEL); | |
2863 | if (ret) | |
c75e8394 | 2864 | return ret; |
ae18c37a JB |
2865 | |
2866 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
2867 | GFP_KERNEL); | |
c75e8394 JB |
2868 | if (!fs_info->delayed_root) |
2869 | return -ENOMEM; | |
ae18c37a JB |
2870 | btrfs_init_delayed_root(fs_info->delayed_root); |
2871 | ||
a0a1db70 FM |
2872 | if (sb_rdonly(sb)) |
2873 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
2874 | ||
c75e8394 | 2875 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
2876 | } |
2877 | ||
97f4dd09 NB |
2878 | static int btrfs_uuid_rescan_kthread(void *data) |
2879 | { | |
0d031dc4 | 2880 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
2881 | int ret; |
2882 | ||
2883 | /* | |
2884 | * 1st step is to iterate through the existing UUID tree and | |
2885 | * to delete all entries that contain outdated data. | |
2886 | * 2nd step is to add all missing entries to the UUID tree. | |
2887 | */ | |
2888 | ret = btrfs_uuid_tree_iterate(fs_info); | |
2889 | if (ret < 0) { | |
c94bec2c JB |
2890 | if (ret != -EINTR) |
2891 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
2892 | ret); | |
97f4dd09 NB |
2893 | up(&fs_info->uuid_tree_rescan_sem); |
2894 | return ret; | |
2895 | } | |
2896 | return btrfs_uuid_scan_kthread(data); | |
2897 | } | |
2898 | ||
2899 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
2900 | { | |
2901 | struct task_struct *task; | |
2902 | ||
2903 | down(&fs_info->uuid_tree_rescan_sem); | |
2904 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
2905 | if (IS_ERR(task)) { | |
2906 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
2907 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
2908 | up(&fs_info->uuid_tree_rescan_sem); | |
2909 | return PTR_ERR(task); | |
2910 | } | |
2911 | ||
2912 | return 0; | |
2913 | } | |
2914 | ||
504b1596 FM |
2915 | static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
2916 | { | |
2917 | u64 root_objectid = 0; | |
2918 | struct btrfs_root *gang[8]; | |
2919 | int i = 0; | |
2920 | int err = 0; | |
2921 | unsigned int ret = 0; | |
2922 | ||
2923 | while (1) { | |
2924 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2925 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2926 | (void **)gang, root_objectid, | |
2927 | ARRAY_SIZE(gang)); | |
2928 | if (!ret) { | |
2929 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2930 | break; | |
2931 | } | |
2932 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
2933 | ||
2934 | for (i = 0; i < ret; i++) { | |
2935 | /* Avoid to grab roots in dead_roots. */ | |
2936 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
2937 | gang[i] = NULL; | |
2938 | continue; | |
2939 | } | |
2940 | /* Grab all the search result for later use. */ | |
2941 | gang[i] = btrfs_grab_root(gang[i]); | |
2942 | } | |
2943 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2944 | ||
2945 | for (i = 0; i < ret; i++) { | |
2946 | if (!gang[i]) | |
2947 | continue; | |
2948 | root_objectid = gang[i]->root_key.objectid; | |
2949 | err = btrfs_orphan_cleanup(gang[i]); | |
2950 | if (err) | |
2951 | goto out; | |
2952 | btrfs_put_root(gang[i]); | |
2953 | } | |
2954 | root_objectid++; | |
2955 | } | |
2956 | out: | |
2957 | /* Release the uncleaned roots due to error. */ | |
2958 | for (; i < ret; i++) { | |
2959 | if (gang[i]) | |
2960 | btrfs_put_root(gang[i]); | |
2961 | } | |
2962 | return err; | |
2963 | } | |
2964 | ||
44c0ca21 BB |
2965 | /* |
2966 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
2967 | * and btrfs_remount when remounting from read-only to read-write. | |
2968 | */ | |
2969 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
2970 | { | |
2971 | int ret; | |
94846229 | 2972 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
1d6a4fc8 | 2973 | bool rebuild_free_space_tree = false; |
8b228324 BB |
2974 | |
2975 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
2976 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
272efa30 JB |
2977 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) |
2978 | btrfs_warn(fs_info, | |
2979 | "'clear_cache' option is ignored with extent tree v2"); | |
2980 | else | |
2981 | rebuild_free_space_tree = true; | |
8b228324 BB |
2982 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && |
2983 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
2984 | btrfs_warn(fs_info, "free space tree is invalid"); | |
1d6a4fc8 | 2985 | rebuild_free_space_tree = true; |
8b228324 BB |
2986 | } |
2987 | ||
1d6a4fc8 QW |
2988 | if (rebuild_free_space_tree) { |
2989 | btrfs_info(fs_info, "rebuilding free space tree"); | |
2990 | ret = btrfs_rebuild_free_space_tree(fs_info); | |
8b228324 BB |
2991 | if (ret) { |
2992 | btrfs_warn(fs_info, | |
1d6a4fc8 QW |
2993 | "failed to rebuild free space tree: %d", ret); |
2994 | goto out; | |
2995 | } | |
2996 | } | |
2997 | ||
2998 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
2999 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { | |
3000 | btrfs_info(fs_info, "disabling free space tree"); | |
3001 | ret = btrfs_delete_free_space_tree(fs_info); | |
3002 | if (ret) { | |
3003 | btrfs_warn(fs_info, | |
3004 | "failed to disable free space tree: %d", ret); | |
8b228324 BB |
3005 | goto out; |
3006 | } | |
3007 | } | |
44c0ca21 | 3008 | |
8d488a8c FM |
3009 | /* |
3010 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3011 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3012 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3013 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3014 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3015 | * item before the root's tree is deleted - this means that if we unmount | |
3016 | * or crash before the deletion completes, on the next mount we will not | |
3017 | * delete what remains of the tree because the orphan item does not | |
3018 | * exists anymore, which is what tells us we have a pending deletion. | |
3019 | */ | |
3020 | ret = btrfs_find_orphan_roots(fs_info); | |
3021 | if (ret) | |
3022 | goto out; | |
3023 | ||
44c0ca21 BB |
3024 | ret = btrfs_cleanup_fs_roots(fs_info); |
3025 | if (ret) | |
3026 | goto out; | |
3027 | ||
8f1c21d7 BB |
3028 | down_read(&fs_info->cleanup_work_sem); |
3029 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3030 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3031 | up_read(&fs_info->cleanup_work_sem); | |
3032 | goto out; | |
3033 | } | |
3034 | up_read(&fs_info->cleanup_work_sem); | |
3035 | ||
44c0ca21 | 3036 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3037 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3038 | mutex_unlock(&fs_info->cleaner_mutex); |
3039 | if (ret < 0) { | |
3040 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3041 | goto out; | |
3042 | } | |
3043 | ||
5011139a BB |
3044 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3045 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3046 | btrfs_info(fs_info, "creating free space tree"); | |
3047 | ret = btrfs_create_free_space_tree(fs_info); | |
3048 | if (ret) { | |
3049 | btrfs_warn(fs_info, | |
3050 | "failed to create free space tree: %d", ret); | |
3051 | goto out; | |
3052 | } | |
3053 | } | |
3054 | ||
94846229 BB |
3055 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3056 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3057 | if (ret) | |
3058 | goto out; | |
3059 | } | |
3060 | ||
44c0ca21 BB |
3061 | ret = btrfs_resume_balance_async(fs_info); |
3062 | if (ret) | |
3063 | goto out; | |
3064 | ||
3065 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3066 | if (ret) { | |
3067 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3068 | goto out; | |
3069 | } | |
3070 | ||
3071 | btrfs_qgroup_rescan_resume(fs_info); | |
3072 | ||
3073 | if (!fs_info->uuid_root) { | |
3074 | btrfs_info(fs_info, "creating UUID tree"); | |
3075 | ret = btrfs_create_uuid_tree(fs_info); | |
3076 | if (ret) { | |
3077 | btrfs_warn(fs_info, | |
3078 | "failed to create the UUID tree %d", ret); | |
3079 | goto out; | |
3080 | } | |
3081 | } | |
3082 | ||
3083 | out: | |
3084 | return ret; | |
3085 | } | |
3086 | ||
d7f67ac9 QW |
3087 | /* |
3088 | * Do various sanity and dependency checks of different features. | |
3089 | * | |
2ba48b20 QW |
3090 | * @is_rw_mount: If the mount is read-write. |
3091 | * | |
d7f67ac9 QW |
3092 | * This is the place for less strict checks (like for subpage or artificial |
3093 | * feature dependencies). | |
3094 | * | |
3095 | * For strict checks or possible corruption detection, see | |
3096 | * btrfs_validate_super(). | |
3097 | * | |
3098 | * This should be called after btrfs_parse_options(), as some mount options | |
3099 | * (space cache related) can modify on-disk format like free space tree and | |
3100 | * screw up certain feature dependencies. | |
3101 | */ | |
2ba48b20 | 3102 | int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) |
d7f67ac9 QW |
3103 | { |
3104 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3105 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3106 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3107 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3108 | ||
3109 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3110 | btrfs_err(fs_info, | |
3111 | "cannot mount because of unknown incompat features (0x%llx)", | |
3112 | incompat); | |
3113 | return -EINVAL; | |
3114 | } | |
3115 | ||
3116 | /* Runtime limitation for mixed block groups. */ | |
3117 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3118 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3119 | btrfs_err(fs_info, | |
3120 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3121 | fs_info->nodesize, fs_info->sectorsize); | |
3122 | return -EINVAL; | |
3123 | } | |
3124 | ||
3125 | /* Mixed backref is an always-enabled feature. */ | |
3126 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3127 | ||
3128 | /* Set compression related flags just in case. */ | |
3129 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3130 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3131 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3132 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3133 | ||
3134 | /* | |
3135 | * An ancient flag, which should really be marked deprecated. | |
3136 | * Such runtime limitation doesn't really need a incompat flag. | |
3137 | */ | |
3138 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3139 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3140 | ||
2ba48b20 | 3141 | if (compat_ro_unsupp && is_rw_mount) { |
d7f67ac9 QW |
3142 | btrfs_err(fs_info, |
3143 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3144 | compat_ro); | |
3145 | return -EINVAL; | |
3146 | } | |
3147 | ||
3148 | /* | |
3149 | * We have unsupported RO compat features, although RO mounted, we | |
3150 | * should not cause any metadata writes, including log replay. | |
3151 | * Or we could screw up whatever the new feature requires. | |
3152 | */ | |
3153 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3154 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3155 | btrfs_err(fs_info, | |
3156 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3157 | compat_ro); | |
3158 | return -EINVAL; | |
3159 | } | |
3160 | ||
3161 | /* | |
3162 | * Artificial limitations for block group tree, to force | |
3163 | * block-group-tree to rely on no-holes and free-space-tree. | |
3164 | */ | |
3165 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3166 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3167 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3168 | btrfs_err(fs_info, | |
3169 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3170 | return -EINVAL; | |
3171 | } | |
3172 | ||
3173 | /* | |
3174 | * Subpage runtime limitation on v1 cache. | |
3175 | * | |
3176 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3177 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3178 | * going to be deprecated anyway. | |
3179 | */ | |
3180 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3181 | btrfs_warn(fs_info, | |
3182 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3183 | PAGE_SIZE, fs_info->sectorsize); | |
3184 | return -EINVAL; | |
3185 | } | |
3186 | ||
3187 | /* This can be called by remount, we need to protect the super block. */ | |
3188 | spin_lock(&fs_info->super_lock); | |
3189 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3190 | spin_unlock(&fs_info->super_lock); | |
3191 | ||
3192 | return 0; | |
3193 | } | |
3194 | ||
ae18c37a JB |
3195 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3196 | char *options) | |
3197 | { | |
3198 | u32 sectorsize; | |
3199 | u32 nodesize; | |
3200 | u32 stripesize; | |
3201 | u64 generation; | |
ae18c37a | 3202 | u16 csum_type; |
ae18c37a JB |
3203 | struct btrfs_super_block *disk_super; |
3204 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3205 | struct btrfs_root *tree_root; | |
3206 | struct btrfs_root *chunk_root; | |
3207 | int ret; | |
ae18c37a JB |
3208 | int level; |
3209 | ||
8260edba | 3210 | ret = init_mount_fs_info(fs_info, sb); |
4871c33b | 3211 | if (ret) |
ae18c37a | 3212 | goto fail; |
53b381b3 | 3213 | |
ae18c37a JB |
3214 | /* These need to be init'ed before we start creating inodes and such. */ |
3215 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3216 | GFP_KERNEL); | |
3217 | fs_info->tree_root = tree_root; | |
3218 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3219 | GFP_KERNEL); | |
3220 | fs_info->chunk_root = chunk_root; | |
3221 | if (!tree_root || !chunk_root) { | |
4871c33b | 3222 | ret = -ENOMEM; |
c75e8394 | 3223 | goto fail; |
ae18c37a JB |
3224 | } |
3225 | ||
dcb2137c | 3226 | ret = btrfs_init_btree_inode(sb); |
4871c33b | 3227 | if (ret) |
c75e8394 | 3228 | goto fail; |
ae18c37a | 3229 | |
d24fa5c1 | 3230 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3231 | |
3232 | /* | |
3233 | * Read super block and check the signature bytes only | |
3234 | */ | |
d24fa5c1 | 3235 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d | 3236 | if (IS_ERR(disk_super)) { |
4871c33b | 3237 | ret = PTR_ERR(disk_super); |
16cdcec7 | 3238 | goto fail_alloc; |
20b45077 | 3239 | } |
39279cc3 | 3240 | |
2db31320 | 3241 | btrfs_info(fs_info, "first mount of filesystem %pU", disk_super->fsid); |
8dc3f22c | 3242 | /* |
260db43c | 3243 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3244 | * corrupted, we'll find out |
3245 | */ | |
8f32380d | 3246 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3247 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3248 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3249 | csum_type); |
4871c33b | 3250 | ret = -EINVAL; |
8f32380d | 3251 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3252 | goto fail_alloc; |
3253 | } | |
3254 | ||
83c68bbc SY |
3255 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3256 | ||
6d97c6e3 JT |
3257 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3258 | if (ret) { | |
8f32380d | 3259 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3260 | goto fail_alloc; |
3261 | } | |
3262 | ||
1104a885 DS |
3263 | /* |
3264 | * We want to check superblock checksum, the type is stored inside. | |
3265 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3266 | */ | |
3d17adea | 3267 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3268 | btrfs_err(fs_info, "superblock checksum mismatch"); |
4871c33b | 3269 | ret = -EINVAL; |
8f32380d | 3270 | btrfs_release_disk_super(disk_super); |
141386e1 | 3271 | goto fail_alloc; |
1104a885 DS |
3272 | } |
3273 | ||
3274 | /* | |
3275 | * super_copy is zeroed at allocation time and we never touch the | |
3276 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3277 | * the whole block of INFO_SIZE | |
3278 | */ | |
8f32380d JT |
3279 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3280 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3281 | |
fbc6feae NB |
3282 | disk_super = fs_info->super_copy; |
3283 | ||
fbc6feae NB |
3284 | memcpy(fs_info->super_for_commit, fs_info->super_copy, |
3285 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3286 | |
069ec957 | 3287 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3288 | if (ret) { |
05135f59 | 3289 | btrfs_err(fs_info, "superblock contains fatal errors"); |
4871c33b | 3290 | ret = -EINVAL; |
141386e1 | 3291 | goto fail_alloc; |
1104a885 DS |
3292 | } |
3293 | ||
4871c33b QW |
3294 | if (!btrfs_super_root(disk_super)) { |
3295 | btrfs_err(fs_info, "invalid superblock tree root bytenr"); | |
3296 | ret = -EINVAL; | |
141386e1 | 3297 | goto fail_alloc; |
4871c33b | 3298 | } |
0f7d52f4 | 3299 | |
acce952b | 3300 | /* check FS state, whether FS is broken. */ |
87533c47 | 3301 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
ae3364e5 | 3302 | WRITE_ONCE(fs_info->fs_error, -EUCLEAN); |
acce952b | 3303 | |
6f93e834 AJ |
3304 | /* Set up fs_info before parsing mount options */ |
3305 | nodesize = btrfs_super_nodesize(disk_super); | |
3306 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3307 | stripesize = sectorsize; | |
3308 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3309 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3310 | ||
3311 | fs_info->nodesize = nodesize; | |
3312 | fs_info->sectorsize = sectorsize; | |
3313 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3314 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3315 | fs_info->stripesize = stripesize; | |
3316 | ||
a6a8f22a JB |
3317 | /* |
3318 | * Handle the space caching options appropriately now that we have the | |
3319 | * super block loaded and validated. | |
3320 | */ | |
3321 | btrfs_set_free_space_cache_settings(fs_info); | |
3322 | ||
ad21f15b JB |
3323 | if (!btrfs_check_options(fs_info, &fs_info->mount_opt, sb->s_flags)) { |
3324 | ret = -EINVAL; | |
141386e1 | 3325 | goto fail_alloc; |
ad21f15b | 3326 | } |
dfe25020 | 3327 | |
2ba48b20 | 3328 | ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); |
4871c33b | 3329 | if (ret < 0) |
dc4d3168 | 3330 | goto fail_alloc; |
dc4d3168 | 3331 | |
ad21f15b JB |
3332 | /* |
3333 | * At this point our mount options are validated, if we set ->max_inline | |
3334 | * to something non-standard make sure we truncate it to sectorsize. | |
3335 | */ | |
3336 | fs_info->max_inline = min_t(u64, fs_info->max_inline, fs_info->sectorsize); | |
3337 | ||
8481dd80 QW |
3338 | if (sectorsize < PAGE_SIZE) { |
3339 | struct btrfs_subpage_info *subpage_info; | |
3340 | ||
95ea0486 QW |
3341 | btrfs_warn(fs_info, |
3342 | "read-write for sector size %u with page size %lu is experimental", | |
3343 | sectorsize, PAGE_SIZE); | |
8481dd80 | 3344 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
4871c33b QW |
3345 | if (!subpage_info) { |
3346 | ret = -ENOMEM; | |
8481dd80 | 3347 | goto fail_alloc; |
4871c33b | 3348 | } |
8481dd80 QW |
3349 | btrfs_init_subpage_info(subpage_info, sectorsize); |
3350 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3351 | } |
0bb3eb3e | 3352 | |
d21deec5 | 3353 | ret = btrfs_init_workqueues(fs_info); |
4871c33b | 3354 | if (ret) |
0dc3b84a | 3355 | goto fail_sb_buffer; |
4543df7e | 3356 | |
9e11ceee JK |
3357 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3358 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3359 | |
4e00422e | 3360 | /* Update the values for the current filesystem. */ |
a061fc8d CM |
3361 | sb->s_blocksize = sectorsize; |
3362 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3363 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3364 | |
925baedd | 3365 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3366 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3367 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3368 | if (ret) { |
05135f59 | 3369 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3370 | goto fail_sb_buffer; |
84eed90f | 3371 | } |
0b86a832 | 3372 | |
84234f3a | 3373 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3374 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3375 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3376 | generation, level); | |
3377 | if (ret) { | |
05135f59 | 3378 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3379 | goto fail_tree_roots; |
83121942 | 3380 | } |
0b86a832 | 3381 | |
e17cade2 | 3382 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3383 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3384 | BTRFS_UUID_SIZE); | |
e17cade2 | 3385 | |
5b4aacef | 3386 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3387 | if (ret) { |
05135f59 | 3388 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3389 | goto fail_tree_roots; |
2b82032c | 3390 | } |
0b86a832 | 3391 | |
8dabb742 | 3392 | /* |
bacce86a AJ |
3393 | * At this point we know all the devices that make this filesystem, |
3394 | * including the seed devices but we don't know yet if the replace | |
3395 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3396 | * filesystem but skip the replace target device which is checked |
bacce86a | 3397 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3398 | */ |
bacce86a | 3399 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3400 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3401 | btrfs_err(fs_info, "failed to read devices"); |
4871c33b | 3402 | ret = -EIO; |
a6b0d5c8 CM |
3403 | goto fail_tree_roots; |
3404 | } | |
3405 | ||
b8522a1e | 3406 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3407 | if (ret) |
b8522a1e | 3408 | goto fail_tree_roots; |
8929ecfa | 3409 | |
73651042 NA |
3410 | /* |
3411 | * Get zone type information of zoned block devices. This will also | |
3412 | * handle emulation of a zoned filesystem if a regular device has the | |
3413 | * zoned incompat feature flag set. | |
3414 | */ | |
3415 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3416 | if (ret) { | |
3417 | btrfs_err(fs_info, | |
4871c33b | 3418 | "zoned: failed to read device zone info: %d", ret); |
73651042 NA |
3419 | goto fail_block_groups; |
3420 | } | |
3421 | ||
75ec1db8 JB |
3422 | /* |
3423 | * If we have a uuid root and we're not being told to rescan we need to | |
3424 | * check the generation here so we can set the | |
3425 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3426 | * transaction during a balance or the log replay without updating the | |
3427 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3428 | * even though it was perfectly fine. | |
3429 | */ | |
3430 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3431 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3432 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3433 | ||
cf90d884 QW |
3434 | ret = btrfs_verify_dev_extents(fs_info); |
3435 | if (ret) { | |
3436 | btrfs_err(fs_info, | |
3437 | "failed to verify dev extents against chunks: %d", | |
3438 | ret); | |
3439 | goto fail_block_groups; | |
3440 | } | |
68310a5e ID |
3441 | ret = btrfs_recover_balance(fs_info); |
3442 | if (ret) { | |
05135f59 | 3443 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3444 | goto fail_block_groups; |
3445 | } | |
3446 | ||
733f4fbb SB |
3447 | ret = btrfs_init_dev_stats(fs_info); |
3448 | if (ret) { | |
05135f59 | 3449 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3450 | goto fail_block_groups; |
3451 | } | |
3452 | ||
8dabb742 SB |
3453 | ret = btrfs_init_dev_replace(fs_info); |
3454 | if (ret) { | |
05135f59 | 3455 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3456 | goto fail_block_groups; |
3457 | } | |
3458 | ||
b70f5097 NA |
3459 | ret = btrfs_check_zoned_mode(fs_info); |
3460 | if (ret) { | |
3461 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3462 | ret); | |
3463 | goto fail_block_groups; | |
3464 | } | |
3465 | ||
c6761a9e | 3466 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3467 | if (ret) { |
05135f59 DS |
3468 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3469 | ret); | |
b7c35e81 AJ |
3470 | goto fail_block_groups; |
3471 | } | |
3472 | ||
96f3136e | 3473 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3474 | if (ret) { |
05135f59 | 3475 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3476 | goto fail_fsdev_sysfs; |
c59021f8 | 3477 | } |
3478 | ||
c59021f8 | 3479 | ret = btrfs_init_space_info(fs_info); |
3480 | if (ret) { | |
05135f59 | 3481 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3482 | goto fail_sysfs; |
c59021f8 | 3483 | } |
3484 | ||
5b4aacef | 3485 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3486 | if (ret) { |
05135f59 | 3487 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3488 | goto fail_sysfs; |
1b1d1f66 | 3489 | } |
4330e183 | 3490 | |
16beac87 NA |
3491 | btrfs_free_zone_cache(fs_info); |
3492 | ||
a7e1ac7b NA |
3493 | btrfs_check_active_zone_reservation(fs_info); |
3494 | ||
5c78a5e7 AJ |
3495 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3496 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3497 | btrfs_warn(fs_info, |
52042d8e | 3498 | "writable mount is not allowed due to too many missing devices"); |
4871c33b | 3499 | ret = -EINVAL; |
2365dd3c | 3500 | goto fail_sysfs; |
292fd7fc | 3501 | } |
9078a3e1 | 3502 | |
33c44184 | 3503 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3504 | "btrfs-cleaner"); |
4871c33b QW |
3505 | if (IS_ERR(fs_info->cleaner_kthread)) { |
3506 | ret = PTR_ERR(fs_info->cleaner_kthread); | |
2365dd3c | 3507 | goto fail_sysfs; |
4871c33b | 3508 | } |
a74a4b97 CM |
3509 | |
3510 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3511 | tree_root, | |
3512 | "btrfs-transaction"); | |
4871c33b QW |
3513 | if (IS_ERR(fs_info->transaction_kthread)) { |
3514 | ret = PTR_ERR(fs_info->transaction_kthread); | |
3f157a2f | 3515 | goto fail_cleaner; |
4871c33b | 3516 | } |
a74a4b97 | 3517 | |
bcef60f2 AJ |
3518 | ret = btrfs_read_qgroup_config(fs_info); |
3519 | if (ret) | |
3520 | goto fail_trans_kthread; | |
21adbd5c | 3521 | |
fd708b81 JB |
3522 | if (btrfs_build_ref_tree(fs_info)) |
3523 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3524 | ||
96da0919 QW |
3525 | /* do not make disk changes in broken FS or nologreplay is given */ |
3526 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3527 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3528 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3529 | ret = btrfs_replay_log(fs_info, fs_devices); |
4871c33b | 3530 | if (ret) |
28c16cbb | 3531 | goto fail_qgroup; |
e02119d5 | 3532 | } |
1a40e23b | 3533 | |
56e9357a | 3534 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 | 3535 | if (IS_ERR(fs_info->fs_root)) { |
4871c33b QW |
3536 | ret = PTR_ERR(fs_info->fs_root); |
3537 | btrfs_warn(fs_info, "failed to read fs tree: %d", ret); | |
315bf8ef | 3538 | fs_info->fs_root = NULL; |
bcef60f2 | 3539 | goto fail_qgroup; |
3140c9a3 | 3540 | } |
c289811c | 3541 | |
bc98a42c | 3542 | if (sb_rdonly(sb)) |
83e3a40a | 3543 | return 0; |
59641015 | 3544 | |
44c0ca21 | 3545 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3546 | if (ret) { |
6bccf3ab | 3547 | close_ctree(fs_info); |
2b6ba629 | 3548 | return ret; |
e3acc2a6 | 3549 | } |
b0643e59 | 3550 | btrfs_discard_resume(fs_info); |
b382a324 | 3551 | |
44c0ca21 BB |
3552 | if (fs_info->uuid_root && |
3553 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3554 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3555 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3556 | ret = btrfs_check_uuid_tree(fs_info); |
3557 | if (ret) { | |
05135f59 DS |
3558 | btrfs_warn(fs_info, |
3559 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3560 | close_ctree(fs_info); |
70f80175 SB |
3561 | return ret; |
3562 | } | |
f7a81ea4 | 3563 | } |
94846229 | 3564 | |
afcdd129 | 3565 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3566 | |
b4be6aef JB |
3567 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3568 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3569 | wake_up_process(fs_info->cleaner_kthread); | |
3570 | ||
ad2b2c80 | 3571 | return 0; |
39279cc3 | 3572 | |
bcef60f2 AJ |
3573 | fail_qgroup: |
3574 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3575 | fail_trans_kthread: |
3576 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3577 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3578 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3579 | fail_cleaner: |
a74a4b97 | 3580 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3581 | |
3582 | /* | |
3583 | * make sure we're done with the btree inode before we stop our | |
3584 | * kthreads | |
3585 | */ | |
3586 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3587 | |
2365dd3c | 3588 | fail_sysfs: |
6618a59b | 3589 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3590 | |
b7c35e81 AJ |
3591 | fail_fsdev_sysfs: |
3592 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3593 | ||
1b1d1f66 | 3594 | fail_block_groups: |
54067ae9 | 3595 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3596 | |
3597 | fail_tree_roots: | |
9e3aa805 JB |
3598 | if (fs_info->data_reloc_root) |
3599 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3600 | free_root_pointers(fs_info, true); |
2b8195bb | 3601 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3602 | |
39279cc3 | 3603 | fail_sb_buffer: |
7abadb64 | 3604 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3605 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3606 | fail_alloc: |
7dc66abb | 3607 | btrfs_mapping_tree_free(fs_info); |
586e46e2 | 3608 | |
4543df7e | 3609 | iput(fs_info->btree_inode); |
7e662854 | 3610 | fail: |
586e46e2 | 3611 | btrfs_close_devices(fs_info->fs_devices); |
4871c33b QW |
3612 | ASSERT(ret < 0); |
3613 | return ret; | |
eb60ceac | 3614 | } |
663faf9f | 3615 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3616 | |
314b6dd0 | 3617 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3618 | { |
314b6dd0 JT |
3619 | struct btrfs_device *device = bio->bi_private; |
3620 | struct bio_vec *bvec; | |
3621 | struct bvec_iter_all iter_all; | |
3622 | struct page *page; | |
3623 | ||
3624 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3625 | page = bvec->bv_page; | |
3626 | ||
3627 | if (bio->bi_status) { | |
3628 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3629 | "lost page write due to IO error on %s (%d)", | |
cb3e217b | 3630 | btrfs_dev_name(device), |
314b6dd0 JT |
3631 | blk_status_to_errno(bio->bi_status)); |
3632 | ClearPageUptodate(page); | |
3633 | SetPageError(page); | |
3634 | btrfs_dev_stat_inc_and_print(device, | |
3635 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3636 | } else { | |
3637 | SetPageUptodate(page); | |
3638 | } | |
3639 | ||
3640 | put_page(page); | |
3641 | unlock_page(page); | |
f2984462 | 3642 | } |
314b6dd0 JT |
3643 | |
3644 | bio_put(bio); | |
f2984462 CM |
3645 | } |
3646 | ||
8f32380d | 3647 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3648 | int copy_num, bool drop_cache) |
29c36d72 | 3649 | { |
29c36d72 | 3650 | struct btrfs_super_block *super; |
8f32380d | 3651 | struct page *page; |
12659251 | 3652 | u64 bytenr, bytenr_orig; |
8f32380d | 3653 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3654 | int ret; |
3655 | ||
3656 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3657 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3658 | if (ret == -ENOENT) | |
3659 | return ERR_PTR(-EINVAL); | |
3660 | else if (ret) | |
3661 | return ERR_PTR(ret); | |
29c36d72 | 3662 | |
cda00eba | 3663 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3664 | return ERR_PTR(-EINVAL); |
29c36d72 | 3665 | |
a05d3c91 QW |
3666 | if (drop_cache) { |
3667 | /* This should only be called with the primary sb. */ | |
3668 | ASSERT(copy_num == 0); | |
3669 | ||
3670 | /* | |
3671 | * Drop the page of the primary superblock, so later read will | |
3672 | * always read from the device. | |
3673 | */ | |
3674 | invalidate_inode_pages2_range(mapping, | |
3675 | bytenr >> PAGE_SHIFT, | |
3676 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3677 | } | |
3678 | ||
8f32380d JT |
3679 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3680 | if (IS_ERR(page)) | |
3681 | return ERR_CAST(page); | |
29c36d72 | 3682 | |
8f32380d | 3683 | super = page_address(page); |
96c2e067 AJ |
3684 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3685 | btrfs_release_disk_super(super); | |
3686 | return ERR_PTR(-ENODATA); | |
3687 | } | |
3688 | ||
12659251 | 3689 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3690 | btrfs_release_disk_super(super); |
3691 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3692 | } |
3693 | ||
8f32380d | 3694 | return super; |
29c36d72 AJ |
3695 | } |
3696 | ||
3697 | ||
8f32380d | 3698 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3699 | { |
8f32380d | 3700 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3701 | int i; |
3702 | u64 transid = 0; | |
a512bbf8 YZ |
3703 | |
3704 | /* we would like to check all the supers, but that would make | |
3705 | * a btrfs mount succeed after a mkfs from a different FS. | |
3706 | * So, we need to add a special mount option to scan for | |
3707 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3708 | */ | |
3709 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3710 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3711 | if (IS_ERR(super)) |
a512bbf8 YZ |
3712 | continue; |
3713 | ||
a512bbf8 | 3714 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3715 | if (latest) |
3716 | btrfs_release_disk_super(super); | |
3717 | ||
3718 | latest = super; | |
a512bbf8 | 3719 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3720 | } |
3721 | } | |
92fc03fb | 3722 | |
8f32380d | 3723 | return super; |
a512bbf8 YZ |
3724 | } |
3725 | ||
4eedeb75 | 3726 | /* |
abbb3b8e | 3727 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3728 | * pages we use for writing are locked. |
4eedeb75 | 3729 | * |
abbb3b8e DS |
3730 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3731 | * the expected device size at commit time. Note that max_mirrors must be | |
3732 | * same for write and wait phases. | |
4eedeb75 | 3733 | * |
314b6dd0 | 3734 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3735 | */ |
a512bbf8 | 3736 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3737 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3738 | { |
d5178578 | 3739 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3740 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3741 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3742 | int i; |
a512bbf8 | 3743 | int errors = 0; |
12659251 NA |
3744 | int ret; |
3745 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3746 | |
3747 | if (max_mirrors == 0) | |
3748 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3749 | ||
d5178578 JT |
3750 | shash->tfm = fs_info->csum_shash; |
3751 | ||
a512bbf8 | 3752 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3753 | struct page *page; |
3754 | struct bio *bio; | |
3755 | struct btrfs_super_block *disk_super; | |
3756 | ||
12659251 NA |
3757 | bytenr_orig = btrfs_sb_offset(i); |
3758 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3759 | if (ret == -ENOENT) { | |
3760 | continue; | |
3761 | } else if (ret < 0) { | |
3762 | btrfs_err(device->fs_info, | |
3763 | "couldn't get super block location for mirror %d", | |
3764 | i); | |
3765 | errors++; | |
3766 | continue; | |
3767 | } | |
935e5cc9 MX |
3768 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3769 | device->commit_total_bytes) | |
a512bbf8 YZ |
3770 | break; |
3771 | ||
12659251 | 3772 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3773 | |
fd08001f EB |
3774 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3775 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3776 | sb->csum); | |
4eedeb75 | 3777 | |
314b6dd0 JT |
3778 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3779 | GFP_NOFS); | |
3780 | if (!page) { | |
abbb3b8e | 3781 | btrfs_err(device->fs_info, |
314b6dd0 | 3782 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3783 | bytenr); |
3784 | errors++; | |
4eedeb75 | 3785 | continue; |
abbb3b8e | 3786 | } |
634554dc | 3787 | |
314b6dd0 JT |
3788 | /* Bump the refcount for wait_dev_supers() */ |
3789 | get_page(page); | |
a512bbf8 | 3790 | |
314b6dd0 JT |
3791 | disk_super = page_address(page); |
3792 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3793 | |
314b6dd0 JT |
3794 | /* |
3795 | * Directly use bios here instead of relying on the page cache | |
3796 | * to do I/O, so we don't lose the ability to do integrity | |
3797 | * checking. | |
3798 | */ | |
07888c66 CH |
3799 | bio = bio_alloc(device->bdev, 1, |
3800 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
3801 | GFP_NOFS); | |
314b6dd0 JT |
3802 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
3803 | bio->bi_private = device; | |
3804 | bio->bi_end_io = btrfs_end_super_write; | |
3805 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3806 | offset_in_page(bytenr)); | |
a512bbf8 | 3807 | |
387125fc | 3808 | /* |
314b6dd0 JT |
3809 | * We FUA only the first super block. The others we allow to |
3810 | * go down lazy and there's a short window where the on-disk | |
3811 | * copies might still contain the older version. | |
387125fc | 3812 | */ |
1b9e619c | 3813 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 | 3814 | bio->bi_opf |= REQ_FUA; |
58ff51f1 | 3815 | submit_bio(bio); |
8376d9e1 NA |
3816 | |
3817 | if (btrfs_advance_sb_log(device, i)) | |
3818 | errors++; | |
a512bbf8 YZ |
3819 | } |
3820 | return errors < i ? 0 : -1; | |
3821 | } | |
3822 | ||
abbb3b8e DS |
3823 | /* |
3824 | * Wait for write completion of superblocks done by write_dev_supers, | |
3825 | * @max_mirrors same for write and wait phases. | |
3826 | * | |
314b6dd0 | 3827 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3828 | * date. |
3829 | */ | |
3830 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3831 | { | |
abbb3b8e DS |
3832 | int i; |
3833 | int errors = 0; | |
b6a535fa | 3834 | bool primary_failed = false; |
12659251 | 3835 | int ret; |
abbb3b8e DS |
3836 | u64 bytenr; |
3837 | ||
3838 | if (max_mirrors == 0) | |
3839 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3840 | ||
3841 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3842 | struct page *page; |
3843 | ||
12659251 NA |
3844 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3845 | if (ret == -ENOENT) { | |
3846 | break; | |
3847 | } else if (ret < 0) { | |
3848 | errors++; | |
3849 | if (i == 0) | |
3850 | primary_failed = true; | |
3851 | continue; | |
3852 | } | |
abbb3b8e DS |
3853 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3854 | device->commit_total_bytes) | |
3855 | break; | |
3856 | ||
314b6dd0 JT |
3857 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3858 | bytenr >> PAGE_SHIFT); | |
3859 | if (!page) { | |
abbb3b8e | 3860 | errors++; |
b6a535fa HM |
3861 | if (i == 0) |
3862 | primary_failed = true; | |
abbb3b8e DS |
3863 | continue; |
3864 | } | |
314b6dd0 JT |
3865 | /* Page is submitted locked and unlocked once the IO completes */ |
3866 | wait_on_page_locked(page); | |
3867 | if (PageError(page)) { | |
abbb3b8e | 3868 | errors++; |
b6a535fa HM |
3869 | if (i == 0) |
3870 | primary_failed = true; | |
3871 | } | |
abbb3b8e | 3872 | |
314b6dd0 JT |
3873 | /* Drop our reference */ |
3874 | put_page(page); | |
abbb3b8e | 3875 | |
314b6dd0 JT |
3876 | /* Drop the reference from the writing run */ |
3877 | put_page(page); | |
abbb3b8e DS |
3878 | } |
3879 | ||
b6a535fa HM |
3880 | /* log error, force error return */ |
3881 | if (primary_failed) { | |
3882 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3883 | device->devid); | |
3884 | return -1; | |
3885 | } | |
3886 | ||
abbb3b8e DS |
3887 | return errors < i ? 0 : -1; |
3888 | } | |
3889 | ||
387125fc CM |
3890 | /* |
3891 | * endio for the write_dev_flush, this will wake anyone waiting | |
3892 | * for the barrier when it is done | |
3893 | */ | |
4246a0b6 | 3894 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3895 | { |
f9e69aa9 | 3896 | bio_uninit(bio); |
e0ae9994 | 3897 | complete(bio->bi_private); |
387125fc CM |
3898 | } |
3899 | ||
3900 | /* | |
4fc6441a AJ |
3901 | * Submit a flush request to the device if it supports it. Error handling is |
3902 | * done in the waiting counterpart. | |
387125fc | 3903 | */ |
4fc6441a | 3904 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3905 | { |
f9e69aa9 | 3906 | struct bio *bio = &device->flush_bio; |
387125fc | 3907 | |
bfd3ea94 AJ |
3908 | device->last_flush_error = BLK_STS_OK; |
3909 | ||
f9e69aa9 CH |
3910 | bio_init(bio, device->bdev, NULL, 0, |
3911 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 3912 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
3913 | init_completion(&device->flush_wait); |
3914 | bio->bi_private = &device->flush_wait; | |
58ff51f1 | 3915 | submit_bio(bio); |
1c3063b6 | 3916 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3917 | } |
387125fc | 3918 | |
4fc6441a AJ |
3919 | /* |
3920 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
1b465784 | 3921 | * Return true for any error, and false otherwise. |
4fc6441a | 3922 | */ |
1b465784 | 3923 | static bool wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3924 | { |
f9e69aa9 | 3925 | struct bio *bio = &device->flush_bio; |
387125fc | 3926 | |
7e812f20 | 3927 | if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
1b465784 | 3928 | return false; |
387125fc | 3929 | |
2980d574 | 3930 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3931 | |
bfd3ea94 AJ |
3932 | if (bio->bi_status) { |
3933 | device->last_flush_error = bio->bi_status; | |
3934 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); | |
1b465784 | 3935 | return true; |
bfd3ea94 AJ |
3936 | } |
3937 | ||
1b465784 | 3938 | return false; |
387125fc | 3939 | } |
387125fc | 3940 | |
387125fc CM |
3941 | /* |
3942 | * send an empty flush down to each device in parallel, | |
3943 | * then wait for them | |
3944 | */ | |
3945 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3946 | { | |
3947 | struct list_head *head; | |
3948 | struct btrfs_device *dev; | |
5af3e8cc | 3949 | int errors_wait = 0; |
387125fc | 3950 | |
1538e6c5 | 3951 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3952 | /* send down all the barriers */ |
3953 | head = &info->fs_devices->devices; | |
1538e6c5 | 3954 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3955 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3956 | continue; |
cea7c8bf | 3957 | if (!dev->bdev) |
387125fc | 3958 | continue; |
e12c9621 | 3959 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3960 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3961 | continue; |
3962 | ||
4fc6441a | 3963 | write_dev_flush(dev); |
387125fc CM |
3964 | } |
3965 | ||
3966 | /* wait for all the barriers */ | |
1538e6c5 | 3967 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3968 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3969 | continue; |
387125fc | 3970 | if (!dev->bdev) { |
5af3e8cc | 3971 | errors_wait++; |
387125fc CM |
3972 | continue; |
3973 | } | |
e12c9621 | 3974 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3975 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3976 | continue; |
3977 | ||
1b465784 | 3978 | if (wait_dev_flush(dev)) |
5af3e8cc | 3979 | errors_wait++; |
401b41e5 AJ |
3980 | } |
3981 | ||
de38a206 AJ |
3982 | /* |
3983 | * Checks last_flush_error of disks in order to determine the device | |
3984 | * state. | |
3985 | */ | |
3986 | if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) | |
3987 | return -EIO; | |
3988 | ||
387125fc CM |
3989 | return 0; |
3990 | } | |
3991 | ||
943c6e99 ZL |
3992 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3993 | { | |
8789f4fe ZL |
3994 | int raid_type; |
3995 | int min_tolerated = INT_MAX; | |
943c6e99 | 3996 | |
8789f4fe ZL |
3997 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3998 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 3999 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4000 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4001 | tolerated_failures); | |
943c6e99 | 4002 | |
8789f4fe ZL |
4003 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4004 | if (raid_type == BTRFS_RAID_SINGLE) | |
4005 | continue; | |
41a6e891 | 4006 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4007 | continue; |
8c3e3582 | 4008 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4009 | btrfs_raid_array[raid_type]. |
4010 | tolerated_failures); | |
4011 | } | |
943c6e99 | 4012 | |
8789f4fe | 4013 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4014 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4015 | min_tolerated = 0; |
4016 | } | |
4017 | ||
4018 | return min_tolerated; | |
943c6e99 ZL |
4019 | } |
4020 | ||
eece6a9c | 4021 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4022 | { |
e5e9a520 | 4023 | struct list_head *head; |
f2984462 | 4024 | struct btrfs_device *dev; |
a061fc8d | 4025 | struct btrfs_super_block *sb; |
f2984462 | 4026 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4027 | int ret; |
4028 | int do_barriers; | |
a236aed1 CM |
4029 | int max_errors; |
4030 | int total_errors = 0; | |
a061fc8d | 4031 | u64 flags; |
f2984462 | 4032 | |
0b246afa | 4033 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4034 | |
4035 | /* | |
4036 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4037 | * not from fsync where the tree roots in fs_info have not | |
4038 | * been consistent on disk. | |
4039 | */ | |
4040 | if (max_mirrors == 0) | |
4041 | backup_super_roots(fs_info); | |
f2984462 | 4042 | |
0b246afa | 4043 | sb = fs_info->super_for_commit; |
a061fc8d | 4044 | dev_item = &sb->dev_item; |
e5e9a520 | 4045 | |
0b246afa JM |
4046 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4047 | head = &fs_info->fs_devices->devices; | |
4048 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4049 | |
5af3e8cc | 4050 | if (do_barriers) { |
0b246afa | 4051 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4052 | if (ret) { |
4053 | mutex_unlock( | |
0b246afa JM |
4054 | &fs_info->fs_devices->device_list_mutex); |
4055 | btrfs_handle_fs_error(fs_info, ret, | |
4056 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4057 | return ret; |
4058 | } | |
4059 | } | |
387125fc | 4060 | |
1538e6c5 | 4061 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4062 | if (!dev->bdev) { |
4063 | total_errors++; | |
4064 | continue; | |
4065 | } | |
e12c9621 | 4066 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4067 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4068 | continue; |
4069 | ||
2b82032c | 4070 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4071 | btrfs_set_stack_device_type(dev_item, dev->type); |
4072 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4073 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4074 | dev->commit_total_bytes); |
ce7213c7 MX |
4075 | btrfs_set_stack_device_bytes_used(dev_item, |
4076 | dev->commit_bytes_used); | |
a061fc8d CM |
4077 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4078 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4079 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4080 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4081 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4082 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4083 | |
a061fc8d CM |
4084 | flags = btrfs_super_flags(sb); |
4085 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4086 | ||
75cb857d QW |
4087 | ret = btrfs_validate_write_super(fs_info, sb); |
4088 | if (ret < 0) { | |
4089 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4090 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4091 | "unexpected superblock corruption detected"); | |
4092 | return -EUCLEAN; | |
4093 | } | |
4094 | ||
abbb3b8e | 4095 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4096 | if (ret) |
4097 | total_errors++; | |
f2984462 | 4098 | } |
a236aed1 | 4099 | if (total_errors > max_errors) { |
0b246afa JM |
4100 | btrfs_err(fs_info, "%d errors while writing supers", |
4101 | total_errors); | |
4102 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4103 | |
9d565ba4 | 4104 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4105 | btrfs_handle_fs_error(fs_info, -EIO, |
4106 | "%d errors while writing supers", | |
4107 | total_errors); | |
9d565ba4 | 4108 | return -EIO; |
a236aed1 | 4109 | } |
f2984462 | 4110 | |
a512bbf8 | 4111 | total_errors = 0; |
1538e6c5 | 4112 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4113 | if (!dev->bdev) |
4114 | continue; | |
e12c9621 | 4115 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4116 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4117 | continue; |
4118 | ||
abbb3b8e | 4119 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4120 | if (ret) |
4121 | total_errors++; | |
f2984462 | 4122 | } |
0b246afa | 4123 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4124 | if (total_errors > max_errors) { |
0b246afa JM |
4125 | btrfs_handle_fs_error(fs_info, -EIO, |
4126 | "%d errors while writing supers", | |
4127 | total_errors); | |
79787eaa | 4128 | return -EIO; |
a236aed1 | 4129 | } |
f2984462 CM |
4130 | return 0; |
4131 | } | |
4132 | ||
cb517eab MX |
4133 | /* Drop a fs root from the radix tree and free it. */ |
4134 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4135 | struct btrfs_root *root) | |
2619ba1f | 4136 | { |
4785e24f JB |
4137 | bool drop_ref = false; |
4138 | ||
fc7cbcd4 DS |
4139 | spin_lock(&fs_info->fs_roots_radix_lock); |
4140 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4141 | (unsigned long)root->root_key.objectid); | |
4142 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4143 | drop_ref = true; |
fc7cbcd4 | 4144 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4145 | |
84961539 | 4146 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4147 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4148 | if (root->reloc_root) { |
00246528 | 4149 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4150 | root->reloc_root = NULL; |
4151 | } | |
4152 | } | |
3321719e | 4153 | |
4785e24f JB |
4154 | if (drop_ref) |
4155 | btrfs_put_root(root); | |
2619ba1f CM |
4156 | } |
4157 | ||
6bccf3ab | 4158 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4159 | { |
6bccf3ab | 4160 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4161 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4162 | |
0b246afa | 4163 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4164 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4165 | mutex_unlock(&fs_info->cleaner_mutex); |
4166 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4167 | |
4168 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4169 | down_write(&fs_info->cleanup_work_sem); |
4170 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4171 | |
7a7eaa40 | 4172 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4173 | if (IS_ERR(trans)) |
4174 | return PTR_ERR(trans); | |
3a45bb20 | 4175 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4176 | } |
4177 | ||
36c86a9e QW |
4178 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4179 | { | |
4180 | struct btrfs_transaction *trans; | |
4181 | struct btrfs_transaction *tmp; | |
4182 | bool found = false; | |
4183 | ||
4184 | if (list_empty(&fs_info->trans_list)) | |
4185 | return; | |
4186 | ||
4187 | /* | |
4188 | * This function is only called at the very end of close_ctree(), | |
4189 | * thus no other running transaction, no need to take trans_lock. | |
4190 | */ | |
4191 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4192 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4193 | struct extent_state *cached = NULL; | |
4194 | u64 dirty_bytes = 0; | |
4195 | u64 cur = 0; | |
4196 | u64 found_start; | |
4197 | u64 found_end; | |
4198 | ||
4199 | found = true; | |
e5860f82 | 4200 | while (find_first_extent_bit(&trans->dirty_pages, cur, |
36c86a9e QW |
4201 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { |
4202 | dirty_bytes += found_end + 1 - found_start; | |
4203 | cur = found_end + 1; | |
4204 | } | |
4205 | btrfs_warn(fs_info, | |
4206 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4207 | trans->transid, dirty_bytes); | |
4208 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4209 | ||
4210 | if (trans == fs_info->running_transaction) | |
4211 | fs_info->running_transaction = NULL; | |
4212 | list_del_init(&trans->list); | |
4213 | ||
4214 | btrfs_put_transaction(trans); | |
4215 | trace_btrfs_transaction_commit(fs_info); | |
4216 | } | |
4217 | ASSERT(!found); | |
4218 | } | |
4219 | ||
b105e927 | 4220 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4221 | { |
c146afad YZ |
4222 | int ret; |
4223 | ||
afcdd129 | 4224 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4225 | |
8a1f1e3d FM |
4226 | /* |
4227 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4228 | * clear that bit and wake up relocation so it can stop. | |
4229 | * We must do this before stopping the block group reclaim task, because | |
4230 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4231 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4232 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4233 | * return 1. | |
4234 | */ | |
4235 | btrfs_wake_unfinished_drop(fs_info); | |
4236 | ||
31e70e52 FM |
4237 | /* |
4238 | * We may have the reclaim task running and relocating a data block group, | |
4239 | * in which case it may create delayed iputs. So stop it before we park | |
4240 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4241 | * parking the cleaner, and that can make the async reclaim task to hang | |
4242 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4243 | * parked and can not run delayed iputs - this will make us hang when | |
4244 | * trying to stop the async reclaim task. | |
4245 | */ | |
4246 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4247 | /* |
4248 | * We don't want the cleaner to start new transactions, add more delayed | |
4249 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4250 | * because that frees the task_struct, and the transaction kthread might | |
4251 | * still try to wake up the cleaner. | |
4252 | */ | |
4253 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4254 | |
7343dd61 | 4255 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4256 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4257 | |
803b2f54 SB |
4258 | /* wait for the uuid_scan task to finish */ |
4259 | down(&fs_info->uuid_tree_rescan_sem); | |
4260 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4261 | up(&fs_info->uuid_tree_rescan_sem); | |
4262 | ||
837d5b6e | 4263 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4264 | btrfs_pause_balance(fs_info); |
837d5b6e | 4265 | |
8dabb742 SB |
4266 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4267 | ||
aa1b8cd4 | 4268 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4269 | |
4270 | /* wait for any defraggers to finish */ | |
4271 | wait_event(fs_info->transaction_wait, | |
4272 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4273 | ||
4274 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4275 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4276 | |
a362bb86 FM |
4277 | /* |
4278 | * After we parked the cleaner kthread, ordered extents may have | |
4279 | * completed and created new delayed iputs. If one of the async reclaim | |
4280 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4281 | * can hang forever trying to stop it, because if a delayed iput is | |
4282 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4283 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4284 | * no one else to run iputs. | |
4285 | * | |
4286 | * So wait for all ongoing ordered extents to complete and then run | |
4287 | * delayed iputs. This works because once we reach this point no one | |
4288 | * can either create new ordered extents nor create delayed iputs | |
4289 | * through some other means. | |
4290 | * | |
4291 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4292 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4293 | * but the delayed iput for the respective inode is made only when doing | |
4294 | * the final btrfs_put_ordered_extent() (which must happen at | |
4295 | * btrfs_finish_ordered_io() when we are unmounting). | |
4296 | */ | |
4297 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4298 | /* Ordered extents for free space inodes. */ | |
4299 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4300 | btrfs_run_delayed_iputs(fs_info); | |
4301 | ||
21c7e756 | 4302 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4303 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4304 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4305 | |
b0643e59 DZ |
4306 | /* Cancel or finish ongoing discard work */ |
4307 | btrfs_discard_cleanup(fs_info); | |
4308 | ||
bc98a42c | 4309 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4310 | /* |
d6fd0ae2 OS |
4311 | * The cleaner kthread is stopped, so do one final pass over |
4312 | * unused block groups. | |
e44163e1 | 4313 | */ |
0b246afa | 4314 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4315 | |
f0cc2cd7 FM |
4316 | /* |
4317 | * There might be existing delayed inode workers still running | |
4318 | * and holding an empty delayed inode item. We must wait for | |
4319 | * them to complete first because they can create a transaction. | |
4320 | * This happens when someone calls btrfs_balance_delayed_items() | |
4321 | * and then a transaction commit runs the same delayed nodes | |
4322 | * before any delayed worker has done something with the nodes. | |
4323 | * We must wait for any worker here and not at transaction | |
4324 | * commit time since that could cause a deadlock. | |
4325 | * This is a very rare case. | |
4326 | */ | |
4327 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4328 | ||
6bccf3ab | 4329 | ret = btrfs_commit_super(fs_info); |
acce952b | 4330 | if (ret) |
04892340 | 4331 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4332 | } |
4333 | ||
84961539 | 4334 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4335 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4336 | |
e3029d9f AV |
4337 | kthread_stop(fs_info->transaction_kthread); |
4338 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4339 | |
e187831e | 4340 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4341 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4342 | |
5958253c QW |
4343 | if (btrfs_check_quota_leak(fs_info)) { |
4344 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4345 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4346 | } | |
4347 | ||
04892340 | 4348 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4349 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4350 | |
963d678b | 4351 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4352 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4353 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4354 | } |
bcc63abb | 4355 | |
5deb17e1 | 4356 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4357 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4358 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4359 | |
6618a59b | 4360 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4361 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4362 | |
1a4319cc LB |
4363 | btrfs_put_block_group_cache(fs_info); |
4364 | ||
de348ee0 WS |
4365 | /* |
4366 | * we must make sure there is not any read request to | |
4367 | * submit after we stopping all workers. | |
4368 | */ | |
4369 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4370 | btrfs_stop_all_workers(fs_info); |
4371 | ||
0a31daa4 | 4372 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4373 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4374 | |
afcdd129 | 4375 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4376 | free_root_pointers(fs_info, true); |
8c38938c | 4377 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4378 | |
4e19443d JB |
4379 | /* |
4380 | * We must free the block groups after dropping the fs_roots as we could | |
4381 | * have had an IO error and have left over tree log blocks that aren't | |
4382 | * cleaned up until the fs roots are freed. This makes the block group | |
4383 | * accounting appear to be wrong because there's pending reserved bytes, | |
4384 | * so make sure we do the block group cleanup afterwards. | |
4385 | */ | |
4386 | btrfs_free_block_groups(fs_info); | |
4387 | ||
13e6c37b | 4388 | iput(fs_info->btree_inode); |
d6bfde87 | 4389 | |
7dc66abb | 4390 | btrfs_mapping_tree_free(fs_info); |
68c94e55 | 4391 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4392 | } |
4393 | ||
50564b65 FM |
4394 | void btrfs_mark_buffer_dirty(struct btrfs_trans_handle *trans, |
4395 | struct extent_buffer *buf) | |
5f39d397 | 4396 | { |
2f4d60df | 4397 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4398 | u64 transid = btrfs_header_generation(buf); |
b4ce94de | 4399 | |
06ea65a3 JB |
4400 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4401 | /* | |
4402 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4403 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4404 | * outside of the sanity tests. |
4405 | */ | |
b0132a3b | 4406 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4407 | return; |
4408 | #endif | |
50564b65 FM |
4409 | /* This is an active transaction (its state < TRANS_STATE_UNBLOCKED). */ |
4410 | ASSERT(trans->transid == fs_info->generation); | |
49d0c642 | 4411 | btrfs_assert_tree_write_locked(buf); |
4ebe8d47 | 4412 | if (unlikely(transid != fs_info->generation)) { |
50564b65 | 4413 | btrfs_abort_transaction(trans, -EUCLEAN); |
20cbe460 FM |
4414 | btrfs_crit(fs_info, |
4415 | "dirty buffer transid mismatch, logical %llu found transid %llu running transid %llu", | |
4416 | buf->start, transid, fs_info->generation); | |
50564b65 | 4417 | } |
f18cc978 | 4418 | set_extent_buffer_dirty(buf); |
eb60ceac CM |
4419 | } |
4420 | ||
2ff7e61e | 4421 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4422 | int flush_delayed) |
16cdcec7 MX |
4423 | { |
4424 | /* | |
4425 | * looks as though older kernels can get into trouble with | |
4426 | * this code, they end up stuck in balance_dirty_pages forever | |
4427 | */ | |
e2d84521 | 4428 | int ret; |
16cdcec7 MX |
4429 | |
4430 | if (current->flags & PF_MEMALLOC) | |
4431 | return; | |
4432 | ||
b53d3f5d | 4433 | if (flush_delayed) |
2ff7e61e | 4434 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4435 | |
d814a491 EL |
4436 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4437 | BTRFS_DIRTY_METADATA_THRESH, | |
4438 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4439 | if (ret > 0) { |
0b246afa | 4440 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4441 | } |
16cdcec7 MX |
4442 | } |
4443 | ||
2ff7e61e | 4444 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4445 | { |
2ff7e61e | 4446 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4447 | } |
585ad2c3 | 4448 | |
2ff7e61e | 4449 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4450 | { |
2ff7e61e | 4451 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4452 | } |
6b80053d | 4453 | |
2ff7e61e | 4454 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4455 | { |
fe816d0f NB |
4456 | /* cleanup FS via transaction */ |
4457 | btrfs_cleanup_transaction(fs_info); | |
4458 | ||
0b246afa | 4459 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4460 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4461 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4462 | |
0b246afa JM |
4463 | down_write(&fs_info->cleanup_work_sem); |
4464 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4465 | } |
4466 | ||
ef67963d JB |
4467 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4468 | { | |
fc7cbcd4 DS |
4469 | struct btrfs_root *gang[8]; |
4470 | u64 root_objectid = 0; | |
4471 | int ret; | |
4472 | ||
4473 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4474 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4475 | (void **)gang, root_objectid, | |
4476 | ARRAY_SIZE(gang))) != 0) { | |
4477 | int i; | |
4478 | ||
4479 | for (i = 0; i < ret; i++) | |
4480 | gang[i] = btrfs_grab_root(gang[i]); | |
4481 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4482 | ||
4483 | for (i = 0; i < ret; i++) { | |
4484 | if (!gang[i]) | |
ef67963d | 4485 | continue; |
fc7cbcd4 DS |
4486 | root_objectid = gang[i]->root_key.objectid; |
4487 | btrfs_free_log(NULL, gang[i]); | |
4488 | btrfs_put_root(gang[i]); | |
ef67963d | 4489 | } |
fc7cbcd4 DS |
4490 | root_objectid++; |
4491 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4492 | } |
fc7cbcd4 | 4493 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4494 | btrfs_free_log_root_tree(NULL, fs_info); |
4495 | } | |
4496 | ||
143bede5 | 4497 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4498 | { |
acce952b | 4499 | struct btrfs_ordered_extent *ordered; |
acce952b | 4500 | |
199c2a9c | 4501 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4502 | /* |
4503 | * This will just short circuit the ordered completion stuff which will | |
4504 | * make sure the ordered extent gets properly cleaned up. | |
4505 | */ | |
199c2a9c | 4506 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4507 | root_extent_list) |
4508 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4509 | spin_unlock(&root->ordered_extent_lock); |
4510 | } | |
4511 | ||
4512 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4513 | { | |
4514 | struct btrfs_root *root; | |
84af994b | 4515 | LIST_HEAD(splice); |
199c2a9c MX |
4516 | |
4517 | spin_lock(&fs_info->ordered_root_lock); | |
4518 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4519 | while (!list_empty(&splice)) { | |
4520 | root = list_first_entry(&splice, struct btrfs_root, | |
4521 | ordered_root); | |
1de2cfde JB |
4522 | list_move_tail(&root->ordered_root, |
4523 | &fs_info->ordered_roots); | |
199c2a9c | 4524 | |
2a85d9ca | 4525 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4526 | btrfs_destroy_ordered_extents(root); |
4527 | ||
2a85d9ca LB |
4528 | cond_resched(); |
4529 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4530 | } |
4531 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4532 | |
4533 | /* | |
4534 | * We need this here because if we've been flipped read-only we won't | |
4535 | * get sync() from the umount, so we need to make sure any ordered | |
4536 | * extents that haven't had their dirty pages IO start writeout yet | |
4537 | * actually get run and error out properly. | |
4538 | */ | |
4539 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4540 | } |
4541 | ||
99f09ce3 FM |
4542 | static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4543 | struct btrfs_fs_info *fs_info) | |
acce952b | 4544 | { |
4545 | struct rb_node *node; | |
4546 | struct btrfs_delayed_ref_root *delayed_refs; | |
4547 | struct btrfs_delayed_ref_node *ref; | |
acce952b | 4548 | |
4549 | delayed_refs = &trans->delayed_refs; | |
4550 | ||
4551 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4552 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4553 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4554 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
99f09ce3 | 4555 | return; |
acce952b | 4556 | } |
4557 | ||
5c9d028b | 4558 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4559 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4560 | struct rb_node *n; |
e78417d1 | 4561 | bool pin_bytes = false; |
acce952b | 4562 | |
d7df2c79 JB |
4563 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4564 | href_node); | |
3069bd26 | 4565 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4566 | continue; |
3069bd26 | 4567 | |
d7df2c79 | 4568 | spin_lock(&head->lock); |
e3d03965 | 4569 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4570 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4571 | ref_node); | |
e3d03965 | 4572 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4573 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4574 | if (!list_empty(&ref->add_list)) |
4575 | list_del(&ref->add_list); | |
d7df2c79 JB |
4576 | atomic_dec(&delayed_refs->num_entries); |
4577 | btrfs_put_delayed_ref(ref); | |
adb86dbe | 4578 | btrfs_delayed_refs_rsv_release(fs_info, 1, 0); |
e78417d1 | 4579 | } |
d7df2c79 JB |
4580 | if (head->must_insert_reserved) |
4581 | pin_bytes = true; | |
4582 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4583 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4584 | spin_unlock(&head->lock); |
4585 | spin_unlock(&delayed_refs->lock); | |
4586 | mutex_unlock(&head->mutex); | |
acce952b | 4587 | |
f603bb94 NB |
4588 | if (pin_bytes) { |
4589 | struct btrfs_block_group *cache; | |
4590 | ||
4591 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4592 | BUG_ON(!cache); | |
4593 | ||
4594 | spin_lock(&cache->space_info->lock); | |
4595 | spin_lock(&cache->lock); | |
4596 | cache->pinned += head->num_bytes; | |
4597 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4598 | cache->space_info, head->num_bytes); | |
4599 | cache->reserved -= head->num_bytes; | |
4600 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4601 | spin_unlock(&cache->lock); | |
4602 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4603 | |
4604 | btrfs_put_block_group(cache); | |
4605 | ||
4606 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4607 | head->bytenr + head->num_bytes - 1); | |
4608 | } | |
31890da0 | 4609 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4610 | btrfs_put_delayed_ref_head(head); |
acce952b | 4611 | cond_resched(); |
4612 | spin_lock(&delayed_refs->lock); | |
4613 | } | |
81f7eb00 | 4614 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4615 | |
4616 | spin_unlock(&delayed_refs->lock); | |
acce952b | 4617 | } |
4618 | ||
143bede5 | 4619 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4620 | { |
4621 | struct btrfs_inode *btrfs_inode; | |
84af994b | 4622 | LIST_HEAD(splice); |
acce952b | 4623 | |
eb73c1b7 MX |
4624 | spin_lock(&root->delalloc_lock); |
4625 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4626 | |
4627 | while (!list_empty(&splice)) { | |
fe816d0f | 4628 | struct inode *inode = NULL; |
eb73c1b7 MX |
4629 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4630 | delalloc_inodes); | |
fe816d0f | 4631 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4632 | spin_unlock(&root->delalloc_lock); |
acce952b | 4633 | |
fe816d0f NB |
4634 | /* |
4635 | * Make sure we get a live inode and that it'll not disappear | |
4636 | * meanwhile. | |
4637 | */ | |
4638 | inode = igrab(&btrfs_inode->vfs_inode); | |
4639 | if (inode) { | |
597441b3 JB |
4640 | unsigned int nofs_flag; |
4641 | ||
4642 | nofs_flag = memalloc_nofs_save(); | |
fe816d0f | 4643 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 | 4644 | memalloc_nofs_restore(nofs_flag); |
fe816d0f NB |
4645 | iput(inode); |
4646 | } | |
eb73c1b7 | 4647 | spin_lock(&root->delalloc_lock); |
acce952b | 4648 | } |
eb73c1b7 MX |
4649 | spin_unlock(&root->delalloc_lock); |
4650 | } | |
4651 | ||
4652 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4653 | { | |
4654 | struct btrfs_root *root; | |
84af994b | 4655 | LIST_HEAD(splice); |
eb73c1b7 MX |
4656 | |
4657 | spin_lock(&fs_info->delalloc_root_lock); | |
4658 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4659 | while (!list_empty(&splice)) { | |
4660 | root = list_first_entry(&splice, struct btrfs_root, | |
4661 | delalloc_root); | |
00246528 | 4662 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4663 | BUG_ON(!root); |
4664 | spin_unlock(&fs_info->delalloc_root_lock); | |
4665 | ||
4666 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4667 | btrfs_put_root(root); |
eb73c1b7 MX |
4668 | |
4669 | spin_lock(&fs_info->delalloc_root_lock); | |
4670 | } | |
4671 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4672 | } |
4673 | ||
aec5716c FM |
4674 | static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
4675 | struct extent_io_tree *dirty_pages, | |
4676 | int mark) | |
acce952b | 4677 | { |
acce952b | 4678 | struct extent_buffer *eb; |
4679 | u64 start = 0; | |
4680 | u64 end; | |
acce952b | 4681 | |
e5860f82 FM |
4682 | while (find_first_extent_bit(dirty_pages, start, &start, &end, |
4683 | mark, NULL)) { | |
91166212 | 4684 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4685 | while (start <= end) { |
0b246afa JM |
4686 | eb = find_extent_buffer(fs_info, start); |
4687 | start += fs_info->nodesize; | |
fd8b2b61 | 4688 | if (!eb) |
acce952b | 4689 | continue; |
c4e54a65 JB |
4690 | |
4691 | btrfs_tree_lock(eb); | |
fd8b2b61 | 4692 | wait_on_extent_buffer_writeback(eb); |
190a8339 | 4693 | btrfs_clear_buffer_dirty(NULL, eb); |
c4e54a65 | 4694 | btrfs_tree_unlock(eb); |
acce952b | 4695 | |
fd8b2b61 | 4696 | free_extent_buffer_stale(eb); |
acce952b | 4697 | } |
4698 | } | |
acce952b | 4699 | } |
4700 | ||
46d81ebd FM |
4701 | static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
4702 | struct extent_io_tree *unpin) | |
acce952b | 4703 | { |
acce952b | 4704 | u64 start; |
4705 | u64 end; | |
acce952b | 4706 | |
acce952b | 4707 | while (1) { |
0e6ec385 FM |
4708 | struct extent_state *cached_state = NULL; |
4709 | ||
fcd5e742 LF |
4710 | /* |
4711 | * The btrfs_finish_extent_commit() may get the same range as | |
4712 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4713 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4714 | * the same extent range. | |
4715 | */ | |
4716 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
e5860f82 FM |
4717 | if (!find_first_extent_bit(unpin, 0, &start, &end, |
4718 | EXTENT_DIRTY, &cached_state)) { | |
fcd5e742 | 4719 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4720 | break; |
fcd5e742 | 4721 | } |
acce952b | 4722 | |
0e6ec385 FM |
4723 | clear_extent_dirty(unpin, start, end, &cached_state); |
4724 | free_extent_state(cached_state); | |
2ff7e61e | 4725 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4726 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4727 | cond_resched(); |
4728 | } | |
acce952b | 4729 | } |
4730 | ||
32da5386 | 4731 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4732 | { |
4733 | struct inode *inode; | |
4734 | ||
4735 | inode = cache->io_ctl.inode; | |
4736 | if (inode) { | |
597441b3 JB |
4737 | unsigned int nofs_flag; |
4738 | ||
4739 | nofs_flag = memalloc_nofs_save(); | |
c79a1751 | 4740 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 JB |
4741 | memalloc_nofs_restore(nofs_flag); |
4742 | ||
c79a1751 LB |
4743 | BTRFS_I(inode)->generation = 0; |
4744 | cache->io_ctl.inode = NULL; | |
4745 | iput(inode); | |
4746 | } | |
bbc37d6e | 4747 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4748 | btrfs_put_block_group(cache); |
4749 | } | |
4750 | ||
4751 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4752 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4753 | { |
32da5386 | 4754 | struct btrfs_block_group *cache; |
c79a1751 LB |
4755 | |
4756 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4757 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4758 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4759 | struct btrfs_block_group, |
c79a1751 | 4760 | dirty_list); |
c79a1751 LB |
4761 | |
4762 | if (!list_empty(&cache->io_list)) { | |
4763 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4764 | list_del_init(&cache->io_list); | |
4765 | btrfs_cleanup_bg_io(cache); | |
4766 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4767 | } | |
4768 | ||
4769 | list_del_init(&cache->dirty_list); | |
4770 | spin_lock(&cache->lock); | |
4771 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4772 | spin_unlock(&cache->lock); | |
4773 | ||
4774 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4775 | btrfs_put_block_group(cache); | |
f66e0209 | 4776 | btrfs_dec_delayed_refs_rsv_bg_updates(fs_info); |
c79a1751 LB |
4777 | spin_lock(&cur_trans->dirty_bgs_lock); |
4778 | } | |
4779 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4780 | ||
45ae2c18 NB |
4781 | /* |
4782 | * Refer to the definition of io_bgs member for details why it's safe | |
4783 | * to use it without any locking | |
4784 | */ | |
c79a1751 LB |
4785 | while (!list_empty(&cur_trans->io_bgs)) { |
4786 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4787 | struct btrfs_block_group, |
c79a1751 | 4788 | io_list); |
c79a1751 LB |
4789 | |
4790 | list_del_init(&cache->io_list); | |
4791 | spin_lock(&cache->lock); | |
4792 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4793 | spin_unlock(&cache->lock); | |
4794 | btrfs_cleanup_bg_io(cache); | |
4795 | } | |
4796 | } | |
4797 | ||
b321a52c BB |
4798 | static void btrfs_free_all_qgroup_pertrans(struct btrfs_fs_info *fs_info) |
4799 | { | |
4800 | struct btrfs_root *gang[8]; | |
4801 | int i; | |
4802 | int ret; | |
4803 | ||
4804 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4805 | while (1) { | |
4806 | ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, | |
4807 | (void **)gang, 0, | |
4808 | ARRAY_SIZE(gang), | |
4809 | BTRFS_ROOT_TRANS_TAG); | |
4810 | if (ret == 0) | |
4811 | break; | |
4812 | for (i = 0; i < ret; i++) { | |
4813 | struct btrfs_root *root = gang[i]; | |
4814 | ||
4815 | btrfs_qgroup_free_meta_all_pertrans(root); | |
4816 | radix_tree_tag_clear(&fs_info->fs_roots_radix, | |
4817 | (unsigned long)root->root_key.objectid, | |
4818 | BTRFS_ROOT_TRANS_TAG); | |
4819 | } | |
4820 | } | |
4821 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4822 | } | |
4823 | ||
49b25e05 | 4824 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4825 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4826 | { |
bbbf7243 NB |
4827 | struct btrfs_device *dev, *tmp; |
4828 | ||
2ff7e61e | 4829 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4830 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4831 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4832 | ||
bbbf7243 NB |
4833 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4834 | post_commit_list) { | |
4835 | list_del_init(&dev->post_commit_list); | |
4836 | } | |
4837 | ||
2ff7e61e | 4838 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4839 | |
4a9d8bde | 4840 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4841 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4842 | |
4a9d8bde | 4843 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4844 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4845 | |
ccdf9b30 | 4846 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4847 | |
2ff7e61e | 4848 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4849 | EXTENT_DIRTY); |
fe119a6e | 4850 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4851 | |
b321a52c BB |
4852 | btrfs_free_all_qgroup_pertrans(fs_info); |
4853 | ||
4a9d8bde MX |
4854 | cur_trans->state =TRANS_STATE_COMPLETED; |
4855 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4856 | } |
4857 | ||
2ff7e61e | 4858 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4859 | { |
4860 | struct btrfs_transaction *t; | |
acce952b | 4861 | |
0b246afa | 4862 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4863 | |
0b246afa JM |
4864 | spin_lock(&fs_info->trans_lock); |
4865 | while (!list_empty(&fs_info->trans_list)) { | |
4866 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 | 4867 | struct btrfs_transaction, list); |
77d20c68 | 4868 | if (t->state >= TRANS_STATE_COMMIT_PREP) { |
9b64f57d | 4869 | refcount_inc(&t->use_count); |
0b246afa | 4870 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4871 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4872 | btrfs_put_transaction(t); |
0b246afa | 4873 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4874 | continue; |
4875 | } | |
0b246afa | 4876 | if (t == fs_info->running_transaction) { |
724e2315 | 4877 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4878 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4879 | /* |
4880 | * We wait for 0 num_writers since we don't hold a trans | |
4881 | * handle open currently for this transaction. | |
4882 | */ | |
4883 | wait_event(t->writer_wait, | |
4884 | atomic_read(&t->num_writers) == 0); | |
4885 | } else { | |
0b246afa | 4886 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4887 | } |
2ff7e61e | 4888 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4889 | |
0b246afa JM |
4890 | spin_lock(&fs_info->trans_lock); |
4891 | if (t == fs_info->running_transaction) | |
4892 | fs_info->running_transaction = NULL; | |
acce952b | 4893 | list_del_init(&t->list); |
0b246afa | 4894 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4895 | |
724e2315 | 4896 | btrfs_put_transaction(t); |
2e4e97ab | 4897 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 4898 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4899 | } |
0b246afa JM |
4900 | spin_unlock(&fs_info->trans_lock); |
4901 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4902 | btrfs_destroy_delayed_inodes(fs_info); |
4903 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4904 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4905 | btrfs_drop_all_logs(fs_info); |
0b246afa | 4906 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4907 | |
4908 | return 0; | |
4909 | } | |
ec7d6dfd | 4910 | |
453e4873 | 4911 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
4912 | { |
4913 | struct btrfs_path *path; | |
4914 | int ret; | |
4915 | struct extent_buffer *l; | |
4916 | struct btrfs_key search_key; | |
4917 | struct btrfs_key found_key; | |
4918 | int slot; | |
4919 | ||
4920 | path = btrfs_alloc_path(); | |
4921 | if (!path) | |
4922 | return -ENOMEM; | |
4923 | ||
4924 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
4925 | search_key.type = -1; | |
4926 | search_key.offset = (u64)-1; | |
4927 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
4928 | if (ret < 0) | |
4929 | goto error; | |
a6724290 DS |
4930 | if (ret == 0) { |
4931 | /* | |
4932 | * Key with offset -1 found, there would have to exist a root | |
4933 | * with such id, but this is out of valid range. | |
4934 | */ | |
4935 | ret = -EUCLEAN; | |
4936 | goto error; | |
4937 | } | |
ec7d6dfd NB |
4938 | if (path->slots[0] > 0) { |
4939 | slot = path->slots[0] - 1; | |
4940 | l = path->nodes[0]; | |
4941 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
4942 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
4943 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 4944 | } else { |
23125104 | 4945 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
4946 | } |
4947 | ret = 0; | |
4948 | error: | |
4949 | btrfs_free_path(path); | |
4950 | return ret; | |
4951 | } | |
4952 | ||
543068a2 | 4953 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
4954 | { |
4955 | int ret; | |
4956 | mutex_lock(&root->objectid_mutex); | |
4957 | ||
6b8fad57 | 4958 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
4959 | btrfs_warn(root->fs_info, |
4960 | "the objectid of root %llu reaches its highest value", | |
4961 | root->root_key.objectid); | |
4962 | ret = -ENOSPC; | |
4963 | goto out; | |
4964 | } | |
4965 | ||
23125104 | 4966 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
4967 | ret = 0; |
4968 | out: | |
4969 | mutex_unlock(&root->objectid_mutex); | |
4970 | return ret; | |
4971 | } |