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> |
0f7d52f4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
d397712b | 10 | #include <linux/buffer_head.h> |
ce9adaa5 | 11 | #include <linux/workqueue.h> |
a74a4b97 | 12 | #include <linux/kthread.h> |
5a0e3ad6 | 13 | #include <linux/slab.h> |
784b4e29 | 14 | #include <linux/migrate.h> |
7a36ddec | 15 | #include <linux/ratelimit.h> |
6463fe58 | 16 | #include <linux/uuid.h> |
803b2f54 | 17 | #include <linux/semaphore.h> |
540adea3 | 18 | #include <linux/error-injection.h> |
9678c543 | 19 | #include <linux/crc32c.h> |
b89f6d1f | 20 | #include <linux/sched/mm.h> |
7e75bf3f | 21 | #include <asm/unaligned.h> |
6d97c6e3 | 22 | #include <crypto/hash.h> |
eb60ceac CM |
23 | #include "ctree.h" |
24 | #include "disk-io.h" | |
e089f05c | 25 | #include "transaction.h" |
0f7d52f4 | 26 | #include "btrfs_inode.h" |
0b86a832 | 27 | #include "volumes.h" |
db94535d | 28 | #include "print-tree.h" |
925baedd | 29 | #include "locking.h" |
e02119d5 | 30 | #include "tree-log.h" |
fa9c0d79 | 31 | #include "free-space-cache.h" |
70f6d82e | 32 | #include "free-space-tree.h" |
581bb050 | 33 | #include "inode-map.h" |
21adbd5c | 34 | #include "check-integrity.h" |
606686ee | 35 | #include "rcu-string.h" |
8dabb742 | 36 | #include "dev-replace.h" |
53b381b3 | 37 | #include "raid56.h" |
5ac1d209 | 38 | #include "sysfs.h" |
fcebe456 | 39 | #include "qgroup.h" |
ebb8765b | 40 | #include "compression.h" |
557ea5dd | 41 | #include "tree-checker.h" |
fd708b81 | 42 | #include "ref-verify.h" |
aac0023c | 43 | #include "block-group.h" |
b0643e59 | 44 | #include "discard.h" |
eb60ceac | 45 | |
319e4d06 QW |
46 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
47 | BTRFS_HEADER_FLAG_RELOC |\ | |
48 | BTRFS_SUPER_FLAG_ERROR |\ | |
49 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
50 | BTRFS_SUPER_FLAG_METADUMP |\ |
51 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 52 | |
e8c9f186 | 53 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 54 | static void end_workqueue_fn(struct btrfs_work *work); |
143bede5 | 55 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 56 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 57 | struct btrfs_fs_info *fs_info); |
143bede5 | 58 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 59 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 60 | struct extent_io_tree *dirty_pages, |
61 | int mark); | |
2ff7e61e | 62 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 63 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
64 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
65 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 66 | |
d352ac68 | 67 | /* |
97eb6b69 DS |
68 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
69 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
70 | * by writes to insert metadata for new file extents after IO is complete. |
71 | */ | |
97eb6b69 | 72 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
73 | struct bio *bio; |
74 | bio_end_io_t *end_io; | |
75 | void *private; | |
76 | struct btrfs_fs_info *info; | |
4e4cbee9 | 77 | blk_status_t status; |
bfebd8b5 | 78 | enum btrfs_wq_endio_type metadata; |
8b712842 | 79 | struct btrfs_work work; |
ce9adaa5 | 80 | }; |
0da5468f | 81 | |
97eb6b69 DS |
82 | static struct kmem_cache *btrfs_end_io_wq_cache; |
83 | ||
84 | int __init btrfs_end_io_wq_init(void) | |
85 | { | |
86 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
87 | sizeof(struct btrfs_end_io_wq), | |
88 | 0, | |
fba4b697 | 89 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
90 | NULL); |
91 | if (!btrfs_end_io_wq_cache) | |
92 | return -ENOMEM; | |
93 | return 0; | |
94 | } | |
95 | ||
e67c718b | 96 | void __cold btrfs_end_io_wq_exit(void) |
97eb6b69 | 97 | { |
5598e900 | 98 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
99 | } |
100 | ||
d352ac68 CM |
101 | /* |
102 | * async submit bios are used to offload expensive checksumming | |
103 | * onto the worker threads. They checksum file and metadata bios | |
104 | * just before they are sent down the IO stack. | |
105 | */ | |
44b8bd7e | 106 | struct async_submit_bio { |
c6100a4b | 107 | void *private_data; |
44b8bd7e | 108 | struct bio *bio; |
a758781d | 109 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 110 | int mirror_num; |
eaf25d93 CM |
111 | /* |
112 | * bio_offset is optional, can be used if the pages in the bio | |
113 | * can't tell us where in the file the bio should go | |
114 | */ | |
115 | u64 bio_offset; | |
8b712842 | 116 | struct btrfs_work work; |
4e4cbee9 | 117 | blk_status_t status; |
44b8bd7e CM |
118 | }; |
119 | ||
85d4e461 CM |
120 | /* |
121 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
122 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
123 | * the level the eb occupies in the tree. | |
124 | * | |
125 | * Different roots are used for different purposes and may nest inside each | |
126 | * other and they require separate keysets. As lockdep keys should be | |
127 | * static, assign keysets according to the purpose of the root as indicated | |
4fd786e6 MT |
128 | * by btrfs_root->root_key.objectid. This ensures that all special purpose |
129 | * roots have separate keysets. | |
4008c04a | 130 | * |
85d4e461 CM |
131 | * Lock-nesting across peer nodes is always done with the immediate parent |
132 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
133 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 134 | * |
85d4e461 CM |
135 | * The key is set by the readpage_end_io_hook after the buffer has passed |
136 | * csum validation but before the pages are unlocked. It is also set by | |
137 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 138 | * |
85d4e461 CM |
139 | * We also add a check to make sure the highest level of the tree is the |
140 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
141 | * needs update as well. | |
4008c04a CM |
142 | */ |
143 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
144 | # if BTRFS_MAX_LEVEL != 8 | |
145 | # error | |
146 | # endif | |
85d4e461 CM |
147 | |
148 | static struct btrfs_lockdep_keyset { | |
149 | u64 id; /* root objectid */ | |
150 | const char *name_stem; /* lock name stem */ | |
151 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
152 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
153 | } btrfs_lockdep_keysets[] = { | |
154 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
155 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
156 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
157 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
158 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
159 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 160 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
161 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
162 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
163 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 164 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 165 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 166 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 167 | }; |
85d4e461 CM |
168 | |
169 | void __init btrfs_init_lockdep(void) | |
170 | { | |
171 | int i, j; | |
172 | ||
173 | /* initialize lockdep class names */ | |
174 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
175 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
176 | ||
177 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
178 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
179 | "btrfs-%s-%02d", ks->name_stem, j); | |
180 | } | |
181 | } | |
182 | ||
183 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
184 | int level) | |
185 | { | |
186 | struct btrfs_lockdep_keyset *ks; | |
187 | ||
188 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
189 | ||
190 | /* find the matching keyset, id 0 is the default entry */ | |
191 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
192 | if (ks->id == objectid) | |
193 | break; | |
194 | ||
195 | lockdep_set_class_and_name(&eb->lock, | |
196 | &ks->keys[level], ks->names[level]); | |
197 | } | |
198 | ||
4008c04a CM |
199 | #endif |
200 | ||
d352ac68 CM |
201 | /* |
202 | * extents on the btree inode are pretty simple, there's one extent | |
203 | * that covers the entire device | |
204 | */ | |
6af49dbd | 205 | struct extent_map *btree_get_extent(struct btrfs_inode *inode, |
39b07b5d OS |
206 | struct page *page, size_t pg_offset, |
207 | u64 start, u64 len) | |
7eccb903 | 208 | { |
fc4f21b1 | 209 | struct extent_map_tree *em_tree = &inode->extent_tree; |
5f39d397 CM |
210 | struct extent_map *em; |
211 | int ret; | |
212 | ||
890871be | 213 | read_lock(&em_tree->lock); |
d1310b2e | 214 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d | 215 | if (em) { |
890871be | 216 | read_unlock(&em_tree->lock); |
5f39d397 | 217 | goto out; |
a061fc8d | 218 | } |
890871be | 219 | read_unlock(&em_tree->lock); |
7b13b7b1 | 220 | |
172ddd60 | 221 | em = alloc_extent_map(); |
5f39d397 CM |
222 | if (!em) { |
223 | em = ERR_PTR(-ENOMEM); | |
224 | goto out; | |
225 | } | |
226 | em->start = 0; | |
0afbaf8c | 227 | em->len = (u64)-1; |
c8b97818 | 228 | em->block_len = (u64)-1; |
5f39d397 | 229 | em->block_start = 0; |
d1310b2e | 230 | |
890871be | 231 | write_lock(&em_tree->lock); |
09a2a8f9 | 232 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
233 | if (ret == -EEXIST) { |
234 | free_extent_map(em); | |
7b13b7b1 | 235 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 236 | if (!em) |
0433f20d | 237 | em = ERR_PTR(-EIO); |
5f39d397 | 238 | } else if (ret) { |
7b13b7b1 | 239 | free_extent_map(em); |
0433f20d | 240 | em = ERR_PTR(ret); |
5f39d397 | 241 | } |
890871be | 242 | write_unlock(&em_tree->lock); |
7b13b7b1 | 243 | |
5f39d397 CM |
244 | out: |
245 | return em; | |
7eccb903 CM |
246 | } |
247 | ||
d352ac68 | 248 | /* |
2996e1f8 JT |
249 | * Compute the csum of a btree block and store the result to provided buffer. |
250 | * | |
251 | * Returns error if the extent buffer cannot be mapped. | |
d352ac68 | 252 | */ |
2996e1f8 | 253 | static int csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 254 | { |
d5178578 JT |
255 | struct btrfs_fs_info *fs_info = buf->fs_info; |
256 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); | |
19c00ddc CM |
257 | unsigned long len; |
258 | unsigned long cur_len; | |
259 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
260 | char *kaddr; |
261 | unsigned long map_start; | |
262 | unsigned long map_len; | |
263 | int err; | |
d5178578 JT |
264 | |
265 | shash->tfm = fs_info->csum_shash; | |
266 | crypto_shash_init(shash); | |
19c00ddc CM |
267 | |
268 | len = buf->len - offset; | |
d5178578 | 269 | |
d397712b | 270 | while (len > 0) { |
d2e174d5 JT |
271 | /* |
272 | * Note: we don't need to check for the err == 1 case here, as | |
273 | * with the given combination of 'start = BTRFS_CSUM_SIZE (32)' | |
274 | * and 'min_len = 32' and the currently implemented mapping | |
275 | * algorithm we cannot cross a page boundary. | |
276 | */ | |
19c00ddc | 277 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 278 | &kaddr, &map_start, &map_len); |
c53839fc | 279 | if (WARN_ON(err)) |
8bd98f0e | 280 | return err; |
19c00ddc | 281 | cur_len = min(len, map_len - (offset - map_start)); |
d5178578 | 282 | crypto_shash_update(shash, kaddr + offset - map_start, cur_len); |
19c00ddc CM |
283 | len -= cur_len; |
284 | offset += cur_len; | |
19c00ddc | 285 | } |
71a63551 | 286 | memset(result, 0, BTRFS_CSUM_SIZE); |
607d432d | 287 | |
d5178578 | 288 | crypto_shash_final(shash, result); |
19c00ddc | 289 | |
19c00ddc CM |
290 | return 0; |
291 | } | |
292 | ||
d352ac68 CM |
293 | /* |
294 | * we can't consider a given block up to date unless the transid of the | |
295 | * block matches the transid in the parent node's pointer. This is how we | |
296 | * detect blocks that either didn't get written at all or got written | |
297 | * in the wrong place. | |
298 | */ | |
1259ab75 | 299 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
300 | struct extent_buffer *eb, u64 parent_transid, |
301 | int atomic) | |
1259ab75 | 302 | { |
2ac55d41 | 303 | struct extent_state *cached_state = NULL; |
1259ab75 | 304 | int ret; |
2755a0de | 305 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
306 | |
307 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
308 | return 0; | |
309 | ||
b9fab919 CM |
310 | if (atomic) |
311 | return -EAGAIN; | |
312 | ||
a26e8c9f JB |
313 | if (need_lock) { |
314 | btrfs_tree_read_lock(eb); | |
300aa896 | 315 | btrfs_set_lock_blocking_read(eb); |
a26e8c9f JB |
316 | } |
317 | ||
2ac55d41 | 318 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 319 | &cached_state); |
0b32f4bb | 320 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
321 | btrfs_header_generation(eb) == parent_transid) { |
322 | ret = 0; | |
323 | goto out; | |
324 | } | |
94647322 DS |
325 | btrfs_err_rl(eb->fs_info, |
326 | "parent transid verify failed on %llu wanted %llu found %llu", | |
327 | eb->start, | |
29549aec | 328 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 329 | ret = 1; |
a26e8c9f JB |
330 | |
331 | /* | |
332 | * Things reading via commit roots that don't have normal protection, | |
333 | * like send, can have a really old block in cache that may point at a | |
01327610 | 334 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
335 | * if we find an eb that is under IO (dirty/writeback) because we could |
336 | * end up reading in the stale data and then writing it back out and | |
337 | * making everybody very sad. | |
338 | */ | |
339 | if (!extent_buffer_under_io(eb)) | |
340 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 341 | out: |
2ac55d41 | 342 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 343 | &cached_state); |
472b909f JB |
344 | if (need_lock) |
345 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 346 | return ret; |
1259ab75 CM |
347 | } |
348 | ||
e7e16f48 JT |
349 | static bool btrfs_supported_super_csum(u16 csum_type) |
350 | { | |
351 | switch (csum_type) { | |
352 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 353 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 354 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 355 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
356 | return true; |
357 | default: | |
358 | return false; | |
359 | } | |
360 | } | |
361 | ||
1104a885 DS |
362 | /* |
363 | * Return 0 if the superblock checksum type matches the checksum value of that | |
364 | * algorithm. Pass the raw disk superblock data. | |
365 | */ | |
ab8d0fc4 JM |
366 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
367 | char *raw_disk_sb) | |
1104a885 DS |
368 | { |
369 | struct btrfs_super_block *disk_sb = | |
370 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 371 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
372 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
373 | ||
374 | shash->tfm = fs_info->csum_shash; | |
375 | crypto_shash_init(shash); | |
1104a885 | 376 | |
51bce6c9 JT |
377 | /* |
378 | * The super_block structure does not span the whole | |
379 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
380 | * filled with zeros and is included in the checksum. | |
381 | */ | |
d5178578 JT |
382 | crypto_shash_update(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
383 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
384 | crypto_shash_final(shash, result); | |
1104a885 | 385 | |
51bce6c9 JT |
386 | if (memcmp(disk_sb->csum, result, btrfs_super_csum_size(disk_sb))) |
387 | return 1; | |
1104a885 | 388 | |
e7e16f48 | 389 | return 0; |
1104a885 DS |
390 | } |
391 | ||
e064d5e9 | 392 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 393 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 394 | { |
e064d5e9 | 395 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
396 | int found_level; |
397 | struct btrfs_key found_key; | |
398 | int ret; | |
399 | ||
400 | found_level = btrfs_header_level(eb); | |
401 | if (found_level != level) { | |
63489055 QW |
402 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
403 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
404 | btrfs_err(fs_info, |
405 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
406 | eb->start, level, found_level); | |
581c1760 QW |
407 | return -EIO; |
408 | } | |
409 | ||
410 | if (!first_key) | |
411 | return 0; | |
412 | ||
5d41be6f QW |
413 | /* |
414 | * For live tree block (new tree blocks in current transaction), | |
415 | * we need proper lock context to avoid race, which is impossible here. | |
416 | * So we only checks tree blocks which is read from disk, whose | |
417 | * generation <= fs_info->last_trans_committed. | |
418 | */ | |
419 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
420 | return 0; | |
62fdaa52 QW |
421 | |
422 | /* We have @first_key, so this @eb must have at least one item */ | |
423 | if (btrfs_header_nritems(eb) == 0) { | |
424 | btrfs_err(fs_info, | |
425 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
426 | eb->start); | |
427 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
428 | return -EUCLEAN; | |
429 | } | |
430 | ||
581c1760 QW |
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 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
436 | ||
581c1760 | 437 | if (ret) { |
63489055 QW |
438 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
439 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 440 | btrfs_err(fs_info, |
ff76a864 LB |
441 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
442 | eb->start, parent_transid, first_key->objectid, | |
443 | first_key->type, first_key->offset, | |
444 | found_key.objectid, found_key.type, | |
445 | found_key.offset); | |
581c1760 | 446 | } |
581c1760 QW |
447 | return ret; |
448 | } | |
449 | ||
d352ac68 CM |
450 | /* |
451 | * helper to read a given tree block, doing retries as required when | |
452 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
453 | * |
454 | * @parent_transid: expected transid, skip check if 0 | |
455 | * @level: expected level, mandatory check | |
456 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 457 | */ |
5ab12d1f | 458 | static int btree_read_extent_buffer_pages(struct extent_buffer *eb, |
581c1760 QW |
459 | u64 parent_transid, int level, |
460 | struct btrfs_key *first_key) | |
f188591e | 461 | { |
5ab12d1f | 462 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 463 | struct extent_io_tree *io_tree; |
ea466794 | 464 | int failed = 0; |
f188591e CM |
465 | int ret; |
466 | int num_copies = 0; | |
467 | int mirror_num = 0; | |
ea466794 | 468 | int failed_mirror = 0; |
f188591e | 469 | |
0b246afa | 470 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 471 | while (1) { |
f8397d69 | 472 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 473 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 474 | if (!ret) { |
581c1760 | 475 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 476 | parent_transid, 0)) |
256dd1bb | 477 | ret = -EIO; |
e064d5e9 | 478 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 479 | first_key, parent_transid)) |
581c1760 QW |
480 | ret = -EUCLEAN; |
481 | else | |
482 | break; | |
256dd1bb | 483 | } |
d397712b | 484 | |
0b246afa | 485 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 486 | eb->start, eb->len); |
4235298e | 487 | if (num_copies == 1) |
ea466794 | 488 | break; |
4235298e | 489 | |
5cf1ab56 JB |
490 | if (!failed_mirror) { |
491 | failed = 1; | |
492 | failed_mirror = eb->read_mirror; | |
493 | } | |
494 | ||
f188591e | 495 | mirror_num++; |
ea466794 JB |
496 | if (mirror_num == failed_mirror) |
497 | mirror_num++; | |
498 | ||
4235298e | 499 | if (mirror_num > num_copies) |
ea466794 | 500 | break; |
f188591e | 501 | } |
ea466794 | 502 | |
c0901581 | 503 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 504 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
505 | |
506 | return ret; | |
f188591e | 507 | } |
19c00ddc | 508 | |
d352ac68 | 509 | /* |
d397712b CM |
510 | * checksum a dirty tree block before IO. This has extra checks to make sure |
511 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 512 | */ |
d397712b | 513 | |
01d58472 | 514 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 515 | { |
4eee4fa4 | 516 | u64 start = page_offset(page); |
19c00ddc | 517 | u64 found_start; |
2996e1f8 JT |
518 | u8 result[BTRFS_CSUM_SIZE]; |
519 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
19c00ddc | 520 | struct extent_buffer *eb; |
8d47a0d8 | 521 | int ret; |
f188591e | 522 | |
4f2de97a JB |
523 | eb = (struct extent_buffer *)page->private; |
524 | if (page != eb->pages[0]) | |
525 | return 0; | |
0f805531 | 526 | |
19c00ddc | 527 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
528 | /* |
529 | * Please do not consolidate these warnings into a single if. | |
530 | * It is useful to know what went wrong. | |
531 | */ | |
532 | if (WARN_ON(found_start != start)) | |
533 | return -EUCLEAN; | |
534 | if (WARN_ON(!PageUptodate(page))) | |
535 | return -EUCLEAN; | |
536 | ||
de37aa51 | 537 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, |
0f805531 AL |
538 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); |
539 | ||
2996e1f8 JT |
540 | if (csum_tree_block(eb, result)) |
541 | return -EINVAL; | |
542 | ||
8d47a0d8 QW |
543 | if (btrfs_header_level(eb)) |
544 | ret = btrfs_check_node(eb); | |
545 | else | |
546 | ret = btrfs_check_leaf_full(eb); | |
547 | ||
548 | if (ret < 0) { | |
c06631b0 | 549 | btrfs_print_tree(eb, 0); |
8d47a0d8 QW |
550 | btrfs_err(fs_info, |
551 | "block=%llu write time tree block corruption detected", | |
552 | eb->start); | |
c06631b0 | 553 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); |
8d47a0d8 QW |
554 | return ret; |
555 | } | |
2996e1f8 | 556 | write_extent_buffer(eb, result, 0, csum_size); |
8d47a0d8 | 557 | |
2996e1f8 | 558 | return 0; |
19c00ddc CM |
559 | } |
560 | ||
b0c9b3b0 | 561 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 562 | { |
b0c9b3b0 | 563 | struct btrfs_fs_info *fs_info = eb->fs_info; |
01d58472 | 564 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
44880fdc | 565 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c YZ |
566 | int ret = 1; |
567 | ||
0a4e5586 | 568 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c | 569 | while (fs_devices) { |
7239ff4b NB |
570 | u8 *metadata_uuid; |
571 | ||
572 | /* | |
573 | * Checking the incompat flag is only valid for the current | |
574 | * fs. For seed devices it's forbidden to have their uuid | |
575 | * changed so reading ->fsid in this case is fine | |
576 | */ | |
577 | if (fs_devices == fs_info->fs_devices && | |
578 | btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
579 | metadata_uuid = fs_devices->metadata_uuid; | |
580 | else | |
581 | metadata_uuid = fs_devices->fsid; | |
582 | ||
583 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) { | |
2b82032c YZ |
584 | ret = 0; |
585 | break; | |
586 | } | |
587 | fs_devices = fs_devices->seed; | |
588 | } | |
589 | return ret; | |
590 | } | |
591 | ||
facc8a22 MX |
592 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
593 | u64 phy_offset, struct page *page, | |
594 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 595 | { |
ce9adaa5 CM |
596 | u64 found_start; |
597 | int found_level; | |
ce9adaa5 CM |
598 | struct extent_buffer *eb; |
599 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 600 | struct btrfs_fs_info *fs_info = root->fs_info; |
2996e1f8 | 601 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
f188591e | 602 | int ret = 0; |
2996e1f8 | 603 | u8 result[BTRFS_CSUM_SIZE]; |
727011e0 | 604 | int reads_done; |
ce9adaa5 | 605 | |
ce9adaa5 CM |
606 | if (!page->private) |
607 | goto out; | |
d397712b | 608 | |
4f2de97a | 609 | eb = (struct extent_buffer *)page->private; |
d397712b | 610 | |
0b32f4bb JB |
611 | /* the pending IO might have been the only thing that kept this buffer |
612 | * in memory. Make sure we have a ref for all this other checks | |
613 | */ | |
67439dad | 614 | atomic_inc(&eb->refs); |
0b32f4bb JB |
615 | |
616 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
617 | if (!reads_done) |
618 | goto err; | |
f188591e | 619 | |
5cf1ab56 | 620 | eb->read_mirror = mirror; |
656f30db | 621 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
622 | ret = -EIO; |
623 | goto err; | |
624 | } | |
625 | ||
ce9adaa5 | 626 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 627 | if (found_start != eb->start) { |
893bf4b1 SY |
628 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
629 | eb->start, found_start); | |
f188591e | 630 | ret = -EIO; |
ce9adaa5 CM |
631 | goto err; |
632 | } | |
b0c9b3b0 | 633 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
634 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
635 | eb->start); | |
1259ab75 CM |
636 | ret = -EIO; |
637 | goto err; | |
638 | } | |
ce9adaa5 | 639 | found_level = btrfs_header_level(eb); |
1c24c3ce | 640 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
641 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
642 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce JB |
643 | ret = -EIO; |
644 | goto err; | |
645 | } | |
ce9adaa5 | 646 | |
85d4e461 CM |
647 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
648 | eb, found_level); | |
4008c04a | 649 | |
2996e1f8 | 650 | ret = csum_tree_block(eb, result); |
8bd98f0e | 651 | if (ret) |
a826d6dc | 652 | goto err; |
a826d6dc | 653 | |
2996e1f8 JT |
654 | if (memcmp_extent_buffer(eb, result, 0, csum_size)) { |
655 | u32 val; | |
656 | u32 found = 0; | |
657 | ||
658 | memcpy(&found, result, csum_size); | |
659 | ||
660 | read_extent_buffer(eb, &val, 0, csum_size); | |
661 | btrfs_warn_rl(fs_info, | |
662 | "%s checksum verify failed on %llu wanted %x found %x level %d", | |
663 | fs_info->sb->s_id, eb->start, | |
664 | val, found, btrfs_header_level(eb)); | |
665 | ret = -EUCLEAN; | |
666 | goto err; | |
667 | } | |
668 | ||
a826d6dc JB |
669 | /* |
670 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
671 | * that we don't try and read the other copies of this block, just | |
672 | * return -EIO. | |
673 | */ | |
1c4360ee | 674 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
675 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
676 | ret = -EIO; | |
677 | } | |
ce9adaa5 | 678 | |
813fd1dc | 679 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
680 | ret = -EIO; |
681 | ||
0b32f4bb JB |
682 | if (!ret) |
683 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
684 | else |
685 | btrfs_err(fs_info, | |
686 | "block=%llu read time tree block corruption detected", | |
687 | eb->start); | |
ce9adaa5 | 688 | err: |
79fb65a1 JB |
689 | if (reads_done && |
690 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 691 | btree_readahead_hook(eb, ret); |
4bb31e92 | 692 | |
53b381b3 DW |
693 | if (ret) { |
694 | /* | |
695 | * our io error hook is going to dec the io pages | |
696 | * again, we have to make sure it has something | |
697 | * to decrement | |
698 | */ | |
699 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 700 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 701 | } |
0b32f4bb | 702 | free_extent_buffer(eb); |
ce9adaa5 | 703 | out: |
f188591e | 704 | return ret; |
ce9adaa5 CM |
705 | } |
706 | ||
4246a0b6 | 707 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 708 | { |
97eb6b69 | 709 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 710 | struct btrfs_fs_info *fs_info; |
9e0af237 | 711 | struct btrfs_workqueue *wq; |
ce9adaa5 | 712 | |
ce9adaa5 | 713 | fs_info = end_io_wq->info; |
4e4cbee9 | 714 | end_io_wq->status = bio->bi_status; |
d20f7043 | 715 | |
37226b21 | 716 | if (bio_op(bio) == REQ_OP_WRITE) { |
a0cac0ec | 717 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 718 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 719 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 720 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 721 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 722 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 723 | else |
9e0af237 | 724 | wq = fs_info->endio_write_workers; |
d20f7043 | 725 | } else { |
a0cac0ec | 726 | if (unlikely(end_io_wq->metadata == BTRFS_WQ_ENDIO_DIO_REPAIR)) |
8b110e39 | 727 | wq = fs_info->endio_repair_workers; |
a0cac0ec | 728 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 729 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 730 | else if (end_io_wq->metadata) |
9e0af237 | 731 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 732 | else |
9e0af237 | 733 | wq = fs_info->endio_workers; |
d20f7043 | 734 | } |
9e0af237 | 735 | |
a0cac0ec | 736 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 737 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
738 | } |
739 | ||
4e4cbee9 | 740 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 741 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 742 | { |
97eb6b69 | 743 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 744 | |
97eb6b69 | 745 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 746 | if (!end_io_wq) |
4e4cbee9 | 747 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
748 | |
749 | end_io_wq->private = bio->bi_private; | |
750 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 751 | end_io_wq->info = info; |
4e4cbee9 | 752 | end_io_wq->status = 0; |
ce9adaa5 | 753 | end_io_wq->bio = bio; |
22c59948 | 754 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
755 | |
756 | bio->bi_private = end_io_wq; | |
757 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
758 | return 0; |
759 | } | |
760 | ||
4a69a410 CM |
761 | static void run_one_async_start(struct btrfs_work *work) |
762 | { | |
4a69a410 | 763 | struct async_submit_bio *async; |
4e4cbee9 | 764 | blk_status_t ret; |
4a69a410 CM |
765 | |
766 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 767 | ret = async->submit_bio_start(async->private_data, async->bio, |
79787eaa JM |
768 | async->bio_offset); |
769 | if (ret) | |
4e4cbee9 | 770 | async->status = ret; |
4a69a410 CM |
771 | } |
772 | ||
06ea01b1 DS |
773 | /* |
774 | * In order to insert checksums into the metadata in large chunks, we wait | |
775 | * until bio submission time. All the pages in the bio are checksummed and | |
776 | * sums are attached onto the ordered extent record. | |
777 | * | |
778 | * At IO completion time the csums attached on the ordered extent record are | |
779 | * inserted into the tree. | |
780 | */ | |
4a69a410 | 781 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 782 | { |
8b712842 | 783 | struct async_submit_bio *async; |
06ea01b1 DS |
784 | struct inode *inode; |
785 | blk_status_t ret; | |
8b712842 CM |
786 | |
787 | async = container_of(work, struct async_submit_bio, work); | |
06ea01b1 | 788 | inode = async->private_data; |
4854ddd0 | 789 | |
bb7ab3b9 | 790 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
791 | if (async->status) { |
792 | async->bio->bi_status = async->status; | |
4246a0b6 | 793 | bio_endio(async->bio); |
79787eaa JM |
794 | return; |
795 | } | |
796 | ||
ec39f769 CM |
797 | /* |
798 | * All of the bios that pass through here are from async helpers. | |
799 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
800 | * This changes nothing when cgroups aren't in use. | |
801 | */ | |
802 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 803 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
804 | if (ret) { |
805 | async->bio->bi_status = ret; | |
806 | bio_endio(async->bio); | |
807 | } | |
4a69a410 CM |
808 | } |
809 | ||
810 | static void run_one_async_free(struct btrfs_work *work) | |
811 | { | |
812 | struct async_submit_bio *async; | |
813 | ||
814 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
815 | kfree(async); |
816 | } | |
817 | ||
8c27cb35 LT |
818 | blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, |
819 | int mirror_num, unsigned long bio_flags, | |
820 | u64 bio_offset, void *private_data, | |
e288c080 | 821 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e CM |
822 | { |
823 | struct async_submit_bio *async; | |
824 | ||
825 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
826 | if (!async) | |
4e4cbee9 | 827 | return BLK_STS_RESOURCE; |
44b8bd7e | 828 | |
c6100a4b | 829 | async->private_data = private_data; |
44b8bd7e CM |
830 | async->bio = bio; |
831 | async->mirror_num = mirror_num; | |
4a69a410 | 832 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 833 | |
a0cac0ec OS |
834 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
835 | run_one_async_free); | |
4a69a410 | 836 | |
eaf25d93 | 837 | async->bio_offset = bio_offset; |
8c8bee1d | 838 | |
4e4cbee9 | 839 | async->status = 0; |
79787eaa | 840 | |
67f055c7 | 841 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 842 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 843 | |
5cdc7ad3 | 844 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
845 | return 0; |
846 | } | |
847 | ||
4e4cbee9 | 848 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 849 | { |
2c30c71b | 850 | struct bio_vec *bvec; |
ce3ed71a | 851 | struct btrfs_root *root; |
2b070cfe | 852 | int ret = 0; |
6dc4f100 | 853 | struct bvec_iter_all iter_all; |
ce3ed71a | 854 | |
c09abff8 | 855 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 856 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 857 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 858 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
859 | if (ret) |
860 | break; | |
ce3ed71a | 861 | } |
2c30c71b | 862 | |
4e4cbee9 | 863 | return errno_to_blk_status(ret); |
ce3ed71a CM |
864 | } |
865 | ||
d0ee3934 | 866 | static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, |
8c27cb35 | 867 | u64 bio_offset) |
22c59948 | 868 | { |
8b712842 CM |
869 | /* |
870 | * when we're called for a write, we're already in the async | |
5443be45 | 871 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 872 | */ |
79787eaa | 873 | return btree_csum_one_bio(bio); |
4a69a410 | 874 | } |
22c59948 | 875 | |
9b4e675a DS |
876 | static int check_async_write(struct btrfs_fs_info *fs_info, |
877 | struct btrfs_inode *bi) | |
de0022b9 | 878 | { |
6300463b LB |
879 | if (atomic_read(&bi->sync_writers)) |
880 | return 0; | |
9b4e675a | 881 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
de0022b9 | 882 | return 0; |
de0022b9 JB |
883 | return 1; |
884 | } | |
885 | ||
a56b1c7b | 886 | static blk_status_t btree_submit_bio_hook(struct inode *inode, struct bio *bio, |
50489a57 NB |
887 | int mirror_num, |
888 | unsigned long bio_flags) | |
44b8bd7e | 889 | { |
0b246afa | 890 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
9b4e675a | 891 | int async = check_async_write(fs_info, BTRFS_I(inode)); |
4e4cbee9 | 892 | blk_status_t ret; |
cad321ad | 893 | |
37226b21 | 894 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
895 | /* |
896 | * called for a read, do the setup so that checksum validation | |
897 | * can happen in the async kernel threads | |
898 | */ | |
0b246afa JM |
899 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
900 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 901 | if (ret) |
61891923 | 902 | goto out_w_error; |
08635bae | 903 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
de0022b9 JB |
904 | } else if (!async) { |
905 | ret = btree_csum_one_bio(bio); | |
906 | if (ret) | |
61891923 | 907 | goto out_w_error; |
08635bae | 908 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
61891923 SB |
909 | } else { |
910 | /* | |
911 | * kthread helpers are used to submit writes so that | |
912 | * checksumming can happen in parallel across all CPUs | |
913 | */ | |
c6100a4b | 914 | ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, |
e68f2ee7 | 915 | 0, inode, btree_submit_bio_start); |
44b8bd7e | 916 | } |
d313d7a3 | 917 | |
4246a0b6 CH |
918 | if (ret) |
919 | goto out_w_error; | |
920 | return 0; | |
921 | ||
61891923 | 922 | out_w_error: |
4e4cbee9 | 923 | bio->bi_status = ret; |
4246a0b6 | 924 | bio_endio(bio); |
61891923 | 925 | return ret; |
44b8bd7e CM |
926 | } |
927 | ||
3dd1462e | 928 | #ifdef CONFIG_MIGRATION |
784b4e29 | 929 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
930 | struct page *newpage, struct page *page, |
931 | enum migrate_mode mode) | |
784b4e29 CM |
932 | { |
933 | /* | |
934 | * we can't safely write a btree page from here, | |
935 | * we haven't done the locking hook | |
936 | */ | |
937 | if (PageDirty(page)) | |
938 | return -EAGAIN; | |
939 | /* | |
940 | * Buffers may be managed in a filesystem specific way. | |
941 | * We must have no buffers or drop them. | |
942 | */ | |
943 | if (page_has_private(page) && | |
944 | !try_to_release_page(page, GFP_KERNEL)) | |
945 | return -EAGAIN; | |
a6bc32b8 | 946 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 947 | } |
3dd1462e | 948 | #endif |
784b4e29 | 949 | |
0da5468f CM |
950 | |
951 | static int btree_writepages(struct address_space *mapping, | |
952 | struct writeback_control *wbc) | |
953 | { | |
e2d84521 MX |
954 | struct btrfs_fs_info *fs_info; |
955 | int ret; | |
956 | ||
d8d5f3e1 | 957 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
958 | |
959 | if (wbc->for_kupdate) | |
960 | return 0; | |
961 | ||
e2d84521 | 962 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 963 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
964 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
965 | BTRFS_DIRTY_METADATA_THRESH, | |
966 | fs_info->dirty_metadata_batch); | |
e2d84521 | 967 | if (ret < 0) |
793955bc | 968 | return 0; |
793955bc | 969 | } |
0b32f4bb | 970 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
971 | } |
972 | ||
b2950863 | 973 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 974 | { |
d1310b2e CM |
975 | struct extent_io_tree *tree; |
976 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 977 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 978 | } |
22b0ebda | 979 | |
70dec807 | 980 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 981 | { |
98509cfc | 982 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 983 | return 0; |
0c4e538b | 984 | |
f7a52a40 | 985 | return try_release_extent_buffer(page); |
d98237b3 CM |
986 | } |
987 | ||
d47992f8 LC |
988 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
989 | unsigned int length) | |
d98237b3 | 990 | { |
d1310b2e CM |
991 | struct extent_io_tree *tree; |
992 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
993 | extent_invalidatepage(tree, page, offset); |
994 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 995 | if (PagePrivate(page)) { |
efe120a0 FH |
996 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
997 | "page private not zero on page %llu", | |
998 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
999 | ClearPagePrivate(page); |
1000 | set_page_private(page, 0); | |
09cbfeaf | 1001 | put_page(page); |
9ad6b7bc | 1002 | } |
d98237b3 CM |
1003 | } |
1004 | ||
0b32f4bb JB |
1005 | static int btree_set_page_dirty(struct page *page) |
1006 | { | |
bb146eb2 | 1007 | #ifdef DEBUG |
0b32f4bb JB |
1008 | struct extent_buffer *eb; |
1009 | ||
1010 | BUG_ON(!PagePrivate(page)); | |
1011 | eb = (struct extent_buffer *)page->private; | |
1012 | BUG_ON(!eb); | |
1013 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1014 | BUG_ON(!atomic_read(&eb->refs)); | |
1015 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1016 | #endif |
0b32f4bb JB |
1017 | return __set_page_dirty_nobuffers(page); |
1018 | } | |
1019 | ||
7f09410b | 1020 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1021 | .readpage = btree_readpage, |
0da5468f | 1022 | .writepages = btree_writepages, |
5f39d397 CM |
1023 | .releasepage = btree_releasepage, |
1024 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1025 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1026 | .migratepage = btree_migratepage, |
5a92bc88 | 1027 | #endif |
0b32f4bb | 1028 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1029 | }; |
1030 | ||
2ff7e61e | 1031 | void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) |
090d1875 | 1032 | { |
5f39d397 | 1033 | struct extent_buffer *buf = NULL; |
537f38f0 | 1034 | int ret; |
090d1875 | 1035 | |
2ff7e61e | 1036 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1037 | if (IS_ERR(buf)) |
6197d86e | 1038 | return; |
537f38f0 | 1039 | |
c2ccfbc6 | 1040 | ret = read_extent_buffer_pages(buf, WAIT_NONE, 0); |
537f38f0 NB |
1041 | if (ret < 0) |
1042 | free_extent_buffer_stale(buf); | |
1043 | else | |
1044 | free_extent_buffer(buf); | |
090d1875 CM |
1045 | } |
1046 | ||
2ff7e61e JM |
1047 | struct extent_buffer *btrfs_find_create_tree_block( |
1048 | struct btrfs_fs_info *fs_info, | |
1049 | u64 bytenr) | |
0999df54 | 1050 | { |
0b246afa JM |
1051 | if (btrfs_is_testing(fs_info)) |
1052 | return alloc_test_extent_buffer(fs_info, bytenr); | |
1053 | return alloc_extent_buffer(fs_info, bytenr); | |
0999df54 CM |
1054 | } |
1055 | ||
581c1760 QW |
1056 | /* |
1057 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1058 | * verification. | |
1059 | * | |
1060 | * @parent_transid: expected transid of this tree block, skip check if 0 | |
1061 | * @level: expected level, mandatory check | |
1062 | * @first_key: expected key in slot 0, skip check if NULL | |
1063 | */ | |
2ff7e61e | 1064 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
581c1760 QW |
1065 | u64 parent_transid, int level, |
1066 | struct btrfs_key *first_key) | |
0999df54 CM |
1067 | { |
1068 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1069 | int ret; |
1070 | ||
2ff7e61e | 1071 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 LB |
1072 | if (IS_ERR(buf)) |
1073 | return buf; | |
0999df54 | 1074 | |
5ab12d1f | 1075 | ret = btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 1076 | level, first_key); |
0f0fe8f7 | 1077 | if (ret) { |
537f38f0 | 1078 | free_extent_buffer_stale(buf); |
64c043de | 1079 | return ERR_PTR(ret); |
0f0fe8f7 | 1080 | } |
5f39d397 | 1081 | return buf; |
ce9adaa5 | 1082 | |
eb60ceac CM |
1083 | } |
1084 | ||
6a884d7d | 1085 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1086 | { |
6a884d7d | 1087 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1088 | if (btrfs_header_generation(buf) == |
e2d84521 | 1089 | fs_info->running_transaction->transid) { |
b9447ef8 | 1090 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1091 | |
b9473439 | 1092 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1093 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1094 | -buf->len, | |
1095 | fs_info->dirty_metadata_batch); | |
ed7b63eb | 1096 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
8bead258 | 1097 | btrfs_set_lock_blocking_write(buf); |
ed7b63eb JB |
1098 | clear_extent_buffer_dirty(buf); |
1099 | } | |
925baedd | 1100 | } |
5f39d397 CM |
1101 | } |
1102 | ||
8257b2dc MX |
1103 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1104 | { | |
1105 | struct btrfs_subvolume_writers *writers; | |
1106 | int ret; | |
1107 | ||
1108 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1109 | if (!writers) | |
1110 | return ERR_PTR(-ENOMEM); | |
1111 | ||
8a5a916d | 1112 | ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS); |
8257b2dc MX |
1113 | if (ret < 0) { |
1114 | kfree(writers); | |
1115 | return ERR_PTR(ret); | |
1116 | } | |
1117 | ||
1118 | init_waitqueue_head(&writers->wait); | |
1119 | return writers; | |
1120 | } | |
1121 | ||
1122 | static void | |
1123 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1124 | { | |
1125 | percpu_counter_destroy(&writers->counter); | |
1126 | kfree(writers); | |
1127 | } | |
1128 | ||
da17066c | 1129 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1130 | u64 objectid) |
d97e63b6 | 1131 | { |
7c0260ee | 1132 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
96dfcb46 | 1133 | root->fs_info = fs_info; |
cfaa7295 | 1134 | root->node = NULL; |
a28ec197 | 1135 | root->commit_root = NULL; |
27cdeb70 | 1136 | root->state = 0; |
d68fc57b | 1137 | root->orphan_cleanup_state = 0; |
0b86a832 | 1138 | |
0f7d52f4 | 1139 | root->last_trans = 0; |
13a8a7c8 | 1140 | root->highest_objectid = 0; |
eb73c1b7 | 1141 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1142 | root->nr_ordered_extents = 0; |
6bef4d31 | 1143 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1144 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1145 | root->block_rsv = NULL; |
0b86a832 CM |
1146 | |
1147 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1148 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1149 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1150 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1151 | INIT_LIST_HEAD(&root->ordered_extents); |
1152 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1153 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1154 | INIT_LIST_HEAD(&root->logged_list[0]); |
1155 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1156 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1157 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1158 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1159 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1160 | spin_lock_init(&root->log_extents_lock[0]); |
1161 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1162 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1163 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1164 | mutex_init(&root->log_mutex); |
31f3d255 | 1165 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1166 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1167 | init_waitqueue_head(&root->log_writer_wait); |
1168 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1169 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1170 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1171 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1172 | atomic_set(&root->log_commit[0], 0); |
1173 | atomic_set(&root->log_commit[1], 0); | |
1174 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1175 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1176 | refcount_set(&root->refs, 1); |
ea14b57f | 1177 | atomic_set(&root->will_be_snapshotted, 0); |
8ecebf4d | 1178 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1179 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1180 | root->log_transid = 0; |
d1433deb | 1181 | root->log_transid_committed = -1; |
257c62e1 | 1182 | root->last_log_commit = 0; |
7c0260ee | 1183 | if (!dummy) |
43eb5f29 QW |
1184 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1185 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
017e5369 | 1186 | |
3768f368 CM |
1187 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1188 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1189 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1190 | if (!dummy) |
06ea65a3 JB |
1191 | root->defrag_trans_start = fs_info->generation; |
1192 | else | |
1193 | root->defrag_trans_start = 0; | |
4d775673 | 1194 | root->root_key.objectid = objectid; |
0ee5dc67 | 1195 | root->anon_dev = 0; |
8ea05e3a | 1196 | |
5f3ab90a | 1197 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1198 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
3768f368 CM |
1199 | } |
1200 | ||
74e4d827 | 1201 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1202 | u64 objectid, gfp_t flags) |
6f07e42e | 1203 | { |
74e4d827 | 1204 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1205 | if (root) |
96dfcb46 | 1206 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1207 | return root; |
1208 | } | |
1209 | ||
06ea65a3 JB |
1210 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1211 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1212 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1213 | { |
1214 | struct btrfs_root *root; | |
1215 | ||
7c0260ee JM |
1216 | if (!fs_info) |
1217 | return ERR_PTR(-EINVAL); | |
1218 | ||
96dfcb46 | 1219 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1220 | if (!root) |
1221 | return ERR_PTR(-ENOMEM); | |
da17066c | 1222 | |
b9ef22de | 1223 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1224 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1225 | |
1226 | return root; | |
1227 | } | |
1228 | #endif | |
1229 | ||
20897f5c | 1230 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1231 | u64 objectid) |
1232 | { | |
9b7a2440 | 1233 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1234 | struct extent_buffer *leaf; |
1235 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1236 | struct btrfs_root *root; | |
1237 | struct btrfs_key key; | |
b89f6d1f | 1238 | unsigned int nofs_flag; |
20897f5c | 1239 | int ret = 0; |
33d85fda | 1240 | uuid_le uuid = NULL_UUID_LE; |
20897f5c | 1241 | |
b89f6d1f FM |
1242 | /* |
1243 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1244 | * context to avoid deadlock if reclaim happens. | |
1245 | */ | |
1246 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1247 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1248 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1249 | if (!root) |
1250 | return ERR_PTR(-ENOMEM); | |
1251 | ||
20897f5c AJ |
1252 | root->root_key.objectid = objectid; |
1253 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1254 | root->root_key.offset = 0; | |
1255 | ||
4d75f8a9 | 1256 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1257 | if (IS_ERR(leaf)) { |
1258 | ret = PTR_ERR(leaf); | |
1dd05682 | 1259 | leaf = NULL; |
20897f5c AJ |
1260 | goto fail; |
1261 | } | |
1262 | ||
20897f5c | 1263 | root->node = leaf; |
20897f5c AJ |
1264 | btrfs_mark_buffer_dirty(leaf); |
1265 | ||
1266 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1267 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1268 | |
1269 | root->root_item.flags = 0; | |
1270 | root->root_item.byte_limit = 0; | |
1271 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1272 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1273 | btrfs_set_root_level(&root->root_item, 0); | |
1274 | btrfs_set_root_refs(&root->root_item, 1); | |
1275 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1276 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1277 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda QW |
1278 | if (is_fstree(objectid)) |
1279 | uuid_le_gen(&uuid); | |
6463fe58 | 1280 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); |
20897f5c AJ |
1281 | root->root_item.drop_level = 0; |
1282 | ||
1283 | key.objectid = objectid; | |
1284 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1285 | key.offset = 0; | |
1286 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1287 | if (ret) | |
1288 | goto fail; | |
1289 | ||
1290 | btrfs_tree_unlock(leaf); | |
1291 | ||
1dd05682 TI |
1292 | return root; |
1293 | ||
20897f5c | 1294 | fail: |
1dd05682 TI |
1295 | if (leaf) { |
1296 | btrfs_tree_unlock(leaf); | |
59885b39 | 1297 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1298 | free_extent_buffer(leaf); |
1299 | } | |
81f096ed | 1300 | btrfs_put_fs_root(root); |
20897f5c | 1301 | |
1dd05682 | 1302 | return ERR_PTR(ret); |
20897f5c AJ |
1303 | } |
1304 | ||
7237f183 YZ |
1305 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1306 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1307 | { |
1308 | struct btrfs_root *root; | |
7237f183 | 1309 | struct extent_buffer *leaf; |
e02119d5 | 1310 | |
96dfcb46 | 1311 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1312 | if (!root) |
7237f183 | 1313 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1314 | |
e02119d5 CM |
1315 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1316 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1317 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1318 | |
7237f183 | 1319 | /* |
27cdeb70 MX |
1320 | * DON'T set REF_COWS for log trees |
1321 | * | |
7237f183 YZ |
1322 | * log trees do not get reference counted because they go away |
1323 | * before a real commit is actually done. They do store pointers | |
1324 | * to file data extents, and those reference counts still get | |
1325 | * updated (along with back refs to the log tree). | |
1326 | */ | |
e02119d5 | 1327 | |
4d75f8a9 DS |
1328 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1329 | NULL, 0, 0, 0); | |
7237f183 | 1330 | if (IS_ERR(leaf)) { |
81f096ed | 1331 | btrfs_put_fs_root(root); |
7237f183 YZ |
1332 | return ERR_CAST(leaf); |
1333 | } | |
e02119d5 | 1334 | |
7237f183 | 1335 | root->node = leaf; |
e02119d5 | 1336 | |
e02119d5 CM |
1337 | btrfs_mark_buffer_dirty(root->node); |
1338 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1339 | return root; |
1340 | } | |
1341 | ||
1342 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1343 | struct btrfs_fs_info *fs_info) | |
1344 | { | |
1345 | struct btrfs_root *log_root; | |
1346 | ||
1347 | log_root = alloc_log_tree(trans, fs_info); | |
1348 | if (IS_ERR(log_root)) | |
1349 | return PTR_ERR(log_root); | |
1350 | WARN_ON(fs_info->log_root_tree); | |
1351 | fs_info->log_root_tree = log_root; | |
1352 | return 0; | |
1353 | } | |
1354 | ||
1355 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1356 | struct btrfs_root *root) | |
1357 | { | |
0b246afa | 1358 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1359 | struct btrfs_root *log_root; |
1360 | struct btrfs_inode_item *inode_item; | |
1361 | ||
0b246afa | 1362 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1363 | if (IS_ERR(log_root)) |
1364 | return PTR_ERR(log_root); | |
1365 | ||
1366 | log_root->last_trans = trans->transid; | |
1367 | log_root->root_key.offset = root->root_key.objectid; | |
1368 | ||
1369 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1370 | btrfs_set_stack_inode_generation(inode_item, 1); |
1371 | btrfs_set_stack_inode_size(inode_item, 3); | |
1372 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1373 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1374 | fs_info->nodesize); |
3cae210f | 1375 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1376 | |
5d4f98a2 | 1377 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1378 | |
1379 | WARN_ON(root->log_root); | |
1380 | root->log_root = log_root; | |
1381 | root->log_transid = 0; | |
d1433deb | 1382 | root->log_transid_committed = -1; |
257c62e1 | 1383 | root->last_log_commit = 0; |
e02119d5 CM |
1384 | return 0; |
1385 | } | |
1386 | ||
62a2c73e JB |
1387 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1388 | struct btrfs_key *key) | |
e02119d5 CM |
1389 | { |
1390 | struct btrfs_root *root; | |
1391 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1392 | struct btrfs_path *path; |
84234f3a | 1393 | u64 generation; |
cb517eab | 1394 | int ret; |
581c1760 | 1395 | int level; |
0f7d52f4 | 1396 | |
cb517eab MX |
1397 | path = btrfs_alloc_path(); |
1398 | if (!path) | |
0f7d52f4 | 1399 | return ERR_PTR(-ENOMEM); |
cb517eab | 1400 | |
96dfcb46 | 1401 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
cb517eab MX |
1402 | if (!root) { |
1403 | ret = -ENOMEM; | |
1404 | goto alloc_fail; | |
0f7d52f4 CM |
1405 | } |
1406 | ||
cb517eab MX |
1407 | ret = btrfs_find_root(tree_root, key, path, |
1408 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1409 | if (ret) { |
13a8a7c8 YZ |
1410 | if (ret > 0) |
1411 | ret = -ENOENT; | |
cb517eab | 1412 | goto find_fail; |
0f7d52f4 | 1413 | } |
13a8a7c8 | 1414 | |
84234f3a | 1415 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1416 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1417 | root->node = read_tree_block(fs_info, |
1418 | btrfs_root_bytenr(&root->root_item), | |
581c1760 | 1419 | generation, level, NULL); |
64c043de LB |
1420 | if (IS_ERR(root->node)) { |
1421 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1422 | goto find_fail; |
1423 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1424 | ret = -EIO; | |
64c043de LB |
1425 | free_extent_buffer(root->node); |
1426 | goto find_fail; | |
416bc658 | 1427 | } |
5d4f98a2 | 1428 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1429 | out: |
cb517eab MX |
1430 | btrfs_free_path(path); |
1431 | return root; | |
1432 | ||
cb517eab | 1433 | find_fail: |
81f096ed | 1434 | btrfs_put_fs_root(root); |
cb517eab MX |
1435 | alloc_fail: |
1436 | root = ERR_PTR(ret); | |
1437 | goto out; | |
1438 | } | |
1439 | ||
a98db0f3 | 1440 | static int btrfs_init_fs_root(struct btrfs_root *root) |
cb517eab MX |
1441 | { |
1442 | int ret; | |
8257b2dc | 1443 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1444 | |
1445 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1446 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1447 | GFP_NOFS); | |
1448 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1449 | ret = -ENOMEM; | |
1450 | goto fail; | |
1451 | } | |
1452 | ||
8257b2dc MX |
1453 | writers = btrfs_alloc_subvolume_writers(); |
1454 | if (IS_ERR(writers)) { | |
1455 | ret = PTR_ERR(writers); | |
1456 | goto fail; | |
1457 | } | |
1458 | root->subv_writers = writers; | |
1459 | ||
f39e4571 JB |
1460 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { |
1461 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); | |
1462 | btrfs_check_and_init_root_item(&root->root_item); | |
1463 | } | |
1464 | ||
cb517eab | 1465 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1466 | spin_lock_init(&root->ino_cache_lock); |
1467 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1468 | |
1469 | ret = get_anon_bdev(&root->anon_dev); | |
1470 | if (ret) | |
876d2cf1 | 1471 | goto fail; |
f32e48e9 CR |
1472 | |
1473 | mutex_lock(&root->objectid_mutex); | |
1474 | ret = btrfs_find_highest_objectid(root, | |
1475 | &root->highest_objectid); | |
1476 | if (ret) { | |
1477 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1478 | goto fail; |
f32e48e9 CR |
1479 | } |
1480 | ||
1481 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1482 | ||
1483 | mutex_unlock(&root->objectid_mutex); | |
1484 | ||
cb517eab MX |
1485 | return 0; |
1486 | fail: | |
84db5ccf | 1487 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1488 | return ret; |
1489 | } | |
1490 | ||
a98db0f3 JB |
1491 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1492 | u64 root_id) | |
cb517eab MX |
1493 | { |
1494 | struct btrfs_root *root; | |
1495 | ||
1496 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1497 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1498 | (unsigned long)root_id); | |
1499 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1500 | return root; | |
1501 | } | |
1502 | ||
1503 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1504 | struct btrfs_root *root) | |
1505 | { | |
1506 | int ret; | |
1507 | ||
e1860a77 | 1508 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1509 | if (ret) |
1510 | return ret; | |
1511 | ||
1512 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1513 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1514 | (unsigned long)root->root_key.objectid, | |
1515 | root); | |
af01d2e5 JB |
1516 | if (ret == 0) { |
1517 | btrfs_grab_fs_root(root); | |
27cdeb70 | 1518 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1519 | } |
cb517eab MX |
1520 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1521 | radix_tree_preload_end(); | |
1522 | ||
1523 | return ret; | |
1524 | } | |
1525 | ||
0d4b0463 JB |
1526 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1527 | { | |
1528 | kfree(fs_info->balance_ctl); | |
1529 | kfree(fs_info->delayed_root); | |
81f096ed JB |
1530 | btrfs_put_fs_root(fs_info->extent_root); |
1531 | btrfs_put_fs_root(fs_info->tree_root); | |
1532 | btrfs_put_fs_root(fs_info->chunk_root); | |
1533 | btrfs_put_fs_root(fs_info->dev_root); | |
1534 | btrfs_put_fs_root(fs_info->csum_root); | |
1535 | btrfs_put_fs_root(fs_info->quota_root); | |
1536 | btrfs_put_fs_root(fs_info->uuid_root); | |
1537 | btrfs_put_fs_root(fs_info->free_space_root); | |
1538 | btrfs_put_fs_root(fs_info->fs_root); | |
0d4b0463 JB |
1539 | kfree(fs_info->super_copy); |
1540 | kfree(fs_info->super_for_commit); | |
1541 | kvfree(fs_info); | |
1542 | } | |
1543 | ||
1544 | ||
c00869f1 MX |
1545 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1546 | struct btrfs_key *location, | |
1547 | bool check_ref) | |
5eda7b5e CM |
1548 | { |
1549 | struct btrfs_root *root; | |
381cf658 | 1550 | struct btrfs_path *path; |
1d4c08e0 | 1551 | struct btrfs_key key; |
5eda7b5e CM |
1552 | int ret; |
1553 | ||
edbd8d4e CM |
1554 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1555 | return fs_info->tree_root; | |
1556 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1557 | return fs_info->extent_root; | |
8f18cf13 CM |
1558 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1559 | return fs_info->chunk_root; | |
1560 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1561 | return fs_info->dev_root; | |
0403e47e YZ |
1562 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1563 | return fs_info->csum_root; | |
bcef60f2 AJ |
1564 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1565 | return fs_info->quota_root ? fs_info->quota_root : | |
1566 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1567 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1568 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1569 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1570 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1571 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1572 | ERR_PTR(-ENOENT); | |
4df27c4d | 1573 | again: |
cb517eab | 1574 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1575 | if (root) { |
c00869f1 | 1576 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1577 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1578 | return root; |
48475471 | 1579 | } |
5eda7b5e | 1580 | |
83db2aad | 1581 | root = btrfs_read_tree_root(fs_info->tree_root, location); |
5eda7b5e CM |
1582 | if (IS_ERR(root)) |
1583 | return root; | |
3394e160 | 1584 | |
c00869f1 | 1585 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1586 | ret = -ENOENT; |
581bb050 | 1587 | goto fail; |
35a30d7c | 1588 | } |
581bb050 | 1589 | |
cb517eab | 1590 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1591 | if (ret) |
1592 | goto fail; | |
3394e160 | 1593 | |
381cf658 DS |
1594 | path = btrfs_alloc_path(); |
1595 | if (!path) { | |
1596 | ret = -ENOMEM; | |
1597 | goto fail; | |
1598 | } | |
1d4c08e0 DS |
1599 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1600 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1601 | key.offset = location->objectid; | |
1602 | ||
1603 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1604 | btrfs_free_path(path); |
d68fc57b YZ |
1605 | if (ret < 0) |
1606 | goto fail; | |
1607 | if (ret == 0) | |
27cdeb70 | 1608 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1609 | |
cb517eab | 1610 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1611 | if (ret) { |
4df27c4d | 1612 | if (ret == -EEXIST) { |
84db5ccf | 1613 | btrfs_free_fs_root(root); |
4df27c4d YZ |
1614 | goto again; |
1615 | } | |
1616 | goto fail; | |
0f7d52f4 | 1617 | } |
edbd8d4e | 1618 | return root; |
4df27c4d | 1619 | fail: |
84db5ccf | 1620 | btrfs_free_fs_root(root); |
4df27c4d | 1621 | return ERR_PTR(ret); |
edbd8d4e CM |
1622 | } |
1623 | ||
04160088 CM |
1624 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1625 | { | |
1626 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1627 | int ret = 0; | |
04160088 CM |
1628 | struct btrfs_device *device; |
1629 | struct backing_dev_info *bdi; | |
b7967db7 | 1630 | |
1f78160c XG |
1631 | rcu_read_lock(); |
1632 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1633 | if (!device->bdev) |
1634 | continue; | |
efa7c9f9 | 1635 | bdi = device->bdev->bd_bdi; |
ff9ea323 | 1636 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1637 | ret = 1; |
1638 | break; | |
1639 | } | |
1640 | } | |
1f78160c | 1641 | rcu_read_unlock(); |
04160088 CM |
1642 | return ret; |
1643 | } | |
1644 | ||
8b712842 CM |
1645 | /* |
1646 | * called by the kthread helper functions to finally call the bio end_io | |
1647 | * functions. This is where read checksum verification actually happens | |
1648 | */ | |
1649 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1650 | { |
ce9adaa5 | 1651 | struct bio *bio; |
97eb6b69 | 1652 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1653 | |
97eb6b69 | 1654 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1655 | bio = end_io_wq->bio; |
ce9adaa5 | 1656 | |
4e4cbee9 | 1657 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1658 | bio->bi_private = end_io_wq->private; |
1659 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1660 | bio_endio(bio); |
9be490f1 | 1661 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1662 | } |
1663 | ||
a74a4b97 CM |
1664 | static int cleaner_kthread(void *arg) |
1665 | { | |
1666 | struct btrfs_root *root = arg; | |
0b246afa | 1667 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1668 | int again; |
a74a4b97 | 1669 | |
d6fd0ae2 | 1670 | while (1) { |
d0278245 | 1671 | again = 0; |
a74a4b97 | 1672 | |
fd340d0f JB |
1673 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1674 | ||
d0278245 | 1675 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1676 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1677 | goto sleep; |
1678 | ||
90c711ab ZB |
1679 | /* |
1680 | * Do not do anything if we might cause open_ctree() to block | |
1681 | * before we have finished mounting the filesystem. | |
1682 | */ | |
0b246afa | 1683 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1684 | goto sleep; |
1685 | ||
0b246afa | 1686 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1687 | goto sleep; |
1688 | ||
dc7f370c MX |
1689 | /* |
1690 | * Avoid the problem that we change the status of the fs | |
1691 | * during the above check and trylock. | |
1692 | */ | |
2ff7e61e | 1693 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1694 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1695 | goto sleep; |
76dda93c | 1696 | } |
a74a4b97 | 1697 | |
2ff7e61e | 1698 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1699 | |
d0278245 | 1700 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1701 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1702 | |
1703 | /* | |
05323cd1 MX |
1704 | * The defragger has dealt with the R/O remount and umount, |
1705 | * needn't do anything special here. | |
d0278245 | 1706 | */ |
0b246afa | 1707 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1708 | |
1709 | /* | |
1710 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1711 | * with relocation (btrfs_relocate_chunk) and relocation | |
1712 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1713 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1714 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1715 | * unused block groups. | |
1716 | */ | |
0b246afa | 1717 | btrfs_delete_unused_bgs(fs_info); |
d0278245 | 1718 | sleep: |
fd340d0f | 1719 | clear_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1720 | if (kthread_should_park()) |
1721 | kthread_parkme(); | |
1722 | if (kthread_should_stop()) | |
1723 | return 0; | |
838fe188 | 1724 | if (!again) { |
a74a4b97 | 1725 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1726 | schedule(); |
a74a4b97 CM |
1727 | __set_current_state(TASK_RUNNING); |
1728 | } | |
da288d28 | 1729 | } |
a74a4b97 CM |
1730 | } |
1731 | ||
1732 | static int transaction_kthread(void *arg) | |
1733 | { | |
1734 | struct btrfs_root *root = arg; | |
0b246afa | 1735 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1736 | struct btrfs_trans_handle *trans; |
1737 | struct btrfs_transaction *cur; | |
8929ecfa | 1738 | u64 transid; |
a944442c | 1739 | time64_t now; |
a74a4b97 | 1740 | unsigned long delay; |
914b2007 | 1741 | bool cannot_commit; |
a74a4b97 CM |
1742 | |
1743 | do { | |
914b2007 | 1744 | cannot_commit = false; |
0b246afa JM |
1745 | delay = HZ * fs_info->commit_interval; |
1746 | mutex_lock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1747 | |
0b246afa JM |
1748 | spin_lock(&fs_info->trans_lock); |
1749 | cur = fs_info->running_transaction; | |
a74a4b97 | 1750 | if (!cur) { |
0b246afa | 1751 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1752 | goto sleep; |
1753 | } | |
31153d81 | 1754 | |
afd48513 | 1755 | now = ktime_get_seconds(); |
3296bf56 | 1756 | if (cur->state < TRANS_STATE_COMMIT_START && |
a514d638 | 1757 | !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) && |
8b87dc17 | 1758 | (now < cur->start_time || |
0b246afa JM |
1759 | now - cur->start_time < fs_info->commit_interval)) { |
1760 | spin_unlock(&fs_info->trans_lock); | |
a74a4b97 CM |
1761 | delay = HZ * 5; |
1762 | goto sleep; | |
1763 | } | |
8929ecfa | 1764 | transid = cur->transid; |
0b246afa | 1765 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1766 | |
79787eaa | 1767 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1768 | trans = btrfs_attach_transaction(root); |
914b2007 | 1769 | if (IS_ERR(trans)) { |
354aa0fb MX |
1770 | if (PTR_ERR(trans) != -ENOENT) |
1771 | cannot_commit = true; | |
79787eaa | 1772 | goto sleep; |
914b2007 | 1773 | } |
8929ecfa | 1774 | if (transid == trans->transid) { |
3a45bb20 | 1775 | btrfs_commit_transaction(trans); |
8929ecfa | 1776 | } else { |
3a45bb20 | 1777 | btrfs_end_transaction(trans); |
8929ecfa | 1778 | } |
a74a4b97 | 1779 | sleep: |
0b246afa JM |
1780 | wake_up_process(fs_info->cleaner_kthread); |
1781 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1782 | |
4e121c06 | 1783 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1784 | &fs_info->fs_state))) |
2ff7e61e | 1785 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1786 | if (!kthread_should_stop() && |
0b246afa | 1787 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1788 | cannot_commit)) |
bc5511d0 | 1789 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1790 | } while (!kthread_should_stop()); |
1791 | return 0; | |
1792 | } | |
1793 | ||
af31f5e5 | 1794 | /* |
01f0f9da NB |
1795 | * This will find the highest generation in the array of root backups. The |
1796 | * index of the highest array is returned, or -EINVAL if we can't find | |
1797 | * anything. | |
af31f5e5 CM |
1798 | * |
1799 | * We check to make sure the array is valid by comparing the | |
1800 | * generation of the latest root in the array with the generation | |
1801 | * in the super block. If they don't match we pitch it. | |
1802 | */ | |
01f0f9da | 1803 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1804 | { |
01f0f9da | 1805 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1806 | u64 cur; |
af31f5e5 CM |
1807 | struct btrfs_root_backup *root_backup; |
1808 | int i; | |
1809 | ||
1810 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1811 | root_backup = info->super_copy->super_roots + i; | |
1812 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1813 | if (cur == newest_gen) | |
01f0f9da | 1814 | return i; |
af31f5e5 CM |
1815 | } |
1816 | ||
01f0f9da | 1817 | return -EINVAL; |
af31f5e5 CM |
1818 | } |
1819 | ||
af31f5e5 CM |
1820 | /* |
1821 | * copy all the root pointers into the super backup array. | |
1822 | * this will bump the backup pointer by one when it is | |
1823 | * done | |
1824 | */ | |
1825 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1826 | { | |
6ef108dd | 1827 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1828 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1829 | |
1830 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1831 | ||
1832 | /* | |
1833 | * make sure all of our padding and empty slots get zero filled | |
1834 | * regardless of which ones we use today | |
1835 | */ | |
1836 | memset(root_backup, 0, sizeof(*root_backup)); | |
1837 | ||
1838 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1839 | ||
1840 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1841 | btrfs_set_backup_tree_root_gen(root_backup, | |
1842 | btrfs_header_generation(info->tree_root->node)); | |
1843 | ||
1844 | btrfs_set_backup_tree_root_level(root_backup, | |
1845 | btrfs_header_level(info->tree_root->node)); | |
1846 | ||
1847 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1848 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1849 | btrfs_header_generation(info->chunk_root->node)); | |
1850 | btrfs_set_backup_chunk_root_level(root_backup, | |
1851 | btrfs_header_level(info->chunk_root->node)); | |
1852 | ||
1853 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1854 | btrfs_set_backup_extent_root_gen(root_backup, | |
1855 | btrfs_header_generation(info->extent_root->node)); | |
1856 | btrfs_set_backup_extent_root_level(root_backup, | |
1857 | btrfs_header_level(info->extent_root->node)); | |
1858 | ||
7c7e82a7 CM |
1859 | /* |
1860 | * we might commit during log recovery, which happens before we set | |
1861 | * the fs_root. Make sure it is valid before we fill it in. | |
1862 | */ | |
1863 | if (info->fs_root && info->fs_root->node) { | |
1864 | btrfs_set_backup_fs_root(root_backup, | |
1865 | info->fs_root->node->start); | |
1866 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1867 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1868 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1869 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1870 | } |
af31f5e5 CM |
1871 | |
1872 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1873 | btrfs_set_backup_dev_root_gen(root_backup, | |
1874 | btrfs_header_generation(info->dev_root->node)); | |
1875 | btrfs_set_backup_dev_root_level(root_backup, | |
1876 | btrfs_header_level(info->dev_root->node)); | |
1877 | ||
1878 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1879 | btrfs_set_backup_csum_root_gen(root_backup, | |
1880 | btrfs_header_generation(info->csum_root->node)); | |
1881 | btrfs_set_backup_csum_root_level(root_backup, | |
1882 | btrfs_header_level(info->csum_root->node)); | |
1883 | ||
1884 | btrfs_set_backup_total_bytes(root_backup, | |
1885 | btrfs_super_total_bytes(info->super_copy)); | |
1886 | btrfs_set_backup_bytes_used(root_backup, | |
1887 | btrfs_super_bytes_used(info->super_copy)); | |
1888 | btrfs_set_backup_num_devices(root_backup, | |
1889 | btrfs_super_num_devices(info->super_copy)); | |
1890 | ||
1891 | /* | |
1892 | * if we don't copy this out to the super_copy, it won't get remembered | |
1893 | * for the next commit | |
1894 | */ | |
1895 | memcpy(&info->super_copy->super_roots, | |
1896 | &info->super_for_commit->super_roots, | |
1897 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1898 | } | |
1899 | ||
bd2336b2 NB |
1900 | /* |
1901 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
1902 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
1903 | * | |
1904 | * fs_info - filesystem whose backup roots need to be read | |
1905 | * priority - priority of backup root required | |
1906 | * | |
1907 | * Returns backup root index on success and -EINVAL otherwise. | |
1908 | */ | |
1909 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1910 | { | |
1911 | int backup_index = find_newest_super_backup(fs_info); | |
1912 | struct btrfs_super_block *super = fs_info->super_copy; | |
1913 | struct btrfs_root_backup *root_backup; | |
1914 | ||
1915 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1916 | if (priority == 0) | |
1917 | return backup_index; | |
1918 | ||
1919 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1920 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1921 | } else { | |
1922 | return -EINVAL; | |
1923 | } | |
1924 | ||
1925 | root_backup = super->super_roots + backup_index; | |
1926 | ||
1927 | btrfs_set_super_generation(super, | |
1928 | btrfs_backup_tree_root_gen(root_backup)); | |
1929 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1930 | btrfs_set_super_root_level(super, | |
1931 | btrfs_backup_tree_root_level(root_backup)); | |
1932 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1933 | ||
1934 | /* | |
1935 | * Fixme: the total bytes and num_devices need to match or we should | |
1936 | * need a fsck | |
1937 | */ | |
1938 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1939 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1940 | ||
1941 | return backup_index; | |
1942 | } | |
1943 | ||
7abadb64 LB |
1944 | /* helper to cleanup workers */ |
1945 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1946 | { | |
dc6e3209 | 1947 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 1948 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 1949 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 1950 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 1951 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
8b110e39 | 1952 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 1953 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
1954 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
1955 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 1956 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 1957 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 1958 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 1959 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 1960 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
1961 | if (fs_info->discard_ctl.discard_workers) |
1962 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
1963 | /* |
1964 | * Now that all other work queues are destroyed, we can safely destroy | |
1965 | * the queues used for metadata I/O, since tasks from those other work | |
1966 | * queues can do metadata I/O operations. | |
1967 | */ | |
1968 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
1969 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
1970 | } |
1971 | ||
2e9f5954 R |
1972 | static void free_root_extent_buffers(struct btrfs_root *root) |
1973 | { | |
1974 | if (root) { | |
1975 | free_extent_buffer(root->node); | |
1976 | free_extent_buffer(root->commit_root); | |
1977 | root->node = NULL; | |
1978 | root->commit_root = NULL; | |
1979 | } | |
1980 | } | |
1981 | ||
af31f5e5 | 1982 | /* helper to cleanup tree roots */ |
4273eaff | 1983 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 1984 | { |
2e9f5954 | 1985 | free_root_extent_buffers(info->tree_root); |
655b09fe | 1986 | |
2e9f5954 R |
1987 | free_root_extent_buffers(info->dev_root); |
1988 | free_root_extent_buffers(info->extent_root); | |
1989 | free_root_extent_buffers(info->csum_root); | |
1990 | free_root_extent_buffers(info->quota_root); | |
1991 | free_root_extent_buffers(info->uuid_root); | |
4273eaff | 1992 | if (free_chunk_root) |
2e9f5954 | 1993 | free_root_extent_buffers(info->chunk_root); |
70f6d82e | 1994 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
1995 | } |
1996 | ||
faa2dbf0 | 1997 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
1998 | { |
1999 | int ret; | |
2000 | struct btrfs_root *gang[8]; | |
2001 | int i; | |
2002 | ||
2003 | while (!list_empty(&fs_info->dead_roots)) { | |
2004 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2005 | struct btrfs_root, root_list); | |
2006 | list_del(&gang[0]->root_list); | |
2007 | ||
27cdeb70 | 2008 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2009 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2010 | } else { |
2011 | free_extent_buffer(gang[0]->node); | |
2012 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2013 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2014 | } |
2015 | } | |
2016 | ||
2017 | while (1) { | |
2018 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2019 | (void **)gang, 0, | |
2020 | ARRAY_SIZE(gang)); | |
2021 | if (!ret) | |
2022 | break; | |
2023 | for (i = 0; i < ret; i++) | |
cb517eab | 2024 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2025 | } |
1a4319cc LB |
2026 | |
2027 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2028 | btrfs_free_log_root_tree(NULL, fs_info); | |
2ff7e61e | 2029 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
1a4319cc | 2030 | } |
171f6537 | 2031 | } |
af31f5e5 | 2032 | |
638aa7ed ES |
2033 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2034 | { | |
2035 | mutex_init(&fs_info->scrub_lock); | |
2036 | atomic_set(&fs_info->scrubs_running, 0); | |
2037 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2038 | atomic_set(&fs_info->scrubs_paused, 0); | |
2039 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2040 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2041 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2042 | } |
2043 | ||
779a65a4 ES |
2044 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2045 | { | |
2046 | spin_lock_init(&fs_info->balance_lock); | |
2047 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2048 | atomic_set(&fs_info->balance_pause_req, 0); |
2049 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2050 | fs_info->balance_ctl = NULL; | |
2051 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2052 | } | |
2053 | ||
6bccf3ab | 2054 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2055 | { |
2ff7e61e JM |
2056 | struct inode *inode = fs_info->btree_inode; |
2057 | ||
2058 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2059 | set_nlink(inode, 1); | |
f37938e0 ES |
2060 | /* |
2061 | * we set the i_size on the btree inode to the max possible int. | |
2062 | * the real end of the address space is determined by all of | |
2063 | * the devices in the system | |
2064 | */ | |
2ff7e61e JM |
2065 | inode->i_size = OFFSET_MAX; |
2066 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2067 | |
2ff7e61e | 2068 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 QW |
2069 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2070 | IO_TREE_INODE_IO, inode); | |
7b439738 | 2071 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2072 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2073 | |
2ff7e61e | 2074 | BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; |
f37938e0 | 2075 | |
2ff7e61e JM |
2076 | BTRFS_I(inode)->root = fs_info->tree_root; |
2077 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); | |
2078 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2079 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2080 | } |
2081 | ||
ad618368 ES |
2082 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2083 | { | |
ad618368 | 2084 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2085 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2086 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2087 | } |
2088 | ||
f9e92e40 ES |
2089 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2090 | { | |
2091 | spin_lock_init(&fs_info->qgroup_lock); | |
2092 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2093 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2094 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2095 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2096 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2097 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2098 | mutex_init(&fs_info->qgroup_rescan_lock); |
2099 | } | |
2100 | ||
2a458198 ES |
2101 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2102 | struct btrfs_fs_devices *fs_devices) | |
2103 | { | |
f7b885be | 2104 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2105 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2106 | |
2107 | fs_info->workers = | |
cb001095 JM |
2108 | btrfs_alloc_workqueue(fs_info, "worker", |
2109 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2110 | |
2111 | fs_info->delalloc_workers = | |
cb001095 JM |
2112 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2113 | flags, max_active, 2); | |
2a458198 ES |
2114 | |
2115 | fs_info->flush_workers = | |
cb001095 JM |
2116 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2117 | flags, max_active, 0); | |
2a458198 ES |
2118 | |
2119 | fs_info->caching_workers = | |
cb001095 | 2120 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2121 | |
2a458198 | 2122 | fs_info->fixup_workers = |
cb001095 | 2123 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2124 | |
2125 | /* | |
2126 | * endios are largely parallel and should have a very | |
2127 | * low idle thresh | |
2128 | */ | |
2129 | fs_info->endio_workers = | |
cb001095 | 2130 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2131 | fs_info->endio_meta_workers = |
cb001095 JM |
2132 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2133 | max_active, 4); | |
2a458198 | 2134 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2135 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2136 | max_active, 2); | |
2a458198 | 2137 | fs_info->endio_raid56_workers = |
cb001095 JM |
2138 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2139 | max_active, 4); | |
2a458198 | 2140 | fs_info->endio_repair_workers = |
cb001095 | 2141 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2142 | fs_info->rmw_workers = |
cb001095 | 2143 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2144 | fs_info->endio_write_workers = |
cb001095 JM |
2145 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2146 | max_active, 2); | |
2a458198 | 2147 | fs_info->endio_freespace_worker = |
cb001095 JM |
2148 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2149 | max_active, 0); | |
2a458198 | 2150 | fs_info->delayed_workers = |
cb001095 JM |
2151 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2152 | max_active, 0); | |
2a458198 | 2153 | fs_info->readahead_workers = |
cb001095 JM |
2154 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2155 | max_active, 2); | |
2a458198 | 2156 | fs_info->qgroup_rescan_workers = |
cb001095 | 2157 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2158 | fs_info->discard_ctl.discard_workers = |
2159 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 ES |
2160 | |
2161 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
ba8a9d07 | 2162 | fs_info->flush_workers && |
2a458198 ES |
2163 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2164 | fs_info->endio_meta_write_workers && | |
2165 | fs_info->endio_repair_workers && | |
2166 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2167 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2168 | fs_info->caching_workers && fs_info->readahead_workers && | |
2169 | fs_info->fixup_workers && fs_info->delayed_workers && | |
b0643e59 DZ |
2170 | fs_info->qgroup_rescan_workers && |
2171 | fs_info->discard_ctl.discard_workers)) { | |
2a458198 ES |
2172 | return -ENOMEM; |
2173 | } | |
2174 | ||
2175 | return 0; | |
2176 | } | |
2177 | ||
6d97c6e3 JT |
2178 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2179 | { | |
2180 | struct crypto_shash *csum_shash; | |
b4e967be | 2181 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2182 | |
b4e967be | 2183 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2184 | |
2185 | if (IS_ERR(csum_shash)) { | |
2186 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2187 | csum_driver); |
6d97c6e3 JT |
2188 | return PTR_ERR(csum_shash); |
2189 | } | |
2190 | ||
2191 | fs_info->csum_shash = csum_shash; | |
2192 | ||
2193 | return 0; | |
2194 | } | |
2195 | ||
2196 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) | |
2197 | { | |
2198 | crypto_free_shash(fs_info->csum_shash); | |
2199 | } | |
2200 | ||
63443bf5 ES |
2201 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2202 | struct btrfs_fs_devices *fs_devices) | |
2203 | { | |
2204 | int ret; | |
63443bf5 ES |
2205 | struct btrfs_root *log_tree_root; |
2206 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2207 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2208 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2209 | |
2210 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2211 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2212 | return -EIO; |
2213 | } | |
2214 | ||
96dfcb46 JB |
2215 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2216 | GFP_KERNEL); | |
63443bf5 ES |
2217 | if (!log_tree_root) |
2218 | return -ENOMEM; | |
2219 | ||
2ff7e61e | 2220 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
581c1760 QW |
2221 | fs_info->generation + 1, |
2222 | level, NULL); | |
64c043de | 2223 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2224 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2225 | ret = PTR_ERR(log_tree_root->node); |
81f096ed | 2226 | btrfs_put_fs_root(log_tree_root); |
0eeff236 | 2227 | return ret; |
64c043de | 2228 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2229 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 | 2230 | free_extent_buffer(log_tree_root->node); |
81f096ed | 2231 | btrfs_put_fs_root(log_tree_root); |
63443bf5 ES |
2232 | return -EIO; |
2233 | } | |
2234 | /* returns with log_tree_root freed on success */ | |
2235 | ret = btrfs_recover_log_trees(log_tree_root); | |
2236 | if (ret) { | |
0b246afa JM |
2237 | btrfs_handle_fs_error(fs_info, ret, |
2238 | "Failed to recover log tree"); | |
63443bf5 | 2239 | free_extent_buffer(log_tree_root->node); |
81f096ed | 2240 | btrfs_put_fs_root(log_tree_root); |
63443bf5 ES |
2241 | return ret; |
2242 | } | |
2243 | ||
bc98a42c | 2244 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2245 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2246 | if (ret) |
2247 | return ret; | |
2248 | } | |
2249 | ||
2250 | return 0; | |
2251 | } | |
2252 | ||
6bccf3ab | 2253 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2254 | { |
6bccf3ab | 2255 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2256 | struct btrfs_root *root; |
4bbcaa64 ES |
2257 | struct btrfs_key location; |
2258 | int ret; | |
2259 | ||
6bccf3ab JM |
2260 | BUG_ON(!fs_info->tree_root); |
2261 | ||
4bbcaa64 ES |
2262 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2263 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2264 | location.offset = 0; | |
2265 | ||
a4f3d2c4 | 2266 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2267 | if (IS_ERR(root)) { |
2268 | ret = PTR_ERR(root); | |
2269 | goto out; | |
2270 | } | |
a4f3d2c4 DS |
2271 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2272 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2273 | |
2274 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2275 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2276 | if (IS_ERR(root)) { |
2277 | ret = PTR_ERR(root); | |
2278 | goto out; | |
2279 | } | |
a4f3d2c4 DS |
2280 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2281 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2282 | btrfs_init_devices_late(fs_info); |
2283 | ||
2284 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 | 2285 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2286 | if (IS_ERR(root)) { |
2287 | ret = PTR_ERR(root); | |
2288 | goto out; | |
2289 | } | |
a4f3d2c4 DS |
2290 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2291 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2292 | |
2293 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2294 | root = btrfs_read_tree_root(tree_root, &location); |
2295 | if (!IS_ERR(root)) { | |
2296 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2297 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2298 | fs_info->quota_root = root; |
4bbcaa64 ES |
2299 | } |
2300 | ||
2301 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2302 | root = btrfs_read_tree_root(tree_root, &location); |
2303 | if (IS_ERR(root)) { | |
2304 | ret = PTR_ERR(root); | |
4bbcaa64 | 2305 | if (ret != -ENOENT) |
f50f4353 | 2306 | goto out; |
4bbcaa64 | 2307 | } else { |
a4f3d2c4 DS |
2308 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2309 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2310 | } |
2311 | ||
70f6d82e OS |
2312 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2313 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2314 | root = btrfs_read_tree_root(tree_root, &location); | |
f50f4353 LB |
2315 | if (IS_ERR(root)) { |
2316 | ret = PTR_ERR(root); | |
2317 | goto out; | |
2318 | } | |
70f6d82e OS |
2319 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2320 | fs_info->free_space_root = root; | |
2321 | } | |
2322 | ||
4bbcaa64 | 2323 | return 0; |
f50f4353 LB |
2324 | out: |
2325 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2326 | location.objectid, ret); | |
2327 | return ret; | |
4bbcaa64 ES |
2328 | } |
2329 | ||
069ec957 QW |
2330 | /* |
2331 | * Real super block validation | |
2332 | * NOTE: super csum type and incompat features will not be checked here. | |
2333 | * | |
2334 | * @sb: super block to check | |
2335 | * @mirror_num: the super block number to check its bytenr: | |
2336 | * 0 the primary (1st) sb | |
2337 | * 1, 2 2nd and 3rd backup copy | |
2338 | * -1 skip bytenr check | |
2339 | */ | |
2340 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2341 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2342 | { |
21a852b0 QW |
2343 | u64 nodesize = btrfs_super_nodesize(sb); |
2344 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2345 | int ret = 0; | |
2346 | ||
2347 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2348 | btrfs_err(fs_info, "no valid FS found"); | |
2349 | ret = -EINVAL; | |
2350 | } | |
2351 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2352 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2353 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2354 | ret = -EINVAL; | |
2355 | } | |
2356 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2357 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2358 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2359 | ret = -EINVAL; | |
2360 | } | |
2361 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2362 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2363 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2364 | ret = -EINVAL; | |
2365 | } | |
2366 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2367 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2368 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2369 | ret = -EINVAL; | |
2370 | } | |
2371 | ||
2372 | /* | |
2373 | * Check sectorsize and nodesize first, other check will need it. | |
2374 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2375 | */ | |
2376 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2377 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2378 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2379 | ret = -EINVAL; | |
2380 | } | |
2381 | /* Only PAGE SIZE is supported yet */ | |
2382 | if (sectorsize != PAGE_SIZE) { | |
2383 | btrfs_err(fs_info, | |
2384 | "sectorsize %llu not supported yet, only support %lu", | |
2385 | sectorsize, PAGE_SIZE); | |
2386 | ret = -EINVAL; | |
2387 | } | |
2388 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
2389 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2390 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2391 | ret = -EINVAL; | |
2392 | } | |
2393 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2394 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2395 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2396 | ret = -EINVAL; | |
2397 | } | |
2398 | ||
2399 | /* Root alignment check */ | |
2400 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2401 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2402 | btrfs_super_root(sb)); | |
2403 | ret = -EINVAL; | |
2404 | } | |
2405 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2406 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2407 | btrfs_super_chunk_root(sb)); | |
2408 | ret = -EINVAL; | |
2409 | } | |
2410 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2411 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2412 | btrfs_super_log_root(sb)); | |
2413 | ret = -EINVAL; | |
2414 | } | |
2415 | ||
de37aa51 | 2416 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2417 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2418 | btrfs_err(fs_info, |
7239ff4b | 2419 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2420 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2421 | ret = -EINVAL; |
2422 | } | |
2423 | ||
2424 | /* | |
2425 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2426 | * done later | |
2427 | */ | |
2428 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2429 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2430 | btrfs_super_bytes_used(sb)); | |
2431 | ret = -EINVAL; | |
2432 | } | |
2433 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2434 | btrfs_err(fs_info, "invalid stripesize %u", | |
2435 | btrfs_super_stripesize(sb)); | |
2436 | ret = -EINVAL; | |
2437 | } | |
2438 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2439 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2440 | btrfs_super_num_devices(sb)); | |
2441 | if (btrfs_super_num_devices(sb) == 0) { | |
2442 | btrfs_err(fs_info, "number of devices is 0"); | |
2443 | ret = -EINVAL; | |
2444 | } | |
2445 | ||
069ec957 QW |
2446 | if (mirror_num >= 0 && |
2447 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2448 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2449 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2450 | ret = -EINVAL; | |
2451 | } | |
2452 | ||
2453 | /* | |
2454 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2455 | * and one chunk | |
2456 | */ | |
2457 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2458 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2459 | btrfs_super_sys_array_size(sb), | |
2460 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2461 | ret = -EINVAL; | |
2462 | } | |
2463 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2464 | + sizeof(struct btrfs_chunk)) { | |
2465 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2466 | btrfs_super_sys_array_size(sb), | |
2467 | sizeof(struct btrfs_disk_key) | |
2468 | + sizeof(struct btrfs_chunk)); | |
2469 | ret = -EINVAL; | |
2470 | } | |
2471 | ||
2472 | /* | |
2473 | * The generation is a global counter, we'll trust it more than the others | |
2474 | * but it's still possible that it's the one that's wrong. | |
2475 | */ | |
2476 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2477 | btrfs_warn(fs_info, | |
2478 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2479 | btrfs_super_generation(sb), | |
2480 | btrfs_super_chunk_root_generation(sb)); | |
2481 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2482 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2483 | btrfs_warn(fs_info, | |
2484 | "suspicious: generation < cache_generation: %llu < %llu", | |
2485 | btrfs_super_generation(sb), | |
2486 | btrfs_super_cache_generation(sb)); | |
2487 | ||
2488 | return ret; | |
2489 | } | |
2490 | ||
069ec957 QW |
2491 | /* |
2492 | * Validation of super block at mount time. | |
2493 | * Some checks already done early at mount time, like csum type and incompat | |
2494 | * flags will be skipped. | |
2495 | */ | |
2496 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2497 | { | |
2498 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2499 | } | |
2500 | ||
75cb857d QW |
2501 | /* |
2502 | * Validation of super block at write time. | |
2503 | * Some checks like bytenr check will be skipped as their values will be | |
2504 | * overwritten soon. | |
2505 | * Extra checks like csum type and incompat flags will be done here. | |
2506 | */ | |
2507 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2508 | struct btrfs_super_block *sb) | |
2509 | { | |
2510 | int ret; | |
2511 | ||
2512 | ret = validate_super(fs_info, sb, -1); | |
2513 | if (ret < 0) | |
2514 | goto out; | |
e7e16f48 | 2515 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2516 | ret = -EUCLEAN; |
2517 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2518 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2519 | goto out; | |
2520 | } | |
2521 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2522 | ret = -EUCLEAN; | |
2523 | btrfs_err(fs_info, | |
2524 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2525 | btrfs_super_incompat_flags(sb), | |
2526 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2527 | goto out; | |
2528 | } | |
2529 | out: | |
2530 | if (ret < 0) | |
2531 | btrfs_err(fs_info, | |
2532 | "super block corruption detected before writing it to disk"); | |
2533 | return ret; | |
2534 | } | |
2535 | ||
6ef108dd | 2536 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2537 | { |
6ef108dd | 2538 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2539 | struct btrfs_super_block *sb = fs_info->super_copy; |
2540 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2541 | bool handle_error = false; | |
2542 | int ret = 0; | |
2543 | int i; | |
2544 | ||
2545 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2546 | u64 generation; | |
2547 | int level; | |
2548 | ||
2549 | if (handle_error) { | |
2550 | if (!IS_ERR(tree_root->node)) | |
2551 | free_extent_buffer(tree_root->node); | |
2552 | tree_root->node = NULL; | |
2553 | ||
2554 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2555 | break; | |
2556 | ||
2557 | free_root_pointers(fs_info, 0); | |
2558 | ||
2559 | /* | |
2560 | * Don't use the log in recovery mode, it won't be | |
2561 | * valid | |
2562 | */ | |
2563 | btrfs_set_super_log_root(sb, 0); | |
2564 | ||
2565 | /* We can't trust the free space cache either */ | |
2566 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2567 | ||
2568 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2569 | backup_index = ret; |
b8522a1e NB |
2570 | if (ret < 0) |
2571 | return ret; | |
2572 | } | |
2573 | generation = btrfs_super_generation(sb); | |
2574 | level = btrfs_super_root_level(sb); | |
2575 | tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), | |
2576 | generation, level, NULL); | |
2577 | if (IS_ERR(tree_root->node) || | |
2578 | !extent_buffer_uptodate(tree_root->node)) { | |
2579 | handle_error = true; | |
2580 | ||
2581 | if (IS_ERR(tree_root->node)) | |
2582 | ret = PTR_ERR(tree_root->node); | |
2583 | else if (!extent_buffer_uptodate(tree_root->node)) | |
2584 | ret = -EUCLEAN; | |
2585 | ||
2586 | btrfs_warn(fs_info, "failed to read tree root"); | |
2587 | continue; | |
2588 | } | |
2589 | ||
2590 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); | |
2591 | tree_root->commit_root = btrfs_root_node(tree_root); | |
2592 | btrfs_set_root_refs(&tree_root->root_item, 1); | |
2593 | ||
336a0d8d NB |
2594 | /* |
2595 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2596 | * hasn't been fully initialised and we are the only user | |
2597 | */ | |
b8522a1e NB |
2598 | ret = btrfs_find_highest_objectid(tree_root, |
2599 | &tree_root->highest_objectid); | |
2600 | if (ret < 0) { | |
b8522a1e NB |
2601 | handle_error = true; |
2602 | continue; | |
2603 | } | |
2604 | ||
2605 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
b8522a1e NB |
2606 | |
2607 | ret = btrfs_read_roots(fs_info); | |
2608 | if (ret < 0) { | |
2609 | handle_error = true; | |
2610 | continue; | |
2611 | } | |
2612 | ||
2613 | /* All successful */ | |
2614 | fs_info->generation = generation; | |
2615 | fs_info->last_trans_committed = generation; | |
6ef108dd NB |
2616 | |
2617 | /* Always begin writing backup roots after the one being used */ | |
2618 | if (backup_index < 0) { | |
2619 | fs_info->backup_root_index = 0; | |
2620 | } else { | |
2621 | fs_info->backup_root_index = backup_index + 1; | |
2622 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2623 | } | |
b8522a1e NB |
2624 | break; |
2625 | } | |
2626 | ||
2627 | return ret; | |
2628 | } | |
2629 | ||
b105e927 | 2630 | int __cold open_ctree(struct super_block *sb, |
ad2b2c80 AV |
2631 | struct btrfs_fs_devices *fs_devices, |
2632 | char *options) | |
2e635a27 | 2633 | { |
db94535d CM |
2634 | u32 sectorsize; |
2635 | u32 nodesize; | |
87ee04eb | 2636 | u32 stripesize; |
84234f3a | 2637 | u64 generation; |
f2b636e8 | 2638 | u64 features; |
51bce6c9 | 2639 | u16 csum_type; |
3de4586c | 2640 | struct btrfs_key location; |
a061fc8d | 2641 | struct buffer_head *bh; |
4d34b278 | 2642 | struct btrfs_super_block *disk_super; |
815745cf | 2643 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2644 | struct btrfs_root *tree_root; |
4d34b278 | 2645 | struct btrfs_root *chunk_root; |
eb60ceac | 2646 | int ret; |
e58ca020 | 2647 | int err = -EINVAL; |
6675df31 | 2648 | int clear_free_space_tree = 0; |
581c1760 | 2649 | int level; |
4543df7e | 2650 | |
96dfcb46 JB |
2651 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, |
2652 | GFP_KERNEL); | |
2653 | fs_info->tree_root = tree_root; | |
2654 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
2655 | GFP_KERNEL); | |
2656 | fs_info->chunk_root = chunk_root; | |
cb517eab | 2657 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2658 | err = -ENOMEM; |
2659 | goto fail; | |
2660 | } | |
76dda93c YZ |
2661 | |
2662 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2663 | if (ret) { | |
2664 | err = ret; | |
2665 | goto fail; | |
2666 | } | |
2667 | ||
4297ff84 | 2668 | ret = percpu_counter_init(&fs_info->dio_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2669 | if (ret) { |
2670 | err = ret; | |
9e11ceee | 2671 | goto fail_srcu; |
e2d84521 | 2672 | } |
4297ff84 JB |
2673 | |
2674 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
2675 | if (ret) { | |
2676 | err = ret; | |
2677 | goto fail_dio_bytes; | |
2678 | } | |
09cbfeaf | 2679 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2680 | (1 + ilog2(nr_cpu_ids)); |
2681 | ||
908c7f19 | 2682 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2683 | if (ret) { |
2684 | err = ret; | |
2685 | goto fail_dirty_metadata_bytes; | |
2686 | } | |
2687 | ||
7f8d236a DS |
2688 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, |
2689 | GFP_KERNEL); | |
c404e0dc MX |
2690 | if (ret) { |
2691 | err = ret; | |
2692 | goto fail_delalloc_bytes; | |
2693 | } | |
2694 | ||
76dda93c | 2695 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2696 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2697 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2698 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2699 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2700 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2701 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2702 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2703 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2704 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2705 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2706 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2707 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2708 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2709 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2710 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2711 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2712 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2713 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2714 | mutex_init(&fs_info->delalloc_root_mutex); |
de98ced9 | 2715 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2716 | |
0b86a832 | 2717 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2718 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2719 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2720 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
c8bf1b67 | 2721 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2722 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2723 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2724 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2725 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2726 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2727 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2728 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2729 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2730 | BTRFS_BLOCK_RSV_DELREFS); | |
2731 | ||
771ed689 | 2732 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2733 | atomic_set(&fs_info->defrag_running, 0); |
2fefd558 | 2734 | atomic_set(&fs_info->reada_works_cnt, 0); |
034f784d | 2735 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2736 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2737 | fs_info->sb = sb; |
95ac567a | 2738 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2739 | fs_info->metadata_ratio = 0; |
4cb5300b | 2740 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2741 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2742 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2743 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2744 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2745 | /* readahead state */ |
d0164adc | 2746 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2747 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2748 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2749 | |
b34b086c CM |
2750 | fs_info->thread_pool_size = min_t(unsigned long, |
2751 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2752 | |
199c2a9c MX |
2753 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2754 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 JB |
2755 | |
2756 | fs_info->btree_inode = new_inode(sb); | |
2757 | if (!fs_info->btree_inode) { | |
2758 | err = -ENOMEM; | |
2759 | goto fail_bio_counter; | |
2760 | } | |
2761 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
2762 | ||
16cdcec7 | 2763 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2764 | GFP_KERNEL); |
16cdcec7 MX |
2765 | if (!fs_info->delayed_root) { |
2766 | err = -ENOMEM; | |
2767 | goto fail_iput; | |
2768 | } | |
2769 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2770 | |
638aa7ed | 2771 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2772 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2773 | fs_info->check_integrity_print_mask = 0; | |
2774 | #endif | |
779a65a4 | 2775 | btrfs_init_balance(fs_info); |
21c7e756 | 2776 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2777 | |
9f6d2510 DS |
2778 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2779 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
a061fc8d | 2780 | |
6bccf3ab | 2781 | btrfs_init_btree_inode(fs_info); |
76dda93c | 2782 | |
0f9dd46c | 2783 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2784 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2785 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2786 | |
43eb5f29 QW |
2787 | extent_io_tree_init(fs_info, &fs_info->freed_extents[0], |
2788 | IO_TREE_FS_INFO_FREED_EXTENTS0, NULL); | |
2789 | extent_io_tree_init(fs_info, &fs_info->freed_extents[1], | |
2790 | IO_TREE_FS_INFO_FREED_EXTENTS1, NULL); | |
11833d66 | 2791 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
afcdd129 | 2792 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2793 | |
5a3f23d5 | 2794 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2795 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2796 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2797 | mutex_init(&fs_info->transaction_kthread_mutex); |
2798 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2799 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2800 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2801 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2802 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2803 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2804 | |
ad618368 | 2805 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2806 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2807 | btrfs_discard_init(fs_info); |
416ac51d | 2808 | |
fa9c0d79 CM |
2809 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2810 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2811 | ||
e6dcd2dc | 2812 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2813 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2814 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2815 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2816 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2817 | |
da17066c JM |
2818 | /* Usable values until the real ones are cached from the superblock */ |
2819 | fs_info->nodesize = 4096; | |
2820 | fs_info->sectorsize = 4096; | |
2821 | fs_info->stripesize = 4096; | |
2822 | ||
eede2bf3 OS |
2823 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2824 | fs_info->swapfile_pins = RB_ROOT; | |
2825 | ||
9e967495 FM |
2826 | fs_info->send_in_progress = 0; |
2827 | ||
53b381b3 DW |
2828 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2829 | if (ret) { | |
83c8266a | 2830 | err = ret; |
53b381b3 DW |
2831 | goto fail_alloc; |
2832 | } | |
2833 | ||
3c4bb26b | 2834 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2835 | |
2836 | /* | |
2837 | * Read super block and check the signature bytes only | |
2838 | */ | |
a512bbf8 | 2839 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2840 | if (IS_ERR(bh)) { |
2841 | err = PTR_ERR(bh); | |
16cdcec7 | 2842 | goto fail_alloc; |
20b45077 | 2843 | } |
39279cc3 | 2844 | |
8dc3f22c JT |
2845 | /* |
2846 | * Verify the type first, if that or the the checksum value are | |
2847 | * corrupted, we'll find out | |
2848 | */ | |
51bce6c9 JT |
2849 | csum_type = btrfs_super_csum_type((struct btrfs_super_block *)bh->b_data); |
2850 | if (!btrfs_supported_super_csum(csum_type)) { | |
8dc3f22c | 2851 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 2852 | csum_type); |
8dc3f22c JT |
2853 | err = -EINVAL; |
2854 | brelse(bh); | |
2855 | goto fail_alloc; | |
2856 | } | |
2857 | ||
6d97c6e3 JT |
2858 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
2859 | if (ret) { | |
2860 | err = ret; | |
2861 | goto fail_alloc; | |
2862 | } | |
2863 | ||
1104a885 DS |
2864 | /* |
2865 | * We want to check superblock checksum, the type is stored inside. | |
2866 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2867 | */ | |
ab8d0fc4 | 2868 | if (btrfs_check_super_csum(fs_info, bh->b_data)) { |
05135f59 | 2869 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2870 | err = -EINVAL; |
b2acdddf | 2871 | brelse(bh); |
6d97c6e3 | 2872 | goto fail_csum; |
1104a885 DS |
2873 | } |
2874 | ||
2875 | /* | |
2876 | * super_copy is zeroed at allocation time and we never touch the | |
2877 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2878 | * the whole block of INFO_SIZE | |
2879 | */ | |
6c41761f | 2880 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
a061fc8d | 2881 | brelse(bh); |
5f39d397 | 2882 | |
fbc6feae NB |
2883 | disk_super = fs_info->super_copy; |
2884 | ||
de37aa51 NB |
2885 | ASSERT(!memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2886 | BTRFS_FSID_SIZE)); | |
2887 | ||
7239ff4b | 2888 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) { |
de37aa51 NB |
2889 | ASSERT(!memcmp(fs_info->fs_devices->metadata_uuid, |
2890 | fs_info->super_copy->metadata_uuid, | |
2891 | BTRFS_FSID_SIZE)); | |
7239ff4b | 2892 | } |
0b86a832 | 2893 | |
fbc6feae NB |
2894 | features = btrfs_super_flags(disk_super); |
2895 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
2896 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
2897 | btrfs_set_super_flags(disk_super, features); | |
2898 | btrfs_info(fs_info, | |
2899 | "found metadata UUID change in progress flag, clearing"); | |
2900 | } | |
2901 | ||
2902 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2903 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 2904 | |
069ec957 | 2905 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 2906 | if (ret) { |
05135f59 | 2907 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 2908 | err = -EINVAL; |
6d97c6e3 | 2909 | goto fail_csum; |
1104a885 DS |
2910 | } |
2911 | ||
0f7d52f4 | 2912 | if (!btrfs_super_root(disk_super)) |
6d97c6e3 | 2913 | goto fail_csum; |
0f7d52f4 | 2914 | |
acce952b | 2915 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2916 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2917 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2918 | |
75e7cb7f LB |
2919 | /* |
2920 | * In the long term, we'll store the compression type in the super | |
2921 | * block, and it'll be used for per file compression control. | |
2922 | */ | |
2923 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2924 | ||
2ff7e61e | 2925 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
2926 | if (ret) { |
2927 | err = ret; | |
6d97c6e3 | 2928 | goto fail_csum; |
2b82032c | 2929 | } |
dfe25020 | 2930 | |
f2b636e8 JB |
2931 | features = btrfs_super_incompat_flags(disk_super) & |
2932 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2933 | if (features) { | |
05135f59 DS |
2934 | btrfs_err(fs_info, |
2935 | "cannot mount because of unsupported optional features (%llx)", | |
2936 | features); | |
f2b636e8 | 2937 | err = -EINVAL; |
6d97c6e3 | 2938 | goto fail_csum; |
f2b636e8 JB |
2939 | } |
2940 | ||
5d4f98a2 | 2941 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2942 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 2943 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2944 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
2945 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
2946 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 2947 | |
3173a18f | 2948 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2949 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2950 | |
727011e0 CM |
2951 | /* |
2952 | * flag our filesystem as having big metadata blocks if | |
2953 | * they are bigger than the page size | |
2954 | */ | |
09cbfeaf | 2955 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2956 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2957 | btrfs_info(fs_info, |
2958 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2959 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2960 | } | |
2961 | ||
bc3f116f | 2962 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2963 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2964 | stripesize = sectorsize; |
707e8a07 | 2965 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2966 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f | 2967 | |
da17066c JM |
2968 | /* Cache block sizes */ |
2969 | fs_info->nodesize = nodesize; | |
2970 | fs_info->sectorsize = sectorsize; | |
2971 | fs_info->stripesize = stripesize; | |
2972 | ||
bc3f116f CM |
2973 | /* |
2974 | * mixed block groups end up with duplicate but slightly offset | |
2975 | * extent buffers for the same range. It leads to corruptions | |
2976 | */ | |
2977 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2978 | (sectorsize != nodesize)) { |
05135f59 DS |
2979 | btrfs_err(fs_info, |
2980 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2981 | nodesize, sectorsize); | |
6d97c6e3 | 2982 | goto fail_csum; |
bc3f116f CM |
2983 | } |
2984 | ||
ceda0864 MX |
2985 | /* |
2986 | * Needn't use the lock because there is no other task which will | |
2987 | * update the flag. | |
2988 | */ | |
a6fa6fae | 2989 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2990 | |
f2b636e8 JB |
2991 | features = btrfs_super_compat_ro_flags(disk_super) & |
2992 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 2993 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
2994 | btrfs_err(fs_info, |
2995 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2996 | features); |
f2b636e8 | 2997 | err = -EINVAL; |
6d97c6e3 | 2998 | goto fail_csum; |
f2b636e8 | 2999 | } |
61d92c32 | 3000 | |
2a458198 ES |
3001 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
3002 | if (ret) { | |
3003 | err = ret; | |
0dc3b84a JB |
3004 | goto fail_sb_buffer; |
3005 | } | |
4543df7e | 3006 | |
9e11ceee JK |
3007 | sb->s_bdi->congested_fn = btrfs_congested_fn; |
3008 | sb->s_bdi->congested_data = fs_info; | |
3009 | sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; | |
b5420237 | 3010 | sb->s_bdi->ra_pages = VM_READAHEAD_PAGES; |
9e11ceee JK |
3011 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3012 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3013 | |
a061fc8d CM |
3014 | sb->s_blocksize = sectorsize; |
3015 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3016 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3017 | |
925baedd | 3018 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3019 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3020 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3021 | if (ret) { |
05135f59 | 3022 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3023 | goto fail_sb_buffer; |
84eed90f | 3024 | } |
0b86a832 | 3025 | |
84234f3a | 3026 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3027 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3028 | |
2ff7e61e | 3029 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3030 | btrfs_super_chunk_root(disk_super), |
581c1760 | 3031 | generation, level, NULL); |
64c043de LB |
3032 | if (IS_ERR(chunk_root->node) || |
3033 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3034 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3035 | if (!IS_ERR(chunk_root->node)) |
3036 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3037 | chunk_root->node = NULL; |
af31f5e5 | 3038 | goto fail_tree_roots; |
83121942 | 3039 | } |
5d4f98a2 YZ |
3040 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3041 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3042 | |
e17cade2 | 3043 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 3044 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 3045 | |
5b4aacef | 3046 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3047 | if (ret) { |
05135f59 | 3048 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3049 | goto fail_tree_roots; |
2b82032c | 3050 | } |
0b86a832 | 3051 | |
8dabb742 | 3052 | /* |
9b99b115 AJ |
3053 | * Keep the devid that is marked to be the target device for the |
3054 | * device replace procedure | |
8dabb742 | 3055 | */ |
9b99b115 | 3056 | btrfs_free_extra_devids(fs_devices, 0); |
dfe25020 | 3057 | |
a6b0d5c8 | 3058 | if (!fs_devices->latest_bdev) { |
05135f59 | 3059 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3060 | goto fail_tree_roots; |
3061 | } | |
3062 | ||
b8522a1e | 3063 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3064 | if (ret) |
b8522a1e | 3065 | goto fail_tree_roots; |
8929ecfa | 3066 | |
cf90d884 QW |
3067 | ret = btrfs_verify_dev_extents(fs_info); |
3068 | if (ret) { | |
3069 | btrfs_err(fs_info, | |
3070 | "failed to verify dev extents against chunks: %d", | |
3071 | ret); | |
3072 | goto fail_block_groups; | |
3073 | } | |
68310a5e ID |
3074 | ret = btrfs_recover_balance(fs_info); |
3075 | if (ret) { | |
05135f59 | 3076 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3077 | goto fail_block_groups; |
3078 | } | |
3079 | ||
733f4fbb SB |
3080 | ret = btrfs_init_dev_stats(fs_info); |
3081 | if (ret) { | |
05135f59 | 3082 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3083 | goto fail_block_groups; |
3084 | } | |
3085 | ||
8dabb742 SB |
3086 | ret = btrfs_init_dev_replace(fs_info); |
3087 | if (ret) { | |
05135f59 | 3088 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3089 | goto fail_block_groups; |
3090 | } | |
3091 | ||
9b99b115 | 3092 | btrfs_free_extra_devids(fs_devices, 1); |
8dabb742 | 3093 | |
c6761a9e | 3094 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3095 | if (ret) { |
05135f59 DS |
3096 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3097 | ret); | |
b7c35e81 AJ |
3098 | goto fail_block_groups; |
3099 | } | |
3100 | ||
96f3136e | 3101 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3102 | if (ret) { |
05135f59 | 3103 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3104 | goto fail_fsdev_sysfs; |
c59021f8 | 3105 | } |
3106 | ||
c59021f8 | 3107 | ret = btrfs_init_space_info(fs_info); |
3108 | if (ret) { | |
05135f59 | 3109 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3110 | goto fail_sysfs; |
c59021f8 | 3111 | } |
3112 | ||
5b4aacef | 3113 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3114 | if (ret) { |
05135f59 | 3115 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3116 | goto fail_sysfs; |
1b1d1f66 | 3117 | } |
4330e183 | 3118 | |
6528b99d | 3119 | if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { |
05135f59 | 3120 | btrfs_warn(fs_info, |
52042d8e | 3121 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3122 | goto fail_sysfs; |
292fd7fc | 3123 | } |
9078a3e1 | 3124 | |
a74a4b97 CM |
3125 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3126 | "btrfs-cleaner"); | |
57506d50 | 3127 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3128 | goto fail_sysfs; |
a74a4b97 CM |
3129 | |
3130 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3131 | tree_root, | |
3132 | "btrfs-transaction"); | |
57506d50 | 3133 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3134 | goto fail_cleaner; |
a74a4b97 | 3135 | |
583b7231 | 3136 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3137 | !fs_info->fs_devices->rotating) { |
583b7231 | 3138 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3139 | } |
3140 | ||
572d9ab7 | 3141 | /* |
01327610 | 3142 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3143 | * not have to wait for transaction commit |
3144 | */ | |
3145 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3146 | |
21adbd5c | 3147 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3148 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3149 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3150 | btrfs_test_opt(fs_info, |
21adbd5c SB |
3151 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3152 | 1 : 0, | |
3153 | fs_info->check_integrity_print_mask); | |
3154 | if (ret) | |
05135f59 DS |
3155 | btrfs_warn(fs_info, |
3156 | "failed to initialize integrity check module: %d", | |
3157 | ret); | |
21adbd5c SB |
3158 | } |
3159 | #endif | |
bcef60f2 AJ |
3160 | ret = btrfs_read_qgroup_config(fs_info); |
3161 | if (ret) | |
3162 | goto fail_trans_kthread; | |
21adbd5c | 3163 | |
fd708b81 JB |
3164 | if (btrfs_build_ref_tree(fs_info)) |
3165 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3166 | ||
96da0919 QW |
3167 | /* do not make disk changes in broken FS or nologreplay is given */ |
3168 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3169 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3170 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3171 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3172 | if (ret) { |
63443bf5 | 3173 | err = ret; |
28c16cbb | 3174 | goto fail_qgroup; |
79787eaa | 3175 | } |
e02119d5 | 3176 | } |
1a40e23b | 3177 | |
6bccf3ab | 3178 | ret = btrfs_find_orphan_roots(fs_info); |
79787eaa | 3179 | if (ret) |
28c16cbb | 3180 | goto fail_qgroup; |
76dda93c | 3181 | |
bc98a42c | 3182 | if (!sb_rdonly(sb)) { |
d68fc57b | 3183 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3184 | if (ret) |
28c16cbb | 3185 | goto fail_qgroup; |
90c711ab ZB |
3186 | |
3187 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3188 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3189 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3190 | if (ret < 0) { |
05135f59 DS |
3191 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3192 | ret); | |
d7ce5843 | 3193 | err = -EINVAL; |
bcef60f2 | 3194 | goto fail_qgroup; |
d7ce5843 | 3195 | } |
7c2ca468 | 3196 | } |
1a40e23b | 3197 | |
3de4586c CM |
3198 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3199 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3200 | location.offset = 0; |
3de4586c | 3201 | |
3619c94f | 3202 | fs_info->fs_root = btrfs_get_fs_root(fs_info, &location, true); |
3140c9a3 DC |
3203 | if (IS_ERR(fs_info->fs_root)) { |
3204 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3205 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3206 | fs_info->fs_root = NULL; |
bcef60f2 | 3207 | goto fail_qgroup; |
3140c9a3 | 3208 | } |
c289811c | 3209 | |
4c78e9f5 JB |
3210 | if (!btrfs_grab_fs_root(fs_info->fs_root)) { |
3211 | fs_info->fs_root = NULL; | |
3212 | err = -ENOENT; | |
3213 | btrfs_warn(fs_info, "failed to grab a ref on the fs tree"); | |
3214 | goto fail_qgroup; | |
3215 | } | |
3216 | ||
bc98a42c | 3217 | if (sb_rdonly(sb)) |
2b6ba629 | 3218 | return 0; |
59641015 | 3219 | |
f8d468a1 OS |
3220 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && |
3221 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
6675df31 OS |
3222 | clear_free_space_tree = 1; |
3223 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3224 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3225 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3226 | clear_free_space_tree = 1; | |
3227 | } | |
3228 | ||
3229 | if (clear_free_space_tree) { | |
f8d468a1 OS |
3230 | btrfs_info(fs_info, "clearing free space tree"); |
3231 | ret = btrfs_clear_free_space_tree(fs_info); | |
3232 | if (ret) { | |
3233 | btrfs_warn(fs_info, | |
3234 | "failed to clear free space tree: %d", ret); | |
6bccf3ab | 3235 | close_ctree(fs_info); |
f8d468a1 OS |
3236 | return ret; |
3237 | } | |
3238 | } | |
3239 | ||
0b246afa | 3240 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
511711af | 3241 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3242 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3243 | ret = btrfs_create_free_space_tree(fs_info); |
3244 | if (ret) { | |
05135f59 DS |
3245 | btrfs_warn(fs_info, |
3246 | "failed to create free space tree: %d", ret); | |
6bccf3ab | 3247 | close_ctree(fs_info); |
511711af CM |
3248 | return ret; |
3249 | } | |
3250 | } | |
3251 | ||
2b6ba629 ID |
3252 | down_read(&fs_info->cleanup_work_sem); |
3253 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3254 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3255 | up_read(&fs_info->cleanup_work_sem); |
6bccf3ab | 3256 | close_ctree(fs_info); |
2b6ba629 ID |
3257 | return ret; |
3258 | } | |
3259 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3260 | |
2b6ba629 ID |
3261 | ret = btrfs_resume_balance_async(fs_info); |
3262 | if (ret) { | |
05135f59 | 3263 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
6bccf3ab | 3264 | close_ctree(fs_info); |
2b6ba629 | 3265 | return ret; |
e3acc2a6 JB |
3266 | } |
3267 | ||
8dabb742 SB |
3268 | ret = btrfs_resume_dev_replace_async(fs_info); |
3269 | if (ret) { | |
05135f59 | 3270 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
6bccf3ab | 3271 | close_ctree(fs_info); |
8dabb742 SB |
3272 | return ret; |
3273 | } | |
3274 | ||
b382a324 | 3275 | btrfs_qgroup_rescan_resume(fs_info); |
b0643e59 | 3276 | btrfs_discard_resume(fs_info); |
b382a324 | 3277 | |
4bbcaa64 | 3278 | if (!fs_info->uuid_root) { |
05135f59 | 3279 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3280 | ret = btrfs_create_uuid_tree(fs_info); |
3281 | if (ret) { | |
05135f59 DS |
3282 | btrfs_warn(fs_info, |
3283 | "failed to create the UUID tree: %d", ret); | |
6bccf3ab | 3284 | close_ctree(fs_info); |
f7a81ea4 SB |
3285 | return ret; |
3286 | } | |
0b246afa | 3287 | } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3288 | fs_info->generation != |
3289 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3290 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3291 | ret = btrfs_check_uuid_tree(fs_info); |
3292 | if (ret) { | |
05135f59 DS |
3293 | btrfs_warn(fs_info, |
3294 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3295 | close_ctree(fs_info); |
70f80175 SB |
3296 | return ret; |
3297 | } | |
3298 | } else { | |
afcdd129 | 3299 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); |
f7a81ea4 | 3300 | } |
afcdd129 | 3301 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3302 | |
8dcddfa0 QW |
3303 | /* |
3304 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3305 | * no need to keep the flag | |
3306 | */ | |
3307 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3308 | ||
ad2b2c80 | 3309 | return 0; |
39279cc3 | 3310 | |
bcef60f2 AJ |
3311 | fail_qgroup: |
3312 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3313 | fail_trans_kthread: |
3314 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3315 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3316 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3317 | fail_cleaner: |
a74a4b97 | 3318 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3319 | |
3320 | /* | |
3321 | * make sure we're done with the btree inode before we stop our | |
3322 | * kthreads | |
3323 | */ | |
3324 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3325 | |
2365dd3c | 3326 | fail_sysfs: |
6618a59b | 3327 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3328 | |
b7c35e81 AJ |
3329 | fail_fsdev_sysfs: |
3330 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3331 | ||
1b1d1f66 | 3332 | fail_block_groups: |
54067ae9 | 3333 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3334 | |
3335 | fail_tree_roots: | |
4273eaff | 3336 | free_root_pointers(fs_info, true); |
2b8195bb | 3337 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3338 | |
39279cc3 | 3339 | fail_sb_buffer: |
7abadb64 | 3340 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3341 | btrfs_free_block_groups(fs_info); |
6d97c6e3 JT |
3342 | fail_csum: |
3343 | btrfs_free_csum_hash(fs_info); | |
16cdcec7 | 3344 | fail_alloc: |
4543df7e | 3345 | fail_iput: |
586e46e2 ID |
3346 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3347 | ||
4543df7e | 3348 | iput(fs_info->btree_inode); |
c404e0dc | 3349 | fail_bio_counter: |
7f8d236a | 3350 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
963d678b MX |
3351 | fail_delalloc_bytes: |
3352 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3353 | fail_dirty_metadata_bytes: |
3354 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
4297ff84 JB |
3355 | fail_dio_bytes: |
3356 | percpu_counter_destroy(&fs_info->dio_bytes); | |
76dda93c YZ |
3357 | fail_srcu: |
3358 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3359 | fail: |
53b381b3 | 3360 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3361 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3362 | return err; |
eb60ceac | 3363 | } |
663faf9f | 3364 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3365 | |
f2984462 CM |
3366 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3367 | { | |
f2984462 CM |
3368 | if (uptodate) { |
3369 | set_buffer_uptodate(bh); | |
3370 | } else { | |
442a4f63 SB |
3371 | struct btrfs_device *device = (struct btrfs_device *) |
3372 | bh->b_private; | |
3373 | ||
fb456252 | 3374 | btrfs_warn_rl_in_rcu(device->fs_info, |
b14af3b4 | 3375 | "lost page write due to IO error on %s", |
606686ee | 3376 | rcu_str_deref(device->name)); |
01327610 | 3377 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3378 | * our own ways of dealing with the IO errors |
3379 | */ | |
f2984462 | 3380 | clear_buffer_uptodate(bh); |
442a4f63 | 3381 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3382 | } |
3383 | unlock_buffer(bh); | |
3384 | put_bh(bh); | |
3385 | } | |
3386 | ||
29c36d72 AJ |
3387 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3388 | struct buffer_head **bh_ret) | |
3389 | { | |
3390 | struct buffer_head *bh; | |
3391 | struct btrfs_super_block *super; | |
3392 | u64 bytenr; | |
3393 | ||
3394 | bytenr = btrfs_sb_offset(copy_num); | |
3395 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3396 | return -EINVAL; | |
3397 | ||
9f6d2510 | 3398 | bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); |
29c36d72 AJ |
3399 | /* |
3400 | * If we fail to read from the underlying devices, as of now | |
3401 | * the best option we have is to mark it EIO. | |
3402 | */ | |
3403 | if (!bh) | |
3404 | return -EIO; | |
3405 | ||
3406 | super = (struct btrfs_super_block *)bh->b_data; | |
3407 | if (btrfs_super_bytenr(super) != bytenr || | |
3408 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3409 | brelse(bh); | |
3410 | return -EINVAL; | |
3411 | } | |
3412 | ||
3413 | *bh_ret = bh; | |
3414 | return 0; | |
3415 | } | |
3416 | ||
3417 | ||
a512bbf8 YZ |
3418 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3419 | { | |
3420 | struct buffer_head *bh; | |
3421 | struct buffer_head *latest = NULL; | |
3422 | struct btrfs_super_block *super; | |
3423 | int i; | |
3424 | u64 transid = 0; | |
92fc03fb | 3425 | int ret = -EINVAL; |
a512bbf8 YZ |
3426 | |
3427 | /* we would like to check all the supers, but that would make | |
3428 | * a btrfs mount succeed after a mkfs from a different FS. | |
3429 | * So, we need to add a special mount option to scan for | |
3430 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3431 | */ | |
3432 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3433 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3434 | if (ret) | |
a512bbf8 YZ |
3435 | continue; |
3436 | ||
3437 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3438 | |
3439 | if (!latest || btrfs_super_generation(super) > transid) { | |
3440 | brelse(latest); | |
3441 | latest = bh; | |
3442 | transid = btrfs_super_generation(super); | |
3443 | } else { | |
3444 | brelse(bh); | |
3445 | } | |
3446 | } | |
92fc03fb AJ |
3447 | |
3448 | if (!latest) | |
3449 | return ERR_PTR(ret); | |
3450 | ||
a512bbf8 YZ |
3451 | return latest; |
3452 | } | |
3453 | ||
4eedeb75 | 3454 | /* |
abbb3b8e DS |
3455 | * Write superblock @sb to the @device. Do not wait for completion, all the |
3456 | * buffer heads we write are pinned. | |
4eedeb75 | 3457 | * |
abbb3b8e DS |
3458 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3459 | * the expected device size at commit time. Note that max_mirrors must be | |
3460 | * same for write and wait phases. | |
4eedeb75 | 3461 | * |
abbb3b8e | 3462 | * Return number of errors when buffer head is not found or submission fails. |
4eedeb75 | 3463 | */ |
a512bbf8 | 3464 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3465 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3466 | { |
d5178578 JT |
3467 | struct btrfs_fs_info *fs_info = device->fs_info; |
3468 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); | |
a512bbf8 YZ |
3469 | struct buffer_head *bh; |
3470 | int i; | |
3471 | int ret; | |
3472 | int errors = 0; | |
a512bbf8 | 3473 | u64 bytenr; |
1b9e619c | 3474 | int op_flags; |
a512bbf8 YZ |
3475 | |
3476 | if (max_mirrors == 0) | |
3477 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3478 | ||
d5178578 JT |
3479 | shash->tfm = fs_info->csum_shash; |
3480 | ||
a512bbf8 YZ |
3481 | for (i = 0; i < max_mirrors; i++) { |
3482 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3483 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3484 | device->commit_total_bytes) | |
a512bbf8 YZ |
3485 | break; |
3486 | ||
abbb3b8e | 3487 | btrfs_set_super_bytenr(sb, bytenr); |
4eedeb75 | 3488 | |
d5178578 JT |
3489 | crypto_shash_init(shash); |
3490 | crypto_shash_update(shash, (const char *)sb + BTRFS_CSUM_SIZE, | |
3491 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
3492 | crypto_shash_final(shash, sb->csum); | |
4eedeb75 | 3493 | |
abbb3b8e | 3494 | /* One reference for us, and we leave it for the caller */ |
9f6d2510 | 3495 | bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, |
abbb3b8e DS |
3496 | BTRFS_SUPER_INFO_SIZE); |
3497 | if (!bh) { | |
3498 | btrfs_err(device->fs_info, | |
3499 | "couldn't get super buffer head for bytenr %llu", | |
3500 | bytenr); | |
3501 | errors++; | |
4eedeb75 | 3502 | continue; |
abbb3b8e | 3503 | } |
634554dc | 3504 | |
abbb3b8e | 3505 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
a512bbf8 | 3506 | |
abbb3b8e DS |
3507 | /* one reference for submit_bh */ |
3508 | get_bh(bh); | |
4eedeb75 | 3509 | |
abbb3b8e DS |
3510 | set_buffer_uptodate(bh); |
3511 | lock_buffer(bh); | |
3512 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
3513 | bh->b_private = device; | |
a512bbf8 | 3514 | |
387125fc CM |
3515 | /* |
3516 | * we fua the first super. The others we allow | |
3517 | * to go down lazy. | |
3518 | */ | |
1b9e619c OS |
3519 | op_flags = REQ_SYNC | REQ_META | REQ_PRIO; |
3520 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) | |
3521 | op_flags |= REQ_FUA; | |
3522 | ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); | |
4eedeb75 | 3523 | if (ret) |
a512bbf8 | 3524 | errors++; |
a512bbf8 YZ |
3525 | } |
3526 | return errors < i ? 0 : -1; | |
3527 | } | |
3528 | ||
abbb3b8e DS |
3529 | /* |
3530 | * Wait for write completion of superblocks done by write_dev_supers, | |
3531 | * @max_mirrors same for write and wait phases. | |
3532 | * | |
3533 | * Return number of errors when buffer head is not found or not marked up to | |
3534 | * date. | |
3535 | */ | |
3536 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3537 | { | |
3538 | struct buffer_head *bh; | |
3539 | int i; | |
3540 | int errors = 0; | |
b6a535fa | 3541 | bool primary_failed = false; |
abbb3b8e DS |
3542 | u64 bytenr; |
3543 | ||
3544 | if (max_mirrors == 0) | |
3545 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3546 | ||
3547 | for (i = 0; i < max_mirrors; i++) { | |
3548 | bytenr = btrfs_sb_offset(i); | |
3549 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
3550 | device->commit_total_bytes) | |
3551 | break; | |
3552 | ||
9f6d2510 DS |
3553 | bh = __find_get_block(device->bdev, |
3554 | bytenr / BTRFS_BDEV_BLOCKSIZE, | |
abbb3b8e DS |
3555 | BTRFS_SUPER_INFO_SIZE); |
3556 | if (!bh) { | |
3557 | errors++; | |
b6a535fa HM |
3558 | if (i == 0) |
3559 | primary_failed = true; | |
abbb3b8e DS |
3560 | continue; |
3561 | } | |
3562 | wait_on_buffer(bh); | |
b6a535fa | 3563 | if (!buffer_uptodate(bh)) { |
abbb3b8e | 3564 | errors++; |
b6a535fa HM |
3565 | if (i == 0) |
3566 | primary_failed = true; | |
3567 | } | |
abbb3b8e DS |
3568 | |
3569 | /* drop our reference */ | |
3570 | brelse(bh); | |
3571 | ||
3572 | /* drop the reference from the writing run */ | |
3573 | brelse(bh); | |
3574 | } | |
3575 | ||
b6a535fa HM |
3576 | /* log error, force error return */ |
3577 | if (primary_failed) { | |
3578 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3579 | device->devid); | |
3580 | return -1; | |
3581 | } | |
3582 | ||
abbb3b8e DS |
3583 | return errors < i ? 0 : -1; |
3584 | } | |
3585 | ||
387125fc CM |
3586 | /* |
3587 | * endio for the write_dev_flush, this will wake anyone waiting | |
3588 | * for the barrier when it is done | |
3589 | */ | |
4246a0b6 | 3590 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3591 | { |
e0ae9994 | 3592 | complete(bio->bi_private); |
387125fc CM |
3593 | } |
3594 | ||
3595 | /* | |
4fc6441a AJ |
3596 | * Submit a flush request to the device if it supports it. Error handling is |
3597 | * done in the waiting counterpart. | |
387125fc | 3598 | */ |
4fc6441a | 3599 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3600 | { |
c2a9c7ab | 3601 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3602 | struct bio *bio = device->flush_bio; |
387125fc | 3603 | |
c2a9c7ab | 3604 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3605 | return; |
387125fc | 3606 | |
e0ae9994 | 3607 | bio_reset(bio); |
387125fc | 3608 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 3609 | bio_set_dev(bio, device->bdev); |
8d910125 | 3610 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3611 | init_completion(&device->flush_wait); |
3612 | bio->bi_private = &device->flush_wait; | |
387125fc | 3613 | |
43a01111 | 3614 | btrfsic_submit_bio(bio); |
1c3063b6 | 3615 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3616 | } |
387125fc | 3617 | |
4fc6441a AJ |
3618 | /* |
3619 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
3620 | */ | |
8c27cb35 | 3621 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3622 | { |
4fc6441a | 3623 | struct bio *bio = device->flush_bio; |
387125fc | 3624 | |
1c3063b6 | 3625 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 3626 | return BLK_STS_OK; |
387125fc | 3627 | |
1c3063b6 | 3628 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 3629 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3630 | |
8c27cb35 | 3631 | return bio->bi_status; |
387125fc | 3632 | } |
387125fc | 3633 | |
d10b82fe | 3634 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 3635 | { |
6528b99d | 3636 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 3637 | return -EIO; |
387125fc CM |
3638 | return 0; |
3639 | } | |
3640 | ||
3641 | /* | |
3642 | * send an empty flush down to each device in parallel, | |
3643 | * then wait for them | |
3644 | */ | |
3645 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3646 | { | |
3647 | struct list_head *head; | |
3648 | struct btrfs_device *dev; | |
5af3e8cc | 3649 | int errors_wait = 0; |
4e4cbee9 | 3650 | blk_status_t ret; |
387125fc | 3651 | |
1538e6c5 | 3652 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3653 | /* send down all the barriers */ |
3654 | head = &info->fs_devices->devices; | |
1538e6c5 | 3655 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3656 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3657 | continue; |
cea7c8bf | 3658 | if (!dev->bdev) |
387125fc | 3659 | continue; |
e12c9621 | 3660 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3661 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3662 | continue; |
3663 | ||
4fc6441a | 3664 | write_dev_flush(dev); |
58efbc9f | 3665 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
3666 | } |
3667 | ||
3668 | /* wait for all the barriers */ | |
1538e6c5 | 3669 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3670 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3671 | continue; |
387125fc | 3672 | if (!dev->bdev) { |
5af3e8cc | 3673 | errors_wait++; |
387125fc CM |
3674 | continue; |
3675 | } | |
e12c9621 | 3676 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3677 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3678 | continue; |
3679 | ||
4fc6441a | 3680 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
3681 | if (ret) { |
3682 | dev->last_flush_error = ret; | |
66b4993e DS |
3683 | btrfs_dev_stat_inc_and_print(dev, |
3684 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 3685 | errors_wait++; |
401b41e5 AJ |
3686 | } |
3687 | } | |
3688 | ||
cea7c8bf | 3689 | if (errors_wait) { |
401b41e5 AJ |
3690 | /* |
3691 | * At some point we need the status of all disks | |
3692 | * to arrive at the volume status. So error checking | |
3693 | * is being pushed to a separate loop. | |
3694 | */ | |
d10b82fe | 3695 | return check_barrier_error(info); |
387125fc | 3696 | } |
387125fc CM |
3697 | return 0; |
3698 | } | |
3699 | ||
943c6e99 ZL |
3700 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3701 | { | |
8789f4fe ZL |
3702 | int raid_type; |
3703 | int min_tolerated = INT_MAX; | |
943c6e99 | 3704 | |
8789f4fe ZL |
3705 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3706 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 3707 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
3708 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
3709 | tolerated_failures); | |
943c6e99 | 3710 | |
8789f4fe ZL |
3711 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3712 | if (raid_type == BTRFS_RAID_SINGLE) | |
3713 | continue; | |
41a6e891 | 3714 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 3715 | continue; |
8c3e3582 | 3716 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
3717 | btrfs_raid_array[raid_type]. |
3718 | tolerated_failures); | |
3719 | } | |
943c6e99 | 3720 | |
8789f4fe | 3721 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 3722 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
3723 | min_tolerated = 0; |
3724 | } | |
3725 | ||
3726 | return min_tolerated; | |
943c6e99 ZL |
3727 | } |
3728 | ||
eece6a9c | 3729 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 3730 | { |
e5e9a520 | 3731 | struct list_head *head; |
f2984462 | 3732 | struct btrfs_device *dev; |
a061fc8d | 3733 | struct btrfs_super_block *sb; |
f2984462 | 3734 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3735 | int ret; |
3736 | int do_barriers; | |
a236aed1 CM |
3737 | int max_errors; |
3738 | int total_errors = 0; | |
a061fc8d | 3739 | u64 flags; |
f2984462 | 3740 | |
0b246afa | 3741 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
3742 | |
3743 | /* | |
3744 | * max_mirrors == 0 indicates we're from commit_transaction, | |
3745 | * not from fsync where the tree roots in fs_info have not | |
3746 | * been consistent on disk. | |
3747 | */ | |
3748 | if (max_mirrors == 0) | |
3749 | backup_super_roots(fs_info); | |
f2984462 | 3750 | |
0b246afa | 3751 | sb = fs_info->super_for_commit; |
a061fc8d | 3752 | dev_item = &sb->dev_item; |
e5e9a520 | 3753 | |
0b246afa JM |
3754 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
3755 | head = &fs_info->fs_devices->devices; | |
3756 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 3757 | |
5af3e8cc | 3758 | if (do_barriers) { |
0b246afa | 3759 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
3760 | if (ret) { |
3761 | mutex_unlock( | |
0b246afa JM |
3762 | &fs_info->fs_devices->device_list_mutex); |
3763 | btrfs_handle_fs_error(fs_info, ret, | |
3764 | "errors while submitting device barriers."); | |
5af3e8cc SB |
3765 | return ret; |
3766 | } | |
3767 | } | |
387125fc | 3768 | |
1538e6c5 | 3769 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3770 | if (!dev->bdev) { |
3771 | total_errors++; | |
3772 | continue; | |
3773 | } | |
e12c9621 | 3774 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3775 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3776 | continue; |
3777 | ||
2b82032c | 3778 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3779 | btrfs_set_stack_device_type(dev_item, dev->type); |
3780 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3781 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3782 | dev->commit_total_bytes); |
ce7213c7 MX |
3783 | btrfs_set_stack_device_bytes_used(dev_item, |
3784 | dev->commit_bytes_used); | |
a061fc8d CM |
3785 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3786 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3787 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3788 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
3789 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
3790 | BTRFS_FSID_SIZE); | |
a512bbf8 | 3791 | |
a061fc8d CM |
3792 | flags = btrfs_super_flags(sb); |
3793 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3794 | ||
75cb857d QW |
3795 | ret = btrfs_validate_write_super(fs_info, sb); |
3796 | if (ret < 0) { | |
3797 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
3798 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
3799 | "unexpected superblock corruption detected"); | |
3800 | return -EUCLEAN; | |
3801 | } | |
3802 | ||
abbb3b8e | 3803 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
3804 | if (ret) |
3805 | total_errors++; | |
f2984462 | 3806 | } |
a236aed1 | 3807 | if (total_errors > max_errors) { |
0b246afa JM |
3808 | btrfs_err(fs_info, "%d errors while writing supers", |
3809 | total_errors); | |
3810 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 3811 | |
9d565ba4 | 3812 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
3813 | btrfs_handle_fs_error(fs_info, -EIO, |
3814 | "%d errors while writing supers", | |
3815 | total_errors); | |
9d565ba4 | 3816 | return -EIO; |
a236aed1 | 3817 | } |
f2984462 | 3818 | |
a512bbf8 | 3819 | total_errors = 0; |
1538e6c5 | 3820 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3821 | if (!dev->bdev) |
3822 | continue; | |
e12c9621 | 3823 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3824 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3825 | continue; |
3826 | ||
abbb3b8e | 3827 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
3828 | if (ret) |
3829 | total_errors++; | |
f2984462 | 3830 | } |
0b246afa | 3831 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3832 | if (total_errors > max_errors) { |
0b246afa JM |
3833 | btrfs_handle_fs_error(fs_info, -EIO, |
3834 | "%d errors while writing supers", | |
3835 | total_errors); | |
79787eaa | 3836 | return -EIO; |
a236aed1 | 3837 | } |
f2984462 CM |
3838 | return 0; |
3839 | } | |
3840 | ||
cb517eab MX |
3841 | /* Drop a fs root from the radix tree and free it. */ |
3842 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3843 | struct btrfs_root *root) | |
2619ba1f | 3844 | { |
4df27c4d | 3845 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3846 | radix_tree_delete(&fs_info->fs_roots_radix, |
3847 | (unsigned long)root->root_key.objectid); | |
af01d2e5 JB |
3848 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
3849 | btrfs_put_fs_root(root); | |
4df27c4d | 3850 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3851 | |
3852 | if (btrfs_root_refs(&root->root_item) == 0) | |
3853 | synchronize_srcu(&fs_info->subvol_srcu); | |
3854 | ||
1c1ea4f7 | 3855 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3856 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3857 | if (root->reloc_root) { |
3858 | free_extent_buffer(root->reloc_root->node); | |
3859 | free_extent_buffer(root->reloc_root->commit_root); | |
3860 | btrfs_put_fs_root(root->reloc_root); | |
3861 | root->reloc_root = NULL; | |
3862 | } | |
3863 | } | |
3321719e | 3864 | |
faa2dbf0 JB |
3865 | if (root->free_ino_pinned) |
3866 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3867 | if (root->free_ino_ctl) | |
3868 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
84db5ccf | 3869 | btrfs_free_fs_root(root); |
4df27c4d YZ |
3870 | } |
3871 | ||
84db5ccf | 3872 | void btrfs_free_fs_root(struct btrfs_root *root) |
4df27c4d | 3873 | { |
57cdc8db | 3874 | iput(root->ino_cache_inode); |
4df27c4d | 3875 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3876 | if (root->anon_dev) |
3877 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3878 | if (root->subv_writers) |
3879 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3880 | free_extent_buffer(root->node); |
3881 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3882 | kfree(root->free_ino_ctl); |
3883 | kfree(root->free_ino_pinned); | |
b0feb9d9 | 3884 | btrfs_put_fs_root(root); |
2619ba1f CM |
3885 | } |
3886 | ||
c146afad | 3887 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3888 | { |
c146afad YZ |
3889 | u64 root_objectid = 0; |
3890 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3891 | int i = 0; |
3892 | int err = 0; | |
3893 | unsigned int ret = 0; | |
3894 | int index; | |
e089f05c | 3895 | |
c146afad | 3896 | while (1) { |
65d33fd7 | 3897 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3898 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3899 | (void **)gang, root_objectid, | |
3900 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3901 | if (!ret) { |
3902 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3903 | break; |
65d33fd7 | 3904 | } |
5d4f98a2 | 3905 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3906 | |
c146afad | 3907 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3908 | /* Avoid to grab roots in dead_roots */ |
3909 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3910 | gang[i] = NULL; | |
3911 | continue; | |
3912 | } | |
3913 | /* grab all the search result for later use */ | |
3914 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3915 | } | |
3916 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3917 | |
65d33fd7 QW |
3918 | for (i = 0; i < ret; i++) { |
3919 | if (!gang[i]) | |
3920 | continue; | |
c146afad | 3921 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3922 | err = btrfs_orphan_cleanup(gang[i]); |
3923 | if (err) | |
65d33fd7 QW |
3924 | break; |
3925 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3926 | } |
3927 | root_objectid++; | |
3928 | } | |
65d33fd7 QW |
3929 | |
3930 | /* release the uncleaned roots due to error */ | |
3931 | for (; i < ret; i++) { | |
3932 | if (gang[i]) | |
3933 | btrfs_put_fs_root(gang[i]); | |
3934 | } | |
3935 | return err; | |
c146afad | 3936 | } |
a2135011 | 3937 | |
6bccf3ab | 3938 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 3939 | { |
6bccf3ab | 3940 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 3941 | struct btrfs_trans_handle *trans; |
a74a4b97 | 3942 | |
0b246afa | 3943 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 3944 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
3945 | mutex_unlock(&fs_info->cleaner_mutex); |
3946 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
3947 | |
3948 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
3949 | down_write(&fs_info->cleanup_work_sem); |
3950 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 3951 | |
7a7eaa40 | 3952 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3953 | if (IS_ERR(trans)) |
3954 | return PTR_ERR(trans); | |
3a45bb20 | 3955 | return btrfs_commit_transaction(trans); |
c146afad YZ |
3956 | } |
3957 | ||
b105e927 | 3958 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 3959 | { |
c146afad YZ |
3960 | int ret; |
3961 | ||
afcdd129 | 3962 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
d6fd0ae2 OS |
3963 | /* |
3964 | * We don't want the cleaner to start new transactions, add more delayed | |
3965 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
3966 | * because that frees the task_struct, and the transaction kthread might | |
3967 | * still try to wake up the cleaner. | |
3968 | */ | |
3969 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 3970 | |
7343dd61 | 3971 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3972 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3973 | |
803b2f54 SB |
3974 | /* wait for the uuid_scan task to finish */ |
3975 | down(&fs_info->uuid_tree_rescan_sem); | |
3976 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3977 | up(&fs_info->uuid_tree_rescan_sem); | |
3978 | ||
837d5b6e | 3979 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3980 | btrfs_pause_balance(fs_info); |
837d5b6e | 3981 | |
8dabb742 SB |
3982 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3983 | ||
aa1b8cd4 | 3984 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3985 | |
3986 | /* wait for any defraggers to finish */ | |
3987 | wait_event(fs_info->transaction_wait, | |
3988 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3989 | ||
3990 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3991 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3992 | |
21c7e756 MX |
3993 | cancel_work_sync(&fs_info->async_reclaim_work); |
3994 | ||
b0643e59 DZ |
3995 | /* Cancel or finish ongoing discard work */ |
3996 | btrfs_discard_cleanup(fs_info); | |
3997 | ||
bc98a42c | 3998 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 3999 | /* |
d6fd0ae2 OS |
4000 | * The cleaner kthread is stopped, so do one final pass over |
4001 | * unused block groups. | |
e44163e1 | 4002 | */ |
0b246afa | 4003 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4004 | |
6bccf3ab | 4005 | ret = btrfs_commit_super(fs_info); |
acce952b | 4006 | if (ret) |
04892340 | 4007 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4008 | } |
4009 | ||
af722733 LB |
4010 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || |
4011 | test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) | |
2ff7e61e | 4012 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4013 | |
e3029d9f AV |
4014 | kthread_stop(fs_info->transaction_kthread); |
4015 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4016 | |
e187831e | 4017 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4018 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4019 | |
04892340 | 4020 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4021 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4022 | |
963d678b | 4023 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4024 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4025 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4026 | } |
bcc63abb | 4027 | |
4297ff84 JB |
4028 | if (percpu_counter_sum(&fs_info->dio_bytes)) |
4029 | btrfs_info(fs_info, "at unmount dio bytes count %lld", | |
4030 | percpu_counter_sum(&fs_info->dio_bytes)); | |
4031 | ||
6618a59b | 4032 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4033 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4034 | |
faa2dbf0 | 4035 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 4036 | |
1a4319cc LB |
4037 | btrfs_put_block_group_cache(fs_info); |
4038 | ||
de348ee0 WS |
4039 | /* |
4040 | * we must make sure there is not any read request to | |
4041 | * submit after we stopping all workers. | |
4042 | */ | |
4043 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4044 | btrfs_stop_all_workers(fs_info); |
4045 | ||
afcdd129 | 4046 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4047 | free_root_pointers(fs_info, true); |
9ad6b7bc | 4048 | |
4e19443d JB |
4049 | /* |
4050 | * We must free the block groups after dropping the fs_roots as we could | |
4051 | * have had an IO error and have left over tree log blocks that aren't | |
4052 | * cleaned up until the fs roots are freed. This makes the block group | |
4053 | * accounting appear to be wrong because there's pending reserved bytes, | |
4054 | * so make sure we do the block group cleanup afterwards. | |
4055 | */ | |
4056 | btrfs_free_block_groups(fs_info); | |
4057 | ||
13e6c37b | 4058 | iput(fs_info->btree_inode); |
d6bfde87 | 4059 | |
21adbd5c | 4060 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4061 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4062 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4063 | #endif |
4064 | ||
0b86a832 | 4065 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4066 | btrfs_close_devices(fs_info->fs_devices); |
b248a415 | 4067 | |
e2d84521 | 4068 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 4069 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
4297ff84 | 4070 | percpu_counter_destroy(&fs_info->dio_bytes); |
7f8d236a | 4071 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
76dda93c | 4072 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 4073 | |
bfcea1c6 | 4074 | btrfs_free_csum_hash(fs_info); |
53b381b3 | 4075 | btrfs_free_stripe_hash_table(fs_info); |
fd708b81 | 4076 | btrfs_free_ref_cache(fs_info); |
eb60ceac CM |
4077 | } |
4078 | ||
b9fab919 CM |
4079 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4080 | int atomic) | |
5f39d397 | 4081 | { |
1259ab75 | 4082 | int ret; |
727011e0 | 4083 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4084 | |
0b32f4bb | 4085 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4086 | if (!ret) |
4087 | return ret; | |
4088 | ||
4089 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4090 | parent_transid, atomic); |
4091 | if (ret == -EAGAIN) | |
4092 | return ret; | |
1259ab75 | 4093 | return !ret; |
5f39d397 CM |
4094 | } |
4095 | ||
5f39d397 CM |
4096 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4097 | { | |
0b246afa | 4098 | struct btrfs_fs_info *fs_info; |
06ea65a3 | 4099 | struct btrfs_root *root; |
5f39d397 | 4100 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4101 | int was_dirty; |
b4ce94de | 4102 | |
06ea65a3 JB |
4103 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4104 | /* | |
4105 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4106 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4107 | * outside of the sanity tests. |
4108 | */ | |
b0132a3b | 4109 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4110 | return; |
4111 | #endif | |
4112 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
0b246afa | 4113 | fs_info = root->fs_info; |
b9447ef8 | 4114 | btrfs_assert_tree_locked(buf); |
0b246afa | 4115 | if (transid != fs_info->generation) |
5d163e0e | 4116 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4117 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4118 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4119 | if (!was_dirty) |
104b4e51 NB |
4120 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4121 | buf->len, | |
4122 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4123 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4124 | /* |
4125 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4126 | * but item data not updated. | |
4127 | * So here we should only check item pointers, not item data. | |
4128 | */ | |
4129 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4130 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4131 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4132 | ASSERT(0); |
4133 | } | |
4134 | #endif | |
eb60ceac CM |
4135 | } |
4136 | ||
2ff7e61e | 4137 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4138 | int flush_delayed) |
16cdcec7 MX |
4139 | { |
4140 | /* | |
4141 | * looks as though older kernels can get into trouble with | |
4142 | * this code, they end up stuck in balance_dirty_pages forever | |
4143 | */ | |
e2d84521 | 4144 | int ret; |
16cdcec7 MX |
4145 | |
4146 | if (current->flags & PF_MEMALLOC) | |
4147 | return; | |
4148 | ||
b53d3f5d | 4149 | if (flush_delayed) |
2ff7e61e | 4150 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4151 | |
d814a491 EL |
4152 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4153 | BTRFS_DIRTY_METADATA_THRESH, | |
4154 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4155 | if (ret > 0) { |
0b246afa | 4156 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4157 | } |
16cdcec7 MX |
4158 | } |
4159 | ||
2ff7e61e | 4160 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4161 | { |
2ff7e61e | 4162 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4163 | } |
585ad2c3 | 4164 | |
2ff7e61e | 4165 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4166 | { |
2ff7e61e | 4167 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4168 | } |
6b80053d | 4169 | |
581c1760 QW |
4170 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4171 | struct btrfs_key *first_key) | |
6b80053d | 4172 | { |
5ab12d1f | 4173 | return btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 4174 | level, first_key); |
6b80053d | 4175 | } |
0da5468f | 4176 | |
2ff7e61e | 4177 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4178 | { |
fe816d0f NB |
4179 | /* cleanup FS via transaction */ |
4180 | btrfs_cleanup_transaction(fs_info); | |
4181 | ||
0b246afa | 4182 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4183 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4184 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4185 | |
0b246afa JM |
4186 | down_write(&fs_info->cleanup_work_sem); |
4187 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4188 | } |
4189 | ||
143bede5 | 4190 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4191 | { |
acce952b | 4192 | struct btrfs_ordered_extent *ordered; |
acce952b | 4193 | |
199c2a9c | 4194 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4195 | /* |
4196 | * This will just short circuit the ordered completion stuff which will | |
4197 | * make sure the ordered extent gets properly cleaned up. | |
4198 | */ | |
199c2a9c | 4199 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4200 | root_extent_list) |
4201 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4202 | spin_unlock(&root->ordered_extent_lock); |
4203 | } | |
4204 | ||
4205 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4206 | { | |
4207 | struct btrfs_root *root; | |
4208 | struct list_head splice; | |
4209 | ||
4210 | INIT_LIST_HEAD(&splice); | |
4211 | ||
4212 | spin_lock(&fs_info->ordered_root_lock); | |
4213 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4214 | while (!list_empty(&splice)) { | |
4215 | root = list_first_entry(&splice, struct btrfs_root, | |
4216 | ordered_root); | |
1de2cfde JB |
4217 | list_move_tail(&root->ordered_root, |
4218 | &fs_info->ordered_roots); | |
199c2a9c | 4219 | |
2a85d9ca | 4220 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4221 | btrfs_destroy_ordered_extents(root); |
4222 | ||
2a85d9ca LB |
4223 | cond_resched(); |
4224 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4225 | } |
4226 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4227 | |
4228 | /* | |
4229 | * We need this here because if we've been flipped read-only we won't | |
4230 | * get sync() from the umount, so we need to make sure any ordered | |
4231 | * extents that haven't had their dirty pages IO start writeout yet | |
4232 | * actually get run and error out properly. | |
4233 | */ | |
4234 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4235 | } |
4236 | ||
35a3621b | 4237 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4238 | struct btrfs_fs_info *fs_info) |
acce952b | 4239 | { |
4240 | struct rb_node *node; | |
4241 | struct btrfs_delayed_ref_root *delayed_refs; | |
4242 | struct btrfs_delayed_ref_node *ref; | |
4243 | int ret = 0; | |
4244 | ||
4245 | delayed_refs = &trans->delayed_refs; | |
4246 | ||
4247 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4248 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4249 | spin_unlock(&delayed_refs->lock); |
0b246afa | 4250 | btrfs_info(fs_info, "delayed_refs has NO entry"); |
acce952b | 4251 | return ret; |
4252 | } | |
4253 | ||
5c9d028b | 4254 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4255 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4256 | struct rb_node *n; |
e78417d1 | 4257 | bool pin_bytes = false; |
acce952b | 4258 | |
d7df2c79 JB |
4259 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4260 | href_node); | |
3069bd26 | 4261 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4262 | continue; |
3069bd26 | 4263 | |
d7df2c79 | 4264 | spin_lock(&head->lock); |
e3d03965 | 4265 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4266 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4267 | ref_node); | |
d7df2c79 | 4268 | ref->in_tree = 0; |
e3d03965 | 4269 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4270 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4271 | if (!list_empty(&ref->add_list)) |
4272 | list_del(&ref->add_list); | |
d7df2c79 JB |
4273 | atomic_dec(&delayed_refs->num_entries); |
4274 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4275 | } |
d7df2c79 JB |
4276 | if (head->must_insert_reserved) |
4277 | pin_bytes = true; | |
4278 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4279 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4280 | spin_unlock(&head->lock); |
4281 | spin_unlock(&delayed_refs->lock); | |
4282 | mutex_unlock(&head->mutex); | |
acce952b | 4283 | |
d7df2c79 | 4284 | if (pin_bytes) |
d278850e JB |
4285 | btrfs_pin_extent(fs_info, head->bytenr, |
4286 | head->num_bytes, 1); | |
31890da0 | 4287 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4288 | btrfs_put_delayed_ref_head(head); |
acce952b | 4289 | cond_resched(); |
4290 | spin_lock(&delayed_refs->lock); | |
4291 | } | |
81f7eb00 | 4292 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4293 | |
4294 | spin_unlock(&delayed_refs->lock); | |
4295 | ||
4296 | return ret; | |
4297 | } | |
4298 | ||
143bede5 | 4299 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4300 | { |
4301 | struct btrfs_inode *btrfs_inode; | |
4302 | struct list_head splice; | |
4303 | ||
4304 | INIT_LIST_HEAD(&splice); | |
4305 | ||
eb73c1b7 MX |
4306 | spin_lock(&root->delalloc_lock); |
4307 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4308 | |
4309 | while (!list_empty(&splice)) { | |
fe816d0f | 4310 | struct inode *inode = NULL; |
eb73c1b7 MX |
4311 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4312 | delalloc_inodes); | |
fe816d0f | 4313 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4314 | spin_unlock(&root->delalloc_lock); |
acce952b | 4315 | |
fe816d0f NB |
4316 | /* |
4317 | * Make sure we get a live inode and that it'll not disappear | |
4318 | * meanwhile. | |
4319 | */ | |
4320 | inode = igrab(&btrfs_inode->vfs_inode); | |
4321 | if (inode) { | |
4322 | invalidate_inode_pages2(inode->i_mapping); | |
4323 | iput(inode); | |
4324 | } | |
eb73c1b7 | 4325 | spin_lock(&root->delalloc_lock); |
acce952b | 4326 | } |
eb73c1b7 MX |
4327 | spin_unlock(&root->delalloc_lock); |
4328 | } | |
4329 | ||
4330 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4331 | { | |
4332 | struct btrfs_root *root; | |
4333 | struct list_head splice; | |
4334 | ||
4335 | INIT_LIST_HEAD(&splice); | |
4336 | ||
4337 | spin_lock(&fs_info->delalloc_root_lock); | |
4338 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4339 | while (!list_empty(&splice)) { | |
4340 | root = list_first_entry(&splice, struct btrfs_root, | |
4341 | delalloc_root); | |
eb73c1b7 MX |
4342 | root = btrfs_grab_fs_root(root); |
4343 | BUG_ON(!root); | |
4344 | spin_unlock(&fs_info->delalloc_root_lock); | |
4345 | ||
4346 | btrfs_destroy_delalloc_inodes(root); | |
4347 | btrfs_put_fs_root(root); | |
4348 | ||
4349 | spin_lock(&fs_info->delalloc_root_lock); | |
4350 | } | |
4351 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4352 | } |
4353 | ||
2ff7e61e | 4354 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4355 | struct extent_io_tree *dirty_pages, |
4356 | int mark) | |
4357 | { | |
4358 | int ret; | |
acce952b | 4359 | struct extent_buffer *eb; |
4360 | u64 start = 0; | |
4361 | u64 end; | |
acce952b | 4362 | |
4363 | while (1) { | |
4364 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4365 | mark, NULL); |
acce952b | 4366 | if (ret) |
4367 | break; | |
4368 | ||
91166212 | 4369 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4370 | while (start <= end) { |
0b246afa JM |
4371 | eb = find_extent_buffer(fs_info, start); |
4372 | start += fs_info->nodesize; | |
fd8b2b61 | 4373 | if (!eb) |
acce952b | 4374 | continue; |
fd8b2b61 | 4375 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4376 | |
fd8b2b61 JB |
4377 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4378 | &eb->bflags)) | |
4379 | clear_extent_buffer_dirty(eb); | |
4380 | free_extent_buffer_stale(eb); | |
acce952b | 4381 | } |
4382 | } | |
4383 | ||
4384 | return ret; | |
4385 | } | |
4386 | ||
2ff7e61e | 4387 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 4388 | struct extent_io_tree *pinned_extents) |
4389 | { | |
4390 | struct extent_io_tree *unpin; | |
4391 | u64 start; | |
4392 | u64 end; | |
4393 | int ret; | |
ed0eaa14 | 4394 | bool loop = true; |
acce952b | 4395 | |
4396 | unpin = pinned_extents; | |
ed0eaa14 | 4397 | again: |
acce952b | 4398 | while (1) { |
0e6ec385 FM |
4399 | struct extent_state *cached_state = NULL; |
4400 | ||
fcd5e742 LF |
4401 | /* |
4402 | * The btrfs_finish_extent_commit() may get the same range as | |
4403 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4404 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4405 | * the same extent range. | |
4406 | */ | |
4407 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4408 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 4409 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
4410 | if (ret) { |
4411 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4412 | break; |
fcd5e742 | 4413 | } |
acce952b | 4414 | |
0e6ec385 FM |
4415 | clear_extent_dirty(unpin, start, end, &cached_state); |
4416 | free_extent_state(cached_state); | |
2ff7e61e | 4417 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4418 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4419 | cond_resched(); |
4420 | } | |
4421 | ||
ed0eaa14 | 4422 | if (loop) { |
0b246afa JM |
4423 | if (unpin == &fs_info->freed_extents[0]) |
4424 | unpin = &fs_info->freed_extents[1]; | |
ed0eaa14 | 4425 | else |
0b246afa | 4426 | unpin = &fs_info->freed_extents[0]; |
ed0eaa14 LB |
4427 | loop = false; |
4428 | goto again; | |
4429 | } | |
4430 | ||
acce952b | 4431 | return 0; |
4432 | } | |
4433 | ||
32da5386 | 4434 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4435 | { |
4436 | struct inode *inode; | |
4437 | ||
4438 | inode = cache->io_ctl.inode; | |
4439 | if (inode) { | |
4440 | invalidate_inode_pages2(inode->i_mapping); | |
4441 | BTRFS_I(inode)->generation = 0; | |
4442 | cache->io_ctl.inode = NULL; | |
4443 | iput(inode); | |
4444 | } | |
4445 | btrfs_put_block_group(cache); | |
4446 | } | |
4447 | ||
4448 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4449 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4450 | { |
32da5386 | 4451 | struct btrfs_block_group *cache; |
c79a1751 LB |
4452 | |
4453 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4454 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4455 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4456 | struct btrfs_block_group, |
c79a1751 | 4457 | dirty_list); |
c79a1751 LB |
4458 | |
4459 | if (!list_empty(&cache->io_list)) { | |
4460 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4461 | list_del_init(&cache->io_list); | |
4462 | btrfs_cleanup_bg_io(cache); | |
4463 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4464 | } | |
4465 | ||
4466 | list_del_init(&cache->dirty_list); | |
4467 | spin_lock(&cache->lock); | |
4468 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4469 | spin_unlock(&cache->lock); | |
4470 | ||
4471 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4472 | btrfs_put_block_group(cache); | |
ba2c4d4e | 4473 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
4474 | spin_lock(&cur_trans->dirty_bgs_lock); |
4475 | } | |
4476 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4477 | ||
45ae2c18 NB |
4478 | /* |
4479 | * Refer to the definition of io_bgs member for details why it's safe | |
4480 | * to use it without any locking | |
4481 | */ | |
c79a1751 LB |
4482 | while (!list_empty(&cur_trans->io_bgs)) { |
4483 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4484 | struct btrfs_block_group, |
c79a1751 | 4485 | io_list); |
c79a1751 LB |
4486 | |
4487 | list_del_init(&cache->io_list); | |
4488 | spin_lock(&cache->lock); | |
4489 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4490 | spin_unlock(&cache->lock); | |
4491 | btrfs_cleanup_bg_io(cache); | |
4492 | } | |
4493 | } | |
4494 | ||
49b25e05 | 4495 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4496 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4497 | { |
bbbf7243 NB |
4498 | struct btrfs_device *dev, *tmp; |
4499 | ||
2ff7e61e | 4500 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4501 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4502 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4503 | ||
bbbf7243 NB |
4504 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4505 | post_commit_list) { | |
4506 | list_del_init(&dev->post_commit_list); | |
4507 | } | |
4508 | ||
2ff7e61e | 4509 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4510 | |
4a9d8bde | 4511 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4512 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4513 | |
4a9d8bde | 4514 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4515 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4516 | |
ccdf9b30 | 4517 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4518 | |
2ff7e61e | 4519 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4520 | EXTENT_DIRTY); |
2ff7e61e | 4521 | btrfs_destroy_pinned_extent(fs_info, |
0b246afa | 4522 | fs_info->pinned_extents); |
49b25e05 | 4523 | |
4a9d8bde MX |
4524 | cur_trans->state =TRANS_STATE_COMPLETED; |
4525 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4526 | } |
4527 | ||
2ff7e61e | 4528 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4529 | { |
4530 | struct btrfs_transaction *t; | |
acce952b | 4531 | |
0b246afa | 4532 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4533 | |
0b246afa JM |
4534 | spin_lock(&fs_info->trans_lock); |
4535 | while (!list_empty(&fs_info->trans_list)) { | |
4536 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4537 | struct btrfs_transaction, list); |
4538 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4539 | refcount_inc(&t->use_count); |
0b246afa | 4540 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4541 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4542 | btrfs_put_transaction(t); |
0b246afa | 4543 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4544 | continue; |
4545 | } | |
0b246afa | 4546 | if (t == fs_info->running_transaction) { |
724e2315 | 4547 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4548 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4549 | /* |
4550 | * We wait for 0 num_writers since we don't hold a trans | |
4551 | * handle open currently for this transaction. | |
4552 | */ | |
4553 | wait_event(t->writer_wait, | |
4554 | atomic_read(&t->num_writers) == 0); | |
4555 | } else { | |
0b246afa | 4556 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4557 | } |
2ff7e61e | 4558 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4559 | |
0b246afa JM |
4560 | spin_lock(&fs_info->trans_lock); |
4561 | if (t == fs_info->running_transaction) | |
4562 | fs_info->running_transaction = NULL; | |
acce952b | 4563 | list_del_init(&t->list); |
0b246afa | 4564 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4565 | |
724e2315 | 4566 | btrfs_put_transaction(t); |
2ff7e61e | 4567 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4568 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4569 | } |
0b246afa JM |
4570 | spin_unlock(&fs_info->trans_lock); |
4571 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4572 | btrfs_destroy_delayed_inodes(fs_info); |
4573 | btrfs_assert_delayed_root_empty(fs_info); | |
2ff7e61e | 4574 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
0b246afa JM |
4575 | btrfs_destroy_all_delalloc_inodes(fs_info); |
4576 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
acce952b | 4577 | |
4578 | return 0; | |
4579 | } | |
4580 | ||
e8c9f186 | 4581 | static const struct extent_io_ops btree_extent_io_ops = { |
4d53dddb | 4582 | /* mandatory callbacks */ |
0b86a832 | 4583 | .submit_bio_hook = btree_submit_bio_hook, |
4d53dddb | 4584 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0da5468f | 4585 | }; |