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> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
0b86a832 | 26 | #include "volumes.h" |
db94535d | 27 | #include "print-tree.h" |
925baedd | 28 | #include "locking.h" |
e02119d5 | 29 | #include "tree-log.h" |
fa9c0d79 | 30 | #include "free-space-cache.h" |
70f6d82e | 31 | #include "free-space-tree.h" |
581bb050 | 32 | #include "inode-map.h" |
21adbd5c | 33 | #include "check-integrity.h" |
606686ee | 34 | #include "rcu-string.h" |
8dabb742 | 35 | #include "dev-replace.h" |
53b381b3 | 36 | #include "raid56.h" |
5ac1d209 | 37 | #include "sysfs.h" |
fcebe456 | 38 | #include "qgroup.h" |
ebb8765b | 39 | #include "compression.h" |
557ea5dd | 40 | #include "tree-checker.h" |
fd708b81 | 41 | #include "ref-verify.h" |
aac0023c | 42 | #include "block-group.h" |
b0643e59 | 43 | #include "discard.h" |
f603bb94 | 44 | #include "space-info.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 | ||
141386e1 JB |
101 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
102 | { | |
103 | if (fs_info->csum_shash) | |
104 | crypto_free_shash(fs_info->csum_shash); | |
105 | } | |
106 | ||
d352ac68 CM |
107 | /* |
108 | * async submit bios are used to offload expensive checksumming | |
109 | * onto the worker threads. They checksum file and metadata bios | |
110 | * just before they are sent down the IO stack. | |
111 | */ | |
44b8bd7e | 112 | struct async_submit_bio { |
c6100a4b | 113 | void *private_data; |
44b8bd7e | 114 | struct bio *bio; |
a758781d | 115 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 116 | int mirror_num; |
eaf25d93 CM |
117 | /* |
118 | * bio_offset is optional, can be used if the pages in the bio | |
119 | * can't tell us where in the file the bio should go | |
120 | */ | |
121 | u64 bio_offset; | |
8b712842 | 122 | struct btrfs_work work; |
4e4cbee9 | 123 | blk_status_t status; |
44b8bd7e CM |
124 | }; |
125 | ||
85d4e461 CM |
126 | /* |
127 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
128 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
129 | * the level the eb occupies in the tree. | |
130 | * | |
131 | * Different roots are used for different purposes and may nest inside each | |
132 | * other and they require separate keysets. As lockdep keys should be | |
133 | * static, assign keysets according to the purpose of the root as indicated | |
4fd786e6 MT |
134 | * by btrfs_root->root_key.objectid. This ensures that all special purpose |
135 | * roots have separate keysets. | |
4008c04a | 136 | * |
85d4e461 CM |
137 | * Lock-nesting across peer nodes is always done with the immediate parent |
138 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
139 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 140 | * |
85d4e461 CM |
141 | * The key is set by the readpage_end_io_hook after the buffer has passed |
142 | * csum validation but before the pages are unlocked. It is also set by | |
143 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 144 | * |
85d4e461 CM |
145 | * We also add a check to make sure the highest level of the tree is the |
146 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
147 | * needs update as well. | |
4008c04a CM |
148 | */ |
149 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
150 | # if BTRFS_MAX_LEVEL != 8 | |
151 | # error | |
152 | # endif | |
85d4e461 CM |
153 | |
154 | static struct btrfs_lockdep_keyset { | |
155 | u64 id; /* root objectid */ | |
156 | const char *name_stem; /* lock name stem */ | |
157 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
158 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
159 | } btrfs_lockdep_keysets[] = { | |
160 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
161 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
162 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
163 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
164 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
165 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 166 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
167 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
168 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
169 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 170 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 171 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 172 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 173 | }; |
85d4e461 CM |
174 | |
175 | void __init btrfs_init_lockdep(void) | |
176 | { | |
177 | int i, j; | |
178 | ||
179 | /* initialize lockdep class names */ | |
180 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
181 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
182 | ||
183 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
184 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
185 | "btrfs-%s-%02d", ks->name_stem, j); | |
186 | } | |
187 | } | |
188 | ||
189 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
190 | int level) | |
191 | { | |
192 | struct btrfs_lockdep_keyset *ks; | |
193 | ||
194 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
195 | ||
196 | /* find the matching keyset, id 0 is the default entry */ | |
197 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
198 | if (ks->id == objectid) | |
199 | break; | |
200 | ||
201 | lockdep_set_class_and_name(&eb->lock, | |
202 | &ks->keys[level], ks->names[level]); | |
203 | } | |
204 | ||
4008c04a CM |
205 | #endif |
206 | ||
d352ac68 CM |
207 | /* |
208 | * extents on the btree inode are pretty simple, there's one extent | |
209 | * that covers the entire device | |
210 | */ | |
6af49dbd | 211 | struct extent_map *btree_get_extent(struct btrfs_inode *inode, |
39b07b5d OS |
212 | struct page *page, size_t pg_offset, |
213 | u64 start, u64 len) | |
7eccb903 | 214 | { |
fc4f21b1 | 215 | struct extent_map_tree *em_tree = &inode->extent_tree; |
5f39d397 CM |
216 | struct extent_map *em; |
217 | int ret; | |
218 | ||
890871be | 219 | read_lock(&em_tree->lock); |
d1310b2e | 220 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d | 221 | if (em) { |
890871be | 222 | read_unlock(&em_tree->lock); |
5f39d397 | 223 | goto out; |
a061fc8d | 224 | } |
890871be | 225 | read_unlock(&em_tree->lock); |
7b13b7b1 | 226 | |
172ddd60 | 227 | em = alloc_extent_map(); |
5f39d397 CM |
228 | if (!em) { |
229 | em = ERR_PTR(-ENOMEM); | |
230 | goto out; | |
231 | } | |
232 | em->start = 0; | |
0afbaf8c | 233 | em->len = (u64)-1; |
c8b97818 | 234 | em->block_len = (u64)-1; |
5f39d397 | 235 | em->block_start = 0; |
d1310b2e | 236 | |
890871be | 237 | write_lock(&em_tree->lock); |
09a2a8f9 | 238 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
239 | if (ret == -EEXIST) { |
240 | free_extent_map(em); | |
7b13b7b1 | 241 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 242 | if (!em) |
0433f20d | 243 | em = ERR_PTR(-EIO); |
5f39d397 | 244 | } else if (ret) { |
7b13b7b1 | 245 | free_extent_map(em); |
0433f20d | 246 | em = ERR_PTR(ret); |
5f39d397 | 247 | } |
890871be | 248 | write_unlock(&em_tree->lock); |
7b13b7b1 | 249 | |
5f39d397 CM |
250 | out: |
251 | return em; | |
7eccb903 CM |
252 | } |
253 | ||
d352ac68 | 254 | /* |
2996e1f8 | 255 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 256 | */ |
c67b3892 | 257 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 258 | { |
d5178578 | 259 | struct btrfs_fs_info *fs_info = buf->fs_info; |
e9be5a30 | 260 | const int num_pages = fs_info->nodesize >> PAGE_SHIFT; |
d5178578 | 261 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 262 | char *kaddr; |
e9be5a30 | 263 | int i; |
d5178578 JT |
264 | |
265 | shash->tfm = fs_info->csum_shash; | |
266 | crypto_shash_init(shash); | |
e9be5a30 DS |
267 | kaddr = page_address(buf->pages[0]); |
268 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, | |
269 | PAGE_SIZE - BTRFS_CSUM_SIZE); | |
19c00ddc | 270 | |
e9be5a30 DS |
271 | for (i = 1; i < num_pages; i++) { |
272 | kaddr = page_address(buf->pages[i]); | |
273 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 274 | } |
71a63551 | 275 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 276 | crypto_shash_final(shash, result); |
19c00ddc CM |
277 | } |
278 | ||
d352ac68 CM |
279 | /* |
280 | * we can't consider a given block up to date unless the transid of the | |
281 | * block matches the transid in the parent node's pointer. This is how we | |
282 | * detect blocks that either didn't get written at all or got written | |
283 | * in the wrong place. | |
284 | */ | |
1259ab75 | 285 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
286 | struct extent_buffer *eb, u64 parent_transid, |
287 | int atomic) | |
1259ab75 | 288 | { |
2ac55d41 | 289 | struct extent_state *cached_state = NULL; |
1259ab75 | 290 | int ret; |
2755a0de | 291 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
292 | |
293 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
294 | return 0; | |
295 | ||
b9fab919 CM |
296 | if (atomic) |
297 | return -EAGAIN; | |
298 | ||
a26e8c9f JB |
299 | if (need_lock) { |
300 | btrfs_tree_read_lock(eb); | |
300aa896 | 301 | btrfs_set_lock_blocking_read(eb); |
a26e8c9f JB |
302 | } |
303 | ||
2ac55d41 | 304 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 305 | &cached_state); |
0b32f4bb | 306 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
307 | btrfs_header_generation(eb) == parent_transid) { |
308 | ret = 0; | |
309 | goto out; | |
310 | } | |
94647322 DS |
311 | btrfs_err_rl(eb->fs_info, |
312 | "parent transid verify failed on %llu wanted %llu found %llu", | |
313 | eb->start, | |
29549aec | 314 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 315 | ret = 1; |
a26e8c9f JB |
316 | |
317 | /* | |
318 | * Things reading via commit roots that don't have normal protection, | |
319 | * like send, can have a really old block in cache that may point at a | |
01327610 | 320 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
321 | * if we find an eb that is under IO (dirty/writeback) because we could |
322 | * end up reading in the stale data and then writing it back out and | |
323 | * making everybody very sad. | |
324 | */ | |
325 | if (!extent_buffer_under_io(eb)) | |
326 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 327 | out: |
2ac55d41 | 328 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 329 | &cached_state); |
472b909f JB |
330 | if (need_lock) |
331 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 332 | return ret; |
1259ab75 CM |
333 | } |
334 | ||
e7e16f48 JT |
335 | static bool btrfs_supported_super_csum(u16 csum_type) |
336 | { | |
337 | switch (csum_type) { | |
338 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 339 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 340 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 341 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
342 | return true; |
343 | default: | |
344 | return false; | |
345 | } | |
346 | } | |
347 | ||
1104a885 DS |
348 | /* |
349 | * Return 0 if the superblock checksum type matches the checksum value of that | |
350 | * algorithm. Pass the raw disk superblock data. | |
351 | */ | |
ab8d0fc4 JM |
352 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
353 | char *raw_disk_sb) | |
1104a885 DS |
354 | { |
355 | struct btrfs_super_block *disk_sb = | |
356 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 357 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
358 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
359 | ||
360 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 361 | |
51bce6c9 JT |
362 | /* |
363 | * The super_block structure does not span the whole | |
364 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
365 | * filled with zeros and is included in the checksum. | |
366 | */ | |
fd08001f EB |
367 | crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
368 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); | |
1104a885 | 369 | |
51bce6c9 JT |
370 | if (memcmp(disk_sb->csum, result, btrfs_super_csum_size(disk_sb))) |
371 | return 1; | |
1104a885 | 372 | |
e7e16f48 | 373 | return 0; |
1104a885 DS |
374 | } |
375 | ||
e064d5e9 | 376 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 377 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 378 | { |
e064d5e9 | 379 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
380 | int found_level; |
381 | struct btrfs_key found_key; | |
382 | int ret; | |
383 | ||
384 | found_level = btrfs_header_level(eb); | |
385 | if (found_level != level) { | |
63489055 QW |
386 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
387 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
388 | btrfs_err(fs_info, |
389 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
390 | eb->start, level, found_level); | |
581c1760 QW |
391 | return -EIO; |
392 | } | |
393 | ||
394 | if (!first_key) | |
395 | return 0; | |
396 | ||
5d41be6f QW |
397 | /* |
398 | * For live tree block (new tree blocks in current transaction), | |
399 | * we need proper lock context to avoid race, which is impossible here. | |
400 | * So we only checks tree blocks which is read from disk, whose | |
401 | * generation <= fs_info->last_trans_committed. | |
402 | */ | |
403 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
404 | return 0; | |
62fdaa52 QW |
405 | |
406 | /* We have @first_key, so this @eb must have at least one item */ | |
407 | if (btrfs_header_nritems(eb) == 0) { | |
408 | btrfs_err(fs_info, | |
409 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
410 | eb->start); | |
411 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
412 | return -EUCLEAN; | |
413 | } | |
414 | ||
581c1760 QW |
415 | if (found_level) |
416 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
417 | else | |
418 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
419 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
420 | ||
581c1760 | 421 | if (ret) { |
63489055 QW |
422 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
423 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 424 | btrfs_err(fs_info, |
ff76a864 LB |
425 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
426 | eb->start, parent_transid, first_key->objectid, | |
427 | first_key->type, first_key->offset, | |
428 | found_key.objectid, found_key.type, | |
429 | found_key.offset); | |
581c1760 | 430 | } |
581c1760 QW |
431 | return ret; |
432 | } | |
433 | ||
d352ac68 CM |
434 | /* |
435 | * helper to read a given tree block, doing retries as required when | |
436 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
437 | * |
438 | * @parent_transid: expected transid, skip check if 0 | |
439 | * @level: expected level, mandatory check | |
440 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 441 | */ |
5ab12d1f | 442 | static int btree_read_extent_buffer_pages(struct extent_buffer *eb, |
581c1760 QW |
443 | u64 parent_transid, int level, |
444 | struct btrfs_key *first_key) | |
f188591e | 445 | { |
5ab12d1f | 446 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 447 | struct extent_io_tree *io_tree; |
ea466794 | 448 | int failed = 0; |
f188591e CM |
449 | int ret; |
450 | int num_copies = 0; | |
451 | int mirror_num = 0; | |
ea466794 | 452 | int failed_mirror = 0; |
f188591e | 453 | |
0b246afa | 454 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 455 | while (1) { |
f8397d69 | 456 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 457 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 458 | if (!ret) { |
581c1760 | 459 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 460 | parent_transid, 0)) |
256dd1bb | 461 | ret = -EIO; |
e064d5e9 | 462 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 463 | first_key, parent_transid)) |
581c1760 QW |
464 | ret = -EUCLEAN; |
465 | else | |
466 | break; | |
256dd1bb | 467 | } |
d397712b | 468 | |
0b246afa | 469 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 470 | eb->start, eb->len); |
4235298e | 471 | if (num_copies == 1) |
ea466794 | 472 | break; |
4235298e | 473 | |
5cf1ab56 JB |
474 | if (!failed_mirror) { |
475 | failed = 1; | |
476 | failed_mirror = eb->read_mirror; | |
477 | } | |
478 | ||
f188591e | 479 | mirror_num++; |
ea466794 JB |
480 | if (mirror_num == failed_mirror) |
481 | mirror_num++; | |
482 | ||
4235298e | 483 | if (mirror_num > num_copies) |
ea466794 | 484 | break; |
f188591e | 485 | } |
ea466794 | 486 | |
c0901581 | 487 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 488 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
489 | |
490 | return ret; | |
f188591e | 491 | } |
19c00ddc | 492 | |
d352ac68 | 493 | /* |
d397712b CM |
494 | * checksum a dirty tree block before IO. This has extra checks to make sure |
495 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 496 | */ |
d397712b | 497 | |
01d58472 | 498 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 499 | { |
4eee4fa4 | 500 | u64 start = page_offset(page); |
19c00ddc | 501 | u64 found_start; |
2996e1f8 JT |
502 | u8 result[BTRFS_CSUM_SIZE]; |
503 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
19c00ddc | 504 | struct extent_buffer *eb; |
8d47a0d8 | 505 | int ret; |
f188591e | 506 | |
4f2de97a JB |
507 | eb = (struct extent_buffer *)page->private; |
508 | if (page != eb->pages[0]) | |
509 | return 0; | |
0f805531 | 510 | |
19c00ddc | 511 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
512 | /* |
513 | * Please do not consolidate these warnings into a single if. | |
514 | * It is useful to know what went wrong. | |
515 | */ | |
516 | if (WARN_ON(found_start != start)) | |
517 | return -EUCLEAN; | |
518 | if (WARN_ON(!PageUptodate(page))) | |
519 | return -EUCLEAN; | |
520 | ||
de37aa51 | 521 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, |
9a8658e3 DS |
522 | offsetof(struct btrfs_header, fsid), |
523 | BTRFS_FSID_SIZE) == 0); | |
0f805531 | 524 | |
c67b3892 | 525 | csum_tree_block(eb, result); |
2996e1f8 | 526 | |
8d47a0d8 QW |
527 | if (btrfs_header_level(eb)) |
528 | ret = btrfs_check_node(eb); | |
529 | else | |
530 | ret = btrfs_check_leaf_full(eb); | |
531 | ||
532 | if (ret < 0) { | |
c06631b0 | 533 | btrfs_print_tree(eb, 0); |
8d47a0d8 QW |
534 | btrfs_err(fs_info, |
535 | "block=%llu write time tree block corruption detected", | |
536 | eb->start); | |
c06631b0 | 537 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); |
8d47a0d8 QW |
538 | return ret; |
539 | } | |
2996e1f8 | 540 | write_extent_buffer(eb, result, 0, csum_size); |
8d47a0d8 | 541 | |
2996e1f8 | 542 | return 0; |
19c00ddc CM |
543 | } |
544 | ||
b0c9b3b0 | 545 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 546 | { |
b0c9b3b0 | 547 | struct btrfs_fs_info *fs_info = eb->fs_info; |
01d58472 | 548 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
44880fdc | 549 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c YZ |
550 | int ret = 1; |
551 | ||
9a8658e3 DS |
552 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
553 | BTRFS_FSID_SIZE); | |
2b82032c | 554 | while (fs_devices) { |
7239ff4b NB |
555 | u8 *metadata_uuid; |
556 | ||
557 | /* | |
558 | * Checking the incompat flag is only valid for the current | |
559 | * fs. For seed devices it's forbidden to have their uuid | |
560 | * changed so reading ->fsid in this case is fine | |
561 | */ | |
562 | if (fs_devices == fs_info->fs_devices && | |
563 | btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
564 | metadata_uuid = fs_devices->metadata_uuid; | |
565 | else | |
566 | metadata_uuid = fs_devices->fsid; | |
567 | ||
568 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) { | |
2b82032c YZ |
569 | ret = 0; |
570 | break; | |
571 | } | |
572 | fs_devices = fs_devices->seed; | |
573 | } | |
574 | return ret; | |
575 | } | |
576 | ||
facc8a22 MX |
577 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
578 | u64 phy_offset, struct page *page, | |
579 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 580 | { |
ce9adaa5 CM |
581 | u64 found_start; |
582 | int found_level; | |
ce9adaa5 | 583 | struct extent_buffer *eb; |
15b6e8a8 DS |
584 | struct btrfs_fs_info *fs_info; |
585 | u16 csum_size; | |
f188591e | 586 | int ret = 0; |
2996e1f8 | 587 | u8 result[BTRFS_CSUM_SIZE]; |
727011e0 | 588 | int reads_done; |
ce9adaa5 | 589 | |
ce9adaa5 CM |
590 | if (!page->private) |
591 | goto out; | |
d397712b | 592 | |
4f2de97a | 593 | eb = (struct extent_buffer *)page->private; |
15b6e8a8 DS |
594 | fs_info = eb->fs_info; |
595 | csum_size = btrfs_super_csum_size(fs_info->super_copy); | |
d397712b | 596 | |
0b32f4bb JB |
597 | /* the pending IO might have been the only thing that kept this buffer |
598 | * in memory. Make sure we have a ref for all this other checks | |
599 | */ | |
67439dad | 600 | atomic_inc(&eb->refs); |
0b32f4bb JB |
601 | |
602 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
603 | if (!reads_done) |
604 | goto err; | |
f188591e | 605 | |
5cf1ab56 | 606 | eb->read_mirror = mirror; |
656f30db | 607 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
608 | ret = -EIO; |
609 | goto err; | |
610 | } | |
611 | ||
ce9adaa5 | 612 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 613 | if (found_start != eb->start) { |
893bf4b1 SY |
614 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
615 | eb->start, found_start); | |
f188591e | 616 | ret = -EIO; |
ce9adaa5 CM |
617 | goto err; |
618 | } | |
b0c9b3b0 | 619 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
620 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
621 | eb->start); | |
1259ab75 CM |
622 | ret = -EIO; |
623 | goto err; | |
624 | } | |
ce9adaa5 | 625 | found_level = btrfs_header_level(eb); |
1c24c3ce | 626 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
627 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
628 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce JB |
629 | ret = -EIO; |
630 | goto err; | |
631 | } | |
ce9adaa5 | 632 | |
85d4e461 CM |
633 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
634 | eb, found_level); | |
4008c04a | 635 | |
c67b3892 | 636 | csum_tree_block(eb, result); |
a826d6dc | 637 | |
2996e1f8 JT |
638 | if (memcmp_extent_buffer(eb, result, 0, csum_size)) { |
639 | u32 val; | |
640 | u32 found = 0; | |
641 | ||
642 | memcpy(&found, result, csum_size); | |
643 | ||
644 | read_extent_buffer(eb, &val, 0, csum_size); | |
645 | btrfs_warn_rl(fs_info, | |
646 | "%s checksum verify failed on %llu wanted %x found %x level %d", | |
647 | fs_info->sb->s_id, eb->start, | |
648 | val, found, btrfs_header_level(eb)); | |
649 | ret = -EUCLEAN; | |
650 | goto err; | |
651 | } | |
652 | ||
a826d6dc JB |
653 | /* |
654 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
655 | * that we don't try and read the other copies of this block, just | |
656 | * return -EIO. | |
657 | */ | |
1c4360ee | 658 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
659 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
660 | ret = -EIO; | |
661 | } | |
ce9adaa5 | 662 | |
813fd1dc | 663 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
664 | ret = -EIO; |
665 | ||
0b32f4bb JB |
666 | if (!ret) |
667 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
668 | else |
669 | btrfs_err(fs_info, | |
670 | "block=%llu read time tree block corruption detected", | |
671 | eb->start); | |
ce9adaa5 | 672 | err: |
79fb65a1 JB |
673 | if (reads_done && |
674 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 675 | btree_readahead_hook(eb, ret); |
4bb31e92 | 676 | |
53b381b3 DW |
677 | if (ret) { |
678 | /* | |
679 | * our io error hook is going to dec the io pages | |
680 | * again, we have to make sure it has something | |
681 | * to decrement | |
682 | */ | |
683 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 684 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 685 | } |
0b32f4bb | 686 | free_extent_buffer(eb); |
ce9adaa5 | 687 | out: |
f188591e | 688 | return ret; |
ce9adaa5 CM |
689 | } |
690 | ||
4246a0b6 | 691 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 692 | { |
97eb6b69 | 693 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 694 | struct btrfs_fs_info *fs_info; |
9e0af237 | 695 | struct btrfs_workqueue *wq; |
ce9adaa5 | 696 | |
ce9adaa5 | 697 | fs_info = end_io_wq->info; |
4e4cbee9 | 698 | end_io_wq->status = bio->bi_status; |
d20f7043 | 699 | |
37226b21 | 700 | if (bio_op(bio) == REQ_OP_WRITE) { |
a0cac0ec | 701 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 702 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 703 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 704 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 705 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 706 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 707 | else |
9e0af237 | 708 | wq = fs_info->endio_write_workers; |
d20f7043 | 709 | } else { |
5c047a69 | 710 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 711 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 712 | else if (end_io_wq->metadata) |
9e0af237 | 713 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 714 | else |
9e0af237 | 715 | wq = fs_info->endio_workers; |
d20f7043 | 716 | } |
9e0af237 | 717 | |
a0cac0ec | 718 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 719 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
720 | } |
721 | ||
4e4cbee9 | 722 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 723 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 724 | { |
97eb6b69 | 725 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 726 | |
97eb6b69 | 727 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 728 | if (!end_io_wq) |
4e4cbee9 | 729 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
730 | |
731 | end_io_wq->private = bio->bi_private; | |
732 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 733 | end_io_wq->info = info; |
4e4cbee9 | 734 | end_io_wq->status = 0; |
ce9adaa5 | 735 | end_io_wq->bio = bio; |
22c59948 | 736 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
737 | |
738 | bio->bi_private = end_io_wq; | |
739 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
740 | return 0; |
741 | } | |
742 | ||
4a69a410 CM |
743 | static void run_one_async_start(struct btrfs_work *work) |
744 | { | |
4a69a410 | 745 | struct async_submit_bio *async; |
4e4cbee9 | 746 | blk_status_t ret; |
4a69a410 CM |
747 | |
748 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 749 | ret = async->submit_bio_start(async->private_data, async->bio, |
79787eaa JM |
750 | async->bio_offset); |
751 | if (ret) | |
4e4cbee9 | 752 | async->status = ret; |
4a69a410 CM |
753 | } |
754 | ||
06ea01b1 DS |
755 | /* |
756 | * In order to insert checksums into the metadata in large chunks, we wait | |
757 | * until bio submission time. All the pages in the bio are checksummed and | |
758 | * sums are attached onto the ordered extent record. | |
759 | * | |
760 | * At IO completion time the csums attached on the ordered extent record are | |
761 | * inserted into the tree. | |
762 | */ | |
4a69a410 | 763 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 764 | { |
8b712842 | 765 | struct async_submit_bio *async; |
06ea01b1 DS |
766 | struct inode *inode; |
767 | blk_status_t ret; | |
8b712842 CM |
768 | |
769 | async = container_of(work, struct async_submit_bio, work); | |
06ea01b1 | 770 | inode = async->private_data; |
4854ddd0 | 771 | |
bb7ab3b9 | 772 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
773 | if (async->status) { |
774 | async->bio->bi_status = async->status; | |
4246a0b6 | 775 | bio_endio(async->bio); |
79787eaa JM |
776 | return; |
777 | } | |
778 | ||
ec39f769 CM |
779 | /* |
780 | * All of the bios that pass through here are from async helpers. | |
781 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
782 | * This changes nothing when cgroups aren't in use. | |
783 | */ | |
784 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 785 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
786 | if (ret) { |
787 | async->bio->bi_status = ret; | |
788 | bio_endio(async->bio); | |
789 | } | |
4a69a410 CM |
790 | } |
791 | ||
792 | static void run_one_async_free(struct btrfs_work *work) | |
793 | { | |
794 | struct async_submit_bio *async; | |
795 | ||
796 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
797 | kfree(async); |
798 | } | |
799 | ||
8c27cb35 LT |
800 | blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, |
801 | int mirror_num, unsigned long bio_flags, | |
802 | u64 bio_offset, void *private_data, | |
e288c080 | 803 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e CM |
804 | { |
805 | struct async_submit_bio *async; | |
806 | ||
807 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
808 | if (!async) | |
4e4cbee9 | 809 | return BLK_STS_RESOURCE; |
44b8bd7e | 810 | |
c6100a4b | 811 | async->private_data = private_data; |
44b8bd7e CM |
812 | async->bio = bio; |
813 | async->mirror_num = mirror_num; | |
4a69a410 | 814 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 815 | |
a0cac0ec OS |
816 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
817 | run_one_async_free); | |
4a69a410 | 818 | |
eaf25d93 | 819 | async->bio_offset = bio_offset; |
8c8bee1d | 820 | |
4e4cbee9 | 821 | async->status = 0; |
79787eaa | 822 | |
67f055c7 | 823 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 824 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 825 | |
5cdc7ad3 | 826 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
827 | return 0; |
828 | } | |
829 | ||
4e4cbee9 | 830 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 831 | { |
2c30c71b | 832 | struct bio_vec *bvec; |
ce3ed71a | 833 | struct btrfs_root *root; |
2b070cfe | 834 | int ret = 0; |
6dc4f100 | 835 | struct bvec_iter_all iter_all; |
ce3ed71a | 836 | |
c09abff8 | 837 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 838 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 839 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 840 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
841 | if (ret) |
842 | break; | |
ce3ed71a | 843 | } |
2c30c71b | 844 | |
4e4cbee9 | 845 | return errno_to_blk_status(ret); |
ce3ed71a CM |
846 | } |
847 | ||
d0ee3934 | 848 | static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, |
8c27cb35 | 849 | u64 bio_offset) |
22c59948 | 850 | { |
8b712842 CM |
851 | /* |
852 | * when we're called for a write, we're already in the async | |
5443be45 | 853 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 854 | */ |
79787eaa | 855 | return btree_csum_one_bio(bio); |
4a69a410 | 856 | } |
22c59948 | 857 | |
9b4e675a DS |
858 | static int check_async_write(struct btrfs_fs_info *fs_info, |
859 | struct btrfs_inode *bi) | |
de0022b9 | 860 | { |
6300463b LB |
861 | if (atomic_read(&bi->sync_writers)) |
862 | return 0; | |
9b4e675a | 863 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
de0022b9 | 864 | return 0; |
de0022b9 JB |
865 | return 1; |
866 | } | |
867 | ||
a56b1c7b | 868 | static blk_status_t btree_submit_bio_hook(struct inode *inode, struct bio *bio, |
50489a57 NB |
869 | int mirror_num, |
870 | unsigned long bio_flags) | |
44b8bd7e | 871 | { |
0b246afa | 872 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
9b4e675a | 873 | int async = check_async_write(fs_info, BTRFS_I(inode)); |
4e4cbee9 | 874 | blk_status_t ret; |
cad321ad | 875 | |
37226b21 | 876 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
877 | /* |
878 | * called for a read, do the setup so that checksum validation | |
879 | * can happen in the async kernel threads | |
880 | */ | |
0b246afa JM |
881 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
882 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 883 | if (ret) |
61891923 | 884 | goto out_w_error; |
08635bae | 885 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
de0022b9 JB |
886 | } else if (!async) { |
887 | ret = btree_csum_one_bio(bio); | |
888 | if (ret) | |
61891923 | 889 | goto out_w_error; |
08635bae | 890 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
61891923 SB |
891 | } else { |
892 | /* | |
893 | * kthread helpers are used to submit writes so that | |
894 | * checksumming can happen in parallel across all CPUs | |
895 | */ | |
c6100a4b | 896 | ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, |
e68f2ee7 | 897 | 0, inode, btree_submit_bio_start); |
44b8bd7e | 898 | } |
d313d7a3 | 899 | |
4246a0b6 CH |
900 | if (ret) |
901 | goto out_w_error; | |
902 | return 0; | |
903 | ||
61891923 | 904 | out_w_error: |
4e4cbee9 | 905 | bio->bi_status = ret; |
4246a0b6 | 906 | bio_endio(bio); |
61891923 | 907 | return ret; |
44b8bd7e CM |
908 | } |
909 | ||
3dd1462e | 910 | #ifdef CONFIG_MIGRATION |
784b4e29 | 911 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
912 | struct page *newpage, struct page *page, |
913 | enum migrate_mode mode) | |
784b4e29 CM |
914 | { |
915 | /* | |
916 | * we can't safely write a btree page from here, | |
917 | * we haven't done the locking hook | |
918 | */ | |
919 | if (PageDirty(page)) | |
920 | return -EAGAIN; | |
921 | /* | |
922 | * Buffers may be managed in a filesystem specific way. | |
923 | * We must have no buffers or drop them. | |
924 | */ | |
925 | if (page_has_private(page) && | |
926 | !try_to_release_page(page, GFP_KERNEL)) | |
927 | return -EAGAIN; | |
a6bc32b8 | 928 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 929 | } |
3dd1462e | 930 | #endif |
784b4e29 | 931 | |
0da5468f CM |
932 | |
933 | static int btree_writepages(struct address_space *mapping, | |
934 | struct writeback_control *wbc) | |
935 | { | |
e2d84521 MX |
936 | struct btrfs_fs_info *fs_info; |
937 | int ret; | |
938 | ||
d8d5f3e1 | 939 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
940 | |
941 | if (wbc->for_kupdate) | |
942 | return 0; | |
943 | ||
e2d84521 | 944 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 945 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
946 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
947 | BTRFS_DIRTY_METADATA_THRESH, | |
948 | fs_info->dirty_metadata_batch); | |
e2d84521 | 949 | if (ret < 0) |
793955bc | 950 | return 0; |
793955bc | 951 | } |
0b32f4bb | 952 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
953 | } |
954 | ||
b2950863 | 955 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 956 | { |
71ad38b4 | 957 | return extent_read_full_page(page, btree_get_extent, 0); |
5f39d397 | 958 | } |
22b0ebda | 959 | |
70dec807 | 960 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 961 | { |
98509cfc | 962 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 963 | return 0; |
0c4e538b | 964 | |
f7a52a40 | 965 | return try_release_extent_buffer(page); |
d98237b3 CM |
966 | } |
967 | ||
d47992f8 LC |
968 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
969 | unsigned int length) | |
d98237b3 | 970 | { |
d1310b2e CM |
971 | struct extent_io_tree *tree; |
972 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
973 | extent_invalidatepage(tree, page, offset); |
974 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 975 | if (PagePrivate(page)) { |
efe120a0 FH |
976 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
977 | "page private not zero on page %llu", | |
978 | (unsigned long long)page_offset(page)); | |
d1b89bc0 | 979 | detach_page_private(page); |
9ad6b7bc | 980 | } |
d98237b3 CM |
981 | } |
982 | ||
0b32f4bb JB |
983 | static int btree_set_page_dirty(struct page *page) |
984 | { | |
bb146eb2 | 985 | #ifdef DEBUG |
0b32f4bb JB |
986 | struct extent_buffer *eb; |
987 | ||
988 | BUG_ON(!PagePrivate(page)); | |
989 | eb = (struct extent_buffer *)page->private; | |
990 | BUG_ON(!eb); | |
991 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
992 | BUG_ON(!atomic_read(&eb->refs)); | |
993 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 994 | #endif |
0b32f4bb JB |
995 | return __set_page_dirty_nobuffers(page); |
996 | } | |
997 | ||
7f09410b | 998 | static const struct address_space_operations btree_aops = { |
d98237b3 | 999 | .readpage = btree_readpage, |
0da5468f | 1000 | .writepages = btree_writepages, |
5f39d397 CM |
1001 | .releasepage = btree_releasepage, |
1002 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1003 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1004 | .migratepage = btree_migratepage, |
5a92bc88 | 1005 | #endif |
0b32f4bb | 1006 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1007 | }; |
1008 | ||
2ff7e61e | 1009 | void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) |
090d1875 | 1010 | { |
5f39d397 | 1011 | struct extent_buffer *buf = NULL; |
537f38f0 | 1012 | int ret; |
090d1875 | 1013 | |
2ff7e61e | 1014 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1015 | if (IS_ERR(buf)) |
6197d86e | 1016 | return; |
537f38f0 | 1017 | |
c2ccfbc6 | 1018 | ret = read_extent_buffer_pages(buf, WAIT_NONE, 0); |
537f38f0 NB |
1019 | if (ret < 0) |
1020 | free_extent_buffer_stale(buf); | |
1021 | else | |
1022 | free_extent_buffer(buf); | |
090d1875 CM |
1023 | } |
1024 | ||
2ff7e61e JM |
1025 | struct extent_buffer *btrfs_find_create_tree_block( |
1026 | struct btrfs_fs_info *fs_info, | |
1027 | u64 bytenr) | |
0999df54 | 1028 | { |
0b246afa JM |
1029 | if (btrfs_is_testing(fs_info)) |
1030 | return alloc_test_extent_buffer(fs_info, bytenr); | |
1031 | return alloc_extent_buffer(fs_info, bytenr); | |
0999df54 CM |
1032 | } |
1033 | ||
581c1760 QW |
1034 | /* |
1035 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1036 | * verification. | |
1037 | * | |
1038 | * @parent_transid: expected transid of this tree block, skip check if 0 | |
1039 | * @level: expected level, mandatory check | |
1040 | * @first_key: expected key in slot 0, skip check if NULL | |
1041 | */ | |
2ff7e61e | 1042 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
581c1760 QW |
1043 | u64 parent_transid, int level, |
1044 | struct btrfs_key *first_key) | |
0999df54 CM |
1045 | { |
1046 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1047 | int ret; |
1048 | ||
2ff7e61e | 1049 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 LB |
1050 | if (IS_ERR(buf)) |
1051 | return buf; | |
0999df54 | 1052 | |
5ab12d1f | 1053 | ret = btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 1054 | level, first_key); |
0f0fe8f7 | 1055 | if (ret) { |
537f38f0 | 1056 | free_extent_buffer_stale(buf); |
64c043de | 1057 | return ERR_PTR(ret); |
0f0fe8f7 | 1058 | } |
5f39d397 | 1059 | return buf; |
ce9adaa5 | 1060 | |
eb60ceac CM |
1061 | } |
1062 | ||
6a884d7d | 1063 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1064 | { |
6a884d7d | 1065 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1066 | if (btrfs_header_generation(buf) == |
e2d84521 | 1067 | fs_info->running_transaction->transid) { |
b9447ef8 | 1068 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1069 | |
b9473439 | 1070 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1071 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1072 | -buf->len, | |
1073 | fs_info->dirty_metadata_batch); | |
ed7b63eb | 1074 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
8bead258 | 1075 | btrfs_set_lock_blocking_write(buf); |
ed7b63eb JB |
1076 | clear_extent_buffer_dirty(buf); |
1077 | } | |
925baedd | 1078 | } |
5f39d397 CM |
1079 | } |
1080 | ||
da17066c | 1081 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1082 | u64 objectid) |
d97e63b6 | 1083 | { |
7c0260ee | 1084 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
96dfcb46 | 1085 | root->fs_info = fs_info; |
cfaa7295 | 1086 | root->node = NULL; |
a28ec197 | 1087 | root->commit_root = NULL; |
27cdeb70 | 1088 | root->state = 0; |
d68fc57b | 1089 | root->orphan_cleanup_state = 0; |
0b86a832 | 1090 | |
0f7d52f4 | 1091 | root->last_trans = 0; |
13a8a7c8 | 1092 | root->highest_objectid = 0; |
eb73c1b7 | 1093 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1094 | root->nr_ordered_extents = 0; |
6bef4d31 | 1095 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1096 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1097 | root->block_rsv = NULL; |
0b86a832 CM |
1098 | |
1099 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1100 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1101 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1102 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1103 | INIT_LIST_HEAD(&root->ordered_extents); |
1104 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1105 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1106 | INIT_LIST_HEAD(&root->logged_list[0]); |
1107 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1108 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1109 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1110 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1111 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1112 | spin_lock_init(&root->log_extents_lock[0]); |
1113 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1114 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1115 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1116 | mutex_init(&root->log_mutex); |
31f3d255 | 1117 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1118 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1119 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1120 | init_waitqueue_head(&root->log_writer_wait); |
1121 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1122 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1123 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1124 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1125 | atomic_set(&root->log_commit[0], 0); |
1126 | atomic_set(&root->log_commit[1], 0); | |
1127 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1128 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1129 | refcount_set(&root->refs, 1); |
8ecebf4d | 1130 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1131 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1132 | root->log_transid = 0; |
d1433deb | 1133 | root->log_transid_committed = -1; |
257c62e1 | 1134 | root->last_log_commit = 0; |
e289f03e | 1135 | if (!dummy) { |
43eb5f29 QW |
1136 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1137 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1138 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1139 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1140 | } | |
017e5369 | 1141 | |
3768f368 CM |
1142 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1143 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1144 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
4d775673 | 1145 | root->root_key.objectid = objectid; |
0ee5dc67 | 1146 | root->anon_dev = 0; |
8ea05e3a | 1147 | |
5f3ab90a | 1148 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1149 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1150 | #ifdef CONFIG_BTRFS_DEBUG |
1151 | INIT_LIST_HEAD(&root->leak_list); | |
1152 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1153 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); | |
1154 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1155 | #endif | |
3768f368 CM |
1156 | } |
1157 | ||
74e4d827 | 1158 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1159 | u64 objectid, gfp_t flags) |
6f07e42e | 1160 | { |
74e4d827 | 1161 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1162 | if (root) |
96dfcb46 | 1163 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1164 | return root; |
1165 | } | |
1166 | ||
06ea65a3 JB |
1167 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1168 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1169 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1170 | { |
1171 | struct btrfs_root *root; | |
1172 | ||
7c0260ee JM |
1173 | if (!fs_info) |
1174 | return ERR_PTR(-EINVAL); | |
1175 | ||
96dfcb46 | 1176 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1177 | if (!root) |
1178 | return ERR_PTR(-ENOMEM); | |
da17066c | 1179 | |
b9ef22de | 1180 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1181 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1182 | |
1183 | return root; | |
1184 | } | |
1185 | #endif | |
1186 | ||
20897f5c | 1187 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1188 | u64 objectid) |
1189 | { | |
9b7a2440 | 1190 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1191 | struct extent_buffer *leaf; |
1192 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1193 | struct btrfs_root *root; | |
1194 | struct btrfs_key key; | |
b89f6d1f | 1195 | unsigned int nofs_flag; |
20897f5c | 1196 | int ret = 0; |
20897f5c | 1197 | |
b89f6d1f FM |
1198 | /* |
1199 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1200 | * context to avoid deadlock if reclaim happens. | |
1201 | */ | |
1202 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1203 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1204 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1205 | if (!root) |
1206 | return ERR_PTR(-ENOMEM); | |
1207 | ||
20897f5c AJ |
1208 | root->root_key.objectid = objectid; |
1209 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1210 | root->root_key.offset = 0; | |
1211 | ||
4d75f8a9 | 1212 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1213 | if (IS_ERR(leaf)) { |
1214 | ret = PTR_ERR(leaf); | |
1dd05682 | 1215 | leaf = NULL; |
20897f5c AJ |
1216 | goto fail; |
1217 | } | |
1218 | ||
20897f5c | 1219 | root->node = leaf; |
20897f5c AJ |
1220 | btrfs_mark_buffer_dirty(leaf); |
1221 | ||
1222 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1223 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1224 | |
1225 | root->root_item.flags = 0; | |
1226 | root->root_item.byte_limit = 0; | |
1227 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1228 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1229 | btrfs_set_root_level(&root->root_item, 0); | |
1230 | btrfs_set_root_refs(&root->root_item, 1); | |
1231 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1232 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1233 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1234 | if (is_fstree(objectid)) |
807fc790 AS |
1235 | generate_random_guid(root->root_item.uuid); |
1236 | else | |
1237 | export_guid(root->root_item.uuid, &guid_null); | |
20897f5c AJ |
1238 | root->root_item.drop_level = 0; |
1239 | ||
1240 | key.objectid = objectid; | |
1241 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1242 | key.offset = 0; | |
1243 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1244 | if (ret) | |
1245 | goto fail; | |
1246 | ||
1247 | btrfs_tree_unlock(leaf); | |
1248 | ||
1dd05682 TI |
1249 | return root; |
1250 | ||
20897f5c | 1251 | fail: |
8c38938c | 1252 | if (leaf) |
1dd05682 | 1253 | btrfs_tree_unlock(leaf); |
00246528 | 1254 | btrfs_put_root(root); |
20897f5c | 1255 | |
1dd05682 | 1256 | return ERR_PTR(ret); |
20897f5c AJ |
1257 | } |
1258 | ||
7237f183 YZ |
1259 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1260 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1261 | { |
1262 | struct btrfs_root *root; | |
7237f183 | 1263 | struct extent_buffer *leaf; |
e02119d5 | 1264 | |
96dfcb46 | 1265 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1266 | if (!root) |
7237f183 | 1267 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1268 | |
e02119d5 CM |
1269 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1270 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1271 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1272 | |
7237f183 | 1273 | /* |
92a7cc42 | 1274 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1275 | * |
92a7cc42 QW |
1276 | * Log trees are not exposed to user space thus can't be snapshotted, |
1277 | * and they go away before a real commit is actually done. | |
1278 | * | |
1279 | * They do store pointers to file data extents, and those reference | |
1280 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1281 | */ |
e02119d5 | 1282 | |
4d75f8a9 DS |
1283 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1284 | NULL, 0, 0, 0); | |
7237f183 | 1285 | if (IS_ERR(leaf)) { |
00246528 | 1286 | btrfs_put_root(root); |
7237f183 YZ |
1287 | return ERR_CAST(leaf); |
1288 | } | |
e02119d5 | 1289 | |
7237f183 | 1290 | root->node = leaf; |
e02119d5 | 1291 | |
e02119d5 CM |
1292 | btrfs_mark_buffer_dirty(root->node); |
1293 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1294 | return root; |
1295 | } | |
1296 | ||
1297 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1298 | struct btrfs_fs_info *fs_info) | |
1299 | { | |
1300 | struct btrfs_root *log_root; | |
1301 | ||
1302 | log_root = alloc_log_tree(trans, fs_info); | |
1303 | if (IS_ERR(log_root)) | |
1304 | return PTR_ERR(log_root); | |
1305 | WARN_ON(fs_info->log_root_tree); | |
1306 | fs_info->log_root_tree = log_root; | |
1307 | return 0; | |
1308 | } | |
1309 | ||
1310 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1311 | struct btrfs_root *root) | |
1312 | { | |
0b246afa | 1313 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1314 | struct btrfs_root *log_root; |
1315 | struct btrfs_inode_item *inode_item; | |
1316 | ||
0b246afa | 1317 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1318 | if (IS_ERR(log_root)) |
1319 | return PTR_ERR(log_root); | |
1320 | ||
1321 | log_root->last_trans = trans->transid; | |
1322 | log_root->root_key.offset = root->root_key.objectid; | |
1323 | ||
1324 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1325 | btrfs_set_stack_inode_generation(inode_item, 1); |
1326 | btrfs_set_stack_inode_size(inode_item, 3); | |
1327 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1328 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1329 | fs_info->nodesize); |
3cae210f | 1330 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1331 | |
5d4f98a2 | 1332 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1333 | |
1334 | WARN_ON(root->log_root); | |
1335 | root->log_root = log_root; | |
1336 | root->log_transid = 0; | |
d1433deb | 1337 | root->log_transid_committed = -1; |
257c62e1 | 1338 | root->last_log_commit = 0; |
e02119d5 CM |
1339 | return 0; |
1340 | } | |
1341 | ||
62a2c73e JB |
1342 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1343 | struct btrfs_key *key) | |
e02119d5 CM |
1344 | { |
1345 | struct btrfs_root *root; | |
1346 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1347 | struct btrfs_path *path; |
84234f3a | 1348 | u64 generation; |
cb517eab | 1349 | int ret; |
581c1760 | 1350 | int level; |
0f7d52f4 | 1351 | |
cb517eab MX |
1352 | path = btrfs_alloc_path(); |
1353 | if (!path) | |
0f7d52f4 | 1354 | return ERR_PTR(-ENOMEM); |
cb517eab | 1355 | |
96dfcb46 | 1356 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
cb517eab MX |
1357 | if (!root) { |
1358 | ret = -ENOMEM; | |
1359 | goto alloc_fail; | |
0f7d52f4 CM |
1360 | } |
1361 | ||
cb517eab MX |
1362 | ret = btrfs_find_root(tree_root, key, path, |
1363 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1364 | if (ret) { |
13a8a7c8 YZ |
1365 | if (ret > 0) |
1366 | ret = -ENOENT; | |
cb517eab | 1367 | goto find_fail; |
0f7d52f4 | 1368 | } |
13a8a7c8 | 1369 | |
84234f3a | 1370 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1371 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1372 | root->node = read_tree_block(fs_info, |
1373 | btrfs_root_bytenr(&root->root_item), | |
581c1760 | 1374 | generation, level, NULL); |
64c043de LB |
1375 | if (IS_ERR(root->node)) { |
1376 | ret = PTR_ERR(root->node); | |
8c38938c | 1377 | root->node = NULL; |
cb517eab MX |
1378 | goto find_fail; |
1379 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1380 | ret = -EIO; | |
64c043de | 1381 | goto find_fail; |
416bc658 | 1382 | } |
5d4f98a2 | 1383 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1384 | out: |
cb517eab MX |
1385 | btrfs_free_path(path); |
1386 | return root; | |
1387 | ||
cb517eab | 1388 | find_fail: |
00246528 | 1389 | btrfs_put_root(root); |
cb517eab MX |
1390 | alloc_fail: |
1391 | root = ERR_PTR(ret); | |
1392 | goto out; | |
1393 | } | |
1394 | ||
2dfb1e43 QW |
1395 | /* |
1396 | * Initialize subvolume root in-memory structure | |
1397 | * | |
1398 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1399 | */ | |
1400 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1401 | { |
1402 | int ret; | |
dcc3eb96 | 1403 | unsigned int nofs_flag; |
cb517eab MX |
1404 | |
1405 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1406 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1407 | GFP_NOFS); | |
1408 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1409 | ret = -ENOMEM; | |
1410 | goto fail; | |
1411 | } | |
1412 | ||
dcc3eb96 NB |
1413 | /* |
1414 | * We might be called under a transaction (e.g. indirect backref | |
1415 | * resolution) which could deadlock if it triggers memory reclaim | |
1416 | */ | |
1417 | nofs_flag = memalloc_nofs_save(); | |
1418 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1419 | memalloc_nofs_restore(nofs_flag); | |
1420 | if (ret) | |
8257b2dc | 1421 | goto fail; |
8257b2dc | 1422 | |
aeb935a4 QW |
1423 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
1424 | root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) { | |
92a7cc42 | 1425 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1426 | btrfs_check_and_init_root_item(&root->root_item); |
1427 | } | |
1428 | ||
cb517eab | 1429 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1430 | spin_lock_init(&root->ino_cache_lock); |
1431 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab | 1432 | |
851fd730 QW |
1433 | /* |
1434 | * Don't assign anonymous block device to roots that are not exposed to | |
1435 | * userspace, the id pool is limited to 1M | |
1436 | */ | |
1437 | if (is_fstree(root->root_key.objectid) && | |
1438 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1439 | if (!anon_dev) { |
1440 | ret = get_anon_bdev(&root->anon_dev); | |
1441 | if (ret) | |
1442 | goto fail; | |
1443 | } else { | |
1444 | root->anon_dev = anon_dev; | |
1445 | } | |
851fd730 | 1446 | } |
f32e48e9 CR |
1447 | |
1448 | mutex_lock(&root->objectid_mutex); | |
1449 | ret = btrfs_find_highest_objectid(root, | |
1450 | &root->highest_objectid); | |
1451 | if (ret) { | |
1452 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1453 | goto fail; |
f32e48e9 CR |
1454 | } |
1455 | ||
1456 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1457 | ||
1458 | mutex_unlock(&root->objectid_mutex); | |
1459 | ||
cb517eab MX |
1460 | return 0; |
1461 | fail: | |
84db5ccf | 1462 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1463 | return ret; |
1464 | } | |
1465 | ||
a98db0f3 JB |
1466 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1467 | u64 root_id) | |
cb517eab MX |
1468 | { |
1469 | struct btrfs_root *root; | |
1470 | ||
1471 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1472 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1473 | (unsigned long)root_id); | |
bc44d7c4 | 1474 | if (root) |
00246528 | 1475 | root = btrfs_grab_root(root); |
cb517eab MX |
1476 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1477 | return root; | |
1478 | } | |
1479 | ||
1480 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1481 | struct btrfs_root *root) | |
1482 | { | |
1483 | int ret; | |
1484 | ||
e1860a77 | 1485 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1486 | if (ret) |
1487 | return ret; | |
1488 | ||
1489 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1490 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1491 | (unsigned long)root->root_key.objectid, | |
1492 | root); | |
af01d2e5 | 1493 | if (ret == 0) { |
00246528 | 1494 | btrfs_grab_root(root); |
27cdeb70 | 1495 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1496 | } |
cb517eab MX |
1497 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1498 | radix_tree_preload_end(); | |
1499 | ||
1500 | return ret; | |
1501 | } | |
1502 | ||
bd647ce3 JB |
1503 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1504 | { | |
1505 | #ifdef CONFIG_BTRFS_DEBUG | |
1506 | struct btrfs_root *root; | |
1507 | ||
1508 | while (!list_empty(&fs_info->allocated_roots)) { | |
1509 | root = list_first_entry(&fs_info->allocated_roots, | |
1510 | struct btrfs_root, leak_list); | |
1511 | btrfs_err(fs_info, "leaked root %llu-%llu refcount %d", | |
1512 | root->root_key.objectid, root->root_key.offset, | |
1513 | refcount_read(&root->refs)); | |
1514 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1515 | btrfs_put_root(root); |
1516 | btrfs_put_root(root); | |
bd647ce3 JB |
1517 | } |
1518 | #endif | |
1519 | } | |
1520 | ||
0d4b0463 JB |
1521 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1522 | { | |
141386e1 JB |
1523 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1524 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
1525 | percpu_counter_destroy(&fs_info->dio_bytes); | |
1526 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); | |
1527 | btrfs_free_csum_hash(fs_info); | |
1528 | btrfs_free_stripe_hash_table(fs_info); | |
1529 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1530 | kfree(fs_info->balance_ctl); |
1531 | kfree(fs_info->delayed_root); | |
00246528 JB |
1532 | btrfs_put_root(fs_info->extent_root); |
1533 | btrfs_put_root(fs_info->tree_root); | |
1534 | btrfs_put_root(fs_info->chunk_root); | |
1535 | btrfs_put_root(fs_info->dev_root); | |
1536 | btrfs_put_root(fs_info->csum_root); | |
1537 | btrfs_put_root(fs_info->quota_root); | |
1538 | btrfs_put_root(fs_info->uuid_root); | |
1539 | btrfs_put_root(fs_info->free_space_root); | |
1540 | btrfs_put_root(fs_info->fs_root); | |
aeb935a4 | 1541 | btrfs_put_root(fs_info->data_reloc_root); |
bd647ce3 | 1542 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1543 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1544 | kfree(fs_info->super_copy); |
1545 | kfree(fs_info->super_for_commit); | |
1546 | kvfree(fs_info); | |
1547 | } | |
1548 | ||
1549 | ||
2dfb1e43 QW |
1550 | /* |
1551 | * Get an in-memory reference of a root structure. | |
1552 | * | |
1553 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1554 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1555 | * found, then read it from disk and add it to cached fs roots. | |
1556 | * | |
1557 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1558 | * | |
1559 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1560 | * same root objectid. | |
1561 | * | |
1562 | * @objectid: root id | |
1563 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1564 | * pass 0 for new allocation. | |
1565 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1566 | * for orphan roots | |
1567 | */ | |
1568 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1569 | u64 objectid, dev_t anon_dev, | |
1570 | bool check_ref) | |
5eda7b5e CM |
1571 | { |
1572 | struct btrfs_root *root; | |
381cf658 | 1573 | struct btrfs_path *path; |
1d4c08e0 | 1574 | struct btrfs_key key; |
5eda7b5e CM |
1575 | int ret; |
1576 | ||
56e9357a | 1577 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
00246528 | 1578 | return btrfs_grab_root(fs_info->tree_root); |
56e9357a | 1579 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
00246528 | 1580 | return btrfs_grab_root(fs_info->extent_root); |
56e9357a | 1581 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
00246528 | 1582 | return btrfs_grab_root(fs_info->chunk_root); |
56e9357a | 1583 | if (objectid == BTRFS_DEV_TREE_OBJECTID) |
00246528 | 1584 | return btrfs_grab_root(fs_info->dev_root); |
56e9357a | 1585 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) |
00246528 | 1586 | return btrfs_grab_root(fs_info->csum_root); |
56e9357a | 1587 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
00246528 | 1588 | return btrfs_grab_root(fs_info->quota_root) ? |
bc44d7c4 | 1589 | fs_info->quota_root : ERR_PTR(-ENOENT); |
56e9357a | 1590 | if (objectid == BTRFS_UUID_TREE_OBJECTID) |
00246528 | 1591 | return btrfs_grab_root(fs_info->uuid_root) ? |
bc44d7c4 | 1592 | fs_info->uuid_root : ERR_PTR(-ENOENT); |
56e9357a | 1593 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
00246528 | 1594 | return btrfs_grab_root(fs_info->free_space_root) ? |
bc44d7c4 | 1595 | fs_info->free_space_root : ERR_PTR(-ENOENT); |
4df27c4d | 1596 | again: |
56e9357a | 1597 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1598 | if (root) { |
2dfb1e43 QW |
1599 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1600 | ASSERT(!anon_dev); | |
bc44d7c4 | 1601 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1602 | btrfs_put_root(root); |
48475471 | 1603 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1604 | } |
5eda7b5e | 1605 | return root; |
48475471 | 1606 | } |
5eda7b5e | 1607 | |
56e9357a DS |
1608 | key.objectid = objectid; |
1609 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1610 | key.offset = (u64)-1; | |
1611 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1612 | if (IS_ERR(root)) |
1613 | return root; | |
3394e160 | 1614 | |
c00869f1 | 1615 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1616 | ret = -ENOENT; |
581bb050 | 1617 | goto fail; |
35a30d7c | 1618 | } |
581bb050 | 1619 | |
2dfb1e43 | 1620 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1621 | if (ret) |
1622 | goto fail; | |
3394e160 | 1623 | |
381cf658 DS |
1624 | path = btrfs_alloc_path(); |
1625 | if (!path) { | |
1626 | ret = -ENOMEM; | |
1627 | goto fail; | |
1628 | } | |
1d4c08e0 DS |
1629 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1630 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1631 | key.offset = objectid; |
1d4c08e0 DS |
1632 | |
1633 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1634 | btrfs_free_path(path); |
d68fc57b YZ |
1635 | if (ret < 0) |
1636 | goto fail; | |
1637 | if (ret == 0) | |
27cdeb70 | 1638 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1639 | |
cb517eab | 1640 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1641 | if (ret) { |
00246528 | 1642 | btrfs_put_root(root); |
4785e24f | 1643 | if (ret == -EEXIST) |
4df27c4d | 1644 | goto again; |
4df27c4d | 1645 | goto fail; |
0f7d52f4 | 1646 | } |
edbd8d4e | 1647 | return root; |
4df27c4d | 1648 | fail: |
8c38938c | 1649 | btrfs_put_root(root); |
4df27c4d | 1650 | return ERR_PTR(ret); |
edbd8d4e CM |
1651 | } |
1652 | ||
2dfb1e43 QW |
1653 | /* |
1654 | * Get in-memory reference of a root structure | |
1655 | * | |
1656 | * @objectid: tree objectid | |
1657 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1658 | * one reference | |
1659 | */ | |
1660 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1661 | u64 objectid, bool check_ref) | |
1662 | { | |
1663 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1664 | } | |
1665 | ||
1666 | /* | |
1667 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1668 | * the anonymous block device id | |
1669 | * | |
1670 | * @objectid: tree objectid | |
1671 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1672 | * parameter value | |
1673 | */ | |
1674 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1675 | u64 objectid, dev_t anon_dev) | |
1676 | { | |
1677 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1678 | } | |
1679 | ||
04160088 CM |
1680 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1681 | { | |
1682 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1683 | int ret = 0; | |
04160088 CM |
1684 | struct btrfs_device *device; |
1685 | struct backing_dev_info *bdi; | |
b7967db7 | 1686 | |
1f78160c XG |
1687 | rcu_read_lock(); |
1688 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1689 | if (!device->bdev) |
1690 | continue; | |
efa7c9f9 | 1691 | bdi = device->bdev->bd_bdi; |
ff9ea323 | 1692 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1693 | ret = 1; |
1694 | break; | |
1695 | } | |
1696 | } | |
1f78160c | 1697 | rcu_read_unlock(); |
04160088 CM |
1698 | return ret; |
1699 | } | |
1700 | ||
8b712842 CM |
1701 | /* |
1702 | * called by the kthread helper functions to finally call the bio end_io | |
1703 | * functions. This is where read checksum verification actually happens | |
1704 | */ | |
1705 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1706 | { |
ce9adaa5 | 1707 | struct bio *bio; |
97eb6b69 | 1708 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1709 | |
97eb6b69 | 1710 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1711 | bio = end_io_wq->bio; |
ce9adaa5 | 1712 | |
4e4cbee9 | 1713 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1714 | bio->bi_private = end_io_wq->private; |
1715 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1716 | bio_endio(bio); |
9be490f1 | 1717 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1718 | } |
1719 | ||
a74a4b97 CM |
1720 | static int cleaner_kthread(void *arg) |
1721 | { | |
1722 | struct btrfs_root *root = arg; | |
0b246afa | 1723 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1724 | int again; |
a74a4b97 | 1725 | |
d6fd0ae2 | 1726 | while (1) { |
d0278245 | 1727 | again = 0; |
a74a4b97 | 1728 | |
fd340d0f JB |
1729 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1730 | ||
d0278245 | 1731 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1732 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1733 | goto sleep; |
1734 | ||
90c711ab ZB |
1735 | /* |
1736 | * Do not do anything if we might cause open_ctree() to block | |
1737 | * before we have finished mounting the filesystem. | |
1738 | */ | |
0b246afa | 1739 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1740 | goto sleep; |
1741 | ||
0b246afa | 1742 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1743 | goto sleep; |
1744 | ||
dc7f370c MX |
1745 | /* |
1746 | * Avoid the problem that we change the status of the fs | |
1747 | * during the above check and trylock. | |
1748 | */ | |
2ff7e61e | 1749 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1750 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1751 | goto sleep; |
76dda93c | 1752 | } |
a74a4b97 | 1753 | |
2ff7e61e | 1754 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1755 | |
d0278245 | 1756 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1757 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1758 | |
1759 | /* | |
05323cd1 MX |
1760 | * The defragger has dealt with the R/O remount and umount, |
1761 | * needn't do anything special here. | |
d0278245 | 1762 | */ |
0b246afa | 1763 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1764 | |
1765 | /* | |
1766 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1767 | * with relocation (btrfs_relocate_chunk) and relocation | |
1768 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1769 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1770 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1771 | * unused block groups. | |
1772 | */ | |
0b246afa | 1773 | btrfs_delete_unused_bgs(fs_info); |
d0278245 | 1774 | sleep: |
fd340d0f | 1775 | clear_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1776 | if (kthread_should_park()) |
1777 | kthread_parkme(); | |
1778 | if (kthread_should_stop()) | |
1779 | return 0; | |
838fe188 | 1780 | if (!again) { |
a74a4b97 | 1781 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1782 | schedule(); |
a74a4b97 CM |
1783 | __set_current_state(TASK_RUNNING); |
1784 | } | |
da288d28 | 1785 | } |
a74a4b97 CM |
1786 | } |
1787 | ||
1788 | static int transaction_kthread(void *arg) | |
1789 | { | |
1790 | struct btrfs_root *root = arg; | |
0b246afa | 1791 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1792 | struct btrfs_trans_handle *trans; |
1793 | struct btrfs_transaction *cur; | |
8929ecfa | 1794 | u64 transid; |
a944442c | 1795 | time64_t now; |
a74a4b97 | 1796 | unsigned long delay; |
914b2007 | 1797 | bool cannot_commit; |
a74a4b97 CM |
1798 | |
1799 | do { | |
914b2007 | 1800 | cannot_commit = false; |
0b246afa JM |
1801 | delay = HZ * fs_info->commit_interval; |
1802 | mutex_lock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1803 | |
0b246afa JM |
1804 | spin_lock(&fs_info->trans_lock); |
1805 | cur = fs_info->running_transaction; | |
a74a4b97 | 1806 | if (!cur) { |
0b246afa | 1807 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1808 | goto sleep; |
1809 | } | |
31153d81 | 1810 | |
afd48513 | 1811 | now = ktime_get_seconds(); |
3296bf56 | 1812 | if (cur->state < TRANS_STATE_COMMIT_START && |
8b87dc17 | 1813 | (now < cur->start_time || |
0b246afa JM |
1814 | now - cur->start_time < fs_info->commit_interval)) { |
1815 | spin_unlock(&fs_info->trans_lock); | |
a74a4b97 CM |
1816 | delay = HZ * 5; |
1817 | goto sleep; | |
1818 | } | |
8929ecfa | 1819 | transid = cur->transid; |
0b246afa | 1820 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1821 | |
79787eaa | 1822 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1823 | trans = btrfs_attach_transaction(root); |
914b2007 | 1824 | if (IS_ERR(trans)) { |
354aa0fb MX |
1825 | if (PTR_ERR(trans) != -ENOENT) |
1826 | cannot_commit = true; | |
79787eaa | 1827 | goto sleep; |
914b2007 | 1828 | } |
8929ecfa | 1829 | if (transid == trans->transid) { |
3a45bb20 | 1830 | btrfs_commit_transaction(trans); |
8929ecfa | 1831 | } else { |
3a45bb20 | 1832 | btrfs_end_transaction(trans); |
8929ecfa | 1833 | } |
a74a4b97 | 1834 | sleep: |
0b246afa JM |
1835 | wake_up_process(fs_info->cleaner_kthread); |
1836 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1837 | |
4e121c06 | 1838 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1839 | &fs_info->fs_state))) |
2ff7e61e | 1840 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1841 | if (!kthread_should_stop() && |
0b246afa | 1842 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1843 | cannot_commit)) |
bc5511d0 | 1844 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1845 | } while (!kthread_should_stop()); |
1846 | return 0; | |
1847 | } | |
1848 | ||
af31f5e5 | 1849 | /* |
01f0f9da NB |
1850 | * This will find the highest generation in the array of root backups. The |
1851 | * index of the highest array is returned, or -EINVAL if we can't find | |
1852 | * anything. | |
af31f5e5 CM |
1853 | * |
1854 | * We check to make sure the array is valid by comparing the | |
1855 | * generation of the latest root in the array with the generation | |
1856 | * in the super block. If they don't match we pitch it. | |
1857 | */ | |
01f0f9da | 1858 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1859 | { |
01f0f9da | 1860 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1861 | u64 cur; |
af31f5e5 CM |
1862 | struct btrfs_root_backup *root_backup; |
1863 | int i; | |
1864 | ||
1865 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1866 | root_backup = info->super_copy->super_roots + i; | |
1867 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1868 | if (cur == newest_gen) | |
01f0f9da | 1869 | return i; |
af31f5e5 CM |
1870 | } |
1871 | ||
01f0f9da | 1872 | return -EINVAL; |
af31f5e5 CM |
1873 | } |
1874 | ||
af31f5e5 CM |
1875 | /* |
1876 | * copy all the root pointers into the super backup array. | |
1877 | * this will bump the backup pointer by one when it is | |
1878 | * done | |
1879 | */ | |
1880 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1881 | { | |
6ef108dd | 1882 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1883 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1884 | |
1885 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1886 | ||
1887 | /* | |
1888 | * make sure all of our padding and empty slots get zero filled | |
1889 | * regardless of which ones we use today | |
1890 | */ | |
1891 | memset(root_backup, 0, sizeof(*root_backup)); | |
1892 | ||
1893 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1894 | ||
1895 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1896 | btrfs_set_backup_tree_root_gen(root_backup, | |
1897 | btrfs_header_generation(info->tree_root->node)); | |
1898 | ||
1899 | btrfs_set_backup_tree_root_level(root_backup, | |
1900 | btrfs_header_level(info->tree_root->node)); | |
1901 | ||
1902 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1903 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1904 | btrfs_header_generation(info->chunk_root->node)); | |
1905 | btrfs_set_backup_chunk_root_level(root_backup, | |
1906 | btrfs_header_level(info->chunk_root->node)); | |
1907 | ||
1908 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1909 | btrfs_set_backup_extent_root_gen(root_backup, | |
1910 | btrfs_header_generation(info->extent_root->node)); | |
1911 | btrfs_set_backup_extent_root_level(root_backup, | |
1912 | btrfs_header_level(info->extent_root->node)); | |
1913 | ||
7c7e82a7 CM |
1914 | /* |
1915 | * we might commit during log recovery, which happens before we set | |
1916 | * the fs_root. Make sure it is valid before we fill it in. | |
1917 | */ | |
1918 | if (info->fs_root && info->fs_root->node) { | |
1919 | btrfs_set_backup_fs_root(root_backup, | |
1920 | info->fs_root->node->start); | |
1921 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1922 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1923 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1924 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1925 | } |
af31f5e5 CM |
1926 | |
1927 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1928 | btrfs_set_backup_dev_root_gen(root_backup, | |
1929 | btrfs_header_generation(info->dev_root->node)); | |
1930 | btrfs_set_backup_dev_root_level(root_backup, | |
1931 | btrfs_header_level(info->dev_root->node)); | |
1932 | ||
1933 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1934 | btrfs_set_backup_csum_root_gen(root_backup, | |
1935 | btrfs_header_generation(info->csum_root->node)); | |
1936 | btrfs_set_backup_csum_root_level(root_backup, | |
1937 | btrfs_header_level(info->csum_root->node)); | |
1938 | ||
1939 | btrfs_set_backup_total_bytes(root_backup, | |
1940 | btrfs_super_total_bytes(info->super_copy)); | |
1941 | btrfs_set_backup_bytes_used(root_backup, | |
1942 | btrfs_super_bytes_used(info->super_copy)); | |
1943 | btrfs_set_backup_num_devices(root_backup, | |
1944 | btrfs_super_num_devices(info->super_copy)); | |
1945 | ||
1946 | /* | |
1947 | * if we don't copy this out to the super_copy, it won't get remembered | |
1948 | * for the next commit | |
1949 | */ | |
1950 | memcpy(&info->super_copy->super_roots, | |
1951 | &info->super_for_commit->super_roots, | |
1952 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1953 | } | |
1954 | ||
bd2336b2 NB |
1955 | /* |
1956 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
1957 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
1958 | * | |
1959 | * fs_info - filesystem whose backup roots need to be read | |
1960 | * priority - priority of backup root required | |
1961 | * | |
1962 | * Returns backup root index on success and -EINVAL otherwise. | |
1963 | */ | |
1964 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1965 | { | |
1966 | int backup_index = find_newest_super_backup(fs_info); | |
1967 | struct btrfs_super_block *super = fs_info->super_copy; | |
1968 | struct btrfs_root_backup *root_backup; | |
1969 | ||
1970 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1971 | if (priority == 0) | |
1972 | return backup_index; | |
1973 | ||
1974 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1975 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1976 | } else { | |
1977 | return -EINVAL; | |
1978 | } | |
1979 | ||
1980 | root_backup = super->super_roots + backup_index; | |
1981 | ||
1982 | btrfs_set_super_generation(super, | |
1983 | btrfs_backup_tree_root_gen(root_backup)); | |
1984 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1985 | btrfs_set_super_root_level(super, | |
1986 | btrfs_backup_tree_root_level(root_backup)); | |
1987 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1988 | ||
1989 | /* | |
1990 | * Fixme: the total bytes and num_devices need to match or we should | |
1991 | * need a fsck | |
1992 | */ | |
1993 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1994 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1995 | ||
1996 | return backup_index; | |
1997 | } | |
1998 | ||
7abadb64 LB |
1999 | /* helper to cleanup workers */ |
2000 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2001 | { | |
dc6e3209 | 2002 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2003 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2004 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2005 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2006 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
d05a33ac | 2007 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2008 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2009 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2010 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2011 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2012 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2013 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2014 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2015 | if (fs_info->discard_ctl.discard_workers) |
2016 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2017 | /* |
2018 | * Now that all other work queues are destroyed, we can safely destroy | |
2019 | * the queues used for metadata I/O, since tasks from those other work | |
2020 | * queues can do metadata I/O operations. | |
2021 | */ | |
2022 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2023 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2024 | } |
2025 | ||
2e9f5954 R |
2026 | static void free_root_extent_buffers(struct btrfs_root *root) |
2027 | { | |
2028 | if (root) { | |
2029 | free_extent_buffer(root->node); | |
2030 | free_extent_buffer(root->commit_root); | |
2031 | root->node = NULL; | |
2032 | root->commit_root = NULL; | |
2033 | } | |
2034 | } | |
2035 | ||
af31f5e5 | 2036 | /* helper to cleanup tree roots */ |
4273eaff | 2037 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2038 | { |
2e9f5954 | 2039 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2040 | |
2e9f5954 R |
2041 | free_root_extent_buffers(info->dev_root); |
2042 | free_root_extent_buffers(info->extent_root); | |
2043 | free_root_extent_buffers(info->csum_root); | |
2044 | free_root_extent_buffers(info->quota_root); | |
2045 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2046 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2047 | free_root_extent_buffers(info->data_reloc_root); |
4273eaff | 2048 | if (free_chunk_root) |
2e9f5954 | 2049 | free_root_extent_buffers(info->chunk_root); |
70f6d82e | 2050 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2051 | } |
2052 | ||
8c38938c JB |
2053 | void btrfs_put_root(struct btrfs_root *root) |
2054 | { | |
2055 | if (!root) | |
2056 | return; | |
2057 | ||
2058 | if (refcount_dec_and_test(&root->refs)) { | |
2059 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2060 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2061 | if (root->anon_dev) |
2062 | free_anon_bdev(root->anon_dev); | |
2063 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2064 | free_root_extent_buffers(root); |
8c38938c JB |
2065 | kfree(root->free_ino_ctl); |
2066 | kfree(root->free_ino_pinned); | |
2067 | #ifdef CONFIG_BTRFS_DEBUG | |
2068 | spin_lock(&root->fs_info->fs_roots_radix_lock); | |
2069 | list_del_init(&root->leak_list); | |
2070 | spin_unlock(&root->fs_info->fs_roots_radix_lock); | |
2071 | #endif | |
2072 | kfree(root); | |
2073 | } | |
2074 | } | |
2075 | ||
faa2dbf0 | 2076 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2077 | { |
2078 | int ret; | |
2079 | struct btrfs_root *gang[8]; | |
2080 | int i; | |
2081 | ||
2082 | while (!list_empty(&fs_info->dead_roots)) { | |
2083 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2084 | struct btrfs_root, root_list); | |
2085 | list_del(&gang[0]->root_list); | |
2086 | ||
8c38938c | 2087 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
cb517eab | 2088 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
dc9492c1 | 2089 | btrfs_put_root(gang[0]); |
171f6537 JB |
2090 | } |
2091 | ||
2092 | while (1) { | |
2093 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2094 | (void **)gang, 0, | |
2095 | ARRAY_SIZE(gang)); | |
2096 | if (!ret) | |
2097 | break; | |
2098 | for (i = 0; i < ret; i++) | |
cb517eab | 2099 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 JB |
2100 | } |
2101 | } | |
af31f5e5 | 2102 | |
638aa7ed ES |
2103 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2104 | { | |
2105 | mutex_init(&fs_info->scrub_lock); | |
2106 | atomic_set(&fs_info->scrubs_running, 0); | |
2107 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2108 | atomic_set(&fs_info->scrubs_paused, 0); | |
2109 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2110 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2111 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2112 | } |
2113 | ||
779a65a4 ES |
2114 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2115 | { | |
2116 | spin_lock_init(&fs_info->balance_lock); | |
2117 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2118 | atomic_set(&fs_info->balance_pause_req, 0); |
2119 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2120 | fs_info->balance_ctl = NULL; | |
2121 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2122 | } | |
2123 | ||
6bccf3ab | 2124 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2125 | { |
2ff7e61e JM |
2126 | struct inode *inode = fs_info->btree_inode; |
2127 | ||
2128 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2129 | set_nlink(inode, 1); | |
f37938e0 ES |
2130 | /* |
2131 | * we set the i_size on the btree inode to the max possible int. | |
2132 | * the real end of the address space is determined by all of | |
2133 | * the devices in the system | |
2134 | */ | |
2ff7e61e JM |
2135 | inode->i_size = OFFSET_MAX; |
2136 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2137 | |
2ff7e61e | 2138 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 QW |
2139 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2140 | IO_TREE_INODE_IO, inode); | |
7b439738 | 2141 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2142 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2143 | |
2ff7e61e | 2144 | BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; |
f37938e0 | 2145 | |
5c8fd99f | 2146 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2147 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2148 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2149 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2150 | } |
2151 | ||
ad618368 ES |
2152 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2153 | { | |
ad618368 | 2154 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2155 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2156 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2157 | } |
2158 | ||
f9e92e40 ES |
2159 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2160 | { | |
2161 | spin_lock_init(&fs_info->qgroup_lock); | |
2162 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2163 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2164 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2165 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2166 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2167 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2168 | mutex_init(&fs_info->qgroup_rescan_lock); |
2169 | } | |
2170 | ||
2a458198 ES |
2171 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2172 | struct btrfs_fs_devices *fs_devices) | |
2173 | { | |
f7b885be | 2174 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2175 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2176 | |
2177 | fs_info->workers = | |
cb001095 JM |
2178 | btrfs_alloc_workqueue(fs_info, "worker", |
2179 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2180 | |
2181 | fs_info->delalloc_workers = | |
cb001095 JM |
2182 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2183 | flags, max_active, 2); | |
2a458198 ES |
2184 | |
2185 | fs_info->flush_workers = | |
cb001095 JM |
2186 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2187 | flags, max_active, 0); | |
2a458198 ES |
2188 | |
2189 | fs_info->caching_workers = | |
cb001095 | 2190 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2191 | |
2a458198 | 2192 | fs_info->fixup_workers = |
cb001095 | 2193 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2194 | |
2195 | /* | |
2196 | * endios are largely parallel and should have a very | |
2197 | * low idle thresh | |
2198 | */ | |
2199 | fs_info->endio_workers = | |
cb001095 | 2200 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2201 | fs_info->endio_meta_workers = |
cb001095 JM |
2202 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2203 | max_active, 4); | |
2a458198 | 2204 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2205 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2206 | max_active, 2); | |
2a458198 | 2207 | fs_info->endio_raid56_workers = |
cb001095 JM |
2208 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2209 | max_active, 4); | |
2a458198 | 2210 | fs_info->rmw_workers = |
cb001095 | 2211 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2212 | fs_info->endio_write_workers = |
cb001095 JM |
2213 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2214 | max_active, 2); | |
2a458198 | 2215 | fs_info->endio_freespace_worker = |
cb001095 JM |
2216 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2217 | max_active, 0); | |
2a458198 | 2218 | fs_info->delayed_workers = |
cb001095 JM |
2219 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2220 | max_active, 0); | |
2a458198 | 2221 | fs_info->readahead_workers = |
cb001095 JM |
2222 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2223 | max_active, 2); | |
2a458198 | 2224 | fs_info->qgroup_rescan_workers = |
cb001095 | 2225 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2226 | fs_info->discard_ctl.discard_workers = |
2227 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 ES |
2228 | |
2229 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
ba8a9d07 | 2230 | fs_info->flush_workers && |
2a458198 ES |
2231 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2232 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2233 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2234 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2235 | fs_info->caching_workers && fs_info->readahead_workers && | |
2236 | fs_info->fixup_workers && fs_info->delayed_workers && | |
b0643e59 DZ |
2237 | fs_info->qgroup_rescan_workers && |
2238 | fs_info->discard_ctl.discard_workers)) { | |
2a458198 ES |
2239 | return -ENOMEM; |
2240 | } | |
2241 | ||
2242 | return 0; | |
2243 | } | |
2244 | ||
6d97c6e3 JT |
2245 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2246 | { | |
2247 | struct crypto_shash *csum_shash; | |
b4e967be | 2248 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2249 | |
b4e967be | 2250 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2251 | |
2252 | if (IS_ERR(csum_shash)) { | |
2253 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2254 | csum_driver); |
6d97c6e3 JT |
2255 | return PTR_ERR(csum_shash); |
2256 | } | |
2257 | ||
2258 | fs_info->csum_shash = csum_shash; | |
2259 | ||
2260 | return 0; | |
2261 | } | |
2262 | ||
63443bf5 ES |
2263 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2264 | struct btrfs_fs_devices *fs_devices) | |
2265 | { | |
2266 | int ret; | |
63443bf5 ES |
2267 | struct btrfs_root *log_tree_root; |
2268 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2269 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2270 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2271 | |
2272 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2273 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2274 | return -EIO; |
2275 | } | |
2276 | ||
96dfcb46 JB |
2277 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2278 | GFP_KERNEL); | |
63443bf5 ES |
2279 | if (!log_tree_root) |
2280 | return -ENOMEM; | |
2281 | ||
2ff7e61e | 2282 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
581c1760 QW |
2283 | fs_info->generation + 1, |
2284 | level, NULL); | |
64c043de | 2285 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2286 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2287 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2288 | log_tree_root->node = NULL; |
00246528 | 2289 | btrfs_put_root(log_tree_root); |
0eeff236 | 2290 | return ret; |
64c043de | 2291 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2292 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2293 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2294 | return -EIO; |
2295 | } | |
2296 | /* returns with log_tree_root freed on success */ | |
2297 | ret = btrfs_recover_log_trees(log_tree_root); | |
2298 | if (ret) { | |
0b246afa JM |
2299 | btrfs_handle_fs_error(fs_info, ret, |
2300 | "Failed to recover log tree"); | |
00246528 | 2301 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2302 | return ret; |
2303 | } | |
2304 | ||
bc98a42c | 2305 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2306 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2307 | if (ret) |
2308 | return ret; | |
2309 | } | |
2310 | ||
2311 | return 0; | |
2312 | } | |
2313 | ||
6bccf3ab | 2314 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2315 | { |
6bccf3ab | 2316 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2317 | struct btrfs_root *root; |
4bbcaa64 ES |
2318 | struct btrfs_key location; |
2319 | int ret; | |
2320 | ||
6bccf3ab JM |
2321 | BUG_ON(!fs_info->tree_root); |
2322 | ||
4bbcaa64 ES |
2323 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2324 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2325 | location.offset = 0; | |
2326 | ||
a4f3d2c4 | 2327 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2328 | if (IS_ERR(root)) { |
2329 | ret = PTR_ERR(root); | |
2330 | goto out; | |
2331 | } | |
a4f3d2c4 DS |
2332 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2333 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2334 | |
2335 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2336 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2337 | if (IS_ERR(root)) { |
2338 | ret = PTR_ERR(root); | |
2339 | goto out; | |
2340 | } | |
a4f3d2c4 DS |
2341 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2342 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2343 | btrfs_init_devices_late(fs_info); |
2344 | ||
2345 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 | 2346 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2347 | if (IS_ERR(root)) { |
2348 | ret = PTR_ERR(root); | |
2349 | goto out; | |
2350 | } | |
a4f3d2c4 DS |
2351 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2352 | fs_info->csum_root = root; | |
4bbcaa64 | 2353 | |
aeb935a4 QW |
2354 | /* |
2355 | * This tree can share blocks with some other fs tree during relocation | |
2356 | * and we need a proper setup by btrfs_get_fs_root | |
2357 | */ | |
56e9357a DS |
2358 | root = btrfs_get_fs_root(tree_root->fs_info, |
2359 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 QW |
2360 | if (IS_ERR(root)) { |
2361 | ret = PTR_ERR(root); | |
2362 | goto out; | |
2363 | } | |
2364 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2365 | fs_info->data_reloc_root = root; | |
2366 | ||
4bbcaa64 | 2367 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2368 | root = btrfs_read_tree_root(tree_root, &location); |
2369 | if (!IS_ERR(root)) { | |
2370 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2371 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2372 | fs_info->quota_root = root; |
4bbcaa64 ES |
2373 | } |
2374 | ||
2375 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2376 | root = btrfs_read_tree_root(tree_root, &location); |
2377 | if (IS_ERR(root)) { | |
2378 | ret = PTR_ERR(root); | |
4bbcaa64 | 2379 | if (ret != -ENOENT) |
f50f4353 | 2380 | goto out; |
4bbcaa64 | 2381 | } else { |
a4f3d2c4 DS |
2382 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2383 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2384 | } |
2385 | ||
70f6d82e OS |
2386 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2387 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2388 | root = btrfs_read_tree_root(tree_root, &location); | |
f50f4353 LB |
2389 | if (IS_ERR(root)) { |
2390 | ret = PTR_ERR(root); | |
2391 | goto out; | |
2392 | } | |
70f6d82e OS |
2393 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2394 | fs_info->free_space_root = root; | |
2395 | } | |
2396 | ||
4bbcaa64 | 2397 | return 0; |
f50f4353 LB |
2398 | out: |
2399 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2400 | location.objectid, ret); | |
2401 | return ret; | |
4bbcaa64 ES |
2402 | } |
2403 | ||
069ec957 QW |
2404 | /* |
2405 | * Real super block validation | |
2406 | * NOTE: super csum type and incompat features will not be checked here. | |
2407 | * | |
2408 | * @sb: super block to check | |
2409 | * @mirror_num: the super block number to check its bytenr: | |
2410 | * 0 the primary (1st) sb | |
2411 | * 1, 2 2nd and 3rd backup copy | |
2412 | * -1 skip bytenr check | |
2413 | */ | |
2414 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2415 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2416 | { |
21a852b0 QW |
2417 | u64 nodesize = btrfs_super_nodesize(sb); |
2418 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2419 | int ret = 0; | |
2420 | ||
2421 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2422 | btrfs_err(fs_info, "no valid FS found"); | |
2423 | ret = -EINVAL; | |
2424 | } | |
2425 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2426 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2427 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2428 | ret = -EINVAL; | |
2429 | } | |
2430 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2431 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2432 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2433 | ret = -EINVAL; | |
2434 | } | |
2435 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2436 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2437 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2438 | ret = -EINVAL; | |
2439 | } | |
2440 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2441 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2442 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2443 | ret = -EINVAL; | |
2444 | } | |
2445 | ||
2446 | /* | |
2447 | * Check sectorsize and nodesize first, other check will need it. | |
2448 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2449 | */ | |
2450 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2451 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2452 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2453 | ret = -EINVAL; | |
2454 | } | |
2455 | /* Only PAGE SIZE is supported yet */ | |
2456 | if (sectorsize != PAGE_SIZE) { | |
2457 | btrfs_err(fs_info, | |
2458 | "sectorsize %llu not supported yet, only support %lu", | |
2459 | sectorsize, PAGE_SIZE); | |
2460 | ret = -EINVAL; | |
2461 | } | |
2462 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
2463 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2464 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2465 | ret = -EINVAL; | |
2466 | } | |
2467 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2468 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2469 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2470 | ret = -EINVAL; | |
2471 | } | |
2472 | ||
2473 | /* Root alignment check */ | |
2474 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2475 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2476 | btrfs_super_root(sb)); | |
2477 | ret = -EINVAL; | |
2478 | } | |
2479 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2480 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2481 | btrfs_super_chunk_root(sb)); | |
2482 | ret = -EINVAL; | |
2483 | } | |
2484 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2485 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2486 | btrfs_super_log_root(sb)); | |
2487 | ret = -EINVAL; | |
2488 | } | |
2489 | ||
de37aa51 | 2490 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2491 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2492 | btrfs_err(fs_info, |
7239ff4b | 2493 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2494 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2495 | ret = -EINVAL; |
2496 | } | |
2497 | ||
2498 | /* | |
2499 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2500 | * done later | |
2501 | */ | |
2502 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2503 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2504 | btrfs_super_bytes_used(sb)); | |
2505 | ret = -EINVAL; | |
2506 | } | |
2507 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2508 | btrfs_err(fs_info, "invalid stripesize %u", | |
2509 | btrfs_super_stripesize(sb)); | |
2510 | ret = -EINVAL; | |
2511 | } | |
2512 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2513 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2514 | btrfs_super_num_devices(sb)); | |
2515 | if (btrfs_super_num_devices(sb) == 0) { | |
2516 | btrfs_err(fs_info, "number of devices is 0"); | |
2517 | ret = -EINVAL; | |
2518 | } | |
2519 | ||
069ec957 QW |
2520 | if (mirror_num >= 0 && |
2521 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2522 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2523 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2524 | ret = -EINVAL; | |
2525 | } | |
2526 | ||
2527 | /* | |
2528 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2529 | * and one chunk | |
2530 | */ | |
2531 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2532 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2533 | btrfs_super_sys_array_size(sb), | |
2534 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2535 | ret = -EINVAL; | |
2536 | } | |
2537 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2538 | + sizeof(struct btrfs_chunk)) { | |
2539 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2540 | btrfs_super_sys_array_size(sb), | |
2541 | sizeof(struct btrfs_disk_key) | |
2542 | + sizeof(struct btrfs_chunk)); | |
2543 | ret = -EINVAL; | |
2544 | } | |
2545 | ||
2546 | /* | |
2547 | * The generation is a global counter, we'll trust it more than the others | |
2548 | * but it's still possible that it's the one that's wrong. | |
2549 | */ | |
2550 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2551 | btrfs_warn(fs_info, | |
2552 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2553 | btrfs_super_generation(sb), | |
2554 | btrfs_super_chunk_root_generation(sb)); | |
2555 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2556 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2557 | btrfs_warn(fs_info, | |
2558 | "suspicious: generation < cache_generation: %llu < %llu", | |
2559 | btrfs_super_generation(sb), | |
2560 | btrfs_super_cache_generation(sb)); | |
2561 | ||
2562 | return ret; | |
2563 | } | |
2564 | ||
069ec957 QW |
2565 | /* |
2566 | * Validation of super block at mount time. | |
2567 | * Some checks already done early at mount time, like csum type and incompat | |
2568 | * flags will be skipped. | |
2569 | */ | |
2570 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2571 | { | |
2572 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2573 | } | |
2574 | ||
75cb857d QW |
2575 | /* |
2576 | * Validation of super block at write time. | |
2577 | * Some checks like bytenr check will be skipped as their values will be | |
2578 | * overwritten soon. | |
2579 | * Extra checks like csum type and incompat flags will be done here. | |
2580 | */ | |
2581 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2582 | struct btrfs_super_block *sb) | |
2583 | { | |
2584 | int ret; | |
2585 | ||
2586 | ret = validate_super(fs_info, sb, -1); | |
2587 | if (ret < 0) | |
2588 | goto out; | |
e7e16f48 | 2589 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2590 | ret = -EUCLEAN; |
2591 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2592 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2593 | goto out; | |
2594 | } | |
2595 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2596 | ret = -EUCLEAN; | |
2597 | btrfs_err(fs_info, | |
2598 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2599 | btrfs_super_incompat_flags(sb), | |
2600 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2601 | goto out; | |
2602 | } | |
2603 | out: | |
2604 | if (ret < 0) | |
2605 | btrfs_err(fs_info, | |
2606 | "super block corruption detected before writing it to disk"); | |
2607 | return ret; | |
2608 | } | |
2609 | ||
6ef108dd | 2610 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2611 | { |
6ef108dd | 2612 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2613 | struct btrfs_super_block *sb = fs_info->super_copy; |
2614 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2615 | bool handle_error = false; | |
2616 | int ret = 0; | |
2617 | int i; | |
2618 | ||
2619 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2620 | u64 generation; | |
2621 | int level; | |
2622 | ||
2623 | if (handle_error) { | |
2624 | if (!IS_ERR(tree_root->node)) | |
2625 | free_extent_buffer(tree_root->node); | |
2626 | tree_root->node = NULL; | |
2627 | ||
2628 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2629 | break; | |
2630 | ||
2631 | free_root_pointers(fs_info, 0); | |
2632 | ||
2633 | /* | |
2634 | * Don't use the log in recovery mode, it won't be | |
2635 | * valid | |
2636 | */ | |
2637 | btrfs_set_super_log_root(sb, 0); | |
2638 | ||
2639 | /* We can't trust the free space cache either */ | |
2640 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2641 | ||
2642 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2643 | backup_index = ret; |
b8522a1e NB |
2644 | if (ret < 0) |
2645 | return ret; | |
2646 | } | |
2647 | generation = btrfs_super_generation(sb); | |
2648 | level = btrfs_super_root_level(sb); | |
2649 | tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), | |
2650 | generation, level, NULL); | |
2651 | if (IS_ERR(tree_root->node) || | |
2652 | !extent_buffer_uptodate(tree_root->node)) { | |
2653 | handle_error = true; | |
2654 | ||
0465337c | 2655 | if (IS_ERR(tree_root->node)) { |
b8522a1e | 2656 | ret = PTR_ERR(tree_root->node); |
0465337c JB |
2657 | tree_root->node = NULL; |
2658 | } else if (!extent_buffer_uptodate(tree_root->node)) { | |
b8522a1e | 2659 | ret = -EUCLEAN; |
0465337c | 2660 | } |
b8522a1e NB |
2661 | |
2662 | btrfs_warn(fs_info, "failed to read tree root"); | |
2663 | continue; | |
2664 | } | |
2665 | ||
2666 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); | |
2667 | tree_root->commit_root = btrfs_root_node(tree_root); | |
2668 | btrfs_set_root_refs(&tree_root->root_item, 1); | |
2669 | ||
336a0d8d NB |
2670 | /* |
2671 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2672 | * hasn't been fully initialised and we are the only user | |
2673 | */ | |
b8522a1e NB |
2674 | ret = btrfs_find_highest_objectid(tree_root, |
2675 | &tree_root->highest_objectid); | |
2676 | if (ret < 0) { | |
b8522a1e NB |
2677 | handle_error = true; |
2678 | continue; | |
2679 | } | |
2680 | ||
2681 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
b8522a1e NB |
2682 | |
2683 | ret = btrfs_read_roots(fs_info); | |
2684 | if (ret < 0) { | |
2685 | handle_error = true; | |
2686 | continue; | |
2687 | } | |
2688 | ||
2689 | /* All successful */ | |
2690 | fs_info->generation = generation; | |
2691 | fs_info->last_trans_committed = generation; | |
6ef108dd NB |
2692 | |
2693 | /* Always begin writing backup roots after the one being used */ | |
2694 | if (backup_index < 0) { | |
2695 | fs_info->backup_root_index = 0; | |
2696 | } else { | |
2697 | fs_info->backup_root_index = backup_index + 1; | |
2698 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2699 | } | |
b8522a1e NB |
2700 | break; |
2701 | } | |
2702 | ||
2703 | return ret; | |
2704 | } | |
2705 | ||
8260edba | 2706 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2707 | { |
76dda93c | 2708 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2709 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2710 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2711 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2712 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2713 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2714 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2715 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2716 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2717 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2718 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2719 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2720 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2721 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2722 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2723 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2724 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2725 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2726 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2727 | mutex_init(&fs_info->delalloc_root_mutex); |
de98ced9 | 2728 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2729 | |
0b86a832 | 2730 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2731 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2732 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2733 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
bd647ce3 JB |
2734 | #ifdef CONFIG_BTRFS_DEBUG |
2735 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2736 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2737 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2738 | #endif |
c8bf1b67 | 2739 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2740 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2741 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2742 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2743 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2744 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2745 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2746 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2747 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2748 | BTRFS_BLOCK_RSV_DELREFS); | |
2749 | ||
771ed689 | 2750 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2751 | atomic_set(&fs_info->defrag_running, 0); |
2fefd558 | 2752 | atomic_set(&fs_info->reada_works_cnt, 0); |
034f784d | 2753 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2754 | atomic64_set(&fs_info->tree_mod_seq, 0); |
95ac567a | 2755 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2756 | fs_info->metadata_ratio = 0; |
4cb5300b | 2757 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2758 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2759 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2760 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2761 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2762 | /* readahead state */ |
d0164adc | 2763 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2764 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2765 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2766 | |
b34b086c CM |
2767 | fs_info->thread_pool_size = min_t(unsigned long, |
2768 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2769 | |
199c2a9c MX |
2770 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2771 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2772 | |
638aa7ed | 2773 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2774 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2775 | fs_info->check_integrity_print_mask = 0; | |
2776 | #endif | |
779a65a4 | 2777 | btrfs_init_balance(fs_info); |
21c7e756 | 2778 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2779 | |
0f9dd46c | 2780 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2781 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2782 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2783 | |
fe119a6e NB |
2784 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
2785 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
afcdd129 | 2786 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2787 | |
5a3f23d5 | 2788 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2789 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2790 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2791 | mutex_init(&fs_info->transaction_kthread_mutex); |
2792 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2793 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2794 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2795 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2796 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2797 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2798 | |
ad618368 | 2799 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2800 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2801 | btrfs_discard_init(fs_info); |
416ac51d | 2802 | |
fa9c0d79 CM |
2803 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2804 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2805 | ||
e6dcd2dc | 2806 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2807 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2808 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2809 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2810 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2811 | |
da17066c JM |
2812 | /* Usable values until the real ones are cached from the superblock */ |
2813 | fs_info->nodesize = 4096; | |
2814 | fs_info->sectorsize = 4096; | |
2815 | fs_info->stripesize = 4096; | |
2816 | ||
eede2bf3 OS |
2817 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2818 | fs_info->swapfile_pins = RB_ROOT; | |
2819 | ||
9e967495 | 2820 | fs_info->send_in_progress = 0; |
8260edba JB |
2821 | } |
2822 | ||
2823 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2824 | { | |
2825 | int ret; | |
2826 | ||
2827 | fs_info->sb = sb; | |
2828 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
2829 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 2830 | |
ae18c37a JB |
2831 | ret = percpu_counter_init(&fs_info->dio_bytes, 0, GFP_KERNEL); |
2832 | if (ret) | |
c75e8394 | 2833 | return ret; |
ae18c37a JB |
2834 | |
2835 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
2836 | if (ret) | |
c75e8394 | 2837 | return ret; |
ae18c37a JB |
2838 | |
2839 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
2840 | (1 + ilog2(nr_cpu_ids)); | |
2841 | ||
2842 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
2843 | if (ret) | |
c75e8394 | 2844 | return ret; |
ae18c37a JB |
2845 | |
2846 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
2847 | GFP_KERNEL); | |
2848 | if (ret) | |
c75e8394 | 2849 | return ret; |
ae18c37a JB |
2850 | |
2851 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
2852 | GFP_KERNEL); | |
c75e8394 JB |
2853 | if (!fs_info->delayed_root) |
2854 | return -ENOMEM; | |
ae18c37a JB |
2855 | btrfs_init_delayed_root(fs_info->delayed_root); |
2856 | ||
c75e8394 | 2857 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
2858 | } |
2859 | ||
97f4dd09 NB |
2860 | static int btrfs_uuid_rescan_kthread(void *data) |
2861 | { | |
2862 | struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; | |
2863 | int ret; | |
2864 | ||
2865 | /* | |
2866 | * 1st step is to iterate through the existing UUID tree and | |
2867 | * to delete all entries that contain outdated data. | |
2868 | * 2nd step is to add all missing entries to the UUID tree. | |
2869 | */ | |
2870 | ret = btrfs_uuid_tree_iterate(fs_info); | |
2871 | if (ret < 0) { | |
c94bec2c JB |
2872 | if (ret != -EINTR) |
2873 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
2874 | ret); | |
97f4dd09 NB |
2875 | up(&fs_info->uuid_tree_rescan_sem); |
2876 | return ret; | |
2877 | } | |
2878 | return btrfs_uuid_scan_kthread(data); | |
2879 | } | |
2880 | ||
2881 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
2882 | { | |
2883 | struct task_struct *task; | |
2884 | ||
2885 | down(&fs_info->uuid_tree_rescan_sem); | |
2886 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
2887 | if (IS_ERR(task)) { | |
2888 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
2889 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
2890 | up(&fs_info->uuid_tree_rescan_sem); | |
2891 | return PTR_ERR(task); | |
2892 | } | |
2893 | ||
2894 | return 0; | |
2895 | } | |
2896 | ||
ae18c37a JB |
2897 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
2898 | char *options) | |
2899 | { | |
2900 | u32 sectorsize; | |
2901 | u32 nodesize; | |
2902 | u32 stripesize; | |
2903 | u64 generation; | |
2904 | u64 features; | |
2905 | u16 csum_type; | |
ae18c37a JB |
2906 | struct btrfs_super_block *disk_super; |
2907 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
2908 | struct btrfs_root *tree_root; | |
2909 | struct btrfs_root *chunk_root; | |
2910 | int ret; | |
2911 | int err = -EINVAL; | |
2912 | int clear_free_space_tree = 0; | |
2913 | int level; | |
2914 | ||
8260edba | 2915 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 2916 | if (ret) { |
83c8266a | 2917 | err = ret; |
ae18c37a | 2918 | goto fail; |
53b381b3 DW |
2919 | } |
2920 | ||
ae18c37a JB |
2921 | /* These need to be init'ed before we start creating inodes and such. */ |
2922 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
2923 | GFP_KERNEL); | |
2924 | fs_info->tree_root = tree_root; | |
2925 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
2926 | GFP_KERNEL); | |
2927 | fs_info->chunk_root = chunk_root; | |
2928 | if (!tree_root || !chunk_root) { | |
2929 | err = -ENOMEM; | |
c75e8394 | 2930 | goto fail; |
ae18c37a JB |
2931 | } |
2932 | ||
2933 | fs_info->btree_inode = new_inode(sb); | |
2934 | if (!fs_info->btree_inode) { | |
2935 | err = -ENOMEM; | |
c75e8394 | 2936 | goto fail; |
ae18c37a JB |
2937 | } |
2938 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
2939 | btrfs_init_btree_inode(fs_info); | |
2940 | ||
3c4bb26b | 2941 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2942 | |
2943 | /* | |
2944 | * Read super block and check the signature bytes only | |
2945 | */ | |
8f32380d JT |
2946 | disk_super = btrfs_read_dev_super(fs_devices->latest_bdev); |
2947 | if (IS_ERR(disk_super)) { | |
2948 | err = PTR_ERR(disk_super); | |
16cdcec7 | 2949 | goto fail_alloc; |
20b45077 | 2950 | } |
39279cc3 | 2951 | |
8dc3f22c JT |
2952 | /* |
2953 | * Verify the type first, if that or the the checksum value are | |
2954 | * corrupted, we'll find out | |
2955 | */ | |
8f32380d | 2956 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 2957 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 2958 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 2959 | csum_type); |
8dc3f22c | 2960 | err = -EINVAL; |
8f32380d | 2961 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
2962 | goto fail_alloc; |
2963 | } | |
2964 | ||
6d97c6e3 JT |
2965 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
2966 | if (ret) { | |
2967 | err = ret; | |
8f32380d | 2968 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
2969 | goto fail_alloc; |
2970 | } | |
2971 | ||
1104a885 DS |
2972 | /* |
2973 | * We want to check superblock checksum, the type is stored inside. | |
2974 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2975 | */ | |
8f32380d | 2976 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 2977 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2978 | err = -EINVAL; |
8f32380d | 2979 | btrfs_release_disk_super(disk_super); |
141386e1 | 2980 | goto fail_alloc; |
1104a885 DS |
2981 | } |
2982 | ||
2983 | /* | |
2984 | * super_copy is zeroed at allocation time and we never touch the | |
2985 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2986 | * the whole block of INFO_SIZE | |
2987 | */ | |
8f32380d JT |
2988 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
2989 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 2990 | |
fbc6feae NB |
2991 | disk_super = fs_info->super_copy; |
2992 | ||
de37aa51 NB |
2993 | ASSERT(!memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2994 | BTRFS_FSID_SIZE)); | |
2995 | ||
7239ff4b | 2996 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) { |
de37aa51 NB |
2997 | ASSERT(!memcmp(fs_info->fs_devices->metadata_uuid, |
2998 | fs_info->super_copy->metadata_uuid, | |
2999 | BTRFS_FSID_SIZE)); | |
7239ff4b | 3000 | } |
0b86a832 | 3001 | |
fbc6feae NB |
3002 | features = btrfs_super_flags(disk_super); |
3003 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3004 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3005 | btrfs_set_super_flags(disk_super, features); | |
3006 | btrfs_info(fs_info, | |
3007 | "found metadata UUID change in progress flag, clearing"); | |
3008 | } | |
3009 | ||
3010 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3011 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3012 | |
069ec957 | 3013 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3014 | if (ret) { |
05135f59 | 3015 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3016 | err = -EINVAL; |
141386e1 | 3017 | goto fail_alloc; |
1104a885 DS |
3018 | } |
3019 | ||
0f7d52f4 | 3020 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3021 | goto fail_alloc; |
0f7d52f4 | 3022 | |
acce952b | 3023 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3024 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3025 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3026 | |
75e7cb7f LB |
3027 | /* |
3028 | * In the long term, we'll store the compression type in the super | |
3029 | * block, and it'll be used for per file compression control. | |
3030 | */ | |
3031 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3032 | ||
2ff7e61e | 3033 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3034 | if (ret) { |
3035 | err = ret; | |
141386e1 | 3036 | goto fail_alloc; |
2b82032c | 3037 | } |
dfe25020 | 3038 | |
f2b636e8 JB |
3039 | features = btrfs_super_incompat_flags(disk_super) & |
3040 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3041 | if (features) { | |
05135f59 DS |
3042 | btrfs_err(fs_info, |
3043 | "cannot mount because of unsupported optional features (%llx)", | |
3044 | features); | |
f2b636e8 | 3045 | err = -EINVAL; |
141386e1 | 3046 | goto fail_alloc; |
f2b636e8 JB |
3047 | } |
3048 | ||
5d4f98a2 | 3049 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3050 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3051 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3052 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3053 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3054 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3055 | |
3173a18f | 3056 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3057 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3058 | |
727011e0 CM |
3059 | /* |
3060 | * flag our filesystem as having big metadata blocks if | |
3061 | * they are bigger than the page size | |
3062 | */ | |
09cbfeaf | 3063 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 3064 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
3065 | btrfs_info(fs_info, |
3066 | "flagging fs with big metadata feature"); | |
727011e0 CM |
3067 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
3068 | } | |
3069 | ||
bc3f116f | 3070 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 3071 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 3072 | stripesize = sectorsize; |
707e8a07 | 3073 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 3074 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f | 3075 | |
da17066c JM |
3076 | /* Cache block sizes */ |
3077 | fs_info->nodesize = nodesize; | |
3078 | fs_info->sectorsize = sectorsize; | |
3079 | fs_info->stripesize = stripesize; | |
3080 | ||
bc3f116f CM |
3081 | /* |
3082 | * mixed block groups end up with duplicate but slightly offset | |
3083 | * extent buffers for the same range. It leads to corruptions | |
3084 | */ | |
3085 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3086 | (sectorsize != nodesize)) { |
05135f59 DS |
3087 | btrfs_err(fs_info, |
3088 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3089 | nodesize, sectorsize); | |
141386e1 | 3090 | goto fail_alloc; |
bc3f116f CM |
3091 | } |
3092 | ||
ceda0864 MX |
3093 | /* |
3094 | * Needn't use the lock because there is no other task which will | |
3095 | * update the flag. | |
3096 | */ | |
a6fa6fae | 3097 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3098 | |
f2b636e8 JB |
3099 | features = btrfs_super_compat_ro_flags(disk_super) & |
3100 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3101 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
3102 | btrfs_err(fs_info, |
3103 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 3104 | features); |
f2b636e8 | 3105 | err = -EINVAL; |
141386e1 | 3106 | goto fail_alloc; |
f2b636e8 | 3107 | } |
61d92c32 | 3108 | |
2a458198 ES |
3109 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
3110 | if (ret) { | |
3111 | err = ret; | |
0dc3b84a JB |
3112 | goto fail_sb_buffer; |
3113 | } | |
4543df7e | 3114 | |
9e11ceee JK |
3115 | sb->s_bdi->congested_fn = btrfs_congested_fn; |
3116 | sb->s_bdi->congested_data = fs_info; | |
3117 | sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; | |
b5420237 | 3118 | sb->s_bdi->ra_pages = VM_READAHEAD_PAGES; |
9e11ceee JK |
3119 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3120 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3121 | |
a061fc8d CM |
3122 | sb->s_blocksize = sectorsize; |
3123 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3124 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3125 | |
925baedd | 3126 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3127 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3128 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3129 | if (ret) { |
05135f59 | 3130 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3131 | goto fail_sb_buffer; |
84eed90f | 3132 | } |
0b86a832 | 3133 | |
84234f3a | 3134 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3135 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3136 | |
2ff7e61e | 3137 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3138 | btrfs_super_chunk_root(disk_super), |
581c1760 | 3139 | generation, level, NULL); |
64c043de LB |
3140 | if (IS_ERR(chunk_root->node) || |
3141 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3142 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3143 | if (!IS_ERR(chunk_root->node)) |
3144 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3145 | chunk_root->node = NULL; |
af31f5e5 | 3146 | goto fail_tree_roots; |
83121942 | 3147 | } |
5d4f98a2 YZ |
3148 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3149 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3150 | |
e17cade2 | 3151 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3152 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3153 | BTRFS_UUID_SIZE); | |
e17cade2 | 3154 | |
5b4aacef | 3155 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3156 | if (ret) { |
05135f59 | 3157 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3158 | goto fail_tree_roots; |
2b82032c | 3159 | } |
0b86a832 | 3160 | |
8dabb742 | 3161 | /* |
9b99b115 AJ |
3162 | * Keep the devid that is marked to be the target device for the |
3163 | * device replace procedure | |
8dabb742 | 3164 | */ |
9b99b115 | 3165 | btrfs_free_extra_devids(fs_devices, 0); |
dfe25020 | 3166 | |
a6b0d5c8 | 3167 | if (!fs_devices->latest_bdev) { |
05135f59 | 3168 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3169 | goto fail_tree_roots; |
3170 | } | |
3171 | ||
b8522a1e | 3172 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3173 | if (ret) |
b8522a1e | 3174 | goto fail_tree_roots; |
8929ecfa | 3175 | |
75ec1db8 JB |
3176 | /* |
3177 | * If we have a uuid root and we're not being told to rescan we need to | |
3178 | * check the generation here so we can set the | |
3179 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3180 | * transaction during a balance or the log replay without updating the | |
3181 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3182 | * even though it was perfectly fine. | |
3183 | */ | |
3184 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3185 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3186 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3187 | ||
cf90d884 QW |
3188 | ret = btrfs_verify_dev_extents(fs_info); |
3189 | if (ret) { | |
3190 | btrfs_err(fs_info, | |
3191 | "failed to verify dev extents against chunks: %d", | |
3192 | ret); | |
3193 | goto fail_block_groups; | |
3194 | } | |
68310a5e ID |
3195 | ret = btrfs_recover_balance(fs_info); |
3196 | if (ret) { | |
05135f59 | 3197 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3198 | goto fail_block_groups; |
3199 | } | |
3200 | ||
733f4fbb SB |
3201 | ret = btrfs_init_dev_stats(fs_info); |
3202 | if (ret) { | |
05135f59 | 3203 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3204 | goto fail_block_groups; |
3205 | } | |
3206 | ||
8dabb742 SB |
3207 | ret = btrfs_init_dev_replace(fs_info); |
3208 | if (ret) { | |
05135f59 | 3209 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3210 | goto fail_block_groups; |
3211 | } | |
3212 | ||
9b99b115 | 3213 | btrfs_free_extra_devids(fs_devices, 1); |
8dabb742 | 3214 | |
c6761a9e | 3215 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3216 | if (ret) { |
05135f59 DS |
3217 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3218 | ret); | |
b7c35e81 AJ |
3219 | goto fail_block_groups; |
3220 | } | |
3221 | ||
96f3136e | 3222 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3223 | if (ret) { |
05135f59 | 3224 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3225 | goto fail_fsdev_sysfs; |
c59021f8 | 3226 | } |
3227 | ||
c59021f8 | 3228 | ret = btrfs_init_space_info(fs_info); |
3229 | if (ret) { | |
05135f59 | 3230 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3231 | goto fail_sysfs; |
c59021f8 | 3232 | } |
3233 | ||
5b4aacef | 3234 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3235 | if (ret) { |
05135f59 | 3236 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3237 | goto fail_sysfs; |
1b1d1f66 | 3238 | } |
4330e183 | 3239 | |
6528b99d | 3240 | if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { |
05135f59 | 3241 | btrfs_warn(fs_info, |
52042d8e | 3242 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3243 | goto fail_sysfs; |
292fd7fc | 3244 | } |
9078a3e1 | 3245 | |
a74a4b97 CM |
3246 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3247 | "btrfs-cleaner"); | |
57506d50 | 3248 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3249 | goto fail_sysfs; |
a74a4b97 CM |
3250 | |
3251 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3252 | tree_root, | |
3253 | "btrfs-transaction"); | |
57506d50 | 3254 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3255 | goto fail_cleaner; |
a74a4b97 | 3256 | |
583b7231 | 3257 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3258 | !fs_info->fs_devices->rotating) { |
583b7231 | 3259 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3260 | } |
3261 | ||
572d9ab7 | 3262 | /* |
01327610 | 3263 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3264 | * not have to wait for transaction commit |
3265 | */ | |
3266 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3267 | |
21adbd5c | 3268 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3269 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3270 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3271 | btrfs_test_opt(fs_info, |
21adbd5c SB |
3272 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3273 | 1 : 0, | |
3274 | fs_info->check_integrity_print_mask); | |
3275 | if (ret) | |
05135f59 DS |
3276 | btrfs_warn(fs_info, |
3277 | "failed to initialize integrity check module: %d", | |
3278 | ret); | |
21adbd5c SB |
3279 | } |
3280 | #endif | |
bcef60f2 AJ |
3281 | ret = btrfs_read_qgroup_config(fs_info); |
3282 | if (ret) | |
3283 | goto fail_trans_kthread; | |
21adbd5c | 3284 | |
fd708b81 JB |
3285 | if (btrfs_build_ref_tree(fs_info)) |
3286 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3287 | ||
96da0919 QW |
3288 | /* do not make disk changes in broken FS or nologreplay is given */ |
3289 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3290 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3291 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3292 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3293 | if (ret) { |
63443bf5 | 3294 | err = ret; |
28c16cbb | 3295 | goto fail_qgroup; |
79787eaa | 3296 | } |
e02119d5 | 3297 | } |
1a40e23b | 3298 | |
6bccf3ab | 3299 | ret = btrfs_find_orphan_roots(fs_info); |
79787eaa | 3300 | if (ret) |
28c16cbb | 3301 | goto fail_qgroup; |
76dda93c | 3302 | |
bc98a42c | 3303 | if (!sb_rdonly(sb)) { |
d68fc57b | 3304 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3305 | if (ret) |
28c16cbb | 3306 | goto fail_qgroup; |
90c711ab ZB |
3307 | |
3308 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3309 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3310 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3311 | if (ret < 0) { |
05135f59 DS |
3312 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3313 | ret); | |
d7ce5843 | 3314 | err = -EINVAL; |
bcef60f2 | 3315 | goto fail_qgroup; |
d7ce5843 | 3316 | } |
7c2ca468 | 3317 | } |
1a40e23b | 3318 | |
56e9357a | 3319 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3320 | if (IS_ERR(fs_info->fs_root)) { |
3321 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3322 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3323 | fs_info->fs_root = NULL; |
bcef60f2 | 3324 | goto fail_qgroup; |
3140c9a3 | 3325 | } |
c289811c | 3326 | |
bc98a42c | 3327 | if (sb_rdonly(sb)) |
2b6ba629 | 3328 | return 0; |
59641015 | 3329 | |
f8d468a1 OS |
3330 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && |
3331 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
6675df31 OS |
3332 | clear_free_space_tree = 1; |
3333 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3334 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3335 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3336 | clear_free_space_tree = 1; | |
3337 | } | |
3338 | ||
3339 | if (clear_free_space_tree) { | |
f8d468a1 OS |
3340 | btrfs_info(fs_info, "clearing free space tree"); |
3341 | ret = btrfs_clear_free_space_tree(fs_info); | |
3342 | if (ret) { | |
3343 | btrfs_warn(fs_info, | |
3344 | "failed to clear free space tree: %d", ret); | |
6bccf3ab | 3345 | close_ctree(fs_info); |
f8d468a1 OS |
3346 | return ret; |
3347 | } | |
3348 | } | |
3349 | ||
0b246afa | 3350 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
511711af | 3351 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3352 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3353 | ret = btrfs_create_free_space_tree(fs_info); |
3354 | if (ret) { | |
05135f59 DS |
3355 | btrfs_warn(fs_info, |
3356 | "failed to create free space tree: %d", ret); | |
6bccf3ab | 3357 | close_ctree(fs_info); |
511711af CM |
3358 | return ret; |
3359 | } | |
3360 | } | |
3361 | ||
2b6ba629 ID |
3362 | down_read(&fs_info->cleanup_work_sem); |
3363 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3364 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3365 | up_read(&fs_info->cleanup_work_sem); |
6bccf3ab | 3366 | close_ctree(fs_info); |
2b6ba629 ID |
3367 | return ret; |
3368 | } | |
3369 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3370 | |
2b6ba629 ID |
3371 | ret = btrfs_resume_balance_async(fs_info); |
3372 | if (ret) { | |
05135f59 | 3373 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
6bccf3ab | 3374 | close_ctree(fs_info); |
2b6ba629 | 3375 | return ret; |
e3acc2a6 JB |
3376 | } |
3377 | ||
8dabb742 SB |
3378 | ret = btrfs_resume_dev_replace_async(fs_info); |
3379 | if (ret) { | |
05135f59 | 3380 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
6bccf3ab | 3381 | close_ctree(fs_info); |
8dabb742 SB |
3382 | return ret; |
3383 | } | |
3384 | ||
b382a324 | 3385 | btrfs_qgroup_rescan_resume(fs_info); |
b0643e59 | 3386 | btrfs_discard_resume(fs_info); |
b382a324 | 3387 | |
4bbcaa64 | 3388 | if (!fs_info->uuid_root) { |
05135f59 | 3389 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3390 | ret = btrfs_create_uuid_tree(fs_info); |
3391 | if (ret) { | |
05135f59 DS |
3392 | btrfs_warn(fs_info, |
3393 | "failed to create the UUID tree: %d", ret); | |
6bccf3ab | 3394 | close_ctree(fs_info); |
f7a81ea4 SB |
3395 | return ret; |
3396 | } | |
0b246afa | 3397 | } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3398 | fs_info->generation != |
3399 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3400 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3401 | ret = btrfs_check_uuid_tree(fs_info); |
3402 | if (ret) { | |
05135f59 DS |
3403 | btrfs_warn(fs_info, |
3404 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3405 | close_ctree(fs_info); |
70f80175 SB |
3406 | return ret; |
3407 | } | |
f7a81ea4 | 3408 | } |
afcdd129 | 3409 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3410 | |
8dcddfa0 QW |
3411 | /* |
3412 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3413 | * no need to keep the flag | |
3414 | */ | |
3415 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3416 | ||
ad2b2c80 | 3417 | return 0; |
39279cc3 | 3418 | |
bcef60f2 AJ |
3419 | fail_qgroup: |
3420 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3421 | fail_trans_kthread: |
3422 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3423 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3424 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3425 | fail_cleaner: |
a74a4b97 | 3426 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3427 | |
3428 | /* | |
3429 | * make sure we're done with the btree inode before we stop our | |
3430 | * kthreads | |
3431 | */ | |
3432 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3433 | |
2365dd3c | 3434 | fail_sysfs: |
6618a59b | 3435 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3436 | |
b7c35e81 AJ |
3437 | fail_fsdev_sysfs: |
3438 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3439 | ||
1b1d1f66 | 3440 | fail_block_groups: |
54067ae9 | 3441 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3442 | |
3443 | fail_tree_roots: | |
4273eaff | 3444 | free_root_pointers(fs_info, true); |
2b8195bb | 3445 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3446 | |
39279cc3 | 3447 | fail_sb_buffer: |
7abadb64 | 3448 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3449 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3450 | fail_alloc: |
586e46e2 ID |
3451 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3452 | ||
4543df7e | 3453 | iput(fs_info->btree_inode); |
7e662854 | 3454 | fail: |
586e46e2 | 3455 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3456 | return err; |
eb60ceac | 3457 | } |
663faf9f | 3458 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3459 | |
314b6dd0 | 3460 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3461 | { |
314b6dd0 JT |
3462 | struct btrfs_device *device = bio->bi_private; |
3463 | struct bio_vec *bvec; | |
3464 | struct bvec_iter_all iter_all; | |
3465 | struct page *page; | |
3466 | ||
3467 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3468 | page = bvec->bv_page; | |
3469 | ||
3470 | if (bio->bi_status) { | |
3471 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3472 | "lost page write due to IO error on %s (%d)", | |
3473 | rcu_str_deref(device->name), | |
3474 | blk_status_to_errno(bio->bi_status)); | |
3475 | ClearPageUptodate(page); | |
3476 | SetPageError(page); | |
3477 | btrfs_dev_stat_inc_and_print(device, | |
3478 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3479 | } else { | |
3480 | SetPageUptodate(page); | |
3481 | } | |
3482 | ||
3483 | put_page(page); | |
3484 | unlock_page(page); | |
f2984462 | 3485 | } |
314b6dd0 JT |
3486 | |
3487 | bio_put(bio); | |
f2984462 CM |
3488 | } |
3489 | ||
8f32380d JT |
3490 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
3491 | int copy_num) | |
29c36d72 | 3492 | { |
29c36d72 | 3493 | struct btrfs_super_block *super; |
8f32380d | 3494 | struct page *page; |
29c36d72 | 3495 | u64 bytenr; |
8f32380d | 3496 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
29c36d72 AJ |
3497 | |
3498 | bytenr = btrfs_sb_offset(copy_num); | |
3499 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
8f32380d | 3500 | return ERR_PTR(-EINVAL); |
29c36d72 | 3501 | |
8f32380d JT |
3502 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3503 | if (IS_ERR(page)) | |
3504 | return ERR_CAST(page); | |
29c36d72 | 3505 | |
8f32380d | 3506 | super = page_address(page); |
29c36d72 AJ |
3507 | if (btrfs_super_bytenr(super) != bytenr || |
3508 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
8f32380d JT |
3509 | btrfs_release_disk_super(super); |
3510 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3511 | } |
3512 | ||
8f32380d | 3513 | return super; |
29c36d72 AJ |
3514 | } |
3515 | ||
3516 | ||
8f32380d | 3517 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3518 | { |
8f32380d | 3519 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3520 | int i; |
3521 | u64 transid = 0; | |
a512bbf8 YZ |
3522 | |
3523 | /* we would like to check all the supers, but that would make | |
3524 | * a btrfs mount succeed after a mkfs from a different FS. | |
3525 | * So, we need to add a special mount option to scan for | |
3526 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3527 | */ | |
3528 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
3529 | super = btrfs_read_dev_one_super(bdev, i); |
3530 | if (IS_ERR(super)) | |
a512bbf8 YZ |
3531 | continue; |
3532 | ||
a512bbf8 | 3533 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3534 | if (latest) |
3535 | btrfs_release_disk_super(super); | |
3536 | ||
3537 | latest = super; | |
a512bbf8 | 3538 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3539 | } |
3540 | } | |
92fc03fb | 3541 | |
8f32380d | 3542 | return super; |
a512bbf8 YZ |
3543 | } |
3544 | ||
4eedeb75 | 3545 | /* |
abbb3b8e | 3546 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3547 | * pages we use for writing are locked. |
4eedeb75 | 3548 | * |
abbb3b8e DS |
3549 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3550 | * the expected device size at commit time. Note that max_mirrors must be | |
3551 | * same for write and wait phases. | |
4eedeb75 | 3552 | * |
314b6dd0 | 3553 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3554 | */ |
a512bbf8 | 3555 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3556 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3557 | { |
d5178578 | 3558 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3559 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3560 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3561 | int i; |
a512bbf8 | 3562 | int errors = 0; |
a512bbf8 | 3563 | u64 bytenr; |
a512bbf8 YZ |
3564 | |
3565 | if (max_mirrors == 0) | |
3566 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3567 | ||
d5178578 JT |
3568 | shash->tfm = fs_info->csum_shash; |
3569 | ||
a512bbf8 | 3570 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3571 | struct page *page; |
3572 | struct bio *bio; | |
3573 | struct btrfs_super_block *disk_super; | |
3574 | ||
a512bbf8 | 3575 | bytenr = btrfs_sb_offset(i); |
935e5cc9 MX |
3576 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3577 | device->commit_total_bytes) | |
a512bbf8 YZ |
3578 | break; |
3579 | ||
abbb3b8e | 3580 | btrfs_set_super_bytenr(sb, bytenr); |
4eedeb75 | 3581 | |
fd08001f EB |
3582 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3583 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3584 | sb->csum); | |
4eedeb75 | 3585 | |
314b6dd0 JT |
3586 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3587 | GFP_NOFS); | |
3588 | if (!page) { | |
abbb3b8e | 3589 | btrfs_err(device->fs_info, |
314b6dd0 | 3590 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3591 | bytenr); |
3592 | errors++; | |
4eedeb75 | 3593 | continue; |
abbb3b8e | 3594 | } |
634554dc | 3595 | |
314b6dd0 JT |
3596 | /* Bump the refcount for wait_dev_supers() */ |
3597 | get_page(page); | |
a512bbf8 | 3598 | |
314b6dd0 JT |
3599 | disk_super = page_address(page); |
3600 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3601 | |
314b6dd0 JT |
3602 | /* |
3603 | * Directly use bios here instead of relying on the page cache | |
3604 | * to do I/O, so we don't lose the ability to do integrity | |
3605 | * checking. | |
3606 | */ | |
3607 | bio = bio_alloc(GFP_NOFS, 1); | |
3608 | bio_set_dev(bio, device->bdev); | |
3609 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; | |
3610 | bio->bi_private = device; | |
3611 | bio->bi_end_io = btrfs_end_super_write; | |
3612 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3613 | offset_in_page(bytenr)); | |
a512bbf8 | 3614 | |
387125fc | 3615 | /* |
314b6dd0 JT |
3616 | * We FUA only the first super block. The others we allow to |
3617 | * go down lazy and there's a short window where the on-disk | |
3618 | * copies might still contain the older version. | |
387125fc | 3619 | */ |
314b6dd0 | 3620 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; |
1b9e619c | 3621 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
3622 | bio->bi_opf |= REQ_FUA; |
3623 | ||
3624 | btrfsic_submit_bio(bio); | |
a512bbf8 YZ |
3625 | } |
3626 | return errors < i ? 0 : -1; | |
3627 | } | |
3628 | ||
abbb3b8e DS |
3629 | /* |
3630 | * Wait for write completion of superblocks done by write_dev_supers, | |
3631 | * @max_mirrors same for write and wait phases. | |
3632 | * | |
314b6dd0 | 3633 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3634 | * date. |
3635 | */ | |
3636 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3637 | { | |
abbb3b8e DS |
3638 | int i; |
3639 | int errors = 0; | |
b6a535fa | 3640 | bool primary_failed = false; |
abbb3b8e DS |
3641 | u64 bytenr; |
3642 | ||
3643 | if (max_mirrors == 0) | |
3644 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3645 | ||
3646 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3647 | struct page *page; |
3648 | ||
abbb3b8e DS |
3649 | bytenr = btrfs_sb_offset(i); |
3650 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
3651 | device->commit_total_bytes) | |
3652 | break; | |
3653 | ||
314b6dd0 JT |
3654 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3655 | bytenr >> PAGE_SHIFT); | |
3656 | if (!page) { | |
abbb3b8e | 3657 | errors++; |
b6a535fa HM |
3658 | if (i == 0) |
3659 | primary_failed = true; | |
abbb3b8e DS |
3660 | continue; |
3661 | } | |
314b6dd0 JT |
3662 | /* Page is submitted locked and unlocked once the IO completes */ |
3663 | wait_on_page_locked(page); | |
3664 | if (PageError(page)) { | |
abbb3b8e | 3665 | errors++; |
b6a535fa HM |
3666 | if (i == 0) |
3667 | primary_failed = true; | |
3668 | } | |
abbb3b8e | 3669 | |
314b6dd0 JT |
3670 | /* Drop our reference */ |
3671 | put_page(page); | |
abbb3b8e | 3672 | |
314b6dd0 JT |
3673 | /* Drop the reference from the writing run */ |
3674 | put_page(page); | |
abbb3b8e DS |
3675 | } |
3676 | ||
b6a535fa HM |
3677 | /* log error, force error return */ |
3678 | if (primary_failed) { | |
3679 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3680 | device->devid); | |
3681 | return -1; | |
3682 | } | |
3683 | ||
abbb3b8e DS |
3684 | return errors < i ? 0 : -1; |
3685 | } | |
3686 | ||
387125fc CM |
3687 | /* |
3688 | * endio for the write_dev_flush, this will wake anyone waiting | |
3689 | * for the barrier when it is done | |
3690 | */ | |
4246a0b6 | 3691 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3692 | { |
e0ae9994 | 3693 | complete(bio->bi_private); |
387125fc CM |
3694 | } |
3695 | ||
3696 | /* | |
4fc6441a AJ |
3697 | * Submit a flush request to the device if it supports it. Error handling is |
3698 | * done in the waiting counterpart. | |
387125fc | 3699 | */ |
4fc6441a | 3700 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3701 | { |
c2a9c7ab | 3702 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3703 | struct bio *bio = device->flush_bio; |
387125fc | 3704 | |
c2a9c7ab | 3705 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3706 | return; |
387125fc | 3707 | |
e0ae9994 | 3708 | bio_reset(bio); |
387125fc | 3709 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 3710 | bio_set_dev(bio, device->bdev); |
8d910125 | 3711 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3712 | init_completion(&device->flush_wait); |
3713 | bio->bi_private = &device->flush_wait; | |
387125fc | 3714 | |
43a01111 | 3715 | btrfsic_submit_bio(bio); |
1c3063b6 | 3716 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3717 | } |
387125fc | 3718 | |
4fc6441a AJ |
3719 | /* |
3720 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
3721 | */ | |
8c27cb35 | 3722 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3723 | { |
4fc6441a | 3724 | struct bio *bio = device->flush_bio; |
387125fc | 3725 | |
1c3063b6 | 3726 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 3727 | return BLK_STS_OK; |
387125fc | 3728 | |
1c3063b6 | 3729 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 3730 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3731 | |
8c27cb35 | 3732 | return bio->bi_status; |
387125fc | 3733 | } |
387125fc | 3734 | |
d10b82fe | 3735 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 3736 | { |
6528b99d | 3737 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 3738 | return -EIO; |
387125fc CM |
3739 | return 0; |
3740 | } | |
3741 | ||
3742 | /* | |
3743 | * send an empty flush down to each device in parallel, | |
3744 | * then wait for them | |
3745 | */ | |
3746 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3747 | { | |
3748 | struct list_head *head; | |
3749 | struct btrfs_device *dev; | |
5af3e8cc | 3750 | int errors_wait = 0; |
4e4cbee9 | 3751 | blk_status_t ret; |
387125fc | 3752 | |
1538e6c5 | 3753 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3754 | /* send down all the barriers */ |
3755 | head = &info->fs_devices->devices; | |
1538e6c5 | 3756 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3757 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3758 | continue; |
cea7c8bf | 3759 | if (!dev->bdev) |
387125fc | 3760 | continue; |
e12c9621 | 3761 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3762 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3763 | continue; |
3764 | ||
4fc6441a | 3765 | write_dev_flush(dev); |
58efbc9f | 3766 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
3767 | } |
3768 | ||
3769 | /* wait for all the barriers */ | |
1538e6c5 | 3770 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3771 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3772 | continue; |
387125fc | 3773 | if (!dev->bdev) { |
5af3e8cc | 3774 | errors_wait++; |
387125fc CM |
3775 | continue; |
3776 | } | |
e12c9621 | 3777 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3778 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3779 | continue; |
3780 | ||
4fc6441a | 3781 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
3782 | if (ret) { |
3783 | dev->last_flush_error = ret; | |
66b4993e DS |
3784 | btrfs_dev_stat_inc_and_print(dev, |
3785 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 3786 | errors_wait++; |
401b41e5 AJ |
3787 | } |
3788 | } | |
3789 | ||
cea7c8bf | 3790 | if (errors_wait) { |
401b41e5 AJ |
3791 | /* |
3792 | * At some point we need the status of all disks | |
3793 | * to arrive at the volume status. So error checking | |
3794 | * is being pushed to a separate loop. | |
3795 | */ | |
d10b82fe | 3796 | return check_barrier_error(info); |
387125fc | 3797 | } |
387125fc CM |
3798 | return 0; |
3799 | } | |
3800 | ||
943c6e99 ZL |
3801 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3802 | { | |
8789f4fe ZL |
3803 | int raid_type; |
3804 | int min_tolerated = INT_MAX; | |
943c6e99 | 3805 | |
8789f4fe ZL |
3806 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3807 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 3808 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
3809 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
3810 | tolerated_failures); | |
943c6e99 | 3811 | |
8789f4fe ZL |
3812 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3813 | if (raid_type == BTRFS_RAID_SINGLE) | |
3814 | continue; | |
41a6e891 | 3815 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 3816 | continue; |
8c3e3582 | 3817 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
3818 | btrfs_raid_array[raid_type]. |
3819 | tolerated_failures); | |
3820 | } | |
943c6e99 | 3821 | |
8789f4fe | 3822 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 3823 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
3824 | min_tolerated = 0; |
3825 | } | |
3826 | ||
3827 | return min_tolerated; | |
943c6e99 ZL |
3828 | } |
3829 | ||
eece6a9c | 3830 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 3831 | { |
e5e9a520 | 3832 | struct list_head *head; |
f2984462 | 3833 | struct btrfs_device *dev; |
a061fc8d | 3834 | struct btrfs_super_block *sb; |
f2984462 | 3835 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3836 | int ret; |
3837 | int do_barriers; | |
a236aed1 CM |
3838 | int max_errors; |
3839 | int total_errors = 0; | |
a061fc8d | 3840 | u64 flags; |
f2984462 | 3841 | |
0b246afa | 3842 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
3843 | |
3844 | /* | |
3845 | * max_mirrors == 0 indicates we're from commit_transaction, | |
3846 | * not from fsync where the tree roots in fs_info have not | |
3847 | * been consistent on disk. | |
3848 | */ | |
3849 | if (max_mirrors == 0) | |
3850 | backup_super_roots(fs_info); | |
f2984462 | 3851 | |
0b246afa | 3852 | sb = fs_info->super_for_commit; |
a061fc8d | 3853 | dev_item = &sb->dev_item; |
e5e9a520 | 3854 | |
0b246afa JM |
3855 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
3856 | head = &fs_info->fs_devices->devices; | |
3857 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 3858 | |
5af3e8cc | 3859 | if (do_barriers) { |
0b246afa | 3860 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
3861 | if (ret) { |
3862 | mutex_unlock( | |
0b246afa JM |
3863 | &fs_info->fs_devices->device_list_mutex); |
3864 | btrfs_handle_fs_error(fs_info, ret, | |
3865 | "errors while submitting device barriers."); | |
5af3e8cc SB |
3866 | return ret; |
3867 | } | |
3868 | } | |
387125fc | 3869 | |
1538e6c5 | 3870 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3871 | if (!dev->bdev) { |
3872 | total_errors++; | |
3873 | continue; | |
3874 | } | |
e12c9621 | 3875 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3876 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3877 | continue; |
3878 | ||
2b82032c | 3879 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3880 | btrfs_set_stack_device_type(dev_item, dev->type); |
3881 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3882 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3883 | dev->commit_total_bytes); |
ce7213c7 MX |
3884 | btrfs_set_stack_device_bytes_used(dev_item, |
3885 | dev->commit_bytes_used); | |
a061fc8d CM |
3886 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3887 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3888 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3889 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
3890 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
3891 | BTRFS_FSID_SIZE); | |
a512bbf8 | 3892 | |
a061fc8d CM |
3893 | flags = btrfs_super_flags(sb); |
3894 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3895 | ||
75cb857d QW |
3896 | ret = btrfs_validate_write_super(fs_info, sb); |
3897 | if (ret < 0) { | |
3898 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
3899 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
3900 | "unexpected superblock corruption detected"); | |
3901 | return -EUCLEAN; | |
3902 | } | |
3903 | ||
abbb3b8e | 3904 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
3905 | if (ret) |
3906 | total_errors++; | |
f2984462 | 3907 | } |
a236aed1 | 3908 | if (total_errors > max_errors) { |
0b246afa JM |
3909 | btrfs_err(fs_info, "%d errors while writing supers", |
3910 | total_errors); | |
3911 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 3912 | |
9d565ba4 | 3913 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
3914 | btrfs_handle_fs_error(fs_info, -EIO, |
3915 | "%d errors while writing supers", | |
3916 | total_errors); | |
9d565ba4 | 3917 | return -EIO; |
a236aed1 | 3918 | } |
f2984462 | 3919 | |
a512bbf8 | 3920 | total_errors = 0; |
1538e6c5 | 3921 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3922 | if (!dev->bdev) |
3923 | continue; | |
e12c9621 | 3924 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3925 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3926 | continue; |
3927 | ||
abbb3b8e | 3928 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
3929 | if (ret) |
3930 | total_errors++; | |
f2984462 | 3931 | } |
0b246afa | 3932 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3933 | if (total_errors > max_errors) { |
0b246afa JM |
3934 | btrfs_handle_fs_error(fs_info, -EIO, |
3935 | "%d errors while writing supers", | |
3936 | total_errors); | |
79787eaa | 3937 | return -EIO; |
a236aed1 | 3938 | } |
f2984462 CM |
3939 | return 0; |
3940 | } | |
3941 | ||
cb517eab MX |
3942 | /* Drop a fs root from the radix tree and free it. */ |
3943 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3944 | struct btrfs_root *root) | |
2619ba1f | 3945 | { |
4785e24f JB |
3946 | bool drop_ref = false; |
3947 | ||
4df27c4d | 3948 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3949 | radix_tree_delete(&fs_info->fs_roots_radix, |
3950 | (unsigned long)root->root_key.objectid); | |
af01d2e5 | 3951 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
4785e24f | 3952 | drop_ref = true; |
4df27c4d | 3953 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 3954 | |
1c1ea4f7 | 3955 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
ef67963d | 3956 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 3957 | if (root->reloc_root) { |
00246528 | 3958 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
3959 | root->reloc_root = NULL; |
3960 | } | |
3961 | } | |
3321719e | 3962 | |
faa2dbf0 JB |
3963 | if (root->free_ino_pinned) |
3964 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3965 | if (root->free_ino_ctl) | |
3966 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
0e996e7f JB |
3967 | if (root->ino_cache_inode) { |
3968 | iput(root->ino_cache_inode); | |
3969 | root->ino_cache_inode = NULL; | |
3970 | } | |
4785e24f JB |
3971 | if (drop_ref) |
3972 | btrfs_put_root(root); | |
2619ba1f CM |
3973 | } |
3974 | ||
c146afad | 3975 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3976 | { |
c146afad YZ |
3977 | u64 root_objectid = 0; |
3978 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3979 | int i = 0; |
3980 | int err = 0; | |
3981 | unsigned int ret = 0; | |
e089f05c | 3982 | |
c146afad | 3983 | while (1) { |
efc34534 | 3984 | spin_lock(&fs_info->fs_roots_radix_lock); |
c146afad YZ |
3985 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3986 | (void **)gang, root_objectid, | |
3987 | ARRAY_SIZE(gang)); | |
65d33fd7 | 3988 | if (!ret) { |
efc34534 | 3989 | spin_unlock(&fs_info->fs_roots_radix_lock); |
c146afad | 3990 | break; |
65d33fd7 | 3991 | } |
5d4f98a2 | 3992 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3993 | |
c146afad | 3994 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3995 | /* Avoid to grab roots in dead_roots */ |
3996 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3997 | gang[i] = NULL; | |
3998 | continue; | |
3999 | } | |
4000 | /* grab all the search result for later use */ | |
00246528 | 4001 | gang[i] = btrfs_grab_root(gang[i]); |
65d33fd7 | 4002 | } |
efc34534 | 4003 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4004 | |
65d33fd7 QW |
4005 | for (i = 0; i < ret; i++) { |
4006 | if (!gang[i]) | |
4007 | continue; | |
c146afad | 4008 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
4009 | err = btrfs_orphan_cleanup(gang[i]); |
4010 | if (err) | |
65d33fd7 | 4011 | break; |
00246528 | 4012 | btrfs_put_root(gang[i]); |
c146afad YZ |
4013 | } |
4014 | root_objectid++; | |
4015 | } | |
65d33fd7 QW |
4016 | |
4017 | /* release the uncleaned roots due to error */ | |
4018 | for (; i < ret; i++) { | |
4019 | if (gang[i]) | |
00246528 | 4020 | btrfs_put_root(gang[i]); |
65d33fd7 QW |
4021 | } |
4022 | return err; | |
c146afad | 4023 | } |
a2135011 | 4024 | |
6bccf3ab | 4025 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4026 | { |
6bccf3ab | 4027 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4028 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4029 | |
0b246afa | 4030 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4031 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4032 | mutex_unlock(&fs_info->cleaner_mutex); |
4033 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4034 | |
4035 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4036 | down_write(&fs_info->cleanup_work_sem); |
4037 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4038 | |
7a7eaa40 | 4039 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4040 | if (IS_ERR(trans)) |
4041 | return PTR_ERR(trans); | |
3a45bb20 | 4042 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4043 | } |
4044 | ||
b105e927 | 4045 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4046 | { |
c146afad YZ |
4047 | int ret; |
4048 | ||
afcdd129 | 4049 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
d6fd0ae2 OS |
4050 | /* |
4051 | * We don't want the cleaner to start new transactions, add more delayed | |
4052 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4053 | * because that frees the task_struct, and the transaction kthread might | |
4054 | * still try to wake up the cleaner. | |
4055 | */ | |
4056 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4057 | |
7343dd61 | 4058 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4059 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4060 | |
803b2f54 SB |
4061 | /* wait for the uuid_scan task to finish */ |
4062 | down(&fs_info->uuid_tree_rescan_sem); | |
4063 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4064 | up(&fs_info->uuid_tree_rescan_sem); | |
4065 | ||
837d5b6e | 4066 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4067 | btrfs_pause_balance(fs_info); |
837d5b6e | 4068 | |
8dabb742 SB |
4069 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4070 | ||
aa1b8cd4 | 4071 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4072 | |
4073 | /* wait for any defraggers to finish */ | |
4074 | wait_event(fs_info->transaction_wait, | |
4075 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4076 | ||
4077 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4078 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4079 | |
21c7e756 MX |
4080 | cancel_work_sync(&fs_info->async_reclaim_work); |
4081 | ||
b0643e59 DZ |
4082 | /* Cancel or finish ongoing discard work */ |
4083 | btrfs_discard_cleanup(fs_info); | |
4084 | ||
bc98a42c | 4085 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4086 | /* |
d6fd0ae2 OS |
4087 | * The cleaner kthread is stopped, so do one final pass over |
4088 | * unused block groups. | |
e44163e1 | 4089 | */ |
0b246afa | 4090 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4091 | |
f0cc2cd7 FM |
4092 | /* |
4093 | * There might be existing delayed inode workers still running | |
4094 | * and holding an empty delayed inode item. We must wait for | |
4095 | * them to complete first because they can create a transaction. | |
4096 | * This happens when someone calls btrfs_balance_delayed_items() | |
4097 | * and then a transaction commit runs the same delayed nodes | |
4098 | * before any delayed worker has done something with the nodes. | |
4099 | * We must wait for any worker here and not at transaction | |
4100 | * commit time since that could cause a deadlock. | |
4101 | * This is a very rare case. | |
4102 | */ | |
4103 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4104 | ||
6bccf3ab | 4105 | ret = btrfs_commit_super(fs_info); |
acce952b | 4106 | if (ret) |
04892340 | 4107 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4108 | } |
4109 | ||
af722733 LB |
4110 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || |
4111 | test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) | |
2ff7e61e | 4112 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4113 | |
e3029d9f AV |
4114 | kthread_stop(fs_info->transaction_kthread); |
4115 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4116 | |
e187831e | 4117 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4118 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4119 | |
5958253c QW |
4120 | if (btrfs_check_quota_leak(fs_info)) { |
4121 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4122 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4123 | } | |
4124 | ||
04892340 | 4125 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4126 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4127 | |
963d678b | 4128 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4129 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4130 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4131 | } |
bcc63abb | 4132 | |
4297ff84 JB |
4133 | if (percpu_counter_sum(&fs_info->dio_bytes)) |
4134 | btrfs_info(fs_info, "at unmount dio bytes count %lld", | |
4135 | percpu_counter_sum(&fs_info->dio_bytes)); | |
4136 | ||
6618a59b | 4137 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4138 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4139 | |
1a4319cc LB |
4140 | btrfs_put_block_group_cache(fs_info); |
4141 | ||
de348ee0 WS |
4142 | /* |
4143 | * we must make sure there is not any read request to | |
4144 | * submit after we stopping all workers. | |
4145 | */ | |
4146 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4147 | btrfs_stop_all_workers(fs_info); |
4148 | ||
afcdd129 | 4149 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4150 | free_root_pointers(fs_info, true); |
8c38938c | 4151 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4152 | |
4e19443d JB |
4153 | /* |
4154 | * We must free the block groups after dropping the fs_roots as we could | |
4155 | * have had an IO error and have left over tree log blocks that aren't | |
4156 | * cleaned up until the fs roots are freed. This makes the block group | |
4157 | * accounting appear to be wrong because there's pending reserved bytes, | |
4158 | * so make sure we do the block group cleanup afterwards. | |
4159 | */ | |
4160 | btrfs_free_block_groups(fs_info); | |
4161 | ||
13e6c37b | 4162 | iput(fs_info->btree_inode); |
d6bfde87 | 4163 | |
21adbd5c | 4164 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4165 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4166 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4167 | #endif |
4168 | ||
0b86a832 | 4169 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4170 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4171 | } |
4172 | ||
b9fab919 CM |
4173 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4174 | int atomic) | |
5f39d397 | 4175 | { |
1259ab75 | 4176 | int ret; |
727011e0 | 4177 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4178 | |
0b32f4bb | 4179 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4180 | if (!ret) |
4181 | return ret; | |
4182 | ||
4183 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4184 | parent_transid, atomic); |
4185 | if (ret == -EAGAIN) | |
4186 | return ret; | |
1259ab75 | 4187 | return !ret; |
5f39d397 CM |
4188 | } |
4189 | ||
5f39d397 CM |
4190 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4191 | { | |
0b246afa | 4192 | struct btrfs_fs_info *fs_info; |
06ea65a3 | 4193 | struct btrfs_root *root; |
5f39d397 | 4194 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4195 | int was_dirty; |
b4ce94de | 4196 | |
06ea65a3 JB |
4197 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4198 | /* | |
4199 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4200 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4201 | * outside of the sanity tests. |
4202 | */ | |
b0132a3b | 4203 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4204 | return; |
4205 | #endif | |
4206 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
0b246afa | 4207 | fs_info = root->fs_info; |
b9447ef8 | 4208 | btrfs_assert_tree_locked(buf); |
0b246afa | 4209 | if (transid != fs_info->generation) |
5d163e0e | 4210 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4211 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4212 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4213 | if (!was_dirty) |
104b4e51 NB |
4214 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4215 | buf->len, | |
4216 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4217 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4218 | /* |
4219 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4220 | * but item data not updated. | |
4221 | * So here we should only check item pointers, not item data. | |
4222 | */ | |
4223 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4224 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4225 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4226 | ASSERT(0); |
4227 | } | |
4228 | #endif | |
eb60ceac CM |
4229 | } |
4230 | ||
2ff7e61e | 4231 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4232 | int flush_delayed) |
16cdcec7 MX |
4233 | { |
4234 | /* | |
4235 | * looks as though older kernels can get into trouble with | |
4236 | * this code, they end up stuck in balance_dirty_pages forever | |
4237 | */ | |
e2d84521 | 4238 | int ret; |
16cdcec7 MX |
4239 | |
4240 | if (current->flags & PF_MEMALLOC) | |
4241 | return; | |
4242 | ||
b53d3f5d | 4243 | if (flush_delayed) |
2ff7e61e | 4244 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4245 | |
d814a491 EL |
4246 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4247 | BTRFS_DIRTY_METADATA_THRESH, | |
4248 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4249 | if (ret > 0) { |
0b246afa | 4250 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4251 | } |
16cdcec7 MX |
4252 | } |
4253 | ||
2ff7e61e | 4254 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4255 | { |
2ff7e61e | 4256 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4257 | } |
585ad2c3 | 4258 | |
2ff7e61e | 4259 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4260 | { |
2ff7e61e | 4261 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4262 | } |
6b80053d | 4263 | |
581c1760 QW |
4264 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4265 | struct btrfs_key *first_key) | |
6b80053d | 4266 | { |
5ab12d1f | 4267 | return btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 4268 | level, first_key); |
6b80053d | 4269 | } |
0da5468f | 4270 | |
2ff7e61e | 4271 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4272 | { |
fe816d0f NB |
4273 | /* cleanup FS via transaction */ |
4274 | btrfs_cleanup_transaction(fs_info); | |
4275 | ||
0b246afa | 4276 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4277 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4278 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4279 | |
0b246afa JM |
4280 | down_write(&fs_info->cleanup_work_sem); |
4281 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4282 | } |
4283 | ||
ef67963d JB |
4284 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4285 | { | |
4286 | struct btrfs_root *gang[8]; | |
4287 | u64 root_objectid = 0; | |
4288 | int ret; | |
4289 | ||
4290 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4291 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4292 | (void **)gang, root_objectid, | |
4293 | ARRAY_SIZE(gang))) != 0) { | |
4294 | int i; | |
4295 | ||
4296 | for (i = 0; i < ret; i++) | |
4297 | gang[i] = btrfs_grab_root(gang[i]); | |
4298 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4299 | ||
4300 | for (i = 0; i < ret; i++) { | |
4301 | if (!gang[i]) | |
4302 | continue; | |
4303 | root_objectid = gang[i]->root_key.objectid; | |
4304 | btrfs_free_log(NULL, gang[i]); | |
4305 | btrfs_put_root(gang[i]); | |
4306 | } | |
4307 | root_objectid++; | |
4308 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4309 | } | |
4310 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4311 | btrfs_free_log_root_tree(NULL, fs_info); | |
4312 | } | |
4313 | ||
143bede5 | 4314 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4315 | { |
acce952b | 4316 | struct btrfs_ordered_extent *ordered; |
acce952b | 4317 | |
199c2a9c | 4318 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4319 | /* |
4320 | * This will just short circuit the ordered completion stuff which will | |
4321 | * make sure the ordered extent gets properly cleaned up. | |
4322 | */ | |
199c2a9c | 4323 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4324 | root_extent_list) |
4325 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4326 | spin_unlock(&root->ordered_extent_lock); |
4327 | } | |
4328 | ||
4329 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4330 | { | |
4331 | struct btrfs_root *root; | |
4332 | struct list_head splice; | |
4333 | ||
4334 | INIT_LIST_HEAD(&splice); | |
4335 | ||
4336 | spin_lock(&fs_info->ordered_root_lock); | |
4337 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4338 | while (!list_empty(&splice)) { | |
4339 | root = list_first_entry(&splice, struct btrfs_root, | |
4340 | ordered_root); | |
1de2cfde JB |
4341 | list_move_tail(&root->ordered_root, |
4342 | &fs_info->ordered_roots); | |
199c2a9c | 4343 | |
2a85d9ca | 4344 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4345 | btrfs_destroy_ordered_extents(root); |
4346 | ||
2a85d9ca LB |
4347 | cond_resched(); |
4348 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4349 | } |
4350 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4351 | |
4352 | /* | |
4353 | * We need this here because if we've been flipped read-only we won't | |
4354 | * get sync() from the umount, so we need to make sure any ordered | |
4355 | * extents that haven't had their dirty pages IO start writeout yet | |
4356 | * actually get run and error out properly. | |
4357 | */ | |
4358 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4359 | } |
4360 | ||
35a3621b | 4361 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4362 | struct btrfs_fs_info *fs_info) |
acce952b | 4363 | { |
4364 | struct rb_node *node; | |
4365 | struct btrfs_delayed_ref_root *delayed_refs; | |
4366 | struct btrfs_delayed_ref_node *ref; | |
4367 | int ret = 0; | |
4368 | ||
4369 | delayed_refs = &trans->delayed_refs; | |
4370 | ||
4371 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4372 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4373 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4374 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4375 | return ret; |
4376 | } | |
4377 | ||
5c9d028b | 4378 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4379 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4380 | struct rb_node *n; |
e78417d1 | 4381 | bool pin_bytes = false; |
acce952b | 4382 | |
d7df2c79 JB |
4383 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4384 | href_node); | |
3069bd26 | 4385 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4386 | continue; |
3069bd26 | 4387 | |
d7df2c79 | 4388 | spin_lock(&head->lock); |
e3d03965 | 4389 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4390 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4391 | ref_node); | |
d7df2c79 | 4392 | ref->in_tree = 0; |
e3d03965 | 4393 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4394 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4395 | if (!list_empty(&ref->add_list)) |
4396 | list_del(&ref->add_list); | |
d7df2c79 JB |
4397 | atomic_dec(&delayed_refs->num_entries); |
4398 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4399 | } |
d7df2c79 JB |
4400 | if (head->must_insert_reserved) |
4401 | pin_bytes = true; | |
4402 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4403 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4404 | spin_unlock(&head->lock); |
4405 | spin_unlock(&delayed_refs->lock); | |
4406 | mutex_unlock(&head->mutex); | |
acce952b | 4407 | |
f603bb94 NB |
4408 | if (pin_bytes) { |
4409 | struct btrfs_block_group *cache; | |
4410 | ||
4411 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4412 | BUG_ON(!cache); | |
4413 | ||
4414 | spin_lock(&cache->space_info->lock); | |
4415 | spin_lock(&cache->lock); | |
4416 | cache->pinned += head->num_bytes; | |
4417 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4418 | cache->space_info, head->num_bytes); | |
4419 | cache->reserved -= head->num_bytes; | |
4420 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4421 | spin_unlock(&cache->lock); | |
4422 | spin_unlock(&cache->space_info->lock); | |
4423 | percpu_counter_add_batch( | |
4424 | &cache->space_info->total_bytes_pinned, | |
4425 | head->num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH); | |
4426 | ||
4427 | btrfs_put_block_group(cache); | |
4428 | ||
4429 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4430 | head->bytenr + head->num_bytes - 1); | |
4431 | } | |
31890da0 | 4432 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4433 | btrfs_put_delayed_ref_head(head); |
acce952b | 4434 | cond_resched(); |
4435 | spin_lock(&delayed_refs->lock); | |
4436 | } | |
81f7eb00 | 4437 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4438 | |
4439 | spin_unlock(&delayed_refs->lock); | |
4440 | ||
4441 | return ret; | |
4442 | } | |
4443 | ||
143bede5 | 4444 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4445 | { |
4446 | struct btrfs_inode *btrfs_inode; | |
4447 | struct list_head splice; | |
4448 | ||
4449 | INIT_LIST_HEAD(&splice); | |
4450 | ||
eb73c1b7 MX |
4451 | spin_lock(&root->delalloc_lock); |
4452 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4453 | |
4454 | while (!list_empty(&splice)) { | |
fe816d0f | 4455 | struct inode *inode = NULL; |
eb73c1b7 MX |
4456 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4457 | delalloc_inodes); | |
fe816d0f | 4458 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4459 | spin_unlock(&root->delalloc_lock); |
acce952b | 4460 | |
fe816d0f NB |
4461 | /* |
4462 | * Make sure we get a live inode and that it'll not disappear | |
4463 | * meanwhile. | |
4464 | */ | |
4465 | inode = igrab(&btrfs_inode->vfs_inode); | |
4466 | if (inode) { | |
4467 | invalidate_inode_pages2(inode->i_mapping); | |
4468 | iput(inode); | |
4469 | } | |
eb73c1b7 | 4470 | spin_lock(&root->delalloc_lock); |
acce952b | 4471 | } |
eb73c1b7 MX |
4472 | spin_unlock(&root->delalloc_lock); |
4473 | } | |
4474 | ||
4475 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4476 | { | |
4477 | struct btrfs_root *root; | |
4478 | struct list_head splice; | |
4479 | ||
4480 | INIT_LIST_HEAD(&splice); | |
4481 | ||
4482 | spin_lock(&fs_info->delalloc_root_lock); | |
4483 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4484 | while (!list_empty(&splice)) { | |
4485 | root = list_first_entry(&splice, struct btrfs_root, | |
4486 | delalloc_root); | |
00246528 | 4487 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4488 | BUG_ON(!root); |
4489 | spin_unlock(&fs_info->delalloc_root_lock); | |
4490 | ||
4491 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4492 | btrfs_put_root(root); |
eb73c1b7 MX |
4493 | |
4494 | spin_lock(&fs_info->delalloc_root_lock); | |
4495 | } | |
4496 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4497 | } |
4498 | ||
2ff7e61e | 4499 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4500 | struct extent_io_tree *dirty_pages, |
4501 | int mark) | |
4502 | { | |
4503 | int ret; | |
acce952b | 4504 | struct extent_buffer *eb; |
4505 | u64 start = 0; | |
4506 | u64 end; | |
acce952b | 4507 | |
4508 | while (1) { | |
4509 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4510 | mark, NULL); |
acce952b | 4511 | if (ret) |
4512 | break; | |
4513 | ||
91166212 | 4514 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4515 | while (start <= end) { |
0b246afa JM |
4516 | eb = find_extent_buffer(fs_info, start); |
4517 | start += fs_info->nodesize; | |
fd8b2b61 | 4518 | if (!eb) |
acce952b | 4519 | continue; |
fd8b2b61 | 4520 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4521 | |
fd8b2b61 JB |
4522 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4523 | &eb->bflags)) | |
4524 | clear_extent_buffer_dirty(eb); | |
4525 | free_extent_buffer_stale(eb); | |
acce952b | 4526 | } |
4527 | } | |
4528 | ||
4529 | return ret; | |
4530 | } | |
4531 | ||
2ff7e61e | 4532 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 4533 | struct extent_io_tree *unpin) |
acce952b | 4534 | { |
acce952b | 4535 | u64 start; |
4536 | u64 end; | |
4537 | int ret; | |
4538 | ||
acce952b | 4539 | while (1) { |
0e6ec385 FM |
4540 | struct extent_state *cached_state = NULL; |
4541 | ||
fcd5e742 LF |
4542 | /* |
4543 | * The btrfs_finish_extent_commit() may get the same range as | |
4544 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4545 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4546 | * the same extent range. | |
4547 | */ | |
4548 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4549 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 4550 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
4551 | if (ret) { |
4552 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4553 | break; |
fcd5e742 | 4554 | } |
acce952b | 4555 | |
0e6ec385 FM |
4556 | clear_extent_dirty(unpin, start, end, &cached_state); |
4557 | free_extent_state(cached_state); | |
2ff7e61e | 4558 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4559 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4560 | cond_resched(); |
4561 | } | |
4562 | ||
4563 | return 0; | |
4564 | } | |
4565 | ||
32da5386 | 4566 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4567 | { |
4568 | struct inode *inode; | |
4569 | ||
4570 | inode = cache->io_ctl.inode; | |
4571 | if (inode) { | |
4572 | invalidate_inode_pages2(inode->i_mapping); | |
4573 | BTRFS_I(inode)->generation = 0; | |
4574 | cache->io_ctl.inode = NULL; | |
4575 | iput(inode); | |
4576 | } | |
4577 | btrfs_put_block_group(cache); | |
4578 | } | |
4579 | ||
4580 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4581 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4582 | { |
32da5386 | 4583 | struct btrfs_block_group *cache; |
c79a1751 LB |
4584 | |
4585 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4586 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4587 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4588 | struct btrfs_block_group, |
c79a1751 | 4589 | dirty_list); |
c79a1751 LB |
4590 | |
4591 | if (!list_empty(&cache->io_list)) { | |
4592 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4593 | list_del_init(&cache->io_list); | |
4594 | btrfs_cleanup_bg_io(cache); | |
4595 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4596 | } | |
4597 | ||
4598 | list_del_init(&cache->dirty_list); | |
4599 | spin_lock(&cache->lock); | |
4600 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4601 | spin_unlock(&cache->lock); | |
4602 | ||
4603 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4604 | btrfs_put_block_group(cache); | |
ba2c4d4e | 4605 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
4606 | spin_lock(&cur_trans->dirty_bgs_lock); |
4607 | } | |
4608 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4609 | ||
45ae2c18 NB |
4610 | /* |
4611 | * Refer to the definition of io_bgs member for details why it's safe | |
4612 | * to use it without any locking | |
4613 | */ | |
c79a1751 LB |
4614 | while (!list_empty(&cur_trans->io_bgs)) { |
4615 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4616 | struct btrfs_block_group, |
c79a1751 | 4617 | io_list); |
c79a1751 LB |
4618 | |
4619 | list_del_init(&cache->io_list); | |
4620 | spin_lock(&cache->lock); | |
4621 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4622 | spin_unlock(&cache->lock); | |
4623 | btrfs_cleanup_bg_io(cache); | |
4624 | } | |
4625 | } | |
4626 | ||
49b25e05 | 4627 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4628 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4629 | { |
bbbf7243 NB |
4630 | struct btrfs_device *dev, *tmp; |
4631 | ||
2ff7e61e | 4632 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4633 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4634 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4635 | ||
bbbf7243 NB |
4636 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4637 | post_commit_list) { | |
4638 | list_del_init(&dev->post_commit_list); | |
4639 | } | |
4640 | ||
2ff7e61e | 4641 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4642 | |
4a9d8bde | 4643 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4644 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4645 | |
4a9d8bde | 4646 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4647 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4648 | |
ccdf9b30 | 4649 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4650 | |
2ff7e61e | 4651 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4652 | EXTENT_DIRTY); |
fe119a6e | 4653 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4654 | |
4a9d8bde MX |
4655 | cur_trans->state =TRANS_STATE_COMPLETED; |
4656 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4657 | } |
4658 | ||
2ff7e61e | 4659 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4660 | { |
4661 | struct btrfs_transaction *t; | |
acce952b | 4662 | |
0b246afa | 4663 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4664 | |
0b246afa JM |
4665 | spin_lock(&fs_info->trans_lock); |
4666 | while (!list_empty(&fs_info->trans_list)) { | |
4667 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4668 | struct btrfs_transaction, list); |
4669 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4670 | refcount_inc(&t->use_count); |
0b246afa | 4671 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4672 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4673 | btrfs_put_transaction(t); |
0b246afa | 4674 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4675 | continue; |
4676 | } | |
0b246afa | 4677 | if (t == fs_info->running_transaction) { |
724e2315 | 4678 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4679 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4680 | /* |
4681 | * We wait for 0 num_writers since we don't hold a trans | |
4682 | * handle open currently for this transaction. | |
4683 | */ | |
4684 | wait_event(t->writer_wait, | |
4685 | atomic_read(&t->num_writers) == 0); | |
4686 | } else { | |
0b246afa | 4687 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4688 | } |
2ff7e61e | 4689 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4690 | |
0b246afa JM |
4691 | spin_lock(&fs_info->trans_lock); |
4692 | if (t == fs_info->running_transaction) | |
4693 | fs_info->running_transaction = NULL; | |
acce952b | 4694 | list_del_init(&t->list); |
0b246afa | 4695 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4696 | |
724e2315 | 4697 | btrfs_put_transaction(t); |
2ff7e61e | 4698 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4699 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4700 | } |
0b246afa JM |
4701 | spin_unlock(&fs_info->trans_lock); |
4702 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4703 | btrfs_destroy_delayed_inodes(fs_info); |
4704 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4705 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4706 | btrfs_drop_all_logs(fs_info); |
0b246afa | 4707 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4708 | |
4709 | return 0; | |
4710 | } | |
4711 | ||
e8c9f186 | 4712 | static const struct extent_io_ops btree_extent_io_ops = { |
4d53dddb | 4713 | /* mandatory callbacks */ |
0b86a832 | 4714 | .submit_bio_hook = btree_submit_bio_hook, |
4d53dddb | 4715 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0da5468f | 4716 | }; |