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" |
21adbd5c | 32 | #include "check-integrity.h" |
606686ee | 33 | #include "rcu-string.h" |
8dabb742 | 34 | #include "dev-replace.h" |
53b381b3 | 35 | #include "raid56.h" |
5ac1d209 | 36 | #include "sysfs.h" |
fcebe456 | 37 | #include "qgroup.h" |
ebb8765b | 38 | #include "compression.h" |
557ea5dd | 39 | #include "tree-checker.h" |
fd708b81 | 40 | #include "ref-verify.h" |
aac0023c | 41 | #include "block-group.h" |
b0643e59 | 42 | #include "discard.h" |
f603bb94 | 43 | #include "space-info.h" |
b70f5097 | 44 | #include "zoned.h" |
139e8cd3 | 45 | #include "subpage.h" |
eb60ceac | 46 | |
319e4d06 QW |
47 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
48 | BTRFS_HEADER_FLAG_RELOC |\ | |
49 | BTRFS_SUPER_FLAG_ERROR |\ | |
50 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
51 | BTRFS_SUPER_FLAG_METADUMP |\ |
52 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 53 | |
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 { |
8896a08d | 113 | struct inode *inode; |
44b8bd7e | 114 | struct bio *bio; |
a758781d | 115 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 116 | int mirror_num; |
1941b64b QW |
117 | |
118 | /* Optional parameter for submit_bio_start used by direct io */ | |
119 | u64 dio_file_offset; | |
8b712842 | 120 | struct btrfs_work work; |
4e4cbee9 | 121 | blk_status_t status; |
44b8bd7e CM |
122 | }; |
123 | ||
85d4e461 CM |
124 | /* |
125 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
126 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
127 | * the level the eb occupies in the tree. | |
128 | * | |
129 | * Different roots are used for different purposes and may nest inside each | |
130 | * other and they require separate keysets. As lockdep keys should be | |
131 | * static, assign keysets according to the purpose of the root as indicated | |
4fd786e6 MT |
132 | * by btrfs_root->root_key.objectid. This ensures that all special purpose |
133 | * roots have separate keysets. | |
4008c04a | 134 | * |
85d4e461 CM |
135 | * Lock-nesting across peer nodes is always done with the immediate parent |
136 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
137 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 138 | * |
85d4e461 CM |
139 | * The key is set by the readpage_end_io_hook after the buffer has passed |
140 | * csum validation but before the pages are unlocked. It is also set by | |
141 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 142 | * |
85d4e461 CM |
143 | * We also add a check to make sure the highest level of the tree is the |
144 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
145 | * needs update as well. | |
4008c04a CM |
146 | */ |
147 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
148 | # if BTRFS_MAX_LEVEL != 8 | |
149 | # error | |
150 | # endif | |
85d4e461 | 151 | |
ab1405aa DS |
152 | #define DEFINE_LEVEL(stem, level) \ |
153 | .names[level] = "btrfs-" stem "-0" #level, | |
154 | ||
155 | #define DEFINE_NAME(stem) \ | |
156 | DEFINE_LEVEL(stem, 0) \ | |
157 | DEFINE_LEVEL(stem, 1) \ | |
158 | DEFINE_LEVEL(stem, 2) \ | |
159 | DEFINE_LEVEL(stem, 3) \ | |
160 | DEFINE_LEVEL(stem, 4) \ | |
161 | DEFINE_LEVEL(stem, 5) \ | |
162 | DEFINE_LEVEL(stem, 6) \ | |
163 | DEFINE_LEVEL(stem, 7) | |
164 | ||
85d4e461 CM |
165 | static struct btrfs_lockdep_keyset { |
166 | u64 id; /* root objectid */ | |
ab1405aa | 167 | /* Longest entry: btrfs-free-space-00 */ |
387824af DS |
168 | char names[BTRFS_MAX_LEVEL][20]; |
169 | struct lock_class_key keys[BTRFS_MAX_LEVEL]; | |
85d4e461 | 170 | } btrfs_lockdep_keysets[] = { |
ab1405aa DS |
171 | { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, |
172 | { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, | |
173 | { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, | |
174 | { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, | |
ab1405aa DS |
175 | { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, |
176 | { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, | |
177 | { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, | |
178 | { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, | |
179 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, | |
180 | { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, | |
181 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, | |
182 | { .id = 0, DEFINE_NAME("tree") }, | |
4008c04a | 183 | }; |
85d4e461 | 184 | |
ab1405aa DS |
185 | #undef DEFINE_LEVEL |
186 | #undef DEFINE_NAME | |
85d4e461 CM |
187 | |
188 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
189 | int level) | |
190 | { | |
191 | struct btrfs_lockdep_keyset *ks; | |
192 | ||
193 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
194 | ||
195 | /* find the matching keyset, id 0 is the default entry */ | |
196 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
197 | if (ks->id == objectid) | |
198 | break; | |
199 | ||
200 | lockdep_set_class_and_name(&eb->lock, | |
201 | &ks->keys[level], ks->names[level]); | |
202 | } | |
203 | ||
4008c04a CM |
204 | #endif |
205 | ||
d352ac68 | 206 | /* |
2996e1f8 | 207 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 208 | */ |
c67b3892 | 209 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 210 | { |
d5178578 | 211 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 212 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 213 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 214 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 215 | char *kaddr; |
e9be5a30 | 216 | int i; |
d5178578 JT |
217 | |
218 | shash->tfm = fs_info->csum_shash; | |
219 | crypto_shash_init(shash); | |
a26663e7 | 220 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 221 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 222 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 223 | |
e9be5a30 DS |
224 | for (i = 1; i < num_pages; i++) { |
225 | kaddr = page_address(buf->pages[i]); | |
226 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 227 | } |
71a63551 | 228 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 229 | crypto_shash_final(shash, result); |
19c00ddc CM |
230 | } |
231 | ||
d352ac68 CM |
232 | /* |
233 | * we can't consider a given block up to date unless the transid of the | |
234 | * block matches the transid in the parent node's pointer. This is how we | |
235 | * detect blocks that either didn't get written at all or got written | |
236 | * in the wrong place. | |
237 | */ | |
1259ab75 | 238 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
239 | struct extent_buffer *eb, u64 parent_transid, |
240 | int atomic) | |
1259ab75 | 241 | { |
2ac55d41 | 242 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
243 | int ret; |
244 | ||
245 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
246 | return 0; | |
247 | ||
b9fab919 CM |
248 | if (atomic) |
249 | return -EAGAIN; | |
250 | ||
2ac55d41 | 251 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 252 | &cached_state); |
0b32f4bb | 253 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
254 | btrfs_header_generation(eb) == parent_transid) { |
255 | ret = 0; | |
256 | goto out; | |
257 | } | |
94647322 DS |
258 | btrfs_err_rl(eb->fs_info, |
259 | "parent transid verify failed on %llu wanted %llu found %llu", | |
260 | eb->start, | |
29549aec | 261 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 262 | ret = 1; |
35b22c19 | 263 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 264 | out: |
2ac55d41 | 265 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 266 | &cached_state); |
1259ab75 | 267 | return ret; |
1259ab75 CM |
268 | } |
269 | ||
e7e16f48 JT |
270 | static bool btrfs_supported_super_csum(u16 csum_type) |
271 | { | |
272 | switch (csum_type) { | |
273 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 274 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 275 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 276 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
277 | return true; |
278 | default: | |
279 | return false; | |
280 | } | |
281 | } | |
282 | ||
1104a885 DS |
283 | /* |
284 | * Return 0 if the superblock checksum type matches the checksum value of that | |
285 | * algorithm. Pass the raw disk superblock data. | |
286 | */ | |
ab8d0fc4 JM |
287 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
288 | char *raw_disk_sb) | |
1104a885 DS |
289 | { |
290 | struct btrfs_super_block *disk_sb = | |
291 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 292 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
293 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
294 | ||
295 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 296 | |
51bce6c9 JT |
297 | /* |
298 | * The super_block structure does not span the whole | |
299 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
300 | * filled with zeros and is included in the checksum. | |
301 | */ | |
fd08001f EB |
302 | crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
303 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); | |
1104a885 | 304 | |
55fc29be | 305 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 306 | return 1; |
1104a885 | 307 | |
e7e16f48 | 308 | return 0; |
1104a885 DS |
309 | } |
310 | ||
e064d5e9 | 311 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 312 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 313 | { |
e064d5e9 | 314 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
315 | int found_level; |
316 | struct btrfs_key found_key; | |
317 | int ret; | |
318 | ||
319 | found_level = btrfs_header_level(eb); | |
320 | if (found_level != level) { | |
63489055 QW |
321 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
322 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
323 | btrfs_err(fs_info, |
324 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
325 | eb->start, level, found_level); | |
581c1760 QW |
326 | return -EIO; |
327 | } | |
328 | ||
329 | if (!first_key) | |
330 | return 0; | |
331 | ||
5d41be6f QW |
332 | /* |
333 | * For live tree block (new tree blocks in current transaction), | |
334 | * we need proper lock context to avoid race, which is impossible here. | |
335 | * So we only checks tree blocks which is read from disk, whose | |
336 | * generation <= fs_info->last_trans_committed. | |
337 | */ | |
338 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
339 | return 0; | |
62fdaa52 QW |
340 | |
341 | /* We have @first_key, so this @eb must have at least one item */ | |
342 | if (btrfs_header_nritems(eb) == 0) { | |
343 | btrfs_err(fs_info, | |
344 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
345 | eb->start); | |
346 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
347 | return -EUCLEAN; | |
348 | } | |
349 | ||
581c1760 QW |
350 | if (found_level) |
351 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
352 | else | |
353 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
354 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
355 | ||
581c1760 | 356 | if (ret) { |
63489055 QW |
357 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
358 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 359 | btrfs_err(fs_info, |
ff76a864 LB |
360 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
361 | eb->start, parent_transid, first_key->objectid, | |
362 | first_key->type, first_key->offset, | |
363 | found_key.objectid, found_key.type, | |
364 | found_key.offset); | |
581c1760 | 365 | } |
581c1760 QW |
366 | return ret; |
367 | } | |
368 | ||
d352ac68 CM |
369 | /* |
370 | * helper to read a given tree block, doing retries as required when | |
371 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
372 | * |
373 | * @parent_transid: expected transid, skip check if 0 | |
374 | * @level: expected level, mandatory check | |
375 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 376 | */ |
5ab12d1f | 377 | static int btree_read_extent_buffer_pages(struct extent_buffer *eb, |
581c1760 QW |
378 | u64 parent_transid, int level, |
379 | struct btrfs_key *first_key) | |
f188591e | 380 | { |
5ab12d1f | 381 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 382 | struct extent_io_tree *io_tree; |
ea466794 | 383 | int failed = 0; |
f188591e CM |
384 | int ret; |
385 | int num_copies = 0; | |
386 | int mirror_num = 0; | |
ea466794 | 387 | int failed_mirror = 0; |
f188591e | 388 | |
0b246afa | 389 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 390 | while (1) { |
f8397d69 | 391 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 392 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 393 | if (!ret) { |
581c1760 | 394 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 395 | parent_transid, 0)) |
256dd1bb | 396 | ret = -EIO; |
e064d5e9 | 397 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 398 | first_key, parent_transid)) |
581c1760 QW |
399 | ret = -EUCLEAN; |
400 | else | |
401 | break; | |
256dd1bb | 402 | } |
d397712b | 403 | |
0b246afa | 404 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 405 | eb->start, eb->len); |
4235298e | 406 | if (num_copies == 1) |
ea466794 | 407 | break; |
4235298e | 408 | |
5cf1ab56 JB |
409 | if (!failed_mirror) { |
410 | failed = 1; | |
411 | failed_mirror = eb->read_mirror; | |
412 | } | |
413 | ||
f188591e | 414 | mirror_num++; |
ea466794 JB |
415 | if (mirror_num == failed_mirror) |
416 | mirror_num++; | |
417 | ||
4235298e | 418 | if (mirror_num > num_copies) |
ea466794 | 419 | break; |
f188591e | 420 | } |
ea466794 | 421 | |
c0901581 | 422 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 423 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
424 | |
425 | return ret; | |
f188591e | 426 | } |
19c00ddc | 427 | |
eca0f6f6 QW |
428 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
429 | { | |
430 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
431 | u8 result[BTRFS_CSUM_SIZE]; | |
432 | int ret; | |
433 | ||
434 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
435 | offsetof(struct btrfs_header, fsid), | |
436 | BTRFS_FSID_SIZE) == 0); | |
437 | csum_tree_block(eb, result); | |
438 | ||
439 | if (btrfs_header_level(eb)) | |
440 | ret = btrfs_check_node(eb); | |
441 | else | |
442 | ret = btrfs_check_leaf_full(eb); | |
443 | ||
444 | if (ret < 0) { | |
445 | btrfs_print_tree(eb, 0); | |
446 | btrfs_err(fs_info, | |
447 | "block=%llu write time tree block corruption detected", | |
448 | eb->start); | |
449 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
450 | return ret; | |
451 | } | |
452 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
453 | ||
454 | return 0; | |
455 | } | |
456 | ||
457 | /* Checksum all dirty extent buffers in one bio_vec */ | |
458 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
459 | struct bio_vec *bvec) | |
460 | { | |
461 | struct page *page = bvec->bv_page; | |
462 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
463 | u64 cur; | |
464 | int ret = 0; | |
465 | ||
466 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
467 | cur += fs_info->nodesize) { | |
468 | struct extent_buffer *eb; | |
469 | bool uptodate; | |
470 | ||
471 | eb = find_extent_buffer(fs_info, cur); | |
472 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
473 | fs_info->nodesize); | |
474 | ||
475 | /* A dirty eb shouldn't disappear from buffer_radix */ | |
476 | if (WARN_ON(!eb)) | |
477 | return -EUCLEAN; | |
478 | ||
479 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
480 | free_extent_buffer(eb); | |
481 | return -EUCLEAN; | |
482 | } | |
483 | if (WARN_ON(!uptodate)) { | |
484 | free_extent_buffer(eb); | |
485 | return -EUCLEAN; | |
486 | } | |
487 | ||
488 | ret = csum_one_extent_buffer(eb); | |
489 | free_extent_buffer(eb); | |
490 | if (ret < 0) | |
491 | return ret; | |
492 | } | |
493 | return ret; | |
494 | } | |
495 | ||
d352ac68 | 496 | /* |
ac303b69 QW |
497 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
498 | * we only fill in the checksum field in the first page of a multi-page block. | |
499 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 500 | */ |
ac303b69 | 501 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 502 | { |
ac303b69 | 503 | struct page *page = bvec->bv_page; |
4eee4fa4 | 504 | u64 start = page_offset(page); |
19c00ddc | 505 | u64 found_start; |
19c00ddc | 506 | struct extent_buffer *eb; |
eca0f6f6 QW |
507 | |
508 | if (fs_info->sectorsize < PAGE_SIZE) | |
509 | return csum_dirty_subpage_buffers(fs_info, bvec); | |
f188591e | 510 | |
4f2de97a JB |
511 | eb = (struct extent_buffer *)page->private; |
512 | if (page != eb->pages[0]) | |
513 | return 0; | |
0f805531 | 514 | |
19c00ddc | 515 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
516 | |
517 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
518 | WARN_ON(found_start != 0); | |
519 | return 0; | |
520 | } | |
521 | ||
0f805531 AL |
522 | /* |
523 | * Please do not consolidate these warnings into a single if. | |
524 | * It is useful to know what went wrong. | |
525 | */ | |
526 | if (WARN_ON(found_start != start)) | |
527 | return -EUCLEAN; | |
528 | if (WARN_ON(!PageUptodate(page))) | |
529 | return -EUCLEAN; | |
530 | ||
eca0f6f6 | 531 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
532 | } |
533 | ||
b0c9b3b0 | 534 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 535 | { |
b0c9b3b0 | 536 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 537 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 538 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 539 | u8 *metadata_uuid; |
2b82032c | 540 | |
9a8658e3 DS |
541 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
542 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
543 | /* |
544 | * Checking the incompat flag is only valid for the current fs. For | |
545 | * seed devices it's forbidden to have their uuid changed so reading | |
546 | * ->fsid in this case is fine | |
547 | */ | |
548 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
549 | metadata_uuid = fs_devices->metadata_uuid; | |
550 | else | |
551 | metadata_uuid = fs_devices->fsid; | |
552 | ||
553 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
554 | return 0; | |
555 | ||
556 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
557 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
558 | return 0; | |
559 | ||
560 | return 1; | |
2b82032c YZ |
561 | } |
562 | ||
77bf40a2 QW |
563 | /* Do basic extent buffer checks at read time */ |
564 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 565 | { |
77bf40a2 | 566 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 567 | u64 found_start; |
77bf40a2 QW |
568 | const u32 csum_size = fs_info->csum_size; |
569 | u8 found_level; | |
2996e1f8 | 570 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 571 | const u8 *header_csum; |
77bf40a2 | 572 | int ret = 0; |
ea466794 | 573 | |
ce9adaa5 | 574 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 575 | if (found_start != eb->start) { |
893bf4b1 SY |
576 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
577 | eb->start, found_start); | |
f188591e | 578 | ret = -EIO; |
77bf40a2 | 579 | goto out; |
ce9adaa5 | 580 | } |
b0c9b3b0 | 581 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
582 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
583 | eb->start); | |
1259ab75 | 584 | ret = -EIO; |
77bf40a2 | 585 | goto out; |
1259ab75 | 586 | } |
ce9adaa5 | 587 | found_level = btrfs_header_level(eb); |
1c24c3ce | 588 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
589 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
590 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce | 591 | ret = -EIO; |
77bf40a2 | 592 | goto out; |
1c24c3ce | 593 | } |
ce9adaa5 | 594 | |
c67b3892 | 595 | csum_tree_block(eb, result); |
dfd29eed DS |
596 | header_csum = page_address(eb->pages[0]) + |
597 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 598 | |
dfd29eed | 599 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 600 | btrfs_warn_rl(fs_info, |
ff14aa79 DS |
601 | "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
602 | eb->start, | |
dfd29eed | 603 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
604 | CSUM_FMT_VALUE(csum_size, result), |
605 | btrfs_header_level(eb)); | |
2996e1f8 | 606 | ret = -EUCLEAN; |
77bf40a2 | 607 | goto out; |
2996e1f8 JT |
608 | } |
609 | ||
a826d6dc JB |
610 | /* |
611 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
612 | * that we don't try and read the other copies of this block, just | |
613 | * return -EIO. | |
614 | */ | |
1c4360ee | 615 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
616 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
617 | ret = -EIO; | |
618 | } | |
ce9adaa5 | 619 | |
813fd1dc | 620 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
621 | ret = -EIO; |
622 | ||
0b32f4bb JB |
623 | if (!ret) |
624 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
625 | else |
626 | btrfs_err(fs_info, | |
627 | "block=%llu read time tree block corruption detected", | |
628 | eb->start); | |
77bf40a2 QW |
629 | out: |
630 | return ret; | |
631 | } | |
632 | ||
371cdc07 QW |
633 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
634 | int mirror) | |
635 | { | |
636 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
637 | struct extent_buffer *eb; | |
638 | bool reads_done; | |
639 | int ret = 0; | |
640 | ||
641 | /* | |
642 | * We don't allow bio merge for subpage metadata read, so we should | |
643 | * only get one eb for each endio hook. | |
644 | */ | |
645 | ASSERT(end == start + fs_info->nodesize - 1); | |
646 | ASSERT(PagePrivate(page)); | |
647 | ||
648 | eb = find_extent_buffer(fs_info, start); | |
649 | /* | |
650 | * When we are reading one tree block, eb must have been inserted into | |
651 | * the radix tree. If not, something is wrong. | |
652 | */ | |
653 | ASSERT(eb); | |
654 | ||
655 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
656 | /* Subpage read must finish in page read */ | |
657 | ASSERT(reads_done); | |
658 | ||
659 | eb->read_mirror = mirror; | |
660 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
661 | ret = -EIO; | |
662 | goto err; | |
663 | } | |
664 | ret = validate_extent_buffer(eb); | |
665 | if (ret < 0) | |
666 | goto err; | |
667 | ||
668 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
669 | btree_readahead_hook(eb, ret); | |
670 | ||
671 | set_extent_buffer_uptodate(eb); | |
672 | ||
673 | free_extent_buffer(eb); | |
674 | return ret; | |
675 | err: | |
676 | /* | |
677 | * end_bio_extent_readpage decrements io_pages in case of error, | |
678 | * make sure it has something to decrement. | |
679 | */ | |
680 | atomic_inc(&eb->io_pages); | |
681 | clear_extent_buffer_uptodate(eb); | |
682 | free_extent_buffer(eb); | |
683 | return ret; | |
684 | } | |
685 | ||
8e1dc982 | 686 | int btrfs_validate_metadata_buffer(struct btrfs_io_bio *io_bio, |
77bf40a2 QW |
687 | struct page *page, u64 start, u64 end, |
688 | int mirror) | |
689 | { | |
690 | struct extent_buffer *eb; | |
691 | int ret = 0; | |
692 | int reads_done; | |
693 | ||
694 | ASSERT(page->private); | |
371cdc07 QW |
695 | |
696 | if (btrfs_sb(page->mapping->host->i_sb)->sectorsize < PAGE_SIZE) | |
697 | return validate_subpage_buffer(page, start, end, mirror); | |
698 | ||
77bf40a2 QW |
699 | eb = (struct extent_buffer *)page->private; |
700 | ||
701 | /* | |
702 | * The pending IO might have been the only thing that kept this buffer | |
703 | * in memory. Make sure we have a ref for all this other checks | |
704 | */ | |
705 | atomic_inc(&eb->refs); | |
706 | ||
707 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
708 | if (!reads_done) | |
709 | goto err; | |
710 | ||
711 | eb->read_mirror = mirror; | |
712 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
713 | ret = -EIO; | |
714 | goto err; | |
715 | } | |
716 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 717 | err: |
79fb65a1 JB |
718 | if (reads_done && |
719 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 720 | btree_readahead_hook(eb, ret); |
4bb31e92 | 721 | |
53b381b3 DW |
722 | if (ret) { |
723 | /* | |
724 | * our io error hook is going to dec the io pages | |
725 | * again, we have to make sure it has something | |
726 | * to decrement | |
727 | */ | |
728 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 729 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 730 | } |
0b32f4bb | 731 | free_extent_buffer(eb); |
77bf40a2 | 732 | |
f188591e | 733 | return ret; |
ce9adaa5 CM |
734 | } |
735 | ||
4246a0b6 | 736 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 737 | { |
97eb6b69 | 738 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 739 | struct btrfs_fs_info *fs_info; |
9e0af237 | 740 | struct btrfs_workqueue *wq; |
ce9adaa5 | 741 | |
ce9adaa5 | 742 | fs_info = end_io_wq->info; |
4e4cbee9 | 743 | end_io_wq->status = bio->bi_status; |
d20f7043 | 744 | |
cfe94440 | 745 | if (btrfs_op(bio) == BTRFS_MAP_WRITE) { |
a0cac0ec | 746 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 747 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 748 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 749 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 750 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 751 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 752 | else |
9e0af237 | 753 | wq = fs_info->endio_write_workers; |
d20f7043 | 754 | } else { |
5c047a69 | 755 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 756 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 757 | else if (end_io_wq->metadata) |
9e0af237 | 758 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 759 | else |
9e0af237 | 760 | wq = fs_info->endio_workers; |
d20f7043 | 761 | } |
9e0af237 | 762 | |
a0cac0ec | 763 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 764 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
765 | } |
766 | ||
4e4cbee9 | 767 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 768 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 769 | { |
97eb6b69 | 770 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 771 | |
97eb6b69 | 772 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 773 | if (!end_io_wq) |
4e4cbee9 | 774 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
775 | |
776 | end_io_wq->private = bio->bi_private; | |
777 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 778 | end_io_wq->info = info; |
4e4cbee9 | 779 | end_io_wq->status = 0; |
ce9adaa5 | 780 | end_io_wq->bio = bio; |
22c59948 | 781 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
782 | |
783 | bio->bi_private = end_io_wq; | |
784 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
785 | return 0; |
786 | } | |
787 | ||
4a69a410 CM |
788 | static void run_one_async_start(struct btrfs_work *work) |
789 | { | |
4a69a410 | 790 | struct async_submit_bio *async; |
4e4cbee9 | 791 | blk_status_t ret; |
4a69a410 CM |
792 | |
793 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
794 | ret = async->submit_bio_start(async->inode, async->bio, |
795 | async->dio_file_offset); | |
79787eaa | 796 | if (ret) |
4e4cbee9 | 797 | async->status = ret; |
4a69a410 CM |
798 | } |
799 | ||
06ea01b1 DS |
800 | /* |
801 | * In order to insert checksums into the metadata in large chunks, we wait | |
802 | * until bio submission time. All the pages in the bio are checksummed and | |
803 | * sums are attached onto the ordered extent record. | |
804 | * | |
805 | * At IO completion time the csums attached on the ordered extent record are | |
806 | * inserted into the tree. | |
807 | */ | |
4a69a410 | 808 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 809 | { |
8b712842 | 810 | struct async_submit_bio *async; |
06ea01b1 DS |
811 | struct inode *inode; |
812 | blk_status_t ret; | |
8b712842 CM |
813 | |
814 | async = container_of(work, struct async_submit_bio, work); | |
8896a08d | 815 | inode = async->inode; |
4854ddd0 | 816 | |
bb7ab3b9 | 817 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
818 | if (async->status) { |
819 | async->bio->bi_status = async->status; | |
4246a0b6 | 820 | bio_endio(async->bio); |
79787eaa JM |
821 | return; |
822 | } | |
823 | ||
ec39f769 CM |
824 | /* |
825 | * All of the bios that pass through here are from async helpers. | |
826 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
827 | * This changes nothing when cgroups aren't in use. | |
828 | */ | |
829 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 830 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
831 | if (ret) { |
832 | async->bio->bi_status = ret; | |
833 | bio_endio(async->bio); | |
834 | } | |
4a69a410 CM |
835 | } |
836 | ||
837 | static void run_one_async_free(struct btrfs_work *work) | |
838 | { | |
839 | struct async_submit_bio *async; | |
840 | ||
841 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
842 | kfree(async); |
843 | } | |
844 | ||
8896a08d | 845 | blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, |
8c27cb35 | 846 | int mirror_num, unsigned long bio_flags, |
1941b64b | 847 | u64 dio_file_offset, |
e288c080 | 848 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e | 849 | { |
8896a08d | 850 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
851 | struct async_submit_bio *async; |
852 | ||
853 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
854 | if (!async) | |
4e4cbee9 | 855 | return BLK_STS_RESOURCE; |
44b8bd7e | 856 | |
8896a08d | 857 | async->inode = inode; |
44b8bd7e CM |
858 | async->bio = bio; |
859 | async->mirror_num = mirror_num; | |
4a69a410 | 860 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 861 | |
a0cac0ec OS |
862 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
863 | run_one_async_free); | |
4a69a410 | 864 | |
1941b64b | 865 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 866 | |
4e4cbee9 | 867 | async->status = 0; |
79787eaa | 868 | |
67f055c7 | 869 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 870 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 871 | |
5cdc7ad3 | 872 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
873 | return 0; |
874 | } | |
875 | ||
4e4cbee9 | 876 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 877 | { |
2c30c71b | 878 | struct bio_vec *bvec; |
ce3ed71a | 879 | struct btrfs_root *root; |
2b070cfe | 880 | int ret = 0; |
6dc4f100 | 881 | struct bvec_iter_all iter_all; |
ce3ed71a | 882 | |
c09abff8 | 883 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 884 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 885 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 886 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
887 | if (ret) |
888 | break; | |
ce3ed71a | 889 | } |
2c30c71b | 890 | |
4e4cbee9 | 891 | return errno_to_blk_status(ret); |
ce3ed71a CM |
892 | } |
893 | ||
8896a08d | 894 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 895 | u64 dio_file_offset) |
22c59948 | 896 | { |
8b712842 CM |
897 | /* |
898 | * when we're called for a write, we're already in the async | |
5443be45 | 899 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 900 | */ |
79787eaa | 901 | return btree_csum_one_bio(bio); |
4a69a410 | 902 | } |
22c59948 | 903 | |
f4dcfb30 | 904 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 905 | struct btrfs_inode *bi) |
de0022b9 | 906 | { |
4eef29ef | 907 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 908 | return false; |
6300463b | 909 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 910 | return false; |
9b4e675a | 911 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
912 | return false; |
913 | return true; | |
de0022b9 JB |
914 | } |
915 | ||
1b36294a NB |
916 | blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, |
917 | int mirror_num, unsigned long bio_flags) | |
44b8bd7e | 918 | { |
0b246afa | 919 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4e4cbee9 | 920 | blk_status_t ret; |
cad321ad | 921 | |
cfe94440 | 922 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
4a69a410 CM |
923 | /* |
924 | * called for a read, do the setup so that checksum validation | |
925 | * can happen in the async kernel threads | |
926 | */ | |
0b246afa JM |
927 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
928 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 929 | if (ret) |
61891923 | 930 | goto out_w_error; |
08635bae | 931 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
f4dcfb30 | 932 | } else if (!should_async_write(fs_info, BTRFS_I(inode))) { |
de0022b9 JB |
933 | ret = btree_csum_one_bio(bio); |
934 | if (ret) | |
61891923 | 935 | goto out_w_error; |
08635bae | 936 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
61891923 SB |
937 | } else { |
938 | /* | |
939 | * kthread helpers are used to submit writes so that | |
940 | * checksumming can happen in parallel across all CPUs | |
941 | */ | |
8896a08d QW |
942 | ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, |
943 | 0, btree_submit_bio_start); | |
44b8bd7e | 944 | } |
d313d7a3 | 945 | |
4246a0b6 CH |
946 | if (ret) |
947 | goto out_w_error; | |
948 | return 0; | |
949 | ||
61891923 | 950 | out_w_error: |
4e4cbee9 | 951 | bio->bi_status = ret; |
4246a0b6 | 952 | bio_endio(bio); |
61891923 | 953 | return ret; |
44b8bd7e CM |
954 | } |
955 | ||
3dd1462e | 956 | #ifdef CONFIG_MIGRATION |
784b4e29 | 957 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
958 | struct page *newpage, struct page *page, |
959 | enum migrate_mode mode) | |
784b4e29 CM |
960 | { |
961 | /* | |
962 | * we can't safely write a btree page from here, | |
963 | * we haven't done the locking hook | |
964 | */ | |
965 | if (PageDirty(page)) | |
966 | return -EAGAIN; | |
967 | /* | |
968 | * Buffers may be managed in a filesystem specific way. | |
969 | * We must have no buffers or drop them. | |
970 | */ | |
971 | if (page_has_private(page) && | |
972 | !try_to_release_page(page, GFP_KERNEL)) | |
973 | return -EAGAIN; | |
a6bc32b8 | 974 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 975 | } |
3dd1462e | 976 | #endif |
784b4e29 | 977 | |
0da5468f CM |
978 | |
979 | static int btree_writepages(struct address_space *mapping, | |
980 | struct writeback_control *wbc) | |
981 | { | |
e2d84521 MX |
982 | struct btrfs_fs_info *fs_info; |
983 | int ret; | |
984 | ||
d8d5f3e1 | 985 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
986 | |
987 | if (wbc->for_kupdate) | |
988 | return 0; | |
989 | ||
e2d84521 | 990 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 991 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
992 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
993 | BTRFS_DIRTY_METADATA_THRESH, | |
994 | fs_info->dirty_metadata_batch); | |
e2d84521 | 995 | if (ret < 0) |
793955bc | 996 | return 0; |
793955bc | 997 | } |
0b32f4bb | 998 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
999 | } |
1000 | ||
70dec807 | 1001 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1002 | { |
98509cfc | 1003 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1004 | return 0; |
0c4e538b | 1005 | |
f7a52a40 | 1006 | return try_release_extent_buffer(page); |
d98237b3 CM |
1007 | } |
1008 | ||
d47992f8 LC |
1009 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1010 | unsigned int length) | |
d98237b3 | 1011 | { |
d1310b2e CM |
1012 | struct extent_io_tree *tree; |
1013 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1014 | extent_invalidatepage(tree, page, offset); |
1015 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1016 | if (PagePrivate(page)) { |
efe120a0 FH |
1017 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1018 | "page private not zero on page %llu", | |
1019 | (unsigned long long)page_offset(page)); | |
d1b89bc0 | 1020 | detach_page_private(page); |
9ad6b7bc | 1021 | } |
d98237b3 CM |
1022 | } |
1023 | ||
0b32f4bb JB |
1024 | static int btree_set_page_dirty(struct page *page) |
1025 | { | |
bb146eb2 | 1026 | #ifdef DEBUG |
139e8cd3 QW |
1027 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); |
1028 | struct btrfs_subpage *subpage; | |
0b32f4bb | 1029 | struct extent_buffer *eb; |
139e8cd3 QW |
1030 | int cur_bit = 0; |
1031 | u64 page_start = page_offset(page); | |
1032 | ||
1033 | if (fs_info->sectorsize == PAGE_SIZE) { | |
1034 | BUG_ON(!PagePrivate(page)); | |
1035 | eb = (struct extent_buffer *)page->private; | |
1036 | BUG_ON(!eb); | |
1037 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1038 | BUG_ON(!atomic_read(&eb->refs)); | |
1039 | btrfs_assert_tree_locked(eb); | |
1040 | return __set_page_dirty_nobuffers(page); | |
1041 | } | |
1042 | ASSERT(PagePrivate(page) && page->private); | |
1043 | subpage = (struct btrfs_subpage *)page->private; | |
1044 | ||
1045 | ASSERT(subpage->dirty_bitmap); | |
1046 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
1047 | unsigned long flags; | |
1048 | u64 cur; | |
1049 | u16 tmp = (1 << cur_bit); | |
1050 | ||
1051 | spin_lock_irqsave(&subpage->lock, flags); | |
1052 | if (!(tmp & subpage->dirty_bitmap)) { | |
1053 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1054 | cur_bit++; | |
1055 | continue; | |
1056 | } | |
1057 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1058 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 1059 | |
139e8cd3 QW |
1060 | eb = find_extent_buffer(fs_info, cur); |
1061 | ASSERT(eb); | |
1062 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1063 | ASSERT(atomic_read(&eb->refs)); | |
1064 | btrfs_assert_tree_locked(eb); | |
1065 | free_extent_buffer(eb); | |
1066 | ||
1067 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
1068 | } | |
bb146eb2 | 1069 | #endif |
0b32f4bb JB |
1070 | return __set_page_dirty_nobuffers(page); |
1071 | } | |
1072 | ||
7f09410b | 1073 | static const struct address_space_operations btree_aops = { |
0da5468f | 1074 | .writepages = btree_writepages, |
5f39d397 CM |
1075 | .releasepage = btree_releasepage, |
1076 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1077 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1078 | .migratepage = btree_migratepage, |
5a92bc88 | 1079 | #endif |
0b32f4bb | 1080 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1081 | }; |
1082 | ||
2ff7e61e JM |
1083 | struct extent_buffer *btrfs_find_create_tree_block( |
1084 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
1085 | u64 bytenr, u64 owner_root, |
1086 | int level) | |
0999df54 | 1087 | { |
0b246afa JM |
1088 | if (btrfs_is_testing(fs_info)) |
1089 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 1090 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
1091 | } |
1092 | ||
581c1760 QW |
1093 | /* |
1094 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1095 | * verification. | |
1096 | * | |
1b7ec85e | 1097 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
1098 | * @parent_transid: expected transid of this tree block, skip check if 0 |
1099 | * @level: expected level, mandatory check | |
1100 | * @first_key: expected key in slot 0, skip check if NULL | |
1101 | */ | |
2ff7e61e | 1102 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
1103 | u64 owner_root, u64 parent_transid, |
1104 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
1105 | { |
1106 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1107 | int ret; |
1108 | ||
3fbaf258 | 1109 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
1110 | if (IS_ERR(buf)) |
1111 | return buf; | |
0999df54 | 1112 | |
5ab12d1f | 1113 | ret = btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 1114 | level, first_key); |
0f0fe8f7 | 1115 | if (ret) { |
537f38f0 | 1116 | free_extent_buffer_stale(buf); |
64c043de | 1117 | return ERR_PTR(ret); |
0f0fe8f7 | 1118 | } |
5f39d397 | 1119 | return buf; |
ce9adaa5 | 1120 | |
eb60ceac CM |
1121 | } |
1122 | ||
6a884d7d | 1123 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1124 | { |
6a884d7d | 1125 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1126 | if (btrfs_header_generation(buf) == |
e2d84521 | 1127 | fs_info->running_transaction->transid) { |
b9447ef8 | 1128 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1129 | |
b9473439 | 1130 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1131 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1132 | -buf->len, | |
1133 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1134 | clear_extent_buffer_dirty(buf); |
1135 | } | |
925baedd | 1136 | } |
5f39d397 CM |
1137 | } |
1138 | ||
da17066c | 1139 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1140 | u64 objectid) |
d97e63b6 | 1141 | { |
7c0260ee | 1142 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
96dfcb46 | 1143 | root->fs_info = fs_info; |
cfaa7295 | 1144 | root->node = NULL; |
a28ec197 | 1145 | root->commit_root = NULL; |
27cdeb70 | 1146 | root->state = 0; |
d68fc57b | 1147 | root->orphan_cleanup_state = 0; |
0b86a832 | 1148 | |
0f7d52f4 | 1149 | root->last_trans = 0; |
6b8fad57 | 1150 | root->free_objectid = 0; |
eb73c1b7 | 1151 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1152 | root->nr_ordered_extents = 0; |
6bef4d31 | 1153 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1154 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1155 | root->block_rsv = NULL; |
0b86a832 CM |
1156 | |
1157 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1158 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1159 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1160 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1161 | INIT_LIST_HEAD(&root->ordered_extents); |
1162 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1163 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1164 | INIT_LIST_HEAD(&root->logged_list[0]); |
1165 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1166 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1167 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1168 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1169 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1170 | spin_lock_init(&root->log_extents_lock[0]); |
1171 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1172 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1173 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1174 | mutex_init(&root->log_mutex); |
31f3d255 | 1175 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1176 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1177 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1178 | init_waitqueue_head(&root->log_writer_wait); |
1179 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1180 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1181 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1182 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1183 | atomic_set(&root->log_commit[0], 0); |
1184 | atomic_set(&root->log_commit[1], 0); | |
1185 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1186 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1187 | refcount_set(&root->refs, 1); |
8ecebf4d | 1188 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1189 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1190 | root->log_transid = 0; |
d1433deb | 1191 | root->log_transid_committed = -1; |
257c62e1 | 1192 | root->last_log_commit = 0; |
e289f03e | 1193 | if (!dummy) { |
43eb5f29 QW |
1194 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1195 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1196 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1197 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1198 | } | |
017e5369 | 1199 | |
3768f368 CM |
1200 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1201 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1202 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
4d775673 | 1203 | root->root_key.objectid = objectid; |
0ee5dc67 | 1204 | root->anon_dev = 0; |
8ea05e3a | 1205 | |
5f3ab90a | 1206 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1207 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1208 | #ifdef CONFIG_BTRFS_DEBUG |
1209 | INIT_LIST_HEAD(&root->leak_list); | |
1210 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1211 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); | |
1212 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1213 | #endif | |
3768f368 CM |
1214 | } |
1215 | ||
74e4d827 | 1216 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1217 | u64 objectid, gfp_t flags) |
6f07e42e | 1218 | { |
74e4d827 | 1219 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1220 | if (root) |
96dfcb46 | 1221 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1222 | return root; |
1223 | } | |
1224 | ||
06ea65a3 JB |
1225 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1226 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1227 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1228 | { |
1229 | struct btrfs_root *root; | |
1230 | ||
7c0260ee JM |
1231 | if (!fs_info) |
1232 | return ERR_PTR(-EINVAL); | |
1233 | ||
96dfcb46 | 1234 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1235 | if (!root) |
1236 | return ERR_PTR(-ENOMEM); | |
da17066c | 1237 | |
b9ef22de | 1238 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1239 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1240 | |
1241 | return root; | |
1242 | } | |
1243 | #endif | |
1244 | ||
20897f5c | 1245 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1246 | u64 objectid) |
1247 | { | |
9b7a2440 | 1248 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1249 | struct extent_buffer *leaf; |
1250 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1251 | struct btrfs_root *root; | |
1252 | struct btrfs_key key; | |
b89f6d1f | 1253 | unsigned int nofs_flag; |
20897f5c | 1254 | int ret = 0; |
20897f5c | 1255 | |
b89f6d1f FM |
1256 | /* |
1257 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1258 | * context to avoid deadlock if reclaim happens. | |
1259 | */ | |
1260 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1261 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1262 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1263 | if (!root) |
1264 | return ERR_PTR(-ENOMEM); | |
1265 | ||
20897f5c AJ |
1266 | root->root_key.objectid = objectid; |
1267 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1268 | root->root_key.offset = 0; | |
1269 | ||
9631e4cc JB |
1270 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1271 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1272 | if (IS_ERR(leaf)) { |
1273 | ret = PTR_ERR(leaf); | |
1dd05682 | 1274 | leaf = NULL; |
8a6a87cd | 1275 | goto fail_unlock; |
20897f5c AJ |
1276 | } |
1277 | ||
20897f5c | 1278 | root->node = leaf; |
20897f5c AJ |
1279 | btrfs_mark_buffer_dirty(leaf); |
1280 | ||
1281 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1282 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1283 | |
f944d2cb DS |
1284 | btrfs_set_root_flags(&root->root_item, 0); |
1285 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1286 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1287 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1288 | btrfs_set_root_level(&root->root_item, 0); | |
1289 | btrfs_set_root_refs(&root->root_item, 1); | |
1290 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1291 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1292 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1293 | if (is_fstree(objectid)) |
807fc790 AS |
1294 | generate_random_guid(root->root_item.uuid); |
1295 | else | |
1296 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1297 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1298 | |
8a6a87cd BB |
1299 | btrfs_tree_unlock(leaf); |
1300 | ||
20897f5c AJ |
1301 | key.objectid = objectid; |
1302 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1303 | key.offset = 0; | |
1304 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1305 | if (ret) | |
1306 | goto fail; | |
1307 | ||
1dd05682 TI |
1308 | return root; |
1309 | ||
8a6a87cd | 1310 | fail_unlock: |
8c38938c | 1311 | if (leaf) |
1dd05682 | 1312 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1313 | fail: |
00246528 | 1314 | btrfs_put_root(root); |
20897f5c | 1315 | |
1dd05682 | 1316 | return ERR_PTR(ret); |
20897f5c AJ |
1317 | } |
1318 | ||
7237f183 YZ |
1319 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1320 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1321 | { |
1322 | struct btrfs_root *root; | |
e02119d5 | 1323 | |
96dfcb46 | 1324 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1325 | if (!root) |
7237f183 | 1326 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1327 | |
e02119d5 CM |
1328 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1329 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1330 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1331 | |
6ab6ebb7 NA |
1332 | return root; |
1333 | } | |
1334 | ||
1335 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1336 | struct btrfs_root *root) | |
1337 | { | |
1338 | struct extent_buffer *leaf; | |
1339 | ||
7237f183 | 1340 | /* |
92a7cc42 | 1341 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1342 | * |
92a7cc42 QW |
1343 | * Log trees are not exposed to user space thus can't be snapshotted, |
1344 | * and they go away before a real commit is actually done. | |
1345 | * | |
1346 | * They do store pointers to file data extents, and those reference | |
1347 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1348 | */ |
e02119d5 | 1349 | |
4d75f8a9 | 1350 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1351 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1352 | if (IS_ERR(leaf)) |
1353 | return PTR_ERR(leaf); | |
e02119d5 | 1354 | |
7237f183 | 1355 | root->node = leaf; |
e02119d5 | 1356 | |
e02119d5 CM |
1357 | btrfs_mark_buffer_dirty(root->node); |
1358 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1359 | |
1360 | return 0; | |
7237f183 YZ |
1361 | } |
1362 | ||
1363 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1364 | struct btrfs_fs_info *fs_info) | |
1365 | { | |
1366 | struct btrfs_root *log_root; | |
1367 | ||
1368 | log_root = alloc_log_tree(trans, fs_info); | |
1369 | if (IS_ERR(log_root)) | |
1370 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1371 | |
3ddebf27 NA |
1372 | if (!btrfs_is_zoned(fs_info)) { |
1373 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1374 | ||
1375 | if (ret) { | |
1376 | btrfs_put_root(log_root); | |
1377 | return ret; | |
1378 | } | |
6ab6ebb7 NA |
1379 | } |
1380 | ||
7237f183 YZ |
1381 | WARN_ON(fs_info->log_root_tree); |
1382 | fs_info->log_root_tree = log_root; | |
1383 | return 0; | |
1384 | } | |
1385 | ||
1386 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1387 | struct btrfs_root *root) | |
1388 | { | |
0b246afa | 1389 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1390 | struct btrfs_root *log_root; |
1391 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1392 | int ret; |
7237f183 | 1393 | |
0b246afa | 1394 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1395 | if (IS_ERR(log_root)) |
1396 | return PTR_ERR(log_root); | |
1397 | ||
6ab6ebb7 NA |
1398 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1399 | if (ret) { | |
1400 | btrfs_put_root(log_root); | |
1401 | return ret; | |
1402 | } | |
1403 | ||
7237f183 YZ |
1404 | log_root->last_trans = trans->transid; |
1405 | log_root->root_key.offset = root->root_key.objectid; | |
1406 | ||
1407 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1408 | btrfs_set_stack_inode_generation(inode_item, 1); |
1409 | btrfs_set_stack_inode_size(inode_item, 3); | |
1410 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1411 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1412 | fs_info->nodesize); |
3cae210f | 1413 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1414 | |
5d4f98a2 | 1415 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1416 | |
1417 | WARN_ON(root->log_root); | |
1418 | root->log_root = log_root; | |
1419 | root->log_transid = 0; | |
d1433deb | 1420 | root->log_transid_committed = -1; |
257c62e1 | 1421 | root->last_log_commit = 0; |
e02119d5 CM |
1422 | return 0; |
1423 | } | |
1424 | ||
49d11bea JB |
1425 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1426 | struct btrfs_path *path, | |
1427 | struct btrfs_key *key) | |
e02119d5 CM |
1428 | { |
1429 | struct btrfs_root *root; | |
1430 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1431 | u64 generation; |
cb517eab | 1432 | int ret; |
581c1760 | 1433 | int level; |
0f7d52f4 | 1434 | |
96dfcb46 | 1435 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1436 | if (!root) |
1437 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1438 | |
cb517eab MX |
1439 | ret = btrfs_find_root(tree_root, key, path, |
1440 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1441 | if (ret) { |
13a8a7c8 YZ |
1442 | if (ret > 0) |
1443 | ret = -ENOENT; | |
49d11bea | 1444 | goto fail; |
0f7d52f4 | 1445 | } |
13a8a7c8 | 1446 | |
84234f3a | 1447 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1448 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1449 | root->node = read_tree_block(fs_info, |
1450 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1451 | key->objectid, generation, level, NULL); |
64c043de LB |
1452 | if (IS_ERR(root->node)) { |
1453 | ret = PTR_ERR(root->node); | |
8c38938c | 1454 | root->node = NULL; |
49d11bea | 1455 | goto fail; |
cb517eab MX |
1456 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { |
1457 | ret = -EIO; | |
49d11bea | 1458 | goto fail; |
416bc658 | 1459 | } |
5d4f98a2 | 1460 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1461 | return root; |
49d11bea | 1462 | fail: |
00246528 | 1463 | btrfs_put_root(root); |
49d11bea JB |
1464 | return ERR_PTR(ret); |
1465 | } | |
1466 | ||
1467 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1468 | struct btrfs_key *key) | |
1469 | { | |
1470 | struct btrfs_root *root; | |
1471 | struct btrfs_path *path; | |
1472 | ||
1473 | path = btrfs_alloc_path(); | |
1474 | if (!path) | |
1475 | return ERR_PTR(-ENOMEM); | |
1476 | root = read_tree_root_path(tree_root, path, key); | |
1477 | btrfs_free_path(path); | |
1478 | ||
1479 | return root; | |
cb517eab MX |
1480 | } |
1481 | ||
2dfb1e43 QW |
1482 | /* |
1483 | * Initialize subvolume root in-memory structure | |
1484 | * | |
1485 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1486 | */ | |
1487 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1488 | { |
1489 | int ret; | |
dcc3eb96 | 1490 | unsigned int nofs_flag; |
cb517eab | 1491 | |
dcc3eb96 NB |
1492 | /* |
1493 | * We might be called under a transaction (e.g. indirect backref | |
1494 | * resolution) which could deadlock if it triggers memory reclaim | |
1495 | */ | |
1496 | nofs_flag = memalloc_nofs_save(); | |
1497 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1498 | memalloc_nofs_restore(nofs_flag); | |
1499 | if (ret) | |
8257b2dc | 1500 | goto fail; |
8257b2dc | 1501 | |
aeb935a4 | 1502 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1503 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1504 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1505 | btrfs_check_and_init_root_item(&root->root_item); |
1506 | } | |
1507 | ||
851fd730 QW |
1508 | /* |
1509 | * Don't assign anonymous block device to roots that are not exposed to | |
1510 | * userspace, the id pool is limited to 1M | |
1511 | */ | |
1512 | if (is_fstree(root->root_key.objectid) && | |
1513 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1514 | if (!anon_dev) { |
1515 | ret = get_anon_bdev(&root->anon_dev); | |
1516 | if (ret) | |
1517 | goto fail; | |
1518 | } else { | |
1519 | root->anon_dev = anon_dev; | |
1520 | } | |
851fd730 | 1521 | } |
f32e48e9 CR |
1522 | |
1523 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1524 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1525 | if (ret) { |
1526 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1527 | goto fail; |
f32e48e9 CR |
1528 | } |
1529 | ||
6b8fad57 | 1530 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1531 | |
1532 | mutex_unlock(&root->objectid_mutex); | |
1533 | ||
cb517eab MX |
1534 | return 0; |
1535 | fail: | |
84db5ccf | 1536 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1537 | return ret; |
1538 | } | |
1539 | ||
a98db0f3 JB |
1540 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1541 | u64 root_id) | |
cb517eab MX |
1542 | { |
1543 | struct btrfs_root *root; | |
1544 | ||
1545 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1546 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1547 | (unsigned long)root_id); | |
bc44d7c4 | 1548 | if (root) |
00246528 | 1549 | root = btrfs_grab_root(root); |
cb517eab MX |
1550 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1551 | return root; | |
1552 | } | |
1553 | ||
49d11bea JB |
1554 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1555 | u64 objectid) | |
1556 | { | |
1557 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1558 | return btrfs_grab_root(fs_info->tree_root); | |
1559 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1560 | return btrfs_grab_root(fs_info->extent_root); | |
1561 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) | |
1562 | return btrfs_grab_root(fs_info->chunk_root); | |
1563 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1564 | return btrfs_grab_root(fs_info->dev_root); | |
1565 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
1566 | return btrfs_grab_root(fs_info->csum_root); | |
1567 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) | |
1568 | return btrfs_grab_root(fs_info->quota_root) ? | |
1569 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1570 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1571 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1572 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
1573 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) | |
1574 | return btrfs_grab_root(fs_info->free_space_root) ? | |
1575 | fs_info->free_space_root : ERR_PTR(-ENOENT); | |
1576 | return NULL; | |
1577 | } | |
1578 | ||
cb517eab MX |
1579 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1580 | struct btrfs_root *root) | |
1581 | { | |
1582 | int ret; | |
1583 | ||
e1860a77 | 1584 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1585 | if (ret) |
1586 | return ret; | |
1587 | ||
1588 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1589 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1590 | (unsigned long)root->root_key.objectid, | |
1591 | root); | |
af01d2e5 | 1592 | if (ret == 0) { |
00246528 | 1593 | btrfs_grab_root(root); |
27cdeb70 | 1594 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1595 | } |
cb517eab MX |
1596 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1597 | radix_tree_preload_end(); | |
1598 | ||
1599 | return ret; | |
1600 | } | |
1601 | ||
bd647ce3 JB |
1602 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1603 | { | |
1604 | #ifdef CONFIG_BTRFS_DEBUG | |
1605 | struct btrfs_root *root; | |
1606 | ||
1607 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1608 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1609 | ||
bd647ce3 JB |
1610 | root = list_first_entry(&fs_info->allocated_roots, |
1611 | struct btrfs_root, leak_list); | |
457f1864 | 1612 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1613 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1614 | refcount_read(&root->refs)); |
1615 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1616 | btrfs_put_root(root); |
1617 | btrfs_put_root(root); | |
bd647ce3 JB |
1618 | } |
1619 | #endif | |
1620 | } | |
1621 | ||
0d4b0463 JB |
1622 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1623 | { | |
141386e1 JB |
1624 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1625 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1626 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1627 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1628 | btrfs_free_csum_hash(fs_info); | |
1629 | btrfs_free_stripe_hash_table(fs_info); | |
1630 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1631 | kfree(fs_info->balance_ctl); |
1632 | kfree(fs_info->delayed_root); | |
00246528 JB |
1633 | btrfs_put_root(fs_info->extent_root); |
1634 | btrfs_put_root(fs_info->tree_root); | |
1635 | btrfs_put_root(fs_info->chunk_root); | |
1636 | btrfs_put_root(fs_info->dev_root); | |
1637 | btrfs_put_root(fs_info->csum_root); | |
1638 | btrfs_put_root(fs_info->quota_root); | |
1639 | btrfs_put_root(fs_info->uuid_root); | |
1640 | btrfs_put_root(fs_info->free_space_root); | |
1641 | btrfs_put_root(fs_info->fs_root); | |
aeb935a4 | 1642 | btrfs_put_root(fs_info->data_reloc_root); |
bd647ce3 | 1643 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1644 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1645 | kfree(fs_info->super_copy); |
1646 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1647 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1648 | kvfree(fs_info); |
1649 | } | |
1650 | ||
1651 | ||
2dfb1e43 QW |
1652 | /* |
1653 | * Get an in-memory reference of a root structure. | |
1654 | * | |
1655 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1656 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1657 | * found, then read it from disk and add it to cached fs roots. | |
1658 | * | |
1659 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1660 | * | |
1661 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1662 | * same root objectid. | |
1663 | * | |
1664 | * @objectid: root id | |
1665 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1666 | * pass 0 for new allocation. | |
1667 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1668 | * for orphan roots | |
1669 | */ | |
1670 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1671 | u64 objectid, dev_t anon_dev, | |
1672 | bool check_ref) | |
5eda7b5e CM |
1673 | { |
1674 | struct btrfs_root *root; | |
381cf658 | 1675 | struct btrfs_path *path; |
1d4c08e0 | 1676 | struct btrfs_key key; |
5eda7b5e CM |
1677 | int ret; |
1678 | ||
49d11bea JB |
1679 | root = btrfs_get_global_root(fs_info, objectid); |
1680 | if (root) | |
1681 | return root; | |
4df27c4d | 1682 | again: |
56e9357a | 1683 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1684 | if (root) { |
2dfb1e43 QW |
1685 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1686 | ASSERT(!anon_dev); | |
bc44d7c4 | 1687 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1688 | btrfs_put_root(root); |
48475471 | 1689 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1690 | } |
5eda7b5e | 1691 | return root; |
48475471 | 1692 | } |
5eda7b5e | 1693 | |
56e9357a DS |
1694 | key.objectid = objectid; |
1695 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1696 | key.offset = (u64)-1; | |
1697 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1698 | if (IS_ERR(root)) |
1699 | return root; | |
3394e160 | 1700 | |
c00869f1 | 1701 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1702 | ret = -ENOENT; |
581bb050 | 1703 | goto fail; |
35a30d7c | 1704 | } |
581bb050 | 1705 | |
2dfb1e43 | 1706 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1707 | if (ret) |
1708 | goto fail; | |
3394e160 | 1709 | |
381cf658 DS |
1710 | path = btrfs_alloc_path(); |
1711 | if (!path) { | |
1712 | ret = -ENOMEM; | |
1713 | goto fail; | |
1714 | } | |
1d4c08e0 DS |
1715 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1716 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1717 | key.offset = objectid; |
1d4c08e0 DS |
1718 | |
1719 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1720 | btrfs_free_path(path); |
d68fc57b YZ |
1721 | if (ret < 0) |
1722 | goto fail; | |
1723 | if (ret == 0) | |
27cdeb70 | 1724 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1725 | |
cb517eab | 1726 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1727 | if (ret) { |
00246528 | 1728 | btrfs_put_root(root); |
4785e24f | 1729 | if (ret == -EEXIST) |
4df27c4d | 1730 | goto again; |
4df27c4d | 1731 | goto fail; |
0f7d52f4 | 1732 | } |
edbd8d4e | 1733 | return root; |
4df27c4d | 1734 | fail: |
8c38938c | 1735 | btrfs_put_root(root); |
4df27c4d | 1736 | return ERR_PTR(ret); |
edbd8d4e CM |
1737 | } |
1738 | ||
2dfb1e43 QW |
1739 | /* |
1740 | * Get in-memory reference of a root structure | |
1741 | * | |
1742 | * @objectid: tree objectid | |
1743 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1744 | * one reference | |
1745 | */ | |
1746 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1747 | u64 objectid, bool check_ref) | |
1748 | { | |
1749 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1750 | } | |
1751 | ||
1752 | /* | |
1753 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1754 | * the anonymous block device id | |
1755 | * | |
1756 | * @objectid: tree objectid | |
1757 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1758 | * parameter value | |
1759 | */ | |
1760 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1761 | u64 objectid, dev_t anon_dev) | |
1762 | { | |
1763 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1764 | } | |
1765 | ||
49d11bea JB |
1766 | /* |
1767 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1768 | * @fs_info: the fs_info | |
1769 | * @objectid: the objectid we need to lookup | |
1770 | * | |
1771 | * This is exclusively used for backref walking, and exists specifically because | |
1772 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1773 | * creation time, which means we may have to read the tree_root in order to look | |
1774 | * up a fs root that is not in memory. If the root is not in memory we will | |
1775 | * read the tree root commit root and look up the fs root from there. This is a | |
1776 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1777 | * have the most uptodate information, it'll simply be discarded once the | |
1778 | * backref code is finished using the root. | |
1779 | */ | |
1780 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1781 | struct btrfs_path *path, | |
1782 | u64 objectid) | |
1783 | { | |
1784 | struct btrfs_root *root; | |
1785 | struct btrfs_key key; | |
1786 | ||
1787 | ASSERT(path->search_commit_root && path->skip_locking); | |
1788 | ||
1789 | /* | |
1790 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1791 | * since this is called via the backref walking code we won't be looking | |
1792 | * up a root that doesn't exist, unless there's corruption. So if root | |
1793 | * != NULL just return it. | |
1794 | */ | |
1795 | root = btrfs_get_global_root(fs_info, objectid); | |
1796 | if (root) | |
1797 | return root; | |
1798 | ||
1799 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1800 | if (root) | |
1801 | return root; | |
1802 | ||
1803 | key.objectid = objectid; | |
1804 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1805 | key.offset = (u64)-1; | |
1806 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1807 | btrfs_release_path(path); | |
1808 | ||
1809 | return root; | |
1810 | } | |
1811 | ||
8b712842 CM |
1812 | /* |
1813 | * called by the kthread helper functions to finally call the bio end_io | |
1814 | * functions. This is where read checksum verification actually happens | |
1815 | */ | |
1816 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1817 | { |
ce9adaa5 | 1818 | struct bio *bio; |
97eb6b69 | 1819 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1820 | |
97eb6b69 | 1821 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1822 | bio = end_io_wq->bio; |
ce9adaa5 | 1823 | |
4e4cbee9 | 1824 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1825 | bio->bi_private = end_io_wq->private; |
1826 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1827 | bio_endio(bio); |
9be490f1 | 1828 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1829 | } |
1830 | ||
a74a4b97 CM |
1831 | static int cleaner_kthread(void *arg) |
1832 | { | |
1833 | struct btrfs_root *root = arg; | |
0b246afa | 1834 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1835 | int again; |
a74a4b97 | 1836 | |
d6fd0ae2 | 1837 | while (1) { |
d0278245 | 1838 | again = 0; |
a74a4b97 | 1839 | |
fd340d0f JB |
1840 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1841 | ||
d0278245 | 1842 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1843 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1844 | goto sleep; |
1845 | ||
90c711ab ZB |
1846 | /* |
1847 | * Do not do anything if we might cause open_ctree() to block | |
1848 | * before we have finished mounting the filesystem. | |
1849 | */ | |
0b246afa | 1850 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1851 | goto sleep; |
1852 | ||
0b246afa | 1853 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1854 | goto sleep; |
1855 | ||
dc7f370c MX |
1856 | /* |
1857 | * Avoid the problem that we change the status of the fs | |
1858 | * during the above check and trylock. | |
1859 | */ | |
2ff7e61e | 1860 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1861 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1862 | goto sleep; |
76dda93c | 1863 | } |
a74a4b97 | 1864 | |
2ff7e61e | 1865 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1866 | |
d0278245 | 1867 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1868 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1869 | |
1870 | /* | |
05323cd1 MX |
1871 | * The defragger has dealt with the R/O remount and umount, |
1872 | * needn't do anything special here. | |
d0278245 | 1873 | */ |
0b246afa | 1874 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1875 | |
1876 | /* | |
f3372065 | 1877 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1878 | * with relocation (btrfs_relocate_chunk) and relocation |
1879 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1880 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1881 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1882 | * unused block groups. | |
1883 | */ | |
0b246afa | 1884 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1885 | |
1886 | /* | |
1887 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1888 | * all unused block_groups. This possibly gives us some more free | |
1889 | * space. | |
1890 | */ | |
1891 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1892 | sleep: |
a0a1db70 | 1893 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1894 | if (kthread_should_park()) |
1895 | kthread_parkme(); | |
1896 | if (kthread_should_stop()) | |
1897 | return 0; | |
838fe188 | 1898 | if (!again) { |
a74a4b97 | 1899 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1900 | schedule(); |
a74a4b97 CM |
1901 | __set_current_state(TASK_RUNNING); |
1902 | } | |
da288d28 | 1903 | } |
a74a4b97 CM |
1904 | } |
1905 | ||
1906 | static int transaction_kthread(void *arg) | |
1907 | { | |
1908 | struct btrfs_root *root = arg; | |
0b246afa | 1909 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1910 | struct btrfs_trans_handle *trans; |
1911 | struct btrfs_transaction *cur; | |
8929ecfa | 1912 | u64 transid; |
643900be | 1913 | time64_t delta; |
a74a4b97 | 1914 | unsigned long delay; |
914b2007 | 1915 | bool cannot_commit; |
a74a4b97 CM |
1916 | |
1917 | do { | |
914b2007 | 1918 | cannot_commit = false; |
ba1bc00f | 1919 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1920 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1921 | |
0b246afa JM |
1922 | spin_lock(&fs_info->trans_lock); |
1923 | cur = fs_info->running_transaction; | |
a74a4b97 | 1924 | if (!cur) { |
0b246afa | 1925 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1926 | goto sleep; |
1927 | } | |
31153d81 | 1928 | |
643900be | 1929 | delta = ktime_get_seconds() - cur->start_time; |
3296bf56 | 1930 | if (cur->state < TRANS_STATE_COMMIT_START && |
643900be | 1931 | delta < fs_info->commit_interval) { |
0b246afa | 1932 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1933 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1934 | delay = min(delay, | |
1935 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1936 | goto sleep; |
1937 | } | |
8929ecfa | 1938 | transid = cur->transid; |
0b246afa | 1939 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1940 | |
79787eaa | 1941 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1942 | trans = btrfs_attach_transaction(root); |
914b2007 | 1943 | if (IS_ERR(trans)) { |
354aa0fb MX |
1944 | if (PTR_ERR(trans) != -ENOENT) |
1945 | cannot_commit = true; | |
79787eaa | 1946 | goto sleep; |
914b2007 | 1947 | } |
8929ecfa | 1948 | if (transid == trans->transid) { |
3a45bb20 | 1949 | btrfs_commit_transaction(trans); |
8929ecfa | 1950 | } else { |
3a45bb20 | 1951 | btrfs_end_transaction(trans); |
8929ecfa | 1952 | } |
a74a4b97 | 1953 | sleep: |
0b246afa JM |
1954 | wake_up_process(fs_info->cleaner_kthread); |
1955 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1956 | |
4e121c06 | 1957 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1958 | &fs_info->fs_state))) |
2ff7e61e | 1959 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1960 | if (!kthread_should_stop() && |
0b246afa | 1961 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1962 | cannot_commit)) |
bc5511d0 | 1963 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1964 | } while (!kthread_should_stop()); |
1965 | return 0; | |
1966 | } | |
1967 | ||
af31f5e5 | 1968 | /* |
01f0f9da NB |
1969 | * This will find the highest generation in the array of root backups. The |
1970 | * index of the highest array is returned, or -EINVAL if we can't find | |
1971 | * anything. | |
af31f5e5 CM |
1972 | * |
1973 | * We check to make sure the array is valid by comparing the | |
1974 | * generation of the latest root in the array with the generation | |
1975 | * in the super block. If they don't match we pitch it. | |
1976 | */ | |
01f0f9da | 1977 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1978 | { |
01f0f9da | 1979 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1980 | u64 cur; |
af31f5e5 CM |
1981 | struct btrfs_root_backup *root_backup; |
1982 | int i; | |
1983 | ||
1984 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1985 | root_backup = info->super_copy->super_roots + i; | |
1986 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1987 | if (cur == newest_gen) | |
01f0f9da | 1988 | return i; |
af31f5e5 CM |
1989 | } |
1990 | ||
01f0f9da | 1991 | return -EINVAL; |
af31f5e5 CM |
1992 | } |
1993 | ||
af31f5e5 CM |
1994 | /* |
1995 | * copy all the root pointers into the super backup array. | |
1996 | * this will bump the backup pointer by one when it is | |
1997 | * done | |
1998 | */ | |
1999 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2000 | { | |
6ef108dd | 2001 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2002 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
2003 | |
2004 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2005 | ||
2006 | /* | |
2007 | * make sure all of our padding and empty slots get zero filled | |
2008 | * regardless of which ones we use today | |
2009 | */ | |
2010 | memset(root_backup, 0, sizeof(*root_backup)); | |
2011 | ||
2012 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2013 | ||
2014 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2015 | btrfs_set_backup_tree_root_gen(root_backup, | |
2016 | btrfs_header_generation(info->tree_root->node)); | |
2017 | ||
2018 | btrfs_set_backup_tree_root_level(root_backup, | |
2019 | btrfs_header_level(info->tree_root->node)); | |
2020 | ||
2021 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2022 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2023 | btrfs_header_generation(info->chunk_root->node)); | |
2024 | btrfs_set_backup_chunk_root_level(root_backup, | |
2025 | btrfs_header_level(info->chunk_root->node)); | |
2026 | ||
2027 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2028 | btrfs_set_backup_extent_root_gen(root_backup, | |
2029 | btrfs_header_generation(info->extent_root->node)); | |
2030 | btrfs_set_backup_extent_root_level(root_backup, | |
2031 | btrfs_header_level(info->extent_root->node)); | |
2032 | ||
7c7e82a7 CM |
2033 | /* |
2034 | * we might commit during log recovery, which happens before we set | |
2035 | * the fs_root. Make sure it is valid before we fill it in. | |
2036 | */ | |
2037 | if (info->fs_root && info->fs_root->node) { | |
2038 | btrfs_set_backup_fs_root(root_backup, | |
2039 | info->fs_root->node->start); | |
2040 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2041 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2042 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2043 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2044 | } |
af31f5e5 CM |
2045 | |
2046 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2047 | btrfs_set_backup_dev_root_gen(root_backup, | |
2048 | btrfs_header_generation(info->dev_root->node)); | |
2049 | btrfs_set_backup_dev_root_level(root_backup, | |
2050 | btrfs_header_level(info->dev_root->node)); | |
2051 | ||
2052 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2053 | btrfs_set_backup_csum_root_gen(root_backup, | |
2054 | btrfs_header_generation(info->csum_root->node)); | |
2055 | btrfs_set_backup_csum_root_level(root_backup, | |
2056 | btrfs_header_level(info->csum_root->node)); | |
2057 | ||
2058 | btrfs_set_backup_total_bytes(root_backup, | |
2059 | btrfs_super_total_bytes(info->super_copy)); | |
2060 | btrfs_set_backup_bytes_used(root_backup, | |
2061 | btrfs_super_bytes_used(info->super_copy)); | |
2062 | btrfs_set_backup_num_devices(root_backup, | |
2063 | btrfs_super_num_devices(info->super_copy)); | |
2064 | ||
2065 | /* | |
2066 | * if we don't copy this out to the super_copy, it won't get remembered | |
2067 | * for the next commit | |
2068 | */ | |
2069 | memcpy(&info->super_copy->super_roots, | |
2070 | &info->super_for_commit->super_roots, | |
2071 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2072 | } | |
2073 | ||
bd2336b2 NB |
2074 | /* |
2075 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2076 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2077 | * | |
2078 | * fs_info - filesystem whose backup roots need to be read | |
2079 | * priority - priority of backup root required | |
2080 | * | |
2081 | * Returns backup root index on success and -EINVAL otherwise. | |
2082 | */ | |
2083 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2084 | { | |
2085 | int backup_index = find_newest_super_backup(fs_info); | |
2086 | struct btrfs_super_block *super = fs_info->super_copy; | |
2087 | struct btrfs_root_backup *root_backup; | |
2088 | ||
2089 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2090 | if (priority == 0) | |
2091 | return backup_index; | |
2092 | ||
2093 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2094 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2095 | } else { | |
2096 | return -EINVAL; | |
2097 | } | |
2098 | ||
2099 | root_backup = super->super_roots + backup_index; | |
2100 | ||
2101 | btrfs_set_super_generation(super, | |
2102 | btrfs_backup_tree_root_gen(root_backup)); | |
2103 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2104 | btrfs_set_super_root_level(super, | |
2105 | btrfs_backup_tree_root_level(root_backup)); | |
2106 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2107 | ||
2108 | /* | |
2109 | * Fixme: the total bytes and num_devices need to match or we should | |
2110 | * need a fsck | |
2111 | */ | |
2112 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2113 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2114 | ||
2115 | return backup_index; | |
2116 | } | |
2117 | ||
7abadb64 LB |
2118 | /* helper to cleanup workers */ |
2119 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2120 | { | |
dc6e3209 | 2121 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2122 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2123 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2124 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2125 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
d05a33ac | 2126 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2127 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2128 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2129 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2130 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2131 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2132 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2133 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2134 | if (fs_info->discard_ctl.discard_workers) |
2135 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2136 | /* |
2137 | * Now that all other work queues are destroyed, we can safely destroy | |
2138 | * the queues used for metadata I/O, since tasks from those other work | |
2139 | * queues can do metadata I/O operations. | |
2140 | */ | |
2141 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2142 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2143 | } |
2144 | ||
2e9f5954 R |
2145 | static void free_root_extent_buffers(struct btrfs_root *root) |
2146 | { | |
2147 | if (root) { | |
2148 | free_extent_buffer(root->node); | |
2149 | free_extent_buffer(root->commit_root); | |
2150 | root->node = NULL; | |
2151 | root->commit_root = NULL; | |
2152 | } | |
2153 | } | |
2154 | ||
af31f5e5 | 2155 | /* helper to cleanup tree roots */ |
4273eaff | 2156 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2157 | { |
2e9f5954 | 2158 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2159 | |
2e9f5954 R |
2160 | free_root_extent_buffers(info->dev_root); |
2161 | free_root_extent_buffers(info->extent_root); | |
2162 | free_root_extent_buffers(info->csum_root); | |
2163 | free_root_extent_buffers(info->quota_root); | |
2164 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2165 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2166 | free_root_extent_buffers(info->data_reloc_root); |
4273eaff | 2167 | if (free_chunk_root) |
2e9f5954 | 2168 | free_root_extent_buffers(info->chunk_root); |
70f6d82e | 2169 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2170 | } |
2171 | ||
8c38938c JB |
2172 | void btrfs_put_root(struct btrfs_root *root) |
2173 | { | |
2174 | if (!root) | |
2175 | return; | |
2176 | ||
2177 | if (refcount_dec_and_test(&root->refs)) { | |
2178 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2179 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2180 | if (root->anon_dev) |
2181 | free_anon_bdev(root->anon_dev); | |
2182 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2183 | free_root_extent_buffers(root); |
8c38938c JB |
2184 | #ifdef CONFIG_BTRFS_DEBUG |
2185 | spin_lock(&root->fs_info->fs_roots_radix_lock); | |
2186 | list_del_init(&root->leak_list); | |
2187 | spin_unlock(&root->fs_info->fs_roots_radix_lock); | |
2188 | #endif | |
2189 | kfree(root); | |
2190 | } | |
2191 | } | |
2192 | ||
faa2dbf0 | 2193 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2194 | { |
2195 | int ret; | |
2196 | struct btrfs_root *gang[8]; | |
2197 | int i; | |
2198 | ||
2199 | while (!list_empty(&fs_info->dead_roots)) { | |
2200 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2201 | struct btrfs_root, root_list); | |
2202 | list_del(&gang[0]->root_list); | |
2203 | ||
8c38938c | 2204 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
cb517eab | 2205 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
dc9492c1 | 2206 | btrfs_put_root(gang[0]); |
171f6537 JB |
2207 | } |
2208 | ||
2209 | while (1) { | |
2210 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2211 | (void **)gang, 0, | |
2212 | ARRAY_SIZE(gang)); | |
2213 | if (!ret) | |
2214 | break; | |
2215 | for (i = 0; i < ret; i++) | |
cb517eab | 2216 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 JB |
2217 | } |
2218 | } | |
af31f5e5 | 2219 | |
638aa7ed ES |
2220 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2221 | { | |
2222 | mutex_init(&fs_info->scrub_lock); | |
2223 | atomic_set(&fs_info->scrubs_running, 0); | |
2224 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2225 | atomic_set(&fs_info->scrubs_paused, 0); | |
2226 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2227 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2228 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2229 | } |
2230 | ||
779a65a4 ES |
2231 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2232 | { | |
2233 | spin_lock_init(&fs_info->balance_lock); | |
2234 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2235 | atomic_set(&fs_info->balance_pause_req, 0); |
2236 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2237 | fs_info->balance_ctl = NULL; | |
2238 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2239 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2240 | } |
2241 | ||
6bccf3ab | 2242 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2243 | { |
2ff7e61e JM |
2244 | struct inode *inode = fs_info->btree_inode; |
2245 | ||
2246 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2247 | set_nlink(inode, 1); | |
f37938e0 ES |
2248 | /* |
2249 | * we set the i_size on the btree inode to the max possible int. | |
2250 | * the real end of the address space is determined by all of | |
2251 | * the devices in the system | |
2252 | */ | |
2ff7e61e JM |
2253 | inode->i_size = OFFSET_MAX; |
2254 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2255 | |
2ff7e61e | 2256 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2257 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2c53a14d | 2258 | IO_TREE_BTREE_INODE_IO, inode); |
7b439738 | 2259 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2260 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2261 | |
5c8fd99f | 2262 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2263 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2264 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2265 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2266 | } |
2267 | ||
ad618368 ES |
2268 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2269 | { | |
ad618368 | 2270 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2271 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2272 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2273 | } |
2274 | ||
f9e92e40 ES |
2275 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2276 | { | |
2277 | spin_lock_init(&fs_info->qgroup_lock); | |
2278 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2279 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2280 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2281 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2282 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2283 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2284 | mutex_init(&fs_info->qgroup_rescan_lock); |
2285 | } | |
2286 | ||
2a458198 ES |
2287 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2288 | struct btrfs_fs_devices *fs_devices) | |
2289 | { | |
f7b885be | 2290 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2291 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2292 | |
2293 | fs_info->workers = | |
cb001095 JM |
2294 | btrfs_alloc_workqueue(fs_info, "worker", |
2295 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2296 | |
2297 | fs_info->delalloc_workers = | |
cb001095 JM |
2298 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2299 | flags, max_active, 2); | |
2a458198 ES |
2300 | |
2301 | fs_info->flush_workers = | |
cb001095 JM |
2302 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2303 | flags, max_active, 0); | |
2a458198 ES |
2304 | |
2305 | fs_info->caching_workers = | |
cb001095 | 2306 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2307 | |
2a458198 | 2308 | fs_info->fixup_workers = |
cb001095 | 2309 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2310 | |
2311 | /* | |
2312 | * endios are largely parallel and should have a very | |
2313 | * low idle thresh | |
2314 | */ | |
2315 | fs_info->endio_workers = | |
cb001095 | 2316 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2317 | fs_info->endio_meta_workers = |
cb001095 JM |
2318 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2319 | max_active, 4); | |
2a458198 | 2320 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2321 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2322 | max_active, 2); | |
2a458198 | 2323 | fs_info->endio_raid56_workers = |
cb001095 JM |
2324 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2325 | max_active, 4); | |
2a458198 | 2326 | fs_info->rmw_workers = |
cb001095 | 2327 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2328 | fs_info->endio_write_workers = |
cb001095 JM |
2329 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2330 | max_active, 2); | |
2a458198 | 2331 | fs_info->endio_freespace_worker = |
cb001095 JM |
2332 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2333 | max_active, 0); | |
2a458198 | 2334 | fs_info->delayed_workers = |
cb001095 JM |
2335 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2336 | max_active, 0); | |
2a458198 | 2337 | fs_info->readahead_workers = |
cb001095 JM |
2338 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2339 | max_active, 2); | |
2a458198 | 2340 | fs_info->qgroup_rescan_workers = |
cb001095 | 2341 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2342 | fs_info->discard_ctl.discard_workers = |
2343 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 ES |
2344 | |
2345 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
ba8a9d07 | 2346 | fs_info->flush_workers && |
2a458198 ES |
2347 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2348 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2349 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2350 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2351 | fs_info->caching_workers && fs_info->readahead_workers && | |
2352 | fs_info->fixup_workers && fs_info->delayed_workers && | |
b0643e59 DZ |
2353 | fs_info->qgroup_rescan_workers && |
2354 | fs_info->discard_ctl.discard_workers)) { | |
2a458198 ES |
2355 | return -ENOMEM; |
2356 | } | |
2357 | ||
2358 | return 0; | |
2359 | } | |
2360 | ||
6d97c6e3 JT |
2361 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2362 | { | |
2363 | struct crypto_shash *csum_shash; | |
b4e967be | 2364 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2365 | |
b4e967be | 2366 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2367 | |
2368 | if (IS_ERR(csum_shash)) { | |
2369 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2370 | csum_driver); |
6d97c6e3 JT |
2371 | return PTR_ERR(csum_shash); |
2372 | } | |
2373 | ||
2374 | fs_info->csum_shash = csum_shash; | |
2375 | ||
2376 | return 0; | |
2377 | } | |
2378 | ||
63443bf5 ES |
2379 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2380 | struct btrfs_fs_devices *fs_devices) | |
2381 | { | |
2382 | int ret; | |
63443bf5 ES |
2383 | struct btrfs_root *log_tree_root; |
2384 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2385 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2386 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2387 | |
2388 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2389 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2390 | return -EIO; |
2391 | } | |
2392 | ||
96dfcb46 JB |
2393 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2394 | GFP_KERNEL); | |
63443bf5 ES |
2395 | if (!log_tree_root) |
2396 | return -ENOMEM; | |
2397 | ||
2ff7e61e | 2398 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2399 | BTRFS_TREE_LOG_OBJECTID, |
2400 | fs_info->generation + 1, level, | |
2401 | NULL); | |
64c043de | 2402 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2403 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2404 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2405 | log_tree_root->node = NULL; |
00246528 | 2406 | btrfs_put_root(log_tree_root); |
0eeff236 | 2407 | return ret; |
64c043de | 2408 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2409 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2410 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2411 | return -EIO; |
2412 | } | |
2413 | /* returns with log_tree_root freed on success */ | |
2414 | ret = btrfs_recover_log_trees(log_tree_root); | |
2415 | if (ret) { | |
0b246afa JM |
2416 | btrfs_handle_fs_error(fs_info, ret, |
2417 | "Failed to recover log tree"); | |
00246528 | 2418 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2419 | return ret; |
2420 | } | |
2421 | ||
bc98a42c | 2422 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2423 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2424 | if (ret) |
2425 | return ret; | |
2426 | } | |
2427 | ||
2428 | return 0; | |
2429 | } | |
2430 | ||
6bccf3ab | 2431 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2432 | { |
6bccf3ab | 2433 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2434 | struct btrfs_root *root; |
4bbcaa64 ES |
2435 | struct btrfs_key location; |
2436 | int ret; | |
2437 | ||
6bccf3ab JM |
2438 | BUG_ON(!fs_info->tree_root); |
2439 | ||
4bbcaa64 ES |
2440 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2441 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2442 | location.offset = 0; | |
2443 | ||
a4f3d2c4 | 2444 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2445 | if (IS_ERR(root)) { |
42437a63 JB |
2446 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2447 | ret = PTR_ERR(root); | |
2448 | goto out; | |
2449 | } | |
2450 | } else { | |
2451 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2452 | fs_info->extent_root = root; | |
f50f4353 | 2453 | } |
4bbcaa64 ES |
2454 | |
2455 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2456 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2457 | if (IS_ERR(root)) { |
42437a63 JB |
2458 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2459 | ret = PTR_ERR(root); | |
2460 | goto out; | |
2461 | } | |
2462 | } else { | |
2463 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2464 | fs_info->dev_root = root; | |
f50f4353 | 2465 | } |
820a49da JB |
2466 | /* Initialize fs_info for all devices in any case */ |
2467 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2468 | |
882dbe0c JB |
2469 | /* If IGNOREDATACSUMS is set don't bother reading the csum root. */ |
2470 | if (!btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2471 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
2472 | root = btrfs_read_tree_root(tree_root, &location); | |
2473 | if (IS_ERR(root)) { | |
2474 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2475 | ret = PTR_ERR(root); | |
2476 | goto out; | |
2477 | } | |
2478 | } else { | |
2479 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2480 | fs_info->csum_root = root; | |
42437a63 | 2481 | } |
f50f4353 | 2482 | } |
4bbcaa64 | 2483 | |
aeb935a4 QW |
2484 | /* |
2485 | * This tree can share blocks with some other fs tree during relocation | |
2486 | * and we need a proper setup by btrfs_get_fs_root | |
2487 | */ | |
56e9357a DS |
2488 | root = btrfs_get_fs_root(tree_root->fs_info, |
2489 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2490 | if (IS_ERR(root)) { |
42437a63 JB |
2491 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2492 | ret = PTR_ERR(root); | |
2493 | goto out; | |
2494 | } | |
2495 | } else { | |
2496 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2497 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2498 | } |
aeb935a4 | 2499 | |
4bbcaa64 | 2500 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2501 | root = btrfs_read_tree_root(tree_root, &location); |
2502 | if (!IS_ERR(root)) { | |
2503 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2504 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2505 | fs_info->quota_root = root; |
4bbcaa64 ES |
2506 | } |
2507 | ||
2508 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2509 | root = btrfs_read_tree_root(tree_root, &location); |
2510 | if (IS_ERR(root)) { | |
42437a63 JB |
2511 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2512 | ret = PTR_ERR(root); | |
2513 | if (ret != -ENOENT) | |
2514 | goto out; | |
2515 | } | |
4bbcaa64 | 2516 | } else { |
a4f3d2c4 DS |
2517 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2518 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2519 | } |
2520 | ||
70f6d82e OS |
2521 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2522 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2523 | root = btrfs_read_tree_root(tree_root, &location); | |
f50f4353 | 2524 | if (IS_ERR(root)) { |
42437a63 JB |
2525 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2526 | ret = PTR_ERR(root); | |
2527 | goto out; | |
2528 | } | |
2529 | } else { | |
2530 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2531 | fs_info->free_space_root = root; | |
f50f4353 | 2532 | } |
70f6d82e OS |
2533 | } |
2534 | ||
4bbcaa64 | 2535 | return 0; |
f50f4353 LB |
2536 | out: |
2537 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2538 | location.objectid, ret); | |
2539 | return ret; | |
4bbcaa64 ES |
2540 | } |
2541 | ||
069ec957 QW |
2542 | /* |
2543 | * Real super block validation | |
2544 | * NOTE: super csum type and incompat features will not be checked here. | |
2545 | * | |
2546 | * @sb: super block to check | |
2547 | * @mirror_num: the super block number to check its bytenr: | |
2548 | * 0 the primary (1st) sb | |
2549 | * 1, 2 2nd and 3rd backup copy | |
2550 | * -1 skip bytenr check | |
2551 | */ | |
2552 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2553 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2554 | { |
21a852b0 QW |
2555 | u64 nodesize = btrfs_super_nodesize(sb); |
2556 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2557 | int ret = 0; | |
2558 | ||
2559 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2560 | btrfs_err(fs_info, "no valid FS found"); | |
2561 | ret = -EINVAL; | |
2562 | } | |
2563 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2564 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2565 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2566 | ret = -EINVAL; | |
2567 | } | |
2568 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2569 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2570 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2571 | ret = -EINVAL; | |
2572 | } | |
2573 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2574 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2575 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2576 | ret = -EINVAL; | |
2577 | } | |
2578 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2579 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2580 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2581 | ret = -EINVAL; | |
2582 | } | |
2583 | ||
2584 | /* | |
2585 | * Check sectorsize and nodesize first, other check will need it. | |
2586 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2587 | */ | |
2588 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2589 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2590 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2591 | ret = -EINVAL; | |
2592 | } | |
0bb3eb3e QW |
2593 | |
2594 | /* | |
2595 | * For 4K page size, we only support 4K sector size. | |
2596 | * For 64K page size, we support read-write for 64K sector size, and | |
2597 | * read-only for 4K sector size. | |
2598 | */ | |
2599 | if ((PAGE_SIZE == SZ_4K && sectorsize != PAGE_SIZE) || | |
2600 | (PAGE_SIZE == SZ_64K && (sectorsize != SZ_4K && | |
2601 | sectorsize != SZ_64K))) { | |
21a852b0 | 2602 | btrfs_err(fs_info, |
0bb3eb3e | 2603 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2604 | sectorsize, PAGE_SIZE); |
2605 | ret = -EINVAL; | |
2606 | } | |
0bb3eb3e | 2607 | |
21a852b0 QW |
2608 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2609 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2610 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2611 | ret = -EINVAL; | |
2612 | } | |
2613 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2614 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2615 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2616 | ret = -EINVAL; | |
2617 | } | |
2618 | ||
2619 | /* Root alignment check */ | |
2620 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2621 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2622 | btrfs_super_root(sb)); | |
2623 | ret = -EINVAL; | |
2624 | } | |
2625 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2626 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2627 | btrfs_super_chunk_root(sb)); | |
2628 | ret = -EINVAL; | |
2629 | } | |
2630 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2631 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2632 | btrfs_super_log_root(sb)); | |
2633 | ret = -EINVAL; | |
2634 | } | |
2635 | ||
aefd7f70 NB |
2636 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2637 | BTRFS_FSID_SIZE)) { | |
2638 | btrfs_err(fs_info, | |
2639 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2640 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2641 | ret = -EINVAL; | |
2642 | } | |
2643 | ||
2644 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2645 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2646 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2647 | btrfs_err(fs_info, | |
2648 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2649 | fs_info->super_copy->metadata_uuid, | |
2650 | fs_info->fs_devices->metadata_uuid); | |
2651 | ret = -EINVAL; | |
2652 | } | |
2653 | ||
de37aa51 | 2654 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2655 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2656 | btrfs_err(fs_info, |
7239ff4b | 2657 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2658 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2659 | ret = -EINVAL; |
2660 | } | |
2661 | ||
2662 | /* | |
2663 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2664 | * done later | |
2665 | */ | |
2666 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2667 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2668 | btrfs_super_bytes_used(sb)); | |
2669 | ret = -EINVAL; | |
2670 | } | |
2671 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2672 | btrfs_err(fs_info, "invalid stripesize %u", | |
2673 | btrfs_super_stripesize(sb)); | |
2674 | ret = -EINVAL; | |
2675 | } | |
2676 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2677 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2678 | btrfs_super_num_devices(sb)); | |
2679 | if (btrfs_super_num_devices(sb) == 0) { | |
2680 | btrfs_err(fs_info, "number of devices is 0"); | |
2681 | ret = -EINVAL; | |
2682 | } | |
2683 | ||
069ec957 QW |
2684 | if (mirror_num >= 0 && |
2685 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2686 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2687 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2688 | ret = -EINVAL; | |
2689 | } | |
2690 | ||
2691 | /* | |
2692 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2693 | * and one chunk | |
2694 | */ | |
2695 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2696 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2697 | btrfs_super_sys_array_size(sb), | |
2698 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2699 | ret = -EINVAL; | |
2700 | } | |
2701 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2702 | + sizeof(struct btrfs_chunk)) { | |
2703 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2704 | btrfs_super_sys_array_size(sb), | |
2705 | sizeof(struct btrfs_disk_key) | |
2706 | + sizeof(struct btrfs_chunk)); | |
2707 | ret = -EINVAL; | |
2708 | } | |
2709 | ||
2710 | /* | |
2711 | * The generation is a global counter, we'll trust it more than the others | |
2712 | * but it's still possible that it's the one that's wrong. | |
2713 | */ | |
2714 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2715 | btrfs_warn(fs_info, | |
2716 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2717 | btrfs_super_generation(sb), | |
2718 | btrfs_super_chunk_root_generation(sb)); | |
2719 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2720 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2721 | btrfs_warn(fs_info, | |
2722 | "suspicious: generation < cache_generation: %llu < %llu", | |
2723 | btrfs_super_generation(sb), | |
2724 | btrfs_super_cache_generation(sb)); | |
2725 | ||
2726 | return ret; | |
2727 | } | |
2728 | ||
069ec957 QW |
2729 | /* |
2730 | * Validation of super block at mount time. | |
2731 | * Some checks already done early at mount time, like csum type and incompat | |
2732 | * flags will be skipped. | |
2733 | */ | |
2734 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2735 | { | |
2736 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2737 | } | |
2738 | ||
75cb857d QW |
2739 | /* |
2740 | * Validation of super block at write time. | |
2741 | * Some checks like bytenr check will be skipped as their values will be | |
2742 | * overwritten soon. | |
2743 | * Extra checks like csum type and incompat flags will be done here. | |
2744 | */ | |
2745 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2746 | struct btrfs_super_block *sb) | |
2747 | { | |
2748 | int ret; | |
2749 | ||
2750 | ret = validate_super(fs_info, sb, -1); | |
2751 | if (ret < 0) | |
2752 | goto out; | |
e7e16f48 | 2753 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2754 | ret = -EUCLEAN; |
2755 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2756 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2757 | goto out; | |
2758 | } | |
2759 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2760 | ret = -EUCLEAN; | |
2761 | btrfs_err(fs_info, | |
2762 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2763 | btrfs_super_incompat_flags(sb), | |
2764 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2765 | goto out; | |
2766 | } | |
2767 | out: | |
2768 | if (ret < 0) | |
2769 | btrfs_err(fs_info, | |
2770 | "super block corruption detected before writing it to disk"); | |
2771 | return ret; | |
2772 | } | |
2773 | ||
6ef108dd | 2774 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2775 | { |
6ef108dd | 2776 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2777 | struct btrfs_super_block *sb = fs_info->super_copy; |
2778 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2779 | bool handle_error = false; | |
2780 | int ret = 0; | |
2781 | int i; | |
2782 | ||
2783 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2784 | u64 generation; | |
2785 | int level; | |
2786 | ||
2787 | if (handle_error) { | |
2788 | if (!IS_ERR(tree_root->node)) | |
2789 | free_extent_buffer(tree_root->node); | |
2790 | tree_root->node = NULL; | |
2791 | ||
2792 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2793 | break; | |
2794 | ||
2795 | free_root_pointers(fs_info, 0); | |
2796 | ||
2797 | /* | |
2798 | * Don't use the log in recovery mode, it won't be | |
2799 | * valid | |
2800 | */ | |
2801 | btrfs_set_super_log_root(sb, 0); | |
2802 | ||
2803 | /* We can't trust the free space cache either */ | |
2804 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2805 | ||
2806 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2807 | backup_index = ret; |
b8522a1e NB |
2808 | if (ret < 0) |
2809 | return ret; | |
2810 | } | |
2811 | generation = btrfs_super_generation(sb); | |
2812 | level = btrfs_super_root_level(sb); | |
2813 | tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), | |
1b7ec85e | 2814 | BTRFS_ROOT_TREE_OBJECTID, |
b8522a1e | 2815 | generation, level, NULL); |
217f5004 | 2816 | if (IS_ERR(tree_root->node)) { |
b8522a1e | 2817 | handle_error = true; |
217f5004 NB |
2818 | ret = PTR_ERR(tree_root->node); |
2819 | tree_root->node = NULL; | |
2820 | btrfs_warn(fs_info, "couldn't read tree root"); | |
2821 | continue; | |
b8522a1e | 2822 | |
217f5004 NB |
2823 | } else if (!extent_buffer_uptodate(tree_root->node)) { |
2824 | handle_error = true; | |
2825 | ret = -EIO; | |
2826 | btrfs_warn(fs_info, "error while reading tree root"); | |
b8522a1e NB |
2827 | continue; |
2828 | } | |
2829 | ||
2830 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); | |
2831 | tree_root->commit_root = btrfs_root_node(tree_root); | |
2832 | btrfs_set_root_refs(&tree_root->root_item, 1); | |
2833 | ||
336a0d8d NB |
2834 | /* |
2835 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2836 | * hasn't been fully initialised and we are the only user | |
2837 | */ | |
453e4873 | 2838 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2839 | if (ret < 0) { |
b8522a1e NB |
2840 | handle_error = true; |
2841 | continue; | |
2842 | } | |
2843 | ||
6b8fad57 | 2844 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2845 | |
2846 | ret = btrfs_read_roots(fs_info); | |
2847 | if (ret < 0) { | |
2848 | handle_error = true; | |
2849 | continue; | |
2850 | } | |
2851 | ||
2852 | /* All successful */ | |
2853 | fs_info->generation = generation; | |
2854 | fs_info->last_trans_committed = generation; | |
6ef108dd NB |
2855 | |
2856 | /* Always begin writing backup roots after the one being used */ | |
2857 | if (backup_index < 0) { | |
2858 | fs_info->backup_root_index = 0; | |
2859 | } else { | |
2860 | fs_info->backup_root_index = backup_index + 1; | |
2861 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2862 | } | |
b8522a1e NB |
2863 | break; |
2864 | } | |
2865 | ||
2866 | return ret; | |
2867 | } | |
2868 | ||
8260edba | 2869 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2870 | { |
76dda93c | 2871 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2872 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2873 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2874 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2875 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2876 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2877 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2878 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2879 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2880 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2881 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2882 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2883 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2884 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2885 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2886 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2887 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
f29021b2 | 2888 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2889 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2890 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2891 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2892 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2893 | mutex_init(&fs_info->zoned_meta_io_lock); |
de98ced9 | 2894 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2895 | |
0b86a832 | 2896 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2897 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2898 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2899 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2900 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 2901 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
2902 | #ifdef CONFIG_BTRFS_DEBUG |
2903 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2904 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2905 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2906 | #endif |
c8bf1b67 | 2907 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2908 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2909 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2910 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2911 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2912 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2913 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2914 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2915 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2916 | BTRFS_BLOCK_RSV_DELREFS); | |
2917 | ||
771ed689 | 2918 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2919 | atomic_set(&fs_info->defrag_running, 0); |
2fefd558 | 2920 | atomic_set(&fs_info->reada_works_cnt, 0); |
034f784d | 2921 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2922 | atomic64_set(&fs_info->tree_mod_seq, 0); |
95ac567a | 2923 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2924 | fs_info->metadata_ratio = 0; |
4cb5300b | 2925 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2926 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2927 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2928 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2929 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2930 | /* readahead state */ |
d0164adc | 2931 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2932 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2933 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2934 | |
b34b086c CM |
2935 | fs_info->thread_pool_size = min_t(unsigned long, |
2936 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2937 | |
199c2a9c MX |
2938 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2939 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2940 | |
638aa7ed | 2941 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2942 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2943 | fs_info->check_integrity_print_mask = 0; | |
2944 | #endif | |
779a65a4 | 2945 | btrfs_init_balance(fs_info); |
57056740 | 2946 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2947 | |
0f9dd46c | 2948 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2949 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2950 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2951 | |
fe119a6e NB |
2952 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
2953 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
afcdd129 | 2954 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2955 | |
5a3f23d5 | 2956 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2957 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2958 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2959 | mutex_init(&fs_info->transaction_kthread_mutex); |
2960 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2961 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2962 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2963 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2964 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2965 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2966 | |
ad618368 | 2967 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2968 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2969 | btrfs_discard_init(fs_info); |
416ac51d | 2970 | |
fa9c0d79 CM |
2971 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2972 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2973 | ||
e6dcd2dc | 2974 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2975 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2976 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2977 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2978 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2979 | |
da17066c JM |
2980 | /* Usable values until the real ones are cached from the superblock */ |
2981 | fs_info->nodesize = 4096; | |
2982 | fs_info->sectorsize = 4096; | |
ab108d99 | 2983 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
2984 | fs_info->stripesize = 4096; |
2985 | ||
eede2bf3 OS |
2986 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2987 | fs_info->swapfile_pins = RB_ROOT; | |
2988 | ||
1cea5cf0 | 2989 | spin_lock_init(&fs_info->send_reloc_lock); |
9e967495 | 2990 | fs_info->send_in_progress = 0; |
18bb8bbf JT |
2991 | |
2992 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; | |
2993 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
2994 | } |
2995 | ||
2996 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2997 | { | |
2998 | int ret; | |
2999 | ||
3000 | fs_info->sb = sb; | |
3001 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3002 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3003 | |
5deb17e1 | 3004 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3005 | if (ret) |
c75e8394 | 3006 | return ret; |
ae18c37a JB |
3007 | |
3008 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3009 | if (ret) | |
c75e8394 | 3010 | return ret; |
ae18c37a JB |
3011 | |
3012 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3013 | (1 + ilog2(nr_cpu_ids)); | |
3014 | ||
3015 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3016 | if (ret) | |
c75e8394 | 3017 | return ret; |
ae18c37a JB |
3018 | |
3019 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3020 | GFP_KERNEL); | |
3021 | if (ret) | |
c75e8394 | 3022 | return ret; |
ae18c37a JB |
3023 | |
3024 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3025 | GFP_KERNEL); | |
c75e8394 JB |
3026 | if (!fs_info->delayed_root) |
3027 | return -ENOMEM; | |
ae18c37a JB |
3028 | btrfs_init_delayed_root(fs_info->delayed_root); |
3029 | ||
a0a1db70 FM |
3030 | if (sb_rdonly(sb)) |
3031 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3032 | ||
c75e8394 | 3033 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3034 | } |
3035 | ||
97f4dd09 NB |
3036 | static int btrfs_uuid_rescan_kthread(void *data) |
3037 | { | |
3038 | struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; | |
3039 | int ret; | |
3040 | ||
3041 | /* | |
3042 | * 1st step is to iterate through the existing UUID tree and | |
3043 | * to delete all entries that contain outdated data. | |
3044 | * 2nd step is to add all missing entries to the UUID tree. | |
3045 | */ | |
3046 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3047 | if (ret < 0) { | |
c94bec2c JB |
3048 | if (ret != -EINTR) |
3049 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3050 | ret); | |
97f4dd09 NB |
3051 | up(&fs_info->uuid_tree_rescan_sem); |
3052 | return ret; | |
3053 | } | |
3054 | return btrfs_uuid_scan_kthread(data); | |
3055 | } | |
3056 | ||
3057 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3058 | { | |
3059 | struct task_struct *task; | |
3060 | ||
3061 | down(&fs_info->uuid_tree_rescan_sem); | |
3062 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3063 | if (IS_ERR(task)) { | |
3064 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3065 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3066 | up(&fs_info->uuid_tree_rescan_sem); | |
3067 | return PTR_ERR(task); | |
3068 | } | |
3069 | ||
3070 | return 0; | |
3071 | } | |
3072 | ||
8cd29088 BB |
3073 | /* |
3074 | * Some options only have meaning at mount time and shouldn't persist across | |
3075 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3076 | * code paths. | |
3077 | */ | |
3078 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3079 | { | |
3080 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3081 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3082 | } |
3083 | ||
44c0ca21 BB |
3084 | /* |
3085 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3086 | * and btrfs_remount when remounting from read-only to read-write. | |
3087 | */ | |
3088 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3089 | { | |
3090 | int ret; | |
94846229 | 3091 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3092 | bool clear_free_space_tree = false; |
3093 | ||
3094 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3095 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3096 | clear_free_space_tree = true; | |
3097 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3098 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3099 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3100 | clear_free_space_tree = true; | |
3101 | } | |
3102 | ||
3103 | if (clear_free_space_tree) { | |
3104 | btrfs_info(fs_info, "clearing free space tree"); | |
3105 | ret = btrfs_clear_free_space_tree(fs_info); | |
3106 | if (ret) { | |
3107 | btrfs_warn(fs_info, | |
3108 | "failed to clear free space tree: %d", ret); | |
3109 | goto out; | |
3110 | } | |
3111 | } | |
44c0ca21 | 3112 | |
8d488a8c FM |
3113 | /* |
3114 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3115 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
3116 | * them into the fs_info->fs_roots_radix tree. This must be done before | |
3117 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it | |
3118 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3119 | * item before the root's tree is deleted - this means that if we unmount | |
3120 | * or crash before the deletion completes, on the next mount we will not | |
3121 | * delete what remains of the tree because the orphan item does not | |
3122 | * exists anymore, which is what tells us we have a pending deletion. | |
3123 | */ | |
3124 | ret = btrfs_find_orphan_roots(fs_info); | |
3125 | if (ret) | |
3126 | goto out; | |
3127 | ||
44c0ca21 BB |
3128 | ret = btrfs_cleanup_fs_roots(fs_info); |
3129 | if (ret) | |
3130 | goto out; | |
3131 | ||
8f1c21d7 BB |
3132 | down_read(&fs_info->cleanup_work_sem); |
3133 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3134 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3135 | up_read(&fs_info->cleanup_work_sem); | |
3136 | goto out; | |
3137 | } | |
3138 | up_read(&fs_info->cleanup_work_sem); | |
3139 | ||
44c0ca21 BB |
3140 | mutex_lock(&fs_info->cleaner_mutex); |
3141 | ret = btrfs_recover_relocation(fs_info->tree_root); | |
3142 | mutex_unlock(&fs_info->cleaner_mutex); | |
3143 | if (ret < 0) { | |
3144 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3145 | goto out; | |
3146 | } | |
3147 | ||
5011139a BB |
3148 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3149 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3150 | btrfs_info(fs_info, "creating free space tree"); | |
3151 | ret = btrfs_create_free_space_tree(fs_info); | |
3152 | if (ret) { | |
3153 | btrfs_warn(fs_info, | |
3154 | "failed to create free space tree: %d", ret); | |
3155 | goto out; | |
3156 | } | |
3157 | } | |
3158 | ||
94846229 BB |
3159 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3160 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3161 | if (ret) | |
3162 | goto out; | |
3163 | } | |
3164 | ||
44c0ca21 BB |
3165 | ret = btrfs_resume_balance_async(fs_info); |
3166 | if (ret) | |
3167 | goto out; | |
3168 | ||
3169 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3170 | if (ret) { | |
3171 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3172 | goto out; | |
3173 | } | |
3174 | ||
3175 | btrfs_qgroup_rescan_resume(fs_info); | |
3176 | ||
3177 | if (!fs_info->uuid_root) { | |
3178 | btrfs_info(fs_info, "creating UUID tree"); | |
3179 | ret = btrfs_create_uuid_tree(fs_info); | |
3180 | if (ret) { | |
3181 | btrfs_warn(fs_info, | |
3182 | "failed to create the UUID tree %d", ret); | |
3183 | goto out; | |
3184 | } | |
3185 | } | |
3186 | ||
3187 | out: | |
3188 | return ret; | |
3189 | } | |
3190 | ||
ae18c37a JB |
3191 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3192 | char *options) | |
3193 | { | |
3194 | u32 sectorsize; | |
3195 | u32 nodesize; | |
3196 | u32 stripesize; | |
3197 | u64 generation; | |
3198 | u64 features; | |
3199 | u16 csum_type; | |
ae18c37a JB |
3200 | struct btrfs_super_block *disk_super; |
3201 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3202 | struct btrfs_root *tree_root; | |
3203 | struct btrfs_root *chunk_root; | |
3204 | int ret; | |
3205 | int err = -EINVAL; | |
ae18c37a JB |
3206 | int level; |
3207 | ||
8260edba | 3208 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3209 | if (ret) { |
83c8266a | 3210 | err = ret; |
ae18c37a | 3211 | goto fail; |
53b381b3 DW |
3212 | } |
3213 | ||
ae18c37a JB |
3214 | /* These need to be init'ed before we start creating inodes and such. */ |
3215 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3216 | GFP_KERNEL); | |
3217 | fs_info->tree_root = tree_root; | |
3218 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3219 | GFP_KERNEL); | |
3220 | fs_info->chunk_root = chunk_root; | |
3221 | if (!tree_root || !chunk_root) { | |
3222 | err = -ENOMEM; | |
c75e8394 | 3223 | goto fail; |
ae18c37a JB |
3224 | } |
3225 | ||
3226 | fs_info->btree_inode = new_inode(sb); | |
3227 | if (!fs_info->btree_inode) { | |
3228 | err = -ENOMEM; | |
c75e8394 | 3229 | goto fail; |
ae18c37a JB |
3230 | } |
3231 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3232 | btrfs_init_btree_inode(fs_info); | |
3233 | ||
d24fa5c1 | 3234 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3235 | |
3236 | /* | |
3237 | * Read super block and check the signature bytes only | |
3238 | */ | |
d24fa5c1 | 3239 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3240 | if (IS_ERR(disk_super)) { |
3241 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3242 | goto fail_alloc; |
20b45077 | 3243 | } |
39279cc3 | 3244 | |
8dc3f22c | 3245 | /* |
260db43c | 3246 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3247 | * corrupted, we'll find out |
3248 | */ | |
8f32380d | 3249 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3250 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3251 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3252 | csum_type); |
8dc3f22c | 3253 | err = -EINVAL; |
8f32380d | 3254 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3255 | goto fail_alloc; |
3256 | } | |
3257 | ||
83c68bbc SY |
3258 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3259 | ||
6d97c6e3 JT |
3260 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3261 | if (ret) { | |
3262 | err = ret; | |
8f32380d | 3263 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3264 | goto fail_alloc; |
3265 | } | |
3266 | ||
1104a885 DS |
3267 | /* |
3268 | * We want to check superblock checksum, the type is stored inside. | |
3269 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3270 | */ | |
8f32380d | 3271 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3272 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3273 | err = -EINVAL; |
8f32380d | 3274 | btrfs_release_disk_super(disk_super); |
141386e1 | 3275 | goto fail_alloc; |
1104a885 DS |
3276 | } |
3277 | ||
3278 | /* | |
3279 | * super_copy is zeroed at allocation time and we never touch the | |
3280 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3281 | * the whole block of INFO_SIZE | |
3282 | */ | |
8f32380d JT |
3283 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3284 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3285 | |
fbc6feae NB |
3286 | disk_super = fs_info->super_copy; |
3287 | ||
0b86a832 | 3288 | |
fbc6feae NB |
3289 | features = btrfs_super_flags(disk_super); |
3290 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3291 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3292 | btrfs_set_super_flags(disk_super, features); | |
3293 | btrfs_info(fs_info, | |
3294 | "found metadata UUID change in progress flag, clearing"); | |
3295 | } | |
3296 | ||
3297 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3298 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3299 | |
069ec957 | 3300 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3301 | if (ret) { |
05135f59 | 3302 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3303 | err = -EINVAL; |
141386e1 | 3304 | goto fail_alloc; |
1104a885 DS |
3305 | } |
3306 | ||
0f7d52f4 | 3307 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3308 | goto fail_alloc; |
0f7d52f4 | 3309 | |
acce952b | 3310 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3311 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3312 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3313 | |
75e7cb7f LB |
3314 | /* |
3315 | * In the long term, we'll store the compression type in the super | |
3316 | * block, and it'll be used for per file compression control. | |
3317 | */ | |
3318 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3319 | ||
6f93e834 AJ |
3320 | /* |
3321 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3322 | * than the page size. | |
3323 | */ | |
3324 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { | |
3325 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
3326 | btrfs_info(fs_info, | |
3327 | "flagging fs with big metadata feature"); | |
3328 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3329 | } | |
3330 | ||
3331 | /* Set up fs_info before parsing mount options */ | |
3332 | nodesize = btrfs_super_nodesize(disk_super); | |
3333 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3334 | stripesize = sectorsize; | |
3335 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3336 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3337 | ||
3338 | fs_info->nodesize = nodesize; | |
3339 | fs_info->sectorsize = sectorsize; | |
3340 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3341 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3342 | fs_info->stripesize = stripesize; | |
3343 | ||
2ff7e61e | 3344 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3345 | if (ret) { |
3346 | err = ret; | |
141386e1 | 3347 | goto fail_alloc; |
2b82032c | 3348 | } |
dfe25020 | 3349 | |
f2b636e8 JB |
3350 | features = btrfs_super_incompat_flags(disk_super) & |
3351 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3352 | if (features) { | |
05135f59 DS |
3353 | btrfs_err(fs_info, |
3354 | "cannot mount because of unsupported optional features (%llx)", | |
3355 | features); | |
f2b636e8 | 3356 | err = -EINVAL; |
141386e1 | 3357 | goto fail_alloc; |
f2b636e8 JB |
3358 | } |
3359 | ||
5d4f98a2 | 3360 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3361 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3362 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3363 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3364 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3365 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3366 | |
3173a18f | 3367 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3368 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3369 | |
bc3f116f CM |
3370 | /* |
3371 | * mixed block groups end up with duplicate but slightly offset | |
3372 | * extent buffers for the same range. It leads to corruptions | |
3373 | */ | |
3374 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3375 | (sectorsize != nodesize)) { |
05135f59 DS |
3376 | btrfs_err(fs_info, |
3377 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3378 | nodesize, sectorsize); | |
141386e1 | 3379 | goto fail_alloc; |
bc3f116f CM |
3380 | } |
3381 | ||
ceda0864 MX |
3382 | /* |
3383 | * Needn't use the lock because there is no other task which will | |
3384 | * update the flag. | |
3385 | */ | |
a6fa6fae | 3386 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3387 | |
f2b636e8 JB |
3388 | features = btrfs_super_compat_ro_flags(disk_super) & |
3389 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3390 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
3391 | btrfs_err(fs_info, |
3392 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 3393 | features); |
f2b636e8 | 3394 | err = -EINVAL; |
141386e1 | 3395 | goto fail_alloc; |
f2b636e8 | 3396 | } |
61d92c32 | 3397 | |
8481dd80 QW |
3398 | if (sectorsize < PAGE_SIZE) { |
3399 | struct btrfs_subpage_info *subpage_info; | |
3400 | ||
95ea0486 QW |
3401 | btrfs_warn(fs_info, |
3402 | "read-write for sector size %u with page size %lu is experimental", | |
3403 | sectorsize, PAGE_SIZE); | |
c8050b3b QW |
3404 | if (btrfs_super_incompat_flags(fs_info->super_copy) & |
3405 | BTRFS_FEATURE_INCOMPAT_RAID56) { | |
3406 | btrfs_err(fs_info, | |
3407 | "RAID56 is not yet supported for sector size %u with page size %lu", | |
3408 | sectorsize, PAGE_SIZE); | |
3409 | err = -EINVAL; | |
3410 | goto fail_alloc; | |
3411 | } | |
8481dd80 QW |
3412 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3413 | if (!subpage_info) | |
3414 | goto fail_alloc; | |
3415 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3416 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3417 | } |
0bb3eb3e | 3418 | |
2a458198 ES |
3419 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
3420 | if (ret) { | |
3421 | err = ret; | |
0dc3b84a JB |
3422 | goto fail_sb_buffer; |
3423 | } | |
4543df7e | 3424 | |
9e11ceee JK |
3425 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3426 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3427 | |
a061fc8d CM |
3428 | sb->s_blocksize = sectorsize; |
3429 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3430 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3431 | |
925baedd | 3432 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3433 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3434 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3435 | if (ret) { |
05135f59 | 3436 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3437 | goto fail_sb_buffer; |
84eed90f | 3438 | } |
0b86a832 | 3439 | |
84234f3a | 3440 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3441 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3442 | |
2ff7e61e | 3443 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3444 | btrfs_super_chunk_root(disk_super), |
1b7ec85e | 3445 | BTRFS_CHUNK_TREE_OBJECTID, |
581c1760 | 3446 | generation, level, NULL); |
64c043de LB |
3447 | if (IS_ERR(chunk_root->node) || |
3448 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3449 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3450 | if (!IS_ERR(chunk_root->node)) |
3451 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3452 | chunk_root->node = NULL; |
af31f5e5 | 3453 | goto fail_tree_roots; |
83121942 | 3454 | } |
5d4f98a2 YZ |
3455 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3456 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3457 | |
e17cade2 | 3458 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3459 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3460 | BTRFS_UUID_SIZE); | |
e17cade2 | 3461 | |
5b4aacef | 3462 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3463 | if (ret) { |
05135f59 | 3464 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3465 | goto fail_tree_roots; |
2b82032c | 3466 | } |
0b86a832 | 3467 | |
8dabb742 | 3468 | /* |
bacce86a AJ |
3469 | * At this point we know all the devices that make this filesystem, |
3470 | * including the seed devices but we don't know yet if the replace | |
3471 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3472 | * filesystem but skip the replace target device which is checked |
bacce86a | 3473 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3474 | */ |
bacce86a | 3475 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3476 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3477 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3478 | goto fail_tree_roots; |
3479 | } | |
3480 | ||
b8522a1e | 3481 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3482 | if (ret) |
b8522a1e | 3483 | goto fail_tree_roots; |
8929ecfa | 3484 | |
73651042 NA |
3485 | /* |
3486 | * Get zone type information of zoned block devices. This will also | |
3487 | * handle emulation of a zoned filesystem if a regular device has the | |
3488 | * zoned incompat feature flag set. | |
3489 | */ | |
3490 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3491 | if (ret) { | |
3492 | btrfs_err(fs_info, | |
3493 | "zoned: failed to read device zone info: %d", | |
3494 | ret); | |
3495 | goto fail_block_groups; | |
3496 | } | |
3497 | ||
75ec1db8 JB |
3498 | /* |
3499 | * If we have a uuid root and we're not being told to rescan we need to | |
3500 | * check the generation here so we can set the | |
3501 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3502 | * transaction during a balance or the log replay without updating the | |
3503 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3504 | * even though it was perfectly fine. | |
3505 | */ | |
3506 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3507 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3508 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3509 | ||
cf90d884 QW |
3510 | ret = btrfs_verify_dev_extents(fs_info); |
3511 | if (ret) { | |
3512 | btrfs_err(fs_info, | |
3513 | "failed to verify dev extents against chunks: %d", | |
3514 | ret); | |
3515 | goto fail_block_groups; | |
3516 | } | |
68310a5e ID |
3517 | ret = btrfs_recover_balance(fs_info); |
3518 | if (ret) { | |
05135f59 | 3519 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3520 | goto fail_block_groups; |
3521 | } | |
3522 | ||
733f4fbb SB |
3523 | ret = btrfs_init_dev_stats(fs_info); |
3524 | if (ret) { | |
05135f59 | 3525 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3526 | goto fail_block_groups; |
3527 | } | |
3528 | ||
8dabb742 SB |
3529 | ret = btrfs_init_dev_replace(fs_info); |
3530 | if (ret) { | |
05135f59 | 3531 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3532 | goto fail_block_groups; |
3533 | } | |
3534 | ||
b70f5097 NA |
3535 | ret = btrfs_check_zoned_mode(fs_info); |
3536 | if (ret) { | |
3537 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3538 | ret); | |
3539 | goto fail_block_groups; | |
3540 | } | |
3541 | ||
c6761a9e | 3542 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3543 | if (ret) { |
05135f59 DS |
3544 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3545 | ret); | |
b7c35e81 AJ |
3546 | goto fail_block_groups; |
3547 | } | |
3548 | ||
96f3136e | 3549 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3550 | if (ret) { |
05135f59 | 3551 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3552 | goto fail_fsdev_sysfs; |
c59021f8 | 3553 | } |
3554 | ||
c59021f8 | 3555 | ret = btrfs_init_space_info(fs_info); |
3556 | if (ret) { | |
05135f59 | 3557 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3558 | goto fail_sysfs; |
c59021f8 | 3559 | } |
3560 | ||
5b4aacef | 3561 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3562 | if (ret) { |
05135f59 | 3563 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3564 | goto fail_sysfs; |
1b1d1f66 | 3565 | } |
4330e183 | 3566 | |
6528b99d | 3567 | if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { |
05135f59 | 3568 | btrfs_warn(fs_info, |
52042d8e | 3569 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3570 | goto fail_sysfs; |
292fd7fc | 3571 | } |
9078a3e1 | 3572 | |
a74a4b97 CM |
3573 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3574 | "btrfs-cleaner"); | |
57506d50 | 3575 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3576 | goto fail_sysfs; |
a74a4b97 CM |
3577 | |
3578 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3579 | tree_root, | |
3580 | "btrfs-transaction"); | |
57506d50 | 3581 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3582 | goto fail_cleaner; |
a74a4b97 | 3583 | |
583b7231 | 3584 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3585 | !fs_info->fs_devices->rotating) { |
583b7231 | 3586 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3587 | } |
3588 | ||
572d9ab7 | 3589 | /* |
01327610 | 3590 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3591 | * not have to wait for transaction commit |
3592 | */ | |
3593 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3594 | |
21adbd5c | 3595 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3596 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3597 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3598 | btrfs_test_opt(fs_info, |
cbeaae4f | 3599 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3600 | fs_info->check_integrity_print_mask); |
3601 | if (ret) | |
05135f59 DS |
3602 | btrfs_warn(fs_info, |
3603 | "failed to initialize integrity check module: %d", | |
3604 | ret); | |
21adbd5c SB |
3605 | } |
3606 | #endif | |
bcef60f2 AJ |
3607 | ret = btrfs_read_qgroup_config(fs_info); |
3608 | if (ret) | |
3609 | goto fail_trans_kthread; | |
21adbd5c | 3610 | |
fd708b81 JB |
3611 | if (btrfs_build_ref_tree(fs_info)) |
3612 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3613 | ||
96da0919 QW |
3614 | /* do not make disk changes in broken FS or nologreplay is given */ |
3615 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3616 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3617 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3618 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3619 | if (ret) { |
63443bf5 | 3620 | err = ret; |
28c16cbb | 3621 | goto fail_qgroup; |
79787eaa | 3622 | } |
e02119d5 | 3623 | } |
1a40e23b | 3624 | |
56e9357a | 3625 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3626 | if (IS_ERR(fs_info->fs_root)) { |
3627 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3628 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3629 | fs_info->fs_root = NULL; |
bcef60f2 | 3630 | goto fail_qgroup; |
3140c9a3 | 3631 | } |
c289811c | 3632 | |
bc98a42c | 3633 | if (sb_rdonly(sb)) |
8cd29088 | 3634 | goto clear_oneshot; |
59641015 | 3635 | |
44c0ca21 | 3636 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3637 | if (ret) { |
6bccf3ab | 3638 | close_ctree(fs_info); |
2b6ba629 | 3639 | return ret; |
e3acc2a6 | 3640 | } |
b0643e59 | 3641 | btrfs_discard_resume(fs_info); |
b382a324 | 3642 | |
44c0ca21 BB |
3643 | if (fs_info->uuid_root && |
3644 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3645 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3646 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3647 | ret = btrfs_check_uuid_tree(fs_info); |
3648 | if (ret) { | |
05135f59 DS |
3649 | btrfs_warn(fs_info, |
3650 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3651 | close_ctree(fs_info); |
70f80175 SB |
3652 | return ret; |
3653 | } | |
f7a81ea4 | 3654 | } |
94846229 | 3655 | |
afcdd129 | 3656 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3657 | |
8cd29088 BB |
3658 | clear_oneshot: |
3659 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3660 | return 0; |
39279cc3 | 3661 | |
bcef60f2 AJ |
3662 | fail_qgroup: |
3663 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3664 | fail_trans_kthread: |
3665 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3666 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3667 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3668 | fail_cleaner: |
a74a4b97 | 3669 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3670 | |
3671 | /* | |
3672 | * make sure we're done with the btree inode before we stop our | |
3673 | * kthreads | |
3674 | */ | |
3675 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3676 | |
2365dd3c | 3677 | fail_sysfs: |
6618a59b | 3678 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3679 | |
b7c35e81 AJ |
3680 | fail_fsdev_sysfs: |
3681 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3682 | ||
1b1d1f66 | 3683 | fail_block_groups: |
54067ae9 | 3684 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3685 | |
3686 | fail_tree_roots: | |
9e3aa805 JB |
3687 | if (fs_info->data_reloc_root) |
3688 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3689 | free_root_pointers(fs_info, true); |
2b8195bb | 3690 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3691 | |
39279cc3 | 3692 | fail_sb_buffer: |
7abadb64 | 3693 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3694 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3695 | fail_alloc: |
586e46e2 ID |
3696 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3697 | ||
4543df7e | 3698 | iput(fs_info->btree_inode); |
7e662854 | 3699 | fail: |
586e46e2 | 3700 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3701 | return err; |
eb60ceac | 3702 | } |
663faf9f | 3703 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3704 | |
314b6dd0 | 3705 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3706 | { |
314b6dd0 JT |
3707 | struct btrfs_device *device = bio->bi_private; |
3708 | struct bio_vec *bvec; | |
3709 | struct bvec_iter_all iter_all; | |
3710 | struct page *page; | |
3711 | ||
3712 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3713 | page = bvec->bv_page; | |
3714 | ||
3715 | if (bio->bi_status) { | |
3716 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3717 | "lost page write due to IO error on %s (%d)", | |
3718 | rcu_str_deref(device->name), | |
3719 | blk_status_to_errno(bio->bi_status)); | |
3720 | ClearPageUptodate(page); | |
3721 | SetPageError(page); | |
3722 | btrfs_dev_stat_inc_and_print(device, | |
3723 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3724 | } else { | |
3725 | SetPageUptodate(page); | |
3726 | } | |
3727 | ||
3728 | put_page(page); | |
3729 | unlock_page(page); | |
f2984462 | 3730 | } |
314b6dd0 JT |
3731 | |
3732 | bio_put(bio); | |
f2984462 CM |
3733 | } |
3734 | ||
8f32380d JT |
3735 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
3736 | int copy_num) | |
29c36d72 | 3737 | { |
29c36d72 | 3738 | struct btrfs_super_block *super; |
8f32380d | 3739 | struct page *page; |
12659251 | 3740 | u64 bytenr, bytenr_orig; |
8f32380d | 3741 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3742 | int ret; |
3743 | ||
3744 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3745 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3746 | if (ret == -ENOENT) | |
3747 | return ERR_PTR(-EINVAL); | |
3748 | else if (ret) | |
3749 | return ERR_PTR(ret); | |
29c36d72 | 3750 | |
29c36d72 | 3751 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) |
8f32380d | 3752 | return ERR_PTR(-EINVAL); |
29c36d72 | 3753 | |
8f32380d JT |
3754 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3755 | if (IS_ERR(page)) | |
3756 | return ERR_CAST(page); | |
29c36d72 | 3757 | |
8f32380d | 3758 | super = page_address(page); |
96c2e067 AJ |
3759 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3760 | btrfs_release_disk_super(super); | |
3761 | return ERR_PTR(-ENODATA); | |
3762 | } | |
3763 | ||
12659251 | 3764 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3765 | btrfs_release_disk_super(super); |
3766 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3767 | } |
3768 | ||
8f32380d | 3769 | return super; |
29c36d72 AJ |
3770 | } |
3771 | ||
3772 | ||
8f32380d | 3773 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3774 | { |
8f32380d | 3775 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3776 | int i; |
3777 | u64 transid = 0; | |
a512bbf8 YZ |
3778 | |
3779 | /* we would like to check all the supers, but that would make | |
3780 | * a btrfs mount succeed after a mkfs from a different FS. | |
3781 | * So, we need to add a special mount option to scan for | |
3782 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3783 | */ | |
3784 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
3785 | super = btrfs_read_dev_one_super(bdev, i); |
3786 | if (IS_ERR(super)) | |
a512bbf8 YZ |
3787 | continue; |
3788 | ||
a512bbf8 | 3789 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3790 | if (latest) |
3791 | btrfs_release_disk_super(super); | |
3792 | ||
3793 | latest = super; | |
a512bbf8 | 3794 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3795 | } |
3796 | } | |
92fc03fb | 3797 | |
8f32380d | 3798 | return super; |
a512bbf8 YZ |
3799 | } |
3800 | ||
4eedeb75 | 3801 | /* |
abbb3b8e | 3802 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3803 | * pages we use for writing are locked. |
4eedeb75 | 3804 | * |
abbb3b8e DS |
3805 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3806 | * the expected device size at commit time. Note that max_mirrors must be | |
3807 | * same for write and wait phases. | |
4eedeb75 | 3808 | * |
314b6dd0 | 3809 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3810 | */ |
a512bbf8 | 3811 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3812 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3813 | { |
d5178578 | 3814 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3815 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3816 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3817 | int i; |
a512bbf8 | 3818 | int errors = 0; |
12659251 NA |
3819 | int ret; |
3820 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3821 | |
3822 | if (max_mirrors == 0) | |
3823 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3824 | ||
d5178578 JT |
3825 | shash->tfm = fs_info->csum_shash; |
3826 | ||
a512bbf8 | 3827 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3828 | struct page *page; |
3829 | struct bio *bio; | |
3830 | struct btrfs_super_block *disk_super; | |
3831 | ||
12659251 NA |
3832 | bytenr_orig = btrfs_sb_offset(i); |
3833 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3834 | if (ret == -ENOENT) { | |
3835 | continue; | |
3836 | } else if (ret < 0) { | |
3837 | btrfs_err(device->fs_info, | |
3838 | "couldn't get super block location for mirror %d", | |
3839 | i); | |
3840 | errors++; | |
3841 | continue; | |
3842 | } | |
935e5cc9 MX |
3843 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3844 | device->commit_total_bytes) | |
a512bbf8 YZ |
3845 | break; |
3846 | ||
12659251 | 3847 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3848 | |
fd08001f EB |
3849 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3850 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3851 | sb->csum); | |
4eedeb75 | 3852 | |
314b6dd0 JT |
3853 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3854 | GFP_NOFS); | |
3855 | if (!page) { | |
abbb3b8e | 3856 | btrfs_err(device->fs_info, |
314b6dd0 | 3857 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3858 | bytenr); |
3859 | errors++; | |
4eedeb75 | 3860 | continue; |
abbb3b8e | 3861 | } |
634554dc | 3862 | |
314b6dd0 JT |
3863 | /* Bump the refcount for wait_dev_supers() */ |
3864 | get_page(page); | |
a512bbf8 | 3865 | |
314b6dd0 JT |
3866 | disk_super = page_address(page); |
3867 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3868 | |
314b6dd0 JT |
3869 | /* |
3870 | * Directly use bios here instead of relying on the page cache | |
3871 | * to do I/O, so we don't lose the ability to do integrity | |
3872 | * checking. | |
3873 | */ | |
3874 | bio = bio_alloc(GFP_NOFS, 1); | |
3875 | bio_set_dev(bio, device->bdev); | |
3876 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; | |
3877 | bio->bi_private = device; | |
3878 | bio->bi_end_io = btrfs_end_super_write; | |
3879 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3880 | offset_in_page(bytenr)); | |
a512bbf8 | 3881 | |
387125fc | 3882 | /* |
314b6dd0 JT |
3883 | * We FUA only the first super block. The others we allow to |
3884 | * go down lazy and there's a short window where the on-disk | |
3885 | * copies might still contain the older version. | |
387125fc | 3886 | */ |
314b6dd0 | 3887 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; |
1b9e619c | 3888 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
3889 | bio->bi_opf |= REQ_FUA; |
3890 | ||
3891 | btrfsic_submit_bio(bio); | |
8376d9e1 NA |
3892 | |
3893 | if (btrfs_advance_sb_log(device, i)) | |
3894 | errors++; | |
a512bbf8 YZ |
3895 | } |
3896 | return errors < i ? 0 : -1; | |
3897 | } | |
3898 | ||
abbb3b8e DS |
3899 | /* |
3900 | * Wait for write completion of superblocks done by write_dev_supers, | |
3901 | * @max_mirrors same for write and wait phases. | |
3902 | * | |
314b6dd0 | 3903 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3904 | * date. |
3905 | */ | |
3906 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3907 | { | |
abbb3b8e DS |
3908 | int i; |
3909 | int errors = 0; | |
b6a535fa | 3910 | bool primary_failed = false; |
12659251 | 3911 | int ret; |
abbb3b8e DS |
3912 | u64 bytenr; |
3913 | ||
3914 | if (max_mirrors == 0) | |
3915 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3916 | ||
3917 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3918 | struct page *page; |
3919 | ||
12659251 NA |
3920 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3921 | if (ret == -ENOENT) { | |
3922 | break; | |
3923 | } else if (ret < 0) { | |
3924 | errors++; | |
3925 | if (i == 0) | |
3926 | primary_failed = true; | |
3927 | continue; | |
3928 | } | |
abbb3b8e DS |
3929 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3930 | device->commit_total_bytes) | |
3931 | break; | |
3932 | ||
314b6dd0 JT |
3933 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3934 | bytenr >> PAGE_SHIFT); | |
3935 | if (!page) { | |
abbb3b8e | 3936 | errors++; |
b6a535fa HM |
3937 | if (i == 0) |
3938 | primary_failed = true; | |
abbb3b8e DS |
3939 | continue; |
3940 | } | |
314b6dd0 JT |
3941 | /* Page is submitted locked and unlocked once the IO completes */ |
3942 | wait_on_page_locked(page); | |
3943 | if (PageError(page)) { | |
abbb3b8e | 3944 | errors++; |
b6a535fa HM |
3945 | if (i == 0) |
3946 | primary_failed = true; | |
3947 | } | |
abbb3b8e | 3948 | |
314b6dd0 JT |
3949 | /* Drop our reference */ |
3950 | put_page(page); | |
abbb3b8e | 3951 | |
314b6dd0 JT |
3952 | /* Drop the reference from the writing run */ |
3953 | put_page(page); | |
abbb3b8e DS |
3954 | } |
3955 | ||
b6a535fa HM |
3956 | /* log error, force error return */ |
3957 | if (primary_failed) { | |
3958 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3959 | device->devid); | |
3960 | return -1; | |
3961 | } | |
3962 | ||
abbb3b8e DS |
3963 | return errors < i ? 0 : -1; |
3964 | } | |
3965 | ||
387125fc CM |
3966 | /* |
3967 | * endio for the write_dev_flush, this will wake anyone waiting | |
3968 | * for the barrier when it is done | |
3969 | */ | |
4246a0b6 | 3970 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3971 | { |
e0ae9994 | 3972 | complete(bio->bi_private); |
387125fc CM |
3973 | } |
3974 | ||
3975 | /* | |
4fc6441a AJ |
3976 | * Submit a flush request to the device if it supports it. Error handling is |
3977 | * done in the waiting counterpart. | |
387125fc | 3978 | */ |
4fc6441a | 3979 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3980 | { |
c2a9c7ab | 3981 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3982 | struct bio *bio = device->flush_bio; |
387125fc | 3983 | |
c2a9c7ab | 3984 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3985 | return; |
387125fc | 3986 | |
e0ae9994 | 3987 | bio_reset(bio); |
387125fc | 3988 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 3989 | bio_set_dev(bio, device->bdev); |
8d910125 | 3990 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3991 | init_completion(&device->flush_wait); |
3992 | bio->bi_private = &device->flush_wait; | |
387125fc | 3993 | |
43a01111 | 3994 | btrfsic_submit_bio(bio); |
1c3063b6 | 3995 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3996 | } |
387125fc | 3997 | |
4fc6441a AJ |
3998 | /* |
3999 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4000 | */ | |
8c27cb35 | 4001 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4002 | { |
4fc6441a | 4003 | struct bio *bio = device->flush_bio; |
387125fc | 4004 | |
1c3063b6 | 4005 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4006 | return BLK_STS_OK; |
387125fc | 4007 | |
1c3063b6 | 4008 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4009 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4010 | |
8c27cb35 | 4011 | return bio->bi_status; |
387125fc | 4012 | } |
387125fc | 4013 | |
d10b82fe | 4014 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4015 | { |
6528b99d | 4016 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4017 | return -EIO; |
387125fc CM |
4018 | return 0; |
4019 | } | |
4020 | ||
4021 | /* | |
4022 | * send an empty flush down to each device in parallel, | |
4023 | * then wait for them | |
4024 | */ | |
4025 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4026 | { | |
4027 | struct list_head *head; | |
4028 | struct btrfs_device *dev; | |
5af3e8cc | 4029 | int errors_wait = 0; |
4e4cbee9 | 4030 | blk_status_t ret; |
387125fc | 4031 | |
1538e6c5 | 4032 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4033 | /* send down all the barriers */ |
4034 | head = &info->fs_devices->devices; | |
1538e6c5 | 4035 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4036 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4037 | continue; |
cea7c8bf | 4038 | if (!dev->bdev) |
387125fc | 4039 | continue; |
e12c9621 | 4040 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4041 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4042 | continue; |
4043 | ||
4fc6441a | 4044 | write_dev_flush(dev); |
58efbc9f | 4045 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4046 | } |
4047 | ||
4048 | /* wait for all the barriers */ | |
1538e6c5 | 4049 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4050 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4051 | continue; |
387125fc | 4052 | if (!dev->bdev) { |
5af3e8cc | 4053 | errors_wait++; |
387125fc CM |
4054 | continue; |
4055 | } | |
e12c9621 | 4056 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4057 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4058 | continue; |
4059 | ||
4fc6441a | 4060 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4061 | if (ret) { |
4062 | dev->last_flush_error = ret; | |
66b4993e DS |
4063 | btrfs_dev_stat_inc_and_print(dev, |
4064 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4065 | errors_wait++; |
401b41e5 AJ |
4066 | } |
4067 | } | |
4068 | ||
cea7c8bf | 4069 | if (errors_wait) { |
401b41e5 AJ |
4070 | /* |
4071 | * At some point we need the status of all disks | |
4072 | * to arrive at the volume status. So error checking | |
4073 | * is being pushed to a separate loop. | |
4074 | */ | |
d10b82fe | 4075 | return check_barrier_error(info); |
387125fc | 4076 | } |
387125fc CM |
4077 | return 0; |
4078 | } | |
4079 | ||
943c6e99 ZL |
4080 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4081 | { | |
8789f4fe ZL |
4082 | int raid_type; |
4083 | int min_tolerated = INT_MAX; | |
943c6e99 | 4084 | |
8789f4fe ZL |
4085 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4086 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4087 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4088 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4089 | tolerated_failures); | |
943c6e99 | 4090 | |
8789f4fe ZL |
4091 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4092 | if (raid_type == BTRFS_RAID_SINGLE) | |
4093 | continue; | |
41a6e891 | 4094 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4095 | continue; |
8c3e3582 | 4096 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4097 | btrfs_raid_array[raid_type]. |
4098 | tolerated_failures); | |
4099 | } | |
943c6e99 | 4100 | |
8789f4fe | 4101 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4102 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4103 | min_tolerated = 0; |
4104 | } | |
4105 | ||
4106 | return min_tolerated; | |
943c6e99 ZL |
4107 | } |
4108 | ||
eece6a9c | 4109 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4110 | { |
e5e9a520 | 4111 | struct list_head *head; |
f2984462 | 4112 | struct btrfs_device *dev; |
a061fc8d | 4113 | struct btrfs_super_block *sb; |
f2984462 | 4114 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4115 | int ret; |
4116 | int do_barriers; | |
a236aed1 CM |
4117 | int max_errors; |
4118 | int total_errors = 0; | |
a061fc8d | 4119 | u64 flags; |
f2984462 | 4120 | |
0b246afa | 4121 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4122 | |
4123 | /* | |
4124 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4125 | * not from fsync where the tree roots in fs_info have not | |
4126 | * been consistent on disk. | |
4127 | */ | |
4128 | if (max_mirrors == 0) | |
4129 | backup_super_roots(fs_info); | |
f2984462 | 4130 | |
0b246afa | 4131 | sb = fs_info->super_for_commit; |
a061fc8d | 4132 | dev_item = &sb->dev_item; |
e5e9a520 | 4133 | |
0b246afa JM |
4134 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4135 | head = &fs_info->fs_devices->devices; | |
4136 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4137 | |
5af3e8cc | 4138 | if (do_barriers) { |
0b246afa | 4139 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4140 | if (ret) { |
4141 | mutex_unlock( | |
0b246afa JM |
4142 | &fs_info->fs_devices->device_list_mutex); |
4143 | btrfs_handle_fs_error(fs_info, ret, | |
4144 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4145 | return ret; |
4146 | } | |
4147 | } | |
387125fc | 4148 | |
1538e6c5 | 4149 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4150 | if (!dev->bdev) { |
4151 | total_errors++; | |
4152 | continue; | |
4153 | } | |
e12c9621 | 4154 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4155 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4156 | continue; |
4157 | ||
2b82032c | 4158 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4159 | btrfs_set_stack_device_type(dev_item, dev->type); |
4160 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4161 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4162 | dev->commit_total_bytes); |
ce7213c7 MX |
4163 | btrfs_set_stack_device_bytes_used(dev_item, |
4164 | dev->commit_bytes_used); | |
a061fc8d CM |
4165 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4166 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4167 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4168 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4169 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4170 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4171 | |
a061fc8d CM |
4172 | flags = btrfs_super_flags(sb); |
4173 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4174 | ||
75cb857d QW |
4175 | ret = btrfs_validate_write_super(fs_info, sb); |
4176 | if (ret < 0) { | |
4177 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4178 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4179 | "unexpected superblock corruption detected"); | |
4180 | return -EUCLEAN; | |
4181 | } | |
4182 | ||
abbb3b8e | 4183 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4184 | if (ret) |
4185 | total_errors++; | |
f2984462 | 4186 | } |
a236aed1 | 4187 | if (total_errors > max_errors) { |
0b246afa JM |
4188 | btrfs_err(fs_info, "%d errors while writing supers", |
4189 | total_errors); | |
4190 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4191 | |
9d565ba4 | 4192 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4193 | btrfs_handle_fs_error(fs_info, -EIO, |
4194 | "%d errors while writing supers", | |
4195 | total_errors); | |
9d565ba4 | 4196 | return -EIO; |
a236aed1 | 4197 | } |
f2984462 | 4198 | |
a512bbf8 | 4199 | total_errors = 0; |
1538e6c5 | 4200 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4201 | if (!dev->bdev) |
4202 | continue; | |
e12c9621 | 4203 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4204 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4205 | continue; |
4206 | ||
abbb3b8e | 4207 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4208 | if (ret) |
4209 | total_errors++; | |
f2984462 | 4210 | } |
0b246afa | 4211 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4212 | if (total_errors > max_errors) { |
0b246afa JM |
4213 | btrfs_handle_fs_error(fs_info, -EIO, |
4214 | "%d errors while writing supers", | |
4215 | total_errors); | |
79787eaa | 4216 | return -EIO; |
a236aed1 | 4217 | } |
f2984462 CM |
4218 | return 0; |
4219 | } | |
4220 | ||
cb517eab MX |
4221 | /* Drop a fs root from the radix tree and free it. */ |
4222 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4223 | struct btrfs_root *root) | |
2619ba1f | 4224 | { |
4785e24f JB |
4225 | bool drop_ref = false; |
4226 | ||
4df27c4d | 4227 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
4228 | radix_tree_delete(&fs_info->fs_roots_radix, |
4229 | (unsigned long)root->root_key.objectid); | |
af01d2e5 | 4230 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
4785e24f | 4231 | drop_ref = true; |
4df27c4d | 4232 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4233 | |
1c1ea4f7 | 4234 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
ef67963d | 4235 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4236 | if (root->reloc_root) { |
00246528 | 4237 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4238 | root->reloc_root = NULL; |
4239 | } | |
4240 | } | |
3321719e | 4241 | |
4785e24f JB |
4242 | if (drop_ref) |
4243 | btrfs_put_root(root); | |
2619ba1f CM |
4244 | } |
4245 | ||
c146afad | 4246 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4247 | { |
c146afad YZ |
4248 | u64 root_objectid = 0; |
4249 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
4250 | int i = 0; |
4251 | int err = 0; | |
4252 | unsigned int ret = 0; | |
e089f05c | 4253 | |
c146afad | 4254 | while (1) { |
efc34534 | 4255 | spin_lock(&fs_info->fs_roots_radix_lock); |
c146afad YZ |
4256 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
4257 | (void **)gang, root_objectid, | |
4258 | ARRAY_SIZE(gang)); | |
65d33fd7 | 4259 | if (!ret) { |
efc34534 | 4260 | spin_unlock(&fs_info->fs_roots_radix_lock); |
c146afad | 4261 | break; |
65d33fd7 | 4262 | } |
5d4f98a2 | 4263 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4264 | |
c146afad | 4265 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
4266 | /* Avoid to grab roots in dead_roots */ |
4267 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4268 | gang[i] = NULL; | |
4269 | continue; | |
4270 | } | |
4271 | /* grab all the search result for later use */ | |
00246528 | 4272 | gang[i] = btrfs_grab_root(gang[i]); |
65d33fd7 | 4273 | } |
efc34534 | 4274 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4275 | |
65d33fd7 QW |
4276 | for (i = 0; i < ret; i++) { |
4277 | if (!gang[i]) | |
4278 | continue; | |
c146afad | 4279 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
4280 | err = btrfs_orphan_cleanup(gang[i]); |
4281 | if (err) | |
65d33fd7 | 4282 | break; |
00246528 | 4283 | btrfs_put_root(gang[i]); |
c146afad YZ |
4284 | } |
4285 | root_objectid++; | |
4286 | } | |
65d33fd7 QW |
4287 | |
4288 | /* release the uncleaned roots due to error */ | |
4289 | for (; i < ret; i++) { | |
4290 | if (gang[i]) | |
00246528 | 4291 | btrfs_put_root(gang[i]); |
65d33fd7 QW |
4292 | } |
4293 | return err; | |
c146afad | 4294 | } |
a2135011 | 4295 | |
6bccf3ab | 4296 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4297 | { |
6bccf3ab | 4298 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4299 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4300 | |
0b246afa | 4301 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4302 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4303 | mutex_unlock(&fs_info->cleaner_mutex); |
4304 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4305 | |
4306 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4307 | down_write(&fs_info->cleanup_work_sem); |
4308 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4309 | |
7a7eaa40 | 4310 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4311 | if (IS_ERR(trans)) |
4312 | return PTR_ERR(trans); | |
3a45bb20 | 4313 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4314 | } |
4315 | ||
b105e927 | 4316 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4317 | { |
c146afad YZ |
4318 | int ret; |
4319 | ||
afcdd129 | 4320 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
d6fd0ae2 OS |
4321 | /* |
4322 | * We don't want the cleaner to start new transactions, add more delayed | |
4323 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4324 | * because that frees the task_struct, and the transaction kthread might | |
4325 | * still try to wake up the cleaner. | |
4326 | */ | |
4327 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4328 | |
7343dd61 | 4329 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4330 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4331 | |
803b2f54 SB |
4332 | /* wait for the uuid_scan task to finish */ |
4333 | down(&fs_info->uuid_tree_rescan_sem); | |
4334 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4335 | up(&fs_info->uuid_tree_rescan_sem); | |
4336 | ||
837d5b6e | 4337 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4338 | btrfs_pause_balance(fs_info); |
837d5b6e | 4339 | |
8dabb742 SB |
4340 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4341 | ||
aa1b8cd4 | 4342 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4343 | |
4344 | /* wait for any defraggers to finish */ | |
4345 | wait_event(fs_info->transaction_wait, | |
4346 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4347 | ||
4348 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4349 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4350 | |
21c7e756 | 4351 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4352 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4353 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4354 | |
18bb8bbf JT |
4355 | cancel_work_sync(&fs_info->reclaim_bgs_work); |
4356 | ||
b0643e59 DZ |
4357 | /* Cancel or finish ongoing discard work */ |
4358 | btrfs_discard_cleanup(fs_info); | |
4359 | ||
bc98a42c | 4360 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4361 | /* |
d6fd0ae2 OS |
4362 | * The cleaner kthread is stopped, so do one final pass over |
4363 | * unused block groups. | |
e44163e1 | 4364 | */ |
0b246afa | 4365 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4366 | |
f0cc2cd7 FM |
4367 | /* |
4368 | * There might be existing delayed inode workers still running | |
4369 | * and holding an empty delayed inode item. We must wait for | |
4370 | * them to complete first because they can create a transaction. | |
4371 | * This happens when someone calls btrfs_balance_delayed_items() | |
4372 | * and then a transaction commit runs the same delayed nodes | |
4373 | * before any delayed worker has done something with the nodes. | |
4374 | * We must wait for any worker here and not at transaction | |
4375 | * commit time since that could cause a deadlock. | |
4376 | * This is a very rare case. | |
4377 | */ | |
4378 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4379 | ||
6bccf3ab | 4380 | ret = btrfs_commit_super(fs_info); |
acce952b | 4381 | if (ret) |
04892340 | 4382 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4383 | } |
4384 | ||
af722733 LB |
4385 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || |
4386 | test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) | |
2ff7e61e | 4387 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4388 | |
e3029d9f AV |
4389 | kthread_stop(fs_info->transaction_kthread); |
4390 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4391 | |
e187831e | 4392 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4393 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4394 | |
5958253c QW |
4395 | if (btrfs_check_quota_leak(fs_info)) { |
4396 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4397 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4398 | } | |
4399 | ||
04892340 | 4400 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4401 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4402 | |
963d678b | 4403 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4404 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4405 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4406 | } |
bcc63abb | 4407 | |
5deb17e1 | 4408 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4409 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4410 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4411 | |
6618a59b | 4412 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4413 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4414 | |
1a4319cc LB |
4415 | btrfs_put_block_group_cache(fs_info); |
4416 | ||
de348ee0 WS |
4417 | /* |
4418 | * we must make sure there is not any read request to | |
4419 | * submit after we stopping all workers. | |
4420 | */ | |
4421 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4422 | btrfs_stop_all_workers(fs_info); |
4423 | ||
0a31daa4 FM |
4424 | /* We shouldn't have any transaction open at this point */ |
4425 | ASSERT(list_empty(&fs_info->trans_list)); | |
4426 | ||
afcdd129 | 4427 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4428 | free_root_pointers(fs_info, true); |
8c38938c | 4429 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4430 | |
4e19443d JB |
4431 | /* |
4432 | * We must free the block groups after dropping the fs_roots as we could | |
4433 | * have had an IO error and have left over tree log blocks that aren't | |
4434 | * cleaned up until the fs roots are freed. This makes the block group | |
4435 | * accounting appear to be wrong because there's pending reserved bytes, | |
4436 | * so make sure we do the block group cleanup afterwards. | |
4437 | */ | |
4438 | btrfs_free_block_groups(fs_info); | |
4439 | ||
13e6c37b | 4440 | iput(fs_info->btree_inode); |
d6bfde87 | 4441 | |
21adbd5c | 4442 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4443 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4444 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4445 | #endif |
4446 | ||
0b86a832 | 4447 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4448 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4449 | } |
4450 | ||
b9fab919 CM |
4451 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4452 | int atomic) | |
5f39d397 | 4453 | { |
1259ab75 | 4454 | int ret; |
727011e0 | 4455 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4456 | |
0b32f4bb | 4457 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4458 | if (!ret) |
4459 | return ret; | |
4460 | ||
4461 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4462 | parent_transid, atomic); |
4463 | if (ret == -EAGAIN) | |
4464 | return ret; | |
1259ab75 | 4465 | return !ret; |
5f39d397 CM |
4466 | } |
4467 | ||
5f39d397 CM |
4468 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4469 | { | |
2f4d60df | 4470 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4471 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4472 | int was_dirty; |
b4ce94de | 4473 | |
06ea65a3 JB |
4474 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4475 | /* | |
4476 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4477 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4478 | * outside of the sanity tests. |
4479 | */ | |
b0132a3b | 4480 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4481 | return; |
4482 | #endif | |
b9447ef8 | 4483 | btrfs_assert_tree_locked(buf); |
0b246afa | 4484 | if (transid != fs_info->generation) |
5d163e0e | 4485 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4486 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4487 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4488 | if (!was_dirty) |
104b4e51 NB |
4489 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4490 | buf->len, | |
4491 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4492 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4493 | /* |
4494 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4495 | * but item data not updated. | |
4496 | * So here we should only check item pointers, not item data. | |
4497 | */ | |
4498 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4499 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4500 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4501 | ASSERT(0); |
4502 | } | |
4503 | #endif | |
eb60ceac CM |
4504 | } |
4505 | ||
2ff7e61e | 4506 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4507 | int flush_delayed) |
16cdcec7 MX |
4508 | { |
4509 | /* | |
4510 | * looks as though older kernels can get into trouble with | |
4511 | * this code, they end up stuck in balance_dirty_pages forever | |
4512 | */ | |
e2d84521 | 4513 | int ret; |
16cdcec7 MX |
4514 | |
4515 | if (current->flags & PF_MEMALLOC) | |
4516 | return; | |
4517 | ||
b53d3f5d | 4518 | if (flush_delayed) |
2ff7e61e | 4519 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4520 | |
d814a491 EL |
4521 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4522 | BTRFS_DIRTY_METADATA_THRESH, | |
4523 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4524 | if (ret > 0) { |
0b246afa | 4525 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4526 | } |
16cdcec7 MX |
4527 | } |
4528 | ||
2ff7e61e | 4529 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4530 | { |
2ff7e61e | 4531 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4532 | } |
585ad2c3 | 4533 | |
2ff7e61e | 4534 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4535 | { |
2ff7e61e | 4536 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4537 | } |
6b80053d | 4538 | |
581c1760 QW |
4539 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4540 | struct btrfs_key *first_key) | |
6b80053d | 4541 | { |
5ab12d1f | 4542 | return btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 4543 | level, first_key); |
6b80053d | 4544 | } |
0da5468f | 4545 | |
2ff7e61e | 4546 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4547 | { |
fe816d0f NB |
4548 | /* cleanup FS via transaction */ |
4549 | btrfs_cleanup_transaction(fs_info); | |
4550 | ||
0b246afa | 4551 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4552 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4553 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4554 | |
0b246afa JM |
4555 | down_write(&fs_info->cleanup_work_sem); |
4556 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4557 | } |
4558 | ||
ef67963d JB |
4559 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4560 | { | |
4561 | struct btrfs_root *gang[8]; | |
4562 | u64 root_objectid = 0; | |
4563 | int ret; | |
4564 | ||
4565 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4566 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4567 | (void **)gang, root_objectid, | |
4568 | ARRAY_SIZE(gang))) != 0) { | |
4569 | int i; | |
4570 | ||
4571 | for (i = 0; i < ret; i++) | |
4572 | gang[i] = btrfs_grab_root(gang[i]); | |
4573 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4574 | ||
4575 | for (i = 0; i < ret; i++) { | |
4576 | if (!gang[i]) | |
4577 | continue; | |
4578 | root_objectid = gang[i]->root_key.objectid; | |
4579 | btrfs_free_log(NULL, gang[i]); | |
4580 | btrfs_put_root(gang[i]); | |
4581 | } | |
4582 | root_objectid++; | |
4583 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4584 | } | |
4585 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4586 | btrfs_free_log_root_tree(NULL, fs_info); | |
4587 | } | |
4588 | ||
143bede5 | 4589 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4590 | { |
acce952b | 4591 | struct btrfs_ordered_extent *ordered; |
acce952b | 4592 | |
199c2a9c | 4593 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4594 | /* |
4595 | * This will just short circuit the ordered completion stuff which will | |
4596 | * make sure the ordered extent gets properly cleaned up. | |
4597 | */ | |
199c2a9c | 4598 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4599 | root_extent_list) |
4600 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4601 | spin_unlock(&root->ordered_extent_lock); |
4602 | } | |
4603 | ||
4604 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4605 | { | |
4606 | struct btrfs_root *root; | |
4607 | struct list_head splice; | |
4608 | ||
4609 | INIT_LIST_HEAD(&splice); | |
4610 | ||
4611 | spin_lock(&fs_info->ordered_root_lock); | |
4612 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4613 | while (!list_empty(&splice)) { | |
4614 | root = list_first_entry(&splice, struct btrfs_root, | |
4615 | ordered_root); | |
1de2cfde JB |
4616 | list_move_tail(&root->ordered_root, |
4617 | &fs_info->ordered_roots); | |
199c2a9c | 4618 | |
2a85d9ca | 4619 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4620 | btrfs_destroy_ordered_extents(root); |
4621 | ||
2a85d9ca LB |
4622 | cond_resched(); |
4623 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4624 | } |
4625 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4626 | |
4627 | /* | |
4628 | * We need this here because if we've been flipped read-only we won't | |
4629 | * get sync() from the umount, so we need to make sure any ordered | |
4630 | * extents that haven't had their dirty pages IO start writeout yet | |
4631 | * actually get run and error out properly. | |
4632 | */ | |
4633 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4634 | } |
4635 | ||
35a3621b | 4636 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4637 | struct btrfs_fs_info *fs_info) |
acce952b | 4638 | { |
4639 | struct rb_node *node; | |
4640 | struct btrfs_delayed_ref_root *delayed_refs; | |
4641 | struct btrfs_delayed_ref_node *ref; | |
4642 | int ret = 0; | |
4643 | ||
4644 | delayed_refs = &trans->delayed_refs; | |
4645 | ||
4646 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4647 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4648 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4649 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4650 | return ret; |
4651 | } | |
4652 | ||
5c9d028b | 4653 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4654 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4655 | struct rb_node *n; |
e78417d1 | 4656 | bool pin_bytes = false; |
acce952b | 4657 | |
d7df2c79 JB |
4658 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4659 | href_node); | |
3069bd26 | 4660 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4661 | continue; |
3069bd26 | 4662 | |
d7df2c79 | 4663 | spin_lock(&head->lock); |
e3d03965 | 4664 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4665 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4666 | ref_node); | |
d7df2c79 | 4667 | ref->in_tree = 0; |
e3d03965 | 4668 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4669 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4670 | if (!list_empty(&ref->add_list)) |
4671 | list_del(&ref->add_list); | |
d7df2c79 JB |
4672 | atomic_dec(&delayed_refs->num_entries); |
4673 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4674 | } |
d7df2c79 JB |
4675 | if (head->must_insert_reserved) |
4676 | pin_bytes = true; | |
4677 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4678 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4679 | spin_unlock(&head->lock); |
4680 | spin_unlock(&delayed_refs->lock); | |
4681 | mutex_unlock(&head->mutex); | |
acce952b | 4682 | |
f603bb94 NB |
4683 | if (pin_bytes) { |
4684 | struct btrfs_block_group *cache; | |
4685 | ||
4686 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4687 | BUG_ON(!cache); | |
4688 | ||
4689 | spin_lock(&cache->space_info->lock); | |
4690 | spin_lock(&cache->lock); | |
4691 | cache->pinned += head->num_bytes; | |
4692 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4693 | cache->space_info, head->num_bytes); | |
4694 | cache->reserved -= head->num_bytes; | |
4695 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4696 | spin_unlock(&cache->lock); | |
4697 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4698 | |
4699 | btrfs_put_block_group(cache); | |
4700 | ||
4701 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4702 | head->bytenr + head->num_bytes - 1); | |
4703 | } | |
31890da0 | 4704 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4705 | btrfs_put_delayed_ref_head(head); |
acce952b | 4706 | cond_resched(); |
4707 | spin_lock(&delayed_refs->lock); | |
4708 | } | |
81f7eb00 | 4709 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4710 | |
4711 | spin_unlock(&delayed_refs->lock); | |
4712 | ||
4713 | return ret; | |
4714 | } | |
4715 | ||
143bede5 | 4716 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4717 | { |
4718 | struct btrfs_inode *btrfs_inode; | |
4719 | struct list_head splice; | |
4720 | ||
4721 | INIT_LIST_HEAD(&splice); | |
4722 | ||
eb73c1b7 MX |
4723 | spin_lock(&root->delalloc_lock); |
4724 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4725 | |
4726 | while (!list_empty(&splice)) { | |
fe816d0f | 4727 | struct inode *inode = NULL; |
eb73c1b7 MX |
4728 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4729 | delalloc_inodes); | |
fe816d0f | 4730 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4731 | spin_unlock(&root->delalloc_lock); |
acce952b | 4732 | |
fe816d0f NB |
4733 | /* |
4734 | * Make sure we get a live inode and that it'll not disappear | |
4735 | * meanwhile. | |
4736 | */ | |
4737 | inode = igrab(&btrfs_inode->vfs_inode); | |
4738 | if (inode) { | |
4739 | invalidate_inode_pages2(inode->i_mapping); | |
4740 | iput(inode); | |
4741 | } | |
eb73c1b7 | 4742 | spin_lock(&root->delalloc_lock); |
acce952b | 4743 | } |
eb73c1b7 MX |
4744 | spin_unlock(&root->delalloc_lock); |
4745 | } | |
4746 | ||
4747 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4748 | { | |
4749 | struct btrfs_root *root; | |
4750 | struct list_head splice; | |
4751 | ||
4752 | INIT_LIST_HEAD(&splice); | |
4753 | ||
4754 | spin_lock(&fs_info->delalloc_root_lock); | |
4755 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4756 | while (!list_empty(&splice)) { | |
4757 | root = list_first_entry(&splice, struct btrfs_root, | |
4758 | delalloc_root); | |
00246528 | 4759 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4760 | BUG_ON(!root); |
4761 | spin_unlock(&fs_info->delalloc_root_lock); | |
4762 | ||
4763 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4764 | btrfs_put_root(root); |
eb73c1b7 MX |
4765 | |
4766 | spin_lock(&fs_info->delalloc_root_lock); | |
4767 | } | |
4768 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4769 | } |
4770 | ||
2ff7e61e | 4771 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4772 | struct extent_io_tree *dirty_pages, |
4773 | int mark) | |
4774 | { | |
4775 | int ret; | |
acce952b | 4776 | struct extent_buffer *eb; |
4777 | u64 start = 0; | |
4778 | u64 end; | |
acce952b | 4779 | |
4780 | while (1) { | |
4781 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4782 | mark, NULL); |
acce952b | 4783 | if (ret) |
4784 | break; | |
4785 | ||
91166212 | 4786 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4787 | while (start <= end) { |
0b246afa JM |
4788 | eb = find_extent_buffer(fs_info, start); |
4789 | start += fs_info->nodesize; | |
fd8b2b61 | 4790 | if (!eb) |
acce952b | 4791 | continue; |
fd8b2b61 | 4792 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4793 | |
fd8b2b61 JB |
4794 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4795 | &eb->bflags)) | |
4796 | clear_extent_buffer_dirty(eb); | |
4797 | free_extent_buffer_stale(eb); | |
acce952b | 4798 | } |
4799 | } | |
4800 | ||
4801 | return ret; | |
4802 | } | |
4803 | ||
2ff7e61e | 4804 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 4805 | struct extent_io_tree *unpin) |
acce952b | 4806 | { |
acce952b | 4807 | u64 start; |
4808 | u64 end; | |
4809 | int ret; | |
4810 | ||
acce952b | 4811 | while (1) { |
0e6ec385 FM |
4812 | struct extent_state *cached_state = NULL; |
4813 | ||
fcd5e742 LF |
4814 | /* |
4815 | * The btrfs_finish_extent_commit() may get the same range as | |
4816 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4817 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4818 | * the same extent range. | |
4819 | */ | |
4820 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4821 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 4822 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
4823 | if (ret) { |
4824 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4825 | break; |
fcd5e742 | 4826 | } |
acce952b | 4827 | |
0e6ec385 FM |
4828 | clear_extent_dirty(unpin, start, end, &cached_state); |
4829 | free_extent_state(cached_state); | |
2ff7e61e | 4830 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4831 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4832 | cond_resched(); |
4833 | } | |
4834 | ||
4835 | return 0; | |
4836 | } | |
4837 | ||
32da5386 | 4838 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4839 | { |
4840 | struct inode *inode; | |
4841 | ||
4842 | inode = cache->io_ctl.inode; | |
4843 | if (inode) { | |
4844 | invalidate_inode_pages2(inode->i_mapping); | |
4845 | BTRFS_I(inode)->generation = 0; | |
4846 | cache->io_ctl.inode = NULL; | |
4847 | iput(inode); | |
4848 | } | |
bbc37d6e | 4849 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4850 | btrfs_put_block_group(cache); |
4851 | } | |
4852 | ||
4853 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4854 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4855 | { |
32da5386 | 4856 | struct btrfs_block_group *cache; |
c79a1751 LB |
4857 | |
4858 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4859 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4860 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4861 | struct btrfs_block_group, |
c79a1751 | 4862 | dirty_list); |
c79a1751 LB |
4863 | |
4864 | if (!list_empty(&cache->io_list)) { | |
4865 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4866 | list_del_init(&cache->io_list); | |
4867 | btrfs_cleanup_bg_io(cache); | |
4868 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4869 | } | |
4870 | ||
4871 | list_del_init(&cache->dirty_list); | |
4872 | spin_lock(&cache->lock); | |
4873 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4874 | spin_unlock(&cache->lock); | |
4875 | ||
4876 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4877 | btrfs_put_block_group(cache); | |
ba2c4d4e | 4878 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
4879 | spin_lock(&cur_trans->dirty_bgs_lock); |
4880 | } | |
4881 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4882 | ||
45ae2c18 NB |
4883 | /* |
4884 | * Refer to the definition of io_bgs member for details why it's safe | |
4885 | * to use it without any locking | |
4886 | */ | |
c79a1751 LB |
4887 | while (!list_empty(&cur_trans->io_bgs)) { |
4888 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4889 | struct btrfs_block_group, |
c79a1751 | 4890 | io_list); |
c79a1751 LB |
4891 | |
4892 | list_del_init(&cache->io_list); | |
4893 | spin_lock(&cache->lock); | |
4894 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4895 | spin_unlock(&cache->lock); | |
4896 | btrfs_cleanup_bg_io(cache); | |
4897 | } | |
4898 | } | |
4899 | ||
49b25e05 | 4900 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4901 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4902 | { |
bbbf7243 NB |
4903 | struct btrfs_device *dev, *tmp; |
4904 | ||
2ff7e61e | 4905 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4906 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4907 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4908 | ||
bbbf7243 NB |
4909 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4910 | post_commit_list) { | |
4911 | list_del_init(&dev->post_commit_list); | |
4912 | } | |
4913 | ||
2ff7e61e | 4914 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4915 | |
4a9d8bde | 4916 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4917 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4918 | |
4a9d8bde | 4919 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4920 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4921 | |
ccdf9b30 | 4922 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4923 | |
2ff7e61e | 4924 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4925 | EXTENT_DIRTY); |
fe119a6e | 4926 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4927 | |
d3575156 NA |
4928 | btrfs_free_redirty_list(cur_trans); |
4929 | ||
4a9d8bde MX |
4930 | cur_trans->state =TRANS_STATE_COMPLETED; |
4931 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4932 | } |
4933 | ||
2ff7e61e | 4934 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4935 | { |
4936 | struct btrfs_transaction *t; | |
acce952b | 4937 | |
0b246afa | 4938 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4939 | |
0b246afa JM |
4940 | spin_lock(&fs_info->trans_lock); |
4941 | while (!list_empty(&fs_info->trans_list)) { | |
4942 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4943 | struct btrfs_transaction, list); |
4944 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4945 | refcount_inc(&t->use_count); |
0b246afa | 4946 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4947 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4948 | btrfs_put_transaction(t); |
0b246afa | 4949 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4950 | continue; |
4951 | } | |
0b246afa | 4952 | if (t == fs_info->running_transaction) { |
724e2315 | 4953 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4954 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4955 | /* |
4956 | * We wait for 0 num_writers since we don't hold a trans | |
4957 | * handle open currently for this transaction. | |
4958 | */ | |
4959 | wait_event(t->writer_wait, | |
4960 | atomic_read(&t->num_writers) == 0); | |
4961 | } else { | |
0b246afa | 4962 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4963 | } |
2ff7e61e | 4964 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4965 | |
0b246afa JM |
4966 | spin_lock(&fs_info->trans_lock); |
4967 | if (t == fs_info->running_transaction) | |
4968 | fs_info->running_transaction = NULL; | |
acce952b | 4969 | list_del_init(&t->list); |
0b246afa | 4970 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4971 | |
724e2315 | 4972 | btrfs_put_transaction(t); |
2ff7e61e | 4973 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4974 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4975 | } |
0b246afa JM |
4976 | spin_unlock(&fs_info->trans_lock); |
4977 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4978 | btrfs_destroy_delayed_inodes(fs_info); |
4979 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4980 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4981 | btrfs_drop_all_logs(fs_info); |
0b246afa | 4982 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4983 | |
4984 | return 0; | |
4985 | } | |
ec7d6dfd | 4986 | |
453e4873 | 4987 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
4988 | { |
4989 | struct btrfs_path *path; | |
4990 | int ret; | |
4991 | struct extent_buffer *l; | |
4992 | struct btrfs_key search_key; | |
4993 | struct btrfs_key found_key; | |
4994 | int slot; | |
4995 | ||
4996 | path = btrfs_alloc_path(); | |
4997 | if (!path) | |
4998 | return -ENOMEM; | |
4999 | ||
5000 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5001 | search_key.type = -1; | |
5002 | search_key.offset = (u64)-1; | |
5003 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5004 | if (ret < 0) | |
5005 | goto error; | |
5006 | BUG_ON(ret == 0); /* Corruption */ | |
5007 | if (path->slots[0] > 0) { | |
5008 | slot = path->slots[0] - 1; | |
5009 | l = path->nodes[0]; | |
5010 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5011 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5012 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5013 | } else { |
23125104 | 5014 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5015 | } |
5016 | ret = 0; | |
5017 | error: | |
5018 | btrfs_free_path(path); | |
5019 | return ret; | |
5020 | } | |
5021 | ||
543068a2 | 5022 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5023 | { |
5024 | int ret; | |
5025 | mutex_lock(&root->objectid_mutex); | |
5026 | ||
6b8fad57 | 5027 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5028 | btrfs_warn(root->fs_info, |
5029 | "the objectid of root %llu reaches its highest value", | |
5030 | root->root_key.objectid); | |
5031 | ret = -ENOSPC; | |
5032 | goto out; | |
5033 | } | |
5034 | ||
23125104 | 5035 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5036 | ret = 0; |
5037 | out: | |
5038 | mutex_unlock(&root->objectid_mutex); | |
5039 | return ret; | |
5040 | } |