]>
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 | ||
6ffd7be5 QW |
444 | if (ret < 0) |
445 | goto error; | |
446 | ||
447 | /* | |
448 | * Also check the generation, the eb reached here must be newer than | |
449 | * last committed. Or something seriously wrong happened. | |
450 | */ | |
451 | if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { | |
452 | ret = -EUCLEAN; | |
eca0f6f6 | 453 | btrfs_err(fs_info, |
6ffd7be5 QW |
454 | "block=%llu bad generation, have %llu expect > %llu", |
455 | eb->start, btrfs_header_generation(eb), | |
456 | fs_info->last_trans_committed); | |
457 | goto error; | |
eca0f6f6 QW |
458 | } |
459 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
460 | ||
461 | return 0; | |
6ffd7be5 QW |
462 | |
463 | error: | |
464 | btrfs_print_tree(eb, 0); | |
465 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
466 | eb->start); | |
467 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
468 | return ret; | |
eca0f6f6 QW |
469 | } |
470 | ||
471 | /* Checksum all dirty extent buffers in one bio_vec */ | |
472 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
473 | struct bio_vec *bvec) | |
474 | { | |
475 | struct page *page = bvec->bv_page; | |
476 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
477 | u64 cur; | |
478 | int ret = 0; | |
479 | ||
480 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
481 | cur += fs_info->nodesize) { | |
482 | struct extent_buffer *eb; | |
483 | bool uptodate; | |
484 | ||
485 | eb = find_extent_buffer(fs_info, cur); | |
486 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
487 | fs_info->nodesize); | |
488 | ||
489 | /* A dirty eb shouldn't disappear from buffer_radix */ | |
490 | if (WARN_ON(!eb)) | |
491 | return -EUCLEAN; | |
492 | ||
493 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
494 | free_extent_buffer(eb); | |
495 | return -EUCLEAN; | |
496 | } | |
497 | if (WARN_ON(!uptodate)) { | |
498 | free_extent_buffer(eb); | |
499 | return -EUCLEAN; | |
500 | } | |
501 | ||
502 | ret = csum_one_extent_buffer(eb); | |
503 | free_extent_buffer(eb); | |
504 | if (ret < 0) | |
505 | return ret; | |
506 | } | |
507 | return ret; | |
508 | } | |
509 | ||
d352ac68 | 510 | /* |
ac303b69 QW |
511 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
512 | * we only fill in the checksum field in the first page of a multi-page block. | |
513 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 514 | */ |
ac303b69 | 515 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 516 | { |
ac303b69 | 517 | struct page *page = bvec->bv_page; |
4eee4fa4 | 518 | u64 start = page_offset(page); |
19c00ddc | 519 | u64 found_start; |
19c00ddc | 520 | struct extent_buffer *eb; |
eca0f6f6 QW |
521 | |
522 | if (fs_info->sectorsize < PAGE_SIZE) | |
523 | return csum_dirty_subpage_buffers(fs_info, bvec); | |
f188591e | 524 | |
4f2de97a JB |
525 | eb = (struct extent_buffer *)page->private; |
526 | if (page != eb->pages[0]) | |
527 | return 0; | |
0f805531 | 528 | |
19c00ddc | 529 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
530 | |
531 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
532 | WARN_ON(found_start != 0); | |
533 | return 0; | |
534 | } | |
535 | ||
0f805531 AL |
536 | /* |
537 | * Please do not consolidate these warnings into a single if. | |
538 | * It is useful to know what went wrong. | |
539 | */ | |
540 | if (WARN_ON(found_start != start)) | |
541 | return -EUCLEAN; | |
542 | if (WARN_ON(!PageUptodate(page))) | |
543 | return -EUCLEAN; | |
544 | ||
eca0f6f6 | 545 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
546 | } |
547 | ||
b0c9b3b0 | 548 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 549 | { |
b0c9b3b0 | 550 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 551 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 552 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 553 | u8 *metadata_uuid; |
2b82032c | 554 | |
9a8658e3 DS |
555 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
556 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
557 | /* |
558 | * Checking the incompat flag is only valid for the current fs. For | |
559 | * seed devices it's forbidden to have their uuid changed so reading | |
560 | * ->fsid in this case is fine | |
561 | */ | |
562 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
563 | metadata_uuid = fs_devices->metadata_uuid; | |
564 | else | |
565 | metadata_uuid = fs_devices->fsid; | |
566 | ||
567 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
568 | return 0; | |
569 | ||
570 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
571 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
572 | return 0; | |
573 | ||
574 | return 1; | |
2b82032c YZ |
575 | } |
576 | ||
77bf40a2 QW |
577 | /* Do basic extent buffer checks at read time */ |
578 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 579 | { |
77bf40a2 | 580 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 581 | u64 found_start; |
77bf40a2 QW |
582 | const u32 csum_size = fs_info->csum_size; |
583 | u8 found_level; | |
2996e1f8 | 584 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 585 | const u8 *header_csum; |
77bf40a2 | 586 | int ret = 0; |
ea466794 | 587 | |
ce9adaa5 | 588 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 589 | if (found_start != eb->start) { |
893bf4b1 SY |
590 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
591 | eb->start, found_start); | |
f188591e | 592 | ret = -EIO; |
77bf40a2 | 593 | goto out; |
ce9adaa5 | 594 | } |
b0c9b3b0 | 595 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
596 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
597 | eb->start); | |
1259ab75 | 598 | ret = -EIO; |
77bf40a2 | 599 | goto out; |
1259ab75 | 600 | } |
ce9adaa5 | 601 | found_level = btrfs_header_level(eb); |
1c24c3ce | 602 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
603 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
604 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce | 605 | ret = -EIO; |
77bf40a2 | 606 | goto out; |
1c24c3ce | 607 | } |
ce9adaa5 | 608 | |
c67b3892 | 609 | csum_tree_block(eb, result); |
dfd29eed DS |
610 | header_csum = page_address(eb->pages[0]) + |
611 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 612 | |
dfd29eed | 613 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 614 | btrfs_warn_rl(fs_info, |
ff14aa79 DS |
615 | "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
616 | eb->start, | |
dfd29eed | 617 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
618 | CSUM_FMT_VALUE(csum_size, result), |
619 | btrfs_header_level(eb)); | |
2996e1f8 | 620 | ret = -EUCLEAN; |
77bf40a2 | 621 | goto out; |
2996e1f8 JT |
622 | } |
623 | ||
a826d6dc JB |
624 | /* |
625 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
626 | * that we don't try and read the other copies of this block, just | |
627 | * return -EIO. | |
628 | */ | |
1c4360ee | 629 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
630 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
631 | ret = -EIO; | |
632 | } | |
ce9adaa5 | 633 | |
813fd1dc | 634 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
635 | ret = -EIO; |
636 | ||
0b32f4bb JB |
637 | if (!ret) |
638 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
639 | else |
640 | btrfs_err(fs_info, | |
641 | "block=%llu read time tree block corruption detected", | |
642 | eb->start); | |
77bf40a2 QW |
643 | out: |
644 | return ret; | |
645 | } | |
646 | ||
371cdc07 QW |
647 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
648 | int mirror) | |
649 | { | |
650 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
651 | struct extent_buffer *eb; | |
652 | bool reads_done; | |
653 | int ret = 0; | |
654 | ||
655 | /* | |
656 | * We don't allow bio merge for subpage metadata read, so we should | |
657 | * only get one eb for each endio hook. | |
658 | */ | |
659 | ASSERT(end == start + fs_info->nodesize - 1); | |
660 | ASSERT(PagePrivate(page)); | |
661 | ||
662 | eb = find_extent_buffer(fs_info, start); | |
663 | /* | |
664 | * When we are reading one tree block, eb must have been inserted into | |
665 | * the radix tree. If not, something is wrong. | |
666 | */ | |
667 | ASSERT(eb); | |
668 | ||
669 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
670 | /* Subpage read must finish in page read */ | |
671 | ASSERT(reads_done); | |
672 | ||
673 | eb->read_mirror = mirror; | |
674 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
675 | ret = -EIO; | |
676 | goto err; | |
677 | } | |
678 | ret = validate_extent_buffer(eb); | |
679 | if (ret < 0) | |
680 | goto err; | |
681 | ||
682 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
683 | btree_readahead_hook(eb, ret); | |
684 | ||
685 | set_extent_buffer_uptodate(eb); | |
686 | ||
687 | free_extent_buffer(eb); | |
688 | return ret; | |
689 | err: | |
690 | /* | |
691 | * end_bio_extent_readpage decrements io_pages in case of error, | |
692 | * make sure it has something to decrement. | |
693 | */ | |
694 | atomic_inc(&eb->io_pages); | |
695 | clear_extent_buffer_uptodate(eb); | |
696 | free_extent_buffer(eb); | |
697 | return ret; | |
698 | } | |
699 | ||
8e1dc982 | 700 | int btrfs_validate_metadata_buffer(struct btrfs_io_bio *io_bio, |
77bf40a2 QW |
701 | struct page *page, u64 start, u64 end, |
702 | int mirror) | |
703 | { | |
704 | struct extent_buffer *eb; | |
705 | int ret = 0; | |
706 | int reads_done; | |
707 | ||
708 | ASSERT(page->private); | |
371cdc07 QW |
709 | |
710 | if (btrfs_sb(page->mapping->host->i_sb)->sectorsize < PAGE_SIZE) | |
711 | return validate_subpage_buffer(page, start, end, mirror); | |
712 | ||
77bf40a2 QW |
713 | eb = (struct extent_buffer *)page->private; |
714 | ||
715 | /* | |
716 | * The pending IO might have been the only thing that kept this buffer | |
717 | * in memory. Make sure we have a ref for all this other checks | |
718 | */ | |
719 | atomic_inc(&eb->refs); | |
720 | ||
721 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
722 | if (!reads_done) | |
723 | goto err; | |
724 | ||
725 | eb->read_mirror = mirror; | |
726 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
727 | ret = -EIO; | |
728 | goto err; | |
729 | } | |
730 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 731 | err: |
79fb65a1 JB |
732 | if (reads_done && |
733 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 734 | btree_readahead_hook(eb, ret); |
4bb31e92 | 735 | |
53b381b3 DW |
736 | if (ret) { |
737 | /* | |
738 | * our io error hook is going to dec the io pages | |
739 | * again, we have to make sure it has something | |
740 | * to decrement | |
741 | */ | |
742 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 743 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 744 | } |
0b32f4bb | 745 | free_extent_buffer(eb); |
77bf40a2 | 746 | |
f188591e | 747 | return ret; |
ce9adaa5 CM |
748 | } |
749 | ||
4246a0b6 | 750 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 751 | { |
97eb6b69 | 752 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 753 | struct btrfs_fs_info *fs_info; |
9e0af237 | 754 | struct btrfs_workqueue *wq; |
ce9adaa5 | 755 | |
ce9adaa5 | 756 | fs_info = end_io_wq->info; |
4e4cbee9 | 757 | end_io_wq->status = bio->bi_status; |
d20f7043 | 758 | |
cfe94440 | 759 | if (btrfs_op(bio) == BTRFS_MAP_WRITE) { |
a0cac0ec | 760 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 761 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 762 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 763 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 764 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 765 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 766 | else |
9e0af237 | 767 | wq = fs_info->endio_write_workers; |
d20f7043 | 768 | } else { |
5c047a69 | 769 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 770 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 771 | else if (end_io_wq->metadata) |
9e0af237 | 772 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 773 | else |
9e0af237 | 774 | wq = fs_info->endio_workers; |
d20f7043 | 775 | } |
9e0af237 | 776 | |
a0cac0ec | 777 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 778 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
779 | } |
780 | ||
4e4cbee9 | 781 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 782 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 783 | { |
97eb6b69 | 784 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 785 | |
97eb6b69 | 786 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 787 | if (!end_io_wq) |
4e4cbee9 | 788 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
789 | |
790 | end_io_wq->private = bio->bi_private; | |
791 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 792 | end_io_wq->info = info; |
4e4cbee9 | 793 | end_io_wq->status = 0; |
ce9adaa5 | 794 | end_io_wq->bio = bio; |
22c59948 | 795 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
796 | |
797 | bio->bi_private = end_io_wq; | |
798 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
799 | return 0; |
800 | } | |
801 | ||
4a69a410 CM |
802 | static void run_one_async_start(struct btrfs_work *work) |
803 | { | |
4a69a410 | 804 | struct async_submit_bio *async; |
4e4cbee9 | 805 | blk_status_t ret; |
4a69a410 CM |
806 | |
807 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
808 | ret = async->submit_bio_start(async->inode, async->bio, |
809 | async->dio_file_offset); | |
79787eaa | 810 | if (ret) |
4e4cbee9 | 811 | async->status = ret; |
4a69a410 CM |
812 | } |
813 | ||
06ea01b1 DS |
814 | /* |
815 | * In order to insert checksums into the metadata in large chunks, we wait | |
816 | * until bio submission time. All the pages in the bio are checksummed and | |
817 | * sums are attached onto the ordered extent record. | |
818 | * | |
819 | * At IO completion time the csums attached on the ordered extent record are | |
820 | * inserted into the tree. | |
821 | */ | |
4a69a410 | 822 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 823 | { |
8b712842 | 824 | struct async_submit_bio *async; |
06ea01b1 DS |
825 | struct inode *inode; |
826 | blk_status_t ret; | |
8b712842 CM |
827 | |
828 | async = container_of(work, struct async_submit_bio, work); | |
8896a08d | 829 | inode = async->inode; |
4854ddd0 | 830 | |
bb7ab3b9 | 831 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
832 | if (async->status) { |
833 | async->bio->bi_status = async->status; | |
4246a0b6 | 834 | bio_endio(async->bio); |
79787eaa JM |
835 | return; |
836 | } | |
837 | ||
ec39f769 CM |
838 | /* |
839 | * All of the bios that pass through here are from async helpers. | |
840 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
841 | * This changes nothing when cgroups aren't in use. | |
842 | */ | |
843 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 844 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
845 | if (ret) { |
846 | async->bio->bi_status = ret; | |
847 | bio_endio(async->bio); | |
848 | } | |
4a69a410 CM |
849 | } |
850 | ||
851 | static void run_one_async_free(struct btrfs_work *work) | |
852 | { | |
853 | struct async_submit_bio *async; | |
854 | ||
855 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
856 | kfree(async); |
857 | } | |
858 | ||
8896a08d | 859 | blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, |
8c27cb35 | 860 | int mirror_num, unsigned long bio_flags, |
1941b64b | 861 | u64 dio_file_offset, |
e288c080 | 862 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e | 863 | { |
8896a08d | 864 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
865 | struct async_submit_bio *async; |
866 | ||
867 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
868 | if (!async) | |
4e4cbee9 | 869 | return BLK_STS_RESOURCE; |
44b8bd7e | 870 | |
8896a08d | 871 | async->inode = inode; |
44b8bd7e CM |
872 | async->bio = bio; |
873 | async->mirror_num = mirror_num; | |
4a69a410 | 874 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 875 | |
a0cac0ec OS |
876 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
877 | run_one_async_free); | |
4a69a410 | 878 | |
1941b64b | 879 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 880 | |
4e4cbee9 | 881 | async->status = 0; |
79787eaa | 882 | |
67f055c7 | 883 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 884 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 885 | |
5cdc7ad3 | 886 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
887 | return 0; |
888 | } | |
889 | ||
4e4cbee9 | 890 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 891 | { |
2c30c71b | 892 | struct bio_vec *bvec; |
ce3ed71a | 893 | struct btrfs_root *root; |
2b070cfe | 894 | int ret = 0; |
6dc4f100 | 895 | struct bvec_iter_all iter_all; |
ce3ed71a | 896 | |
c09abff8 | 897 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 898 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 899 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 900 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
901 | if (ret) |
902 | break; | |
ce3ed71a | 903 | } |
2c30c71b | 904 | |
4e4cbee9 | 905 | return errno_to_blk_status(ret); |
ce3ed71a CM |
906 | } |
907 | ||
8896a08d | 908 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 909 | u64 dio_file_offset) |
22c59948 | 910 | { |
8b712842 CM |
911 | /* |
912 | * when we're called for a write, we're already in the async | |
5443be45 | 913 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 914 | */ |
79787eaa | 915 | return btree_csum_one_bio(bio); |
4a69a410 | 916 | } |
22c59948 | 917 | |
f4dcfb30 | 918 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 919 | struct btrfs_inode *bi) |
de0022b9 | 920 | { |
4eef29ef | 921 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 922 | return false; |
6300463b | 923 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 924 | return false; |
9b4e675a | 925 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
926 | return false; |
927 | return true; | |
de0022b9 JB |
928 | } |
929 | ||
1b36294a NB |
930 | blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, |
931 | int mirror_num, unsigned long bio_flags) | |
44b8bd7e | 932 | { |
0b246afa | 933 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4e4cbee9 | 934 | blk_status_t ret; |
cad321ad | 935 | |
cfe94440 | 936 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
4a69a410 CM |
937 | /* |
938 | * called for a read, do the setup so that checksum validation | |
939 | * can happen in the async kernel threads | |
940 | */ | |
0b246afa JM |
941 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
942 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 943 | if (ret) |
61891923 | 944 | goto out_w_error; |
08635bae | 945 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
f4dcfb30 | 946 | } else if (!should_async_write(fs_info, BTRFS_I(inode))) { |
de0022b9 JB |
947 | ret = btree_csum_one_bio(bio); |
948 | if (ret) | |
61891923 | 949 | goto out_w_error; |
08635bae | 950 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
61891923 SB |
951 | } else { |
952 | /* | |
953 | * kthread helpers are used to submit writes so that | |
954 | * checksumming can happen in parallel across all CPUs | |
955 | */ | |
8896a08d QW |
956 | ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, |
957 | 0, btree_submit_bio_start); | |
44b8bd7e | 958 | } |
d313d7a3 | 959 | |
4246a0b6 CH |
960 | if (ret) |
961 | goto out_w_error; | |
962 | return 0; | |
963 | ||
61891923 | 964 | out_w_error: |
4e4cbee9 | 965 | bio->bi_status = ret; |
4246a0b6 | 966 | bio_endio(bio); |
61891923 | 967 | return ret; |
44b8bd7e CM |
968 | } |
969 | ||
3dd1462e | 970 | #ifdef CONFIG_MIGRATION |
784b4e29 | 971 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
972 | struct page *newpage, struct page *page, |
973 | enum migrate_mode mode) | |
784b4e29 CM |
974 | { |
975 | /* | |
976 | * we can't safely write a btree page from here, | |
977 | * we haven't done the locking hook | |
978 | */ | |
979 | if (PageDirty(page)) | |
980 | return -EAGAIN; | |
981 | /* | |
982 | * Buffers may be managed in a filesystem specific way. | |
983 | * We must have no buffers or drop them. | |
984 | */ | |
985 | if (page_has_private(page) && | |
986 | !try_to_release_page(page, GFP_KERNEL)) | |
987 | return -EAGAIN; | |
a6bc32b8 | 988 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 989 | } |
3dd1462e | 990 | #endif |
784b4e29 | 991 | |
0da5468f CM |
992 | |
993 | static int btree_writepages(struct address_space *mapping, | |
994 | struct writeback_control *wbc) | |
995 | { | |
e2d84521 MX |
996 | struct btrfs_fs_info *fs_info; |
997 | int ret; | |
998 | ||
d8d5f3e1 | 999 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1000 | |
1001 | if (wbc->for_kupdate) | |
1002 | return 0; | |
1003 | ||
e2d84521 | 1004 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1005 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
1006 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1007 | BTRFS_DIRTY_METADATA_THRESH, | |
1008 | fs_info->dirty_metadata_batch); | |
e2d84521 | 1009 | if (ret < 0) |
793955bc | 1010 | return 0; |
793955bc | 1011 | } |
0b32f4bb | 1012 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1013 | } |
1014 | ||
70dec807 | 1015 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1016 | { |
98509cfc | 1017 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1018 | return 0; |
0c4e538b | 1019 | |
f7a52a40 | 1020 | return try_release_extent_buffer(page); |
d98237b3 CM |
1021 | } |
1022 | ||
d47992f8 LC |
1023 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1024 | unsigned int length) | |
d98237b3 | 1025 | { |
d1310b2e CM |
1026 | struct extent_io_tree *tree; |
1027 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1028 | extent_invalidatepage(tree, page, offset); |
1029 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1030 | if (PagePrivate(page)) { |
efe120a0 FH |
1031 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1032 | "page private not zero on page %llu", | |
1033 | (unsigned long long)page_offset(page)); | |
d1b89bc0 | 1034 | detach_page_private(page); |
9ad6b7bc | 1035 | } |
d98237b3 CM |
1036 | } |
1037 | ||
0b32f4bb JB |
1038 | static int btree_set_page_dirty(struct page *page) |
1039 | { | |
bb146eb2 | 1040 | #ifdef DEBUG |
139e8cd3 QW |
1041 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); |
1042 | struct btrfs_subpage *subpage; | |
0b32f4bb | 1043 | struct extent_buffer *eb; |
139e8cd3 QW |
1044 | int cur_bit = 0; |
1045 | u64 page_start = page_offset(page); | |
1046 | ||
1047 | if (fs_info->sectorsize == PAGE_SIZE) { | |
1048 | BUG_ON(!PagePrivate(page)); | |
1049 | eb = (struct extent_buffer *)page->private; | |
1050 | BUG_ON(!eb); | |
1051 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1052 | BUG_ON(!atomic_read(&eb->refs)); | |
1053 | btrfs_assert_tree_locked(eb); | |
1054 | return __set_page_dirty_nobuffers(page); | |
1055 | } | |
1056 | ASSERT(PagePrivate(page) && page->private); | |
1057 | subpage = (struct btrfs_subpage *)page->private; | |
1058 | ||
1059 | ASSERT(subpage->dirty_bitmap); | |
1060 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
1061 | unsigned long flags; | |
1062 | u64 cur; | |
1063 | u16 tmp = (1 << cur_bit); | |
1064 | ||
1065 | spin_lock_irqsave(&subpage->lock, flags); | |
1066 | if (!(tmp & subpage->dirty_bitmap)) { | |
1067 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1068 | cur_bit++; | |
1069 | continue; | |
1070 | } | |
1071 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1072 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 1073 | |
139e8cd3 QW |
1074 | eb = find_extent_buffer(fs_info, cur); |
1075 | ASSERT(eb); | |
1076 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1077 | ASSERT(atomic_read(&eb->refs)); | |
1078 | btrfs_assert_tree_locked(eb); | |
1079 | free_extent_buffer(eb); | |
1080 | ||
1081 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
1082 | } | |
bb146eb2 | 1083 | #endif |
0b32f4bb JB |
1084 | return __set_page_dirty_nobuffers(page); |
1085 | } | |
1086 | ||
7f09410b | 1087 | static const struct address_space_operations btree_aops = { |
0da5468f | 1088 | .writepages = btree_writepages, |
5f39d397 CM |
1089 | .releasepage = btree_releasepage, |
1090 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1091 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1092 | .migratepage = btree_migratepage, |
5a92bc88 | 1093 | #endif |
0b32f4bb | 1094 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1095 | }; |
1096 | ||
2ff7e61e JM |
1097 | struct extent_buffer *btrfs_find_create_tree_block( |
1098 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
1099 | u64 bytenr, u64 owner_root, |
1100 | int level) | |
0999df54 | 1101 | { |
0b246afa JM |
1102 | if (btrfs_is_testing(fs_info)) |
1103 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 1104 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
1105 | } |
1106 | ||
581c1760 QW |
1107 | /* |
1108 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1109 | * verification. | |
1110 | * | |
1b7ec85e | 1111 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
1112 | * @parent_transid: expected transid of this tree block, skip check if 0 |
1113 | * @level: expected level, mandatory check | |
1114 | * @first_key: expected key in slot 0, skip check if NULL | |
1115 | */ | |
2ff7e61e | 1116 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
1117 | u64 owner_root, u64 parent_transid, |
1118 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
1119 | { |
1120 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1121 | int ret; |
1122 | ||
3fbaf258 | 1123 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
1124 | if (IS_ERR(buf)) |
1125 | return buf; | |
0999df54 | 1126 | |
5ab12d1f | 1127 | ret = btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 1128 | level, first_key); |
0f0fe8f7 | 1129 | if (ret) { |
537f38f0 | 1130 | free_extent_buffer_stale(buf); |
64c043de | 1131 | return ERR_PTR(ret); |
0f0fe8f7 | 1132 | } |
5f39d397 | 1133 | return buf; |
ce9adaa5 | 1134 | |
eb60ceac CM |
1135 | } |
1136 | ||
6a884d7d | 1137 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1138 | { |
6a884d7d | 1139 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1140 | if (btrfs_header_generation(buf) == |
e2d84521 | 1141 | fs_info->running_transaction->transid) { |
b9447ef8 | 1142 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1143 | |
b9473439 | 1144 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1145 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1146 | -buf->len, | |
1147 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1148 | clear_extent_buffer_dirty(buf); |
1149 | } | |
925baedd | 1150 | } |
5f39d397 CM |
1151 | } |
1152 | ||
da17066c | 1153 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1154 | u64 objectid) |
d97e63b6 | 1155 | { |
7c0260ee | 1156 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
96dfcb46 | 1157 | root->fs_info = fs_info; |
cfaa7295 | 1158 | root->node = NULL; |
a28ec197 | 1159 | root->commit_root = NULL; |
27cdeb70 | 1160 | root->state = 0; |
d68fc57b | 1161 | root->orphan_cleanup_state = 0; |
0b86a832 | 1162 | |
0f7d52f4 | 1163 | root->last_trans = 0; |
6b8fad57 | 1164 | root->free_objectid = 0; |
eb73c1b7 | 1165 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1166 | root->nr_ordered_extents = 0; |
6bef4d31 | 1167 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1168 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1169 | root->block_rsv = NULL; |
0b86a832 CM |
1170 | |
1171 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1172 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1173 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1174 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1175 | INIT_LIST_HEAD(&root->ordered_extents); |
1176 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1177 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1178 | INIT_LIST_HEAD(&root->logged_list[0]); |
1179 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1180 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1181 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1182 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1183 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1184 | spin_lock_init(&root->log_extents_lock[0]); |
1185 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1186 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1187 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1188 | mutex_init(&root->log_mutex); |
31f3d255 | 1189 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1190 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1191 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1192 | init_waitqueue_head(&root->log_writer_wait); |
1193 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1194 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1195 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1196 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1197 | atomic_set(&root->log_commit[0], 0); |
1198 | atomic_set(&root->log_commit[1], 0); | |
1199 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1200 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1201 | refcount_set(&root->refs, 1); |
8ecebf4d | 1202 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1203 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1204 | root->log_transid = 0; |
d1433deb | 1205 | root->log_transid_committed = -1; |
257c62e1 | 1206 | root->last_log_commit = 0; |
e289f03e | 1207 | if (!dummy) { |
43eb5f29 QW |
1208 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1209 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1210 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1211 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1212 | } | |
017e5369 | 1213 | |
3768f368 CM |
1214 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1215 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1216 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
4d775673 | 1217 | root->root_key.objectid = objectid; |
0ee5dc67 | 1218 | root->anon_dev = 0; |
8ea05e3a | 1219 | |
5f3ab90a | 1220 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1221 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1222 | #ifdef CONFIG_BTRFS_DEBUG |
1223 | INIT_LIST_HEAD(&root->leak_list); | |
1224 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1225 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); | |
1226 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1227 | #endif | |
3768f368 CM |
1228 | } |
1229 | ||
74e4d827 | 1230 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1231 | u64 objectid, gfp_t flags) |
6f07e42e | 1232 | { |
74e4d827 | 1233 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1234 | if (root) |
96dfcb46 | 1235 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1236 | return root; |
1237 | } | |
1238 | ||
06ea65a3 JB |
1239 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1240 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1241 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1242 | { |
1243 | struct btrfs_root *root; | |
1244 | ||
7c0260ee JM |
1245 | if (!fs_info) |
1246 | return ERR_PTR(-EINVAL); | |
1247 | ||
96dfcb46 | 1248 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1249 | if (!root) |
1250 | return ERR_PTR(-ENOMEM); | |
da17066c | 1251 | |
b9ef22de | 1252 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1253 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1254 | |
1255 | return root; | |
1256 | } | |
1257 | #endif | |
1258 | ||
20897f5c | 1259 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1260 | u64 objectid) |
1261 | { | |
9b7a2440 | 1262 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1263 | struct extent_buffer *leaf; |
1264 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1265 | struct btrfs_root *root; | |
1266 | struct btrfs_key key; | |
b89f6d1f | 1267 | unsigned int nofs_flag; |
20897f5c | 1268 | int ret = 0; |
20897f5c | 1269 | |
b89f6d1f FM |
1270 | /* |
1271 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1272 | * context to avoid deadlock if reclaim happens. | |
1273 | */ | |
1274 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1275 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1276 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1277 | if (!root) |
1278 | return ERR_PTR(-ENOMEM); | |
1279 | ||
20897f5c AJ |
1280 | root->root_key.objectid = objectid; |
1281 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1282 | root->root_key.offset = 0; | |
1283 | ||
9631e4cc JB |
1284 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1285 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1286 | if (IS_ERR(leaf)) { |
1287 | ret = PTR_ERR(leaf); | |
1dd05682 | 1288 | leaf = NULL; |
8a6a87cd | 1289 | goto fail_unlock; |
20897f5c AJ |
1290 | } |
1291 | ||
20897f5c | 1292 | root->node = leaf; |
20897f5c AJ |
1293 | btrfs_mark_buffer_dirty(leaf); |
1294 | ||
1295 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1296 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1297 | |
f944d2cb DS |
1298 | btrfs_set_root_flags(&root->root_item, 0); |
1299 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1300 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1301 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1302 | btrfs_set_root_level(&root->root_item, 0); | |
1303 | btrfs_set_root_refs(&root->root_item, 1); | |
1304 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1305 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1306 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1307 | if (is_fstree(objectid)) |
807fc790 AS |
1308 | generate_random_guid(root->root_item.uuid); |
1309 | else | |
1310 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1311 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1312 | |
8a6a87cd BB |
1313 | btrfs_tree_unlock(leaf); |
1314 | ||
20897f5c AJ |
1315 | key.objectid = objectid; |
1316 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1317 | key.offset = 0; | |
1318 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1319 | if (ret) | |
1320 | goto fail; | |
1321 | ||
1dd05682 TI |
1322 | return root; |
1323 | ||
8a6a87cd | 1324 | fail_unlock: |
8c38938c | 1325 | if (leaf) |
1dd05682 | 1326 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1327 | fail: |
00246528 | 1328 | btrfs_put_root(root); |
20897f5c | 1329 | |
1dd05682 | 1330 | return ERR_PTR(ret); |
20897f5c AJ |
1331 | } |
1332 | ||
7237f183 YZ |
1333 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1334 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1335 | { |
1336 | struct btrfs_root *root; | |
e02119d5 | 1337 | |
96dfcb46 | 1338 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1339 | if (!root) |
7237f183 | 1340 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1341 | |
e02119d5 CM |
1342 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1343 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1344 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1345 | |
6ab6ebb7 NA |
1346 | return root; |
1347 | } | |
1348 | ||
1349 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1350 | struct btrfs_root *root) | |
1351 | { | |
1352 | struct extent_buffer *leaf; | |
1353 | ||
7237f183 | 1354 | /* |
92a7cc42 | 1355 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1356 | * |
92a7cc42 QW |
1357 | * Log trees are not exposed to user space thus can't be snapshotted, |
1358 | * and they go away before a real commit is actually done. | |
1359 | * | |
1360 | * They do store pointers to file data extents, and those reference | |
1361 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1362 | */ |
e02119d5 | 1363 | |
4d75f8a9 | 1364 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1365 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1366 | if (IS_ERR(leaf)) |
1367 | return PTR_ERR(leaf); | |
e02119d5 | 1368 | |
7237f183 | 1369 | root->node = leaf; |
e02119d5 | 1370 | |
e02119d5 CM |
1371 | btrfs_mark_buffer_dirty(root->node); |
1372 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1373 | |
1374 | return 0; | |
7237f183 YZ |
1375 | } |
1376 | ||
1377 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1378 | struct btrfs_fs_info *fs_info) | |
1379 | { | |
1380 | struct btrfs_root *log_root; | |
1381 | ||
1382 | log_root = alloc_log_tree(trans, fs_info); | |
1383 | if (IS_ERR(log_root)) | |
1384 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1385 | |
3ddebf27 NA |
1386 | if (!btrfs_is_zoned(fs_info)) { |
1387 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1388 | ||
1389 | if (ret) { | |
1390 | btrfs_put_root(log_root); | |
1391 | return ret; | |
1392 | } | |
6ab6ebb7 NA |
1393 | } |
1394 | ||
7237f183 YZ |
1395 | WARN_ON(fs_info->log_root_tree); |
1396 | fs_info->log_root_tree = log_root; | |
1397 | return 0; | |
1398 | } | |
1399 | ||
1400 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1401 | struct btrfs_root *root) | |
1402 | { | |
0b246afa | 1403 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1404 | struct btrfs_root *log_root; |
1405 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1406 | int ret; |
7237f183 | 1407 | |
0b246afa | 1408 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1409 | if (IS_ERR(log_root)) |
1410 | return PTR_ERR(log_root); | |
1411 | ||
6ab6ebb7 NA |
1412 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1413 | if (ret) { | |
1414 | btrfs_put_root(log_root); | |
1415 | return ret; | |
1416 | } | |
1417 | ||
7237f183 YZ |
1418 | log_root->last_trans = trans->transid; |
1419 | log_root->root_key.offset = root->root_key.objectid; | |
1420 | ||
1421 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1422 | btrfs_set_stack_inode_generation(inode_item, 1); |
1423 | btrfs_set_stack_inode_size(inode_item, 3); | |
1424 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1425 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1426 | fs_info->nodesize); |
3cae210f | 1427 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1428 | |
5d4f98a2 | 1429 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1430 | |
1431 | WARN_ON(root->log_root); | |
1432 | root->log_root = log_root; | |
1433 | root->log_transid = 0; | |
d1433deb | 1434 | root->log_transid_committed = -1; |
257c62e1 | 1435 | root->last_log_commit = 0; |
e02119d5 CM |
1436 | return 0; |
1437 | } | |
1438 | ||
49d11bea JB |
1439 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1440 | struct btrfs_path *path, | |
1441 | struct btrfs_key *key) | |
e02119d5 CM |
1442 | { |
1443 | struct btrfs_root *root; | |
1444 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1445 | u64 generation; |
cb517eab | 1446 | int ret; |
581c1760 | 1447 | int level; |
0f7d52f4 | 1448 | |
96dfcb46 | 1449 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1450 | if (!root) |
1451 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1452 | |
cb517eab MX |
1453 | ret = btrfs_find_root(tree_root, key, path, |
1454 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1455 | if (ret) { |
13a8a7c8 YZ |
1456 | if (ret > 0) |
1457 | ret = -ENOENT; | |
49d11bea | 1458 | goto fail; |
0f7d52f4 | 1459 | } |
13a8a7c8 | 1460 | |
84234f3a | 1461 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1462 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1463 | root->node = read_tree_block(fs_info, |
1464 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1465 | key->objectid, generation, level, NULL); |
64c043de LB |
1466 | if (IS_ERR(root->node)) { |
1467 | ret = PTR_ERR(root->node); | |
8c38938c | 1468 | root->node = NULL; |
49d11bea | 1469 | goto fail; |
cb517eab MX |
1470 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { |
1471 | ret = -EIO; | |
49d11bea | 1472 | goto fail; |
416bc658 | 1473 | } |
5d4f98a2 | 1474 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1475 | return root; |
49d11bea | 1476 | fail: |
00246528 | 1477 | btrfs_put_root(root); |
49d11bea JB |
1478 | return ERR_PTR(ret); |
1479 | } | |
1480 | ||
1481 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1482 | struct btrfs_key *key) | |
1483 | { | |
1484 | struct btrfs_root *root; | |
1485 | struct btrfs_path *path; | |
1486 | ||
1487 | path = btrfs_alloc_path(); | |
1488 | if (!path) | |
1489 | return ERR_PTR(-ENOMEM); | |
1490 | root = read_tree_root_path(tree_root, path, key); | |
1491 | btrfs_free_path(path); | |
1492 | ||
1493 | return root; | |
cb517eab MX |
1494 | } |
1495 | ||
2dfb1e43 QW |
1496 | /* |
1497 | * Initialize subvolume root in-memory structure | |
1498 | * | |
1499 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1500 | */ | |
1501 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1502 | { |
1503 | int ret; | |
dcc3eb96 | 1504 | unsigned int nofs_flag; |
cb517eab | 1505 | |
dcc3eb96 NB |
1506 | /* |
1507 | * We might be called under a transaction (e.g. indirect backref | |
1508 | * resolution) which could deadlock if it triggers memory reclaim | |
1509 | */ | |
1510 | nofs_flag = memalloc_nofs_save(); | |
1511 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1512 | memalloc_nofs_restore(nofs_flag); | |
1513 | if (ret) | |
8257b2dc | 1514 | goto fail; |
8257b2dc | 1515 | |
aeb935a4 | 1516 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
b8f762a2 | 1517 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1518 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1519 | btrfs_check_and_init_root_item(&root->root_item); |
1520 | } | |
1521 | ||
851fd730 QW |
1522 | /* |
1523 | * Don't assign anonymous block device to roots that are not exposed to | |
1524 | * userspace, the id pool is limited to 1M | |
1525 | */ | |
1526 | if (is_fstree(root->root_key.objectid) && | |
1527 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1528 | if (!anon_dev) { |
1529 | ret = get_anon_bdev(&root->anon_dev); | |
1530 | if (ret) | |
1531 | goto fail; | |
1532 | } else { | |
1533 | root->anon_dev = anon_dev; | |
1534 | } | |
851fd730 | 1535 | } |
f32e48e9 CR |
1536 | |
1537 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1538 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1539 | if (ret) { |
1540 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1541 | goto fail; |
f32e48e9 CR |
1542 | } |
1543 | ||
6b8fad57 | 1544 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1545 | |
1546 | mutex_unlock(&root->objectid_mutex); | |
1547 | ||
cb517eab MX |
1548 | return 0; |
1549 | fail: | |
84db5ccf | 1550 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1551 | return ret; |
1552 | } | |
1553 | ||
a98db0f3 JB |
1554 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1555 | u64 root_id) | |
cb517eab MX |
1556 | { |
1557 | struct btrfs_root *root; | |
1558 | ||
1559 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1560 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1561 | (unsigned long)root_id); | |
bc44d7c4 | 1562 | if (root) |
00246528 | 1563 | root = btrfs_grab_root(root); |
cb517eab MX |
1564 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1565 | return root; | |
1566 | } | |
1567 | ||
49d11bea JB |
1568 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1569 | u64 objectid) | |
1570 | { | |
1571 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1572 | return btrfs_grab_root(fs_info->tree_root); | |
1573 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1574 | return btrfs_grab_root(fs_info->extent_root); | |
1575 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) | |
1576 | return btrfs_grab_root(fs_info->chunk_root); | |
1577 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1578 | return btrfs_grab_root(fs_info->dev_root); | |
1579 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
1580 | return btrfs_grab_root(fs_info->csum_root); | |
1581 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) | |
1582 | return btrfs_grab_root(fs_info->quota_root) ? | |
1583 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1584 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1585 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1586 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
1587 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) | |
1588 | return btrfs_grab_root(fs_info->free_space_root) ? | |
1589 | fs_info->free_space_root : ERR_PTR(-ENOENT); | |
1590 | return NULL; | |
1591 | } | |
1592 | ||
cb517eab MX |
1593 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1594 | struct btrfs_root *root) | |
1595 | { | |
1596 | int ret; | |
1597 | ||
e1860a77 | 1598 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1599 | if (ret) |
1600 | return ret; | |
1601 | ||
1602 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1603 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1604 | (unsigned long)root->root_key.objectid, | |
1605 | root); | |
af01d2e5 | 1606 | if (ret == 0) { |
00246528 | 1607 | btrfs_grab_root(root); |
27cdeb70 | 1608 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1609 | } |
cb517eab MX |
1610 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1611 | radix_tree_preload_end(); | |
1612 | ||
1613 | return ret; | |
1614 | } | |
1615 | ||
bd647ce3 JB |
1616 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1617 | { | |
1618 | #ifdef CONFIG_BTRFS_DEBUG | |
1619 | struct btrfs_root *root; | |
1620 | ||
1621 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1622 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1623 | ||
bd647ce3 JB |
1624 | root = list_first_entry(&fs_info->allocated_roots, |
1625 | struct btrfs_root, leak_list); | |
457f1864 | 1626 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1627 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1628 | refcount_read(&root->refs)); |
1629 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1630 | btrfs_put_root(root); |
1631 | btrfs_put_root(root); | |
bd647ce3 JB |
1632 | } |
1633 | #endif | |
1634 | } | |
1635 | ||
0d4b0463 JB |
1636 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1637 | { | |
141386e1 JB |
1638 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1639 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1640 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1641 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1642 | btrfs_free_csum_hash(fs_info); | |
1643 | btrfs_free_stripe_hash_table(fs_info); | |
1644 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1645 | kfree(fs_info->balance_ctl); |
1646 | kfree(fs_info->delayed_root); | |
00246528 JB |
1647 | btrfs_put_root(fs_info->extent_root); |
1648 | btrfs_put_root(fs_info->tree_root); | |
1649 | btrfs_put_root(fs_info->chunk_root); | |
1650 | btrfs_put_root(fs_info->dev_root); | |
1651 | btrfs_put_root(fs_info->csum_root); | |
1652 | btrfs_put_root(fs_info->quota_root); | |
1653 | btrfs_put_root(fs_info->uuid_root); | |
1654 | btrfs_put_root(fs_info->free_space_root); | |
1655 | btrfs_put_root(fs_info->fs_root); | |
aeb935a4 | 1656 | btrfs_put_root(fs_info->data_reloc_root); |
bd647ce3 | 1657 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1658 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1659 | kfree(fs_info->super_copy); |
1660 | kfree(fs_info->super_for_commit); | |
1661 | kvfree(fs_info); | |
1662 | } | |
1663 | ||
1664 | ||
2dfb1e43 QW |
1665 | /* |
1666 | * Get an in-memory reference of a root structure. | |
1667 | * | |
1668 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1669 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1670 | * found, then read it from disk and add it to cached fs roots. | |
1671 | * | |
1672 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1673 | * | |
1674 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1675 | * same root objectid. | |
1676 | * | |
1677 | * @objectid: root id | |
1678 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1679 | * pass 0 for new allocation. | |
1680 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1681 | * for orphan roots | |
1682 | */ | |
1683 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1684 | u64 objectid, dev_t anon_dev, | |
1685 | bool check_ref) | |
5eda7b5e CM |
1686 | { |
1687 | struct btrfs_root *root; | |
381cf658 | 1688 | struct btrfs_path *path; |
1d4c08e0 | 1689 | struct btrfs_key key; |
5eda7b5e CM |
1690 | int ret; |
1691 | ||
49d11bea JB |
1692 | root = btrfs_get_global_root(fs_info, objectid); |
1693 | if (root) | |
1694 | return root; | |
4df27c4d | 1695 | again: |
56e9357a | 1696 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1697 | if (root) { |
2dfb1e43 QW |
1698 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1699 | ASSERT(!anon_dev); | |
bc44d7c4 | 1700 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1701 | btrfs_put_root(root); |
48475471 | 1702 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1703 | } |
5eda7b5e | 1704 | return root; |
48475471 | 1705 | } |
5eda7b5e | 1706 | |
56e9357a DS |
1707 | key.objectid = objectid; |
1708 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1709 | key.offset = (u64)-1; | |
1710 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1711 | if (IS_ERR(root)) |
1712 | return root; | |
3394e160 | 1713 | |
c00869f1 | 1714 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1715 | ret = -ENOENT; |
581bb050 | 1716 | goto fail; |
35a30d7c | 1717 | } |
581bb050 | 1718 | |
2dfb1e43 | 1719 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1720 | if (ret) |
1721 | goto fail; | |
3394e160 | 1722 | |
381cf658 DS |
1723 | path = btrfs_alloc_path(); |
1724 | if (!path) { | |
1725 | ret = -ENOMEM; | |
1726 | goto fail; | |
1727 | } | |
1d4c08e0 DS |
1728 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1729 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1730 | key.offset = objectid; |
1d4c08e0 DS |
1731 | |
1732 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1733 | btrfs_free_path(path); |
d68fc57b YZ |
1734 | if (ret < 0) |
1735 | goto fail; | |
1736 | if (ret == 0) | |
27cdeb70 | 1737 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1738 | |
cb517eab | 1739 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1740 | if (ret) { |
dbf1575a JJB |
1741 | if (ret == -EEXIST) { |
1742 | btrfs_put_root(root); | |
4df27c4d | 1743 | goto again; |
dbf1575a | 1744 | } |
4df27c4d | 1745 | goto fail; |
0f7d52f4 | 1746 | } |
edbd8d4e | 1747 | return root; |
4df27c4d | 1748 | fail: |
e019bdc9 FM |
1749 | /* |
1750 | * If our caller provided us an anonymous device, then it's his | |
1751 | * responsability to free it in case we fail. So we have to set our | |
1752 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() | |
1753 | * and once again by our caller. | |
1754 | */ | |
1755 | if (anon_dev) | |
1756 | root->anon_dev = 0; | |
8c38938c | 1757 | btrfs_put_root(root); |
4df27c4d | 1758 | return ERR_PTR(ret); |
edbd8d4e CM |
1759 | } |
1760 | ||
2dfb1e43 QW |
1761 | /* |
1762 | * Get in-memory reference of a root structure | |
1763 | * | |
1764 | * @objectid: tree objectid | |
1765 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1766 | * one reference | |
1767 | */ | |
1768 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1769 | u64 objectid, bool check_ref) | |
1770 | { | |
1771 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1772 | } | |
1773 | ||
1774 | /* | |
1775 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1776 | * the anonymous block device id | |
1777 | * | |
1778 | * @objectid: tree objectid | |
1779 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1780 | * parameter value | |
1781 | */ | |
1782 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1783 | u64 objectid, dev_t anon_dev) | |
1784 | { | |
1785 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1786 | } | |
1787 | ||
49d11bea JB |
1788 | /* |
1789 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1790 | * @fs_info: the fs_info | |
1791 | * @objectid: the objectid we need to lookup | |
1792 | * | |
1793 | * This is exclusively used for backref walking, and exists specifically because | |
1794 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1795 | * creation time, which means we may have to read the tree_root in order to look | |
1796 | * up a fs root that is not in memory. If the root is not in memory we will | |
1797 | * read the tree root commit root and look up the fs root from there. This is a | |
1798 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1799 | * have the most uptodate information, it'll simply be discarded once the | |
1800 | * backref code is finished using the root. | |
1801 | */ | |
1802 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1803 | struct btrfs_path *path, | |
1804 | u64 objectid) | |
1805 | { | |
1806 | struct btrfs_root *root; | |
1807 | struct btrfs_key key; | |
1808 | ||
1809 | ASSERT(path->search_commit_root && path->skip_locking); | |
1810 | ||
1811 | /* | |
1812 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1813 | * since this is called via the backref walking code we won't be looking | |
1814 | * up a root that doesn't exist, unless there's corruption. So if root | |
1815 | * != NULL just return it. | |
1816 | */ | |
1817 | root = btrfs_get_global_root(fs_info, objectid); | |
1818 | if (root) | |
1819 | return root; | |
1820 | ||
1821 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1822 | if (root) | |
1823 | return root; | |
1824 | ||
1825 | key.objectid = objectid; | |
1826 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1827 | key.offset = (u64)-1; | |
1828 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1829 | btrfs_release_path(path); | |
1830 | ||
1831 | return root; | |
1832 | } | |
1833 | ||
8b712842 CM |
1834 | /* |
1835 | * called by the kthread helper functions to finally call the bio end_io | |
1836 | * functions. This is where read checksum verification actually happens | |
1837 | */ | |
1838 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1839 | { |
ce9adaa5 | 1840 | struct bio *bio; |
97eb6b69 | 1841 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1842 | |
97eb6b69 | 1843 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1844 | bio = end_io_wq->bio; |
ce9adaa5 | 1845 | |
4e4cbee9 | 1846 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1847 | bio->bi_private = end_io_wq->private; |
1848 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1849 | bio_endio(bio); |
9be490f1 | 1850 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1851 | } |
1852 | ||
a74a4b97 CM |
1853 | static int cleaner_kthread(void *arg) |
1854 | { | |
1855 | struct btrfs_root *root = arg; | |
0b246afa | 1856 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1857 | int again; |
a74a4b97 | 1858 | |
d6fd0ae2 | 1859 | while (1) { |
d0278245 | 1860 | again = 0; |
a74a4b97 | 1861 | |
fd340d0f JB |
1862 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1863 | ||
d0278245 | 1864 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1865 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1866 | goto sleep; |
1867 | ||
90c711ab ZB |
1868 | /* |
1869 | * Do not do anything if we might cause open_ctree() to block | |
1870 | * before we have finished mounting the filesystem. | |
1871 | */ | |
0b246afa | 1872 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1873 | goto sleep; |
1874 | ||
0b246afa | 1875 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1876 | goto sleep; |
1877 | ||
dc7f370c MX |
1878 | /* |
1879 | * Avoid the problem that we change the status of the fs | |
1880 | * during the above check and trylock. | |
1881 | */ | |
2ff7e61e | 1882 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1883 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1884 | goto sleep; |
76dda93c | 1885 | } |
a74a4b97 | 1886 | |
2ff7e61e | 1887 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1888 | |
d0278245 | 1889 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1890 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1891 | |
1892 | /* | |
05323cd1 MX |
1893 | * The defragger has dealt with the R/O remount and umount, |
1894 | * needn't do anything special here. | |
d0278245 | 1895 | */ |
0b246afa | 1896 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1897 | |
1898 | /* | |
f3372065 | 1899 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1900 | * with relocation (btrfs_relocate_chunk) and relocation |
1901 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1902 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1903 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1904 | * unused block groups. | |
1905 | */ | |
0b246afa | 1906 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1907 | |
1908 | /* | |
1909 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1910 | * all unused block_groups. This possibly gives us some more free | |
1911 | * space. | |
1912 | */ | |
1913 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1914 | sleep: |
a0a1db70 | 1915 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1916 | if (kthread_should_park()) |
1917 | kthread_parkme(); | |
1918 | if (kthread_should_stop()) | |
1919 | return 0; | |
838fe188 | 1920 | if (!again) { |
a74a4b97 | 1921 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1922 | schedule(); |
a74a4b97 CM |
1923 | __set_current_state(TASK_RUNNING); |
1924 | } | |
da288d28 | 1925 | } |
a74a4b97 CM |
1926 | } |
1927 | ||
1928 | static int transaction_kthread(void *arg) | |
1929 | { | |
1930 | struct btrfs_root *root = arg; | |
0b246afa | 1931 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1932 | struct btrfs_trans_handle *trans; |
1933 | struct btrfs_transaction *cur; | |
8929ecfa | 1934 | u64 transid; |
643900be | 1935 | time64_t delta; |
a74a4b97 | 1936 | unsigned long delay; |
914b2007 | 1937 | bool cannot_commit; |
a74a4b97 CM |
1938 | |
1939 | do { | |
914b2007 | 1940 | cannot_commit = false; |
ba1bc00f | 1941 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1942 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1943 | |
0b246afa JM |
1944 | spin_lock(&fs_info->trans_lock); |
1945 | cur = fs_info->running_transaction; | |
a74a4b97 | 1946 | if (!cur) { |
0b246afa | 1947 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1948 | goto sleep; |
1949 | } | |
31153d81 | 1950 | |
643900be | 1951 | delta = ktime_get_seconds() - cur->start_time; |
3296bf56 | 1952 | if (cur->state < TRANS_STATE_COMMIT_START && |
643900be | 1953 | delta < fs_info->commit_interval) { |
0b246afa | 1954 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1955 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1956 | delay = min(delay, | |
1957 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1958 | goto sleep; |
1959 | } | |
8929ecfa | 1960 | transid = cur->transid; |
0b246afa | 1961 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1962 | |
79787eaa | 1963 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1964 | trans = btrfs_attach_transaction(root); |
914b2007 | 1965 | if (IS_ERR(trans)) { |
354aa0fb MX |
1966 | if (PTR_ERR(trans) != -ENOENT) |
1967 | cannot_commit = true; | |
79787eaa | 1968 | goto sleep; |
914b2007 | 1969 | } |
8929ecfa | 1970 | if (transid == trans->transid) { |
3a45bb20 | 1971 | btrfs_commit_transaction(trans); |
8929ecfa | 1972 | } else { |
3a45bb20 | 1973 | btrfs_end_transaction(trans); |
8929ecfa | 1974 | } |
a74a4b97 | 1975 | sleep: |
0b246afa JM |
1976 | wake_up_process(fs_info->cleaner_kthread); |
1977 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1978 | |
4e121c06 | 1979 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1980 | &fs_info->fs_state))) |
2ff7e61e | 1981 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1982 | if (!kthread_should_stop() && |
0b246afa | 1983 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1984 | cannot_commit)) |
bc5511d0 | 1985 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1986 | } while (!kthread_should_stop()); |
1987 | return 0; | |
1988 | } | |
1989 | ||
af31f5e5 | 1990 | /* |
01f0f9da NB |
1991 | * This will find the highest generation in the array of root backups. The |
1992 | * index of the highest array is returned, or -EINVAL if we can't find | |
1993 | * anything. | |
af31f5e5 CM |
1994 | * |
1995 | * We check to make sure the array is valid by comparing the | |
1996 | * generation of the latest root in the array with the generation | |
1997 | * in the super block. If they don't match we pitch it. | |
1998 | */ | |
01f0f9da | 1999 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 2000 | { |
01f0f9da | 2001 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 2002 | u64 cur; |
af31f5e5 CM |
2003 | struct btrfs_root_backup *root_backup; |
2004 | int i; | |
2005 | ||
2006 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2007 | root_backup = info->super_copy->super_roots + i; | |
2008 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2009 | if (cur == newest_gen) | |
01f0f9da | 2010 | return i; |
af31f5e5 CM |
2011 | } |
2012 | ||
01f0f9da | 2013 | return -EINVAL; |
af31f5e5 CM |
2014 | } |
2015 | ||
af31f5e5 CM |
2016 | /* |
2017 | * copy all the root pointers into the super backup array. | |
2018 | * this will bump the backup pointer by one when it is | |
2019 | * done | |
2020 | */ | |
2021 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2022 | { | |
6ef108dd | 2023 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2024 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
2025 | |
2026 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2027 | ||
2028 | /* | |
2029 | * make sure all of our padding and empty slots get zero filled | |
2030 | * regardless of which ones we use today | |
2031 | */ | |
2032 | memset(root_backup, 0, sizeof(*root_backup)); | |
2033 | ||
2034 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2035 | ||
2036 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2037 | btrfs_set_backup_tree_root_gen(root_backup, | |
2038 | btrfs_header_generation(info->tree_root->node)); | |
2039 | ||
2040 | btrfs_set_backup_tree_root_level(root_backup, | |
2041 | btrfs_header_level(info->tree_root->node)); | |
2042 | ||
2043 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2044 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2045 | btrfs_header_generation(info->chunk_root->node)); | |
2046 | btrfs_set_backup_chunk_root_level(root_backup, | |
2047 | btrfs_header_level(info->chunk_root->node)); | |
2048 | ||
2049 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2050 | btrfs_set_backup_extent_root_gen(root_backup, | |
2051 | btrfs_header_generation(info->extent_root->node)); | |
2052 | btrfs_set_backup_extent_root_level(root_backup, | |
2053 | btrfs_header_level(info->extent_root->node)); | |
2054 | ||
7c7e82a7 CM |
2055 | /* |
2056 | * we might commit during log recovery, which happens before we set | |
2057 | * the fs_root. Make sure it is valid before we fill it in. | |
2058 | */ | |
2059 | if (info->fs_root && info->fs_root->node) { | |
2060 | btrfs_set_backup_fs_root(root_backup, | |
2061 | info->fs_root->node->start); | |
2062 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2063 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2064 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2065 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2066 | } |
af31f5e5 CM |
2067 | |
2068 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2069 | btrfs_set_backup_dev_root_gen(root_backup, | |
2070 | btrfs_header_generation(info->dev_root->node)); | |
2071 | btrfs_set_backup_dev_root_level(root_backup, | |
2072 | btrfs_header_level(info->dev_root->node)); | |
2073 | ||
2074 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2075 | btrfs_set_backup_csum_root_gen(root_backup, | |
2076 | btrfs_header_generation(info->csum_root->node)); | |
2077 | btrfs_set_backup_csum_root_level(root_backup, | |
2078 | btrfs_header_level(info->csum_root->node)); | |
2079 | ||
2080 | btrfs_set_backup_total_bytes(root_backup, | |
2081 | btrfs_super_total_bytes(info->super_copy)); | |
2082 | btrfs_set_backup_bytes_used(root_backup, | |
2083 | btrfs_super_bytes_used(info->super_copy)); | |
2084 | btrfs_set_backup_num_devices(root_backup, | |
2085 | btrfs_super_num_devices(info->super_copy)); | |
2086 | ||
2087 | /* | |
2088 | * if we don't copy this out to the super_copy, it won't get remembered | |
2089 | * for the next commit | |
2090 | */ | |
2091 | memcpy(&info->super_copy->super_roots, | |
2092 | &info->super_for_commit->super_roots, | |
2093 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2094 | } | |
2095 | ||
bd2336b2 NB |
2096 | /* |
2097 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2098 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2099 | * | |
2100 | * fs_info - filesystem whose backup roots need to be read | |
2101 | * priority - priority of backup root required | |
2102 | * | |
2103 | * Returns backup root index on success and -EINVAL otherwise. | |
2104 | */ | |
2105 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2106 | { | |
2107 | int backup_index = find_newest_super_backup(fs_info); | |
2108 | struct btrfs_super_block *super = fs_info->super_copy; | |
2109 | struct btrfs_root_backup *root_backup; | |
2110 | ||
2111 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2112 | if (priority == 0) | |
2113 | return backup_index; | |
2114 | ||
2115 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2116 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2117 | } else { | |
2118 | return -EINVAL; | |
2119 | } | |
2120 | ||
2121 | root_backup = super->super_roots + backup_index; | |
2122 | ||
2123 | btrfs_set_super_generation(super, | |
2124 | btrfs_backup_tree_root_gen(root_backup)); | |
2125 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2126 | btrfs_set_super_root_level(super, | |
2127 | btrfs_backup_tree_root_level(root_backup)); | |
2128 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2129 | ||
2130 | /* | |
2131 | * Fixme: the total bytes and num_devices need to match or we should | |
2132 | * need a fsck | |
2133 | */ | |
2134 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2135 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2136 | ||
2137 | return backup_index; | |
2138 | } | |
2139 | ||
7abadb64 LB |
2140 | /* helper to cleanup workers */ |
2141 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2142 | { | |
dc6e3209 | 2143 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2144 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2145 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2146 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2147 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
d05a33ac | 2148 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2149 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2150 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2151 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2152 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2153 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2154 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2155 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2156 | if (fs_info->discard_ctl.discard_workers) |
2157 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2158 | /* |
2159 | * Now that all other work queues are destroyed, we can safely destroy | |
2160 | * the queues used for metadata I/O, since tasks from those other work | |
2161 | * queues can do metadata I/O operations. | |
2162 | */ | |
2163 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2164 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2165 | } |
2166 | ||
2e9f5954 R |
2167 | static void free_root_extent_buffers(struct btrfs_root *root) |
2168 | { | |
2169 | if (root) { | |
2170 | free_extent_buffer(root->node); | |
2171 | free_extent_buffer(root->commit_root); | |
2172 | root->node = NULL; | |
2173 | root->commit_root = NULL; | |
2174 | } | |
2175 | } | |
2176 | ||
af31f5e5 | 2177 | /* helper to cleanup tree roots */ |
4273eaff | 2178 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2179 | { |
2e9f5954 | 2180 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2181 | |
2e9f5954 R |
2182 | free_root_extent_buffers(info->dev_root); |
2183 | free_root_extent_buffers(info->extent_root); | |
2184 | free_root_extent_buffers(info->csum_root); | |
2185 | free_root_extent_buffers(info->quota_root); | |
2186 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2187 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2188 | free_root_extent_buffers(info->data_reloc_root); |
4273eaff | 2189 | if (free_chunk_root) |
2e9f5954 | 2190 | free_root_extent_buffers(info->chunk_root); |
70f6d82e | 2191 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2192 | } |
2193 | ||
8c38938c JB |
2194 | void btrfs_put_root(struct btrfs_root *root) |
2195 | { | |
2196 | if (!root) | |
2197 | return; | |
2198 | ||
2199 | if (refcount_dec_and_test(&root->refs)) { | |
2200 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2201 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2202 | if (root->anon_dev) |
2203 | free_anon_bdev(root->anon_dev); | |
2204 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2205 | free_root_extent_buffers(root); |
8c38938c JB |
2206 | #ifdef CONFIG_BTRFS_DEBUG |
2207 | spin_lock(&root->fs_info->fs_roots_radix_lock); | |
2208 | list_del_init(&root->leak_list); | |
2209 | spin_unlock(&root->fs_info->fs_roots_radix_lock); | |
2210 | #endif | |
2211 | kfree(root); | |
2212 | } | |
2213 | } | |
2214 | ||
faa2dbf0 | 2215 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2216 | { |
2217 | int ret; | |
2218 | struct btrfs_root *gang[8]; | |
2219 | int i; | |
2220 | ||
2221 | while (!list_empty(&fs_info->dead_roots)) { | |
2222 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2223 | struct btrfs_root, root_list); | |
2224 | list_del(&gang[0]->root_list); | |
2225 | ||
8c38938c | 2226 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
cb517eab | 2227 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
dc9492c1 | 2228 | btrfs_put_root(gang[0]); |
171f6537 JB |
2229 | } |
2230 | ||
2231 | while (1) { | |
2232 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2233 | (void **)gang, 0, | |
2234 | ARRAY_SIZE(gang)); | |
2235 | if (!ret) | |
2236 | break; | |
2237 | for (i = 0; i < ret; i++) | |
cb517eab | 2238 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 JB |
2239 | } |
2240 | } | |
af31f5e5 | 2241 | |
638aa7ed ES |
2242 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2243 | { | |
2244 | mutex_init(&fs_info->scrub_lock); | |
2245 | atomic_set(&fs_info->scrubs_running, 0); | |
2246 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2247 | atomic_set(&fs_info->scrubs_paused, 0); | |
2248 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2249 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2250 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2251 | } |
2252 | ||
779a65a4 ES |
2253 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2254 | { | |
2255 | spin_lock_init(&fs_info->balance_lock); | |
2256 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2257 | atomic_set(&fs_info->balance_pause_req, 0); |
2258 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2259 | fs_info->balance_ctl = NULL; | |
2260 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2261 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2262 | } |
2263 | ||
6bccf3ab | 2264 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2265 | { |
2ff7e61e JM |
2266 | struct inode *inode = fs_info->btree_inode; |
2267 | ||
2268 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2269 | set_nlink(inode, 1); | |
f37938e0 ES |
2270 | /* |
2271 | * we set the i_size on the btree inode to the max possible int. | |
2272 | * the real end of the address space is determined by all of | |
2273 | * the devices in the system | |
2274 | */ | |
2ff7e61e JM |
2275 | inode->i_size = OFFSET_MAX; |
2276 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2277 | |
2ff7e61e | 2278 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2279 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2c53a14d | 2280 | IO_TREE_BTREE_INODE_IO, inode); |
7b439738 | 2281 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2282 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2283 | |
5c8fd99f | 2284 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2285 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2286 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2287 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2288 | } |
2289 | ||
ad618368 ES |
2290 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2291 | { | |
ad618368 | 2292 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2293 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2294 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2295 | } |
2296 | ||
f9e92e40 ES |
2297 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2298 | { | |
2299 | spin_lock_init(&fs_info->qgroup_lock); | |
2300 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2301 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2302 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2303 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2304 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2305 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2306 | mutex_init(&fs_info->qgroup_rescan_lock); |
2307 | } | |
2308 | ||
2a458198 ES |
2309 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2310 | struct btrfs_fs_devices *fs_devices) | |
2311 | { | |
f7b885be | 2312 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2313 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2314 | |
2315 | fs_info->workers = | |
cb001095 JM |
2316 | btrfs_alloc_workqueue(fs_info, "worker", |
2317 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2318 | |
2319 | fs_info->delalloc_workers = | |
cb001095 JM |
2320 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2321 | flags, max_active, 2); | |
2a458198 ES |
2322 | |
2323 | fs_info->flush_workers = | |
cb001095 JM |
2324 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2325 | flags, max_active, 0); | |
2a458198 ES |
2326 | |
2327 | fs_info->caching_workers = | |
cb001095 | 2328 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2329 | |
2a458198 | 2330 | fs_info->fixup_workers = |
cb001095 | 2331 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2332 | |
2333 | /* | |
2334 | * endios are largely parallel and should have a very | |
2335 | * low idle thresh | |
2336 | */ | |
2337 | fs_info->endio_workers = | |
cb001095 | 2338 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2339 | fs_info->endio_meta_workers = |
cb001095 JM |
2340 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2341 | max_active, 4); | |
2a458198 | 2342 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2343 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2344 | max_active, 2); | |
2a458198 | 2345 | fs_info->endio_raid56_workers = |
cb001095 JM |
2346 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2347 | max_active, 4); | |
2a458198 | 2348 | fs_info->rmw_workers = |
cb001095 | 2349 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2350 | fs_info->endio_write_workers = |
cb001095 JM |
2351 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2352 | max_active, 2); | |
2a458198 | 2353 | fs_info->endio_freespace_worker = |
cb001095 JM |
2354 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2355 | max_active, 0); | |
2a458198 | 2356 | fs_info->delayed_workers = |
cb001095 JM |
2357 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2358 | max_active, 0); | |
2a458198 | 2359 | fs_info->readahead_workers = |
cb001095 JM |
2360 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2361 | max_active, 2); | |
2a458198 | 2362 | fs_info->qgroup_rescan_workers = |
cb001095 | 2363 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2364 | fs_info->discard_ctl.discard_workers = |
2365 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 ES |
2366 | |
2367 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
ba8a9d07 | 2368 | fs_info->flush_workers && |
2a458198 ES |
2369 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2370 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2371 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2372 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2373 | fs_info->caching_workers && fs_info->readahead_workers && | |
2374 | fs_info->fixup_workers && fs_info->delayed_workers && | |
b0643e59 DZ |
2375 | fs_info->qgroup_rescan_workers && |
2376 | fs_info->discard_ctl.discard_workers)) { | |
2a458198 ES |
2377 | return -ENOMEM; |
2378 | } | |
2379 | ||
2380 | return 0; | |
2381 | } | |
2382 | ||
6d97c6e3 JT |
2383 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2384 | { | |
2385 | struct crypto_shash *csum_shash; | |
b4e967be | 2386 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2387 | |
b4e967be | 2388 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2389 | |
2390 | if (IS_ERR(csum_shash)) { | |
2391 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2392 | csum_driver); |
6d97c6e3 JT |
2393 | return PTR_ERR(csum_shash); |
2394 | } | |
2395 | ||
2396 | fs_info->csum_shash = csum_shash; | |
2397 | ||
2398 | return 0; | |
2399 | } | |
2400 | ||
63443bf5 ES |
2401 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2402 | struct btrfs_fs_devices *fs_devices) | |
2403 | { | |
2404 | int ret; | |
63443bf5 ES |
2405 | struct btrfs_root *log_tree_root; |
2406 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2407 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2408 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2409 | |
2410 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2411 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2412 | return -EIO; |
2413 | } | |
2414 | ||
96dfcb46 JB |
2415 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2416 | GFP_KERNEL); | |
63443bf5 ES |
2417 | if (!log_tree_root) |
2418 | return -ENOMEM; | |
2419 | ||
2ff7e61e | 2420 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2421 | BTRFS_TREE_LOG_OBJECTID, |
2422 | fs_info->generation + 1, level, | |
2423 | NULL); | |
64c043de | 2424 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2425 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2426 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2427 | log_tree_root->node = NULL; |
00246528 | 2428 | btrfs_put_root(log_tree_root); |
0eeff236 | 2429 | return ret; |
64c043de | 2430 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2431 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2432 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2433 | return -EIO; |
2434 | } | |
2435 | /* returns with log_tree_root freed on success */ | |
2436 | ret = btrfs_recover_log_trees(log_tree_root); | |
2437 | if (ret) { | |
0b246afa JM |
2438 | btrfs_handle_fs_error(fs_info, ret, |
2439 | "Failed to recover log tree"); | |
00246528 | 2440 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2441 | return ret; |
2442 | } | |
2443 | ||
bc98a42c | 2444 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2445 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2446 | if (ret) |
2447 | return ret; | |
2448 | } | |
2449 | ||
2450 | return 0; | |
2451 | } | |
2452 | ||
6bccf3ab | 2453 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2454 | { |
6bccf3ab | 2455 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2456 | struct btrfs_root *root; |
4bbcaa64 ES |
2457 | struct btrfs_key location; |
2458 | int ret; | |
2459 | ||
6bccf3ab JM |
2460 | BUG_ON(!fs_info->tree_root); |
2461 | ||
4bbcaa64 ES |
2462 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2463 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2464 | location.offset = 0; | |
2465 | ||
a4f3d2c4 | 2466 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2467 | if (IS_ERR(root)) { |
42437a63 JB |
2468 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2469 | ret = PTR_ERR(root); | |
2470 | goto out; | |
2471 | } | |
2472 | } else { | |
2473 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2474 | fs_info->extent_root = root; | |
f50f4353 | 2475 | } |
4bbcaa64 ES |
2476 | |
2477 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2478 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2479 | if (IS_ERR(root)) { |
42437a63 JB |
2480 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2481 | ret = PTR_ERR(root); | |
2482 | goto out; | |
2483 | } | |
2484 | } else { | |
2485 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2486 | fs_info->dev_root = root; | |
f50f4353 | 2487 | } |
820a49da JB |
2488 | /* Initialize fs_info for all devices in any case */ |
2489 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2490 | |
882dbe0c JB |
2491 | /* If IGNOREDATACSUMS is set don't bother reading the csum root. */ |
2492 | if (!btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2493 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
2494 | root = btrfs_read_tree_root(tree_root, &location); | |
2495 | if (IS_ERR(root)) { | |
2496 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2497 | ret = PTR_ERR(root); | |
2498 | goto out; | |
2499 | } | |
2500 | } else { | |
2501 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2502 | fs_info->csum_root = root; | |
42437a63 | 2503 | } |
f50f4353 | 2504 | } |
4bbcaa64 | 2505 | |
aeb935a4 QW |
2506 | /* |
2507 | * This tree can share blocks with some other fs tree during relocation | |
2508 | * and we need a proper setup by btrfs_get_fs_root | |
2509 | */ | |
56e9357a DS |
2510 | root = btrfs_get_fs_root(tree_root->fs_info, |
2511 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2512 | if (IS_ERR(root)) { |
42437a63 JB |
2513 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2514 | ret = PTR_ERR(root); | |
2515 | goto out; | |
2516 | } | |
2517 | } else { | |
2518 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2519 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2520 | } |
aeb935a4 | 2521 | |
4bbcaa64 | 2522 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2523 | root = btrfs_read_tree_root(tree_root, &location); |
2524 | if (!IS_ERR(root)) { | |
2525 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2526 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2527 | fs_info->quota_root = root; |
4bbcaa64 ES |
2528 | } |
2529 | ||
2530 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2531 | root = btrfs_read_tree_root(tree_root, &location); |
2532 | if (IS_ERR(root)) { | |
42437a63 JB |
2533 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2534 | ret = PTR_ERR(root); | |
2535 | if (ret != -ENOENT) | |
2536 | goto out; | |
2537 | } | |
4bbcaa64 | 2538 | } else { |
a4f3d2c4 DS |
2539 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2540 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2541 | } |
2542 | ||
70f6d82e OS |
2543 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2544 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2545 | root = btrfs_read_tree_root(tree_root, &location); | |
f50f4353 | 2546 | if (IS_ERR(root)) { |
42437a63 JB |
2547 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2548 | ret = PTR_ERR(root); | |
2549 | goto out; | |
2550 | } | |
2551 | } else { | |
2552 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2553 | fs_info->free_space_root = root; | |
f50f4353 | 2554 | } |
70f6d82e OS |
2555 | } |
2556 | ||
4bbcaa64 | 2557 | return 0; |
f50f4353 LB |
2558 | out: |
2559 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2560 | location.objectid, ret); | |
2561 | return ret; | |
4bbcaa64 ES |
2562 | } |
2563 | ||
069ec957 QW |
2564 | /* |
2565 | * Real super block validation | |
2566 | * NOTE: super csum type and incompat features will not be checked here. | |
2567 | * | |
2568 | * @sb: super block to check | |
2569 | * @mirror_num: the super block number to check its bytenr: | |
2570 | * 0 the primary (1st) sb | |
2571 | * 1, 2 2nd and 3rd backup copy | |
2572 | * -1 skip bytenr check | |
2573 | */ | |
2574 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2575 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2576 | { |
21a852b0 QW |
2577 | u64 nodesize = btrfs_super_nodesize(sb); |
2578 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2579 | int ret = 0; | |
2580 | ||
2581 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2582 | btrfs_err(fs_info, "no valid FS found"); | |
2583 | ret = -EINVAL; | |
2584 | } | |
2585 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2586 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2587 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2588 | ret = -EINVAL; | |
2589 | } | |
2590 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2591 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2592 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2593 | ret = -EINVAL; | |
2594 | } | |
2595 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2596 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2597 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2598 | ret = -EINVAL; | |
2599 | } | |
2600 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2601 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2602 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2603 | ret = -EINVAL; | |
2604 | } | |
2605 | ||
2606 | /* | |
2607 | * Check sectorsize and nodesize first, other check will need it. | |
2608 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2609 | */ | |
2610 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2611 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2612 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2613 | ret = -EINVAL; | |
2614 | } | |
0bb3eb3e QW |
2615 | |
2616 | /* | |
2617 | * For 4K page size, we only support 4K sector size. | |
2618 | * For 64K page size, we support read-write for 64K sector size, and | |
2619 | * read-only for 4K sector size. | |
2620 | */ | |
2621 | if ((PAGE_SIZE == SZ_4K && sectorsize != PAGE_SIZE) || | |
2622 | (PAGE_SIZE == SZ_64K && (sectorsize != SZ_4K && | |
2623 | sectorsize != SZ_64K))) { | |
21a852b0 | 2624 | btrfs_err(fs_info, |
0bb3eb3e | 2625 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2626 | sectorsize, PAGE_SIZE); |
2627 | ret = -EINVAL; | |
2628 | } | |
0bb3eb3e | 2629 | |
21a852b0 QW |
2630 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2631 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2632 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2633 | ret = -EINVAL; | |
2634 | } | |
2635 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2636 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2637 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2638 | ret = -EINVAL; | |
2639 | } | |
2640 | ||
2641 | /* Root alignment check */ | |
2642 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2643 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2644 | btrfs_super_root(sb)); | |
2645 | ret = -EINVAL; | |
2646 | } | |
2647 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2648 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2649 | btrfs_super_chunk_root(sb)); | |
2650 | ret = -EINVAL; | |
2651 | } | |
2652 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2653 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2654 | btrfs_super_log_root(sb)); | |
2655 | ret = -EINVAL; | |
2656 | } | |
2657 | ||
aefd7f70 NB |
2658 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2659 | BTRFS_FSID_SIZE)) { | |
2660 | btrfs_err(fs_info, | |
2661 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2662 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2663 | ret = -EINVAL; | |
2664 | } | |
2665 | ||
2666 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2667 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2668 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2669 | btrfs_err(fs_info, | |
2670 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2671 | fs_info->super_copy->metadata_uuid, | |
2672 | fs_info->fs_devices->metadata_uuid); | |
2673 | ret = -EINVAL; | |
2674 | } | |
2675 | ||
de37aa51 | 2676 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2677 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2678 | btrfs_err(fs_info, |
7239ff4b | 2679 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2680 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2681 | ret = -EINVAL; |
2682 | } | |
2683 | ||
2684 | /* | |
2685 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2686 | * done later | |
2687 | */ | |
2688 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2689 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2690 | btrfs_super_bytes_used(sb)); | |
2691 | ret = -EINVAL; | |
2692 | } | |
2693 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2694 | btrfs_err(fs_info, "invalid stripesize %u", | |
2695 | btrfs_super_stripesize(sb)); | |
2696 | ret = -EINVAL; | |
2697 | } | |
2698 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2699 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2700 | btrfs_super_num_devices(sb)); | |
2701 | if (btrfs_super_num_devices(sb) == 0) { | |
2702 | btrfs_err(fs_info, "number of devices is 0"); | |
2703 | ret = -EINVAL; | |
2704 | } | |
2705 | ||
069ec957 QW |
2706 | if (mirror_num >= 0 && |
2707 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2708 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2709 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2710 | ret = -EINVAL; | |
2711 | } | |
2712 | ||
2713 | /* | |
2714 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2715 | * and one chunk | |
2716 | */ | |
2717 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2718 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2719 | btrfs_super_sys_array_size(sb), | |
2720 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2721 | ret = -EINVAL; | |
2722 | } | |
2723 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2724 | + sizeof(struct btrfs_chunk)) { | |
2725 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2726 | btrfs_super_sys_array_size(sb), | |
2727 | sizeof(struct btrfs_disk_key) | |
2728 | + sizeof(struct btrfs_chunk)); | |
2729 | ret = -EINVAL; | |
2730 | } | |
2731 | ||
2732 | /* | |
2733 | * The generation is a global counter, we'll trust it more than the others | |
2734 | * but it's still possible that it's the one that's wrong. | |
2735 | */ | |
2736 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2737 | btrfs_warn(fs_info, | |
2738 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2739 | btrfs_super_generation(sb), | |
2740 | btrfs_super_chunk_root_generation(sb)); | |
2741 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2742 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2743 | btrfs_warn(fs_info, | |
2744 | "suspicious: generation < cache_generation: %llu < %llu", | |
2745 | btrfs_super_generation(sb), | |
2746 | btrfs_super_cache_generation(sb)); | |
2747 | ||
2748 | return ret; | |
2749 | } | |
2750 | ||
069ec957 QW |
2751 | /* |
2752 | * Validation of super block at mount time. | |
2753 | * Some checks already done early at mount time, like csum type and incompat | |
2754 | * flags will be skipped. | |
2755 | */ | |
2756 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2757 | { | |
2758 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2759 | } | |
2760 | ||
75cb857d QW |
2761 | /* |
2762 | * Validation of super block at write time. | |
2763 | * Some checks like bytenr check will be skipped as their values will be | |
2764 | * overwritten soon. | |
2765 | * Extra checks like csum type and incompat flags will be done here. | |
2766 | */ | |
2767 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2768 | struct btrfs_super_block *sb) | |
2769 | { | |
2770 | int ret; | |
2771 | ||
2772 | ret = validate_super(fs_info, sb, -1); | |
2773 | if (ret < 0) | |
2774 | goto out; | |
e7e16f48 | 2775 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2776 | ret = -EUCLEAN; |
2777 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2778 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2779 | goto out; | |
2780 | } | |
2781 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2782 | ret = -EUCLEAN; | |
2783 | btrfs_err(fs_info, | |
2784 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2785 | btrfs_super_incompat_flags(sb), | |
2786 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2787 | goto out; | |
2788 | } | |
2789 | out: | |
2790 | if (ret < 0) | |
2791 | btrfs_err(fs_info, | |
2792 | "super block corruption detected before writing it to disk"); | |
2793 | return ret; | |
2794 | } | |
2795 | ||
6ef108dd | 2796 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2797 | { |
6ef108dd | 2798 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2799 | struct btrfs_super_block *sb = fs_info->super_copy; |
2800 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2801 | bool handle_error = false; | |
2802 | int ret = 0; | |
2803 | int i; | |
2804 | ||
2805 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2806 | u64 generation; | |
2807 | int level; | |
2808 | ||
2809 | if (handle_error) { | |
2810 | if (!IS_ERR(tree_root->node)) | |
2811 | free_extent_buffer(tree_root->node); | |
2812 | tree_root->node = NULL; | |
2813 | ||
2814 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2815 | break; | |
2816 | ||
2817 | free_root_pointers(fs_info, 0); | |
2818 | ||
2819 | /* | |
2820 | * Don't use the log in recovery mode, it won't be | |
2821 | * valid | |
2822 | */ | |
2823 | btrfs_set_super_log_root(sb, 0); | |
2824 | ||
2825 | /* We can't trust the free space cache either */ | |
2826 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2827 | ||
2828 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2829 | backup_index = ret; |
b8522a1e NB |
2830 | if (ret < 0) |
2831 | return ret; | |
2832 | } | |
2833 | generation = btrfs_super_generation(sb); | |
2834 | level = btrfs_super_root_level(sb); | |
2835 | tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), | |
1b7ec85e | 2836 | BTRFS_ROOT_TREE_OBJECTID, |
b8522a1e | 2837 | generation, level, NULL); |
217f5004 | 2838 | if (IS_ERR(tree_root->node)) { |
b8522a1e | 2839 | handle_error = true; |
217f5004 NB |
2840 | ret = PTR_ERR(tree_root->node); |
2841 | tree_root->node = NULL; | |
2842 | btrfs_warn(fs_info, "couldn't read tree root"); | |
2843 | continue; | |
b8522a1e | 2844 | |
217f5004 NB |
2845 | } else if (!extent_buffer_uptodate(tree_root->node)) { |
2846 | handle_error = true; | |
2847 | ret = -EIO; | |
2848 | btrfs_warn(fs_info, "error while reading tree root"); | |
b8522a1e NB |
2849 | continue; |
2850 | } | |
2851 | ||
2852 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); | |
2853 | tree_root->commit_root = btrfs_root_node(tree_root); | |
2854 | btrfs_set_root_refs(&tree_root->root_item, 1); | |
2855 | ||
336a0d8d NB |
2856 | /* |
2857 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2858 | * hasn't been fully initialised and we are the only user | |
2859 | */ | |
453e4873 | 2860 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2861 | if (ret < 0) { |
b8522a1e NB |
2862 | handle_error = true; |
2863 | continue; | |
2864 | } | |
2865 | ||
6b8fad57 | 2866 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2867 | |
2868 | ret = btrfs_read_roots(fs_info); | |
2869 | if (ret < 0) { | |
2870 | handle_error = true; | |
2871 | continue; | |
2872 | } | |
2873 | ||
2874 | /* All successful */ | |
2875 | fs_info->generation = generation; | |
2876 | fs_info->last_trans_committed = generation; | |
ebde420a | 2877 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2878 | |
2879 | /* Always begin writing backup roots after the one being used */ | |
2880 | if (backup_index < 0) { | |
2881 | fs_info->backup_root_index = 0; | |
2882 | } else { | |
2883 | fs_info->backup_root_index = backup_index + 1; | |
2884 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2885 | } | |
b8522a1e NB |
2886 | break; |
2887 | } | |
2888 | ||
2889 | return ret; | |
2890 | } | |
2891 | ||
8260edba | 2892 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2893 | { |
76dda93c | 2894 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2895 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2896 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2897 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2898 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2899 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2900 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2901 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2902 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2903 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2904 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2905 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2906 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2907 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2908 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2909 | spin_lock_init(&fs_info->treelog_bg_lock); |
763e2672 | 2910 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2911 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2912 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2913 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2914 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2915 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2916 | mutex_init(&fs_info->zoned_meta_io_lock); |
de98ced9 | 2917 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2918 | |
0b86a832 | 2919 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2920 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2921 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2922 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2923 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
bd647ce3 JB |
2924 | #ifdef CONFIG_BTRFS_DEBUG |
2925 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2926 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2927 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2928 | #endif |
c8bf1b67 | 2929 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2930 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2931 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2932 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2933 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2934 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2935 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2936 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2937 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2938 | BTRFS_BLOCK_RSV_DELREFS); | |
2939 | ||
771ed689 | 2940 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2941 | atomic_set(&fs_info->defrag_running, 0); |
2fefd558 | 2942 | atomic_set(&fs_info->reada_works_cnt, 0); |
034f784d | 2943 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2944 | atomic64_set(&fs_info->tree_mod_seq, 0); |
95ac567a | 2945 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2946 | fs_info->metadata_ratio = 0; |
4cb5300b | 2947 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2948 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2949 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2950 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2951 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2952 | /* readahead state */ |
d0164adc | 2953 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2954 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2955 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2956 | |
b34b086c CM |
2957 | fs_info->thread_pool_size = min_t(unsigned long, |
2958 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2959 | |
199c2a9c MX |
2960 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2961 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2962 | |
638aa7ed | 2963 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2964 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2965 | fs_info->check_integrity_print_mask = 0; | |
2966 | #endif | |
779a65a4 | 2967 | btrfs_init_balance(fs_info); |
57056740 | 2968 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2969 | |
0f9dd46c | 2970 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2971 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2972 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2973 | |
fe119a6e NB |
2974 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
2975 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
afcdd129 | 2976 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2977 | |
5a3f23d5 | 2978 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2979 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2980 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2981 | mutex_init(&fs_info->transaction_kthread_mutex); |
2982 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2983 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2984 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2985 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2986 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2987 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2988 | |
ad618368 | 2989 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2990 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2991 | btrfs_discard_init(fs_info); |
416ac51d | 2992 | |
fa9c0d79 CM |
2993 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2994 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2995 | ||
e6dcd2dc | 2996 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2997 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2998 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2999 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3000 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3001 | |
da17066c JM |
3002 | /* Usable values until the real ones are cached from the superblock */ |
3003 | fs_info->nodesize = 4096; | |
3004 | fs_info->sectorsize = 4096; | |
ab108d99 | 3005 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3006 | fs_info->stripesize = 4096; |
3007 | ||
eede2bf3 OS |
3008 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3009 | fs_info->swapfile_pins = RB_ROOT; | |
3010 | ||
18bb8bbf JT |
3011 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3012 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3013 | } |
3014 | ||
3015 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3016 | { | |
3017 | int ret; | |
3018 | ||
3019 | fs_info->sb = sb; | |
3020 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3021 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3022 | |
5deb17e1 | 3023 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3024 | if (ret) |
c75e8394 | 3025 | return ret; |
ae18c37a JB |
3026 | |
3027 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3028 | if (ret) | |
c75e8394 | 3029 | return ret; |
ae18c37a JB |
3030 | |
3031 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3032 | (1 + ilog2(nr_cpu_ids)); | |
3033 | ||
3034 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3035 | if (ret) | |
c75e8394 | 3036 | return ret; |
ae18c37a JB |
3037 | |
3038 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3039 | GFP_KERNEL); | |
3040 | if (ret) | |
c75e8394 | 3041 | return ret; |
ae18c37a JB |
3042 | |
3043 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3044 | GFP_KERNEL); | |
c75e8394 JB |
3045 | if (!fs_info->delayed_root) |
3046 | return -ENOMEM; | |
ae18c37a JB |
3047 | btrfs_init_delayed_root(fs_info->delayed_root); |
3048 | ||
a0a1db70 FM |
3049 | if (sb_rdonly(sb)) |
3050 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3051 | ||
c75e8394 | 3052 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3053 | } |
3054 | ||
97f4dd09 NB |
3055 | static int btrfs_uuid_rescan_kthread(void *data) |
3056 | { | |
3057 | struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; | |
3058 | int ret; | |
3059 | ||
3060 | /* | |
3061 | * 1st step is to iterate through the existing UUID tree and | |
3062 | * to delete all entries that contain outdated data. | |
3063 | * 2nd step is to add all missing entries to the UUID tree. | |
3064 | */ | |
3065 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3066 | if (ret < 0) { | |
c94bec2c JB |
3067 | if (ret != -EINTR) |
3068 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3069 | ret); | |
97f4dd09 NB |
3070 | up(&fs_info->uuid_tree_rescan_sem); |
3071 | return ret; | |
3072 | } | |
3073 | return btrfs_uuid_scan_kthread(data); | |
3074 | } | |
3075 | ||
3076 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3077 | { | |
3078 | struct task_struct *task; | |
3079 | ||
3080 | down(&fs_info->uuid_tree_rescan_sem); | |
3081 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3082 | if (IS_ERR(task)) { | |
3083 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3084 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3085 | up(&fs_info->uuid_tree_rescan_sem); | |
3086 | return PTR_ERR(task); | |
3087 | } | |
3088 | ||
3089 | return 0; | |
3090 | } | |
3091 | ||
8cd29088 BB |
3092 | /* |
3093 | * Some options only have meaning at mount time and shouldn't persist across | |
3094 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3095 | * code paths. | |
3096 | */ | |
3097 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3098 | { | |
3099 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3100 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3101 | } |
3102 | ||
44c0ca21 BB |
3103 | /* |
3104 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3105 | * and btrfs_remount when remounting from read-only to read-write. | |
3106 | */ | |
3107 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3108 | { | |
3109 | int ret; | |
94846229 | 3110 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3111 | bool clear_free_space_tree = false; |
3112 | ||
3113 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3114 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3115 | clear_free_space_tree = true; | |
3116 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3117 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3118 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3119 | clear_free_space_tree = true; | |
3120 | } | |
3121 | ||
3122 | if (clear_free_space_tree) { | |
3123 | btrfs_info(fs_info, "clearing free space tree"); | |
3124 | ret = btrfs_clear_free_space_tree(fs_info); | |
3125 | if (ret) { | |
3126 | btrfs_warn(fs_info, | |
3127 | "failed to clear free space tree: %d", ret); | |
3128 | goto out; | |
3129 | } | |
3130 | } | |
44c0ca21 | 3131 | |
8d488a8c FM |
3132 | /* |
3133 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3134 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
3135 | * them into the fs_info->fs_roots_radix tree. This must be done before | |
3136 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it | |
3137 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3138 | * item before the root's tree is deleted - this means that if we unmount | |
3139 | * or crash before the deletion completes, on the next mount we will not | |
3140 | * delete what remains of the tree because the orphan item does not | |
3141 | * exists anymore, which is what tells us we have a pending deletion. | |
3142 | */ | |
3143 | ret = btrfs_find_orphan_roots(fs_info); | |
3144 | if (ret) | |
3145 | goto out; | |
3146 | ||
44c0ca21 BB |
3147 | ret = btrfs_cleanup_fs_roots(fs_info); |
3148 | if (ret) | |
3149 | goto out; | |
3150 | ||
8f1c21d7 BB |
3151 | down_read(&fs_info->cleanup_work_sem); |
3152 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3153 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3154 | up_read(&fs_info->cleanup_work_sem); | |
3155 | goto out; | |
3156 | } | |
3157 | up_read(&fs_info->cleanup_work_sem); | |
3158 | ||
44c0ca21 BB |
3159 | mutex_lock(&fs_info->cleaner_mutex); |
3160 | ret = btrfs_recover_relocation(fs_info->tree_root); | |
3161 | mutex_unlock(&fs_info->cleaner_mutex); | |
3162 | if (ret < 0) { | |
3163 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3164 | goto out; | |
3165 | } | |
3166 | ||
5011139a BB |
3167 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3168 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3169 | btrfs_info(fs_info, "creating free space tree"); | |
3170 | ret = btrfs_create_free_space_tree(fs_info); | |
3171 | if (ret) { | |
3172 | btrfs_warn(fs_info, | |
3173 | "failed to create free space tree: %d", ret); | |
3174 | goto out; | |
3175 | } | |
3176 | } | |
3177 | ||
94846229 BB |
3178 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3179 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3180 | if (ret) | |
3181 | goto out; | |
3182 | } | |
3183 | ||
44c0ca21 BB |
3184 | ret = btrfs_resume_balance_async(fs_info); |
3185 | if (ret) | |
3186 | goto out; | |
3187 | ||
3188 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3189 | if (ret) { | |
3190 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3191 | goto out; | |
3192 | } | |
3193 | ||
3194 | btrfs_qgroup_rescan_resume(fs_info); | |
3195 | ||
3196 | if (!fs_info->uuid_root) { | |
3197 | btrfs_info(fs_info, "creating UUID tree"); | |
3198 | ret = btrfs_create_uuid_tree(fs_info); | |
3199 | if (ret) { | |
3200 | btrfs_warn(fs_info, | |
3201 | "failed to create the UUID tree %d", ret); | |
3202 | goto out; | |
3203 | } | |
3204 | } | |
3205 | ||
3206 | out: | |
3207 | return ret; | |
3208 | } | |
3209 | ||
ae18c37a JB |
3210 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3211 | char *options) | |
3212 | { | |
3213 | u32 sectorsize; | |
3214 | u32 nodesize; | |
3215 | u32 stripesize; | |
3216 | u64 generation; | |
3217 | u64 features; | |
3218 | u16 csum_type; | |
ae18c37a JB |
3219 | struct btrfs_super_block *disk_super; |
3220 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3221 | struct btrfs_root *tree_root; | |
3222 | struct btrfs_root *chunk_root; | |
3223 | int ret; | |
3224 | int err = -EINVAL; | |
ae18c37a JB |
3225 | int level; |
3226 | ||
8260edba | 3227 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3228 | if (ret) { |
83c8266a | 3229 | err = ret; |
ae18c37a | 3230 | goto fail; |
53b381b3 DW |
3231 | } |
3232 | ||
ae18c37a JB |
3233 | /* These need to be init'ed before we start creating inodes and such. */ |
3234 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3235 | GFP_KERNEL); | |
3236 | fs_info->tree_root = tree_root; | |
3237 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3238 | GFP_KERNEL); | |
3239 | fs_info->chunk_root = chunk_root; | |
3240 | if (!tree_root || !chunk_root) { | |
3241 | err = -ENOMEM; | |
c75e8394 | 3242 | goto fail; |
ae18c37a JB |
3243 | } |
3244 | ||
3245 | fs_info->btree_inode = new_inode(sb); | |
3246 | if (!fs_info->btree_inode) { | |
3247 | err = -ENOMEM; | |
c75e8394 | 3248 | goto fail; |
ae18c37a JB |
3249 | } |
3250 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3251 | btrfs_init_btree_inode(fs_info); | |
3252 | ||
678e71f9 | 3253 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3254 | |
3255 | /* | |
3256 | * Read super block and check the signature bytes only | |
3257 | */ | |
678e71f9 | 3258 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3259 | if (IS_ERR(disk_super)) { |
3260 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3261 | goto fail_alloc; |
20b45077 | 3262 | } |
39279cc3 | 3263 | |
8dc3f22c | 3264 | /* |
260db43c | 3265 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3266 | * corrupted, we'll find out |
3267 | */ | |
8f32380d | 3268 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3269 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3270 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3271 | csum_type); |
8dc3f22c | 3272 | err = -EINVAL; |
8f32380d | 3273 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3274 | goto fail_alloc; |
3275 | } | |
3276 | ||
83c68bbc SY |
3277 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3278 | ||
6d97c6e3 JT |
3279 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3280 | if (ret) { | |
3281 | err = ret; | |
8f32380d | 3282 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3283 | goto fail_alloc; |
3284 | } | |
3285 | ||
1104a885 DS |
3286 | /* |
3287 | * We want to check superblock checksum, the type is stored inside. | |
3288 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3289 | */ | |
8f32380d | 3290 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3291 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3292 | err = -EINVAL; |
8f32380d | 3293 | btrfs_release_disk_super(disk_super); |
141386e1 | 3294 | goto fail_alloc; |
1104a885 DS |
3295 | } |
3296 | ||
3297 | /* | |
3298 | * super_copy is zeroed at allocation time and we never touch the | |
3299 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3300 | * the whole block of INFO_SIZE | |
3301 | */ | |
8f32380d JT |
3302 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3303 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3304 | |
fbc6feae NB |
3305 | disk_super = fs_info->super_copy; |
3306 | ||
0b86a832 | 3307 | |
fbc6feae NB |
3308 | features = btrfs_super_flags(disk_super); |
3309 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3310 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3311 | btrfs_set_super_flags(disk_super, features); | |
3312 | btrfs_info(fs_info, | |
3313 | "found metadata UUID change in progress flag, clearing"); | |
3314 | } | |
3315 | ||
3316 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3317 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3318 | |
069ec957 | 3319 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3320 | if (ret) { |
05135f59 | 3321 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3322 | err = -EINVAL; |
141386e1 | 3323 | goto fail_alloc; |
1104a885 DS |
3324 | } |
3325 | ||
0f7d52f4 | 3326 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3327 | goto fail_alloc; |
0f7d52f4 | 3328 | |
acce952b | 3329 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3330 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3331 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3332 | |
75e7cb7f LB |
3333 | /* |
3334 | * In the long term, we'll store the compression type in the super | |
3335 | * block, and it'll be used for per file compression control. | |
3336 | */ | |
3337 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3338 | ||
6f93e834 AJ |
3339 | /* |
3340 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3341 | * than the page size. | |
3342 | */ | |
3343 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { | |
3344 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
3345 | btrfs_info(fs_info, | |
3346 | "flagging fs with big metadata feature"); | |
3347 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3348 | } | |
3349 | ||
3350 | /* Set up fs_info before parsing mount options */ | |
3351 | nodesize = btrfs_super_nodesize(disk_super); | |
3352 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3353 | stripesize = sectorsize; | |
3354 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3355 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3356 | ||
3357 | fs_info->nodesize = nodesize; | |
3358 | fs_info->sectorsize = sectorsize; | |
3359 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3360 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3361 | fs_info->stripesize = stripesize; | |
3362 | ||
2ff7e61e | 3363 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3364 | if (ret) { |
3365 | err = ret; | |
141386e1 | 3366 | goto fail_alloc; |
2b82032c | 3367 | } |
dfe25020 | 3368 | |
f2b636e8 JB |
3369 | features = btrfs_super_incompat_flags(disk_super) & |
3370 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3371 | if (features) { | |
05135f59 DS |
3372 | btrfs_err(fs_info, |
3373 | "cannot mount because of unsupported optional features (%llx)", | |
3374 | features); | |
f2b636e8 | 3375 | err = -EINVAL; |
141386e1 | 3376 | goto fail_alloc; |
f2b636e8 JB |
3377 | } |
3378 | ||
5d4f98a2 | 3379 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3380 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3381 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3382 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3383 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3384 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3385 | |
3173a18f | 3386 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3387 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3388 | |
bc3f116f CM |
3389 | /* |
3390 | * mixed block groups end up with duplicate but slightly offset | |
3391 | * extent buffers for the same range. It leads to corruptions | |
3392 | */ | |
3393 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3394 | (sectorsize != nodesize)) { |
05135f59 DS |
3395 | btrfs_err(fs_info, |
3396 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3397 | nodesize, sectorsize); | |
141386e1 | 3398 | goto fail_alloc; |
bc3f116f CM |
3399 | } |
3400 | ||
ceda0864 MX |
3401 | /* |
3402 | * Needn't use the lock because there is no other task which will | |
3403 | * update the flag. | |
3404 | */ | |
a6fa6fae | 3405 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3406 | |
f2b636e8 JB |
3407 | features = btrfs_super_compat_ro_flags(disk_super) & |
3408 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3409 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
3410 | btrfs_err(fs_info, |
3411 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 3412 | features); |
f2b636e8 | 3413 | err = -EINVAL; |
141386e1 | 3414 | goto fail_alloc; |
f2b636e8 | 3415 | } |
61d92c32 | 3416 | |
95ea0486 QW |
3417 | if (sectorsize != PAGE_SIZE) { |
3418 | btrfs_warn(fs_info, | |
3419 | "read-write for sector size %u with page size %lu is experimental", | |
3420 | sectorsize, PAGE_SIZE); | |
0bb3eb3e | 3421 | } |
c8050b3b QW |
3422 | if (sectorsize != PAGE_SIZE) { |
3423 | if (btrfs_super_incompat_flags(fs_info->super_copy) & | |
3424 | BTRFS_FEATURE_INCOMPAT_RAID56) { | |
3425 | btrfs_err(fs_info, | |
3426 | "RAID56 is not yet supported for sector size %u with page size %lu", | |
3427 | sectorsize, PAGE_SIZE); | |
3428 | err = -EINVAL; | |
3429 | goto fail_alloc; | |
3430 | } | |
3431 | } | |
0bb3eb3e | 3432 | |
2a458198 ES |
3433 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
3434 | if (ret) { | |
3435 | err = ret; | |
0dc3b84a JB |
3436 | goto fail_sb_buffer; |
3437 | } | |
4543df7e | 3438 | |
9e11ceee JK |
3439 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3440 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3441 | |
a061fc8d CM |
3442 | sb->s_blocksize = sectorsize; |
3443 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3444 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3445 | |
925baedd | 3446 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3447 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3448 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3449 | if (ret) { |
05135f59 | 3450 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3451 | goto fail_sb_buffer; |
84eed90f | 3452 | } |
0b86a832 | 3453 | |
84234f3a | 3454 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3455 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3456 | |
2ff7e61e | 3457 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3458 | btrfs_super_chunk_root(disk_super), |
1b7ec85e | 3459 | BTRFS_CHUNK_TREE_OBJECTID, |
581c1760 | 3460 | generation, level, NULL); |
64c043de LB |
3461 | if (IS_ERR(chunk_root->node) || |
3462 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3463 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3464 | if (!IS_ERR(chunk_root->node)) |
3465 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3466 | chunk_root->node = NULL; |
af31f5e5 | 3467 | goto fail_tree_roots; |
83121942 | 3468 | } |
5d4f98a2 YZ |
3469 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3470 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3471 | |
e17cade2 | 3472 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3473 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3474 | BTRFS_UUID_SIZE); | |
e17cade2 | 3475 | |
5b4aacef | 3476 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3477 | if (ret) { |
05135f59 | 3478 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3479 | goto fail_tree_roots; |
2b82032c | 3480 | } |
0b86a832 | 3481 | |
8dabb742 | 3482 | /* |
bacce86a AJ |
3483 | * At this point we know all the devices that make this filesystem, |
3484 | * including the seed devices but we don't know yet if the replace | |
3485 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3486 | * filesystem but skip the replace target device which is checked |
bacce86a | 3487 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3488 | */ |
bacce86a | 3489 | btrfs_free_extra_devids(fs_devices); |
678e71f9 | 3490 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3491 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3492 | goto fail_tree_roots; |
3493 | } | |
3494 | ||
b8522a1e | 3495 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3496 | if (ret) |
b8522a1e | 3497 | goto fail_tree_roots; |
8929ecfa | 3498 | |
73651042 NA |
3499 | /* |
3500 | * Get zone type information of zoned block devices. This will also | |
3501 | * handle emulation of a zoned filesystem if a regular device has the | |
3502 | * zoned incompat feature flag set. | |
3503 | */ | |
3504 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3505 | if (ret) { | |
3506 | btrfs_err(fs_info, | |
3507 | "zoned: failed to read device zone info: %d", | |
3508 | ret); | |
3509 | goto fail_block_groups; | |
3510 | } | |
3511 | ||
75ec1db8 JB |
3512 | /* |
3513 | * If we have a uuid root and we're not being told to rescan we need to | |
3514 | * check the generation here so we can set the | |
3515 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3516 | * transaction during a balance or the log replay without updating the | |
3517 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3518 | * even though it was perfectly fine. | |
3519 | */ | |
3520 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3521 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3522 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3523 | ||
cf90d884 QW |
3524 | ret = btrfs_verify_dev_extents(fs_info); |
3525 | if (ret) { | |
3526 | btrfs_err(fs_info, | |
3527 | "failed to verify dev extents against chunks: %d", | |
3528 | ret); | |
3529 | goto fail_block_groups; | |
3530 | } | |
68310a5e ID |
3531 | ret = btrfs_recover_balance(fs_info); |
3532 | if (ret) { | |
05135f59 | 3533 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3534 | goto fail_block_groups; |
3535 | } | |
3536 | ||
733f4fbb SB |
3537 | ret = btrfs_init_dev_stats(fs_info); |
3538 | if (ret) { | |
05135f59 | 3539 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3540 | goto fail_block_groups; |
3541 | } | |
3542 | ||
8dabb742 SB |
3543 | ret = btrfs_init_dev_replace(fs_info); |
3544 | if (ret) { | |
05135f59 | 3545 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3546 | goto fail_block_groups; |
3547 | } | |
3548 | ||
b70f5097 NA |
3549 | ret = btrfs_check_zoned_mode(fs_info); |
3550 | if (ret) { | |
3551 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3552 | ret); | |
3553 | goto fail_block_groups; | |
3554 | } | |
3555 | ||
c6761a9e | 3556 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3557 | if (ret) { |
05135f59 DS |
3558 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3559 | ret); | |
b7c35e81 AJ |
3560 | goto fail_block_groups; |
3561 | } | |
3562 | ||
96f3136e | 3563 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3564 | if (ret) { |
05135f59 | 3565 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3566 | goto fail_fsdev_sysfs; |
c59021f8 | 3567 | } |
3568 | ||
c59021f8 | 3569 | ret = btrfs_init_space_info(fs_info); |
3570 | if (ret) { | |
05135f59 | 3571 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3572 | goto fail_sysfs; |
c59021f8 | 3573 | } |
3574 | ||
5b4aacef | 3575 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3576 | if (ret) { |
05135f59 | 3577 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3578 | goto fail_sysfs; |
1b1d1f66 | 3579 | } |
4330e183 | 3580 | |
007648db NA |
3581 | btrfs_free_zone_cache(fs_info); |
3582 | ||
510bfab0 AJ |
3583 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3584 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3585 | btrfs_warn(fs_info, |
52042d8e | 3586 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3587 | goto fail_sysfs; |
292fd7fc | 3588 | } |
9078a3e1 | 3589 | |
a74a4b97 CM |
3590 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3591 | "btrfs-cleaner"); | |
57506d50 | 3592 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3593 | goto fail_sysfs; |
a74a4b97 CM |
3594 | |
3595 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3596 | tree_root, | |
3597 | "btrfs-transaction"); | |
57506d50 | 3598 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3599 | goto fail_cleaner; |
a74a4b97 | 3600 | |
583b7231 | 3601 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3602 | !fs_info->fs_devices->rotating) { |
583b7231 | 3603 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3604 | } |
3605 | ||
572d9ab7 | 3606 | /* |
01327610 | 3607 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3608 | * not have to wait for transaction commit |
3609 | */ | |
3610 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3611 | |
21adbd5c | 3612 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3613 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3614 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3615 | btrfs_test_opt(fs_info, |
cbeaae4f | 3616 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3617 | fs_info->check_integrity_print_mask); |
3618 | if (ret) | |
05135f59 DS |
3619 | btrfs_warn(fs_info, |
3620 | "failed to initialize integrity check module: %d", | |
3621 | ret); | |
21adbd5c SB |
3622 | } |
3623 | #endif | |
bcef60f2 AJ |
3624 | ret = btrfs_read_qgroup_config(fs_info); |
3625 | if (ret) | |
3626 | goto fail_trans_kthread; | |
21adbd5c | 3627 | |
fd708b81 JB |
3628 | if (btrfs_build_ref_tree(fs_info)) |
3629 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3630 | ||
96da0919 QW |
3631 | /* do not make disk changes in broken FS or nologreplay is given */ |
3632 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3633 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3634 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3635 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3636 | if (ret) { |
63443bf5 | 3637 | err = ret; |
28c16cbb | 3638 | goto fail_qgroup; |
79787eaa | 3639 | } |
e02119d5 | 3640 | } |
1a40e23b | 3641 | |
56e9357a | 3642 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3643 | if (IS_ERR(fs_info->fs_root)) { |
3644 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3645 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3646 | fs_info->fs_root = NULL; |
bcef60f2 | 3647 | goto fail_qgroup; |
3140c9a3 | 3648 | } |
c289811c | 3649 | |
bc98a42c | 3650 | if (sb_rdonly(sb)) |
8cd29088 | 3651 | goto clear_oneshot; |
59641015 | 3652 | |
44c0ca21 | 3653 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3654 | if (ret) { |
6bccf3ab | 3655 | close_ctree(fs_info); |
2b6ba629 | 3656 | return ret; |
e3acc2a6 | 3657 | } |
b0643e59 | 3658 | btrfs_discard_resume(fs_info); |
b382a324 | 3659 | |
44c0ca21 BB |
3660 | if (fs_info->uuid_root && |
3661 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3662 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3663 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3664 | ret = btrfs_check_uuid_tree(fs_info); |
3665 | if (ret) { | |
05135f59 DS |
3666 | btrfs_warn(fs_info, |
3667 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3668 | close_ctree(fs_info); |
70f80175 SB |
3669 | return ret; |
3670 | } | |
f7a81ea4 | 3671 | } |
94846229 | 3672 | |
afcdd129 | 3673 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3674 | |
78db1921 JB |
3675 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3676 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3677 | wake_up_process(fs_info->cleaner_kthread); | |
3678 | ||
8cd29088 BB |
3679 | clear_oneshot: |
3680 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3681 | return 0; |
39279cc3 | 3682 | |
bcef60f2 AJ |
3683 | fail_qgroup: |
3684 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3685 | fail_trans_kthread: |
3686 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3687 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3688 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3689 | fail_cleaner: |
a74a4b97 | 3690 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3691 | |
3692 | /* | |
3693 | * make sure we're done with the btree inode before we stop our | |
3694 | * kthreads | |
3695 | */ | |
3696 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3697 | |
2365dd3c | 3698 | fail_sysfs: |
6618a59b | 3699 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3700 | |
b7c35e81 AJ |
3701 | fail_fsdev_sysfs: |
3702 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3703 | ||
1b1d1f66 | 3704 | fail_block_groups: |
54067ae9 | 3705 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3706 | |
3707 | fail_tree_roots: | |
9e3aa805 JB |
3708 | if (fs_info->data_reloc_root) |
3709 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3710 | free_root_pointers(fs_info, true); |
2b8195bb | 3711 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3712 | |
39279cc3 | 3713 | fail_sb_buffer: |
7abadb64 | 3714 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3715 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3716 | fail_alloc: |
586e46e2 ID |
3717 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3718 | ||
4543df7e | 3719 | iput(fs_info->btree_inode); |
7e662854 | 3720 | fail: |
586e46e2 | 3721 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3722 | return err; |
eb60ceac | 3723 | } |
663faf9f | 3724 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3725 | |
314b6dd0 | 3726 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3727 | { |
314b6dd0 JT |
3728 | struct btrfs_device *device = bio->bi_private; |
3729 | struct bio_vec *bvec; | |
3730 | struct bvec_iter_all iter_all; | |
3731 | struct page *page; | |
3732 | ||
3733 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3734 | page = bvec->bv_page; | |
3735 | ||
3736 | if (bio->bi_status) { | |
3737 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3738 | "lost page write due to IO error on %s (%d)", | |
3739 | rcu_str_deref(device->name), | |
3740 | blk_status_to_errno(bio->bi_status)); | |
3741 | ClearPageUptodate(page); | |
3742 | SetPageError(page); | |
3743 | btrfs_dev_stat_inc_and_print(device, | |
3744 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3745 | } else { | |
3746 | SetPageUptodate(page); | |
3747 | } | |
3748 | ||
3749 | put_page(page); | |
3750 | unlock_page(page); | |
f2984462 | 3751 | } |
314b6dd0 JT |
3752 | |
3753 | bio_put(bio); | |
f2984462 CM |
3754 | } |
3755 | ||
8f32380d JT |
3756 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
3757 | int copy_num) | |
29c36d72 | 3758 | { |
29c36d72 | 3759 | struct btrfs_super_block *super; |
8f32380d | 3760 | struct page *page; |
12659251 | 3761 | u64 bytenr, bytenr_orig; |
8f32380d | 3762 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3763 | int ret; |
3764 | ||
3765 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3766 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3767 | if (ret == -ENOENT) | |
3768 | return ERR_PTR(-EINVAL); | |
3769 | else if (ret) | |
3770 | return ERR_PTR(ret); | |
29c36d72 | 3771 | |
29c36d72 | 3772 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) |
8f32380d | 3773 | return ERR_PTR(-EINVAL); |
29c36d72 | 3774 | |
8f32380d JT |
3775 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3776 | if (IS_ERR(page)) | |
3777 | return ERR_CAST(page); | |
29c36d72 | 3778 | |
8f32380d | 3779 | super = page_address(page); |
96c2e067 AJ |
3780 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3781 | btrfs_release_disk_super(super); | |
3782 | return ERR_PTR(-ENODATA); | |
3783 | } | |
3784 | ||
12659251 | 3785 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3786 | btrfs_release_disk_super(super); |
3787 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3788 | } |
3789 | ||
8f32380d | 3790 | return super; |
29c36d72 AJ |
3791 | } |
3792 | ||
3793 | ||
8f32380d | 3794 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3795 | { |
8f32380d | 3796 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3797 | int i; |
3798 | u64 transid = 0; | |
a512bbf8 YZ |
3799 | |
3800 | /* we would like to check all the supers, but that would make | |
3801 | * a btrfs mount succeed after a mkfs from a different FS. | |
3802 | * So, we need to add a special mount option to scan for | |
3803 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3804 | */ | |
3805 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
3806 | super = btrfs_read_dev_one_super(bdev, i); |
3807 | if (IS_ERR(super)) | |
a512bbf8 YZ |
3808 | continue; |
3809 | ||
a512bbf8 | 3810 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3811 | if (latest) |
3812 | btrfs_release_disk_super(super); | |
3813 | ||
3814 | latest = super; | |
a512bbf8 | 3815 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3816 | } |
3817 | } | |
92fc03fb | 3818 | |
8f32380d | 3819 | return super; |
a512bbf8 YZ |
3820 | } |
3821 | ||
4eedeb75 | 3822 | /* |
abbb3b8e | 3823 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3824 | * pages we use for writing are locked. |
4eedeb75 | 3825 | * |
abbb3b8e DS |
3826 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3827 | * the expected device size at commit time. Note that max_mirrors must be | |
3828 | * same for write and wait phases. | |
4eedeb75 | 3829 | * |
314b6dd0 | 3830 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3831 | */ |
a512bbf8 | 3832 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3833 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3834 | { |
d5178578 | 3835 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3836 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3837 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3838 | int i; |
a512bbf8 | 3839 | int errors = 0; |
12659251 NA |
3840 | int ret; |
3841 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3842 | |
3843 | if (max_mirrors == 0) | |
3844 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3845 | ||
d5178578 JT |
3846 | shash->tfm = fs_info->csum_shash; |
3847 | ||
a512bbf8 | 3848 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3849 | struct page *page; |
3850 | struct bio *bio; | |
3851 | struct btrfs_super_block *disk_super; | |
3852 | ||
12659251 NA |
3853 | bytenr_orig = btrfs_sb_offset(i); |
3854 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3855 | if (ret == -ENOENT) { | |
3856 | continue; | |
3857 | } else if (ret < 0) { | |
3858 | btrfs_err(device->fs_info, | |
3859 | "couldn't get super block location for mirror %d", | |
3860 | i); | |
3861 | errors++; | |
3862 | continue; | |
3863 | } | |
935e5cc9 MX |
3864 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3865 | device->commit_total_bytes) | |
a512bbf8 YZ |
3866 | break; |
3867 | ||
12659251 | 3868 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3869 | |
fd08001f EB |
3870 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3871 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3872 | sb->csum); | |
4eedeb75 | 3873 | |
314b6dd0 JT |
3874 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3875 | GFP_NOFS); | |
3876 | if (!page) { | |
abbb3b8e | 3877 | btrfs_err(device->fs_info, |
314b6dd0 | 3878 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3879 | bytenr); |
3880 | errors++; | |
4eedeb75 | 3881 | continue; |
abbb3b8e | 3882 | } |
634554dc | 3883 | |
314b6dd0 JT |
3884 | /* Bump the refcount for wait_dev_supers() */ |
3885 | get_page(page); | |
a512bbf8 | 3886 | |
314b6dd0 JT |
3887 | disk_super = page_address(page); |
3888 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3889 | |
314b6dd0 JT |
3890 | /* |
3891 | * Directly use bios here instead of relying on the page cache | |
3892 | * to do I/O, so we don't lose the ability to do integrity | |
3893 | * checking. | |
3894 | */ | |
3895 | bio = bio_alloc(GFP_NOFS, 1); | |
3896 | bio_set_dev(bio, device->bdev); | |
3897 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; | |
3898 | bio->bi_private = device; | |
3899 | bio->bi_end_io = btrfs_end_super_write; | |
3900 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3901 | offset_in_page(bytenr)); | |
a512bbf8 | 3902 | |
387125fc | 3903 | /* |
314b6dd0 JT |
3904 | * We FUA only the first super block. The others we allow to |
3905 | * go down lazy and there's a short window where the on-disk | |
3906 | * copies might still contain the older version. | |
387125fc | 3907 | */ |
314b6dd0 | 3908 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; |
1b9e619c | 3909 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
3910 | bio->bi_opf |= REQ_FUA; |
3911 | ||
3912 | btrfsic_submit_bio(bio); | |
12659251 | 3913 | btrfs_advance_sb_log(device, i); |
a512bbf8 YZ |
3914 | } |
3915 | return errors < i ? 0 : -1; | |
3916 | } | |
3917 | ||
abbb3b8e DS |
3918 | /* |
3919 | * Wait for write completion of superblocks done by write_dev_supers, | |
3920 | * @max_mirrors same for write and wait phases. | |
3921 | * | |
314b6dd0 | 3922 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3923 | * date. |
3924 | */ | |
3925 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3926 | { | |
abbb3b8e DS |
3927 | int i; |
3928 | int errors = 0; | |
b6a535fa | 3929 | bool primary_failed = false; |
12659251 | 3930 | int ret; |
abbb3b8e DS |
3931 | u64 bytenr; |
3932 | ||
3933 | if (max_mirrors == 0) | |
3934 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3935 | ||
3936 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3937 | struct page *page; |
3938 | ||
12659251 NA |
3939 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3940 | if (ret == -ENOENT) { | |
3941 | break; | |
3942 | } else if (ret < 0) { | |
3943 | errors++; | |
3944 | if (i == 0) | |
3945 | primary_failed = true; | |
3946 | continue; | |
3947 | } | |
abbb3b8e DS |
3948 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3949 | device->commit_total_bytes) | |
3950 | break; | |
3951 | ||
314b6dd0 JT |
3952 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3953 | bytenr >> PAGE_SHIFT); | |
3954 | if (!page) { | |
abbb3b8e | 3955 | errors++; |
b6a535fa HM |
3956 | if (i == 0) |
3957 | primary_failed = true; | |
abbb3b8e DS |
3958 | continue; |
3959 | } | |
314b6dd0 JT |
3960 | /* Page is submitted locked and unlocked once the IO completes */ |
3961 | wait_on_page_locked(page); | |
3962 | if (PageError(page)) { | |
abbb3b8e | 3963 | errors++; |
b6a535fa HM |
3964 | if (i == 0) |
3965 | primary_failed = true; | |
3966 | } | |
abbb3b8e | 3967 | |
314b6dd0 JT |
3968 | /* Drop our reference */ |
3969 | put_page(page); | |
abbb3b8e | 3970 | |
314b6dd0 JT |
3971 | /* Drop the reference from the writing run */ |
3972 | put_page(page); | |
abbb3b8e DS |
3973 | } |
3974 | ||
b6a535fa HM |
3975 | /* log error, force error return */ |
3976 | if (primary_failed) { | |
3977 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3978 | device->devid); | |
3979 | return -1; | |
3980 | } | |
3981 | ||
abbb3b8e DS |
3982 | return errors < i ? 0 : -1; |
3983 | } | |
3984 | ||
387125fc CM |
3985 | /* |
3986 | * endio for the write_dev_flush, this will wake anyone waiting | |
3987 | * for the barrier when it is done | |
3988 | */ | |
4246a0b6 | 3989 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3990 | { |
e0ae9994 | 3991 | complete(bio->bi_private); |
387125fc CM |
3992 | } |
3993 | ||
3994 | /* | |
4fc6441a AJ |
3995 | * Submit a flush request to the device if it supports it. Error handling is |
3996 | * done in the waiting counterpart. | |
387125fc | 3997 | */ |
4fc6441a | 3998 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3999 | { |
e0ae9994 | 4000 | struct bio *bio = device->flush_bio; |
387125fc | 4001 | |
e26dedd0 WY |
4002 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4003 | /* | |
4004 | * When a disk has write caching disabled, we skip submission of a bio | |
4005 | * with flush and sync requests before writing the superblock, since | |
4006 | * it's not needed. However when the integrity checker is enabled, this | |
4007 | * results in reports that there are metadata blocks referred by a | |
4008 | * superblock that were not properly flushed. So don't skip the bio | |
4009 | * submission only when the integrity checker is enabled for the sake | |
4010 | * of simplicity, since this is a debug tool and not meant for use in | |
4011 | * non-debug builds. | |
4012 | */ | |
4013 | struct request_queue *q = bdev_get_queue(device->bdev); | |
c2a9c7ab | 4014 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 4015 | return; |
e26dedd0 | 4016 | #endif |
387125fc | 4017 | |
e0ae9994 | 4018 | bio_reset(bio); |
387125fc | 4019 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 4020 | bio_set_dev(bio, device->bdev); |
8d910125 | 4021 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
4022 | init_completion(&device->flush_wait); |
4023 | bio->bi_private = &device->flush_wait; | |
387125fc | 4024 | |
43a01111 | 4025 | btrfsic_submit_bio(bio); |
1c3063b6 | 4026 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4027 | } |
387125fc | 4028 | |
4fc6441a AJ |
4029 | /* |
4030 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4031 | */ | |
8c27cb35 | 4032 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4033 | { |
4fc6441a | 4034 | struct bio *bio = device->flush_bio; |
387125fc | 4035 | |
1c3063b6 | 4036 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4037 | return BLK_STS_OK; |
387125fc | 4038 | |
1c3063b6 | 4039 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4040 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4041 | |
8c27cb35 | 4042 | return bio->bi_status; |
387125fc | 4043 | } |
387125fc | 4044 | |
d10b82fe | 4045 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4046 | { |
6528b99d | 4047 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4048 | return -EIO; |
387125fc CM |
4049 | return 0; |
4050 | } | |
4051 | ||
4052 | /* | |
4053 | * send an empty flush down to each device in parallel, | |
4054 | * then wait for them | |
4055 | */ | |
4056 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4057 | { | |
4058 | struct list_head *head; | |
4059 | struct btrfs_device *dev; | |
5af3e8cc | 4060 | int errors_wait = 0; |
4e4cbee9 | 4061 | blk_status_t ret; |
387125fc | 4062 | |
1538e6c5 | 4063 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4064 | /* send down all the barriers */ |
4065 | head = &info->fs_devices->devices; | |
1538e6c5 | 4066 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4067 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4068 | continue; |
cea7c8bf | 4069 | if (!dev->bdev) |
387125fc | 4070 | continue; |
e12c9621 | 4071 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4072 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4073 | continue; |
4074 | ||
4fc6441a | 4075 | write_dev_flush(dev); |
58efbc9f | 4076 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4077 | } |
4078 | ||
4079 | /* wait for all the barriers */ | |
1538e6c5 | 4080 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4081 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4082 | continue; |
387125fc | 4083 | if (!dev->bdev) { |
5af3e8cc | 4084 | errors_wait++; |
387125fc CM |
4085 | continue; |
4086 | } | |
e12c9621 | 4087 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4088 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4089 | continue; |
4090 | ||
4fc6441a | 4091 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4092 | if (ret) { |
4093 | dev->last_flush_error = ret; | |
66b4993e DS |
4094 | btrfs_dev_stat_inc_and_print(dev, |
4095 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4096 | errors_wait++; |
401b41e5 AJ |
4097 | } |
4098 | } | |
4099 | ||
cea7c8bf | 4100 | if (errors_wait) { |
401b41e5 AJ |
4101 | /* |
4102 | * At some point we need the status of all disks | |
4103 | * to arrive at the volume status. So error checking | |
4104 | * is being pushed to a separate loop. | |
4105 | */ | |
d10b82fe | 4106 | return check_barrier_error(info); |
387125fc | 4107 | } |
387125fc CM |
4108 | return 0; |
4109 | } | |
4110 | ||
943c6e99 ZL |
4111 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4112 | { | |
8789f4fe ZL |
4113 | int raid_type; |
4114 | int min_tolerated = INT_MAX; | |
943c6e99 | 4115 | |
8789f4fe ZL |
4116 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4117 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4118 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4119 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4120 | tolerated_failures); | |
943c6e99 | 4121 | |
8789f4fe ZL |
4122 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4123 | if (raid_type == BTRFS_RAID_SINGLE) | |
4124 | continue; | |
41a6e891 | 4125 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4126 | continue; |
8c3e3582 | 4127 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4128 | btrfs_raid_array[raid_type]. |
4129 | tolerated_failures); | |
4130 | } | |
943c6e99 | 4131 | |
8789f4fe | 4132 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4133 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4134 | min_tolerated = 0; |
4135 | } | |
4136 | ||
4137 | return min_tolerated; | |
943c6e99 ZL |
4138 | } |
4139 | ||
eece6a9c | 4140 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4141 | { |
e5e9a520 | 4142 | struct list_head *head; |
f2984462 | 4143 | struct btrfs_device *dev; |
a061fc8d | 4144 | struct btrfs_super_block *sb; |
f2984462 | 4145 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4146 | int ret; |
4147 | int do_barriers; | |
a236aed1 CM |
4148 | int max_errors; |
4149 | int total_errors = 0; | |
a061fc8d | 4150 | u64 flags; |
f2984462 | 4151 | |
0b246afa | 4152 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4153 | |
4154 | /* | |
4155 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4156 | * not from fsync where the tree roots in fs_info have not | |
4157 | * been consistent on disk. | |
4158 | */ | |
4159 | if (max_mirrors == 0) | |
4160 | backup_super_roots(fs_info); | |
f2984462 | 4161 | |
0b246afa | 4162 | sb = fs_info->super_for_commit; |
a061fc8d | 4163 | dev_item = &sb->dev_item; |
e5e9a520 | 4164 | |
0b246afa JM |
4165 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4166 | head = &fs_info->fs_devices->devices; | |
4167 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4168 | |
5af3e8cc | 4169 | if (do_barriers) { |
0b246afa | 4170 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4171 | if (ret) { |
4172 | mutex_unlock( | |
0b246afa JM |
4173 | &fs_info->fs_devices->device_list_mutex); |
4174 | btrfs_handle_fs_error(fs_info, ret, | |
4175 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4176 | return ret; |
4177 | } | |
4178 | } | |
387125fc | 4179 | |
1538e6c5 | 4180 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4181 | if (!dev->bdev) { |
4182 | total_errors++; | |
4183 | continue; | |
4184 | } | |
e12c9621 | 4185 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4186 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4187 | continue; |
4188 | ||
2b82032c | 4189 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4190 | btrfs_set_stack_device_type(dev_item, dev->type); |
4191 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4192 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4193 | dev->commit_total_bytes); |
ce7213c7 MX |
4194 | btrfs_set_stack_device_bytes_used(dev_item, |
4195 | dev->commit_bytes_used); | |
a061fc8d CM |
4196 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4197 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4198 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4199 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4200 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4201 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4202 | |
a061fc8d CM |
4203 | flags = btrfs_super_flags(sb); |
4204 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4205 | ||
75cb857d QW |
4206 | ret = btrfs_validate_write_super(fs_info, sb); |
4207 | if (ret < 0) { | |
4208 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4209 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4210 | "unexpected superblock corruption detected"); | |
4211 | return -EUCLEAN; | |
4212 | } | |
4213 | ||
abbb3b8e | 4214 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4215 | if (ret) |
4216 | total_errors++; | |
f2984462 | 4217 | } |
a236aed1 | 4218 | if (total_errors > max_errors) { |
0b246afa JM |
4219 | btrfs_err(fs_info, "%d errors while writing supers", |
4220 | total_errors); | |
4221 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4222 | |
9d565ba4 | 4223 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4224 | btrfs_handle_fs_error(fs_info, -EIO, |
4225 | "%d errors while writing supers", | |
4226 | total_errors); | |
9d565ba4 | 4227 | return -EIO; |
a236aed1 | 4228 | } |
f2984462 | 4229 | |
a512bbf8 | 4230 | total_errors = 0; |
1538e6c5 | 4231 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4232 | if (!dev->bdev) |
4233 | continue; | |
e12c9621 | 4234 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4235 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4236 | continue; |
4237 | ||
abbb3b8e | 4238 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4239 | if (ret) |
4240 | total_errors++; | |
f2984462 | 4241 | } |
0b246afa | 4242 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4243 | if (total_errors > max_errors) { |
0b246afa JM |
4244 | btrfs_handle_fs_error(fs_info, -EIO, |
4245 | "%d errors while writing supers", | |
4246 | total_errors); | |
79787eaa | 4247 | return -EIO; |
a236aed1 | 4248 | } |
f2984462 CM |
4249 | return 0; |
4250 | } | |
4251 | ||
cb517eab MX |
4252 | /* Drop a fs root from the radix tree and free it. */ |
4253 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4254 | struct btrfs_root *root) | |
2619ba1f | 4255 | { |
4785e24f JB |
4256 | bool drop_ref = false; |
4257 | ||
4df27c4d | 4258 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
4259 | radix_tree_delete(&fs_info->fs_roots_radix, |
4260 | (unsigned long)root->root_key.objectid); | |
af01d2e5 | 4261 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
4785e24f | 4262 | drop_ref = true; |
4df27c4d | 4263 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4264 | |
1c1ea4f7 | 4265 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
ef67963d | 4266 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4267 | if (root->reloc_root) { |
00246528 | 4268 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4269 | root->reloc_root = NULL; |
4270 | } | |
4271 | } | |
3321719e | 4272 | |
4785e24f JB |
4273 | if (drop_ref) |
4274 | btrfs_put_root(root); | |
2619ba1f CM |
4275 | } |
4276 | ||
c146afad | 4277 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4278 | { |
c146afad YZ |
4279 | u64 root_objectid = 0; |
4280 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
4281 | int i = 0; |
4282 | int err = 0; | |
4283 | unsigned int ret = 0; | |
e089f05c | 4284 | |
c146afad | 4285 | while (1) { |
efc34534 | 4286 | spin_lock(&fs_info->fs_roots_radix_lock); |
c146afad YZ |
4287 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
4288 | (void **)gang, root_objectid, | |
4289 | ARRAY_SIZE(gang)); | |
65d33fd7 | 4290 | if (!ret) { |
efc34534 | 4291 | spin_unlock(&fs_info->fs_roots_radix_lock); |
c146afad | 4292 | break; |
65d33fd7 | 4293 | } |
5d4f98a2 | 4294 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4295 | |
c146afad | 4296 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
4297 | /* Avoid to grab roots in dead_roots */ |
4298 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4299 | gang[i] = NULL; | |
4300 | continue; | |
4301 | } | |
4302 | /* grab all the search result for later use */ | |
00246528 | 4303 | gang[i] = btrfs_grab_root(gang[i]); |
65d33fd7 | 4304 | } |
efc34534 | 4305 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4306 | |
65d33fd7 QW |
4307 | for (i = 0; i < ret; i++) { |
4308 | if (!gang[i]) | |
4309 | continue; | |
c146afad | 4310 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
4311 | err = btrfs_orphan_cleanup(gang[i]); |
4312 | if (err) | |
65d33fd7 | 4313 | break; |
00246528 | 4314 | btrfs_put_root(gang[i]); |
c146afad YZ |
4315 | } |
4316 | root_objectid++; | |
4317 | } | |
65d33fd7 QW |
4318 | |
4319 | /* release the uncleaned roots due to error */ | |
4320 | for (; i < ret; i++) { | |
4321 | if (gang[i]) | |
00246528 | 4322 | btrfs_put_root(gang[i]); |
65d33fd7 QW |
4323 | } |
4324 | return err; | |
c146afad | 4325 | } |
a2135011 | 4326 | |
6bccf3ab | 4327 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4328 | { |
6bccf3ab | 4329 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4330 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4331 | |
0b246afa | 4332 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4333 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4334 | mutex_unlock(&fs_info->cleaner_mutex); |
4335 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4336 | |
4337 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4338 | down_write(&fs_info->cleanup_work_sem); |
4339 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4340 | |
7a7eaa40 | 4341 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4342 | if (IS_ERR(trans)) |
4343 | return PTR_ERR(trans); | |
3a45bb20 | 4344 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4345 | } |
4346 | ||
b105e927 | 4347 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4348 | { |
c146afad YZ |
4349 | int ret; |
4350 | ||
afcdd129 | 4351 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
d6fd0ae2 OS |
4352 | /* |
4353 | * We don't want the cleaner to start new transactions, add more delayed | |
4354 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4355 | * because that frees the task_struct, and the transaction kthread might | |
4356 | * still try to wake up the cleaner. | |
4357 | */ | |
4358 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4359 | |
78db1921 JB |
4360 | /* |
4361 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4362 | * clear that bit and wake up relocation so it can stop. | |
4363 | */ | |
4364 | btrfs_wake_unfinished_drop(fs_info); | |
4365 | ||
7343dd61 | 4366 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4367 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4368 | |
803b2f54 SB |
4369 | /* wait for the uuid_scan task to finish */ |
4370 | down(&fs_info->uuid_tree_rescan_sem); | |
4371 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4372 | up(&fs_info->uuid_tree_rescan_sem); | |
4373 | ||
837d5b6e | 4374 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4375 | btrfs_pause_balance(fs_info); |
837d5b6e | 4376 | |
8dabb742 SB |
4377 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4378 | ||
aa1b8cd4 | 4379 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4380 | |
4381 | /* wait for any defraggers to finish */ | |
4382 | wait_event(fs_info->transaction_wait, | |
4383 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4384 | ||
4385 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4386 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4387 | |
21c7e756 | 4388 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4389 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4390 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4391 | |
18bb8bbf JT |
4392 | cancel_work_sync(&fs_info->reclaim_bgs_work); |
4393 | ||
b0643e59 DZ |
4394 | /* Cancel or finish ongoing discard work */ |
4395 | btrfs_discard_cleanup(fs_info); | |
4396 | ||
bc98a42c | 4397 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4398 | /* |
d6fd0ae2 OS |
4399 | * The cleaner kthread is stopped, so do one final pass over |
4400 | * unused block groups. | |
e44163e1 | 4401 | */ |
0b246afa | 4402 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4403 | |
f0cc2cd7 FM |
4404 | /* |
4405 | * There might be existing delayed inode workers still running | |
4406 | * and holding an empty delayed inode item. We must wait for | |
4407 | * them to complete first because they can create a transaction. | |
4408 | * This happens when someone calls btrfs_balance_delayed_items() | |
4409 | * and then a transaction commit runs the same delayed nodes | |
4410 | * before any delayed worker has done something with the nodes. | |
4411 | * We must wait for any worker here and not at transaction | |
4412 | * commit time since that could cause a deadlock. | |
4413 | * This is a very rare case. | |
4414 | */ | |
4415 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4416 | ||
6bccf3ab | 4417 | ret = btrfs_commit_super(fs_info); |
acce952b | 4418 | if (ret) |
04892340 | 4419 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4420 | } |
4421 | ||
af722733 LB |
4422 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || |
4423 | test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) | |
2ff7e61e | 4424 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4425 | |
e3029d9f AV |
4426 | kthread_stop(fs_info->transaction_kthread); |
4427 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4428 | |
e187831e | 4429 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4430 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4431 | |
5958253c QW |
4432 | if (btrfs_check_quota_leak(fs_info)) { |
4433 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4434 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4435 | } | |
4436 | ||
04892340 | 4437 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4438 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4439 | |
963d678b | 4440 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4441 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4442 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4443 | } |
bcc63abb | 4444 | |
5deb17e1 | 4445 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4446 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4447 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4448 | |
6618a59b | 4449 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4450 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4451 | |
1a4319cc LB |
4452 | btrfs_put_block_group_cache(fs_info); |
4453 | ||
de348ee0 WS |
4454 | /* |
4455 | * we must make sure there is not any read request to | |
4456 | * submit after we stopping all workers. | |
4457 | */ | |
4458 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4459 | btrfs_stop_all_workers(fs_info); |
4460 | ||
0a31daa4 FM |
4461 | /* We shouldn't have any transaction open at this point */ |
4462 | ASSERT(list_empty(&fs_info->trans_list)); | |
4463 | ||
afcdd129 | 4464 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4465 | free_root_pointers(fs_info, true); |
8c38938c | 4466 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4467 | |
4e19443d JB |
4468 | /* |
4469 | * We must free the block groups after dropping the fs_roots as we could | |
4470 | * have had an IO error and have left over tree log blocks that aren't | |
4471 | * cleaned up until the fs roots are freed. This makes the block group | |
4472 | * accounting appear to be wrong because there's pending reserved bytes, | |
4473 | * so make sure we do the block group cleanup afterwards. | |
4474 | */ | |
4475 | btrfs_free_block_groups(fs_info); | |
4476 | ||
13e6c37b | 4477 | iput(fs_info->btree_inode); |
d6bfde87 | 4478 | |
21adbd5c | 4479 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4480 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4481 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4482 | #endif |
4483 | ||
0b86a832 | 4484 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4485 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4486 | } |
4487 | ||
b9fab919 CM |
4488 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4489 | int atomic) | |
5f39d397 | 4490 | { |
1259ab75 | 4491 | int ret; |
727011e0 | 4492 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4493 | |
0b32f4bb | 4494 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4495 | if (!ret) |
4496 | return ret; | |
4497 | ||
4498 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4499 | parent_transid, atomic); |
4500 | if (ret == -EAGAIN) | |
4501 | return ret; | |
1259ab75 | 4502 | return !ret; |
5f39d397 CM |
4503 | } |
4504 | ||
5f39d397 CM |
4505 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4506 | { | |
2f4d60df | 4507 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4508 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4509 | int was_dirty; |
b4ce94de | 4510 | |
06ea65a3 JB |
4511 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4512 | /* | |
4513 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4514 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4515 | * outside of the sanity tests. |
4516 | */ | |
b0132a3b | 4517 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4518 | return; |
4519 | #endif | |
b9447ef8 | 4520 | btrfs_assert_tree_locked(buf); |
0b246afa | 4521 | if (transid != fs_info->generation) |
5d163e0e | 4522 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4523 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4524 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4525 | if (!was_dirty) |
104b4e51 NB |
4526 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4527 | buf->len, | |
4528 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4529 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4530 | /* |
4531 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4532 | * but item data not updated. | |
4533 | * So here we should only check item pointers, not item data. | |
4534 | */ | |
4535 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4536 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4537 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4538 | ASSERT(0); |
4539 | } | |
4540 | #endif | |
eb60ceac CM |
4541 | } |
4542 | ||
2ff7e61e | 4543 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4544 | int flush_delayed) |
16cdcec7 MX |
4545 | { |
4546 | /* | |
4547 | * looks as though older kernels can get into trouble with | |
4548 | * this code, they end up stuck in balance_dirty_pages forever | |
4549 | */ | |
e2d84521 | 4550 | int ret; |
16cdcec7 MX |
4551 | |
4552 | if (current->flags & PF_MEMALLOC) | |
4553 | return; | |
4554 | ||
b53d3f5d | 4555 | if (flush_delayed) |
2ff7e61e | 4556 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4557 | |
d814a491 EL |
4558 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4559 | BTRFS_DIRTY_METADATA_THRESH, | |
4560 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4561 | if (ret > 0) { |
0b246afa | 4562 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4563 | } |
16cdcec7 MX |
4564 | } |
4565 | ||
2ff7e61e | 4566 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4567 | { |
2ff7e61e | 4568 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4569 | } |
585ad2c3 | 4570 | |
2ff7e61e | 4571 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4572 | { |
2ff7e61e | 4573 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4574 | } |
6b80053d | 4575 | |
581c1760 QW |
4576 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4577 | struct btrfs_key *first_key) | |
6b80053d | 4578 | { |
5ab12d1f | 4579 | return btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 4580 | level, first_key); |
6b80053d | 4581 | } |
0da5468f | 4582 | |
2ff7e61e | 4583 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4584 | { |
fe816d0f NB |
4585 | /* cleanup FS via transaction */ |
4586 | btrfs_cleanup_transaction(fs_info); | |
4587 | ||
0b246afa | 4588 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4589 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4590 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4591 | |
0b246afa JM |
4592 | down_write(&fs_info->cleanup_work_sem); |
4593 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4594 | } |
4595 | ||
ef67963d JB |
4596 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4597 | { | |
4598 | struct btrfs_root *gang[8]; | |
4599 | u64 root_objectid = 0; | |
4600 | int ret; | |
4601 | ||
4602 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4603 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4604 | (void **)gang, root_objectid, | |
4605 | ARRAY_SIZE(gang))) != 0) { | |
4606 | int i; | |
4607 | ||
4608 | for (i = 0; i < ret; i++) | |
4609 | gang[i] = btrfs_grab_root(gang[i]); | |
4610 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4611 | ||
4612 | for (i = 0; i < ret; i++) { | |
4613 | if (!gang[i]) | |
4614 | continue; | |
4615 | root_objectid = gang[i]->root_key.objectid; | |
4616 | btrfs_free_log(NULL, gang[i]); | |
4617 | btrfs_put_root(gang[i]); | |
4618 | } | |
4619 | root_objectid++; | |
4620 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4621 | } | |
4622 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4623 | btrfs_free_log_root_tree(NULL, fs_info); | |
4624 | } | |
4625 | ||
143bede5 | 4626 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4627 | { |
acce952b | 4628 | struct btrfs_ordered_extent *ordered; |
acce952b | 4629 | |
199c2a9c | 4630 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4631 | /* |
4632 | * This will just short circuit the ordered completion stuff which will | |
4633 | * make sure the ordered extent gets properly cleaned up. | |
4634 | */ | |
199c2a9c | 4635 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4636 | root_extent_list) |
4637 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4638 | spin_unlock(&root->ordered_extent_lock); |
4639 | } | |
4640 | ||
4641 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4642 | { | |
4643 | struct btrfs_root *root; | |
4644 | struct list_head splice; | |
4645 | ||
4646 | INIT_LIST_HEAD(&splice); | |
4647 | ||
4648 | spin_lock(&fs_info->ordered_root_lock); | |
4649 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4650 | while (!list_empty(&splice)) { | |
4651 | root = list_first_entry(&splice, struct btrfs_root, | |
4652 | ordered_root); | |
1de2cfde JB |
4653 | list_move_tail(&root->ordered_root, |
4654 | &fs_info->ordered_roots); | |
199c2a9c | 4655 | |
2a85d9ca | 4656 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4657 | btrfs_destroy_ordered_extents(root); |
4658 | ||
2a85d9ca LB |
4659 | cond_resched(); |
4660 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4661 | } |
4662 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4663 | |
4664 | /* | |
4665 | * We need this here because if we've been flipped read-only we won't | |
4666 | * get sync() from the umount, so we need to make sure any ordered | |
4667 | * extents that haven't had their dirty pages IO start writeout yet | |
4668 | * actually get run and error out properly. | |
4669 | */ | |
4670 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4671 | } |
4672 | ||
35a3621b | 4673 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4674 | struct btrfs_fs_info *fs_info) |
acce952b | 4675 | { |
4676 | struct rb_node *node; | |
4677 | struct btrfs_delayed_ref_root *delayed_refs; | |
4678 | struct btrfs_delayed_ref_node *ref; | |
4679 | int ret = 0; | |
4680 | ||
4681 | delayed_refs = &trans->delayed_refs; | |
4682 | ||
4683 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4684 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4685 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4686 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4687 | return ret; |
4688 | } | |
4689 | ||
5c9d028b | 4690 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4691 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4692 | struct rb_node *n; |
e78417d1 | 4693 | bool pin_bytes = false; |
acce952b | 4694 | |
d7df2c79 JB |
4695 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4696 | href_node); | |
3069bd26 | 4697 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4698 | continue; |
3069bd26 | 4699 | |
d7df2c79 | 4700 | spin_lock(&head->lock); |
e3d03965 | 4701 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4702 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4703 | ref_node); | |
d7df2c79 | 4704 | ref->in_tree = 0; |
e3d03965 | 4705 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4706 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4707 | if (!list_empty(&ref->add_list)) |
4708 | list_del(&ref->add_list); | |
d7df2c79 JB |
4709 | atomic_dec(&delayed_refs->num_entries); |
4710 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4711 | } |
d7df2c79 JB |
4712 | if (head->must_insert_reserved) |
4713 | pin_bytes = true; | |
4714 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4715 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4716 | spin_unlock(&head->lock); |
4717 | spin_unlock(&delayed_refs->lock); | |
4718 | mutex_unlock(&head->mutex); | |
acce952b | 4719 | |
f603bb94 NB |
4720 | if (pin_bytes) { |
4721 | struct btrfs_block_group *cache; | |
4722 | ||
4723 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4724 | BUG_ON(!cache); | |
4725 | ||
4726 | spin_lock(&cache->space_info->lock); | |
4727 | spin_lock(&cache->lock); | |
4728 | cache->pinned += head->num_bytes; | |
4729 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4730 | cache->space_info, head->num_bytes); | |
4731 | cache->reserved -= head->num_bytes; | |
4732 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4733 | spin_unlock(&cache->lock); | |
4734 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4735 | |
4736 | btrfs_put_block_group(cache); | |
4737 | ||
4738 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4739 | head->bytenr + head->num_bytes - 1); | |
4740 | } | |
31890da0 | 4741 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4742 | btrfs_put_delayed_ref_head(head); |
acce952b | 4743 | cond_resched(); |
4744 | spin_lock(&delayed_refs->lock); | |
4745 | } | |
81f7eb00 | 4746 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4747 | |
4748 | spin_unlock(&delayed_refs->lock); | |
4749 | ||
4750 | return ret; | |
4751 | } | |
4752 | ||
143bede5 | 4753 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4754 | { |
4755 | struct btrfs_inode *btrfs_inode; | |
4756 | struct list_head splice; | |
4757 | ||
4758 | INIT_LIST_HEAD(&splice); | |
4759 | ||
eb73c1b7 MX |
4760 | spin_lock(&root->delalloc_lock); |
4761 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4762 | |
4763 | while (!list_empty(&splice)) { | |
fe816d0f | 4764 | struct inode *inode = NULL; |
eb73c1b7 MX |
4765 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4766 | delalloc_inodes); | |
fe816d0f | 4767 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4768 | spin_unlock(&root->delalloc_lock); |
acce952b | 4769 | |
fe816d0f NB |
4770 | /* |
4771 | * Make sure we get a live inode and that it'll not disappear | |
4772 | * meanwhile. | |
4773 | */ | |
4774 | inode = igrab(&btrfs_inode->vfs_inode); | |
4775 | if (inode) { | |
4776 | invalidate_inode_pages2(inode->i_mapping); | |
4777 | iput(inode); | |
4778 | } | |
eb73c1b7 | 4779 | spin_lock(&root->delalloc_lock); |
acce952b | 4780 | } |
eb73c1b7 MX |
4781 | spin_unlock(&root->delalloc_lock); |
4782 | } | |
4783 | ||
4784 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4785 | { | |
4786 | struct btrfs_root *root; | |
4787 | struct list_head splice; | |
4788 | ||
4789 | INIT_LIST_HEAD(&splice); | |
4790 | ||
4791 | spin_lock(&fs_info->delalloc_root_lock); | |
4792 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4793 | while (!list_empty(&splice)) { | |
4794 | root = list_first_entry(&splice, struct btrfs_root, | |
4795 | delalloc_root); | |
00246528 | 4796 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4797 | BUG_ON(!root); |
4798 | spin_unlock(&fs_info->delalloc_root_lock); | |
4799 | ||
4800 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4801 | btrfs_put_root(root); |
eb73c1b7 MX |
4802 | |
4803 | spin_lock(&fs_info->delalloc_root_lock); | |
4804 | } | |
4805 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4806 | } |
4807 | ||
2ff7e61e | 4808 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4809 | struct extent_io_tree *dirty_pages, |
4810 | int mark) | |
4811 | { | |
4812 | int ret; | |
acce952b | 4813 | struct extent_buffer *eb; |
4814 | u64 start = 0; | |
4815 | u64 end; | |
acce952b | 4816 | |
4817 | while (1) { | |
4818 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4819 | mark, NULL); |
acce952b | 4820 | if (ret) |
4821 | break; | |
4822 | ||
91166212 | 4823 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4824 | while (start <= end) { |
0b246afa JM |
4825 | eb = find_extent_buffer(fs_info, start); |
4826 | start += fs_info->nodesize; | |
fd8b2b61 | 4827 | if (!eb) |
acce952b | 4828 | continue; |
fd8b2b61 | 4829 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4830 | |
fd8b2b61 JB |
4831 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4832 | &eb->bflags)) | |
4833 | clear_extent_buffer_dirty(eb); | |
4834 | free_extent_buffer_stale(eb); | |
acce952b | 4835 | } |
4836 | } | |
4837 | ||
4838 | return ret; | |
4839 | } | |
4840 | ||
2ff7e61e | 4841 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 4842 | struct extent_io_tree *unpin) |
acce952b | 4843 | { |
acce952b | 4844 | u64 start; |
4845 | u64 end; | |
4846 | int ret; | |
4847 | ||
acce952b | 4848 | while (1) { |
0e6ec385 FM |
4849 | struct extent_state *cached_state = NULL; |
4850 | ||
fcd5e742 LF |
4851 | /* |
4852 | * The btrfs_finish_extent_commit() may get the same range as | |
4853 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4854 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4855 | * the same extent range. | |
4856 | */ | |
4857 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4858 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 4859 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
4860 | if (ret) { |
4861 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 4862 | break; |
fcd5e742 | 4863 | } |
acce952b | 4864 | |
0e6ec385 FM |
4865 | clear_extent_dirty(unpin, start, end, &cached_state); |
4866 | free_extent_state(cached_state); | |
2ff7e61e | 4867 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4868 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4869 | cond_resched(); |
4870 | } | |
4871 | ||
4872 | return 0; | |
4873 | } | |
4874 | ||
32da5386 | 4875 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4876 | { |
4877 | struct inode *inode; | |
4878 | ||
4879 | inode = cache->io_ctl.inode; | |
4880 | if (inode) { | |
4881 | invalidate_inode_pages2(inode->i_mapping); | |
4882 | BTRFS_I(inode)->generation = 0; | |
4883 | cache->io_ctl.inode = NULL; | |
4884 | iput(inode); | |
4885 | } | |
bbc37d6e | 4886 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4887 | btrfs_put_block_group(cache); |
4888 | } | |
4889 | ||
4890 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4891 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4892 | { |
32da5386 | 4893 | struct btrfs_block_group *cache; |
c79a1751 LB |
4894 | |
4895 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4896 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4897 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4898 | struct btrfs_block_group, |
c79a1751 | 4899 | dirty_list); |
c79a1751 LB |
4900 | |
4901 | if (!list_empty(&cache->io_list)) { | |
4902 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4903 | list_del_init(&cache->io_list); | |
4904 | btrfs_cleanup_bg_io(cache); | |
4905 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4906 | } | |
4907 | ||
4908 | list_del_init(&cache->dirty_list); | |
4909 | spin_lock(&cache->lock); | |
4910 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4911 | spin_unlock(&cache->lock); | |
4912 | ||
4913 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4914 | btrfs_put_block_group(cache); | |
ba2c4d4e | 4915 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
4916 | spin_lock(&cur_trans->dirty_bgs_lock); |
4917 | } | |
4918 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4919 | ||
45ae2c18 NB |
4920 | /* |
4921 | * Refer to the definition of io_bgs member for details why it's safe | |
4922 | * to use it without any locking | |
4923 | */ | |
c79a1751 LB |
4924 | while (!list_empty(&cur_trans->io_bgs)) { |
4925 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4926 | struct btrfs_block_group, |
c79a1751 | 4927 | io_list); |
c79a1751 LB |
4928 | |
4929 | list_del_init(&cache->io_list); | |
4930 | spin_lock(&cache->lock); | |
4931 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4932 | spin_unlock(&cache->lock); | |
4933 | btrfs_cleanup_bg_io(cache); | |
4934 | } | |
4935 | } | |
4936 | ||
49b25e05 | 4937 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4938 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4939 | { |
bbbf7243 NB |
4940 | struct btrfs_device *dev, *tmp; |
4941 | ||
2ff7e61e | 4942 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4943 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4944 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4945 | ||
bbbf7243 NB |
4946 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4947 | post_commit_list) { | |
4948 | list_del_init(&dev->post_commit_list); | |
4949 | } | |
4950 | ||
2ff7e61e | 4951 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4952 | |
4a9d8bde | 4953 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4954 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4955 | |
4a9d8bde | 4956 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4957 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4958 | |
ccdf9b30 | 4959 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4960 | |
2ff7e61e | 4961 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4962 | EXTENT_DIRTY); |
fe119a6e | 4963 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4964 | |
d3575156 NA |
4965 | btrfs_free_redirty_list(cur_trans); |
4966 | ||
4a9d8bde MX |
4967 | cur_trans->state =TRANS_STATE_COMPLETED; |
4968 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4969 | } |
4970 | ||
2ff7e61e | 4971 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4972 | { |
4973 | struct btrfs_transaction *t; | |
acce952b | 4974 | |
0b246afa | 4975 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4976 | |
0b246afa JM |
4977 | spin_lock(&fs_info->trans_lock); |
4978 | while (!list_empty(&fs_info->trans_list)) { | |
4979 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4980 | struct btrfs_transaction, list); |
4981 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4982 | refcount_inc(&t->use_count); |
0b246afa | 4983 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4984 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4985 | btrfs_put_transaction(t); |
0b246afa | 4986 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4987 | continue; |
4988 | } | |
0b246afa | 4989 | if (t == fs_info->running_transaction) { |
724e2315 | 4990 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4991 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4992 | /* |
4993 | * We wait for 0 num_writers since we don't hold a trans | |
4994 | * handle open currently for this transaction. | |
4995 | */ | |
4996 | wait_event(t->writer_wait, | |
4997 | atomic_read(&t->num_writers) == 0); | |
4998 | } else { | |
0b246afa | 4999 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5000 | } |
2ff7e61e | 5001 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5002 | |
0b246afa JM |
5003 | spin_lock(&fs_info->trans_lock); |
5004 | if (t == fs_info->running_transaction) | |
5005 | fs_info->running_transaction = NULL; | |
acce952b | 5006 | list_del_init(&t->list); |
0b246afa | 5007 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5008 | |
724e2315 | 5009 | btrfs_put_transaction(t); |
2ff7e61e | 5010 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 5011 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5012 | } |
0b246afa JM |
5013 | spin_unlock(&fs_info->trans_lock); |
5014 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5015 | btrfs_destroy_delayed_inodes(fs_info); |
5016 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5017 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5018 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5019 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5020 | |
5021 | return 0; | |
5022 | } | |
ec7d6dfd | 5023 | |
453e4873 | 5024 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5025 | { |
5026 | struct btrfs_path *path; | |
5027 | int ret; | |
5028 | struct extent_buffer *l; | |
5029 | struct btrfs_key search_key; | |
5030 | struct btrfs_key found_key; | |
5031 | int slot; | |
5032 | ||
5033 | path = btrfs_alloc_path(); | |
5034 | if (!path) | |
5035 | return -ENOMEM; | |
5036 | ||
5037 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5038 | search_key.type = -1; | |
5039 | search_key.offset = (u64)-1; | |
5040 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5041 | if (ret < 0) | |
5042 | goto error; | |
5043 | BUG_ON(ret == 0); /* Corruption */ | |
5044 | if (path->slots[0] > 0) { | |
5045 | slot = path->slots[0] - 1; | |
5046 | l = path->nodes[0]; | |
5047 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5048 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5049 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5050 | } else { |
23125104 | 5051 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5052 | } |
5053 | ret = 0; | |
5054 | error: | |
5055 | btrfs_free_path(path); | |
5056 | return ret; | |
5057 | } | |
5058 | ||
543068a2 | 5059 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5060 | { |
5061 | int ret; | |
5062 | mutex_lock(&root->objectid_mutex); | |
5063 | ||
6b8fad57 | 5064 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5065 | btrfs_warn(root->fs_info, |
5066 | "the objectid of root %llu reaches its highest value", | |
5067 | root->root_key.objectid); | |
5068 | ret = -ENOSPC; | |
5069 | goto out; | |
5070 | } | |
5071 | ||
23125104 | 5072 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5073 | ret = 0; |
5074 | out: | |
5075 | mutex_unlock(&root->objectid_mutex); | |
5076 | return ret; | |
5077 | } |