]>
Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
e20d96d6 | 6 | #include <linux/fs.h> |
d98237b3 | 7 | #include <linux/blkdev.h> |
fc7cbcd4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
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 | */ |
6a2e9dc4 FM |
377 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
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 | ||
3777369f 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, |
3777369f 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; | |
3777369f 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 | ||
8ee92268 | 489 | /* A dirty eb shouldn't disappear from extent_buffers */ |
eca0f6f6 QW |
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 | 521 | |
fbca46eb | 522 | if (fs_info->nodesize < PAGE_SIZE) |
eca0f6f6 | 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 | ||
371cdc07 QW |
682 | set_extent_buffer_uptodate(eb); |
683 | ||
684 | free_extent_buffer(eb); | |
685 | return ret; | |
686 | err: | |
687 | /* | |
688 | * end_bio_extent_readpage decrements io_pages in case of error, | |
689 | * make sure it has something to decrement. | |
690 | */ | |
691 | atomic_inc(&eb->io_pages); | |
692 | clear_extent_buffer_uptodate(eb); | |
693 | free_extent_buffer(eb); | |
694 | return ret; | |
695 | } | |
696 | ||
c3a3b19b | 697 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
698 | struct page *page, u64 start, u64 end, |
699 | int mirror) | |
700 | { | |
701 | struct extent_buffer *eb; | |
702 | int ret = 0; | |
703 | int reads_done; | |
704 | ||
705 | ASSERT(page->private); | |
371cdc07 | 706 | |
fbca46eb | 707 | if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) |
371cdc07 QW |
708 | return validate_subpage_buffer(page, start, end, mirror); |
709 | ||
77bf40a2 QW |
710 | eb = (struct extent_buffer *)page->private; |
711 | ||
712 | /* | |
713 | * The pending IO might have been the only thing that kept this buffer | |
714 | * in memory. Make sure we have a ref for all this other checks | |
715 | */ | |
716 | atomic_inc(&eb->refs); | |
717 | ||
718 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
719 | if (!reads_done) | |
720 | goto err; | |
721 | ||
722 | eb->read_mirror = mirror; | |
723 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
724 | ret = -EIO; | |
725 | goto err; | |
726 | } | |
727 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 728 | err: |
53b381b3 DW |
729 | if (ret) { |
730 | /* | |
731 | * our io error hook is going to dec the io pages | |
732 | * again, we have to make sure it has something | |
733 | * to decrement | |
734 | */ | |
735 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 736 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 737 | } |
0b32f4bb | 738 | free_extent_buffer(eb); |
77bf40a2 | 739 | |
f188591e | 740 | return ret; |
ce9adaa5 CM |
741 | } |
742 | ||
4246a0b6 | 743 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 744 | { |
97eb6b69 | 745 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 746 | struct btrfs_fs_info *fs_info; |
9e0af237 | 747 | struct btrfs_workqueue *wq; |
ce9adaa5 | 748 | |
ce9adaa5 | 749 | fs_info = end_io_wq->info; |
4e4cbee9 | 750 | end_io_wq->status = bio->bi_status; |
d20f7043 | 751 | |
cfe94440 | 752 | if (btrfs_op(bio) == BTRFS_MAP_WRITE) { |
a0cac0ec | 753 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 754 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 755 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 756 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 757 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 758 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 759 | else |
9e0af237 | 760 | wq = fs_info->endio_write_workers; |
d20f7043 | 761 | } else { |
5c047a69 | 762 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 763 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 764 | else if (end_io_wq->metadata) |
9e0af237 | 765 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 766 | else |
9e0af237 | 767 | wq = fs_info->endio_workers; |
d20f7043 | 768 | } |
9e0af237 | 769 | |
a0cac0ec | 770 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 771 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
772 | } |
773 | ||
4e4cbee9 | 774 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 775 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 776 | { |
97eb6b69 | 777 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 778 | |
97eb6b69 | 779 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 780 | if (!end_io_wq) |
4e4cbee9 | 781 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
782 | |
783 | end_io_wq->private = bio->bi_private; | |
784 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 785 | end_io_wq->info = info; |
4e4cbee9 | 786 | end_io_wq->status = 0; |
ce9adaa5 | 787 | end_io_wq->bio = bio; |
22c59948 | 788 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
789 | |
790 | bio->bi_private = end_io_wq; | |
791 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
792 | return 0; |
793 | } | |
794 | ||
4a69a410 CM |
795 | static void run_one_async_start(struct btrfs_work *work) |
796 | { | |
4a69a410 | 797 | struct async_submit_bio *async; |
4e4cbee9 | 798 | blk_status_t ret; |
4a69a410 CM |
799 | |
800 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
801 | ret = async->submit_bio_start(async->inode, async->bio, |
802 | async->dio_file_offset); | |
79787eaa | 803 | if (ret) |
4e4cbee9 | 804 | async->status = ret; |
4a69a410 CM |
805 | } |
806 | ||
06ea01b1 DS |
807 | /* |
808 | * In order to insert checksums into the metadata in large chunks, we wait | |
809 | * until bio submission time. All the pages in the bio are checksummed and | |
810 | * sums are attached onto the ordered extent record. | |
811 | * | |
812 | * At IO completion time the csums attached on the ordered extent record are | |
813 | * inserted into the tree. | |
814 | */ | |
4a69a410 | 815 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 816 | { |
8b712842 | 817 | struct async_submit_bio *async; |
06ea01b1 DS |
818 | struct inode *inode; |
819 | blk_status_t ret; | |
8b712842 CM |
820 | |
821 | async = container_of(work, struct async_submit_bio, work); | |
8896a08d | 822 | inode = async->inode; |
4854ddd0 | 823 | |
bb7ab3b9 | 824 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
825 | if (async->status) { |
826 | async->bio->bi_status = async->status; | |
4246a0b6 | 827 | bio_endio(async->bio); |
79787eaa JM |
828 | return; |
829 | } | |
830 | ||
ec39f769 CM |
831 | /* |
832 | * All of the bios that pass through here are from async helpers. | |
833 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
834 | * This changes nothing when cgroups aren't in use. | |
835 | */ | |
836 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 837 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
838 | if (ret) { |
839 | async->bio->bi_status = ret; | |
840 | bio_endio(async->bio); | |
841 | } | |
4a69a410 CM |
842 | } |
843 | ||
844 | static void run_one_async_free(struct btrfs_work *work) | |
845 | { | |
846 | struct async_submit_bio *async; | |
847 | ||
848 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
849 | kfree(async); |
850 | } | |
851 | ||
8896a08d | 852 | blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, |
a6f5e39e | 853 | int mirror_num, u64 dio_file_offset, |
e288c080 | 854 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e | 855 | { |
8896a08d | 856 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
857 | struct async_submit_bio *async; |
858 | ||
859 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
860 | if (!async) | |
4e4cbee9 | 861 | return BLK_STS_RESOURCE; |
44b8bd7e | 862 | |
8896a08d | 863 | async->inode = inode; |
44b8bd7e CM |
864 | async->bio = bio; |
865 | async->mirror_num = mirror_num; | |
4a69a410 | 866 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 867 | |
a0cac0ec OS |
868 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
869 | run_one_async_free); | |
4a69a410 | 870 | |
1941b64b | 871 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 872 | |
4e4cbee9 | 873 | async->status = 0; |
79787eaa | 874 | |
67f055c7 | 875 | if (op_is_sync(bio->bi_opf)) |
a31b4a43 CH |
876 | btrfs_queue_work(fs_info->hipri_workers, &async->work); |
877 | else | |
878 | btrfs_queue_work(fs_info->workers, &async->work); | |
44b8bd7e CM |
879 | return 0; |
880 | } | |
881 | ||
4e4cbee9 | 882 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 883 | { |
2c30c71b | 884 | struct bio_vec *bvec; |
ce3ed71a | 885 | struct btrfs_root *root; |
2b070cfe | 886 | int ret = 0; |
6dc4f100 | 887 | struct bvec_iter_all iter_all; |
ce3ed71a | 888 | |
c09abff8 | 889 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 890 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 891 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 892 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
893 | if (ret) |
894 | break; | |
ce3ed71a | 895 | } |
2c30c71b | 896 | |
4e4cbee9 | 897 | return errno_to_blk_status(ret); |
ce3ed71a CM |
898 | } |
899 | ||
8896a08d | 900 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 901 | u64 dio_file_offset) |
22c59948 | 902 | { |
8b712842 CM |
903 | /* |
904 | * when we're called for a write, we're already in the async | |
5443be45 | 905 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 906 | */ |
79787eaa | 907 | return btree_csum_one_bio(bio); |
4a69a410 | 908 | } |
22c59948 | 909 | |
f4dcfb30 | 910 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 911 | struct btrfs_inode *bi) |
de0022b9 | 912 | { |
4eef29ef | 913 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 914 | return false; |
6300463b | 915 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 916 | return false; |
9b4e675a | 917 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
918 | return false; |
919 | return true; | |
de0022b9 JB |
920 | } |
921 | ||
94d9e11b | 922 | void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) |
44b8bd7e | 923 | { |
0b246afa | 924 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4e4cbee9 | 925 | blk_status_t ret; |
cad321ad | 926 | |
cfe94440 | 927 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
4a69a410 CM |
928 | /* |
929 | * called for a read, do the setup so that checksum validation | |
930 | * can happen in the async kernel threads | |
931 | */ | |
0b246afa JM |
932 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
933 | BTRFS_WQ_ENDIO_METADATA); | |
94d9e11b CH |
934 | if (!ret) |
935 | ret = btrfs_map_bio(fs_info, bio, mirror_num); | |
f4dcfb30 | 936 | } else if (!should_async_write(fs_info, BTRFS_I(inode))) { |
de0022b9 | 937 | ret = btree_csum_one_bio(bio); |
94d9e11b CH |
938 | if (!ret) |
939 | ret = btrfs_map_bio(fs_info, bio, mirror_num); | |
61891923 SB |
940 | } else { |
941 | /* | |
942 | * kthread helpers are used to submit writes so that | |
943 | * checksumming can happen in parallel across all CPUs | |
944 | */ | |
8896a08d | 945 | ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, |
a6f5e39e | 946 | btree_submit_bio_start); |
44b8bd7e | 947 | } |
d313d7a3 | 948 | |
94d9e11b CH |
949 | if (ret) { |
950 | bio->bi_status = ret; | |
951 | bio_endio(bio); | |
952 | } | |
44b8bd7e CM |
953 | } |
954 | ||
3dd1462e | 955 | #ifdef CONFIG_MIGRATION |
784b4e29 | 956 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
957 | struct page *newpage, struct page *page, |
958 | enum migrate_mode mode) | |
784b4e29 CM |
959 | { |
960 | /* | |
961 | * we can't safely write a btree page from here, | |
962 | * we haven't done the locking hook | |
963 | */ | |
964 | if (PageDirty(page)) | |
965 | return -EAGAIN; | |
966 | /* | |
967 | * Buffers may be managed in a filesystem specific way. | |
968 | * We must have no buffers or drop them. | |
969 | */ | |
970 | if (page_has_private(page) && | |
971 | !try_to_release_page(page, GFP_KERNEL)) | |
972 | return -EAGAIN; | |
a6bc32b8 | 973 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 974 | } |
3dd1462e | 975 | #endif |
784b4e29 | 976 | |
0da5468f CM |
977 | |
978 | static int btree_writepages(struct address_space *mapping, | |
979 | struct writeback_control *wbc) | |
980 | { | |
e2d84521 MX |
981 | struct btrfs_fs_info *fs_info; |
982 | int ret; | |
983 | ||
d8d5f3e1 | 984 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
985 | |
986 | if (wbc->for_kupdate) | |
987 | return 0; | |
988 | ||
e2d84521 | 989 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 990 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
991 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
992 | BTRFS_DIRTY_METADATA_THRESH, | |
993 | fs_info->dirty_metadata_batch); | |
e2d84521 | 994 | if (ret < 0) |
793955bc | 995 | return 0; |
793955bc | 996 | } |
0b32f4bb | 997 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
998 | } |
999 | ||
70dec807 | 1000 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1001 | { |
98509cfc | 1002 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1003 | return 0; |
0c4e538b | 1004 | |
f7a52a40 | 1005 | return try_release_extent_buffer(page); |
d98237b3 CM |
1006 | } |
1007 | ||
895586eb MWO |
1008 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
1009 | size_t length) | |
d98237b3 | 1010 | { |
d1310b2e | 1011 | struct extent_io_tree *tree; |
895586eb MWO |
1012 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
1013 | extent_invalidate_folio(tree, folio, offset); | |
1014 | btree_releasepage(&folio->page, GFP_NOFS); | |
1015 | if (folio_get_private(folio)) { | |
1016 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
1017 | "folio private not zero on folio %llu", | |
1018 | (unsigned long long)folio_pos(folio)); | |
1019 | folio_detach_private(folio); | |
9ad6b7bc | 1020 | } |
d98237b3 CM |
1021 | } |
1022 | ||
bb146eb2 | 1023 | #ifdef DEBUG |
0079c3b1 MWO |
1024 | static bool btree_dirty_folio(struct address_space *mapping, |
1025 | struct folio *folio) | |
1026 | { | |
1027 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
139e8cd3 | 1028 | struct btrfs_subpage *subpage; |
0b32f4bb | 1029 | struct extent_buffer *eb; |
139e8cd3 | 1030 | int cur_bit = 0; |
0079c3b1 | 1031 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
1032 | |
1033 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 1034 | eb = folio_get_private(folio); |
139e8cd3 QW |
1035 | BUG_ON(!eb); |
1036 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1037 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 1038 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 1039 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 1040 | } |
0079c3b1 | 1041 | subpage = folio_get_private(folio); |
139e8cd3 QW |
1042 | |
1043 | ASSERT(subpage->dirty_bitmap); | |
1044 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
1045 | unsigned long flags; | |
1046 | u64 cur; | |
1047 | u16 tmp = (1 << cur_bit); | |
1048 | ||
1049 | spin_lock_irqsave(&subpage->lock, flags); | |
1050 | if (!(tmp & subpage->dirty_bitmap)) { | |
1051 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1052 | cur_bit++; | |
1053 | continue; | |
1054 | } | |
1055 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1056 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 1057 | |
139e8cd3 QW |
1058 | eb = find_extent_buffer(fs_info, cur); |
1059 | ASSERT(eb); | |
1060 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1061 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 1062 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
1063 | free_extent_buffer(eb); |
1064 | ||
1065 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
1066 | } | |
0079c3b1 | 1067 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 1068 | } |
0079c3b1 MWO |
1069 | #else |
1070 | #define btree_dirty_folio filemap_dirty_folio | |
1071 | #endif | |
0b32f4bb | 1072 | |
7f09410b | 1073 | static const struct address_space_operations btree_aops = { |
0da5468f | 1074 | .writepages = btree_writepages, |
5f39d397 | 1075 | .releasepage = btree_releasepage, |
895586eb | 1076 | .invalidate_folio = btree_invalidate_folio, |
5a92bc88 | 1077 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1078 | .migratepage = btree_migratepage, |
5a92bc88 | 1079 | #endif |
0079c3b1 | 1080 | .dirty_folio = btree_dirty_folio, |
d98237b3 CM |
1081 | }; |
1082 | ||
2ff7e61e JM |
1083 | struct extent_buffer *btrfs_find_create_tree_block( |
1084 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
1085 | u64 bytenr, u64 owner_root, |
1086 | int level) | |
0999df54 | 1087 | { |
0b246afa JM |
1088 | if (btrfs_is_testing(fs_info)) |
1089 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 1090 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
1091 | } |
1092 | ||
581c1760 QW |
1093 | /* |
1094 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1095 | * verification. | |
1096 | * | |
1b7ec85e | 1097 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
1098 | * @parent_transid: expected transid of this tree block, skip check if 0 |
1099 | * @level: expected level, mandatory check | |
1100 | * @first_key: expected key in slot 0, skip check if NULL | |
1101 | */ | |
2ff7e61e | 1102 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
1103 | u64 owner_root, u64 parent_transid, |
1104 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
1105 | { |
1106 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1107 | int ret; |
1108 | ||
3fbaf258 | 1109 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
1110 | if (IS_ERR(buf)) |
1111 | return buf; | |
0999df54 | 1112 | |
6a2e9dc4 | 1113 | ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); |
0f0fe8f7 | 1114 | if (ret) { |
537f38f0 | 1115 | free_extent_buffer_stale(buf); |
64c043de | 1116 | return ERR_PTR(ret); |
0f0fe8f7 | 1117 | } |
88c602ab QW |
1118 | if (btrfs_check_eb_owner(buf, owner_root)) { |
1119 | free_extent_buffer_stale(buf); | |
1120 | return ERR_PTR(-EUCLEAN); | |
1121 | } | |
5f39d397 | 1122 | return buf; |
ce9adaa5 | 1123 | |
eb60ceac CM |
1124 | } |
1125 | ||
6a884d7d | 1126 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1127 | { |
6a884d7d | 1128 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1129 | if (btrfs_header_generation(buf) == |
e2d84521 | 1130 | fs_info->running_transaction->transid) { |
49d0c642 | 1131 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 1132 | |
b9473439 | 1133 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1134 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1135 | -buf->len, | |
1136 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1137 | clear_extent_buffer_dirty(buf); |
1138 | } | |
925baedd | 1139 | } |
5f39d397 CM |
1140 | } |
1141 | ||
da17066c | 1142 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1143 | u64 objectid) |
d97e63b6 | 1144 | { |
7c0260ee | 1145 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
1146 | |
1147 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
1148 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
1149 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 1150 | root->fs_info = fs_info; |
2e608bd1 | 1151 | root->root_key.objectid = objectid; |
cfaa7295 | 1152 | root->node = NULL; |
a28ec197 | 1153 | root->commit_root = NULL; |
27cdeb70 | 1154 | root->state = 0; |
abed4aaa | 1155 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 1156 | |
0f7d52f4 | 1157 | root->last_trans = 0; |
6b8fad57 | 1158 | root->free_objectid = 0; |
eb73c1b7 | 1159 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1160 | root->nr_ordered_extents = 0; |
6bef4d31 | 1161 | root->inode_tree = RB_ROOT; |
253bf575 | 1162 | xa_init_flags(&root->delayed_nodes, GFP_ATOMIC); |
2e608bd1 JB |
1163 | |
1164 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
1165 | |
1166 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1167 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1168 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1169 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1170 | INIT_LIST_HEAD(&root->ordered_extents); |
1171 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1172 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1173 | INIT_LIST_HEAD(&root->logged_list[0]); |
1174 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1175 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1176 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1177 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1178 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1179 | spin_lock_init(&root->log_extents_lock[0]); |
1180 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1181 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1182 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1183 | mutex_init(&root->log_mutex); |
31f3d255 | 1184 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1185 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1186 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1187 | init_waitqueue_head(&root->log_writer_wait); |
1188 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1189 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1190 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1191 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1192 | atomic_set(&root->log_commit[0], 0); |
1193 | atomic_set(&root->log_commit[1], 0); | |
1194 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1195 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1196 | refcount_set(&root->refs, 1); |
8ecebf4d | 1197 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1198 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1199 | root->log_transid = 0; |
d1433deb | 1200 | root->log_transid_committed = -1; |
257c62e1 | 1201 | root->last_log_commit = 0; |
2e608bd1 | 1202 | root->anon_dev = 0; |
e289f03e | 1203 | if (!dummy) { |
43eb5f29 QW |
1204 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1205 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1206 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1207 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1208 | } | |
017e5369 | 1209 | |
5f3ab90a | 1210 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1211 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1212 | #ifdef CONFIG_BTRFS_DEBUG |
1213 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 1214 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1215 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 1216 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1217 | #endif |
3768f368 CM |
1218 | } |
1219 | ||
74e4d827 | 1220 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1221 | u64 objectid, gfp_t flags) |
6f07e42e | 1222 | { |
74e4d827 | 1223 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1224 | if (root) |
96dfcb46 | 1225 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1226 | return root; |
1227 | } | |
1228 | ||
06ea65a3 JB |
1229 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1230 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1231 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1232 | { |
1233 | struct btrfs_root *root; | |
1234 | ||
7c0260ee JM |
1235 | if (!fs_info) |
1236 | return ERR_PTR(-EINVAL); | |
1237 | ||
96dfcb46 | 1238 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1239 | if (!root) |
1240 | return ERR_PTR(-ENOMEM); | |
da17066c | 1241 | |
b9ef22de | 1242 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1243 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1244 | |
1245 | return root; | |
1246 | } | |
1247 | #endif | |
1248 | ||
abed4aaa JB |
1249 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1250 | { | |
1251 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1252 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1253 | ||
1254 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1255 | } | |
1256 | ||
1257 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1258 | { | |
1259 | const struct btrfs_key *key = k; | |
1260 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1261 | ||
1262 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1263 | } | |
1264 | ||
1265 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1266 | { | |
1267 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1268 | struct rb_node *tmp; | |
1269 | ||
1270 | write_lock(&fs_info->global_root_lock); | |
1271 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1272 | write_unlock(&fs_info->global_root_lock); | |
1273 | ASSERT(!tmp); | |
1274 | ||
1275 | return tmp ? -EEXIST : 0; | |
1276 | } | |
1277 | ||
1278 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1279 | { | |
1280 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1281 | ||
1282 | write_lock(&fs_info->global_root_lock); | |
1283 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1284 | write_unlock(&fs_info->global_root_lock); | |
1285 | } | |
1286 | ||
1287 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1288 | struct btrfs_key *key) | |
1289 | { | |
1290 | struct rb_node *node; | |
1291 | struct btrfs_root *root = NULL; | |
1292 | ||
1293 | read_lock(&fs_info->global_root_lock); | |
1294 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1295 | if (node) | |
1296 | root = container_of(node, struct btrfs_root, rb_node); | |
1297 | read_unlock(&fs_info->global_root_lock); | |
1298 | ||
1299 | return root; | |
1300 | } | |
1301 | ||
f7238e50 JB |
1302 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
1303 | { | |
1304 | struct btrfs_block_group *block_group; | |
1305 | u64 ret; | |
1306 | ||
1307 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1308 | return 0; | |
1309 | ||
1310 | if (bytenr) | |
1311 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
1312 | else | |
1313 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
1314 | ASSERT(block_group); | |
1315 | if (!block_group) | |
1316 | return 0; | |
1317 | ret = block_group->global_root_id; | |
1318 | btrfs_put_block_group(block_group); | |
1319 | ||
1320 | return ret; | |
1321 | } | |
1322 | ||
abed4aaa JB |
1323 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
1324 | { | |
1325 | struct btrfs_key key = { | |
1326 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1327 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1328 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1329 | }; |
1330 | ||
1331 | return btrfs_global_root(fs_info, &key); | |
1332 | } | |
1333 | ||
1334 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1335 | { | |
1336 | struct btrfs_key key = { | |
1337 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1338 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1339 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1340 | }; |
1341 | ||
1342 | return btrfs_global_root(fs_info, &key); | |
1343 | } | |
1344 | ||
20897f5c | 1345 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1346 | u64 objectid) |
1347 | { | |
9b7a2440 | 1348 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1349 | struct extent_buffer *leaf; |
1350 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1351 | struct btrfs_root *root; | |
1352 | struct btrfs_key key; | |
b89f6d1f | 1353 | unsigned int nofs_flag; |
20897f5c | 1354 | int ret = 0; |
20897f5c | 1355 | |
b89f6d1f FM |
1356 | /* |
1357 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1358 | * context to avoid deadlock if reclaim happens. | |
1359 | */ | |
1360 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1361 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1362 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1363 | if (!root) |
1364 | return ERR_PTR(-ENOMEM); | |
1365 | ||
20897f5c AJ |
1366 | root->root_key.objectid = objectid; |
1367 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1368 | root->root_key.offset = 0; | |
1369 | ||
9631e4cc JB |
1370 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1371 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1372 | if (IS_ERR(leaf)) { |
1373 | ret = PTR_ERR(leaf); | |
1dd05682 | 1374 | leaf = NULL; |
8a6a87cd | 1375 | goto fail_unlock; |
20897f5c AJ |
1376 | } |
1377 | ||
20897f5c | 1378 | root->node = leaf; |
20897f5c AJ |
1379 | btrfs_mark_buffer_dirty(leaf); |
1380 | ||
1381 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1382 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1383 | |
f944d2cb DS |
1384 | btrfs_set_root_flags(&root->root_item, 0); |
1385 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1386 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1387 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1388 | btrfs_set_root_level(&root->root_item, 0); | |
1389 | btrfs_set_root_refs(&root->root_item, 1); | |
1390 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1391 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1392 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1393 | if (is_fstree(objectid)) |
807fc790 AS |
1394 | generate_random_guid(root->root_item.uuid); |
1395 | else | |
1396 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1397 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1398 | |
8a6a87cd BB |
1399 | btrfs_tree_unlock(leaf); |
1400 | ||
20897f5c AJ |
1401 | key.objectid = objectid; |
1402 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1403 | key.offset = 0; | |
1404 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1405 | if (ret) | |
1406 | goto fail; | |
1407 | ||
1dd05682 TI |
1408 | return root; |
1409 | ||
8a6a87cd | 1410 | fail_unlock: |
8c38938c | 1411 | if (leaf) |
1dd05682 | 1412 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1413 | fail: |
00246528 | 1414 | btrfs_put_root(root); |
20897f5c | 1415 | |
1dd05682 | 1416 | return ERR_PTR(ret); |
20897f5c AJ |
1417 | } |
1418 | ||
7237f183 YZ |
1419 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1420 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1421 | { |
1422 | struct btrfs_root *root; | |
e02119d5 | 1423 | |
96dfcb46 | 1424 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1425 | if (!root) |
7237f183 | 1426 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1427 | |
e02119d5 CM |
1428 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1429 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1430 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1431 | |
6ab6ebb7 NA |
1432 | return root; |
1433 | } | |
1434 | ||
1435 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1436 | struct btrfs_root *root) | |
1437 | { | |
1438 | struct extent_buffer *leaf; | |
1439 | ||
7237f183 | 1440 | /* |
92a7cc42 | 1441 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1442 | * |
92a7cc42 QW |
1443 | * Log trees are not exposed to user space thus can't be snapshotted, |
1444 | * and they go away before a real commit is actually done. | |
1445 | * | |
1446 | * They do store pointers to file data extents, and those reference | |
1447 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1448 | */ |
e02119d5 | 1449 | |
4d75f8a9 | 1450 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1451 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1452 | if (IS_ERR(leaf)) |
1453 | return PTR_ERR(leaf); | |
e02119d5 | 1454 | |
7237f183 | 1455 | root->node = leaf; |
e02119d5 | 1456 | |
e02119d5 CM |
1457 | btrfs_mark_buffer_dirty(root->node); |
1458 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1459 | |
1460 | return 0; | |
7237f183 YZ |
1461 | } |
1462 | ||
1463 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1464 | struct btrfs_fs_info *fs_info) | |
1465 | { | |
1466 | struct btrfs_root *log_root; | |
1467 | ||
1468 | log_root = alloc_log_tree(trans, fs_info); | |
1469 | if (IS_ERR(log_root)) | |
1470 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1471 | |
3ddebf27 NA |
1472 | if (!btrfs_is_zoned(fs_info)) { |
1473 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1474 | ||
1475 | if (ret) { | |
1476 | btrfs_put_root(log_root); | |
1477 | return ret; | |
1478 | } | |
6ab6ebb7 NA |
1479 | } |
1480 | ||
7237f183 YZ |
1481 | WARN_ON(fs_info->log_root_tree); |
1482 | fs_info->log_root_tree = log_root; | |
1483 | return 0; | |
1484 | } | |
1485 | ||
1486 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1487 | struct btrfs_root *root) | |
1488 | { | |
0b246afa | 1489 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1490 | struct btrfs_root *log_root; |
1491 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1492 | int ret; |
7237f183 | 1493 | |
0b246afa | 1494 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1495 | if (IS_ERR(log_root)) |
1496 | return PTR_ERR(log_root); | |
1497 | ||
6ab6ebb7 NA |
1498 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1499 | if (ret) { | |
1500 | btrfs_put_root(log_root); | |
1501 | return ret; | |
1502 | } | |
1503 | ||
7237f183 YZ |
1504 | log_root->last_trans = trans->transid; |
1505 | log_root->root_key.offset = root->root_key.objectid; | |
1506 | ||
1507 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1508 | btrfs_set_stack_inode_generation(inode_item, 1); |
1509 | btrfs_set_stack_inode_size(inode_item, 3); | |
1510 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1511 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1512 | fs_info->nodesize); |
3cae210f | 1513 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1514 | |
5d4f98a2 | 1515 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1516 | |
1517 | WARN_ON(root->log_root); | |
1518 | root->log_root = log_root; | |
1519 | root->log_transid = 0; | |
d1433deb | 1520 | root->log_transid_committed = -1; |
257c62e1 | 1521 | root->last_log_commit = 0; |
e02119d5 CM |
1522 | return 0; |
1523 | } | |
1524 | ||
49d11bea JB |
1525 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1526 | struct btrfs_path *path, | |
1527 | struct btrfs_key *key) | |
e02119d5 CM |
1528 | { |
1529 | struct btrfs_root *root; | |
1530 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1531 | u64 generation; |
cb517eab | 1532 | int ret; |
581c1760 | 1533 | int level; |
0f7d52f4 | 1534 | |
96dfcb46 | 1535 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1536 | if (!root) |
1537 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1538 | |
cb517eab MX |
1539 | ret = btrfs_find_root(tree_root, key, path, |
1540 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1541 | if (ret) { |
13a8a7c8 YZ |
1542 | if (ret > 0) |
1543 | ret = -ENOENT; | |
49d11bea | 1544 | goto fail; |
0f7d52f4 | 1545 | } |
13a8a7c8 | 1546 | |
84234f3a | 1547 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1548 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1549 | root->node = read_tree_block(fs_info, |
1550 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1551 | key->objectid, generation, level, NULL); |
64c043de LB |
1552 | if (IS_ERR(root->node)) { |
1553 | ret = PTR_ERR(root->node); | |
8c38938c | 1554 | root->node = NULL; |
49d11bea | 1555 | goto fail; |
4eb150d6 QW |
1556 | } |
1557 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1558 | ret = -EIO; |
49d11bea | 1559 | goto fail; |
416bc658 | 1560 | } |
88c602ab QW |
1561 | |
1562 | /* | |
1563 | * For real fs, and not log/reloc trees, root owner must | |
1564 | * match its root node owner | |
1565 | */ | |
1566 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1567 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1568 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1569 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1570 | btrfs_crit(fs_info, | |
1571 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1572 | root->root_key.objectid, root->node->start, | |
1573 | btrfs_header_owner(root->node), | |
1574 | root->root_key.objectid); | |
1575 | ret = -EUCLEAN; | |
1576 | goto fail; | |
1577 | } | |
5d4f98a2 | 1578 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1579 | return root; |
49d11bea | 1580 | fail: |
00246528 | 1581 | btrfs_put_root(root); |
49d11bea JB |
1582 | return ERR_PTR(ret); |
1583 | } | |
1584 | ||
1585 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1586 | struct btrfs_key *key) | |
1587 | { | |
1588 | struct btrfs_root *root; | |
1589 | struct btrfs_path *path; | |
1590 | ||
1591 | path = btrfs_alloc_path(); | |
1592 | if (!path) | |
1593 | return ERR_PTR(-ENOMEM); | |
1594 | root = read_tree_root_path(tree_root, path, key); | |
1595 | btrfs_free_path(path); | |
1596 | ||
1597 | return root; | |
cb517eab MX |
1598 | } |
1599 | ||
2dfb1e43 QW |
1600 | /* |
1601 | * Initialize subvolume root in-memory structure | |
1602 | * | |
1603 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1604 | */ | |
1605 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1606 | { |
1607 | int ret; | |
dcc3eb96 | 1608 | unsigned int nofs_flag; |
cb517eab | 1609 | |
dcc3eb96 NB |
1610 | /* |
1611 | * We might be called under a transaction (e.g. indirect backref | |
1612 | * resolution) which could deadlock if it triggers memory reclaim | |
1613 | */ | |
1614 | nofs_flag = memalloc_nofs_save(); | |
1615 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1616 | memalloc_nofs_restore(nofs_flag); | |
1617 | if (ret) | |
8257b2dc | 1618 | goto fail; |
8257b2dc | 1619 | |
aeb935a4 | 1620 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1621 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1622 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1623 | btrfs_check_and_init_root_item(&root->root_item); |
1624 | } | |
1625 | ||
851fd730 QW |
1626 | /* |
1627 | * Don't assign anonymous block device to roots that are not exposed to | |
1628 | * userspace, the id pool is limited to 1M | |
1629 | */ | |
1630 | if (is_fstree(root->root_key.objectid) && | |
1631 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1632 | if (!anon_dev) { |
1633 | ret = get_anon_bdev(&root->anon_dev); | |
1634 | if (ret) | |
1635 | goto fail; | |
1636 | } else { | |
1637 | root->anon_dev = anon_dev; | |
1638 | } | |
851fd730 | 1639 | } |
f32e48e9 CR |
1640 | |
1641 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1642 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1643 | if (ret) { |
1644 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1645 | goto fail; |
f32e48e9 CR |
1646 | } |
1647 | ||
6b8fad57 | 1648 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1649 | |
1650 | mutex_unlock(&root->objectid_mutex); | |
1651 | ||
cb517eab MX |
1652 | return 0; |
1653 | fail: | |
84db5ccf | 1654 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1655 | return ret; |
1656 | } | |
1657 | ||
a98db0f3 JB |
1658 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1659 | u64 root_id) | |
cb517eab MX |
1660 | { |
1661 | struct btrfs_root *root; | |
1662 | ||
fc7cbcd4 DS |
1663 | spin_lock(&fs_info->fs_roots_radix_lock); |
1664 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1665 | (unsigned long)root_id); | |
bc44d7c4 | 1666 | if (root) |
00246528 | 1667 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1668 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1669 | return root; |
1670 | } | |
1671 | ||
49d11bea JB |
1672 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1673 | u64 objectid) | |
1674 | { | |
abed4aaa JB |
1675 | struct btrfs_key key = { |
1676 | .objectid = objectid, | |
1677 | .type = BTRFS_ROOT_ITEM_KEY, | |
1678 | .offset = 0, | |
1679 | }; | |
1680 | ||
49d11bea JB |
1681 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1682 | return btrfs_grab_root(fs_info->tree_root); | |
1683 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1684 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1685 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1686 | return btrfs_grab_root(fs_info->chunk_root); | |
1687 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1688 | return btrfs_grab_root(fs_info->dev_root); | |
1689 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1690 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1691 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1692 | return btrfs_grab_root(fs_info->quota_root) ? | |
1693 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1694 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1695 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1696 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1697 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1698 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1699 | ||
1700 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1701 | } | |
49d11bea JB |
1702 | return NULL; |
1703 | } | |
1704 | ||
cb517eab MX |
1705 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1706 | struct btrfs_root *root) | |
1707 | { | |
1708 | int ret; | |
1709 | ||
fc7cbcd4 DS |
1710 | ret = radix_tree_preload(GFP_NOFS); |
1711 | if (ret) | |
1712 | return ret; | |
1713 | ||
1714 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1715 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1716 | (unsigned long)root->root_key.objectid, | |
1717 | root); | |
af01d2e5 | 1718 | if (ret == 0) { |
00246528 | 1719 | btrfs_grab_root(root); |
fc7cbcd4 | 1720 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1721 | } |
fc7cbcd4 DS |
1722 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1723 | radix_tree_preload_end(); | |
cb517eab MX |
1724 | |
1725 | return ret; | |
1726 | } | |
1727 | ||
bd647ce3 JB |
1728 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1729 | { | |
1730 | #ifdef CONFIG_BTRFS_DEBUG | |
1731 | struct btrfs_root *root; | |
1732 | ||
1733 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1734 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1735 | ||
bd647ce3 JB |
1736 | root = list_first_entry(&fs_info->allocated_roots, |
1737 | struct btrfs_root, leak_list); | |
457f1864 | 1738 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1739 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1740 | refcount_read(&root->refs)); |
1741 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1742 | btrfs_put_root(root); |
1743 | btrfs_put_root(root); | |
bd647ce3 JB |
1744 | } |
1745 | #endif | |
1746 | } | |
1747 | ||
abed4aaa JB |
1748 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1749 | { | |
1750 | struct btrfs_root *root; | |
1751 | struct rb_node *node; | |
1752 | ||
1753 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1754 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1755 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1756 | btrfs_put_root(root); | |
1757 | } | |
1758 | } | |
1759 | ||
0d4b0463 JB |
1760 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1761 | { | |
141386e1 JB |
1762 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1763 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1764 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1765 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1766 | btrfs_free_csum_hash(fs_info); | |
1767 | btrfs_free_stripe_hash_table(fs_info); | |
1768 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1769 | kfree(fs_info->balance_ctl); |
1770 | kfree(fs_info->delayed_root); | |
abed4aaa | 1771 | free_global_roots(fs_info); |
00246528 JB |
1772 | btrfs_put_root(fs_info->tree_root); |
1773 | btrfs_put_root(fs_info->chunk_root); | |
1774 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1775 | btrfs_put_root(fs_info->quota_root); |
1776 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1777 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1778 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1779 | btrfs_put_root(fs_info->block_group_root); |
bd647ce3 | 1780 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1781 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1782 | kfree(fs_info->super_copy); |
1783 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1784 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1785 | kvfree(fs_info); |
1786 | } | |
1787 | ||
1788 | ||
2dfb1e43 QW |
1789 | /* |
1790 | * Get an in-memory reference of a root structure. | |
1791 | * | |
1792 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1793 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1794 | * found, then read it from disk and add it to cached fs roots. | |
1795 | * | |
1796 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1797 | * | |
1798 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1799 | * same root objectid. | |
1800 | * | |
1801 | * @objectid: root id | |
1802 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1803 | * pass 0 for new allocation. | |
1804 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1805 | * for orphan roots | |
1806 | */ | |
1807 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1808 | u64 objectid, dev_t anon_dev, | |
1809 | bool check_ref) | |
5eda7b5e CM |
1810 | { |
1811 | struct btrfs_root *root; | |
381cf658 | 1812 | struct btrfs_path *path; |
1d4c08e0 | 1813 | struct btrfs_key key; |
5eda7b5e CM |
1814 | int ret; |
1815 | ||
49d11bea JB |
1816 | root = btrfs_get_global_root(fs_info, objectid); |
1817 | if (root) | |
1818 | return root; | |
4df27c4d | 1819 | again: |
56e9357a | 1820 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1821 | if (root) { |
2dfb1e43 QW |
1822 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1823 | ASSERT(!anon_dev); | |
bc44d7c4 | 1824 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1825 | btrfs_put_root(root); |
48475471 | 1826 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1827 | } |
5eda7b5e | 1828 | return root; |
48475471 | 1829 | } |
5eda7b5e | 1830 | |
56e9357a DS |
1831 | key.objectid = objectid; |
1832 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1833 | key.offset = (u64)-1; | |
1834 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1835 | if (IS_ERR(root)) |
1836 | return root; | |
3394e160 | 1837 | |
c00869f1 | 1838 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1839 | ret = -ENOENT; |
581bb050 | 1840 | goto fail; |
35a30d7c | 1841 | } |
581bb050 | 1842 | |
2dfb1e43 | 1843 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1844 | if (ret) |
1845 | goto fail; | |
3394e160 | 1846 | |
381cf658 DS |
1847 | path = btrfs_alloc_path(); |
1848 | if (!path) { | |
1849 | ret = -ENOMEM; | |
1850 | goto fail; | |
1851 | } | |
1d4c08e0 DS |
1852 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1853 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1854 | key.offset = objectid; |
1d4c08e0 DS |
1855 | |
1856 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1857 | btrfs_free_path(path); |
d68fc57b YZ |
1858 | if (ret < 0) |
1859 | goto fail; | |
1860 | if (ret == 0) | |
27cdeb70 | 1861 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1862 | |
cb517eab | 1863 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1864 | if (ret) { |
168a2f77 JJB |
1865 | if (ret == -EEXIST) { |
1866 | btrfs_put_root(root); | |
4df27c4d | 1867 | goto again; |
168a2f77 | 1868 | } |
4df27c4d | 1869 | goto fail; |
0f7d52f4 | 1870 | } |
edbd8d4e | 1871 | return root; |
4df27c4d | 1872 | fail: |
33fab972 FM |
1873 | /* |
1874 | * If our caller provided us an anonymous device, then it's his | |
1875 | * responsability to free it in case we fail. So we have to set our | |
1876 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() | |
1877 | * and once again by our caller. | |
1878 | */ | |
1879 | if (anon_dev) | |
1880 | root->anon_dev = 0; | |
8c38938c | 1881 | btrfs_put_root(root); |
4df27c4d | 1882 | return ERR_PTR(ret); |
edbd8d4e CM |
1883 | } |
1884 | ||
2dfb1e43 QW |
1885 | /* |
1886 | * Get in-memory reference of a root structure | |
1887 | * | |
1888 | * @objectid: tree objectid | |
1889 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1890 | * one reference | |
1891 | */ | |
1892 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1893 | u64 objectid, bool check_ref) | |
1894 | { | |
1895 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1896 | } | |
1897 | ||
1898 | /* | |
1899 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1900 | * the anonymous block device id | |
1901 | * | |
1902 | * @objectid: tree objectid | |
1903 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1904 | * parameter value | |
1905 | */ | |
1906 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1907 | u64 objectid, dev_t anon_dev) | |
1908 | { | |
1909 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1910 | } | |
1911 | ||
49d11bea JB |
1912 | /* |
1913 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1914 | * @fs_info: the fs_info | |
1915 | * @objectid: the objectid we need to lookup | |
1916 | * | |
1917 | * This is exclusively used for backref walking, and exists specifically because | |
1918 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1919 | * creation time, which means we may have to read the tree_root in order to look | |
1920 | * up a fs root that is not in memory. If the root is not in memory we will | |
1921 | * read the tree root commit root and look up the fs root from there. This is a | |
1922 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1923 | * have the most uptodate information, it'll simply be discarded once the | |
1924 | * backref code is finished using the root. | |
1925 | */ | |
1926 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1927 | struct btrfs_path *path, | |
1928 | u64 objectid) | |
1929 | { | |
1930 | struct btrfs_root *root; | |
1931 | struct btrfs_key key; | |
1932 | ||
1933 | ASSERT(path->search_commit_root && path->skip_locking); | |
1934 | ||
1935 | /* | |
1936 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1937 | * since this is called via the backref walking code we won't be looking | |
1938 | * up a root that doesn't exist, unless there's corruption. So if root | |
1939 | * != NULL just return it. | |
1940 | */ | |
1941 | root = btrfs_get_global_root(fs_info, objectid); | |
1942 | if (root) | |
1943 | return root; | |
1944 | ||
1945 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1946 | if (root) | |
1947 | return root; | |
1948 | ||
1949 | key.objectid = objectid; | |
1950 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1951 | key.offset = (u64)-1; | |
1952 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1953 | btrfs_release_path(path); | |
1954 | ||
1955 | return root; | |
1956 | } | |
1957 | ||
8b712842 CM |
1958 | /* |
1959 | * called by the kthread helper functions to finally call the bio end_io | |
1960 | * functions. This is where read checksum verification actually happens | |
1961 | */ | |
1962 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1963 | { |
ce9adaa5 | 1964 | struct bio *bio; |
97eb6b69 | 1965 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1966 | |
97eb6b69 | 1967 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1968 | bio = end_io_wq->bio; |
ce9adaa5 | 1969 | |
4e4cbee9 | 1970 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1971 | bio->bi_private = end_io_wq->private; |
1972 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1973 | bio_endio(bio); |
9be490f1 | 1974 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1975 | } |
1976 | ||
a74a4b97 CM |
1977 | static int cleaner_kthread(void *arg) |
1978 | { | |
0d031dc4 | 1979 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1980 | int again; |
a74a4b97 | 1981 | |
d6fd0ae2 | 1982 | while (1) { |
d0278245 | 1983 | again = 0; |
a74a4b97 | 1984 | |
fd340d0f JB |
1985 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1986 | ||
d0278245 | 1987 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1988 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1989 | goto sleep; |
1990 | ||
90c711ab ZB |
1991 | /* |
1992 | * Do not do anything if we might cause open_ctree() to block | |
1993 | * before we have finished mounting the filesystem. | |
1994 | */ | |
0b246afa | 1995 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1996 | goto sleep; |
1997 | ||
0b246afa | 1998 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1999 | goto sleep; |
2000 | ||
dc7f370c MX |
2001 | /* |
2002 | * Avoid the problem that we change the status of the fs | |
2003 | * during the above check and trylock. | |
2004 | */ | |
2ff7e61e | 2005 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 2006 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 2007 | goto sleep; |
76dda93c | 2008 | } |
a74a4b97 | 2009 | |
2ff7e61e | 2010 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 2011 | |
33c44184 | 2012 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 2013 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
2014 | |
2015 | /* | |
05323cd1 MX |
2016 | * The defragger has dealt with the R/O remount and umount, |
2017 | * needn't do anything special here. | |
d0278245 | 2018 | */ |
0b246afa | 2019 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
2020 | |
2021 | /* | |
f3372065 | 2022 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
2023 | * with relocation (btrfs_relocate_chunk) and relocation |
2024 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 2025 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
2026 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
2027 | * unused block groups. | |
2028 | */ | |
0b246afa | 2029 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
2030 | |
2031 | /* | |
2032 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
2033 | * all unused block_groups. This possibly gives us some more free | |
2034 | * space. | |
2035 | */ | |
2036 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 2037 | sleep: |
a0a1db70 | 2038 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
2039 | if (kthread_should_park()) |
2040 | kthread_parkme(); | |
2041 | if (kthread_should_stop()) | |
2042 | return 0; | |
838fe188 | 2043 | if (!again) { |
a74a4b97 | 2044 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 2045 | schedule(); |
a74a4b97 CM |
2046 | __set_current_state(TASK_RUNNING); |
2047 | } | |
da288d28 | 2048 | } |
a74a4b97 CM |
2049 | } |
2050 | ||
2051 | static int transaction_kthread(void *arg) | |
2052 | { | |
2053 | struct btrfs_root *root = arg; | |
0b246afa | 2054 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
2055 | struct btrfs_trans_handle *trans; |
2056 | struct btrfs_transaction *cur; | |
8929ecfa | 2057 | u64 transid; |
643900be | 2058 | time64_t delta; |
a74a4b97 | 2059 | unsigned long delay; |
914b2007 | 2060 | bool cannot_commit; |
a74a4b97 CM |
2061 | |
2062 | do { | |
914b2007 | 2063 | cannot_commit = false; |
ba1bc00f | 2064 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 2065 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 2066 | |
0b246afa JM |
2067 | spin_lock(&fs_info->trans_lock); |
2068 | cur = fs_info->running_transaction; | |
a74a4b97 | 2069 | if (!cur) { |
0b246afa | 2070 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
2071 | goto sleep; |
2072 | } | |
31153d81 | 2073 | |
643900be | 2074 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
2075 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
2076 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 2077 | delta < fs_info->commit_interval) { |
0b246afa | 2078 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
2079 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
2080 | delay = min(delay, | |
2081 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
2082 | goto sleep; |
2083 | } | |
8929ecfa | 2084 | transid = cur->transid; |
0b246afa | 2085 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 2086 | |
79787eaa | 2087 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 2088 | trans = btrfs_attach_transaction(root); |
914b2007 | 2089 | if (IS_ERR(trans)) { |
354aa0fb MX |
2090 | if (PTR_ERR(trans) != -ENOENT) |
2091 | cannot_commit = true; | |
79787eaa | 2092 | goto sleep; |
914b2007 | 2093 | } |
8929ecfa | 2094 | if (transid == trans->transid) { |
3a45bb20 | 2095 | btrfs_commit_transaction(trans); |
8929ecfa | 2096 | } else { |
3a45bb20 | 2097 | btrfs_end_transaction(trans); |
8929ecfa | 2098 | } |
a74a4b97 | 2099 | sleep: |
0b246afa JM |
2100 | wake_up_process(fs_info->cleaner_kthread); |
2101 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 2102 | |
84961539 | 2103 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 2104 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 2105 | if (!kthread_should_stop() && |
0b246afa | 2106 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 2107 | cannot_commit)) |
bc5511d0 | 2108 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
2109 | } while (!kthread_should_stop()); |
2110 | return 0; | |
2111 | } | |
2112 | ||
af31f5e5 | 2113 | /* |
01f0f9da NB |
2114 | * This will find the highest generation in the array of root backups. The |
2115 | * index of the highest array is returned, or -EINVAL if we can't find | |
2116 | * anything. | |
af31f5e5 CM |
2117 | * |
2118 | * We check to make sure the array is valid by comparing the | |
2119 | * generation of the latest root in the array with the generation | |
2120 | * in the super block. If they don't match we pitch it. | |
2121 | */ | |
01f0f9da | 2122 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 2123 | { |
01f0f9da | 2124 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 2125 | u64 cur; |
af31f5e5 CM |
2126 | struct btrfs_root_backup *root_backup; |
2127 | int i; | |
2128 | ||
2129 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2130 | root_backup = info->super_copy->super_roots + i; | |
2131 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2132 | if (cur == newest_gen) | |
01f0f9da | 2133 | return i; |
af31f5e5 CM |
2134 | } |
2135 | ||
01f0f9da | 2136 | return -EINVAL; |
af31f5e5 CM |
2137 | } |
2138 | ||
af31f5e5 CM |
2139 | /* |
2140 | * copy all the root pointers into the super backup array. | |
2141 | * this will bump the backup pointer by one when it is | |
2142 | * done | |
2143 | */ | |
2144 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2145 | { | |
6ef108dd | 2146 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2147 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
2148 | |
2149 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2150 | ||
2151 | /* | |
2152 | * make sure all of our padding and empty slots get zero filled | |
2153 | * regardless of which ones we use today | |
2154 | */ | |
2155 | memset(root_backup, 0, sizeof(*root_backup)); | |
2156 | ||
2157 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2158 | ||
2159 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2160 | btrfs_set_backup_tree_root_gen(root_backup, | |
2161 | btrfs_header_generation(info->tree_root->node)); | |
2162 | ||
2163 | btrfs_set_backup_tree_root_level(root_backup, | |
2164 | btrfs_header_level(info->tree_root->node)); | |
2165 | ||
2166 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2167 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2168 | btrfs_header_generation(info->chunk_root->node)); | |
2169 | btrfs_set_backup_chunk_root_level(root_backup, | |
2170 | btrfs_header_level(info->chunk_root->node)); | |
2171 | ||
9c54e80d JB |
2172 | if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) { |
2173 | btrfs_set_backup_block_group_root(root_backup, | |
2174 | info->block_group_root->node->start); | |
2175 | btrfs_set_backup_block_group_root_gen(root_backup, | |
2176 | btrfs_header_generation(info->block_group_root->node)); | |
2177 | btrfs_set_backup_block_group_root_level(root_backup, | |
2178 | btrfs_header_level(info->block_group_root->node)); | |
2179 | } else { | |
2180 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); | |
f7238e50 | 2181 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
2182 | |
2183 | btrfs_set_backup_extent_root(root_backup, | |
2184 | extent_root->node->start); | |
2185 | btrfs_set_backup_extent_root_gen(root_backup, | |
2186 | btrfs_header_generation(extent_root->node)); | |
2187 | btrfs_set_backup_extent_root_level(root_backup, | |
2188 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
2189 | |
2190 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
2191 | btrfs_set_backup_csum_root_gen(root_backup, | |
2192 | btrfs_header_generation(csum_root->node)); | |
2193 | btrfs_set_backup_csum_root_level(root_backup, | |
2194 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 2195 | } |
af31f5e5 | 2196 | |
7c7e82a7 CM |
2197 | /* |
2198 | * we might commit during log recovery, which happens before we set | |
2199 | * the fs_root. Make sure it is valid before we fill it in. | |
2200 | */ | |
2201 | if (info->fs_root && info->fs_root->node) { | |
2202 | btrfs_set_backup_fs_root(root_backup, | |
2203 | info->fs_root->node->start); | |
2204 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2205 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2206 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2207 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2208 | } |
af31f5e5 CM |
2209 | |
2210 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2211 | btrfs_set_backup_dev_root_gen(root_backup, | |
2212 | btrfs_header_generation(info->dev_root->node)); | |
2213 | btrfs_set_backup_dev_root_level(root_backup, | |
2214 | btrfs_header_level(info->dev_root->node)); | |
2215 | ||
af31f5e5 CM |
2216 | btrfs_set_backup_total_bytes(root_backup, |
2217 | btrfs_super_total_bytes(info->super_copy)); | |
2218 | btrfs_set_backup_bytes_used(root_backup, | |
2219 | btrfs_super_bytes_used(info->super_copy)); | |
2220 | btrfs_set_backup_num_devices(root_backup, | |
2221 | btrfs_super_num_devices(info->super_copy)); | |
2222 | ||
2223 | /* | |
2224 | * if we don't copy this out to the super_copy, it won't get remembered | |
2225 | * for the next commit | |
2226 | */ | |
2227 | memcpy(&info->super_copy->super_roots, | |
2228 | &info->super_for_commit->super_roots, | |
2229 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2230 | } | |
2231 | ||
bd2336b2 NB |
2232 | /* |
2233 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2234 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2235 | * | |
2236 | * fs_info - filesystem whose backup roots need to be read | |
2237 | * priority - priority of backup root required | |
2238 | * | |
2239 | * Returns backup root index on success and -EINVAL otherwise. | |
2240 | */ | |
2241 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2242 | { | |
2243 | int backup_index = find_newest_super_backup(fs_info); | |
2244 | struct btrfs_super_block *super = fs_info->super_copy; | |
2245 | struct btrfs_root_backup *root_backup; | |
2246 | ||
2247 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2248 | if (priority == 0) | |
2249 | return backup_index; | |
2250 | ||
2251 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2252 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2253 | } else { | |
2254 | return -EINVAL; | |
2255 | } | |
2256 | ||
2257 | root_backup = super->super_roots + backup_index; | |
2258 | ||
2259 | btrfs_set_super_generation(super, | |
2260 | btrfs_backup_tree_root_gen(root_backup)); | |
2261 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2262 | btrfs_set_super_root_level(super, | |
2263 | btrfs_backup_tree_root_level(root_backup)); | |
2264 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2265 | ||
2266 | /* | |
2267 | * Fixme: the total bytes and num_devices need to match or we should | |
2268 | * need a fsck | |
2269 | */ | |
2270 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2271 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2272 | ||
2273 | return backup_index; | |
2274 | } | |
2275 | ||
7abadb64 LB |
2276 | /* helper to cleanup workers */ |
2277 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2278 | { | |
dc6e3209 | 2279 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2280 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
a31b4a43 | 2281 | btrfs_destroy_workqueue(fs_info->hipri_workers); |
5cdc7ad3 | 2282 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2283 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2284 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
385de0ef CH |
2285 | if (fs_info->rmw_workers) |
2286 | destroy_workqueue(fs_info->rmw_workers); | |
fccb5d86 QW |
2287 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2288 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2289 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2290 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2291 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2292 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2293 | if (fs_info->discard_ctl.discard_workers) |
2294 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2295 | /* |
2296 | * Now that all other work queues are destroyed, we can safely destroy | |
2297 | * the queues used for metadata I/O, since tasks from those other work | |
2298 | * queues can do metadata I/O operations. | |
2299 | */ | |
2300 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2301 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2302 | } |
2303 | ||
2e9f5954 R |
2304 | static void free_root_extent_buffers(struct btrfs_root *root) |
2305 | { | |
2306 | if (root) { | |
2307 | free_extent_buffer(root->node); | |
2308 | free_extent_buffer(root->commit_root); | |
2309 | root->node = NULL; | |
2310 | root->commit_root = NULL; | |
2311 | } | |
2312 | } | |
2313 | ||
abed4aaa JB |
2314 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2315 | { | |
2316 | struct btrfs_root *root, *tmp; | |
2317 | ||
2318 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2319 | &fs_info->global_root_tree, | |
2320 | rb_node) | |
2321 | free_root_extent_buffers(root); | |
2322 | } | |
2323 | ||
af31f5e5 | 2324 | /* helper to cleanup tree roots */ |
4273eaff | 2325 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2326 | { |
2e9f5954 | 2327 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2328 | |
abed4aaa | 2329 | free_global_root_pointers(info); |
2e9f5954 | 2330 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2331 | free_root_extent_buffers(info->quota_root); |
2332 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2333 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2334 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 2335 | free_root_extent_buffers(info->block_group_root); |
4273eaff | 2336 | if (free_chunk_root) |
2e9f5954 | 2337 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2338 | } |
2339 | ||
8c38938c JB |
2340 | void btrfs_put_root(struct btrfs_root *root) |
2341 | { | |
2342 | if (!root) | |
2343 | return; | |
2344 | ||
2345 | if (refcount_dec_and_test(&root->refs)) { | |
2346 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2347 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2348 | if (root->anon_dev) |
2349 | free_anon_bdev(root->anon_dev); | |
2350 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2351 | free_root_extent_buffers(root); |
8c38938c | 2352 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 2353 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 2354 | list_del_init(&root->leak_list); |
fc7cbcd4 | 2355 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
2356 | #endif |
2357 | kfree(root); | |
2358 | } | |
2359 | } | |
2360 | ||
faa2dbf0 | 2361 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 2362 | { |
fc7cbcd4 DS |
2363 | int ret; |
2364 | struct btrfs_root *gang[8]; | |
2365 | int i; | |
171f6537 JB |
2366 | |
2367 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
2368 | gang[0] = list_entry(fs_info->dead_roots.next, |
2369 | struct btrfs_root, root_list); | |
2370 | list_del(&gang[0]->root_list); | |
171f6537 | 2371 | |
fc7cbcd4 DS |
2372 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
2373 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
2374 | btrfs_put_root(gang[0]); | |
171f6537 JB |
2375 | } |
2376 | ||
fc7cbcd4 DS |
2377 | while (1) { |
2378 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2379 | (void **)gang, 0, | |
2380 | ARRAY_SIZE(gang)); | |
2381 | if (!ret) | |
2382 | break; | |
2383 | for (i = 0; i < ret; i++) | |
2384 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
2385 | } |
2386 | } | |
af31f5e5 | 2387 | |
638aa7ed ES |
2388 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2389 | { | |
2390 | mutex_init(&fs_info->scrub_lock); | |
2391 | atomic_set(&fs_info->scrubs_running, 0); | |
2392 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2393 | atomic_set(&fs_info->scrubs_paused, 0); | |
2394 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2395 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2396 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2397 | } |
2398 | ||
779a65a4 ES |
2399 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2400 | { | |
2401 | spin_lock_init(&fs_info->balance_lock); | |
2402 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2403 | atomic_set(&fs_info->balance_pause_req, 0); |
2404 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2405 | fs_info->balance_ctl = NULL; | |
2406 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2407 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2408 | } |
2409 | ||
6bccf3ab | 2410 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2411 | { |
2ff7e61e JM |
2412 | struct inode *inode = fs_info->btree_inode; |
2413 | ||
2414 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2415 | set_nlink(inode, 1); | |
f37938e0 ES |
2416 | /* |
2417 | * we set the i_size on the btree inode to the max possible int. | |
2418 | * the real end of the address space is determined by all of | |
2419 | * the devices in the system | |
2420 | */ | |
2ff7e61e JM |
2421 | inode->i_size = OFFSET_MAX; |
2422 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2423 | |
2ff7e61e | 2424 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2425 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2c53a14d | 2426 | IO_TREE_BTREE_INODE_IO, inode); |
7b439738 | 2427 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2428 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2429 | |
5c8fd99f | 2430 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2431 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2432 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2433 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2434 | } |
2435 | ||
ad618368 ES |
2436 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2437 | { | |
ad618368 | 2438 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2439 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2440 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2441 | } |
2442 | ||
f9e92e40 ES |
2443 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2444 | { | |
2445 | spin_lock_init(&fs_info->qgroup_lock); | |
2446 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2447 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2448 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2449 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2450 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2451 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2452 | mutex_init(&fs_info->qgroup_rescan_lock); |
2453 | } | |
2454 | ||
d21deec5 | 2455 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2456 | { |
f7b885be | 2457 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2458 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2459 | |
2460 | fs_info->workers = | |
a31b4a43 CH |
2461 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2462 | fs_info->hipri_workers = | |
2463 | btrfs_alloc_workqueue(fs_info, "worker-high", | |
cb001095 | 2464 | flags | WQ_HIGHPRI, max_active, 16); |
2a458198 ES |
2465 | |
2466 | fs_info->delalloc_workers = | |
cb001095 JM |
2467 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2468 | flags, max_active, 2); | |
2a458198 ES |
2469 | |
2470 | fs_info->flush_workers = | |
cb001095 JM |
2471 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2472 | flags, max_active, 0); | |
2a458198 ES |
2473 | |
2474 | fs_info->caching_workers = | |
cb001095 | 2475 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2476 | |
2a458198 | 2477 | fs_info->fixup_workers = |
cb001095 | 2478 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2479 | |
2480 | /* | |
2481 | * endios are largely parallel and should have a very | |
2482 | * low idle thresh | |
2483 | */ | |
2484 | fs_info->endio_workers = | |
cb001095 | 2485 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2486 | fs_info->endio_meta_workers = |
cb001095 JM |
2487 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2488 | max_active, 4); | |
2a458198 | 2489 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2490 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2491 | max_active, 2); | |
2a458198 | 2492 | fs_info->endio_raid56_workers = |
cb001095 JM |
2493 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2494 | max_active, 4); | |
385de0ef | 2495 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2496 | fs_info->endio_write_workers = |
cb001095 JM |
2497 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2498 | max_active, 2); | |
2a458198 | 2499 | fs_info->endio_freespace_worker = |
cb001095 JM |
2500 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2501 | max_active, 0); | |
2a458198 | 2502 | fs_info->delayed_workers = |
cb001095 JM |
2503 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2504 | max_active, 0); | |
2a458198 | 2505 | fs_info->qgroup_rescan_workers = |
cb001095 | 2506 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2507 | fs_info->discard_ctl.discard_workers = |
2508 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 | 2509 | |
a31b4a43 CH |
2510 | if (!(fs_info->workers && fs_info->hipri_workers && |
2511 | fs_info->delalloc_workers && fs_info->flush_workers && | |
2a458198 ES |
2512 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2513 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2514 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2515 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
f26c9238 QW |
2516 | fs_info->caching_workers && fs_info->fixup_workers && |
2517 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2518 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2519 | return -ENOMEM; |
2520 | } | |
2521 | ||
2522 | return 0; | |
2523 | } | |
2524 | ||
6d97c6e3 JT |
2525 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2526 | { | |
2527 | struct crypto_shash *csum_shash; | |
b4e967be | 2528 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2529 | |
b4e967be | 2530 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2531 | |
2532 | if (IS_ERR(csum_shash)) { | |
2533 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2534 | csum_driver); |
6d97c6e3 JT |
2535 | return PTR_ERR(csum_shash); |
2536 | } | |
2537 | ||
2538 | fs_info->csum_shash = csum_shash; | |
2539 | ||
2540 | return 0; | |
2541 | } | |
2542 | ||
63443bf5 ES |
2543 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2544 | struct btrfs_fs_devices *fs_devices) | |
2545 | { | |
2546 | int ret; | |
63443bf5 ES |
2547 | struct btrfs_root *log_tree_root; |
2548 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2549 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2550 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2551 | |
2552 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2553 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2554 | return -EIO; |
2555 | } | |
2556 | ||
96dfcb46 JB |
2557 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2558 | GFP_KERNEL); | |
63443bf5 ES |
2559 | if (!log_tree_root) |
2560 | return -ENOMEM; | |
2561 | ||
2ff7e61e | 2562 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2563 | BTRFS_TREE_LOG_OBJECTID, |
2564 | fs_info->generation + 1, level, | |
2565 | NULL); | |
64c043de | 2566 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2567 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2568 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2569 | log_tree_root->node = NULL; |
00246528 | 2570 | btrfs_put_root(log_tree_root); |
0eeff236 | 2571 | return ret; |
4eb150d6 QW |
2572 | } |
2573 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2574 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2575 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2576 | return -EIO; |
2577 | } | |
4eb150d6 | 2578 | |
63443bf5 ES |
2579 | /* returns with log_tree_root freed on success */ |
2580 | ret = btrfs_recover_log_trees(log_tree_root); | |
2581 | if (ret) { | |
0b246afa JM |
2582 | btrfs_handle_fs_error(fs_info, ret, |
2583 | "Failed to recover log tree"); | |
00246528 | 2584 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2585 | return ret; |
2586 | } | |
2587 | ||
bc98a42c | 2588 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2589 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2590 | if (ret) |
2591 | return ret; | |
2592 | } | |
2593 | ||
2594 | return 0; | |
2595 | } | |
2596 | ||
abed4aaa JB |
2597 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2598 | struct btrfs_path *path, u64 objectid, | |
2599 | const char *name) | |
2600 | { | |
2601 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2602 | struct btrfs_root *root; | |
f7238e50 | 2603 | u64 max_global_id = 0; |
abed4aaa JB |
2604 | int ret; |
2605 | struct btrfs_key key = { | |
2606 | .objectid = objectid, | |
2607 | .type = BTRFS_ROOT_ITEM_KEY, | |
2608 | .offset = 0, | |
2609 | }; | |
2610 | bool found = false; | |
2611 | ||
2612 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2613 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2614 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2615 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2616 | return 0; | |
2617 | } | |
2618 | ||
2619 | while (1) { | |
2620 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2621 | if (ret < 0) | |
2622 | break; | |
2623 | ||
2624 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2625 | ret = btrfs_next_leaf(tree_root, path); | |
2626 | if (ret) { | |
2627 | if (ret > 0) | |
2628 | ret = 0; | |
2629 | break; | |
2630 | } | |
2631 | } | |
2632 | ret = 0; | |
2633 | ||
2634 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2635 | if (key.objectid != objectid) | |
2636 | break; | |
2637 | btrfs_release_path(path); | |
2638 | ||
f7238e50 JB |
2639 | /* |
2640 | * Just worry about this for extent tree, it'll be the same for | |
2641 | * everybody. | |
2642 | */ | |
2643 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2644 | max_global_id = max(max_global_id, key.offset); | |
2645 | ||
abed4aaa JB |
2646 | found = true; |
2647 | root = read_tree_root_path(tree_root, path, &key); | |
2648 | if (IS_ERR(root)) { | |
2649 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2650 | ret = PTR_ERR(root); | |
2651 | break; | |
2652 | } | |
2653 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2654 | ret = btrfs_global_root_insert(root); | |
2655 | if (ret) { | |
2656 | btrfs_put_root(root); | |
2657 | break; | |
2658 | } | |
2659 | key.offset++; | |
2660 | } | |
2661 | btrfs_release_path(path); | |
2662 | ||
f7238e50 JB |
2663 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2664 | fs_info->nr_global_roots = max_global_id + 1; | |
2665 | ||
abed4aaa JB |
2666 | if (!found || ret) { |
2667 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2668 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2669 | ||
2670 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2671 | ret = ret ? ret : -ENOENT; | |
2672 | else | |
2673 | ret = 0; | |
2674 | btrfs_err(fs_info, "failed to load root %s", name); | |
2675 | } | |
2676 | return ret; | |
2677 | } | |
2678 | ||
2679 | static int load_global_roots(struct btrfs_root *tree_root) | |
2680 | { | |
2681 | struct btrfs_path *path; | |
2682 | int ret = 0; | |
2683 | ||
2684 | path = btrfs_alloc_path(); | |
2685 | if (!path) | |
2686 | return -ENOMEM; | |
2687 | ||
2688 | ret = load_global_roots_objectid(tree_root, path, | |
2689 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2690 | if (ret) | |
2691 | goto out; | |
2692 | ret = load_global_roots_objectid(tree_root, path, | |
2693 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2694 | if (ret) | |
2695 | goto out; | |
2696 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2697 | goto out; | |
2698 | ret = load_global_roots_objectid(tree_root, path, | |
2699 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2700 | "free space"); | |
2701 | out: | |
2702 | btrfs_free_path(path); | |
2703 | return ret; | |
2704 | } | |
2705 | ||
6bccf3ab | 2706 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2707 | { |
6bccf3ab | 2708 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2709 | struct btrfs_root *root; |
4bbcaa64 ES |
2710 | struct btrfs_key location; |
2711 | int ret; | |
2712 | ||
6bccf3ab JM |
2713 | BUG_ON(!fs_info->tree_root); |
2714 | ||
abed4aaa JB |
2715 | ret = load_global_roots(tree_root); |
2716 | if (ret) | |
2717 | return ret; | |
2718 | ||
2719 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
4bbcaa64 ES |
2720 | location.type = BTRFS_ROOT_ITEM_KEY; |
2721 | location.offset = 0; | |
2722 | ||
a4f3d2c4 | 2723 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2724 | if (IS_ERR(root)) { |
42437a63 JB |
2725 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2726 | ret = PTR_ERR(root); | |
2727 | goto out; | |
2728 | } | |
2729 | } else { | |
2730 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2731 | fs_info->dev_root = root; | |
f50f4353 | 2732 | } |
820a49da JB |
2733 | /* Initialize fs_info for all devices in any case */ |
2734 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2735 | |
aeb935a4 QW |
2736 | /* |
2737 | * This tree can share blocks with some other fs tree during relocation | |
2738 | * and we need a proper setup by btrfs_get_fs_root | |
2739 | */ | |
56e9357a DS |
2740 | root = btrfs_get_fs_root(tree_root->fs_info, |
2741 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2742 | if (IS_ERR(root)) { |
42437a63 JB |
2743 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2744 | ret = PTR_ERR(root); | |
2745 | goto out; | |
2746 | } | |
2747 | } else { | |
2748 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2749 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2750 | } |
aeb935a4 | 2751 | |
4bbcaa64 | 2752 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2753 | root = btrfs_read_tree_root(tree_root, &location); |
2754 | if (!IS_ERR(root)) { | |
2755 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2756 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2757 | fs_info->quota_root = root; |
4bbcaa64 ES |
2758 | } |
2759 | ||
2760 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2761 | root = btrfs_read_tree_root(tree_root, &location); |
2762 | if (IS_ERR(root)) { | |
42437a63 JB |
2763 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2764 | ret = PTR_ERR(root); | |
2765 | if (ret != -ENOENT) | |
2766 | goto out; | |
2767 | } | |
4bbcaa64 | 2768 | } else { |
a4f3d2c4 DS |
2769 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2770 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2771 | } |
2772 | ||
2773 | return 0; | |
f50f4353 LB |
2774 | out: |
2775 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2776 | location.objectid, ret); | |
2777 | return ret; | |
4bbcaa64 ES |
2778 | } |
2779 | ||
069ec957 QW |
2780 | /* |
2781 | * Real super block validation | |
2782 | * NOTE: super csum type and incompat features will not be checked here. | |
2783 | * | |
2784 | * @sb: super block to check | |
2785 | * @mirror_num: the super block number to check its bytenr: | |
2786 | * 0 the primary (1st) sb | |
2787 | * 1, 2 2nd and 3rd backup copy | |
2788 | * -1 skip bytenr check | |
2789 | */ | |
2790 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2791 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2792 | { |
21a852b0 QW |
2793 | u64 nodesize = btrfs_super_nodesize(sb); |
2794 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2795 | int ret = 0; | |
2796 | ||
2797 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2798 | btrfs_err(fs_info, "no valid FS found"); | |
2799 | ret = -EINVAL; | |
2800 | } | |
2801 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2802 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2803 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2804 | ret = -EINVAL; | |
2805 | } | |
2806 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2807 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2808 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2809 | ret = -EINVAL; | |
2810 | } | |
2811 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2812 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2813 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2814 | ret = -EINVAL; | |
2815 | } | |
2816 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2817 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2818 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2819 | ret = -EINVAL; | |
2820 | } | |
2821 | ||
2822 | /* | |
2823 | * Check sectorsize and nodesize first, other check will need it. | |
2824 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2825 | */ | |
2826 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2827 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2828 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2829 | ret = -EINVAL; | |
2830 | } | |
0bb3eb3e QW |
2831 | |
2832 | /* | |
1a42daab QW |
2833 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2834 | * | |
2835 | * We can support 16K sectorsize with 64K page size without problem, | |
2836 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2837 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2838 | * beginning. | |
0bb3eb3e | 2839 | */ |
1a42daab | 2840 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2841 | btrfs_err(fs_info, |
0bb3eb3e | 2842 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2843 | sectorsize, PAGE_SIZE); |
2844 | ret = -EINVAL; | |
2845 | } | |
0bb3eb3e | 2846 | |
21a852b0 QW |
2847 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2848 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2849 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2850 | ret = -EINVAL; | |
2851 | } | |
2852 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2853 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2854 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2855 | ret = -EINVAL; | |
2856 | } | |
2857 | ||
2858 | /* Root alignment check */ | |
2859 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2860 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2861 | btrfs_super_root(sb)); | |
2862 | ret = -EINVAL; | |
2863 | } | |
2864 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2865 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2866 | btrfs_super_chunk_root(sb)); | |
2867 | ret = -EINVAL; | |
2868 | } | |
2869 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2870 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2871 | btrfs_super_log_root(sb)); | |
2872 | ret = -EINVAL; | |
2873 | } | |
2874 | ||
aefd7f70 NB |
2875 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2876 | BTRFS_FSID_SIZE)) { | |
2877 | btrfs_err(fs_info, | |
2878 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2879 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2880 | ret = -EINVAL; | |
2881 | } | |
2882 | ||
2883 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2884 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2885 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2886 | btrfs_err(fs_info, | |
2887 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2888 | fs_info->super_copy->metadata_uuid, | |
2889 | fs_info->fs_devices->metadata_uuid); | |
2890 | ret = -EINVAL; | |
2891 | } | |
2892 | ||
de37aa51 | 2893 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2894 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2895 | btrfs_err(fs_info, |
7239ff4b | 2896 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2897 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2898 | ret = -EINVAL; |
2899 | } | |
2900 | ||
2901 | /* | |
2902 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2903 | * done later | |
2904 | */ | |
2905 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2906 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2907 | btrfs_super_bytes_used(sb)); | |
2908 | ret = -EINVAL; | |
2909 | } | |
2910 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2911 | btrfs_err(fs_info, "invalid stripesize %u", | |
2912 | btrfs_super_stripesize(sb)); | |
2913 | ret = -EINVAL; | |
2914 | } | |
2915 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2916 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2917 | btrfs_super_num_devices(sb)); | |
2918 | if (btrfs_super_num_devices(sb) == 0) { | |
2919 | btrfs_err(fs_info, "number of devices is 0"); | |
2920 | ret = -EINVAL; | |
2921 | } | |
2922 | ||
069ec957 QW |
2923 | if (mirror_num >= 0 && |
2924 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2925 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2926 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2927 | ret = -EINVAL; | |
2928 | } | |
2929 | ||
2930 | /* | |
2931 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2932 | * and one chunk | |
2933 | */ | |
2934 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2935 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2936 | btrfs_super_sys_array_size(sb), | |
2937 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2938 | ret = -EINVAL; | |
2939 | } | |
2940 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2941 | + sizeof(struct btrfs_chunk)) { | |
2942 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2943 | btrfs_super_sys_array_size(sb), | |
2944 | sizeof(struct btrfs_disk_key) | |
2945 | + sizeof(struct btrfs_chunk)); | |
2946 | ret = -EINVAL; | |
2947 | } | |
2948 | ||
2949 | /* | |
2950 | * The generation is a global counter, we'll trust it more than the others | |
2951 | * but it's still possible that it's the one that's wrong. | |
2952 | */ | |
2953 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2954 | btrfs_warn(fs_info, | |
2955 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2956 | btrfs_super_generation(sb), | |
2957 | btrfs_super_chunk_root_generation(sb)); | |
2958 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2959 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2960 | btrfs_warn(fs_info, | |
2961 | "suspicious: generation < cache_generation: %llu < %llu", | |
2962 | btrfs_super_generation(sb), | |
2963 | btrfs_super_cache_generation(sb)); | |
2964 | ||
2965 | return ret; | |
2966 | } | |
2967 | ||
069ec957 QW |
2968 | /* |
2969 | * Validation of super block at mount time. | |
2970 | * Some checks already done early at mount time, like csum type and incompat | |
2971 | * flags will be skipped. | |
2972 | */ | |
2973 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2974 | { | |
2975 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2976 | } | |
2977 | ||
75cb857d QW |
2978 | /* |
2979 | * Validation of super block at write time. | |
2980 | * Some checks like bytenr check will be skipped as their values will be | |
2981 | * overwritten soon. | |
2982 | * Extra checks like csum type and incompat flags will be done here. | |
2983 | */ | |
2984 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2985 | struct btrfs_super_block *sb) | |
2986 | { | |
2987 | int ret; | |
2988 | ||
2989 | ret = validate_super(fs_info, sb, -1); | |
2990 | if (ret < 0) | |
2991 | goto out; | |
e7e16f48 | 2992 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2993 | ret = -EUCLEAN; |
2994 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2995 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2996 | goto out; | |
2997 | } | |
2998 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2999 | ret = -EUCLEAN; | |
3000 | btrfs_err(fs_info, | |
3001 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
3002 | btrfs_super_incompat_flags(sb), | |
3003 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
3004 | goto out; | |
3005 | } | |
3006 | out: | |
3007 | if (ret < 0) | |
3008 | btrfs_err(fs_info, | |
3009 | "super block corruption detected before writing it to disk"); | |
3010 | return ret; | |
3011 | } | |
3012 | ||
bd676446 JB |
3013 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
3014 | { | |
3015 | int ret = 0; | |
3016 | ||
3017 | root->node = read_tree_block(root->fs_info, bytenr, | |
3018 | root->root_key.objectid, gen, level, NULL); | |
3019 | if (IS_ERR(root->node)) { | |
3020 | ret = PTR_ERR(root->node); | |
3021 | root->node = NULL; | |
4eb150d6 QW |
3022 | return ret; |
3023 | } | |
3024 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
3025 | free_extent_buffer(root->node); |
3026 | root->node = NULL; | |
4eb150d6 | 3027 | return -EIO; |
bd676446 JB |
3028 | } |
3029 | ||
bd676446 JB |
3030 | btrfs_set_root_node(&root->root_item, root->node); |
3031 | root->commit_root = btrfs_root_node(root); | |
3032 | btrfs_set_root_refs(&root->root_item, 1); | |
3033 | return ret; | |
3034 | } | |
3035 | ||
3036 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
3037 | { | |
3038 | struct btrfs_super_block *sb = fs_info->super_copy; | |
3039 | u64 gen, bytenr; | |
3040 | int level, ret; | |
3041 | ||
3042 | bytenr = btrfs_super_root(sb); | |
3043 | gen = btrfs_super_generation(sb); | |
3044 | level = btrfs_super_root_level(sb); | |
3045 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 3046 | if (ret) { |
bd676446 | 3047 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
3048 | return ret; |
3049 | } | |
3050 | ||
3051 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
3052 | return 0; | |
3053 | ||
3054 | bytenr = btrfs_super_block_group_root(sb); | |
3055 | gen = btrfs_super_block_group_root_generation(sb); | |
3056 | level = btrfs_super_block_group_root_level(sb); | |
3057 | ret = load_super_root(fs_info->block_group_root, bytenr, gen, level); | |
3058 | if (ret) | |
3059 | btrfs_warn(fs_info, "couldn't read block group root"); | |
bd676446 JB |
3060 | return ret; |
3061 | } | |
3062 | ||
6ef108dd | 3063 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 3064 | { |
6ef108dd | 3065 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
3066 | struct btrfs_super_block *sb = fs_info->super_copy; |
3067 | struct btrfs_root *tree_root = fs_info->tree_root; | |
3068 | bool handle_error = false; | |
3069 | int ret = 0; | |
3070 | int i; | |
3071 | ||
9c54e80d JB |
3072 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { |
3073 | struct btrfs_root *root; | |
3074 | ||
3075 | root = btrfs_alloc_root(fs_info, BTRFS_BLOCK_GROUP_TREE_OBJECTID, | |
3076 | GFP_KERNEL); | |
3077 | if (!root) | |
3078 | return -ENOMEM; | |
3079 | fs_info->block_group_root = root; | |
3080 | } | |
b8522a1e | 3081 | |
b8522a1e | 3082 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { |
b8522a1e NB |
3083 | if (handle_error) { |
3084 | if (!IS_ERR(tree_root->node)) | |
3085 | free_extent_buffer(tree_root->node); | |
3086 | tree_root->node = NULL; | |
3087 | ||
3088 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
3089 | break; | |
3090 | ||
3091 | free_root_pointers(fs_info, 0); | |
3092 | ||
3093 | /* | |
3094 | * Don't use the log in recovery mode, it won't be | |
3095 | * valid | |
3096 | */ | |
3097 | btrfs_set_super_log_root(sb, 0); | |
3098 | ||
3099 | /* We can't trust the free space cache either */ | |
3100 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3101 | ||
3102 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 3103 | backup_index = ret; |
b8522a1e NB |
3104 | if (ret < 0) |
3105 | return ret; | |
3106 | } | |
b8522a1e | 3107 | |
bd676446 JB |
3108 | ret = load_important_roots(fs_info); |
3109 | if (ret) { | |
217f5004 | 3110 | handle_error = true; |
b8522a1e NB |
3111 | continue; |
3112 | } | |
3113 | ||
336a0d8d NB |
3114 | /* |
3115 | * No need to hold btrfs_root::objectid_mutex since the fs | |
3116 | * hasn't been fully initialised and we are the only user | |
3117 | */ | |
453e4873 | 3118 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 3119 | if (ret < 0) { |
b8522a1e NB |
3120 | handle_error = true; |
3121 | continue; | |
3122 | } | |
3123 | ||
6b8fad57 | 3124 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
3125 | |
3126 | ret = btrfs_read_roots(fs_info); | |
3127 | if (ret < 0) { | |
3128 | handle_error = true; | |
3129 | continue; | |
3130 | } | |
3131 | ||
3132 | /* All successful */ | |
bd676446 JB |
3133 | fs_info->generation = btrfs_header_generation(tree_root->node); |
3134 | fs_info->last_trans_committed = fs_info->generation; | |
d96b3424 | 3135 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
3136 | |
3137 | /* Always begin writing backup roots after the one being used */ | |
3138 | if (backup_index < 0) { | |
3139 | fs_info->backup_root_index = 0; | |
3140 | } else { | |
3141 | fs_info->backup_root_index = backup_index + 1; | |
3142 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
3143 | } | |
b8522a1e NB |
3144 | break; |
3145 | } | |
3146 | ||
3147 | return ret; | |
3148 | } | |
3149 | ||
8260edba | 3150 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 3151 | { |
fc7cbcd4 | 3152 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8ee92268 | 3153 | xa_init_flags(&fs_info->extent_buffers, GFP_ATOMIC); |
8fd17795 | 3154 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 3155 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 3156 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 3157 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 3158 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 3159 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 3160 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 3161 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 3162 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 3163 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 3164 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 3165 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 3166 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 3167 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 3168 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 3169 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 3170 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 3171 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 3172 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 3173 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 3174 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 3175 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 3176 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 3177 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 3178 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 3179 | |
0b86a832 | 3180 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3181 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3182 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3183 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3184 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3185 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3186 | #ifdef CONFIG_BTRFS_DEBUG |
3187 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3188 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3189 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3190 | #endif |
c8bf1b67 | 3191 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3192 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3193 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3194 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3195 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3196 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3197 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3198 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3199 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3200 | BTRFS_BLOCK_RSV_DELREFS); | |
3201 | ||
771ed689 | 3202 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3203 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3204 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3205 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3206 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3207 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3208 | fs_info->metadata_ratio = 0; |
4cb5300b | 3209 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3210 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3211 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3212 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3213 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3214 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3215 | |
b34b086c CM |
3216 | fs_info->thread_pool_size = min_t(unsigned long, |
3217 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3218 | |
199c2a9c MX |
3219 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3220 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3221 | |
638aa7ed | 3222 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3223 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3224 | fs_info->check_integrity_print_mask = 0; | |
3225 | #endif | |
779a65a4 | 3226 | btrfs_init_balance(fs_info); |
57056740 | 3227 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3228 | |
16b0c258 | 3229 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 3230 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 3231 | |
fe119a6e NB |
3232 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
3233 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
39279cc3 | 3234 | |
5a3f23d5 | 3235 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3236 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3237 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3238 | mutex_init(&fs_info->transaction_kthread_mutex); |
3239 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3240 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3241 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3242 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3243 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3244 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3245 | |
ad618368 | 3246 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3247 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3248 | btrfs_discard_init(fs_info); |
416ac51d | 3249 | |
fa9c0d79 CM |
3250 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3251 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3252 | ||
e6dcd2dc | 3253 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3254 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3255 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3256 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3257 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3258 | |
da17066c JM |
3259 | /* Usable values until the real ones are cached from the superblock */ |
3260 | fs_info->nodesize = 4096; | |
3261 | fs_info->sectorsize = 4096; | |
ab108d99 | 3262 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3263 | fs_info->stripesize = 4096; |
3264 | ||
eede2bf3 OS |
3265 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3266 | fs_info->swapfile_pins = RB_ROOT; | |
3267 | ||
18bb8bbf JT |
3268 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3269 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3270 | } |
3271 | ||
3272 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3273 | { | |
3274 | int ret; | |
3275 | ||
3276 | fs_info->sb = sb; | |
3277 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3278 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3279 | |
5deb17e1 | 3280 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3281 | if (ret) |
c75e8394 | 3282 | return ret; |
ae18c37a JB |
3283 | |
3284 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3285 | if (ret) | |
c75e8394 | 3286 | return ret; |
ae18c37a JB |
3287 | |
3288 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3289 | (1 + ilog2(nr_cpu_ids)); | |
3290 | ||
3291 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3292 | if (ret) | |
c75e8394 | 3293 | return ret; |
ae18c37a JB |
3294 | |
3295 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3296 | GFP_KERNEL); | |
3297 | if (ret) | |
c75e8394 | 3298 | return ret; |
ae18c37a JB |
3299 | |
3300 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3301 | GFP_KERNEL); | |
c75e8394 JB |
3302 | if (!fs_info->delayed_root) |
3303 | return -ENOMEM; | |
ae18c37a JB |
3304 | btrfs_init_delayed_root(fs_info->delayed_root); |
3305 | ||
a0a1db70 FM |
3306 | if (sb_rdonly(sb)) |
3307 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3308 | ||
c75e8394 | 3309 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3310 | } |
3311 | ||
97f4dd09 NB |
3312 | static int btrfs_uuid_rescan_kthread(void *data) |
3313 | { | |
0d031dc4 | 3314 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
3315 | int ret; |
3316 | ||
3317 | /* | |
3318 | * 1st step is to iterate through the existing UUID tree and | |
3319 | * to delete all entries that contain outdated data. | |
3320 | * 2nd step is to add all missing entries to the UUID tree. | |
3321 | */ | |
3322 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3323 | if (ret < 0) { | |
c94bec2c JB |
3324 | if (ret != -EINTR) |
3325 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3326 | ret); | |
97f4dd09 NB |
3327 | up(&fs_info->uuid_tree_rescan_sem); |
3328 | return ret; | |
3329 | } | |
3330 | return btrfs_uuid_scan_kthread(data); | |
3331 | } | |
3332 | ||
3333 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3334 | { | |
3335 | struct task_struct *task; | |
3336 | ||
3337 | down(&fs_info->uuid_tree_rescan_sem); | |
3338 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3339 | if (IS_ERR(task)) { | |
3340 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3341 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3342 | up(&fs_info->uuid_tree_rescan_sem); | |
3343 | return PTR_ERR(task); | |
3344 | } | |
3345 | ||
3346 | return 0; | |
3347 | } | |
3348 | ||
8cd29088 BB |
3349 | /* |
3350 | * Some options only have meaning at mount time and shouldn't persist across | |
3351 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3352 | * code paths. | |
3353 | */ | |
3354 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3355 | { | |
3356 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3357 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3358 | } |
3359 | ||
44c0ca21 BB |
3360 | /* |
3361 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3362 | * and btrfs_remount when remounting from read-only to read-write. | |
3363 | */ | |
3364 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3365 | { | |
3366 | int ret; | |
94846229 | 3367 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3368 | bool clear_free_space_tree = false; |
3369 | ||
3370 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3371 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3372 | clear_free_space_tree = true; | |
3373 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3374 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3375 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3376 | clear_free_space_tree = true; | |
3377 | } | |
3378 | ||
3379 | if (clear_free_space_tree) { | |
3380 | btrfs_info(fs_info, "clearing free space tree"); | |
3381 | ret = btrfs_clear_free_space_tree(fs_info); | |
3382 | if (ret) { | |
3383 | btrfs_warn(fs_info, | |
3384 | "failed to clear free space tree: %d", ret); | |
3385 | goto out; | |
3386 | } | |
3387 | } | |
44c0ca21 | 3388 | |
8d488a8c FM |
3389 | /* |
3390 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3391 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3392 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3393 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3394 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3395 | * item before the root's tree is deleted - this means that if we unmount | |
3396 | * or crash before the deletion completes, on the next mount we will not | |
3397 | * delete what remains of the tree because the orphan item does not | |
3398 | * exists anymore, which is what tells us we have a pending deletion. | |
3399 | */ | |
3400 | ret = btrfs_find_orphan_roots(fs_info); | |
3401 | if (ret) | |
3402 | goto out; | |
3403 | ||
44c0ca21 BB |
3404 | ret = btrfs_cleanup_fs_roots(fs_info); |
3405 | if (ret) | |
3406 | goto out; | |
3407 | ||
8f1c21d7 BB |
3408 | down_read(&fs_info->cleanup_work_sem); |
3409 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3410 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3411 | up_read(&fs_info->cleanup_work_sem); | |
3412 | goto out; | |
3413 | } | |
3414 | up_read(&fs_info->cleanup_work_sem); | |
3415 | ||
44c0ca21 | 3416 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3417 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3418 | mutex_unlock(&fs_info->cleaner_mutex); |
3419 | if (ret < 0) { | |
3420 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3421 | goto out; | |
3422 | } | |
3423 | ||
5011139a BB |
3424 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3425 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3426 | btrfs_info(fs_info, "creating free space tree"); | |
3427 | ret = btrfs_create_free_space_tree(fs_info); | |
3428 | if (ret) { | |
3429 | btrfs_warn(fs_info, | |
3430 | "failed to create free space tree: %d", ret); | |
3431 | goto out; | |
3432 | } | |
3433 | } | |
3434 | ||
94846229 BB |
3435 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3436 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3437 | if (ret) | |
3438 | goto out; | |
3439 | } | |
3440 | ||
44c0ca21 BB |
3441 | ret = btrfs_resume_balance_async(fs_info); |
3442 | if (ret) | |
3443 | goto out; | |
3444 | ||
3445 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3446 | if (ret) { | |
3447 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3448 | goto out; | |
3449 | } | |
3450 | ||
3451 | btrfs_qgroup_rescan_resume(fs_info); | |
3452 | ||
3453 | if (!fs_info->uuid_root) { | |
3454 | btrfs_info(fs_info, "creating UUID tree"); | |
3455 | ret = btrfs_create_uuid_tree(fs_info); | |
3456 | if (ret) { | |
3457 | btrfs_warn(fs_info, | |
3458 | "failed to create the UUID tree %d", ret); | |
3459 | goto out; | |
3460 | } | |
3461 | } | |
3462 | ||
3463 | out: | |
3464 | return ret; | |
3465 | } | |
3466 | ||
ae18c37a JB |
3467 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3468 | char *options) | |
3469 | { | |
3470 | u32 sectorsize; | |
3471 | u32 nodesize; | |
3472 | u32 stripesize; | |
3473 | u64 generation; | |
3474 | u64 features; | |
3475 | u16 csum_type; | |
ae18c37a JB |
3476 | struct btrfs_super_block *disk_super; |
3477 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3478 | struct btrfs_root *tree_root; | |
3479 | struct btrfs_root *chunk_root; | |
3480 | int ret; | |
3481 | int err = -EINVAL; | |
ae18c37a JB |
3482 | int level; |
3483 | ||
8260edba | 3484 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3485 | if (ret) { |
83c8266a | 3486 | err = ret; |
ae18c37a | 3487 | goto fail; |
53b381b3 DW |
3488 | } |
3489 | ||
ae18c37a JB |
3490 | /* These need to be init'ed before we start creating inodes and such. */ |
3491 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3492 | GFP_KERNEL); | |
3493 | fs_info->tree_root = tree_root; | |
3494 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3495 | GFP_KERNEL); | |
3496 | fs_info->chunk_root = chunk_root; | |
3497 | if (!tree_root || !chunk_root) { | |
3498 | err = -ENOMEM; | |
c75e8394 | 3499 | goto fail; |
ae18c37a JB |
3500 | } |
3501 | ||
3502 | fs_info->btree_inode = new_inode(sb); | |
3503 | if (!fs_info->btree_inode) { | |
3504 | err = -ENOMEM; | |
c75e8394 | 3505 | goto fail; |
ae18c37a JB |
3506 | } |
3507 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3508 | btrfs_init_btree_inode(fs_info); | |
3509 | ||
d24fa5c1 | 3510 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3511 | |
3512 | /* | |
3513 | * Read super block and check the signature bytes only | |
3514 | */ | |
d24fa5c1 | 3515 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3516 | if (IS_ERR(disk_super)) { |
3517 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3518 | goto fail_alloc; |
20b45077 | 3519 | } |
39279cc3 | 3520 | |
8dc3f22c | 3521 | /* |
260db43c | 3522 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3523 | * corrupted, we'll find out |
3524 | */ | |
8f32380d | 3525 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3526 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3527 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3528 | csum_type); |
8dc3f22c | 3529 | err = -EINVAL; |
8f32380d | 3530 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3531 | goto fail_alloc; |
3532 | } | |
3533 | ||
83c68bbc SY |
3534 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3535 | ||
6d97c6e3 JT |
3536 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3537 | if (ret) { | |
3538 | err = ret; | |
8f32380d | 3539 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3540 | goto fail_alloc; |
3541 | } | |
3542 | ||
1104a885 DS |
3543 | /* |
3544 | * We want to check superblock checksum, the type is stored inside. | |
3545 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3546 | */ | |
8f32380d | 3547 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3548 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3549 | err = -EINVAL; |
8f32380d | 3550 | btrfs_release_disk_super(disk_super); |
141386e1 | 3551 | goto fail_alloc; |
1104a885 DS |
3552 | } |
3553 | ||
3554 | /* | |
3555 | * super_copy is zeroed at allocation time and we never touch the | |
3556 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3557 | * the whole block of INFO_SIZE | |
3558 | */ | |
8f32380d JT |
3559 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3560 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3561 | |
fbc6feae NB |
3562 | disk_super = fs_info->super_copy; |
3563 | ||
0b86a832 | 3564 | |
fbc6feae NB |
3565 | features = btrfs_super_flags(disk_super); |
3566 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3567 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3568 | btrfs_set_super_flags(disk_super, features); | |
3569 | btrfs_info(fs_info, | |
3570 | "found metadata UUID change in progress flag, clearing"); | |
3571 | } | |
3572 | ||
3573 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3574 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3575 | |
069ec957 | 3576 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3577 | if (ret) { |
05135f59 | 3578 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3579 | err = -EINVAL; |
141386e1 | 3580 | goto fail_alloc; |
1104a885 DS |
3581 | } |
3582 | ||
0f7d52f4 | 3583 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3584 | goto fail_alloc; |
0f7d52f4 | 3585 | |
acce952b | 3586 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3587 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3588 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3589 | |
75e7cb7f LB |
3590 | /* |
3591 | * In the long term, we'll store the compression type in the super | |
3592 | * block, and it'll be used for per file compression control. | |
3593 | */ | |
3594 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3595 | ||
6f93e834 AJ |
3596 | /* |
3597 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3598 | * than the page size. | |
3599 | */ | |
3600 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { | |
3601 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
3602 | btrfs_info(fs_info, | |
3603 | "flagging fs with big metadata feature"); | |
3604 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3605 | } | |
3606 | ||
3607 | /* Set up fs_info before parsing mount options */ | |
3608 | nodesize = btrfs_super_nodesize(disk_super); | |
3609 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3610 | stripesize = sectorsize; | |
3611 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3612 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3613 | ||
3614 | fs_info->nodesize = nodesize; | |
3615 | fs_info->sectorsize = sectorsize; | |
3616 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3617 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3618 | fs_info->stripesize = stripesize; | |
3619 | ||
2ff7e61e | 3620 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3621 | if (ret) { |
3622 | err = ret; | |
141386e1 | 3623 | goto fail_alloc; |
2b82032c | 3624 | } |
dfe25020 | 3625 | |
f2b636e8 JB |
3626 | features = btrfs_super_incompat_flags(disk_super) & |
3627 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3628 | if (features) { | |
05135f59 | 3629 | btrfs_err(fs_info, |
d5321a0f | 3630 | "cannot mount because of unsupported optional features (0x%llx)", |
05135f59 | 3631 | features); |
f2b636e8 | 3632 | err = -EINVAL; |
141386e1 | 3633 | goto fail_alloc; |
f2b636e8 JB |
3634 | } |
3635 | ||
5d4f98a2 | 3636 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3637 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3638 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3639 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3640 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3641 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3642 | |
3173a18f | 3643 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3644 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3645 | |
bc3f116f CM |
3646 | /* |
3647 | * mixed block groups end up with duplicate but slightly offset | |
3648 | * extent buffers for the same range. It leads to corruptions | |
3649 | */ | |
3650 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3651 | (sectorsize != nodesize)) { |
05135f59 DS |
3652 | btrfs_err(fs_info, |
3653 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3654 | nodesize, sectorsize); | |
141386e1 | 3655 | goto fail_alloc; |
bc3f116f CM |
3656 | } |
3657 | ||
ceda0864 MX |
3658 | /* |
3659 | * Needn't use the lock because there is no other task which will | |
3660 | * update the flag. | |
3661 | */ | |
a6fa6fae | 3662 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3663 | |
f2b636e8 JB |
3664 | features = btrfs_super_compat_ro_flags(disk_super) & |
3665 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3666 | if (!sb_rdonly(sb) && features) { |
05135f59 | 3667 | btrfs_err(fs_info, |
d5321a0f | 3668 | "cannot mount read-write because of unsupported optional features (0x%llx)", |
c1c9ff7c | 3669 | features); |
f2b636e8 | 3670 | err = -EINVAL; |
141386e1 | 3671 | goto fail_alloc; |
f2b636e8 | 3672 | } |
61d92c32 | 3673 | |
8481dd80 QW |
3674 | if (sectorsize < PAGE_SIZE) { |
3675 | struct btrfs_subpage_info *subpage_info; | |
3676 | ||
9f73f1ae QW |
3677 | /* |
3678 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3679 | * going to be deprecated. | |
3680 | * | |
3681 | * Force to use v2 cache for subpage case. | |
3682 | */ | |
3683 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3684 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3685 | "forcing free space tree for sector size %u with page size %lu", | |
3686 | sectorsize, PAGE_SIZE); | |
3687 | ||
95ea0486 QW |
3688 | btrfs_warn(fs_info, |
3689 | "read-write for sector size %u with page size %lu is experimental", | |
3690 | sectorsize, PAGE_SIZE); | |
8481dd80 QW |
3691 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3692 | if (!subpage_info) | |
3693 | goto fail_alloc; | |
3694 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3695 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3696 | } |
0bb3eb3e | 3697 | |
d21deec5 | 3698 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3699 | if (ret) { |
3700 | err = ret; | |
0dc3b84a JB |
3701 | goto fail_sb_buffer; |
3702 | } | |
4543df7e | 3703 | |
9e11ceee JK |
3704 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3705 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3706 | |
a061fc8d CM |
3707 | sb->s_blocksize = sectorsize; |
3708 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3709 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3710 | |
925baedd | 3711 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3712 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3713 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3714 | if (ret) { |
05135f59 | 3715 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3716 | goto fail_sb_buffer; |
84eed90f | 3717 | } |
0b86a832 | 3718 | |
84234f3a | 3719 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3720 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3721 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3722 | generation, level); | |
3723 | if (ret) { | |
05135f59 | 3724 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3725 | goto fail_tree_roots; |
83121942 | 3726 | } |
0b86a832 | 3727 | |
e17cade2 | 3728 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3729 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3730 | BTRFS_UUID_SIZE); | |
e17cade2 | 3731 | |
5b4aacef | 3732 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3733 | if (ret) { |
05135f59 | 3734 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3735 | goto fail_tree_roots; |
2b82032c | 3736 | } |
0b86a832 | 3737 | |
8dabb742 | 3738 | /* |
bacce86a AJ |
3739 | * At this point we know all the devices that make this filesystem, |
3740 | * including the seed devices but we don't know yet if the replace | |
3741 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3742 | * filesystem but skip the replace target device which is checked |
bacce86a | 3743 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3744 | */ |
bacce86a | 3745 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3746 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3747 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3748 | goto fail_tree_roots; |
3749 | } | |
3750 | ||
b8522a1e | 3751 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3752 | if (ret) |
b8522a1e | 3753 | goto fail_tree_roots; |
8929ecfa | 3754 | |
73651042 NA |
3755 | /* |
3756 | * Get zone type information of zoned block devices. This will also | |
3757 | * handle emulation of a zoned filesystem if a regular device has the | |
3758 | * zoned incompat feature flag set. | |
3759 | */ | |
3760 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3761 | if (ret) { | |
3762 | btrfs_err(fs_info, | |
3763 | "zoned: failed to read device zone info: %d", | |
3764 | ret); | |
3765 | goto fail_block_groups; | |
3766 | } | |
3767 | ||
75ec1db8 JB |
3768 | /* |
3769 | * If we have a uuid root and we're not being told to rescan we need to | |
3770 | * check the generation here so we can set the | |
3771 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3772 | * transaction during a balance or the log replay without updating the | |
3773 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3774 | * even though it was perfectly fine. | |
3775 | */ | |
3776 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3777 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3778 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3779 | ||
cf90d884 QW |
3780 | ret = btrfs_verify_dev_extents(fs_info); |
3781 | if (ret) { | |
3782 | btrfs_err(fs_info, | |
3783 | "failed to verify dev extents against chunks: %d", | |
3784 | ret); | |
3785 | goto fail_block_groups; | |
3786 | } | |
68310a5e ID |
3787 | ret = btrfs_recover_balance(fs_info); |
3788 | if (ret) { | |
05135f59 | 3789 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3790 | goto fail_block_groups; |
3791 | } | |
3792 | ||
733f4fbb SB |
3793 | ret = btrfs_init_dev_stats(fs_info); |
3794 | if (ret) { | |
05135f59 | 3795 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3796 | goto fail_block_groups; |
3797 | } | |
3798 | ||
8dabb742 SB |
3799 | ret = btrfs_init_dev_replace(fs_info); |
3800 | if (ret) { | |
05135f59 | 3801 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3802 | goto fail_block_groups; |
3803 | } | |
3804 | ||
b70f5097 NA |
3805 | ret = btrfs_check_zoned_mode(fs_info); |
3806 | if (ret) { | |
3807 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3808 | ret); | |
3809 | goto fail_block_groups; | |
3810 | } | |
3811 | ||
c6761a9e | 3812 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3813 | if (ret) { |
05135f59 DS |
3814 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3815 | ret); | |
b7c35e81 AJ |
3816 | goto fail_block_groups; |
3817 | } | |
3818 | ||
96f3136e | 3819 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3820 | if (ret) { |
05135f59 | 3821 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3822 | goto fail_fsdev_sysfs; |
c59021f8 | 3823 | } |
3824 | ||
c59021f8 | 3825 | ret = btrfs_init_space_info(fs_info); |
3826 | if (ret) { | |
05135f59 | 3827 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3828 | goto fail_sysfs; |
c59021f8 | 3829 | } |
3830 | ||
5b4aacef | 3831 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3832 | if (ret) { |
05135f59 | 3833 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3834 | goto fail_sysfs; |
1b1d1f66 | 3835 | } |
4330e183 | 3836 | |
16beac87 NA |
3837 | btrfs_free_zone_cache(fs_info); |
3838 | ||
5c78a5e7 AJ |
3839 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3840 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3841 | btrfs_warn(fs_info, |
52042d8e | 3842 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3843 | goto fail_sysfs; |
292fd7fc | 3844 | } |
9078a3e1 | 3845 | |
33c44184 | 3846 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3847 | "btrfs-cleaner"); |
57506d50 | 3848 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3849 | goto fail_sysfs; |
a74a4b97 CM |
3850 | |
3851 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3852 | tree_root, | |
3853 | "btrfs-transaction"); | |
57506d50 | 3854 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3855 | goto fail_cleaner; |
a74a4b97 | 3856 | |
583b7231 | 3857 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3858 | !fs_info->fs_devices->rotating) { |
583b7231 | 3859 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3860 | } |
3861 | ||
572d9ab7 | 3862 | /* |
01327610 | 3863 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3864 | * not have to wait for transaction commit |
3865 | */ | |
3866 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3867 | |
21adbd5c | 3868 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3869 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3870 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3871 | btrfs_test_opt(fs_info, |
cbeaae4f | 3872 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3873 | fs_info->check_integrity_print_mask); |
3874 | if (ret) | |
05135f59 DS |
3875 | btrfs_warn(fs_info, |
3876 | "failed to initialize integrity check module: %d", | |
3877 | ret); | |
21adbd5c SB |
3878 | } |
3879 | #endif | |
bcef60f2 AJ |
3880 | ret = btrfs_read_qgroup_config(fs_info); |
3881 | if (ret) | |
3882 | goto fail_trans_kthread; | |
21adbd5c | 3883 | |
fd708b81 JB |
3884 | if (btrfs_build_ref_tree(fs_info)) |
3885 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3886 | ||
96da0919 QW |
3887 | /* do not make disk changes in broken FS or nologreplay is given */ |
3888 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3889 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3890 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3891 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3892 | if (ret) { |
63443bf5 | 3893 | err = ret; |
28c16cbb | 3894 | goto fail_qgroup; |
79787eaa | 3895 | } |
e02119d5 | 3896 | } |
1a40e23b | 3897 | |
56e9357a | 3898 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3899 | if (IS_ERR(fs_info->fs_root)) { |
3900 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3901 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3902 | fs_info->fs_root = NULL; |
bcef60f2 | 3903 | goto fail_qgroup; |
3140c9a3 | 3904 | } |
c289811c | 3905 | |
bc98a42c | 3906 | if (sb_rdonly(sb)) |
8cd29088 | 3907 | goto clear_oneshot; |
59641015 | 3908 | |
44c0ca21 | 3909 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3910 | if (ret) { |
6bccf3ab | 3911 | close_ctree(fs_info); |
2b6ba629 | 3912 | return ret; |
e3acc2a6 | 3913 | } |
b0643e59 | 3914 | btrfs_discard_resume(fs_info); |
b382a324 | 3915 | |
44c0ca21 BB |
3916 | if (fs_info->uuid_root && |
3917 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3918 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3919 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3920 | ret = btrfs_check_uuid_tree(fs_info); |
3921 | if (ret) { | |
05135f59 DS |
3922 | btrfs_warn(fs_info, |
3923 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3924 | close_ctree(fs_info); |
70f80175 SB |
3925 | return ret; |
3926 | } | |
f7a81ea4 | 3927 | } |
94846229 | 3928 | |
afcdd129 | 3929 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3930 | |
b4be6aef JB |
3931 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3932 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3933 | wake_up_process(fs_info->cleaner_kthread); | |
3934 | ||
8cd29088 BB |
3935 | clear_oneshot: |
3936 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3937 | return 0; |
39279cc3 | 3938 | |
bcef60f2 AJ |
3939 | fail_qgroup: |
3940 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3941 | fail_trans_kthread: |
3942 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3943 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3944 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3945 | fail_cleaner: |
a74a4b97 | 3946 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3947 | |
3948 | /* | |
3949 | * make sure we're done with the btree inode before we stop our | |
3950 | * kthreads | |
3951 | */ | |
3952 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3953 | |
2365dd3c | 3954 | fail_sysfs: |
6618a59b | 3955 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3956 | |
b7c35e81 AJ |
3957 | fail_fsdev_sysfs: |
3958 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3959 | ||
1b1d1f66 | 3960 | fail_block_groups: |
54067ae9 | 3961 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3962 | |
3963 | fail_tree_roots: | |
9e3aa805 JB |
3964 | if (fs_info->data_reloc_root) |
3965 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3966 | free_root_pointers(fs_info, true); |
2b8195bb | 3967 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3968 | |
39279cc3 | 3969 | fail_sb_buffer: |
7abadb64 | 3970 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3971 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3972 | fail_alloc: |
586e46e2 ID |
3973 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3974 | ||
4543df7e | 3975 | iput(fs_info->btree_inode); |
7e662854 | 3976 | fail: |
586e46e2 | 3977 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3978 | return err; |
eb60ceac | 3979 | } |
663faf9f | 3980 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3981 | |
314b6dd0 | 3982 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3983 | { |
314b6dd0 JT |
3984 | struct btrfs_device *device = bio->bi_private; |
3985 | struct bio_vec *bvec; | |
3986 | struct bvec_iter_all iter_all; | |
3987 | struct page *page; | |
3988 | ||
3989 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3990 | page = bvec->bv_page; | |
3991 | ||
3992 | if (bio->bi_status) { | |
3993 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3994 | "lost page write due to IO error on %s (%d)", | |
3995 | rcu_str_deref(device->name), | |
3996 | blk_status_to_errno(bio->bi_status)); | |
3997 | ClearPageUptodate(page); | |
3998 | SetPageError(page); | |
3999 | btrfs_dev_stat_inc_and_print(device, | |
4000 | BTRFS_DEV_STAT_WRITE_ERRS); | |
4001 | } else { | |
4002 | SetPageUptodate(page); | |
4003 | } | |
4004 | ||
4005 | put_page(page); | |
4006 | unlock_page(page); | |
f2984462 | 4007 | } |
314b6dd0 JT |
4008 | |
4009 | bio_put(bio); | |
f2984462 CM |
4010 | } |
4011 | ||
8f32380d JT |
4012 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
4013 | int copy_num) | |
29c36d72 | 4014 | { |
29c36d72 | 4015 | struct btrfs_super_block *super; |
8f32380d | 4016 | struct page *page; |
12659251 | 4017 | u64 bytenr, bytenr_orig; |
8f32380d | 4018 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
4019 | int ret; |
4020 | ||
4021 | bytenr_orig = btrfs_sb_offset(copy_num); | |
4022 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
4023 | if (ret == -ENOENT) | |
4024 | return ERR_PTR(-EINVAL); | |
4025 | else if (ret) | |
4026 | return ERR_PTR(ret); | |
29c36d72 | 4027 | |
cda00eba | 4028 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 4029 | return ERR_PTR(-EINVAL); |
29c36d72 | 4030 | |
8f32380d JT |
4031 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
4032 | if (IS_ERR(page)) | |
4033 | return ERR_CAST(page); | |
29c36d72 | 4034 | |
8f32380d | 4035 | super = page_address(page); |
96c2e067 AJ |
4036 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
4037 | btrfs_release_disk_super(super); | |
4038 | return ERR_PTR(-ENODATA); | |
4039 | } | |
4040 | ||
12659251 | 4041 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
4042 | btrfs_release_disk_super(super); |
4043 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
4044 | } |
4045 | ||
8f32380d | 4046 | return super; |
29c36d72 AJ |
4047 | } |
4048 | ||
4049 | ||
8f32380d | 4050 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 4051 | { |
8f32380d | 4052 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
4053 | int i; |
4054 | u64 transid = 0; | |
a512bbf8 YZ |
4055 | |
4056 | /* we would like to check all the supers, but that would make | |
4057 | * a btrfs mount succeed after a mkfs from a different FS. | |
4058 | * So, we need to add a special mount option to scan for | |
4059 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
4060 | */ | |
4061 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
4062 | super = btrfs_read_dev_one_super(bdev, i); |
4063 | if (IS_ERR(super)) | |
a512bbf8 YZ |
4064 | continue; |
4065 | ||
a512bbf8 | 4066 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
4067 | if (latest) |
4068 | btrfs_release_disk_super(super); | |
4069 | ||
4070 | latest = super; | |
a512bbf8 | 4071 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
4072 | } |
4073 | } | |
92fc03fb | 4074 | |
8f32380d | 4075 | return super; |
a512bbf8 YZ |
4076 | } |
4077 | ||
4eedeb75 | 4078 | /* |
abbb3b8e | 4079 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 4080 | * pages we use for writing are locked. |
4eedeb75 | 4081 | * |
abbb3b8e DS |
4082 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
4083 | * the expected device size at commit time. Note that max_mirrors must be | |
4084 | * same for write and wait phases. | |
4eedeb75 | 4085 | * |
314b6dd0 | 4086 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 4087 | */ |
a512bbf8 | 4088 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 4089 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 4090 | { |
d5178578 | 4091 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 4092 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 4093 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 4094 | int i; |
a512bbf8 | 4095 | int errors = 0; |
12659251 NA |
4096 | int ret; |
4097 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
4098 | |
4099 | if (max_mirrors == 0) | |
4100 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4101 | ||
d5178578 JT |
4102 | shash->tfm = fs_info->csum_shash; |
4103 | ||
a512bbf8 | 4104 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
4105 | struct page *page; |
4106 | struct bio *bio; | |
4107 | struct btrfs_super_block *disk_super; | |
4108 | ||
12659251 NA |
4109 | bytenr_orig = btrfs_sb_offset(i); |
4110 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
4111 | if (ret == -ENOENT) { | |
4112 | continue; | |
4113 | } else if (ret < 0) { | |
4114 | btrfs_err(device->fs_info, | |
4115 | "couldn't get super block location for mirror %d", | |
4116 | i); | |
4117 | errors++; | |
4118 | continue; | |
4119 | } | |
935e5cc9 MX |
4120 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4121 | device->commit_total_bytes) | |
a512bbf8 YZ |
4122 | break; |
4123 | ||
12659251 | 4124 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 4125 | |
fd08001f EB |
4126 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
4127 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
4128 | sb->csum); | |
4eedeb75 | 4129 | |
314b6dd0 JT |
4130 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
4131 | GFP_NOFS); | |
4132 | if (!page) { | |
abbb3b8e | 4133 | btrfs_err(device->fs_info, |
314b6dd0 | 4134 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
4135 | bytenr); |
4136 | errors++; | |
4eedeb75 | 4137 | continue; |
abbb3b8e | 4138 | } |
634554dc | 4139 | |
314b6dd0 JT |
4140 | /* Bump the refcount for wait_dev_supers() */ |
4141 | get_page(page); | |
a512bbf8 | 4142 | |
314b6dd0 JT |
4143 | disk_super = page_address(page); |
4144 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 4145 | |
314b6dd0 JT |
4146 | /* |
4147 | * Directly use bios here instead of relying on the page cache | |
4148 | * to do I/O, so we don't lose the ability to do integrity | |
4149 | * checking. | |
4150 | */ | |
07888c66 CH |
4151 | bio = bio_alloc(device->bdev, 1, |
4152 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
4153 | GFP_NOFS); | |
314b6dd0 JT |
4154 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
4155 | bio->bi_private = device; | |
4156 | bio->bi_end_io = btrfs_end_super_write; | |
4157 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4158 | offset_in_page(bytenr)); | |
a512bbf8 | 4159 | |
387125fc | 4160 | /* |
314b6dd0 JT |
4161 | * We FUA only the first super block. The others we allow to |
4162 | * go down lazy and there's a short window where the on-disk | |
4163 | * copies might still contain the older version. | |
387125fc | 4164 | */ |
1b9e619c | 4165 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4166 | bio->bi_opf |= REQ_FUA; |
4167 | ||
58ff51f1 CH |
4168 | btrfsic_check_bio(bio); |
4169 | submit_bio(bio); | |
8376d9e1 NA |
4170 | |
4171 | if (btrfs_advance_sb_log(device, i)) | |
4172 | errors++; | |
a512bbf8 YZ |
4173 | } |
4174 | return errors < i ? 0 : -1; | |
4175 | } | |
4176 | ||
abbb3b8e DS |
4177 | /* |
4178 | * Wait for write completion of superblocks done by write_dev_supers, | |
4179 | * @max_mirrors same for write and wait phases. | |
4180 | * | |
314b6dd0 | 4181 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4182 | * date. |
4183 | */ | |
4184 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4185 | { | |
abbb3b8e DS |
4186 | int i; |
4187 | int errors = 0; | |
b6a535fa | 4188 | bool primary_failed = false; |
12659251 | 4189 | int ret; |
abbb3b8e DS |
4190 | u64 bytenr; |
4191 | ||
4192 | if (max_mirrors == 0) | |
4193 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4194 | ||
4195 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4196 | struct page *page; |
4197 | ||
12659251 NA |
4198 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4199 | if (ret == -ENOENT) { | |
4200 | break; | |
4201 | } else if (ret < 0) { | |
4202 | errors++; | |
4203 | if (i == 0) | |
4204 | primary_failed = true; | |
4205 | continue; | |
4206 | } | |
abbb3b8e DS |
4207 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4208 | device->commit_total_bytes) | |
4209 | break; | |
4210 | ||
314b6dd0 JT |
4211 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4212 | bytenr >> PAGE_SHIFT); | |
4213 | if (!page) { | |
abbb3b8e | 4214 | errors++; |
b6a535fa HM |
4215 | if (i == 0) |
4216 | primary_failed = true; | |
abbb3b8e DS |
4217 | continue; |
4218 | } | |
314b6dd0 JT |
4219 | /* Page is submitted locked and unlocked once the IO completes */ |
4220 | wait_on_page_locked(page); | |
4221 | if (PageError(page)) { | |
abbb3b8e | 4222 | errors++; |
b6a535fa HM |
4223 | if (i == 0) |
4224 | primary_failed = true; | |
4225 | } | |
abbb3b8e | 4226 | |
314b6dd0 JT |
4227 | /* Drop our reference */ |
4228 | put_page(page); | |
abbb3b8e | 4229 | |
314b6dd0 JT |
4230 | /* Drop the reference from the writing run */ |
4231 | put_page(page); | |
abbb3b8e DS |
4232 | } |
4233 | ||
b6a535fa HM |
4234 | /* log error, force error return */ |
4235 | if (primary_failed) { | |
4236 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4237 | device->devid); | |
4238 | return -1; | |
4239 | } | |
4240 | ||
abbb3b8e DS |
4241 | return errors < i ? 0 : -1; |
4242 | } | |
4243 | ||
387125fc CM |
4244 | /* |
4245 | * endio for the write_dev_flush, this will wake anyone waiting | |
4246 | * for the barrier when it is done | |
4247 | */ | |
4246a0b6 | 4248 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4249 | { |
e0ae9994 | 4250 | complete(bio->bi_private); |
387125fc CM |
4251 | } |
4252 | ||
4253 | /* | |
4fc6441a AJ |
4254 | * Submit a flush request to the device if it supports it. Error handling is |
4255 | * done in the waiting counterpart. | |
387125fc | 4256 | */ |
4fc6441a | 4257 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4258 | { |
e0ae9994 | 4259 | struct bio *bio = device->flush_bio; |
387125fc | 4260 | |
a91cf0ff WY |
4261 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4262 | /* | |
4263 | * When a disk has write caching disabled, we skip submission of a bio | |
4264 | * with flush and sync requests before writing the superblock, since | |
4265 | * it's not needed. However when the integrity checker is enabled, this | |
4266 | * results in reports that there are metadata blocks referred by a | |
4267 | * superblock that were not properly flushed. So don't skip the bio | |
4268 | * submission only when the integrity checker is enabled for the sake | |
4269 | * of simplicity, since this is a debug tool and not meant for use in | |
4270 | * non-debug builds. | |
4271 | */ | |
4272 | struct request_queue *q = bdev_get_queue(device->bdev); | |
c2a9c7ab | 4273 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 4274 | return; |
a91cf0ff | 4275 | #endif |
387125fc | 4276 | |
a7c50c94 | 4277 | bio_reset(bio, device->bdev, REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); |
387125fc | 4278 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
4279 | init_completion(&device->flush_wait); |
4280 | bio->bi_private = &device->flush_wait; | |
387125fc | 4281 | |
58ff51f1 CH |
4282 | btrfsic_check_bio(bio); |
4283 | submit_bio(bio); | |
1c3063b6 | 4284 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4285 | } |
387125fc | 4286 | |
4fc6441a AJ |
4287 | /* |
4288 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4289 | */ | |
8c27cb35 | 4290 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4291 | { |
4fc6441a | 4292 | struct bio *bio = device->flush_bio; |
387125fc | 4293 | |
1c3063b6 | 4294 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4295 | return BLK_STS_OK; |
387125fc | 4296 | |
1c3063b6 | 4297 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4298 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4299 | |
8c27cb35 | 4300 | return bio->bi_status; |
387125fc | 4301 | } |
387125fc | 4302 | |
d10b82fe | 4303 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4304 | { |
6528b99d | 4305 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4306 | return -EIO; |
387125fc CM |
4307 | return 0; |
4308 | } | |
4309 | ||
4310 | /* | |
4311 | * send an empty flush down to each device in parallel, | |
4312 | * then wait for them | |
4313 | */ | |
4314 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4315 | { | |
4316 | struct list_head *head; | |
4317 | struct btrfs_device *dev; | |
5af3e8cc | 4318 | int errors_wait = 0; |
4e4cbee9 | 4319 | blk_status_t ret; |
387125fc | 4320 | |
1538e6c5 | 4321 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4322 | /* send down all the barriers */ |
4323 | head = &info->fs_devices->devices; | |
1538e6c5 | 4324 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4325 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4326 | continue; |
cea7c8bf | 4327 | if (!dev->bdev) |
387125fc | 4328 | continue; |
e12c9621 | 4329 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4330 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4331 | continue; |
4332 | ||
4fc6441a | 4333 | write_dev_flush(dev); |
58efbc9f | 4334 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4335 | } |
4336 | ||
4337 | /* wait for all the barriers */ | |
1538e6c5 | 4338 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4339 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4340 | continue; |
387125fc | 4341 | if (!dev->bdev) { |
5af3e8cc | 4342 | errors_wait++; |
387125fc CM |
4343 | continue; |
4344 | } | |
e12c9621 | 4345 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4346 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4347 | continue; |
4348 | ||
4fc6441a | 4349 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4350 | if (ret) { |
4351 | dev->last_flush_error = ret; | |
66b4993e DS |
4352 | btrfs_dev_stat_inc_and_print(dev, |
4353 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4354 | errors_wait++; |
401b41e5 AJ |
4355 | } |
4356 | } | |
4357 | ||
cea7c8bf | 4358 | if (errors_wait) { |
401b41e5 AJ |
4359 | /* |
4360 | * At some point we need the status of all disks | |
4361 | * to arrive at the volume status. So error checking | |
4362 | * is being pushed to a separate loop. | |
4363 | */ | |
d10b82fe | 4364 | return check_barrier_error(info); |
387125fc | 4365 | } |
387125fc CM |
4366 | return 0; |
4367 | } | |
4368 | ||
943c6e99 ZL |
4369 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4370 | { | |
8789f4fe ZL |
4371 | int raid_type; |
4372 | int min_tolerated = INT_MAX; | |
943c6e99 | 4373 | |
8789f4fe ZL |
4374 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4375 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4376 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4377 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4378 | tolerated_failures); | |
943c6e99 | 4379 | |
8789f4fe ZL |
4380 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4381 | if (raid_type == BTRFS_RAID_SINGLE) | |
4382 | continue; | |
41a6e891 | 4383 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4384 | continue; |
8c3e3582 | 4385 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4386 | btrfs_raid_array[raid_type]. |
4387 | tolerated_failures); | |
4388 | } | |
943c6e99 | 4389 | |
8789f4fe | 4390 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4391 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4392 | min_tolerated = 0; |
4393 | } | |
4394 | ||
4395 | return min_tolerated; | |
943c6e99 ZL |
4396 | } |
4397 | ||
eece6a9c | 4398 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4399 | { |
e5e9a520 | 4400 | struct list_head *head; |
f2984462 | 4401 | struct btrfs_device *dev; |
a061fc8d | 4402 | struct btrfs_super_block *sb; |
f2984462 | 4403 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4404 | int ret; |
4405 | int do_barriers; | |
a236aed1 CM |
4406 | int max_errors; |
4407 | int total_errors = 0; | |
a061fc8d | 4408 | u64 flags; |
f2984462 | 4409 | |
0b246afa | 4410 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4411 | |
4412 | /* | |
4413 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4414 | * not from fsync where the tree roots in fs_info have not | |
4415 | * been consistent on disk. | |
4416 | */ | |
4417 | if (max_mirrors == 0) | |
4418 | backup_super_roots(fs_info); | |
f2984462 | 4419 | |
0b246afa | 4420 | sb = fs_info->super_for_commit; |
a061fc8d | 4421 | dev_item = &sb->dev_item; |
e5e9a520 | 4422 | |
0b246afa JM |
4423 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4424 | head = &fs_info->fs_devices->devices; | |
4425 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4426 | |
5af3e8cc | 4427 | if (do_barriers) { |
0b246afa | 4428 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4429 | if (ret) { |
4430 | mutex_unlock( | |
0b246afa JM |
4431 | &fs_info->fs_devices->device_list_mutex); |
4432 | btrfs_handle_fs_error(fs_info, ret, | |
4433 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4434 | return ret; |
4435 | } | |
4436 | } | |
387125fc | 4437 | |
1538e6c5 | 4438 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4439 | if (!dev->bdev) { |
4440 | total_errors++; | |
4441 | continue; | |
4442 | } | |
e12c9621 | 4443 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4444 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4445 | continue; |
4446 | ||
2b82032c | 4447 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4448 | btrfs_set_stack_device_type(dev_item, dev->type); |
4449 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4450 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4451 | dev->commit_total_bytes); |
ce7213c7 MX |
4452 | btrfs_set_stack_device_bytes_used(dev_item, |
4453 | dev->commit_bytes_used); | |
a061fc8d CM |
4454 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4455 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4456 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4457 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4458 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4459 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4460 | |
a061fc8d CM |
4461 | flags = btrfs_super_flags(sb); |
4462 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4463 | ||
75cb857d QW |
4464 | ret = btrfs_validate_write_super(fs_info, sb); |
4465 | if (ret < 0) { | |
4466 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4467 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4468 | "unexpected superblock corruption detected"); | |
4469 | return -EUCLEAN; | |
4470 | } | |
4471 | ||
abbb3b8e | 4472 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4473 | if (ret) |
4474 | total_errors++; | |
f2984462 | 4475 | } |
a236aed1 | 4476 | if (total_errors > max_errors) { |
0b246afa JM |
4477 | btrfs_err(fs_info, "%d errors while writing supers", |
4478 | total_errors); | |
4479 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4480 | |
9d565ba4 | 4481 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4482 | btrfs_handle_fs_error(fs_info, -EIO, |
4483 | "%d errors while writing supers", | |
4484 | total_errors); | |
9d565ba4 | 4485 | return -EIO; |
a236aed1 | 4486 | } |
f2984462 | 4487 | |
a512bbf8 | 4488 | total_errors = 0; |
1538e6c5 | 4489 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4490 | if (!dev->bdev) |
4491 | continue; | |
e12c9621 | 4492 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4493 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4494 | continue; |
4495 | ||
abbb3b8e | 4496 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4497 | if (ret) |
4498 | total_errors++; | |
f2984462 | 4499 | } |
0b246afa | 4500 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4501 | if (total_errors > max_errors) { |
0b246afa JM |
4502 | btrfs_handle_fs_error(fs_info, -EIO, |
4503 | "%d errors while writing supers", | |
4504 | total_errors); | |
79787eaa | 4505 | return -EIO; |
a236aed1 | 4506 | } |
f2984462 CM |
4507 | return 0; |
4508 | } | |
4509 | ||
cb517eab MX |
4510 | /* Drop a fs root from the radix tree and free it. */ |
4511 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4512 | struct btrfs_root *root) | |
2619ba1f | 4513 | { |
4785e24f JB |
4514 | bool drop_ref = false; |
4515 | ||
fc7cbcd4 DS |
4516 | spin_lock(&fs_info->fs_roots_radix_lock); |
4517 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4518 | (unsigned long)root->root_key.objectid); | |
4519 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4520 | drop_ref = true; |
fc7cbcd4 | 4521 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4522 | |
84961539 | 4523 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4524 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4525 | if (root->reloc_root) { |
00246528 | 4526 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4527 | root->reloc_root = NULL; |
4528 | } | |
4529 | } | |
3321719e | 4530 | |
4785e24f JB |
4531 | if (drop_ref) |
4532 | btrfs_put_root(root); | |
2619ba1f CM |
4533 | } |
4534 | ||
c146afad | 4535 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4536 | { |
fc7cbcd4 DS |
4537 | u64 root_objectid = 0; |
4538 | struct btrfs_root *gang[8]; | |
4539 | int i = 0; | |
65d33fd7 | 4540 | int err = 0; |
fc7cbcd4 | 4541 | unsigned int ret = 0; |
e089f05c | 4542 | |
c146afad | 4543 | while (1) { |
fc7cbcd4 DS |
4544 | spin_lock(&fs_info->fs_roots_radix_lock); |
4545 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4546 | (void **)gang, root_objectid, | |
4547 | ARRAY_SIZE(gang)); | |
4548 | if (!ret) { | |
4549 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4550 | break; | |
65d33fd7 | 4551 | } |
fc7cbcd4 | 4552 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4553 | |
fc7cbcd4 DS |
4554 | for (i = 0; i < ret; i++) { |
4555 | /* Avoid to grab roots in dead_roots */ | |
4556 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4557 | gang[i] = NULL; | |
4558 | continue; | |
4559 | } | |
4560 | /* grab all the search result for later use */ | |
4561 | gang[i] = btrfs_grab_root(gang[i]); | |
65d33fd7 | 4562 | } |
fc7cbcd4 | 4563 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4564 | |
fc7cbcd4 DS |
4565 | for (i = 0; i < ret; i++) { |
4566 | if (!gang[i]) | |
65d33fd7 | 4567 | continue; |
fc7cbcd4 DS |
4568 | root_objectid = gang[i]->root_key.objectid; |
4569 | err = btrfs_orphan_cleanup(gang[i]); | |
66b4ffd1 | 4570 | if (err) |
fc7cbcd4 DS |
4571 | break; |
4572 | btrfs_put_root(gang[i]); | |
c146afad | 4573 | } |
fc7cbcd4 | 4574 | root_objectid++; |
c146afad | 4575 | } |
65d33fd7 | 4576 | |
fc7cbcd4 DS |
4577 | /* release the uncleaned roots due to error */ |
4578 | for (; i < ret; i++) { | |
4579 | if (gang[i]) | |
4580 | btrfs_put_root(gang[i]); | |
65d33fd7 QW |
4581 | } |
4582 | return err; | |
c146afad | 4583 | } |
a2135011 | 4584 | |
6bccf3ab | 4585 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4586 | { |
6bccf3ab | 4587 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4588 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4589 | |
0b246afa | 4590 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4591 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4592 | mutex_unlock(&fs_info->cleaner_mutex); |
4593 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4594 | |
4595 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4596 | down_write(&fs_info->cleanup_work_sem); |
4597 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4598 | |
7a7eaa40 | 4599 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4600 | if (IS_ERR(trans)) |
4601 | return PTR_ERR(trans); | |
3a45bb20 | 4602 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4603 | } |
4604 | ||
36c86a9e QW |
4605 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4606 | { | |
4607 | struct btrfs_transaction *trans; | |
4608 | struct btrfs_transaction *tmp; | |
4609 | bool found = false; | |
4610 | ||
4611 | if (list_empty(&fs_info->trans_list)) | |
4612 | return; | |
4613 | ||
4614 | /* | |
4615 | * This function is only called at the very end of close_ctree(), | |
4616 | * thus no other running transaction, no need to take trans_lock. | |
4617 | */ | |
4618 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4619 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4620 | struct extent_state *cached = NULL; | |
4621 | u64 dirty_bytes = 0; | |
4622 | u64 cur = 0; | |
4623 | u64 found_start; | |
4624 | u64 found_end; | |
4625 | ||
4626 | found = true; | |
4627 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4628 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4629 | dirty_bytes += found_end + 1 - found_start; | |
4630 | cur = found_end + 1; | |
4631 | } | |
4632 | btrfs_warn(fs_info, | |
4633 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4634 | trans->transid, dirty_bytes); | |
4635 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4636 | ||
4637 | if (trans == fs_info->running_transaction) | |
4638 | fs_info->running_transaction = NULL; | |
4639 | list_del_init(&trans->list); | |
4640 | ||
4641 | btrfs_put_transaction(trans); | |
4642 | trace_btrfs_transaction_commit(fs_info); | |
4643 | } | |
4644 | ASSERT(!found); | |
4645 | } | |
4646 | ||
b105e927 | 4647 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4648 | { |
c146afad YZ |
4649 | int ret; |
4650 | ||
afcdd129 | 4651 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 FM |
4652 | |
4653 | /* | |
4654 | * We may have the reclaim task running and relocating a data block group, | |
4655 | * in which case it may create delayed iputs. So stop it before we park | |
4656 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4657 | * parking the cleaner, and that can make the async reclaim task to hang | |
4658 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4659 | * parked and can not run delayed iputs - this will make us hang when | |
4660 | * trying to stop the async reclaim task. | |
4661 | */ | |
4662 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4663 | /* |
4664 | * We don't want the cleaner to start new transactions, add more delayed | |
4665 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4666 | * because that frees the task_struct, and the transaction kthread might | |
4667 | * still try to wake up the cleaner. | |
4668 | */ | |
4669 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4670 | |
b4be6aef JB |
4671 | /* |
4672 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4673 | * clear that bit and wake up relocation so it can stop. | |
4674 | */ | |
4675 | btrfs_wake_unfinished_drop(fs_info); | |
4676 | ||
7343dd61 | 4677 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4678 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4679 | |
803b2f54 SB |
4680 | /* wait for the uuid_scan task to finish */ |
4681 | down(&fs_info->uuid_tree_rescan_sem); | |
4682 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4683 | up(&fs_info->uuid_tree_rescan_sem); | |
4684 | ||
837d5b6e | 4685 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4686 | btrfs_pause_balance(fs_info); |
837d5b6e | 4687 | |
8dabb742 SB |
4688 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4689 | ||
aa1b8cd4 | 4690 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4691 | |
4692 | /* wait for any defraggers to finish */ | |
4693 | wait_event(fs_info->transaction_wait, | |
4694 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4695 | ||
4696 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4697 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4698 | |
21c7e756 | 4699 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4700 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4701 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4702 | |
b0643e59 DZ |
4703 | /* Cancel or finish ongoing discard work */ |
4704 | btrfs_discard_cleanup(fs_info); | |
4705 | ||
bc98a42c | 4706 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4707 | /* |
d6fd0ae2 OS |
4708 | * The cleaner kthread is stopped, so do one final pass over |
4709 | * unused block groups. | |
e44163e1 | 4710 | */ |
0b246afa | 4711 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4712 | |
f0cc2cd7 FM |
4713 | /* |
4714 | * There might be existing delayed inode workers still running | |
4715 | * and holding an empty delayed inode item. We must wait for | |
4716 | * them to complete first because they can create a transaction. | |
4717 | * This happens when someone calls btrfs_balance_delayed_items() | |
4718 | * and then a transaction commit runs the same delayed nodes | |
4719 | * before any delayed worker has done something with the nodes. | |
4720 | * We must wait for any worker here and not at transaction | |
4721 | * commit time since that could cause a deadlock. | |
4722 | * This is a very rare case. | |
4723 | */ | |
4724 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4725 | ||
6bccf3ab | 4726 | ret = btrfs_commit_super(fs_info); |
acce952b | 4727 | if (ret) |
04892340 | 4728 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4729 | } |
4730 | ||
84961539 | 4731 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4732 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4733 | |
e3029d9f AV |
4734 | kthread_stop(fs_info->transaction_kthread); |
4735 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4736 | |
e187831e | 4737 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4738 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4739 | |
5958253c QW |
4740 | if (btrfs_check_quota_leak(fs_info)) { |
4741 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4742 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4743 | } | |
4744 | ||
04892340 | 4745 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4746 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4747 | |
963d678b | 4748 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4749 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4750 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4751 | } |
bcc63abb | 4752 | |
5deb17e1 | 4753 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4754 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4755 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4756 | |
6618a59b | 4757 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4758 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4759 | |
1a4319cc LB |
4760 | btrfs_put_block_group_cache(fs_info); |
4761 | ||
de348ee0 WS |
4762 | /* |
4763 | * we must make sure there is not any read request to | |
4764 | * submit after we stopping all workers. | |
4765 | */ | |
4766 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4767 | btrfs_stop_all_workers(fs_info); |
4768 | ||
0a31daa4 | 4769 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4770 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4771 | |
afcdd129 | 4772 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4773 | free_root_pointers(fs_info, true); |
8c38938c | 4774 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4775 | |
4e19443d JB |
4776 | /* |
4777 | * We must free the block groups after dropping the fs_roots as we could | |
4778 | * have had an IO error and have left over tree log blocks that aren't | |
4779 | * cleaned up until the fs roots are freed. This makes the block group | |
4780 | * accounting appear to be wrong because there's pending reserved bytes, | |
4781 | * so make sure we do the block group cleanup afterwards. | |
4782 | */ | |
4783 | btrfs_free_block_groups(fs_info); | |
4784 | ||
13e6c37b | 4785 | iput(fs_info->btree_inode); |
d6bfde87 | 4786 | |
21adbd5c | 4787 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4788 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4789 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4790 | #endif |
4791 | ||
0b86a832 | 4792 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4793 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4794 | } |
4795 | ||
b9fab919 CM |
4796 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4797 | int atomic) | |
5f39d397 | 4798 | { |
1259ab75 | 4799 | int ret; |
727011e0 | 4800 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4801 | |
0b32f4bb | 4802 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4803 | if (!ret) |
4804 | return ret; | |
4805 | ||
4806 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4807 | parent_transid, atomic); |
4808 | if (ret == -EAGAIN) | |
4809 | return ret; | |
1259ab75 | 4810 | return !ret; |
5f39d397 CM |
4811 | } |
4812 | ||
5f39d397 CM |
4813 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4814 | { | |
2f4d60df | 4815 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4816 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4817 | int was_dirty; |
b4ce94de | 4818 | |
06ea65a3 JB |
4819 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4820 | /* | |
4821 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4822 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4823 | * outside of the sanity tests. |
4824 | */ | |
b0132a3b | 4825 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4826 | return; |
4827 | #endif | |
49d0c642 | 4828 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4829 | if (transid != fs_info->generation) |
5d163e0e | 4830 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4831 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4832 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4833 | if (!was_dirty) |
104b4e51 NB |
4834 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4835 | buf->len, | |
4836 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4837 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4838 | /* |
4839 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4840 | * but item data not updated. | |
4841 | * So here we should only check item pointers, not item data. | |
4842 | */ | |
4843 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4844 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4845 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4846 | ASSERT(0); |
4847 | } | |
4848 | #endif | |
eb60ceac CM |
4849 | } |
4850 | ||
2ff7e61e | 4851 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4852 | int flush_delayed) |
16cdcec7 MX |
4853 | { |
4854 | /* | |
4855 | * looks as though older kernels can get into trouble with | |
4856 | * this code, they end up stuck in balance_dirty_pages forever | |
4857 | */ | |
e2d84521 | 4858 | int ret; |
16cdcec7 MX |
4859 | |
4860 | if (current->flags & PF_MEMALLOC) | |
4861 | return; | |
4862 | ||
b53d3f5d | 4863 | if (flush_delayed) |
2ff7e61e | 4864 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4865 | |
d814a491 EL |
4866 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4867 | BTRFS_DIRTY_METADATA_THRESH, | |
4868 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4869 | if (ret > 0) { |
0b246afa | 4870 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4871 | } |
16cdcec7 MX |
4872 | } |
4873 | ||
2ff7e61e | 4874 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4875 | { |
2ff7e61e | 4876 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4877 | } |
585ad2c3 | 4878 | |
2ff7e61e | 4879 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4880 | { |
2ff7e61e | 4881 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4882 | } |
6b80053d | 4883 | |
2ff7e61e | 4884 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4885 | { |
fe816d0f NB |
4886 | /* cleanup FS via transaction */ |
4887 | btrfs_cleanup_transaction(fs_info); | |
4888 | ||
0b246afa | 4889 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4890 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4891 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4892 | |
0b246afa JM |
4893 | down_write(&fs_info->cleanup_work_sem); |
4894 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4895 | } |
4896 | ||
ef67963d JB |
4897 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4898 | { | |
fc7cbcd4 DS |
4899 | struct btrfs_root *gang[8]; |
4900 | u64 root_objectid = 0; | |
4901 | int ret; | |
4902 | ||
4903 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4904 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4905 | (void **)gang, root_objectid, | |
4906 | ARRAY_SIZE(gang))) != 0) { | |
4907 | int i; | |
4908 | ||
4909 | for (i = 0; i < ret; i++) | |
4910 | gang[i] = btrfs_grab_root(gang[i]); | |
4911 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4912 | ||
4913 | for (i = 0; i < ret; i++) { | |
4914 | if (!gang[i]) | |
ef67963d | 4915 | continue; |
fc7cbcd4 DS |
4916 | root_objectid = gang[i]->root_key.objectid; |
4917 | btrfs_free_log(NULL, gang[i]); | |
4918 | btrfs_put_root(gang[i]); | |
ef67963d | 4919 | } |
fc7cbcd4 DS |
4920 | root_objectid++; |
4921 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4922 | } |
fc7cbcd4 | 4923 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4924 | btrfs_free_log_root_tree(NULL, fs_info); |
4925 | } | |
4926 | ||
143bede5 | 4927 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4928 | { |
acce952b | 4929 | struct btrfs_ordered_extent *ordered; |
acce952b | 4930 | |
199c2a9c | 4931 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4932 | /* |
4933 | * This will just short circuit the ordered completion stuff which will | |
4934 | * make sure the ordered extent gets properly cleaned up. | |
4935 | */ | |
199c2a9c | 4936 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4937 | root_extent_list) |
4938 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4939 | spin_unlock(&root->ordered_extent_lock); |
4940 | } | |
4941 | ||
4942 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4943 | { | |
4944 | struct btrfs_root *root; | |
4945 | struct list_head splice; | |
4946 | ||
4947 | INIT_LIST_HEAD(&splice); | |
4948 | ||
4949 | spin_lock(&fs_info->ordered_root_lock); | |
4950 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4951 | while (!list_empty(&splice)) { | |
4952 | root = list_first_entry(&splice, struct btrfs_root, | |
4953 | ordered_root); | |
1de2cfde JB |
4954 | list_move_tail(&root->ordered_root, |
4955 | &fs_info->ordered_roots); | |
199c2a9c | 4956 | |
2a85d9ca | 4957 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4958 | btrfs_destroy_ordered_extents(root); |
4959 | ||
2a85d9ca LB |
4960 | cond_resched(); |
4961 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4962 | } |
4963 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4964 | |
4965 | /* | |
4966 | * We need this here because if we've been flipped read-only we won't | |
4967 | * get sync() from the umount, so we need to make sure any ordered | |
4968 | * extents that haven't had their dirty pages IO start writeout yet | |
4969 | * actually get run and error out properly. | |
4970 | */ | |
4971 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4972 | } |
4973 | ||
35a3621b | 4974 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4975 | struct btrfs_fs_info *fs_info) |
acce952b | 4976 | { |
4977 | struct rb_node *node; | |
4978 | struct btrfs_delayed_ref_root *delayed_refs; | |
4979 | struct btrfs_delayed_ref_node *ref; | |
4980 | int ret = 0; | |
4981 | ||
4982 | delayed_refs = &trans->delayed_refs; | |
4983 | ||
4984 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4985 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4986 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4987 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4988 | return ret; |
4989 | } | |
4990 | ||
5c9d028b | 4991 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4992 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4993 | struct rb_node *n; |
e78417d1 | 4994 | bool pin_bytes = false; |
acce952b | 4995 | |
d7df2c79 JB |
4996 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4997 | href_node); | |
3069bd26 | 4998 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4999 | continue; |
3069bd26 | 5000 | |
d7df2c79 | 5001 | spin_lock(&head->lock); |
e3d03965 | 5002 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
5003 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
5004 | ref_node); | |
d7df2c79 | 5005 | ref->in_tree = 0; |
e3d03965 | 5006 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 5007 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
5008 | if (!list_empty(&ref->add_list)) |
5009 | list_del(&ref->add_list); | |
d7df2c79 JB |
5010 | atomic_dec(&delayed_refs->num_entries); |
5011 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 5012 | } |
d7df2c79 JB |
5013 | if (head->must_insert_reserved) |
5014 | pin_bytes = true; | |
5015 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 5016 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
5017 | spin_unlock(&head->lock); |
5018 | spin_unlock(&delayed_refs->lock); | |
5019 | mutex_unlock(&head->mutex); | |
acce952b | 5020 | |
f603bb94 NB |
5021 | if (pin_bytes) { |
5022 | struct btrfs_block_group *cache; | |
5023 | ||
5024 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
5025 | BUG_ON(!cache); | |
5026 | ||
5027 | spin_lock(&cache->space_info->lock); | |
5028 | spin_lock(&cache->lock); | |
5029 | cache->pinned += head->num_bytes; | |
5030 | btrfs_space_info_update_bytes_pinned(fs_info, | |
5031 | cache->space_info, head->num_bytes); | |
5032 | cache->reserved -= head->num_bytes; | |
5033 | cache->space_info->bytes_reserved -= head->num_bytes; | |
5034 | spin_unlock(&cache->lock); | |
5035 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
5036 | |
5037 | btrfs_put_block_group(cache); | |
5038 | ||
5039 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
5040 | head->bytenr + head->num_bytes - 1); | |
5041 | } | |
31890da0 | 5042 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 5043 | btrfs_put_delayed_ref_head(head); |
acce952b | 5044 | cond_resched(); |
5045 | spin_lock(&delayed_refs->lock); | |
5046 | } | |
81f7eb00 | 5047 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 5048 | |
5049 | spin_unlock(&delayed_refs->lock); | |
5050 | ||
5051 | return ret; | |
5052 | } | |
5053 | ||
143bede5 | 5054 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 5055 | { |
5056 | struct btrfs_inode *btrfs_inode; | |
5057 | struct list_head splice; | |
5058 | ||
5059 | INIT_LIST_HEAD(&splice); | |
5060 | ||
eb73c1b7 MX |
5061 | spin_lock(&root->delalloc_lock); |
5062 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 5063 | |
5064 | while (!list_empty(&splice)) { | |
fe816d0f | 5065 | struct inode *inode = NULL; |
eb73c1b7 MX |
5066 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
5067 | delalloc_inodes); | |
fe816d0f | 5068 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 5069 | spin_unlock(&root->delalloc_lock); |
acce952b | 5070 | |
fe816d0f NB |
5071 | /* |
5072 | * Make sure we get a live inode and that it'll not disappear | |
5073 | * meanwhile. | |
5074 | */ | |
5075 | inode = igrab(&btrfs_inode->vfs_inode); | |
5076 | if (inode) { | |
5077 | invalidate_inode_pages2(inode->i_mapping); | |
5078 | iput(inode); | |
5079 | } | |
eb73c1b7 | 5080 | spin_lock(&root->delalloc_lock); |
acce952b | 5081 | } |
eb73c1b7 MX |
5082 | spin_unlock(&root->delalloc_lock); |
5083 | } | |
5084 | ||
5085 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
5086 | { | |
5087 | struct btrfs_root *root; | |
5088 | struct list_head splice; | |
5089 | ||
5090 | INIT_LIST_HEAD(&splice); | |
5091 | ||
5092 | spin_lock(&fs_info->delalloc_root_lock); | |
5093 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
5094 | while (!list_empty(&splice)) { | |
5095 | root = list_first_entry(&splice, struct btrfs_root, | |
5096 | delalloc_root); | |
00246528 | 5097 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
5098 | BUG_ON(!root); |
5099 | spin_unlock(&fs_info->delalloc_root_lock); | |
5100 | ||
5101 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 5102 | btrfs_put_root(root); |
eb73c1b7 MX |
5103 | |
5104 | spin_lock(&fs_info->delalloc_root_lock); | |
5105 | } | |
5106 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 5107 | } |
5108 | ||
2ff7e61e | 5109 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 5110 | struct extent_io_tree *dirty_pages, |
5111 | int mark) | |
5112 | { | |
5113 | int ret; | |
acce952b | 5114 | struct extent_buffer *eb; |
5115 | u64 start = 0; | |
5116 | u64 end; | |
acce952b | 5117 | |
5118 | while (1) { | |
5119 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5120 | mark, NULL); |
acce952b | 5121 | if (ret) |
5122 | break; | |
5123 | ||
91166212 | 5124 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5125 | while (start <= end) { |
0b246afa JM |
5126 | eb = find_extent_buffer(fs_info, start); |
5127 | start += fs_info->nodesize; | |
fd8b2b61 | 5128 | if (!eb) |
acce952b | 5129 | continue; |
fd8b2b61 | 5130 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5131 | |
fd8b2b61 JB |
5132 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5133 | &eb->bflags)) | |
5134 | clear_extent_buffer_dirty(eb); | |
5135 | free_extent_buffer_stale(eb); | |
acce952b | 5136 | } |
5137 | } | |
5138 | ||
5139 | return ret; | |
5140 | } | |
5141 | ||
2ff7e61e | 5142 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5143 | struct extent_io_tree *unpin) |
acce952b | 5144 | { |
acce952b | 5145 | u64 start; |
5146 | u64 end; | |
5147 | int ret; | |
5148 | ||
acce952b | 5149 | while (1) { |
0e6ec385 FM |
5150 | struct extent_state *cached_state = NULL; |
5151 | ||
fcd5e742 LF |
5152 | /* |
5153 | * The btrfs_finish_extent_commit() may get the same range as | |
5154 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5155 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5156 | * the same extent range. | |
5157 | */ | |
5158 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5159 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5160 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5161 | if (ret) { |
5162 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5163 | break; |
fcd5e742 | 5164 | } |
acce952b | 5165 | |
0e6ec385 FM |
5166 | clear_extent_dirty(unpin, start, end, &cached_state); |
5167 | free_extent_state(cached_state); | |
2ff7e61e | 5168 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5169 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5170 | cond_resched(); |
5171 | } | |
5172 | ||
5173 | return 0; | |
5174 | } | |
5175 | ||
32da5386 | 5176 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5177 | { |
5178 | struct inode *inode; | |
5179 | ||
5180 | inode = cache->io_ctl.inode; | |
5181 | if (inode) { | |
5182 | invalidate_inode_pages2(inode->i_mapping); | |
5183 | BTRFS_I(inode)->generation = 0; | |
5184 | cache->io_ctl.inode = NULL; | |
5185 | iput(inode); | |
5186 | } | |
bbc37d6e | 5187 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5188 | btrfs_put_block_group(cache); |
5189 | } | |
5190 | ||
5191 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5192 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5193 | { |
32da5386 | 5194 | struct btrfs_block_group *cache; |
c79a1751 LB |
5195 | |
5196 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5197 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5198 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5199 | struct btrfs_block_group, |
c79a1751 | 5200 | dirty_list); |
c79a1751 LB |
5201 | |
5202 | if (!list_empty(&cache->io_list)) { | |
5203 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5204 | list_del_init(&cache->io_list); | |
5205 | btrfs_cleanup_bg_io(cache); | |
5206 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5207 | } | |
5208 | ||
5209 | list_del_init(&cache->dirty_list); | |
5210 | spin_lock(&cache->lock); | |
5211 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5212 | spin_unlock(&cache->lock); | |
5213 | ||
5214 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5215 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5216 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5217 | spin_lock(&cur_trans->dirty_bgs_lock); |
5218 | } | |
5219 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5220 | ||
45ae2c18 NB |
5221 | /* |
5222 | * Refer to the definition of io_bgs member for details why it's safe | |
5223 | * to use it without any locking | |
5224 | */ | |
c79a1751 LB |
5225 | while (!list_empty(&cur_trans->io_bgs)) { |
5226 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5227 | struct btrfs_block_group, |
c79a1751 | 5228 | io_list); |
c79a1751 LB |
5229 | |
5230 | list_del_init(&cache->io_list); | |
5231 | spin_lock(&cache->lock); | |
5232 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5233 | spin_unlock(&cache->lock); | |
5234 | btrfs_cleanup_bg_io(cache); | |
5235 | } | |
5236 | } | |
5237 | ||
49b25e05 | 5238 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5239 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5240 | { |
bbbf7243 NB |
5241 | struct btrfs_device *dev, *tmp; |
5242 | ||
2ff7e61e | 5243 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5244 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5245 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5246 | ||
bbbf7243 NB |
5247 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5248 | post_commit_list) { | |
5249 | list_del_init(&dev->post_commit_list); | |
5250 | } | |
5251 | ||
2ff7e61e | 5252 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5253 | |
4a9d8bde | 5254 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5255 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5256 | |
4a9d8bde | 5257 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5258 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5259 | |
ccdf9b30 | 5260 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5261 | |
2ff7e61e | 5262 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5263 | EXTENT_DIRTY); |
fe119a6e | 5264 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5265 | |
d3575156 NA |
5266 | btrfs_free_redirty_list(cur_trans); |
5267 | ||
4a9d8bde MX |
5268 | cur_trans->state =TRANS_STATE_COMPLETED; |
5269 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5270 | } |
5271 | ||
2ff7e61e | 5272 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5273 | { |
5274 | struct btrfs_transaction *t; | |
acce952b | 5275 | |
0b246afa | 5276 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5277 | |
0b246afa JM |
5278 | spin_lock(&fs_info->trans_lock); |
5279 | while (!list_empty(&fs_info->trans_list)) { | |
5280 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5281 | struct btrfs_transaction, list); |
5282 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5283 | refcount_inc(&t->use_count); |
0b246afa | 5284 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5285 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5286 | btrfs_put_transaction(t); |
0b246afa | 5287 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5288 | continue; |
5289 | } | |
0b246afa | 5290 | if (t == fs_info->running_transaction) { |
724e2315 | 5291 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5292 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5293 | /* |
5294 | * We wait for 0 num_writers since we don't hold a trans | |
5295 | * handle open currently for this transaction. | |
5296 | */ | |
5297 | wait_event(t->writer_wait, | |
5298 | atomic_read(&t->num_writers) == 0); | |
5299 | } else { | |
0b246afa | 5300 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5301 | } |
2ff7e61e | 5302 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5303 | |
0b246afa JM |
5304 | spin_lock(&fs_info->trans_lock); |
5305 | if (t == fs_info->running_transaction) | |
5306 | fs_info->running_transaction = NULL; | |
acce952b | 5307 | list_del_init(&t->list); |
0b246afa | 5308 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5309 | |
724e2315 | 5310 | btrfs_put_transaction(t); |
2e4e97ab | 5311 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5312 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5313 | } |
0b246afa JM |
5314 | spin_unlock(&fs_info->trans_lock); |
5315 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5316 | btrfs_destroy_delayed_inodes(fs_info); |
5317 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5318 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5319 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5320 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5321 | |
5322 | return 0; | |
5323 | } | |
ec7d6dfd | 5324 | |
453e4873 | 5325 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5326 | { |
5327 | struct btrfs_path *path; | |
5328 | int ret; | |
5329 | struct extent_buffer *l; | |
5330 | struct btrfs_key search_key; | |
5331 | struct btrfs_key found_key; | |
5332 | int slot; | |
5333 | ||
5334 | path = btrfs_alloc_path(); | |
5335 | if (!path) | |
5336 | return -ENOMEM; | |
5337 | ||
5338 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5339 | search_key.type = -1; | |
5340 | search_key.offset = (u64)-1; | |
5341 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5342 | if (ret < 0) | |
5343 | goto error; | |
5344 | BUG_ON(ret == 0); /* Corruption */ | |
5345 | if (path->slots[0] > 0) { | |
5346 | slot = path->slots[0] - 1; | |
5347 | l = path->nodes[0]; | |
5348 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5349 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5350 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5351 | } else { |
23125104 | 5352 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5353 | } |
5354 | ret = 0; | |
5355 | error: | |
5356 | btrfs_free_path(path); | |
5357 | return ret; | |
5358 | } | |
5359 | ||
543068a2 | 5360 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5361 | { |
5362 | int ret; | |
5363 | mutex_lock(&root->objectid_mutex); | |
5364 | ||
6b8fad57 | 5365 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5366 | btrfs_warn(root->fs_info, |
5367 | "the objectid of root %llu reaches its highest value", | |
5368 | root->root_key.objectid); | |
5369 | ret = -ENOSPC; | |
5370 | goto out; | |
5371 | } | |
5372 | ||
23125104 | 5373 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5374 | ret = 0; |
5375 | out: | |
5376 | mutex_unlock(&root->objectid_mutex); | |
5377 | return ret; | |
5378 | } |