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> |
35b7e476 | 8 | #include <linux/writeback.h> |
ce9adaa5 | 9 | #include <linux/workqueue.h> |
a74a4b97 | 10 | #include <linux/kthread.h> |
5a0e3ad6 | 11 | #include <linux/slab.h> |
784b4e29 | 12 | #include <linux/migrate.h> |
7a36ddec | 13 | #include <linux/ratelimit.h> |
6463fe58 | 14 | #include <linux/uuid.h> |
803b2f54 | 15 | #include <linux/semaphore.h> |
540adea3 | 16 | #include <linux/error-injection.h> |
9678c543 | 17 | #include <linux/crc32c.h> |
b89f6d1f | 18 | #include <linux/sched/mm.h> |
7e75bf3f | 19 | #include <asm/unaligned.h> |
6d97c6e3 | 20 | #include <crypto/hash.h> |
eb60ceac CM |
21 | #include "ctree.h" |
22 | #include "disk-io.h" | |
e089f05c | 23 | #include "transaction.h" |
0f7d52f4 | 24 | #include "btrfs_inode.h" |
0b86a832 | 25 | #include "volumes.h" |
db94535d | 26 | #include "print-tree.h" |
925baedd | 27 | #include "locking.h" |
e02119d5 | 28 | #include "tree-log.h" |
fa9c0d79 | 29 | #include "free-space-cache.h" |
70f6d82e | 30 | #include "free-space-tree.h" |
21adbd5c | 31 | #include "check-integrity.h" |
606686ee | 32 | #include "rcu-string.h" |
8dabb742 | 33 | #include "dev-replace.h" |
53b381b3 | 34 | #include "raid56.h" |
5ac1d209 | 35 | #include "sysfs.h" |
fcebe456 | 36 | #include "qgroup.h" |
ebb8765b | 37 | #include "compression.h" |
557ea5dd | 38 | #include "tree-checker.h" |
fd708b81 | 39 | #include "ref-verify.h" |
aac0023c | 40 | #include "block-group.h" |
b0643e59 | 41 | #include "discard.h" |
f603bb94 | 42 | #include "space-info.h" |
b70f5097 | 43 | #include "zoned.h" |
139e8cd3 | 44 | #include "subpage.h" |
eb60ceac | 45 | |
319e4d06 QW |
46 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
47 | BTRFS_HEADER_FLAG_RELOC |\ | |
48 | BTRFS_SUPER_FLAG_ERROR |\ | |
49 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
50 | BTRFS_SUPER_FLAG_METADUMP |\ |
51 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 52 | |
8b712842 | 53 | static void end_workqueue_fn(struct btrfs_work *work); |
143bede5 | 54 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 56 | struct btrfs_fs_info *fs_info); |
143bede5 | 57 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 58 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 59 | struct extent_io_tree *dirty_pages, |
60 | int mark); | |
2ff7e61e | 61 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 62 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
63 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
64 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 65 | |
d352ac68 | 66 | /* |
97eb6b69 DS |
67 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
68 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
69 | * by writes to insert metadata for new file extents after IO is complete. |
70 | */ | |
97eb6b69 | 71 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
72 | struct bio *bio; |
73 | bio_end_io_t *end_io; | |
74 | void *private; | |
75 | struct btrfs_fs_info *info; | |
4e4cbee9 | 76 | blk_status_t status; |
bfebd8b5 | 77 | enum btrfs_wq_endio_type metadata; |
8b712842 | 78 | struct btrfs_work work; |
ce9adaa5 | 79 | }; |
0da5468f | 80 | |
97eb6b69 DS |
81 | static struct kmem_cache *btrfs_end_io_wq_cache; |
82 | ||
83 | int __init btrfs_end_io_wq_init(void) | |
84 | { | |
85 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
86 | sizeof(struct btrfs_end_io_wq), | |
87 | 0, | |
fba4b697 | 88 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
89 | NULL); |
90 | if (!btrfs_end_io_wq_cache) | |
91 | return -ENOMEM; | |
92 | return 0; | |
93 | } | |
94 | ||
e67c718b | 95 | void __cold btrfs_end_io_wq_exit(void) |
97eb6b69 | 96 | { |
5598e900 | 97 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
98 | } |
99 | ||
141386e1 JB |
100 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
101 | { | |
102 | if (fs_info->csum_shash) | |
103 | crypto_free_shash(fs_info->csum_shash); | |
104 | } | |
105 | ||
d352ac68 CM |
106 | /* |
107 | * async submit bios are used to offload expensive checksumming | |
108 | * onto the worker threads. They checksum file and metadata bios | |
109 | * just before they are sent down the IO stack. | |
110 | */ | |
44b8bd7e | 111 | struct async_submit_bio { |
8896a08d | 112 | struct inode *inode; |
44b8bd7e | 113 | struct bio *bio; |
a758781d | 114 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 115 | int mirror_num; |
1941b64b QW |
116 | |
117 | /* Optional parameter for submit_bio_start used by direct io */ | |
118 | u64 dio_file_offset; | |
8b712842 | 119 | struct btrfs_work work; |
4e4cbee9 | 120 | blk_status_t status; |
44b8bd7e CM |
121 | }; |
122 | ||
85d4e461 CM |
123 | /* |
124 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
125 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
126 | * the level the eb occupies in the tree. | |
127 | * | |
128 | * Different roots are used for different purposes and may nest inside each | |
129 | * other and they require separate keysets. As lockdep keys should be | |
130 | * static, assign keysets according to the purpose of the root as indicated | |
4fd786e6 MT |
131 | * by btrfs_root->root_key.objectid. This ensures that all special purpose |
132 | * roots have separate keysets. | |
4008c04a | 133 | * |
85d4e461 CM |
134 | * Lock-nesting across peer nodes is always done with the immediate parent |
135 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
136 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 137 | * |
85d4e461 CM |
138 | * The key is set by the readpage_end_io_hook after the buffer has passed |
139 | * csum validation but before the pages are unlocked. It is also set by | |
140 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 141 | * |
85d4e461 CM |
142 | * We also add a check to make sure the highest level of the tree is the |
143 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
144 | * needs update as well. | |
4008c04a CM |
145 | */ |
146 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
147 | # if BTRFS_MAX_LEVEL != 8 | |
148 | # error | |
149 | # endif | |
85d4e461 | 150 | |
ab1405aa DS |
151 | #define DEFINE_LEVEL(stem, level) \ |
152 | .names[level] = "btrfs-" stem "-0" #level, | |
153 | ||
154 | #define DEFINE_NAME(stem) \ | |
155 | DEFINE_LEVEL(stem, 0) \ | |
156 | DEFINE_LEVEL(stem, 1) \ | |
157 | DEFINE_LEVEL(stem, 2) \ | |
158 | DEFINE_LEVEL(stem, 3) \ | |
159 | DEFINE_LEVEL(stem, 4) \ | |
160 | DEFINE_LEVEL(stem, 5) \ | |
161 | DEFINE_LEVEL(stem, 6) \ | |
162 | DEFINE_LEVEL(stem, 7) | |
163 | ||
85d4e461 CM |
164 | static struct btrfs_lockdep_keyset { |
165 | u64 id; /* root objectid */ | |
ab1405aa | 166 | /* Longest entry: btrfs-free-space-00 */ |
387824af DS |
167 | char names[BTRFS_MAX_LEVEL][20]; |
168 | struct lock_class_key keys[BTRFS_MAX_LEVEL]; | |
85d4e461 | 169 | } btrfs_lockdep_keysets[] = { |
ab1405aa DS |
170 | { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, |
171 | { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, | |
172 | { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, | |
173 | { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, | |
ab1405aa DS |
174 | { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, |
175 | { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, | |
176 | { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, | |
177 | { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, | |
178 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, | |
179 | { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, | |
180 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, | |
181 | { .id = 0, DEFINE_NAME("tree") }, | |
4008c04a | 182 | }; |
85d4e461 | 183 | |
ab1405aa DS |
184 | #undef DEFINE_LEVEL |
185 | #undef DEFINE_NAME | |
85d4e461 CM |
186 | |
187 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
188 | int level) | |
189 | { | |
190 | struct btrfs_lockdep_keyset *ks; | |
191 | ||
192 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
193 | ||
194 | /* find the matching keyset, id 0 is the default entry */ | |
195 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
196 | if (ks->id == objectid) | |
197 | break; | |
198 | ||
199 | lockdep_set_class_and_name(&eb->lock, | |
200 | &ks->keys[level], ks->names[level]); | |
201 | } | |
202 | ||
4008c04a CM |
203 | #endif |
204 | ||
d352ac68 | 205 | /* |
2996e1f8 | 206 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 207 | */ |
c67b3892 | 208 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 209 | { |
d5178578 | 210 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 211 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 212 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 213 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 214 | char *kaddr; |
e9be5a30 | 215 | int i; |
d5178578 JT |
216 | |
217 | shash->tfm = fs_info->csum_shash; | |
218 | crypto_shash_init(shash); | |
a26663e7 | 219 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 220 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 221 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 222 | |
e9be5a30 DS |
223 | for (i = 1; i < num_pages; i++) { |
224 | kaddr = page_address(buf->pages[i]); | |
225 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 226 | } |
71a63551 | 227 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 228 | crypto_shash_final(shash, result); |
19c00ddc CM |
229 | } |
230 | ||
d352ac68 CM |
231 | /* |
232 | * we can't consider a given block up to date unless the transid of the | |
233 | * block matches the transid in the parent node's pointer. This is how we | |
234 | * detect blocks that either didn't get written at all or got written | |
235 | * in the wrong place. | |
236 | */ | |
1259ab75 | 237 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
238 | struct extent_buffer *eb, u64 parent_transid, |
239 | int atomic) | |
1259ab75 | 240 | { |
2ac55d41 | 241 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
242 | int ret; |
243 | ||
244 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
245 | return 0; | |
246 | ||
b9fab919 CM |
247 | if (atomic) |
248 | return -EAGAIN; | |
249 | ||
2ac55d41 | 250 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 251 | &cached_state); |
0b32f4bb | 252 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
253 | btrfs_header_generation(eb) == parent_transid) { |
254 | ret = 0; | |
255 | goto out; | |
256 | } | |
94647322 DS |
257 | btrfs_err_rl(eb->fs_info, |
258 | "parent transid verify failed on %llu wanted %llu found %llu", | |
259 | eb->start, | |
29549aec | 260 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 261 | ret = 1; |
35b22c19 | 262 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 263 | out: |
2ac55d41 | 264 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 265 | &cached_state); |
1259ab75 | 266 | return ret; |
1259ab75 CM |
267 | } |
268 | ||
e7e16f48 JT |
269 | static bool btrfs_supported_super_csum(u16 csum_type) |
270 | { | |
271 | switch (csum_type) { | |
272 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 273 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 274 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 275 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
276 | return true; |
277 | default: | |
278 | return false; | |
279 | } | |
280 | } | |
281 | ||
1104a885 DS |
282 | /* |
283 | * Return 0 if the superblock checksum type matches the checksum value of that | |
284 | * algorithm. Pass the raw disk superblock data. | |
285 | */ | |
ab8d0fc4 JM |
286 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
287 | char *raw_disk_sb) | |
1104a885 DS |
288 | { |
289 | struct btrfs_super_block *disk_sb = | |
290 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 291 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
292 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
293 | ||
294 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 295 | |
51bce6c9 JT |
296 | /* |
297 | * The super_block structure does not span the whole | |
298 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
299 | * filled with zeros and is included in the checksum. | |
300 | */ | |
fd08001f EB |
301 | crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
302 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); | |
1104a885 | 303 | |
55fc29be | 304 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 305 | return 1; |
1104a885 | 306 | |
e7e16f48 | 307 | return 0; |
1104a885 DS |
308 | } |
309 | ||
e064d5e9 | 310 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 311 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 312 | { |
e064d5e9 | 313 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
314 | int found_level; |
315 | struct btrfs_key found_key; | |
316 | int ret; | |
317 | ||
318 | found_level = btrfs_header_level(eb); | |
319 | if (found_level != level) { | |
63489055 QW |
320 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
321 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
322 | btrfs_err(fs_info, |
323 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
324 | eb->start, level, found_level); | |
581c1760 QW |
325 | return -EIO; |
326 | } | |
327 | ||
328 | if (!first_key) | |
329 | return 0; | |
330 | ||
5d41be6f QW |
331 | /* |
332 | * For live tree block (new tree blocks in current transaction), | |
333 | * we need proper lock context to avoid race, which is impossible here. | |
334 | * So we only checks tree blocks which is read from disk, whose | |
335 | * generation <= fs_info->last_trans_committed. | |
336 | */ | |
337 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
338 | return 0; | |
62fdaa52 QW |
339 | |
340 | /* We have @first_key, so this @eb must have at least one item */ | |
341 | if (btrfs_header_nritems(eb) == 0) { | |
342 | btrfs_err(fs_info, | |
343 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
344 | eb->start); | |
345 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
346 | return -EUCLEAN; | |
347 | } | |
348 | ||
581c1760 QW |
349 | if (found_level) |
350 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
351 | else | |
352 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
353 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
354 | ||
581c1760 | 355 | if (ret) { |
63489055 QW |
356 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
357 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 358 | btrfs_err(fs_info, |
ff76a864 LB |
359 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
360 | eb->start, parent_transid, first_key->objectid, | |
361 | first_key->type, first_key->offset, | |
362 | found_key.objectid, found_key.type, | |
363 | found_key.offset); | |
581c1760 | 364 | } |
581c1760 QW |
365 | return ret; |
366 | } | |
367 | ||
d352ac68 CM |
368 | /* |
369 | * helper to read a given tree block, doing retries as required when | |
370 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
371 | * |
372 | * @parent_transid: expected transid, skip check if 0 | |
373 | * @level: expected level, mandatory check | |
374 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 375 | */ |
6a2e9dc4 FM |
376 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
377 | u64 parent_transid, int level, | |
378 | struct btrfs_key *first_key) | |
f188591e | 379 | { |
5ab12d1f | 380 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 381 | struct extent_io_tree *io_tree; |
ea466794 | 382 | int failed = 0; |
f188591e CM |
383 | int ret; |
384 | int num_copies = 0; | |
385 | int mirror_num = 0; | |
ea466794 | 386 | int failed_mirror = 0; |
f188591e | 387 | |
0b246afa | 388 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 389 | while (1) { |
f8397d69 | 390 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 391 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 392 | if (!ret) { |
581c1760 | 393 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 394 | parent_transid, 0)) |
256dd1bb | 395 | ret = -EIO; |
e064d5e9 | 396 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 397 | first_key, parent_transid)) |
581c1760 QW |
398 | ret = -EUCLEAN; |
399 | else | |
400 | break; | |
256dd1bb | 401 | } |
d397712b | 402 | |
0b246afa | 403 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 404 | eb->start, eb->len); |
4235298e | 405 | if (num_copies == 1) |
ea466794 | 406 | break; |
4235298e | 407 | |
5cf1ab56 JB |
408 | if (!failed_mirror) { |
409 | failed = 1; | |
410 | failed_mirror = eb->read_mirror; | |
411 | } | |
412 | ||
f188591e | 413 | mirror_num++; |
ea466794 JB |
414 | if (mirror_num == failed_mirror) |
415 | mirror_num++; | |
416 | ||
4235298e | 417 | if (mirror_num > num_copies) |
ea466794 | 418 | break; |
f188591e | 419 | } |
ea466794 | 420 | |
c0901581 | 421 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 422 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
423 | |
424 | return ret; | |
f188591e | 425 | } |
19c00ddc | 426 | |
eca0f6f6 QW |
427 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
428 | { | |
429 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
430 | u8 result[BTRFS_CSUM_SIZE]; | |
431 | int ret; | |
432 | ||
433 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
434 | offsetof(struct btrfs_header, fsid), | |
435 | BTRFS_FSID_SIZE) == 0); | |
436 | csum_tree_block(eb, result); | |
437 | ||
438 | if (btrfs_header_level(eb)) | |
439 | ret = btrfs_check_node(eb); | |
440 | else | |
441 | ret = btrfs_check_leaf_full(eb); | |
442 | ||
3777369f QW |
443 | if (ret < 0) |
444 | goto error; | |
445 | ||
446 | /* | |
447 | * Also check the generation, the eb reached here must be newer than | |
448 | * last committed. Or something seriously wrong happened. | |
449 | */ | |
450 | if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { | |
451 | ret = -EUCLEAN; | |
eca0f6f6 | 452 | btrfs_err(fs_info, |
3777369f QW |
453 | "block=%llu bad generation, have %llu expect > %llu", |
454 | eb->start, btrfs_header_generation(eb), | |
455 | fs_info->last_trans_committed); | |
456 | goto error; | |
eca0f6f6 QW |
457 | } |
458 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
459 | ||
460 | return 0; | |
3777369f QW |
461 | |
462 | error: | |
463 | btrfs_print_tree(eb, 0); | |
464 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
465 | eb->start); | |
466 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
467 | return ret; | |
eca0f6f6 QW |
468 | } |
469 | ||
470 | /* Checksum all dirty extent buffers in one bio_vec */ | |
471 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
472 | struct bio_vec *bvec) | |
473 | { | |
474 | struct page *page = bvec->bv_page; | |
475 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
476 | u64 cur; | |
477 | int ret = 0; | |
478 | ||
479 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
480 | cur += fs_info->nodesize) { | |
481 | struct extent_buffer *eb; | |
482 | bool uptodate; | |
483 | ||
484 | eb = find_extent_buffer(fs_info, cur); | |
485 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
486 | fs_info->nodesize); | |
487 | ||
8ee92268 | 488 | /* A dirty eb shouldn't disappear from extent_buffers */ |
eca0f6f6 QW |
489 | if (WARN_ON(!eb)) |
490 | return -EUCLEAN; | |
491 | ||
492 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
493 | free_extent_buffer(eb); | |
494 | return -EUCLEAN; | |
495 | } | |
496 | if (WARN_ON(!uptodate)) { | |
497 | free_extent_buffer(eb); | |
498 | return -EUCLEAN; | |
499 | } | |
500 | ||
501 | ret = csum_one_extent_buffer(eb); | |
502 | free_extent_buffer(eb); | |
503 | if (ret < 0) | |
504 | return ret; | |
505 | } | |
506 | return ret; | |
507 | } | |
508 | ||
d352ac68 | 509 | /* |
ac303b69 QW |
510 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
511 | * we only fill in the checksum field in the first page of a multi-page block. | |
512 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 513 | */ |
ac303b69 | 514 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 515 | { |
ac303b69 | 516 | struct page *page = bvec->bv_page; |
4eee4fa4 | 517 | u64 start = page_offset(page); |
19c00ddc | 518 | u64 found_start; |
19c00ddc | 519 | struct extent_buffer *eb; |
eca0f6f6 | 520 | |
fbca46eb | 521 | if (fs_info->nodesize < PAGE_SIZE) |
eca0f6f6 | 522 | return csum_dirty_subpage_buffers(fs_info, bvec); |
f188591e | 523 | |
4f2de97a JB |
524 | eb = (struct extent_buffer *)page->private; |
525 | if (page != eb->pages[0]) | |
526 | return 0; | |
0f805531 | 527 | |
19c00ddc | 528 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
529 | |
530 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
531 | WARN_ON(found_start != 0); | |
532 | return 0; | |
533 | } | |
534 | ||
0f805531 AL |
535 | /* |
536 | * Please do not consolidate these warnings into a single if. | |
537 | * It is useful to know what went wrong. | |
538 | */ | |
539 | if (WARN_ON(found_start != start)) | |
540 | return -EUCLEAN; | |
541 | if (WARN_ON(!PageUptodate(page))) | |
542 | return -EUCLEAN; | |
543 | ||
eca0f6f6 | 544 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
545 | } |
546 | ||
b0c9b3b0 | 547 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 548 | { |
b0c9b3b0 | 549 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 550 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 551 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 552 | u8 *metadata_uuid; |
2b82032c | 553 | |
9a8658e3 DS |
554 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
555 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
556 | /* |
557 | * Checking the incompat flag is only valid for the current fs. For | |
558 | * seed devices it's forbidden to have their uuid changed so reading | |
559 | * ->fsid in this case is fine | |
560 | */ | |
561 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
562 | metadata_uuid = fs_devices->metadata_uuid; | |
563 | else | |
564 | metadata_uuid = fs_devices->fsid; | |
565 | ||
566 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
567 | return 0; | |
568 | ||
569 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
570 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
571 | return 0; | |
572 | ||
573 | return 1; | |
2b82032c YZ |
574 | } |
575 | ||
77bf40a2 QW |
576 | /* Do basic extent buffer checks at read time */ |
577 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 578 | { |
77bf40a2 | 579 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 580 | u64 found_start; |
77bf40a2 QW |
581 | const u32 csum_size = fs_info->csum_size; |
582 | u8 found_level; | |
2996e1f8 | 583 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 584 | const u8 *header_csum; |
77bf40a2 | 585 | int ret = 0; |
ea466794 | 586 | |
ce9adaa5 | 587 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 588 | if (found_start != eb->start) { |
893bf4b1 SY |
589 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
590 | eb->start, found_start); | |
f188591e | 591 | ret = -EIO; |
77bf40a2 | 592 | goto out; |
ce9adaa5 | 593 | } |
b0c9b3b0 | 594 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
595 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
596 | eb->start); | |
1259ab75 | 597 | ret = -EIO; |
77bf40a2 | 598 | goto out; |
1259ab75 | 599 | } |
ce9adaa5 | 600 | found_level = btrfs_header_level(eb); |
1c24c3ce | 601 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
602 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
603 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce | 604 | ret = -EIO; |
77bf40a2 | 605 | goto out; |
1c24c3ce | 606 | } |
ce9adaa5 | 607 | |
c67b3892 | 608 | csum_tree_block(eb, result); |
dfd29eed DS |
609 | header_csum = page_address(eb->pages[0]) + |
610 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 611 | |
dfd29eed | 612 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 613 | btrfs_warn_rl(fs_info, |
ff14aa79 DS |
614 | "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
615 | eb->start, | |
dfd29eed | 616 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
617 | CSUM_FMT_VALUE(csum_size, result), |
618 | btrfs_header_level(eb)); | |
2996e1f8 | 619 | ret = -EUCLEAN; |
77bf40a2 | 620 | goto out; |
2996e1f8 JT |
621 | } |
622 | ||
a826d6dc JB |
623 | /* |
624 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
625 | * that we don't try and read the other copies of this block, just | |
626 | * return -EIO. | |
627 | */ | |
1c4360ee | 628 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
629 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
630 | ret = -EIO; | |
631 | } | |
ce9adaa5 | 632 | |
813fd1dc | 633 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
634 | ret = -EIO; |
635 | ||
0b32f4bb JB |
636 | if (!ret) |
637 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
638 | else |
639 | btrfs_err(fs_info, | |
640 | "block=%llu read time tree block corruption detected", | |
641 | eb->start); | |
77bf40a2 QW |
642 | out: |
643 | return ret; | |
644 | } | |
645 | ||
371cdc07 QW |
646 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
647 | int mirror) | |
648 | { | |
649 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
650 | struct extent_buffer *eb; | |
651 | bool reads_done; | |
652 | int ret = 0; | |
653 | ||
654 | /* | |
655 | * We don't allow bio merge for subpage metadata read, so we should | |
656 | * only get one eb for each endio hook. | |
657 | */ | |
658 | ASSERT(end == start + fs_info->nodesize - 1); | |
659 | ASSERT(PagePrivate(page)); | |
660 | ||
661 | eb = find_extent_buffer(fs_info, start); | |
662 | /* | |
663 | * When we are reading one tree block, eb must have been inserted into | |
664 | * the radix tree. If not, something is wrong. | |
665 | */ | |
666 | ASSERT(eb); | |
667 | ||
668 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
669 | /* Subpage read must finish in page read */ | |
670 | ASSERT(reads_done); | |
671 | ||
672 | eb->read_mirror = mirror; | |
673 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
674 | ret = -EIO; | |
675 | goto err; | |
676 | } | |
677 | ret = validate_extent_buffer(eb); | |
678 | if (ret < 0) | |
679 | goto err; | |
680 | ||
371cdc07 QW |
681 | set_extent_buffer_uptodate(eb); |
682 | ||
683 | free_extent_buffer(eb); | |
684 | return ret; | |
685 | err: | |
686 | /* | |
687 | * end_bio_extent_readpage decrements io_pages in case of error, | |
688 | * make sure it has something to decrement. | |
689 | */ | |
690 | atomic_inc(&eb->io_pages); | |
691 | clear_extent_buffer_uptodate(eb); | |
692 | free_extent_buffer(eb); | |
693 | return ret; | |
694 | } | |
695 | ||
c3a3b19b | 696 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
697 | struct page *page, u64 start, u64 end, |
698 | int mirror) | |
699 | { | |
700 | struct extent_buffer *eb; | |
701 | int ret = 0; | |
702 | int reads_done; | |
703 | ||
704 | ASSERT(page->private); | |
371cdc07 | 705 | |
fbca46eb | 706 | if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) |
371cdc07 QW |
707 | return validate_subpage_buffer(page, start, end, mirror); |
708 | ||
77bf40a2 QW |
709 | eb = (struct extent_buffer *)page->private; |
710 | ||
711 | /* | |
712 | * The pending IO might have been the only thing that kept this buffer | |
713 | * in memory. Make sure we have a ref for all this other checks | |
714 | */ | |
715 | atomic_inc(&eb->refs); | |
716 | ||
717 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
718 | if (!reads_done) | |
719 | goto err; | |
720 | ||
721 | eb->read_mirror = mirror; | |
722 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
723 | ret = -EIO; | |
724 | goto err; | |
725 | } | |
726 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 727 | err: |
53b381b3 DW |
728 | if (ret) { |
729 | /* | |
730 | * our io error hook is going to dec the io pages | |
731 | * again, we have to make sure it has something | |
732 | * to decrement | |
733 | */ | |
734 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 735 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 736 | } |
0b32f4bb | 737 | free_extent_buffer(eb); |
77bf40a2 | 738 | |
f188591e | 739 | return ret; |
ce9adaa5 CM |
740 | } |
741 | ||
4246a0b6 | 742 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 743 | { |
97eb6b69 | 744 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 745 | struct btrfs_fs_info *fs_info; |
9e0af237 | 746 | struct btrfs_workqueue *wq; |
ce9adaa5 | 747 | |
ce9adaa5 | 748 | fs_info = end_io_wq->info; |
4e4cbee9 | 749 | end_io_wq->status = bio->bi_status; |
d20f7043 | 750 | |
cfe94440 | 751 | if (btrfs_op(bio) == BTRFS_MAP_WRITE) { |
a0cac0ec | 752 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 753 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 754 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 755 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 756 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 757 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 758 | else |
9e0af237 | 759 | wq = fs_info->endio_write_workers; |
d20f7043 | 760 | } else { |
5c047a69 | 761 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 762 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 763 | else if (end_io_wq->metadata) |
9e0af237 | 764 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 765 | else |
9e0af237 | 766 | wq = fs_info->endio_workers; |
d20f7043 | 767 | } |
9e0af237 | 768 | |
a0cac0ec | 769 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 770 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
771 | } |
772 | ||
4e4cbee9 | 773 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 774 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 775 | { |
97eb6b69 | 776 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 777 | |
97eb6b69 | 778 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 779 | if (!end_io_wq) |
4e4cbee9 | 780 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
781 | |
782 | end_io_wq->private = bio->bi_private; | |
783 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 784 | end_io_wq->info = info; |
4e4cbee9 | 785 | end_io_wq->status = 0; |
ce9adaa5 | 786 | end_io_wq->bio = bio; |
22c59948 | 787 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
788 | |
789 | bio->bi_private = end_io_wq; | |
790 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
791 | return 0; |
792 | } | |
793 | ||
4a69a410 CM |
794 | static void run_one_async_start(struct btrfs_work *work) |
795 | { | |
4a69a410 | 796 | struct async_submit_bio *async; |
4e4cbee9 | 797 | blk_status_t ret; |
4a69a410 CM |
798 | |
799 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
800 | ret = async->submit_bio_start(async->inode, async->bio, |
801 | async->dio_file_offset); | |
79787eaa | 802 | if (ret) |
4e4cbee9 | 803 | async->status = ret; |
4a69a410 CM |
804 | } |
805 | ||
06ea01b1 DS |
806 | /* |
807 | * In order to insert checksums into the metadata in large chunks, we wait | |
808 | * until bio submission time. All the pages in the bio are checksummed and | |
809 | * sums are attached onto the ordered extent record. | |
810 | * | |
811 | * At IO completion time the csums attached on the ordered extent record are | |
812 | * inserted into the tree. | |
813 | */ | |
4a69a410 | 814 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 815 | { |
8b712842 | 816 | struct async_submit_bio *async; |
06ea01b1 DS |
817 | struct inode *inode; |
818 | blk_status_t ret; | |
8b712842 CM |
819 | |
820 | async = container_of(work, struct async_submit_bio, work); | |
8896a08d | 821 | inode = async->inode; |
4854ddd0 | 822 | |
bb7ab3b9 | 823 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
824 | if (async->status) { |
825 | async->bio->bi_status = async->status; | |
4246a0b6 | 826 | bio_endio(async->bio); |
79787eaa JM |
827 | return; |
828 | } | |
829 | ||
ec39f769 CM |
830 | /* |
831 | * All of the bios that pass through here are from async helpers. | |
832 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
833 | * This changes nothing when cgroups aren't in use. | |
834 | */ | |
835 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 836 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
837 | if (ret) { |
838 | async->bio->bi_status = ret; | |
839 | bio_endio(async->bio); | |
840 | } | |
4a69a410 CM |
841 | } |
842 | ||
843 | static void run_one_async_free(struct btrfs_work *work) | |
844 | { | |
845 | struct async_submit_bio *async; | |
846 | ||
847 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
848 | kfree(async); |
849 | } | |
850 | ||
8896a08d | 851 | blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, |
a6f5e39e | 852 | int mirror_num, u64 dio_file_offset, |
e288c080 | 853 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e | 854 | { |
8896a08d | 855 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
856 | struct async_submit_bio *async; |
857 | ||
858 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
859 | if (!async) | |
4e4cbee9 | 860 | return BLK_STS_RESOURCE; |
44b8bd7e | 861 | |
8896a08d | 862 | async->inode = inode; |
44b8bd7e CM |
863 | async->bio = bio; |
864 | async->mirror_num = mirror_num; | |
4a69a410 | 865 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 866 | |
a0cac0ec OS |
867 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
868 | run_one_async_free); | |
4a69a410 | 869 | |
1941b64b | 870 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 871 | |
4e4cbee9 | 872 | async->status = 0; |
79787eaa | 873 | |
67f055c7 | 874 | if (op_is_sync(bio->bi_opf)) |
a31b4a43 CH |
875 | btrfs_queue_work(fs_info->hipri_workers, &async->work); |
876 | else | |
877 | btrfs_queue_work(fs_info->workers, &async->work); | |
44b8bd7e CM |
878 | return 0; |
879 | } | |
880 | ||
4e4cbee9 | 881 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 882 | { |
2c30c71b | 883 | struct bio_vec *bvec; |
ce3ed71a | 884 | struct btrfs_root *root; |
2b070cfe | 885 | int ret = 0; |
6dc4f100 | 886 | struct bvec_iter_all iter_all; |
ce3ed71a | 887 | |
c09abff8 | 888 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 889 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 890 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 891 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
892 | if (ret) |
893 | break; | |
ce3ed71a | 894 | } |
2c30c71b | 895 | |
4e4cbee9 | 896 | return errno_to_blk_status(ret); |
ce3ed71a CM |
897 | } |
898 | ||
8896a08d | 899 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 900 | u64 dio_file_offset) |
22c59948 | 901 | { |
8b712842 CM |
902 | /* |
903 | * when we're called for a write, we're already in the async | |
5443be45 | 904 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 905 | */ |
79787eaa | 906 | return btree_csum_one_bio(bio); |
4a69a410 | 907 | } |
22c59948 | 908 | |
f4dcfb30 | 909 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 910 | struct btrfs_inode *bi) |
de0022b9 | 911 | { |
4eef29ef | 912 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 913 | return false; |
6300463b | 914 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 915 | return false; |
9b4e675a | 916 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
917 | return false; |
918 | return true; | |
de0022b9 JB |
919 | } |
920 | ||
94d9e11b | 921 | void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) |
44b8bd7e | 922 | { |
0b246afa | 923 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4e4cbee9 | 924 | blk_status_t ret; |
cad321ad | 925 | |
cfe94440 | 926 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
4a69a410 CM |
927 | /* |
928 | * called for a read, do the setup so that checksum validation | |
929 | * can happen in the async kernel threads | |
930 | */ | |
0b246afa JM |
931 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
932 | BTRFS_WQ_ENDIO_METADATA); | |
94d9e11b CH |
933 | if (!ret) |
934 | ret = btrfs_map_bio(fs_info, bio, mirror_num); | |
f4dcfb30 | 935 | } else if (!should_async_write(fs_info, BTRFS_I(inode))) { |
de0022b9 | 936 | ret = btree_csum_one_bio(bio); |
94d9e11b CH |
937 | if (!ret) |
938 | ret = btrfs_map_bio(fs_info, bio, mirror_num); | |
61891923 SB |
939 | } else { |
940 | /* | |
941 | * kthread helpers are used to submit writes so that | |
942 | * checksumming can happen in parallel across all CPUs | |
943 | */ | |
8896a08d | 944 | ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, |
a6f5e39e | 945 | btree_submit_bio_start); |
44b8bd7e | 946 | } |
d313d7a3 | 947 | |
94d9e11b CH |
948 | if (ret) { |
949 | bio->bi_status = ret; | |
950 | bio_endio(bio); | |
951 | } | |
44b8bd7e CM |
952 | } |
953 | ||
3dd1462e | 954 | #ifdef CONFIG_MIGRATION |
784b4e29 | 955 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
956 | struct page *newpage, struct page *page, |
957 | enum migrate_mode mode) | |
784b4e29 CM |
958 | { |
959 | /* | |
960 | * we can't safely write a btree page from here, | |
961 | * we haven't done the locking hook | |
962 | */ | |
963 | if (PageDirty(page)) | |
964 | return -EAGAIN; | |
965 | /* | |
966 | * Buffers may be managed in a filesystem specific way. | |
967 | * We must have no buffers or drop them. | |
968 | */ | |
969 | if (page_has_private(page) && | |
970 | !try_to_release_page(page, GFP_KERNEL)) | |
971 | return -EAGAIN; | |
a6bc32b8 | 972 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 973 | } |
3dd1462e | 974 | #endif |
784b4e29 | 975 | |
0da5468f CM |
976 | |
977 | static int btree_writepages(struct address_space *mapping, | |
978 | struct writeback_control *wbc) | |
979 | { | |
e2d84521 MX |
980 | struct btrfs_fs_info *fs_info; |
981 | int ret; | |
982 | ||
d8d5f3e1 | 983 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
984 | |
985 | if (wbc->for_kupdate) | |
986 | return 0; | |
987 | ||
e2d84521 | 988 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 989 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
990 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
991 | BTRFS_DIRTY_METADATA_THRESH, | |
992 | fs_info->dirty_metadata_batch); | |
e2d84521 | 993 | if (ret < 0) |
793955bc | 994 | return 0; |
793955bc | 995 | } |
0b32f4bb | 996 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
997 | } |
998 | ||
f913cff3 | 999 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 1000 | { |
f913cff3 MWO |
1001 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
1002 | return false; | |
0c4e538b | 1003 | |
f913cff3 | 1004 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
1005 | } |
1006 | ||
895586eb MWO |
1007 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
1008 | size_t length) | |
d98237b3 | 1009 | { |
d1310b2e | 1010 | struct extent_io_tree *tree; |
895586eb MWO |
1011 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
1012 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 1013 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
1014 | if (folio_get_private(folio)) { |
1015 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
1016 | "folio private not zero on folio %llu", | |
1017 | (unsigned long long)folio_pos(folio)); | |
1018 | folio_detach_private(folio); | |
9ad6b7bc | 1019 | } |
d98237b3 CM |
1020 | } |
1021 | ||
bb146eb2 | 1022 | #ifdef DEBUG |
0079c3b1 MWO |
1023 | static bool btree_dirty_folio(struct address_space *mapping, |
1024 | struct folio *folio) | |
1025 | { | |
1026 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
139e8cd3 | 1027 | struct btrfs_subpage *subpage; |
0b32f4bb | 1028 | struct extent_buffer *eb; |
139e8cd3 | 1029 | int cur_bit = 0; |
0079c3b1 | 1030 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
1031 | |
1032 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 1033 | eb = folio_get_private(folio); |
139e8cd3 QW |
1034 | BUG_ON(!eb); |
1035 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1036 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 1037 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 1038 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 1039 | } |
0079c3b1 | 1040 | subpage = folio_get_private(folio); |
139e8cd3 QW |
1041 | |
1042 | ASSERT(subpage->dirty_bitmap); | |
1043 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
1044 | unsigned long flags; | |
1045 | u64 cur; | |
1046 | u16 tmp = (1 << cur_bit); | |
1047 | ||
1048 | spin_lock_irqsave(&subpage->lock, flags); | |
1049 | if (!(tmp & subpage->dirty_bitmap)) { | |
1050 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1051 | cur_bit++; | |
1052 | continue; | |
1053 | } | |
1054 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1055 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 1056 | |
139e8cd3 QW |
1057 | eb = find_extent_buffer(fs_info, cur); |
1058 | ASSERT(eb); | |
1059 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1060 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 1061 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
1062 | free_extent_buffer(eb); |
1063 | ||
1064 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
1065 | } | |
0079c3b1 | 1066 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 1067 | } |
0079c3b1 MWO |
1068 | #else |
1069 | #define btree_dirty_folio filemap_dirty_folio | |
1070 | #endif | |
0b32f4bb | 1071 | |
7f09410b | 1072 | static const struct address_space_operations btree_aops = { |
0da5468f | 1073 | .writepages = btree_writepages, |
f913cff3 | 1074 | .release_folio = btree_release_folio, |
895586eb | 1075 | .invalidate_folio = btree_invalidate_folio, |
5a92bc88 | 1076 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1077 | .migratepage = btree_migratepage, |
5a92bc88 | 1078 | #endif |
0079c3b1 | 1079 | .dirty_folio = btree_dirty_folio, |
d98237b3 CM |
1080 | }; |
1081 | ||
2ff7e61e JM |
1082 | struct extent_buffer *btrfs_find_create_tree_block( |
1083 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
1084 | u64 bytenr, u64 owner_root, |
1085 | int level) | |
0999df54 | 1086 | { |
0b246afa JM |
1087 | if (btrfs_is_testing(fs_info)) |
1088 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 1089 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
1090 | } |
1091 | ||
581c1760 QW |
1092 | /* |
1093 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1094 | * verification. | |
1095 | * | |
1b7ec85e | 1096 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
1097 | * @parent_transid: expected transid of this tree block, skip check if 0 |
1098 | * @level: expected level, mandatory check | |
1099 | * @first_key: expected key in slot 0, skip check if NULL | |
1100 | */ | |
2ff7e61e | 1101 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
1102 | u64 owner_root, u64 parent_transid, |
1103 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
1104 | { |
1105 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1106 | int ret; |
1107 | ||
3fbaf258 | 1108 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
1109 | if (IS_ERR(buf)) |
1110 | return buf; | |
0999df54 | 1111 | |
6a2e9dc4 | 1112 | ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); |
0f0fe8f7 | 1113 | if (ret) { |
537f38f0 | 1114 | free_extent_buffer_stale(buf); |
64c043de | 1115 | return ERR_PTR(ret); |
0f0fe8f7 | 1116 | } |
88c602ab QW |
1117 | if (btrfs_check_eb_owner(buf, owner_root)) { |
1118 | free_extent_buffer_stale(buf); | |
1119 | return ERR_PTR(-EUCLEAN); | |
1120 | } | |
5f39d397 | 1121 | return buf; |
ce9adaa5 | 1122 | |
eb60ceac CM |
1123 | } |
1124 | ||
6a884d7d | 1125 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1126 | { |
6a884d7d | 1127 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1128 | if (btrfs_header_generation(buf) == |
e2d84521 | 1129 | fs_info->running_transaction->transid) { |
49d0c642 | 1130 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 1131 | |
b9473439 | 1132 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1133 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1134 | -buf->len, | |
1135 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1136 | clear_extent_buffer_dirty(buf); |
1137 | } | |
925baedd | 1138 | } |
5f39d397 CM |
1139 | } |
1140 | ||
da17066c | 1141 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1142 | u64 objectid) |
d97e63b6 | 1143 | { |
7c0260ee | 1144 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
1145 | |
1146 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
1147 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
1148 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 1149 | root->fs_info = fs_info; |
2e608bd1 | 1150 | root->root_key.objectid = objectid; |
cfaa7295 | 1151 | root->node = NULL; |
a28ec197 | 1152 | root->commit_root = NULL; |
27cdeb70 | 1153 | root->state = 0; |
abed4aaa | 1154 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 1155 | |
0f7d52f4 | 1156 | root->last_trans = 0; |
6b8fad57 | 1157 | root->free_objectid = 0; |
eb73c1b7 | 1158 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1159 | root->nr_ordered_extents = 0; |
6bef4d31 | 1160 | root->inode_tree = RB_ROOT; |
253bf575 | 1161 | xa_init_flags(&root->delayed_nodes, GFP_ATOMIC); |
2e608bd1 JB |
1162 | |
1163 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
1164 | |
1165 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1166 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1167 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1168 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1169 | INIT_LIST_HEAD(&root->ordered_extents); |
1170 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1171 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1172 | INIT_LIST_HEAD(&root->logged_list[0]); |
1173 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1174 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1175 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1176 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1177 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1178 | spin_lock_init(&root->log_extents_lock[0]); |
1179 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1180 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1181 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1182 | mutex_init(&root->log_mutex); |
31f3d255 | 1183 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1184 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1185 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1186 | init_waitqueue_head(&root->log_writer_wait); |
1187 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1188 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1189 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1190 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1191 | atomic_set(&root->log_commit[0], 0); |
1192 | atomic_set(&root->log_commit[1], 0); | |
1193 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1194 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1195 | refcount_set(&root->refs, 1); |
8ecebf4d | 1196 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1197 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1198 | root->log_transid = 0; |
d1433deb | 1199 | root->log_transid_committed = -1; |
257c62e1 | 1200 | root->last_log_commit = 0; |
2e608bd1 | 1201 | root->anon_dev = 0; |
e289f03e | 1202 | if (!dummy) { |
43eb5f29 QW |
1203 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1204 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1205 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1206 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1207 | } | |
017e5369 | 1208 | |
5f3ab90a | 1209 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1210 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1211 | #ifdef CONFIG_BTRFS_DEBUG |
1212 | INIT_LIST_HEAD(&root->leak_list); | |
48b36a60 | 1213 | spin_lock(&fs_info->fs_roots_lock); |
bd647ce3 | 1214 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
48b36a60 | 1215 | spin_unlock(&fs_info->fs_roots_lock); |
bd647ce3 | 1216 | #endif |
3768f368 CM |
1217 | } |
1218 | ||
74e4d827 | 1219 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1220 | u64 objectid, gfp_t flags) |
6f07e42e | 1221 | { |
74e4d827 | 1222 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1223 | if (root) |
96dfcb46 | 1224 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1225 | return root; |
1226 | } | |
1227 | ||
06ea65a3 JB |
1228 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1229 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1230 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1231 | { |
1232 | struct btrfs_root *root; | |
1233 | ||
7c0260ee JM |
1234 | if (!fs_info) |
1235 | return ERR_PTR(-EINVAL); | |
1236 | ||
96dfcb46 | 1237 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1238 | if (!root) |
1239 | return ERR_PTR(-ENOMEM); | |
da17066c | 1240 | |
b9ef22de | 1241 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1242 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1243 | |
1244 | return root; | |
1245 | } | |
1246 | #endif | |
1247 | ||
abed4aaa JB |
1248 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1249 | { | |
1250 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1251 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1252 | ||
1253 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1254 | } | |
1255 | ||
1256 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1257 | { | |
1258 | const struct btrfs_key *key = k; | |
1259 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1260 | ||
1261 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1262 | } | |
1263 | ||
1264 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1265 | { | |
1266 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1267 | struct rb_node *tmp; | |
1268 | ||
1269 | write_lock(&fs_info->global_root_lock); | |
1270 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1271 | write_unlock(&fs_info->global_root_lock); | |
1272 | ASSERT(!tmp); | |
1273 | ||
1274 | return tmp ? -EEXIST : 0; | |
1275 | } | |
1276 | ||
1277 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1278 | { | |
1279 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1280 | ||
1281 | write_lock(&fs_info->global_root_lock); | |
1282 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1283 | write_unlock(&fs_info->global_root_lock); | |
1284 | } | |
1285 | ||
1286 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1287 | struct btrfs_key *key) | |
1288 | { | |
1289 | struct rb_node *node; | |
1290 | struct btrfs_root *root = NULL; | |
1291 | ||
1292 | read_lock(&fs_info->global_root_lock); | |
1293 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1294 | if (node) | |
1295 | root = container_of(node, struct btrfs_root, rb_node); | |
1296 | read_unlock(&fs_info->global_root_lock); | |
1297 | ||
1298 | return root; | |
1299 | } | |
1300 | ||
f7238e50 JB |
1301 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
1302 | { | |
1303 | struct btrfs_block_group *block_group; | |
1304 | u64 ret; | |
1305 | ||
1306 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1307 | return 0; | |
1308 | ||
1309 | if (bytenr) | |
1310 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
1311 | else | |
1312 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
1313 | ASSERT(block_group); | |
1314 | if (!block_group) | |
1315 | return 0; | |
1316 | ret = block_group->global_root_id; | |
1317 | btrfs_put_block_group(block_group); | |
1318 | ||
1319 | return ret; | |
1320 | } | |
1321 | ||
abed4aaa JB |
1322 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
1323 | { | |
1324 | struct btrfs_key key = { | |
1325 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1326 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1327 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1328 | }; |
1329 | ||
1330 | return btrfs_global_root(fs_info, &key); | |
1331 | } | |
1332 | ||
1333 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1334 | { | |
1335 | struct btrfs_key key = { | |
1336 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1337 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1338 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1339 | }; |
1340 | ||
1341 | return btrfs_global_root(fs_info, &key); | |
1342 | } | |
1343 | ||
20897f5c | 1344 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1345 | u64 objectid) |
1346 | { | |
9b7a2440 | 1347 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1348 | struct extent_buffer *leaf; |
1349 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1350 | struct btrfs_root *root; | |
1351 | struct btrfs_key key; | |
b89f6d1f | 1352 | unsigned int nofs_flag; |
20897f5c | 1353 | int ret = 0; |
20897f5c | 1354 | |
b89f6d1f FM |
1355 | /* |
1356 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1357 | * context to avoid deadlock if reclaim happens. | |
1358 | */ | |
1359 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1360 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1361 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1362 | if (!root) |
1363 | return ERR_PTR(-ENOMEM); | |
1364 | ||
20897f5c AJ |
1365 | root->root_key.objectid = objectid; |
1366 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1367 | root->root_key.offset = 0; | |
1368 | ||
9631e4cc JB |
1369 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1370 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1371 | if (IS_ERR(leaf)) { |
1372 | ret = PTR_ERR(leaf); | |
1dd05682 | 1373 | leaf = NULL; |
8a6a87cd | 1374 | goto fail_unlock; |
20897f5c AJ |
1375 | } |
1376 | ||
20897f5c | 1377 | root->node = leaf; |
20897f5c AJ |
1378 | btrfs_mark_buffer_dirty(leaf); |
1379 | ||
1380 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1381 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1382 | |
f944d2cb DS |
1383 | btrfs_set_root_flags(&root->root_item, 0); |
1384 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1385 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1386 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1387 | btrfs_set_root_level(&root->root_item, 0); | |
1388 | btrfs_set_root_refs(&root->root_item, 1); | |
1389 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1390 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1391 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1392 | if (is_fstree(objectid)) |
807fc790 AS |
1393 | generate_random_guid(root->root_item.uuid); |
1394 | else | |
1395 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1396 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1397 | |
8a6a87cd BB |
1398 | btrfs_tree_unlock(leaf); |
1399 | ||
20897f5c AJ |
1400 | key.objectid = objectid; |
1401 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1402 | key.offset = 0; | |
1403 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1404 | if (ret) | |
1405 | goto fail; | |
1406 | ||
1dd05682 TI |
1407 | return root; |
1408 | ||
8a6a87cd | 1409 | fail_unlock: |
8c38938c | 1410 | if (leaf) |
1dd05682 | 1411 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1412 | fail: |
00246528 | 1413 | btrfs_put_root(root); |
20897f5c | 1414 | |
1dd05682 | 1415 | return ERR_PTR(ret); |
20897f5c AJ |
1416 | } |
1417 | ||
7237f183 YZ |
1418 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1419 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1420 | { |
1421 | struct btrfs_root *root; | |
e02119d5 | 1422 | |
96dfcb46 | 1423 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1424 | if (!root) |
7237f183 | 1425 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1426 | |
e02119d5 CM |
1427 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1428 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1429 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1430 | |
6ab6ebb7 NA |
1431 | return root; |
1432 | } | |
1433 | ||
1434 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1435 | struct btrfs_root *root) | |
1436 | { | |
1437 | struct extent_buffer *leaf; | |
1438 | ||
7237f183 | 1439 | /* |
92a7cc42 | 1440 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1441 | * |
92a7cc42 QW |
1442 | * Log trees are not exposed to user space thus can't be snapshotted, |
1443 | * and they go away before a real commit is actually done. | |
1444 | * | |
1445 | * They do store pointers to file data extents, and those reference | |
1446 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1447 | */ |
e02119d5 | 1448 | |
4d75f8a9 | 1449 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1450 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1451 | if (IS_ERR(leaf)) |
1452 | return PTR_ERR(leaf); | |
e02119d5 | 1453 | |
7237f183 | 1454 | root->node = leaf; |
e02119d5 | 1455 | |
e02119d5 CM |
1456 | btrfs_mark_buffer_dirty(root->node); |
1457 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1458 | |
1459 | return 0; | |
7237f183 YZ |
1460 | } |
1461 | ||
1462 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1463 | struct btrfs_fs_info *fs_info) | |
1464 | { | |
1465 | struct btrfs_root *log_root; | |
1466 | ||
1467 | log_root = alloc_log_tree(trans, fs_info); | |
1468 | if (IS_ERR(log_root)) | |
1469 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1470 | |
3ddebf27 NA |
1471 | if (!btrfs_is_zoned(fs_info)) { |
1472 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1473 | ||
1474 | if (ret) { | |
1475 | btrfs_put_root(log_root); | |
1476 | return ret; | |
1477 | } | |
6ab6ebb7 NA |
1478 | } |
1479 | ||
7237f183 YZ |
1480 | WARN_ON(fs_info->log_root_tree); |
1481 | fs_info->log_root_tree = log_root; | |
1482 | return 0; | |
1483 | } | |
1484 | ||
1485 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1486 | struct btrfs_root *root) | |
1487 | { | |
0b246afa | 1488 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1489 | struct btrfs_root *log_root; |
1490 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1491 | int ret; |
7237f183 | 1492 | |
0b246afa | 1493 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1494 | if (IS_ERR(log_root)) |
1495 | return PTR_ERR(log_root); | |
1496 | ||
6ab6ebb7 NA |
1497 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1498 | if (ret) { | |
1499 | btrfs_put_root(log_root); | |
1500 | return ret; | |
1501 | } | |
1502 | ||
7237f183 YZ |
1503 | log_root->last_trans = trans->transid; |
1504 | log_root->root_key.offset = root->root_key.objectid; | |
1505 | ||
1506 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1507 | btrfs_set_stack_inode_generation(inode_item, 1); |
1508 | btrfs_set_stack_inode_size(inode_item, 3); | |
1509 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1510 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1511 | fs_info->nodesize); |
3cae210f | 1512 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1513 | |
5d4f98a2 | 1514 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1515 | |
1516 | WARN_ON(root->log_root); | |
1517 | root->log_root = log_root; | |
1518 | root->log_transid = 0; | |
d1433deb | 1519 | root->log_transid_committed = -1; |
257c62e1 | 1520 | root->last_log_commit = 0; |
e02119d5 CM |
1521 | return 0; |
1522 | } | |
1523 | ||
49d11bea JB |
1524 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1525 | struct btrfs_path *path, | |
1526 | struct btrfs_key *key) | |
e02119d5 CM |
1527 | { |
1528 | struct btrfs_root *root; | |
1529 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1530 | u64 generation; |
cb517eab | 1531 | int ret; |
581c1760 | 1532 | int level; |
0f7d52f4 | 1533 | |
96dfcb46 | 1534 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1535 | if (!root) |
1536 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1537 | |
cb517eab MX |
1538 | ret = btrfs_find_root(tree_root, key, path, |
1539 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1540 | if (ret) { |
13a8a7c8 YZ |
1541 | if (ret > 0) |
1542 | ret = -ENOENT; | |
49d11bea | 1543 | goto fail; |
0f7d52f4 | 1544 | } |
13a8a7c8 | 1545 | |
84234f3a | 1546 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1547 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1548 | root->node = read_tree_block(fs_info, |
1549 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1550 | key->objectid, generation, level, NULL); |
64c043de LB |
1551 | if (IS_ERR(root->node)) { |
1552 | ret = PTR_ERR(root->node); | |
8c38938c | 1553 | root->node = NULL; |
49d11bea | 1554 | goto fail; |
4eb150d6 QW |
1555 | } |
1556 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1557 | ret = -EIO; |
49d11bea | 1558 | goto fail; |
416bc658 | 1559 | } |
88c602ab QW |
1560 | |
1561 | /* | |
1562 | * For real fs, and not log/reloc trees, root owner must | |
1563 | * match its root node owner | |
1564 | */ | |
1565 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1566 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1567 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1568 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1569 | btrfs_crit(fs_info, | |
1570 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1571 | root->root_key.objectid, root->node->start, | |
1572 | btrfs_header_owner(root->node), | |
1573 | root->root_key.objectid); | |
1574 | ret = -EUCLEAN; | |
1575 | goto fail; | |
1576 | } | |
5d4f98a2 | 1577 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1578 | return root; |
49d11bea | 1579 | fail: |
00246528 | 1580 | btrfs_put_root(root); |
49d11bea JB |
1581 | return ERR_PTR(ret); |
1582 | } | |
1583 | ||
1584 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1585 | struct btrfs_key *key) | |
1586 | { | |
1587 | struct btrfs_root *root; | |
1588 | struct btrfs_path *path; | |
1589 | ||
1590 | path = btrfs_alloc_path(); | |
1591 | if (!path) | |
1592 | return ERR_PTR(-ENOMEM); | |
1593 | root = read_tree_root_path(tree_root, path, key); | |
1594 | btrfs_free_path(path); | |
1595 | ||
1596 | return root; | |
cb517eab MX |
1597 | } |
1598 | ||
2dfb1e43 QW |
1599 | /* |
1600 | * Initialize subvolume root in-memory structure | |
1601 | * | |
1602 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1603 | */ | |
1604 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1605 | { |
1606 | int ret; | |
dcc3eb96 | 1607 | unsigned int nofs_flag; |
cb517eab | 1608 | |
dcc3eb96 NB |
1609 | /* |
1610 | * We might be called under a transaction (e.g. indirect backref | |
1611 | * resolution) which could deadlock if it triggers memory reclaim | |
1612 | */ | |
1613 | nofs_flag = memalloc_nofs_save(); | |
1614 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1615 | memalloc_nofs_restore(nofs_flag); | |
1616 | if (ret) | |
8257b2dc | 1617 | goto fail; |
8257b2dc | 1618 | |
aeb935a4 | 1619 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1620 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1621 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1622 | btrfs_check_and_init_root_item(&root->root_item); |
1623 | } | |
1624 | ||
851fd730 QW |
1625 | /* |
1626 | * Don't assign anonymous block device to roots that are not exposed to | |
1627 | * userspace, the id pool is limited to 1M | |
1628 | */ | |
1629 | if (is_fstree(root->root_key.objectid) && | |
1630 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1631 | if (!anon_dev) { |
1632 | ret = get_anon_bdev(&root->anon_dev); | |
1633 | if (ret) | |
1634 | goto fail; | |
1635 | } else { | |
1636 | root->anon_dev = anon_dev; | |
1637 | } | |
851fd730 | 1638 | } |
f32e48e9 CR |
1639 | |
1640 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1641 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1642 | if (ret) { |
1643 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1644 | goto fail; |
f32e48e9 CR |
1645 | } |
1646 | ||
6b8fad57 | 1647 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1648 | |
1649 | mutex_unlock(&root->objectid_mutex); | |
1650 | ||
cb517eab MX |
1651 | return 0; |
1652 | fail: | |
84db5ccf | 1653 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1654 | return ret; |
1655 | } | |
1656 | ||
a98db0f3 JB |
1657 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1658 | u64 root_id) | |
cb517eab MX |
1659 | { |
1660 | struct btrfs_root *root; | |
1661 | ||
48b36a60 GN |
1662 | spin_lock(&fs_info->fs_roots_lock); |
1663 | root = xa_load(&fs_info->fs_roots, (unsigned long)root_id); | |
bc44d7c4 | 1664 | if (root) |
00246528 | 1665 | root = btrfs_grab_root(root); |
48b36a60 | 1666 | spin_unlock(&fs_info->fs_roots_lock); |
cb517eab MX |
1667 | return root; |
1668 | } | |
1669 | ||
49d11bea JB |
1670 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1671 | u64 objectid) | |
1672 | { | |
abed4aaa JB |
1673 | struct btrfs_key key = { |
1674 | .objectid = objectid, | |
1675 | .type = BTRFS_ROOT_ITEM_KEY, | |
1676 | .offset = 0, | |
1677 | }; | |
1678 | ||
49d11bea JB |
1679 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1680 | return btrfs_grab_root(fs_info->tree_root); | |
1681 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1682 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1683 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1684 | return btrfs_grab_root(fs_info->chunk_root); | |
1685 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1686 | return btrfs_grab_root(fs_info->dev_root); | |
1687 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1688 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1689 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1690 | return btrfs_grab_root(fs_info->quota_root) ? | |
1691 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1692 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1693 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1694 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1695 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1696 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1697 | ||
1698 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1699 | } | |
49d11bea JB |
1700 | return NULL; |
1701 | } | |
1702 | ||
cb517eab MX |
1703 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1704 | struct btrfs_root *root) | |
1705 | { | |
1706 | int ret; | |
1707 | ||
48b36a60 GN |
1708 | spin_lock(&fs_info->fs_roots_lock); |
1709 | ret = xa_insert(&fs_info->fs_roots, (unsigned long)root->root_key.objectid, | |
1710 | root, GFP_NOFS); | |
af01d2e5 | 1711 | if (ret == 0) { |
00246528 | 1712 | btrfs_grab_root(root); |
48b36a60 | 1713 | set_bit(BTRFS_ROOT_REGISTERED, &root->state); |
af01d2e5 | 1714 | } |
48b36a60 | 1715 | spin_unlock(&fs_info->fs_roots_lock); |
cb517eab MX |
1716 | |
1717 | return ret; | |
1718 | } | |
1719 | ||
bd647ce3 JB |
1720 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1721 | { | |
1722 | #ifdef CONFIG_BTRFS_DEBUG | |
1723 | struct btrfs_root *root; | |
1724 | ||
1725 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1726 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1727 | ||
bd647ce3 JB |
1728 | root = list_first_entry(&fs_info->allocated_roots, |
1729 | struct btrfs_root, leak_list); | |
457f1864 | 1730 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1731 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1732 | refcount_read(&root->refs)); |
1733 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1734 | btrfs_put_root(root); |
1735 | btrfs_put_root(root); | |
bd647ce3 JB |
1736 | } |
1737 | #endif | |
1738 | } | |
1739 | ||
abed4aaa JB |
1740 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1741 | { | |
1742 | struct btrfs_root *root; | |
1743 | struct rb_node *node; | |
1744 | ||
1745 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1746 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1747 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1748 | btrfs_put_root(root); | |
1749 | } | |
1750 | } | |
1751 | ||
0d4b0463 JB |
1752 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1753 | { | |
141386e1 JB |
1754 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1755 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1756 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1757 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1758 | btrfs_free_csum_hash(fs_info); | |
1759 | btrfs_free_stripe_hash_table(fs_info); | |
1760 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1761 | kfree(fs_info->balance_ctl); |
1762 | kfree(fs_info->delayed_root); | |
abed4aaa | 1763 | free_global_roots(fs_info); |
00246528 JB |
1764 | btrfs_put_root(fs_info->tree_root); |
1765 | btrfs_put_root(fs_info->chunk_root); | |
1766 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1767 | btrfs_put_root(fs_info->quota_root); |
1768 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1769 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1770 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1771 | btrfs_put_root(fs_info->block_group_root); |
bd647ce3 | 1772 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1773 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1774 | kfree(fs_info->super_copy); |
1775 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1776 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1777 | kvfree(fs_info); |
1778 | } | |
1779 | ||
1780 | ||
2dfb1e43 QW |
1781 | /* |
1782 | * Get an in-memory reference of a root structure. | |
1783 | * | |
1784 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1785 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1786 | * found, then read it from disk and add it to cached fs roots. | |
1787 | * | |
1788 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1789 | * | |
1790 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1791 | * same root objectid. | |
1792 | * | |
1793 | * @objectid: root id | |
1794 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1795 | * pass 0 for new allocation. | |
1796 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1797 | * for orphan roots | |
1798 | */ | |
1799 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1800 | u64 objectid, dev_t anon_dev, | |
1801 | bool check_ref) | |
5eda7b5e CM |
1802 | { |
1803 | struct btrfs_root *root; | |
381cf658 | 1804 | struct btrfs_path *path; |
1d4c08e0 | 1805 | struct btrfs_key key; |
5eda7b5e CM |
1806 | int ret; |
1807 | ||
49d11bea JB |
1808 | root = btrfs_get_global_root(fs_info, objectid); |
1809 | if (root) | |
1810 | return root; | |
4df27c4d | 1811 | again: |
56e9357a | 1812 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1813 | if (root) { |
2dfb1e43 QW |
1814 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1815 | ASSERT(!anon_dev); | |
bc44d7c4 | 1816 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1817 | btrfs_put_root(root); |
48475471 | 1818 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1819 | } |
5eda7b5e | 1820 | return root; |
48475471 | 1821 | } |
5eda7b5e | 1822 | |
56e9357a DS |
1823 | key.objectid = objectid; |
1824 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1825 | key.offset = (u64)-1; | |
1826 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1827 | if (IS_ERR(root)) |
1828 | return root; | |
3394e160 | 1829 | |
c00869f1 | 1830 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1831 | ret = -ENOENT; |
581bb050 | 1832 | goto fail; |
35a30d7c | 1833 | } |
581bb050 | 1834 | |
2dfb1e43 | 1835 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1836 | if (ret) |
1837 | goto fail; | |
3394e160 | 1838 | |
381cf658 DS |
1839 | path = btrfs_alloc_path(); |
1840 | if (!path) { | |
1841 | ret = -ENOMEM; | |
1842 | goto fail; | |
1843 | } | |
1d4c08e0 DS |
1844 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1845 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1846 | key.offset = objectid; |
1d4c08e0 DS |
1847 | |
1848 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1849 | btrfs_free_path(path); |
d68fc57b YZ |
1850 | if (ret < 0) |
1851 | goto fail; | |
1852 | if (ret == 0) | |
27cdeb70 | 1853 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1854 | |
cb517eab | 1855 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1856 | if (ret) { |
168a2f77 JJB |
1857 | if (ret == -EEXIST) { |
1858 | btrfs_put_root(root); | |
4df27c4d | 1859 | goto again; |
168a2f77 | 1860 | } |
4df27c4d | 1861 | goto fail; |
0f7d52f4 | 1862 | } |
edbd8d4e | 1863 | return root; |
4df27c4d | 1864 | fail: |
33fab972 FM |
1865 | /* |
1866 | * If our caller provided us an anonymous device, then it's his | |
1867 | * responsability to free it in case we fail. So we have to set our | |
1868 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() | |
1869 | * and once again by our caller. | |
1870 | */ | |
1871 | if (anon_dev) | |
1872 | root->anon_dev = 0; | |
8c38938c | 1873 | btrfs_put_root(root); |
4df27c4d | 1874 | return ERR_PTR(ret); |
edbd8d4e CM |
1875 | } |
1876 | ||
2dfb1e43 QW |
1877 | /* |
1878 | * Get in-memory reference of a root structure | |
1879 | * | |
1880 | * @objectid: tree objectid | |
1881 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1882 | * one reference | |
1883 | */ | |
1884 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1885 | u64 objectid, bool check_ref) | |
1886 | { | |
1887 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1888 | } | |
1889 | ||
1890 | /* | |
1891 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1892 | * the anonymous block device id | |
1893 | * | |
1894 | * @objectid: tree objectid | |
1895 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1896 | * parameter value | |
1897 | */ | |
1898 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1899 | u64 objectid, dev_t anon_dev) | |
1900 | { | |
1901 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1902 | } | |
1903 | ||
49d11bea JB |
1904 | /* |
1905 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1906 | * @fs_info: the fs_info | |
1907 | * @objectid: the objectid we need to lookup | |
1908 | * | |
1909 | * This is exclusively used for backref walking, and exists specifically because | |
1910 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1911 | * creation time, which means we may have to read the tree_root in order to look | |
1912 | * up a fs root that is not in memory. If the root is not in memory we will | |
1913 | * read the tree root commit root and look up the fs root from there. This is a | |
1914 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1915 | * have the most uptodate information, it'll simply be discarded once the | |
1916 | * backref code is finished using the root. | |
1917 | */ | |
1918 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1919 | struct btrfs_path *path, | |
1920 | u64 objectid) | |
1921 | { | |
1922 | struct btrfs_root *root; | |
1923 | struct btrfs_key key; | |
1924 | ||
1925 | ASSERT(path->search_commit_root && path->skip_locking); | |
1926 | ||
1927 | /* | |
1928 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1929 | * since this is called via the backref walking code we won't be looking | |
1930 | * up a root that doesn't exist, unless there's corruption. So if root | |
1931 | * != NULL just return it. | |
1932 | */ | |
1933 | root = btrfs_get_global_root(fs_info, objectid); | |
1934 | if (root) | |
1935 | return root; | |
1936 | ||
1937 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1938 | if (root) | |
1939 | return root; | |
1940 | ||
1941 | key.objectid = objectid; | |
1942 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1943 | key.offset = (u64)-1; | |
1944 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1945 | btrfs_release_path(path); | |
1946 | ||
1947 | return root; | |
1948 | } | |
1949 | ||
8b712842 CM |
1950 | /* |
1951 | * called by the kthread helper functions to finally call the bio end_io | |
1952 | * functions. This is where read checksum verification actually happens | |
1953 | */ | |
1954 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1955 | { |
ce9adaa5 | 1956 | struct bio *bio; |
97eb6b69 | 1957 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1958 | |
97eb6b69 | 1959 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1960 | bio = end_io_wq->bio; |
ce9adaa5 | 1961 | |
4e4cbee9 | 1962 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1963 | bio->bi_private = end_io_wq->private; |
1964 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1965 | bio_endio(bio); |
9be490f1 | 1966 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1967 | } |
1968 | ||
a74a4b97 CM |
1969 | static int cleaner_kthread(void *arg) |
1970 | { | |
0d031dc4 | 1971 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1972 | int again; |
a74a4b97 | 1973 | |
d6fd0ae2 | 1974 | while (1) { |
d0278245 | 1975 | again = 0; |
a74a4b97 | 1976 | |
fd340d0f JB |
1977 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1978 | ||
d0278245 | 1979 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1980 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1981 | goto sleep; |
1982 | ||
90c711ab ZB |
1983 | /* |
1984 | * Do not do anything if we might cause open_ctree() to block | |
1985 | * before we have finished mounting the filesystem. | |
1986 | */ | |
0b246afa | 1987 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1988 | goto sleep; |
1989 | ||
0b246afa | 1990 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1991 | goto sleep; |
1992 | ||
dc7f370c MX |
1993 | /* |
1994 | * Avoid the problem that we change the status of the fs | |
1995 | * during the above check and trylock. | |
1996 | */ | |
2ff7e61e | 1997 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1998 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1999 | goto sleep; |
76dda93c | 2000 | } |
a74a4b97 | 2001 | |
2ff7e61e | 2002 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 2003 | |
33c44184 | 2004 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 2005 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
2006 | |
2007 | /* | |
05323cd1 MX |
2008 | * The defragger has dealt with the R/O remount and umount, |
2009 | * needn't do anything special here. | |
d0278245 | 2010 | */ |
0b246afa | 2011 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
2012 | |
2013 | /* | |
f3372065 | 2014 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
2015 | * with relocation (btrfs_relocate_chunk) and relocation |
2016 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 2017 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
2018 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
2019 | * unused block groups. | |
2020 | */ | |
0b246afa | 2021 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
2022 | |
2023 | /* | |
2024 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
2025 | * all unused block_groups. This possibly gives us some more free | |
2026 | * space. | |
2027 | */ | |
2028 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 2029 | sleep: |
a0a1db70 | 2030 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
2031 | if (kthread_should_park()) |
2032 | kthread_parkme(); | |
2033 | if (kthread_should_stop()) | |
2034 | return 0; | |
838fe188 | 2035 | if (!again) { |
a74a4b97 | 2036 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 2037 | schedule(); |
a74a4b97 CM |
2038 | __set_current_state(TASK_RUNNING); |
2039 | } | |
da288d28 | 2040 | } |
a74a4b97 CM |
2041 | } |
2042 | ||
2043 | static int transaction_kthread(void *arg) | |
2044 | { | |
2045 | struct btrfs_root *root = arg; | |
0b246afa | 2046 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
2047 | struct btrfs_trans_handle *trans; |
2048 | struct btrfs_transaction *cur; | |
8929ecfa | 2049 | u64 transid; |
643900be | 2050 | time64_t delta; |
a74a4b97 | 2051 | unsigned long delay; |
914b2007 | 2052 | bool cannot_commit; |
a74a4b97 CM |
2053 | |
2054 | do { | |
914b2007 | 2055 | cannot_commit = false; |
ba1bc00f | 2056 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 2057 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 2058 | |
0b246afa JM |
2059 | spin_lock(&fs_info->trans_lock); |
2060 | cur = fs_info->running_transaction; | |
a74a4b97 | 2061 | if (!cur) { |
0b246afa | 2062 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
2063 | goto sleep; |
2064 | } | |
31153d81 | 2065 | |
643900be | 2066 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
2067 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
2068 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 2069 | delta < fs_info->commit_interval) { |
0b246afa | 2070 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
2071 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
2072 | delay = min(delay, | |
2073 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
2074 | goto sleep; |
2075 | } | |
8929ecfa | 2076 | transid = cur->transid; |
0b246afa | 2077 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 2078 | |
79787eaa | 2079 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 2080 | trans = btrfs_attach_transaction(root); |
914b2007 | 2081 | if (IS_ERR(trans)) { |
354aa0fb MX |
2082 | if (PTR_ERR(trans) != -ENOENT) |
2083 | cannot_commit = true; | |
79787eaa | 2084 | goto sleep; |
914b2007 | 2085 | } |
8929ecfa | 2086 | if (transid == trans->transid) { |
3a45bb20 | 2087 | btrfs_commit_transaction(trans); |
8929ecfa | 2088 | } else { |
3a45bb20 | 2089 | btrfs_end_transaction(trans); |
8929ecfa | 2090 | } |
a74a4b97 | 2091 | sleep: |
0b246afa JM |
2092 | wake_up_process(fs_info->cleaner_kthread); |
2093 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 2094 | |
84961539 | 2095 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 2096 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 2097 | if (!kthread_should_stop() && |
0b246afa | 2098 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 2099 | cannot_commit)) |
bc5511d0 | 2100 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
2101 | } while (!kthread_should_stop()); |
2102 | return 0; | |
2103 | } | |
2104 | ||
af31f5e5 | 2105 | /* |
01f0f9da NB |
2106 | * This will find the highest generation in the array of root backups. The |
2107 | * index of the highest array is returned, or -EINVAL if we can't find | |
2108 | * anything. | |
af31f5e5 CM |
2109 | * |
2110 | * We check to make sure the array is valid by comparing the | |
2111 | * generation of the latest root in the array with the generation | |
2112 | * in the super block. If they don't match we pitch it. | |
2113 | */ | |
01f0f9da | 2114 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 2115 | { |
01f0f9da | 2116 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 2117 | u64 cur; |
af31f5e5 CM |
2118 | struct btrfs_root_backup *root_backup; |
2119 | int i; | |
2120 | ||
2121 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2122 | root_backup = info->super_copy->super_roots + i; | |
2123 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2124 | if (cur == newest_gen) | |
01f0f9da | 2125 | return i; |
af31f5e5 CM |
2126 | } |
2127 | ||
01f0f9da | 2128 | return -EINVAL; |
af31f5e5 CM |
2129 | } |
2130 | ||
af31f5e5 CM |
2131 | /* |
2132 | * copy all the root pointers into the super backup array. | |
2133 | * this will bump the backup pointer by one when it is | |
2134 | * done | |
2135 | */ | |
2136 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2137 | { | |
6ef108dd | 2138 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2139 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
2140 | |
2141 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2142 | ||
2143 | /* | |
2144 | * make sure all of our padding and empty slots get zero filled | |
2145 | * regardless of which ones we use today | |
2146 | */ | |
2147 | memset(root_backup, 0, sizeof(*root_backup)); | |
2148 | ||
2149 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2150 | ||
2151 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2152 | btrfs_set_backup_tree_root_gen(root_backup, | |
2153 | btrfs_header_generation(info->tree_root->node)); | |
2154 | ||
2155 | btrfs_set_backup_tree_root_level(root_backup, | |
2156 | btrfs_header_level(info->tree_root->node)); | |
2157 | ||
2158 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2159 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2160 | btrfs_header_generation(info->chunk_root->node)); | |
2161 | btrfs_set_backup_chunk_root_level(root_backup, | |
2162 | btrfs_header_level(info->chunk_root->node)); | |
2163 | ||
9c54e80d JB |
2164 | if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) { |
2165 | btrfs_set_backup_block_group_root(root_backup, | |
2166 | info->block_group_root->node->start); | |
2167 | btrfs_set_backup_block_group_root_gen(root_backup, | |
2168 | btrfs_header_generation(info->block_group_root->node)); | |
2169 | btrfs_set_backup_block_group_root_level(root_backup, | |
2170 | btrfs_header_level(info->block_group_root->node)); | |
2171 | } else { | |
2172 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); | |
f7238e50 | 2173 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
2174 | |
2175 | btrfs_set_backup_extent_root(root_backup, | |
2176 | extent_root->node->start); | |
2177 | btrfs_set_backup_extent_root_gen(root_backup, | |
2178 | btrfs_header_generation(extent_root->node)); | |
2179 | btrfs_set_backup_extent_root_level(root_backup, | |
2180 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
2181 | |
2182 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
2183 | btrfs_set_backup_csum_root_gen(root_backup, | |
2184 | btrfs_header_generation(csum_root->node)); | |
2185 | btrfs_set_backup_csum_root_level(root_backup, | |
2186 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 2187 | } |
af31f5e5 | 2188 | |
7c7e82a7 CM |
2189 | /* |
2190 | * we might commit during log recovery, which happens before we set | |
2191 | * the fs_root. Make sure it is valid before we fill it in. | |
2192 | */ | |
2193 | if (info->fs_root && info->fs_root->node) { | |
2194 | btrfs_set_backup_fs_root(root_backup, | |
2195 | info->fs_root->node->start); | |
2196 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2197 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2198 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2199 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2200 | } |
af31f5e5 CM |
2201 | |
2202 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2203 | btrfs_set_backup_dev_root_gen(root_backup, | |
2204 | btrfs_header_generation(info->dev_root->node)); | |
2205 | btrfs_set_backup_dev_root_level(root_backup, | |
2206 | btrfs_header_level(info->dev_root->node)); | |
2207 | ||
af31f5e5 CM |
2208 | btrfs_set_backup_total_bytes(root_backup, |
2209 | btrfs_super_total_bytes(info->super_copy)); | |
2210 | btrfs_set_backup_bytes_used(root_backup, | |
2211 | btrfs_super_bytes_used(info->super_copy)); | |
2212 | btrfs_set_backup_num_devices(root_backup, | |
2213 | btrfs_super_num_devices(info->super_copy)); | |
2214 | ||
2215 | /* | |
2216 | * if we don't copy this out to the super_copy, it won't get remembered | |
2217 | * for the next commit | |
2218 | */ | |
2219 | memcpy(&info->super_copy->super_roots, | |
2220 | &info->super_for_commit->super_roots, | |
2221 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2222 | } | |
2223 | ||
bd2336b2 NB |
2224 | /* |
2225 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2226 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2227 | * | |
2228 | * fs_info - filesystem whose backup roots need to be read | |
2229 | * priority - priority of backup root required | |
2230 | * | |
2231 | * Returns backup root index on success and -EINVAL otherwise. | |
2232 | */ | |
2233 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2234 | { | |
2235 | int backup_index = find_newest_super_backup(fs_info); | |
2236 | struct btrfs_super_block *super = fs_info->super_copy; | |
2237 | struct btrfs_root_backup *root_backup; | |
2238 | ||
2239 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2240 | if (priority == 0) | |
2241 | return backup_index; | |
2242 | ||
2243 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2244 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2245 | } else { | |
2246 | return -EINVAL; | |
2247 | } | |
2248 | ||
2249 | root_backup = super->super_roots + backup_index; | |
2250 | ||
2251 | btrfs_set_super_generation(super, | |
2252 | btrfs_backup_tree_root_gen(root_backup)); | |
2253 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2254 | btrfs_set_super_root_level(super, | |
2255 | btrfs_backup_tree_root_level(root_backup)); | |
2256 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2257 | ||
2258 | /* | |
2259 | * Fixme: the total bytes and num_devices need to match or we should | |
2260 | * need a fsck | |
2261 | */ | |
2262 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2263 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2264 | ||
2265 | return backup_index; | |
2266 | } | |
2267 | ||
7abadb64 LB |
2268 | /* helper to cleanup workers */ |
2269 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2270 | { | |
dc6e3209 | 2271 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2272 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
a31b4a43 | 2273 | btrfs_destroy_workqueue(fs_info->hipri_workers); |
5cdc7ad3 | 2274 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2275 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2276 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
385de0ef CH |
2277 | if (fs_info->rmw_workers) |
2278 | destroy_workqueue(fs_info->rmw_workers); | |
fccb5d86 QW |
2279 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2280 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2281 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2282 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2283 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2284 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2285 | if (fs_info->discard_ctl.discard_workers) |
2286 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2287 | /* |
2288 | * Now that all other work queues are destroyed, we can safely destroy | |
2289 | * the queues used for metadata I/O, since tasks from those other work | |
2290 | * queues can do metadata I/O operations. | |
2291 | */ | |
2292 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2293 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2294 | } |
2295 | ||
2e9f5954 R |
2296 | static void free_root_extent_buffers(struct btrfs_root *root) |
2297 | { | |
2298 | if (root) { | |
2299 | free_extent_buffer(root->node); | |
2300 | free_extent_buffer(root->commit_root); | |
2301 | root->node = NULL; | |
2302 | root->commit_root = NULL; | |
2303 | } | |
2304 | } | |
2305 | ||
abed4aaa JB |
2306 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2307 | { | |
2308 | struct btrfs_root *root, *tmp; | |
2309 | ||
2310 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2311 | &fs_info->global_root_tree, | |
2312 | rb_node) | |
2313 | free_root_extent_buffers(root); | |
2314 | } | |
2315 | ||
af31f5e5 | 2316 | /* helper to cleanup tree roots */ |
4273eaff | 2317 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2318 | { |
2e9f5954 | 2319 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2320 | |
abed4aaa | 2321 | free_global_root_pointers(info); |
2e9f5954 | 2322 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2323 | free_root_extent_buffers(info->quota_root); |
2324 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2325 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2326 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 2327 | free_root_extent_buffers(info->block_group_root); |
4273eaff | 2328 | if (free_chunk_root) |
2e9f5954 | 2329 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2330 | } |
2331 | ||
8c38938c JB |
2332 | void btrfs_put_root(struct btrfs_root *root) |
2333 | { | |
2334 | if (!root) | |
2335 | return; | |
2336 | ||
2337 | if (refcount_dec_and_test(&root->refs)) { | |
2338 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2339 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2340 | if (root->anon_dev) |
2341 | free_anon_bdev(root->anon_dev); | |
2342 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2343 | free_root_extent_buffers(root); |
8c38938c | 2344 | #ifdef CONFIG_BTRFS_DEBUG |
48b36a60 | 2345 | spin_lock(&root->fs_info->fs_roots_lock); |
8c38938c | 2346 | list_del_init(&root->leak_list); |
48b36a60 | 2347 | spin_unlock(&root->fs_info->fs_roots_lock); |
8c38938c JB |
2348 | #endif |
2349 | kfree(root); | |
2350 | } | |
2351 | } | |
2352 | ||
faa2dbf0 | 2353 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 2354 | { |
48b36a60 GN |
2355 | struct btrfs_root *root; |
2356 | unsigned long index = 0; | |
171f6537 JB |
2357 | |
2358 | while (!list_empty(&fs_info->dead_roots)) { | |
48b36a60 GN |
2359 | root = list_entry(fs_info->dead_roots.next, |
2360 | struct btrfs_root, root_list); | |
2361 | list_del(&root->root_list); | |
171f6537 | 2362 | |
48b36a60 GN |
2363 | if (test_bit(BTRFS_ROOT_REGISTERED, &root->state)) |
2364 | btrfs_drop_and_free_fs_root(fs_info, root); | |
2365 | btrfs_put_root(root); | |
171f6537 JB |
2366 | } |
2367 | ||
48b36a60 GN |
2368 | xa_for_each(&fs_info->fs_roots, index, root) { |
2369 | btrfs_drop_and_free_fs_root(fs_info, root); | |
171f6537 JB |
2370 | } |
2371 | } | |
af31f5e5 | 2372 | |
638aa7ed ES |
2373 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2374 | { | |
2375 | mutex_init(&fs_info->scrub_lock); | |
2376 | atomic_set(&fs_info->scrubs_running, 0); | |
2377 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2378 | atomic_set(&fs_info->scrubs_paused, 0); | |
2379 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2380 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2381 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2382 | } |
2383 | ||
779a65a4 ES |
2384 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2385 | { | |
2386 | spin_lock_init(&fs_info->balance_lock); | |
2387 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2388 | atomic_set(&fs_info->balance_pause_req, 0); |
2389 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2390 | fs_info->balance_ctl = NULL; | |
2391 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2392 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2393 | } |
2394 | ||
6bccf3ab | 2395 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2396 | { |
2ff7e61e JM |
2397 | struct inode *inode = fs_info->btree_inode; |
2398 | ||
2399 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2400 | set_nlink(inode, 1); | |
f37938e0 ES |
2401 | /* |
2402 | * we set the i_size on the btree inode to the max possible int. | |
2403 | * the real end of the address space is determined by all of | |
2404 | * the devices in the system | |
2405 | */ | |
2ff7e61e JM |
2406 | inode->i_size = OFFSET_MAX; |
2407 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2408 | |
2ff7e61e | 2409 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2410 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2c53a14d | 2411 | IO_TREE_BTREE_INODE_IO, inode); |
7b439738 | 2412 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2413 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2414 | |
5c8fd99f | 2415 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2416 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2417 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2418 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2419 | } |
2420 | ||
ad618368 ES |
2421 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2422 | { | |
ad618368 | 2423 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2424 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2425 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2426 | } |
2427 | ||
f9e92e40 ES |
2428 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2429 | { | |
2430 | spin_lock_init(&fs_info->qgroup_lock); | |
2431 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2432 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2433 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2434 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2435 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2436 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2437 | mutex_init(&fs_info->qgroup_rescan_lock); |
2438 | } | |
2439 | ||
d21deec5 | 2440 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2441 | { |
f7b885be | 2442 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2443 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2444 | |
2445 | fs_info->workers = | |
a31b4a43 CH |
2446 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2447 | fs_info->hipri_workers = | |
2448 | btrfs_alloc_workqueue(fs_info, "worker-high", | |
cb001095 | 2449 | flags | WQ_HIGHPRI, max_active, 16); |
2a458198 ES |
2450 | |
2451 | fs_info->delalloc_workers = | |
cb001095 JM |
2452 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2453 | flags, max_active, 2); | |
2a458198 ES |
2454 | |
2455 | fs_info->flush_workers = | |
cb001095 JM |
2456 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2457 | flags, max_active, 0); | |
2a458198 ES |
2458 | |
2459 | fs_info->caching_workers = | |
cb001095 | 2460 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2461 | |
2a458198 | 2462 | fs_info->fixup_workers = |
cb001095 | 2463 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2464 | |
2465 | /* | |
2466 | * endios are largely parallel and should have a very | |
2467 | * low idle thresh | |
2468 | */ | |
2469 | fs_info->endio_workers = | |
cb001095 | 2470 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2471 | fs_info->endio_meta_workers = |
cb001095 JM |
2472 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2473 | max_active, 4); | |
2a458198 | 2474 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2475 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2476 | max_active, 2); | |
2a458198 | 2477 | fs_info->endio_raid56_workers = |
cb001095 JM |
2478 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2479 | max_active, 4); | |
385de0ef | 2480 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2481 | fs_info->endio_write_workers = |
cb001095 JM |
2482 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2483 | max_active, 2); | |
2a458198 | 2484 | fs_info->endio_freespace_worker = |
cb001095 JM |
2485 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2486 | max_active, 0); | |
2a458198 | 2487 | fs_info->delayed_workers = |
cb001095 JM |
2488 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2489 | max_active, 0); | |
2a458198 | 2490 | fs_info->qgroup_rescan_workers = |
cb001095 | 2491 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2492 | fs_info->discard_ctl.discard_workers = |
2493 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 | 2494 | |
a31b4a43 CH |
2495 | if (!(fs_info->workers && fs_info->hipri_workers && |
2496 | fs_info->delalloc_workers && fs_info->flush_workers && | |
2a458198 ES |
2497 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2498 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2499 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2500 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
f26c9238 QW |
2501 | fs_info->caching_workers && fs_info->fixup_workers && |
2502 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2503 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2504 | return -ENOMEM; |
2505 | } | |
2506 | ||
2507 | return 0; | |
2508 | } | |
2509 | ||
6d97c6e3 JT |
2510 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2511 | { | |
2512 | struct crypto_shash *csum_shash; | |
b4e967be | 2513 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2514 | |
b4e967be | 2515 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2516 | |
2517 | if (IS_ERR(csum_shash)) { | |
2518 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2519 | csum_driver); |
6d97c6e3 JT |
2520 | return PTR_ERR(csum_shash); |
2521 | } | |
2522 | ||
2523 | fs_info->csum_shash = csum_shash; | |
2524 | ||
2525 | return 0; | |
2526 | } | |
2527 | ||
63443bf5 ES |
2528 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2529 | struct btrfs_fs_devices *fs_devices) | |
2530 | { | |
2531 | int ret; | |
63443bf5 ES |
2532 | struct btrfs_root *log_tree_root; |
2533 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2534 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2535 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2536 | |
2537 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2538 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2539 | return -EIO; |
2540 | } | |
2541 | ||
96dfcb46 JB |
2542 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2543 | GFP_KERNEL); | |
63443bf5 ES |
2544 | if (!log_tree_root) |
2545 | return -ENOMEM; | |
2546 | ||
2ff7e61e | 2547 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2548 | BTRFS_TREE_LOG_OBJECTID, |
2549 | fs_info->generation + 1, level, | |
2550 | NULL); | |
64c043de | 2551 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2552 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2553 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2554 | log_tree_root->node = NULL; |
00246528 | 2555 | btrfs_put_root(log_tree_root); |
0eeff236 | 2556 | return ret; |
4eb150d6 QW |
2557 | } |
2558 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2559 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2560 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2561 | return -EIO; |
2562 | } | |
4eb150d6 | 2563 | |
63443bf5 ES |
2564 | /* returns with log_tree_root freed on success */ |
2565 | ret = btrfs_recover_log_trees(log_tree_root); | |
2566 | if (ret) { | |
0b246afa JM |
2567 | btrfs_handle_fs_error(fs_info, ret, |
2568 | "Failed to recover log tree"); | |
00246528 | 2569 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2570 | return ret; |
2571 | } | |
2572 | ||
bc98a42c | 2573 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2574 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2575 | if (ret) |
2576 | return ret; | |
2577 | } | |
2578 | ||
2579 | return 0; | |
2580 | } | |
2581 | ||
abed4aaa JB |
2582 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2583 | struct btrfs_path *path, u64 objectid, | |
2584 | const char *name) | |
2585 | { | |
2586 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2587 | struct btrfs_root *root; | |
f7238e50 | 2588 | u64 max_global_id = 0; |
abed4aaa JB |
2589 | int ret; |
2590 | struct btrfs_key key = { | |
2591 | .objectid = objectid, | |
2592 | .type = BTRFS_ROOT_ITEM_KEY, | |
2593 | .offset = 0, | |
2594 | }; | |
2595 | bool found = false; | |
2596 | ||
2597 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2598 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2599 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2600 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2601 | return 0; | |
2602 | } | |
2603 | ||
2604 | while (1) { | |
2605 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2606 | if (ret < 0) | |
2607 | break; | |
2608 | ||
2609 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2610 | ret = btrfs_next_leaf(tree_root, path); | |
2611 | if (ret) { | |
2612 | if (ret > 0) | |
2613 | ret = 0; | |
2614 | break; | |
2615 | } | |
2616 | } | |
2617 | ret = 0; | |
2618 | ||
2619 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2620 | if (key.objectid != objectid) | |
2621 | break; | |
2622 | btrfs_release_path(path); | |
2623 | ||
f7238e50 JB |
2624 | /* |
2625 | * Just worry about this for extent tree, it'll be the same for | |
2626 | * everybody. | |
2627 | */ | |
2628 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2629 | max_global_id = max(max_global_id, key.offset); | |
2630 | ||
abed4aaa JB |
2631 | found = true; |
2632 | root = read_tree_root_path(tree_root, path, &key); | |
2633 | if (IS_ERR(root)) { | |
2634 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2635 | ret = PTR_ERR(root); | |
2636 | break; | |
2637 | } | |
2638 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2639 | ret = btrfs_global_root_insert(root); | |
2640 | if (ret) { | |
2641 | btrfs_put_root(root); | |
2642 | break; | |
2643 | } | |
2644 | key.offset++; | |
2645 | } | |
2646 | btrfs_release_path(path); | |
2647 | ||
f7238e50 JB |
2648 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2649 | fs_info->nr_global_roots = max_global_id + 1; | |
2650 | ||
abed4aaa JB |
2651 | if (!found || ret) { |
2652 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2653 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2654 | ||
2655 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2656 | ret = ret ? ret : -ENOENT; | |
2657 | else | |
2658 | ret = 0; | |
2659 | btrfs_err(fs_info, "failed to load root %s", name); | |
2660 | } | |
2661 | return ret; | |
2662 | } | |
2663 | ||
2664 | static int load_global_roots(struct btrfs_root *tree_root) | |
2665 | { | |
2666 | struct btrfs_path *path; | |
2667 | int ret = 0; | |
2668 | ||
2669 | path = btrfs_alloc_path(); | |
2670 | if (!path) | |
2671 | return -ENOMEM; | |
2672 | ||
2673 | ret = load_global_roots_objectid(tree_root, path, | |
2674 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2675 | if (ret) | |
2676 | goto out; | |
2677 | ret = load_global_roots_objectid(tree_root, path, | |
2678 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2679 | if (ret) | |
2680 | goto out; | |
2681 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2682 | goto out; | |
2683 | ret = load_global_roots_objectid(tree_root, path, | |
2684 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2685 | "free space"); | |
2686 | out: | |
2687 | btrfs_free_path(path); | |
2688 | return ret; | |
2689 | } | |
2690 | ||
6bccf3ab | 2691 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2692 | { |
6bccf3ab | 2693 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2694 | struct btrfs_root *root; |
4bbcaa64 ES |
2695 | struct btrfs_key location; |
2696 | int ret; | |
2697 | ||
6bccf3ab JM |
2698 | BUG_ON(!fs_info->tree_root); |
2699 | ||
abed4aaa JB |
2700 | ret = load_global_roots(tree_root); |
2701 | if (ret) | |
2702 | return ret; | |
2703 | ||
2704 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
4bbcaa64 ES |
2705 | location.type = BTRFS_ROOT_ITEM_KEY; |
2706 | location.offset = 0; | |
2707 | ||
a4f3d2c4 | 2708 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2709 | if (IS_ERR(root)) { |
42437a63 JB |
2710 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2711 | ret = PTR_ERR(root); | |
2712 | goto out; | |
2713 | } | |
2714 | } else { | |
2715 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2716 | fs_info->dev_root = root; | |
f50f4353 | 2717 | } |
820a49da JB |
2718 | /* Initialize fs_info for all devices in any case */ |
2719 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2720 | |
aeb935a4 QW |
2721 | /* |
2722 | * This tree can share blocks with some other fs tree during relocation | |
2723 | * and we need a proper setup by btrfs_get_fs_root | |
2724 | */ | |
56e9357a DS |
2725 | root = btrfs_get_fs_root(tree_root->fs_info, |
2726 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2727 | if (IS_ERR(root)) { |
42437a63 JB |
2728 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2729 | ret = PTR_ERR(root); | |
2730 | goto out; | |
2731 | } | |
2732 | } else { | |
2733 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2734 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2735 | } |
aeb935a4 | 2736 | |
4bbcaa64 | 2737 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2738 | root = btrfs_read_tree_root(tree_root, &location); |
2739 | if (!IS_ERR(root)) { | |
2740 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2741 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2742 | fs_info->quota_root = root; |
4bbcaa64 ES |
2743 | } |
2744 | ||
2745 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2746 | root = btrfs_read_tree_root(tree_root, &location); |
2747 | if (IS_ERR(root)) { | |
42437a63 JB |
2748 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2749 | ret = PTR_ERR(root); | |
2750 | if (ret != -ENOENT) | |
2751 | goto out; | |
2752 | } | |
4bbcaa64 | 2753 | } else { |
a4f3d2c4 DS |
2754 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2755 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2756 | } |
2757 | ||
2758 | return 0; | |
f50f4353 LB |
2759 | out: |
2760 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2761 | location.objectid, ret); | |
2762 | return ret; | |
4bbcaa64 ES |
2763 | } |
2764 | ||
069ec957 QW |
2765 | /* |
2766 | * Real super block validation | |
2767 | * NOTE: super csum type and incompat features will not be checked here. | |
2768 | * | |
2769 | * @sb: super block to check | |
2770 | * @mirror_num: the super block number to check its bytenr: | |
2771 | * 0 the primary (1st) sb | |
2772 | * 1, 2 2nd and 3rd backup copy | |
2773 | * -1 skip bytenr check | |
2774 | */ | |
2775 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2776 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2777 | { |
21a852b0 QW |
2778 | u64 nodesize = btrfs_super_nodesize(sb); |
2779 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2780 | int ret = 0; | |
2781 | ||
2782 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2783 | btrfs_err(fs_info, "no valid FS found"); | |
2784 | ret = -EINVAL; | |
2785 | } | |
2786 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2787 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2788 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2789 | ret = -EINVAL; | |
2790 | } | |
2791 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2792 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2793 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2794 | ret = -EINVAL; | |
2795 | } | |
2796 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2797 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2798 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2799 | ret = -EINVAL; | |
2800 | } | |
2801 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2802 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2803 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2804 | ret = -EINVAL; | |
2805 | } | |
2806 | ||
2807 | /* | |
2808 | * Check sectorsize and nodesize first, other check will need it. | |
2809 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2810 | */ | |
2811 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2812 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2813 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2814 | ret = -EINVAL; | |
2815 | } | |
0bb3eb3e QW |
2816 | |
2817 | /* | |
1a42daab QW |
2818 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2819 | * | |
2820 | * We can support 16K sectorsize with 64K page size without problem, | |
2821 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2822 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2823 | * beginning. | |
0bb3eb3e | 2824 | */ |
1a42daab | 2825 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2826 | btrfs_err(fs_info, |
0bb3eb3e | 2827 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2828 | sectorsize, PAGE_SIZE); |
2829 | ret = -EINVAL; | |
2830 | } | |
0bb3eb3e | 2831 | |
21a852b0 QW |
2832 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2833 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2834 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2835 | ret = -EINVAL; | |
2836 | } | |
2837 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2838 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2839 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2840 | ret = -EINVAL; | |
2841 | } | |
2842 | ||
2843 | /* Root alignment check */ | |
2844 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2845 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2846 | btrfs_super_root(sb)); | |
2847 | ret = -EINVAL; | |
2848 | } | |
2849 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2850 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2851 | btrfs_super_chunk_root(sb)); | |
2852 | ret = -EINVAL; | |
2853 | } | |
2854 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2855 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2856 | btrfs_super_log_root(sb)); | |
2857 | ret = -EINVAL; | |
2858 | } | |
2859 | ||
aefd7f70 NB |
2860 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2861 | BTRFS_FSID_SIZE)) { | |
2862 | btrfs_err(fs_info, | |
2863 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2864 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2865 | ret = -EINVAL; | |
2866 | } | |
2867 | ||
2868 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2869 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2870 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2871 | btrfs_err(fs_info, | |
2872 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2873 | fs_info->super_copy->metadata_uuid, | |
2874 | fs_info->fs_devices->metadata_uuid); | |
2875 | ret = -EINVAL; | |
2876 | } | |
2877 | ||
de37aa51 | 2878 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2879 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2880 | btrfs_err(fs_info, |
7239ff4b | 2881 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2882 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2883 | ret = -EINVAL; |
2884 | } | |
2885 | ||
2886 | /* | |
2887 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2888 | * done later | |
2889 | */ | |
2890 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2891 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2892 | btrfs_super_bytes_used(sb)); | |
2893 | ret = -EINVAL; | |
2894 | } | |
2895 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2896 | btrfs_err(fs_info, "invalid stripesize %u", | |
2897 | btrfs_super_stripesize(sb)); | |
2898 | ret = -EINVAL; | |
2899 | } | |
2900 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2901 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2902 | btrfs_super_num_devices(sb)); | |
2903 | if (btrfs_super_num_devices(sb) == 0) { | |
2904 | btrfs_err(fs_info, "number of devices is 0"); | |
2905 | ret = -EINVAL; | |
2906 | } | |
2907 | ||
069ec957 QW |
2908 | if (mirror_num >= 0 && |
2909 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2910 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2911 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2912 | ret = -EINVAL; | |
2913 | } | |
2914 | ||
2915 | /* | |
2916 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2917 | * and one chunk | |
2918 | */ | |
2919 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2920 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2921 | btrfs_super_sys_array_size(sb), | |
2922 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2923 | ret = -EINVAL; | |
2924 | } | |
2925 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2926 | + sizeof(struct btrfs_chunk)) { | |
2927 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2928 | btrfs_super_sys_array_size(sb), | |
2929 | sizeof(struct btrfs_disk_key) | |
2930 | + sizeof(struct btrfs_chunk)); | |
2931 | ret = -EINVAL; | |
2932 | } | |
2933 | ||
2934 | /* | |
2935 | * The generation is a global counter, we'll trust it more than the others | |
2936 | * but it's still possible that it's the one that's wrong. | |
2937 | */ | |
2938 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2939 | btrfs_warn(fs_info, | |
2940 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2941 | btrfs_super_generation(sb), | |
2942 | btrfs_super_chunk_root_generation(sb)); | |
2943 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2944 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2945 | btrfs_warn(fs_info, | |
2946 | "suspicious: generation < cache_generation: %llu < %llu", | |
2947 | btrfs_super_generation(sb), | |
2948 | btrfs_super_cache_generation(sb)); | |
2949 | ||
2950 | return ret; | |
2951 | } | |
2952 | ||
069ec957 QW |
2953 | /* |
2954 | * Validation of super block at mount time. | |
2955 | * Some checks already done early at mount time, like csum type and incompat | |
2956 | * flags will be skipped. | |
2957 | */ | |
2958 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2959 | { | |
2960 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2961 | } | |
2962 | ||
75cb857d QW |
2963 | /* |
2964 | * Validation of super block at write time. | |
2965 | * Some checks like bytenr check will be skipped as their values will be | |
2966 | * overwritten soon. | |
2967 | * Extra checks like csum type and incompat flags will be done here. | |
2968 | */ | |
2969 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2970 | struct btrfs_super_block *sb) | |
2971 | { | |
2972 | int ret; | |
2973 | ||
2974 | ret = validate_super(fs_info, sb, -1); | |
2975 | if (ret < 0) | |
2976 | goto out; | |
e7e16f48 | 2977 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2978 | ret = -EUCLEAN; |
2979 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2980 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2981 | goto out; | |
2982 | } | |
2983 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2984 | ret = -EUCLEAN; | |
2985 | btrfs_err(fs_info, | |
2986 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2987 | btrfs_super_incompat_flags(sb), | |
2988 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2989 | goto out; | |
2990 | } | |
2991 | out: | |
2992 | if (ret < 0) | |
2993 | btrfs_err(fs_info, | |
2994 | "super block corruption detected before writing it to disk"); | |
2995 | return ret; | |
2996 | } | |
2997 | ||
bd676446 JB |
2998 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2999 | { | |
3000 | int ret = 0; | |
3001 | ||
3002 | root->node = read_tree_block(root->fs_info, bytenr, | |
3003 | root->root_key.objectid, gen, level, NULL); | |
3004 | if (IS_ERR(root->node)) { | |
3005 | ret = PTR_ERR(root->node); | |
3006 | root->node = NULL; | |
4eb150d6 QW |
3007 | return ret; |
3008 | } | |
3009 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
3010 | free_extent_buffer(root->node); |
3011 | root->node = NULL; | |
4eb150d6 | 3012 | return -EIO; |
bd676446 JB |
3013 | } |
3014 | ||
bd676446 JB |
3015 | btrfs_set_root_node(&root->root_item, root->node); |
3016 | root->commit_root = btrfs_root_node(root); | |
3017 | btrfs_set_root_refs(&root->root_item, 1); | |
3018 | return ret; | |
3019 | } | |
3020 | ||
3021 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
3022 | { | |
3023 | struct btrfs_super_block *sb = fs_info->super_copy; | |
3024 | u64 gen, bytenr; | |
3025 | int level, ret; | |
3026 | ||
3027 | bytenr = btrfs_super_root(sb); | |
3028 | gen = btrfs_super_generation(sb); | |
3029 | level = btrfs_super_root_level(sb); | |
3030 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 3031 | if (ret) { |
bd676446 | 3032 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
3033 | return ret; |
3034 | } | |
3035 | ||
3036 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
3037 | return 0; | |
3038 | ||
3039 | bytenr = btrfs_super_block_group_root(sb); | |
3040 | gen = btrfs_super_block_group_root_generation(sb); | |
3041 | level = btrfs_super_block_group_root_level(sb); | |
3042 | ret = load_super_root(fs_info->block_group_root, bytenr, gen, level); | |
3043 | if (ret) | |
3044 | btrfs_warn(fs_info, "couldn't read block group root"); | |
bd676446 JB |
3045 | return ret; |
3046 | } | |
3047 | ||
6ef108dd | 3048 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 3049 | { |
6ef108dd | 3050 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
3051 | struct btrfs_super_block *sb = fs_info->super_copy; |
3052 | struct btrfs_root *tree_root = fs_info->tree_root; | |
3053 | bool handle_error = false; | |
3054 | int ret = 0; | |
3055 | int i; | |
3056 | ||
9c54e80d JB |
3057 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { |
3058 | struct btrfs_root *root; | |
3059 | ||
3060 | root = btrfs_alloc_root(fs_info, BTRFS_BLOCK_GROUP_TREE_OBJECTID, | |
3061 | GFP_KERNEL); | |
3062 | if (!root) | |
3063 | return -ENOMEM; | |
3064 | fs_info->block_group_root = root; | |
3065 | } | |
b8522a1e | 3066 | |
b8522a1e | 3067 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { |
b8522a1e NB |
3068 | if (handle_error) { |
3069 | if (!IS_ERR(tree_root->node)) | |
3070 | free_extent_buffer(tree_root->node); | |
3071 | tree_root->node = NULL; | |
3072 | ||
3073 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
3074 | break; | |
3075 | ||
3076 | free_root_pointers(fs_info, 0); | |
3077 | ||
3078 | /* | |
3079 | * Don't use the log in recovery mode, it won't be | |
3080 | * valid | |
3081 | */ | |
3082 | btrfs_set_super_log_root(sb, 0); | |
3083 | ||
3084 | /* We can't trust the free space cache either */ | |
3085 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3086 | ||
3087 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 3088 | backup_index = ret; |
b8522a1e NB |
3089 | if (ret < 0) |
3090 | return ret; | |
3091 | } | |
b8522a1e | 3092 | |
bd676446 JB |
3093 | ret = load_important_roots(fs_info); |
3094 | if (ret) { | |
217f5004 | 3095 | handle_error = true; |
b8522a1e NB |
3096 | continue; |
3097 | } | |
3098 | ||
336a0d8d NB |
3099 | /* |
3100 | * No need to hold btrfs_root::objectid_mutex since the fs | |
3101 | * hasn't been fully initialised and we are the only user | |
3102 | */ | |
453e4873 | 3103 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 3104 | if (ret < 0) { |
b8522a1e NB |
3105 | handle_error = true; |
3106 | continue; | |
3107 | } | |
3108 | ||
6b8fad57 | 3109 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
3110 | |
3111 | ret = btrfs_read_roots(fs_info); | |
3112 | if (ret < 0) { | |
3113 | handle_error = true; | |
3114 | continue; | |
3115 | } | |
3116 | ||
3117 | /* All successful */ | |
bd676446 JB |
3118 | fs_info->generation = btrfs_header_generation(tree_root->node); |
3119 | fs_info->last_trans_committed = fs_info->generation; | |
d96b3424 | 3120 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
3121 | |
3122 | /* Always begin writing backup roots after the one being used */ | |
3123 | if (backup_index < 0) { | |
3124 | fs_info->backup_root_index = 0; | |
3125 | } else { | |
3126 | fs_info->backup_root_index = backup_index + 1; | |
3127 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
3128 | } | |
b8522a1e NB |
3129 | break; |
3130 | } | |
3131 | ||
3132 | return ret; | |
3133 | } | |
3134 | ||
8260edba | 3135 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 3136 | { |
48b36a60 | 3137 | xa_init_flags(&fs_info->fs_roots, GFP_ATOMIC); |
8ee92268 | 3138 | xa_init_flags(&fs_info->extent_buffers, GFP_ATOMIC); |
8fd17795 | 3139 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 3140 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 3141 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 3142 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 3143 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 3144 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 3145 | spin_lock_init(&fs_info->trans_lock); |
48b36a60 | 3146 | spin_lock_init(&fs_info->fs_roots_lock); |
24bbcf04 | 3147 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 3148 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 3149 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 3150 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 3151 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 3152 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 3153 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 3154 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 3155 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 3156 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 3157 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 3158 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 3159 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 3160 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 3161 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 3162 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 3163 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 3164 | |
0b86a832 | 3165 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3166 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3167 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3168 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3169 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3170 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3171 | #ifdef CONFIG_BTRFS_DEBUG |
3172 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3173 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3174 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3175 | #endif |
c8bf1b67 | 3176 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3177 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3178 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3179 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3180 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3181 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3182 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3183 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3184 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3185 | BTRFS_BLOCK_RSV_DELREFS); | |
3186 | ||
771ed689 | 3187 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3188 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3189 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3190 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3191 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3192 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3193 | fs_info->metadata_ratio = 0; |
4cb5300b | 3194 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3195 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3196 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3197 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3198 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3199 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3200 | |
b34b086c CM |
3201 | fs_info->thread_pool_size = min_t(unsigned long, |
3202 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3203 | |
199c2a9c MX |
3204 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3205 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3206 | |
638aa7ed | 3207 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3208 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3209 | fs_info->check_integrity_print_mask = 0; | |
3210 | #endif | |
779a65a4 | 3211 | btrfs_init_balance(fs_info); |
57056740 | 3212 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3213 | |
16b0c258 | 3214 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 3215 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 3216 | |
fe119a6e NB |
3217 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
3218 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
39279cc3 | 3219 | |
5a3f23d5 | 3220 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3221 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3222 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3223 | mutex_init(&fs_info->transaction_kthread_mutex); |
3224 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3225 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3226 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3227 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3228 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3229 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3230 | |
ad618368 | 3231 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3232 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3233 | btrfs_discard_init(fs_info); |
416ac51d | 3234 | |
fa9c0d79 CM |
3235 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3236 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3237 | ||
e6dcd2dc | 3238 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3239 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3240 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3241 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3242 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3243 | |
da17066c JM |
3244 | /* Usable values until the real ones are cached from the superblock */ |
3245 | fs_info->nodesize = 4096; | |
3246 | fs_info->sectorsize = 4096; | |
ab108d99 | 3247 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3248 | fs_info->stripesize = 4096; |
3249 | ||
eede2bf3 OS |
3250 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3251 | fs_info->swapfile_pins = RB_ROOT; | |
3252 | ||
18bb8bbf JT |
3253 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3254 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3255 | } |
3256 | ||
3257 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3258 | { | |
3259 | int ret; | |
3260 | ||
3261 | fs_info->sb = sb; | |
3262 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3263 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3264 | |
5deb17e1 | 3265 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3266 | if (ret) |
c75e8394 | 3267 | return ret; |
ae18c37a JB |
3268 | |
3269 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3270 | if (ret) | |
c75e8394 | 3271 | return ret; |
ae18c37a JB |
3272 | |
3273 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3274 | (1 + ilog2(nr_cpu_ids)); | |
3275 | ||
3276 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3277 | if (ret) | |
c75e8394 | 3278 | return ret; |
ae18c37a JB |
3279 | |
3280 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3281 | GFP_KERNEL); | |
3282 | if (ret) | |
c75e8394 | 3283 | return ret; |
ae18c37a JB |
3284 | |
3285 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3286 | GFP_KERNEL); | |
c75e8394 JB |
3287 | if (!fs_info->delayed_root) |
3288 | return -ENOMEM; | |
ae18c37a JB |
3289 | btrfs_init_delayed_root(fs_info->delayed_root); |
3290 | ||
a0a1db70 FM |
3291 | if (sb_rdonly(sb)) |
3292 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3293 | ||
c75e8394 | 3294 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3295 | } |
3296 | ||
97f4dd09 NB |
3297 | static int btrfs_uuid_rescan_kthread(void *data) |
3298 | { | |
0d031dc4 | 3299 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
3300 | int ret; |
3301 | ||
3302 | /* | |
3303 | * 1st step is to iterate through the existing UUID tree and | |
3304 | * to delete all entries that contain outdated data. | |
3305 | * 2nd step is to add all missing entries to the UUID tree. | |
3306 | */ | |
3307 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3308 | if (ret < 0) { | |
c94bec2c JB |
3309 | if (ret != -EINTR) |
3310 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3311 | ret); | |
97f4dd09 NB |
3312 | up(&fs_info->uuid_tree_rescan_sem); |
3313 | return ret; | |
3314 | } | |
3315 | return btrfs_uuid_scan_kthread(data); | |
3316 | } | |
3317 | ||
3318 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3319 | { | |
3320 | struct task_struct *task; | |
3321 | ||
3322 | down(&fs_info->uuid_tree_rescan_sem); | |
3323 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3324 | if (IS_ERR(task)) { | |
3325 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3326 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3327 | up(&fs_info->uuid_tree_rescan_sem); | |
3328 | return PTR_ERR(task); | |
3329 | } | |
3330 | ||
3331 | return 0; | |
3332 | } | |
3333 | ||
8cd29088 BB |
3334 | /* |
3335 | * Some options only have meaning at mount time and shouldn't persist across | |
3336 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3337 | * code paths. | |
3338 | */ | |
3339 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3340 | { | |
3341 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3342 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3343 | } |
3344 | ||
44c0ca21 BB |
3345 | /* |
3346 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3347 | * and btrfs_remount when remounting from read-only to read-write. | |
3348 | */ | |
3349 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3350 | { | |
3351 | int ret; | |
94846229 | 3352 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3353 | bool clear_free_space_tree = false; |
3354 | ||
3355 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3356 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3357 | clear_free_space_tree = true; | |
3358 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3359 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3360 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3361 | clear_free_space_tree = true; | |
3362 | } | |
3363 | ||
3364 | if (clear_free_space_tree) { | |
3365 | btrfs_info(fs_info, "clearing free space tree"); | |
3366 | ret = btrfs_clear_free_space_tree(fs_info); | |
3367 | if (ret) { | |
3368 | btrfs_warn(fs_info, | |
3369 | "failed to clear free space tree: %d", ret); | |
3370 | goto out; | |
3371 | } | |
3372 | } | |
44c0ca21 | 3373 | |
8d488a8c FM |
3374 | /* |
3375 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3376 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
48b36a60 | 3377 | * them into the fs_info->fs_roots. This must be done before |
8d488a8c FM |
3378 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3379 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3380 | * item before the root's tree is deleted - this means that if we unmount | |
3381 | * or crash before the deletion completes, on the next mount we will not | |
3382 | * delete what remains of the tree because the orphan item does not | |
3383 | * exists anymore, which is what tells us we have a pending deletion. | |
3384 | */ | |
3385 | ret = btrfs_find_orphan_roots(fs_info); | |
3386 | if (ret) | |
3387 | goto out; | |
3388 | ||
44c0ca21 BB |
3389 | ret = btrfs_cleanup_fs_roots(fs_info); |
3390 | if (ret) | |
3391 | goto out; | |
3392 | ||
8f1c21d7 BB |
3393 | down_read(&fs_info->cleanup_work_sem); |
3394 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3395 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3396 | up_read(&fs_info->cleanup_work_sem); | |
3397 | goto out; | |
3398 | } | |
3399 | up_read(&fs_info->cleanup_work_sem); | |
3400 | ||
44c0ca21 | 3401 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3402 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3403 | mutex_unlock(&fs_info->cleaner_mutex); |
3404 | if (ret < 0) { | |
3405 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3406 | goto out; | |
3407 | } | |
3408 | ||
5011139a BB |
3409 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3410 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3411 | btrfs_info(fs_info, "creating free space tree"); | |
3412 | ret = btrfs_create_free_space_tree(fs_info); | |
3413 | if (ret) { | |
3414 | btrfs_warn(fs_info, | |
3415 | "failed to create free space tree: %d", ret); | |
3416 | goto out; | |
3417 | } | |
3418 | } | |
3419 | ||
94846229 BB |
3420 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3421 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3422 | if (ret) | |
3423 | goto out; | |
3424 | } | |
3425 | ||
44c0ca21 BB |
3426 | ret = btrfs_resume_balance_async(fs_info); |
3427 | if (ret) | |
3428 | goto out; | |
3429 | ||
3430 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3431 | if (ret) { | |
3432 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3433 | goto out; | |
3434 | } | |
3435 | ||
3436 | btrfs_qgroup_rescan_resume(fs_info); | |
3437 | ||
3438 | if (!fs_info->uuid_root) { | |
3439 | btrfs_info(fs_info, "creating UUID tree"); | |
3440 | ret = btrfs_create_uuid_tree(fs_info); | |
3441 | if (ret) { | |
3442 | btrfs_warn(fs_info, | |
3443 | "failed to create the UUID tree %d", ret); | |
3444 | goto out; | |
3445 | } | |
3446 | } | |
3447 | ||
3448 | out: | |
3449 | return ret; | |
3450 | } | |
3451 | ||
ae18c37a JB |
3452 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3453 | char *options) | |
3454 | { | |
3455 | u32 sectorsize; | |
3456 | u32 nodesize; | |
3457 | u32 stripesize; | |
3458 | u64 generation; | |
3459 | u64 features; | |
3460 | u16 csum_type; | |
ae18c37a JB |
3461 | struct btrfs_super_block *disk_super; |
3462 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3463 | struct btrfs_root *tree_root; | |
3464 | struct btrfs_root *chunk_root; | |
3465 | int ret; | |
3466 | int err = -EINVAL; | |
ae18c37a JB |
3467 | int level; |
3468 | ||
8260edba | 3469 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3470 | if (ret) { |
83c8266a | 3471 | err = ret; |
ae18c37a | 3472 | goto fail; |
53b381b3 DW |
3473 | } |
3474 | ||
ae18c37a JB |
3475 | /* These need to be init'ed before we start creating inodes and such. */ |
3476 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3477 | GFP_KERNEL); | |
3478 | fs_info->tree_root = tree_root; | |
3479 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3480 | GFP_KERNEL); | |
3481 | fs_info->chunk_root = chunk_root; | |
3482 | if (!tree_root || !chunk_root) { | |
3483 | err = -ENOMEM; | |
c75e8394 | 3484 | goto fail; |
ae18c37a JB |
3485 | } |
3486 | ||
3487 | fs_info->btree_inode = new_inode(sb); | |
3488 | if (!fs_info->btree_inode) { | |
3489 | err = -ENOMEM; | |
c75e8394 | 3490 | goto fail; |
ae18c37a JB |
3491 | } |
3492 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3493 | btrfs_init_btree_inode(fs_info); | |
3494 | ||
d24fa5c1 | 3495 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3496 | |
3497 | /* | |
3498 | * Read super block and check the signature bytes only | |
3499 | */ | |
d24fa5c1 | 3500 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3501 | if (IS_ERR(disk_super)) { |
3502 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3503 | goto fail_alloc; |
20b45077 | 3504 | } |
39279cc3 | 3505 | |
8dc3f22c | 3506 | /* |
260db43c | 3507 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3508 | * corrupted, we'll find out |
3509 | */ | |
8f32380d | 3510 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3511 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3512 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3513 | csum_type); |
8dc3f22c | 3514 | err = -EINVAL; |
8f32380d | 3515 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3516 | goto fail_alloc; |
3517 | } | |
3518 | ||
83c68bbc SY |
3519 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3520 | ||
6d97c6e3 JT |
3521 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3522 | if (ret) { | |
3523 | err = ret; | |
8f32380d | 3524 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3525 | goto fail_alloc; |
3526 | } | |
3527 | ||
1104a885 DS |
3528 | /* |
3529 | * We want to check superblock checksum, the type is stored inside. | |
3530 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3531 | */ | |
8f32380d | 3532 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3533 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3534 | err = -EINVAL; |
8f32380d | 3535 | btrfs_release_disk_super(disk_super); |
141386e1 | 3536 | goto fail_alloc; |
1104a885 DS |
3537 | } |
3538 | ||
3539 | /* | |
3540 | * super_copy is zeroed at allocation time and we never touch the | |
3541 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3542 | * the whole block of INFO_SIZE | |
3543 | */ | |
8f32380d JT |
3544 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3545 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3546 | |
fbc6feae NB |
3547 | disk_super = fs_info->super_copy; |
3548 | ||
0b86a832 | 3549 | |
fbc6feae NB |
3550 | features = btrfs_super_flags(disk_super); |
3551 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3552 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3553 | btrfs_set_super_flags(disk_super, features); | |
3554 | btrfs_info(fs_info, | |
3555 | "found metadata UUID change in progress flag, clearing"); | |
3556 | } | |
3557 | ||
3558 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3559 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3560 | |
069ec957 | 3561 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3562 | if (ret) { |
05135f59 | 3563 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3564 | err = -EINVAL; |
141386e1 | 3565 | goto fail_alloc; |
1104a885 DS |
3566 | } |
3567 | ||
0f7d52f4 | 3568 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3569 | goto fail_alloc; |
0f7d52f4 | 3570 | |
acce952b | 3571 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3572 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3573 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3574 | |
75e7cb7f LB |
3575 | /* |
3576 | * In the long term, we'll store the compression type in the super | |
3577 | * block, and it'll be used for per file compression control. | |
3578 | */ | |
3579 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3580 | ||
6f93e834 AJ |
3581 | /* |
3582 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3583 | * than the page size. | |
3584 | */ | |
3585 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { | |
3586 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
3587 | btrfs_info(fs_info, | |
3588 | "flagging fs with big metadata feature"); | |
3589 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3590 | } | |
3591 | ||
3592 | /* Set up fs_info before parsing mount options */ | |
3593 | nodesize = btrfs_super_nodesize(disk_super); | |
3594 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3595 | stripesize = sectorsize; | |
3596 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3597 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3598 | ||
3599 | fs_info->nodesize = nodesize; | |
3600 | fs_info->sectorsize = sectorsize; | |
3601 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3602 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3603 | fs_info->stripesize = stripesize; | |
3604 | ||
2ff7e61e | 3605 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3606 | if (ret) { |
3607 | err = ret; | |
141386e1 | 3608 | goto fail_alloc; |
2b82032c | 3609 | } |
dfe25020 | 3610 | |
f2b636e8 JB |
3611 | features = btrfs_super_incompat_flags(disk_super) & |
3612 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3613 | if (features) { | |
05135f59 | 3614 | btrfs_err(fs_info, |
d5321a0f | 3615 | "cannot mount because of unsupported optional features (0x%llx)", |
05135f59 | 3616 | features); |
f2b636e8 | 3617 | err = -EINVAL; |
141386e1 | 3618 | goto fail_alloc; |
f2b636e8 JB |
3619 | } |
3620 | ||
5d4f98a2 | 3621 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3622 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3623 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3624 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3625 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3626 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3627 | |
3173a18f | 3628 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3629 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3630 | |
bc3f116f CM |
3631 | /* |
3632 | * mixed block groups end up with duplicate but slightly offset | |
3633 | * extent buffers for the same range. It leads to corruptions | |
3634 | */ | |
3635 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3636 | (sectorsize != nodesize)) { |
05135f59 DS |
3637 | btrfs_err(fs_info, |
3638 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3639 | nodesize, sectorsize); | |
141386e1 | 3640 | goto fail_alloc; |
bc3f116f CM |
3641 | } |
3642 | ||
ceda0864 MX |
3643 | /* |
3644 | * Needn't use the lock because there is no other task which will | |
3645 | * update the flag. | |
3646 | */ | |
a6fa6fae | 3647 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3648 | |
f2b636e8 JB |
3649 | features = btrfs_super_compat_ro_flags(disk_super) & |
3650 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3651 | if (!sb_rdonly(sb) && features) { |
05135f59 | 3652 | btrfs_err(fs_info, |
d5321a0f | 3653 | "cannot mount read-write because of unsupported optional features (0x%llx)", |
c1c9ff7c | 3654 | features); |
f2b636e8 | 3655 | err = -EINVAL; |
141386e1 | 3656 | goto fail_alloc; |
f2b636e8 | 3657 | } |
61d92c32 | 3658 | |
8481dd80 QW |
3659 | if (sectorsize < PAGE_SIZE) { |
3660 | struct btrfs_subpage_info *subpage_info; | |
3661 | ||
9f73f1ae QW |
3662 | /* |
3663 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3664 | * going to be deprecated. | |
3665 | * | |
3666 | * Force to use v2 cache for subpage case. | |
3667 | */ | |
3668 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3669 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3670 | "forcing free space tree for sector size %u with page size %lu", | |
3671 | sectorsize, PAGE_SIZE); | |
3672 | ||
95ea0486 QW |
3673 | btrfs_warn(fs_info, |
3674 | "read-write for sector size %u with page size %lu is experimental", | |
3675 | sectorsize, PAGE_SIZE); | |
8481dd80 QW |
3676 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3677 | if (!subpage_info) | |
3678 | goto fail_alloc; | |
3679 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3680 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3681 | } |
0bb3eb3e | 3682 | |
d21deec5 | 3683 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3684 | if (ret) { |
3685 | err = ret; | |
0dc3b84a JB |
3686 | goto fail_sb_buffer; |
3687 | } | |
4543df7e | 3688 | |
9e11ceee JK |
3689 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3690 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3691 | |
a061fc8d CM |
3692 | sb->s_blocksize = sectorsize; |
3693 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3694 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3695 | |
925baedd | 3696 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3697 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3698 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3699 | if (ret) { |
05135f59 | 3700 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3701 | goto fail_sb_buffer; |
84eed90f | 3702 | } |
0b86a832 | 3703 | |
84234f3a | 3704 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3705 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3706 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3707 | generation, level); | |
3708 | if (ret) { | |
05135f59 | 3709 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3710 | goto fail_tree_roots; |
83121942 | 3711 | } |
0b86a832 | 3712 | |
e17cade2 | 3713 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3714 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3715 | BTRFS_UUID_SIZE); | |
e17cade2 | 3716 | |
5b4aacef | 3717 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3718 | if (ret) { |
05135f59 | 3719 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3720 | goto fail_tree_roots; |
2b82032c | 3721 | } |
0b86a832 | 3722 | |
8dabb742 | 3723 | /* |
bacce86a AJ |
3724 | * At this point we know all the devices that make this filesystem, |
3725 | * including the seed devices but we don't know yet if the replace | |
3726 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3727 | * filesystem but skip the replace target device which is checked |
bacce86a | 3728 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3729 | */ |
bacce86a | 3730 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3731 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3732 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3733 | goto fail_tree_roots; |
3734 | } | |
3735 | ||
b8522a1e | 3736 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3737 | if (ret) |
b8522a1e | 3738 | goto fail_tree_roots; |
8929ecfa | 3739 | |
73651042 NA |
3740 | /* |
3741 | * Get zone type information of zoned block devices. This will also | |
3742 | * handle emulation of a zoned filesystem if a regular device has the | |
3743 | * zoned incompat feature flag set. | |
3744 | */ | |
3745 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3746 | if (ret) { | |
3747 | btrfs_err(fs_info, | |
3748 | "zoned: failed to read device zone info: %d", | |
3749 | ret); | |
3750 | goto fail_block_groups; | |
3751 | } | |
3752 | ||
75ec1db8 JB |
3753 | /* |
3754 | * If we have a uuid root and we're not being told to rescan we need to | |
3755 | * check the generation here so we can set the | |
3756 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3757 | * transaction during a balance or the log replay without updating the | |
3758 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3759 | * even though it was perfectly fine. | |
3760 | */ | |
3761 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3762 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3763 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3764 | ||
cf90d884 QW |
3765 | ret = btrfs_verify_dev_extents(fs_info); |
3766 | if (ret) { | |
3767 | btrfs_err(fs_info, | |
3768 | "failed to verify dev extents against chunks: %d", | |
3769 | ret); | |
3770 | goto fail_block_groups; | |
3771 | } | |
68310a5e ID |
3772 | ret = btrfs_recover_balance(fs_info); |
3773 | if (ret) { | |
05135f59 | 3774 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3775 | goto fail_block_groups; |
3776 | } | |
3777 | ||
733f4fbb SB |
3778 | ret = btrfs_init_dev_stats(fs_info); |
3779 | if (ret) { | |
05135f59 | 3780 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3781 | goto fail_block_groups; |
3782 | } | |
3783 | ||
8dabb742 SB |
3784 | ret = btrfs_init_dev_replace(fs_info); |
3785 | if (ret) { | |
05135f59 | 3786 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3787 | goto fail_block_groups; |
3788 | } | |
3789 | ||
b70f5097 NA |
3790 | ret = btrfs_check_zoned_mode(fs_info); |
3791 | if (ret) { | |
3792 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3793 | ret); | |
3794 | goto fail_block_groups; | |
3795 | } | |
3796 | ||
c6761a9e | 3797 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3798 | if (ret) { |
05135f59 DS |
3799 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3800 | ret); | |
b7c35e81 AJ |
3801 | goto fail_block_groups; |
3802 | } | |
3803 | ||
96f3136e | 3804 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3805 | if (ret) { |
05135f59 | 3806 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3807 | goto fail_fsdev_sysfs; |
c59021f8 | 3808 | } |
3809 | ||
c59021f8 | 3810 | ret = btrfs_init_space_info(fs_info); |
3811 | if (ret) { | |
05135f59 | 3812 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3813 | goto fail_sysfs; |
c59021f8 | 3814 | } |
3815 | ||
5b4aacef | 3816 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3817 | if (ret) { |
05135f59 | 3818 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3819 | goto fail_sysfs; |
1b1d1f66 | 3820 | } |
4330e183 | 3821 | |
16beac87 NA |
3822 | btrfs_free_zone_cache(fs_info); |
3823 | ||
5c78a5e7 AJ |
3824 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3825 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3826 | btrfs_warn(fs_info, |
52042d8e | 3827 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3828 | goto fail_sysfs; |
292fd7fc | 3829 | } |
9078a3e1 | 3830 | |
33c44184 | 3831 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3832 | "btrfs-cleaner"); |
57506d50 | 3833 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3834 | goto fail_sysfs; |
a74a4b97 CM |
3835 | |
3836 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3837 | tree_root, | |
3838 | "btrfs-transaction"); | |
57506d50 | 3839 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3840 | goto fail_cleaner; |
a74a4b97 | 3841 | |
583b7231 | 3842 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3843 | !fs_info->fs_devices->rotating) { |
583b7231 | 3844 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3845 | } |
3846 | ||
572d9ab7 | 3847 | /* |
01327610 | 3848 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3849 | * not have to wait for transaction commit |
3850 | */ | |
3851 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3852 | |
21adbd5c | 3853 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3854 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3855 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3856 | btrfs_test_opt(fs_info, |
cbeaae4f | 3857 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3858 | fs_info->check_integrity_print_mask); |
3859 | if (ret) | |
05135f59 DS |
3860 | btrfs_warn(fs_info, |
3861 | "failed to initialize integrity check module: %d", | |
3862 | ret); | |
21adbd5c SB |
3863 | } |
3864 | #endif | |
bcef60f2 AJ |
3865 | ret = btrfs_read_qgroup_config(fs_info); |
3866 | if (ret) | |
3867 | goto fail_trans_kthread; | |
21adbd5c | 3868 | |
fd708b81 JB |
3869 | if (btrfs_build_ref_tree(fs_info)) |
3870 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3871 | ||
96da0919 QW |
3872 | /* do not make disk changes in broken FS or nologreplay is given */ |
3873 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3874 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3875 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3876 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3877 | if (ret) { |
63443bf5 | 3878 | err = ret; |
28c16cbb | 3879 | goto fail_qgroup; |
79787eaa | 3880 | } |
e02119d5 | 3881 | } |
1a40e23b | 3882 | |
56e9357a | 3883 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3884 | if (IS_ERR(fs_info->fs_root)) { |
3885 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3886 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3887 | fs_info->fs_root = NULL; |
bcef60f2 | 3888 | goto fail_qgroup; |
3140c9a3 | 3889 | } |
c289811c | 3890 | |
bc98a42c | 3891 | if (sb_rdonly(sb)) |
8cd29088 | 3892 | goto clear_oneshot; |
59641015 | 3893 | |
44c0ca21 | 3894 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3895 | if (ret) { |
6bccf3ab | 3896 | close_ctree(fs_info); |
2b6ba629 | 3897 | return ret; |
e3acc2a6 | 3898 | } |
b0643e59 | 3899 | btrfs_discard_resume(fs_info); |
b382a324 | 3900 | |
44c0ca21 BB |
3901 | if (fs_info->uuid_root && |
3902 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3903 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3904 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3905 | ret = btrfs_check_uuid_tree(fs_info); |
3906 | if (ret) { | |
05135f59 DS |
3907 | btrfs_warn(fs_info, |
3908 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3909 | close_ctree(fs_info); |
70f80175 SB |
3910 | return ret; |
3911 | } | |
f7a81ea4 | 3912 | } |
94846229 | 3913 | |
afcdd129 | 3914 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3915 | |
b4be6aef JB |
3916 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3917 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3918 | wake_up_process(fs_info->cleaner_kthread); | |
3919 | ||
8cd29088 BB |
3920 | clear_oneshot: |
3921 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3922 | return 0; |
39279cc3 | 3923 | |
bcef60f2 AJ |
3924 | fail_qgroup: |
3925 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3926 | fail_trans_kthread: |
3927 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3928 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3929 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3930 | fail_cleaner: |
a74a4b97 | 3931 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3932 | |
3933 | /* | |
3934 | * make sure we're done with the btree inode before we stop our | |
3935 | * kthreads | |
3936 | */ | |
3937 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3938 | |
2365dd3c | 3939 | fail_sysfs: |
6618a59b | 3940 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3941 | |
b7c35e81 AJ |
3942 | fail_fsdev_sysfs: |
3943 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3944 | ||
1b1d1f66 | 3945 | fail_block_groups: |
54067ae9 | 3946 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3947 | |
3948 | fail_tree_roots: | |
9e3aa805 JB |
3949 | if (fs_info->data_reloc_root) |
3950 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3951 | free_root_pointers(fs_info, true); |
2b8195bb | 3952 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3953 | |
39279cc3 | 3954 | fail_sb_buffer: |
7abadb64 | 3955 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3956 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3957 | fail_alloc: |
586e46e2 ID |
3958 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3959 | ||
4543df7e | 3960 | iput(fs_info->btree_inode); |
7e662854 | 3961 | fail: |
586e46e2 | 3962 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3963 | return err; |
eb60ceac | 3964 | } |
663faf9f | 3965 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3966 | |
314b6dd0 | 3967 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3968 | { |
314b6dd0 JT |
3969 | struct btrfs_device *device = bio->bi_private; |
3970 | struct bio_vec *bvec; | |
3971 | struct bvec_iter_all iter_all; | |
3972 | struct page *page; | |
3973 | ||
3974 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3975 | page = bvec->bv_page; | |
3976 | ||
3977 | if (bio->bi_status) { | |
3978 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3979 | "lost page write due to IO error on %s (%d)", | |
3980 | rcu_str_deref(device->name), | |
3981 | blk_status_to_errno(bio->bi_status)); | |
3982 | ClearPageUptodate(page); | |
3983 | SetPageError(page); | |
3984 | btrfs_dev_stat_inc_and_print(device, | |
3985 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3986 | } else { | |
3987 | SetPageUptodate(page); | |
3988 | } | |
3989 | ||
3990 | put_page(page); | |
3991 | unlock_page(page); | |
f2984462 | 3992 | } |
314b6dd0 JT |
3993 | |
3994 | bio_put(bio); | |
f2984462 CM |
3995 | } |
3996 | ||
8f32380d JT |
3997 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
3998 | int copy_num) | |
29c36d72 | 3999 | { |
29c36d72 | 4000 | struct btrfs_super_block *super; |
8f32380d | 4001 | struct page *page; |
12659251 | 4002 | u64 bytenr, bytenr_orig; |
8f32380d | 4003 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
4004 | int ret; |
4005 | ||
4006 | bytenr_orig = btrfs_sb_offset(copy_num); | |
4007 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
4008 | if (ret == -ENOENT) | |
4009 | return ERR_PTR(-EINVAL); | |
4010 | else if (ret) | |
4011 | return ERR_PTR(ret); | |
29c36d72 | 4012 | |
cda00eba | 4013 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 4014 | return ERR_PTR(-EINVAL); |
29c36d72 | 4015 | |
8f32380d JT |
4016 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
4017 | if (IS_ERR(page)) | |
4018 | return ERR_CAST(page); | |
29c36d72 | 4019 | |
8f32380d | 4020 | super = page_address(page); |
96c2e067 AJ |
4021 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
4022 | btrfs_release_disk_super(super); | |
4023 | return ERR_PTR(-ENODATA); | |
4024 | } | |
4025 | ||
12659251 | 4026 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
4027 | btrfs_release_disk_super(super); |
4028 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
4029 | } |
4030 | ||
8f32380d | 4031 | return super; |
29c36d72 AJ |
4032 | } |
4033 | ||
4034 | ||
8f32380d | 4035 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 4036 | { |
8f32380d | 4037 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
4038 | int i; |
4039 | u64 transid = 0; | |
a512bbf8 YZ |
4040 | |
4041 | /* we would like to check all the supers, but that would make | |
4042 | * a btrfs mount succeed after a mkfs from a different FS. | |
4043 | * So, we need to add a special mount option to scan for | |
4044 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
4045 | */ | |
4046 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
4047 | super = btrfs_read_dev_one_super(bdev, i); |
4048 | if (IS_ERR(super)) | |
a512bbf8 YZ |
4049 | continue; |
4050 | ||
a512bbf8 | 4051 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
4052 | if (latest) |
4053 | btrfs_release_disk_super(super); | |
4054 | ||
4055 | latest = super; | |
a512bbf8 | 4056 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
4057 | } |
4058 | } | |
92fc03fb | 4059 | |
8f32380d | 4060 | return super; |
a512bbf8 YZ |
4061 | } |
4062 | ||
4eedeb75 | 4063 | /* |
abbb3b8e | 4064 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 4065 | * pages we use for writing are locked. |
4eedeb75 | 4066 | * |
abbb3b8e DS |
4067 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
4068 | * the expected device size at commit time. Note that max_mirrors must be | |
4069 | * same for write and wait phases. | |
4eedeb75 | 4070 | * |
314b6dd0 | 4071 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 4072 | */ |
a512bbf8 | 4073 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 4074 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 4075 | { |
d5178578 | 4076 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 4077 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 4078 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 4079 | int i; |
a512bbf8 | 4080 | int errors = 0; |
12659251 NA |
4081 | int ret; |
4082 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
4083 | |
4084 | if (max_mirrors == 0) | |
4085 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4086 | ||
d5178578 JT |
4087 | shash->tfm = fs_info->csum_shash; |
4088 | ||
a512bbf8 | 4089 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
4090 | struct page *page; |
4091 | struct bio *bio; | |
4092 | struct btrfs_super_block *disk_super; | |
4093 | ||
12659251 NA |
4094 | bytenr_orig = btrfs_sb_offset(i); |
4095 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
4096 | if (ret == -ENOENT) { | |
4097 | continue; | |
4098 | } else if (ret < 0) { | |
4099 | btrfs_err(device->fs_info, | |
4100 | "couldn't get super block location for mirror %d", | |
4101 | i); | |
4102 | errors++; | |
4103 | continue; | |
4104 | } | |
935e5cc9 MX |
4105 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4106 | device->commit_total_bytes) | |
a512bbf8 YZ |
4107 | break; |
4108 | ||
12659251 | 4109 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 4110 | |
fd08001f EB |
4111 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
4112 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
4113 | sb->csum); | |
4eedeb75 | 4114 | |
314b6dd0 JT |
4115 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
4116 | GFP_NOFS); | |
4117 | if (!page) { | |
abbb3b8e | 4118 | btrfs_err(device->fs_info, |
314b6dd0 | 4119 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
4120 | bytenr); |
4121 | errors++; | |
4eedeb75 | 4122 | continue; |
abbb3b8e | 4123 | } |
634554dc | 4124 | |
314b6dd0 JT |
4125 | /* Bump the refcount for wait_dev_supers() */ |
4126 | get_page(page); | |
a512bbf8 | 4127 | |
314b6dd0 JT |
4128 | disk_super = page_address(page); |
4129 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 4130 | |
314b6dd0 JT |
4131 | /* |
4132 | * Directly use bios here instead of relying on the page cache | |
4133 | * to do I/O, so we don't lose the ability to do integrity | |
4134 | * checking. | |
4135 | */ | |
07888c66 CH |
4136 | bio = bio_alloc(device->bdev, 1, |
4137 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
4138 | GFP_NOFS); | |
314b6dd0 JT |
4139 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
4140 | bio->bi_private = device; | |
4141 | bio->bi_end_io = btrfs_end_super_write; | |
4142 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4143 | offset_in_page(bytenr)); | |
a512bbf8 | 4144 | |
387125fc | 4145 | /* |
314b6dd0 JT |
4146 | * We FUA only the first super block. The others we allow to |
4147 | * go down lazy and there's a short window where the on-disk | |
4148 | * copies might still contain the older version. | |
387125fc | 4149 | */ |
1b9e619c | 4150 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4151 | bio->bi_opf |= REQ_FUA; |
4152 | ||
58ff51f1 CH |
4153 | btrfsic_check_bio(bio); |
4154 | submit_bio(bio); | |
8376d9e1 NA |
4155 | |
4156 | if (btrfs_advance_sb_log(device, i)) | |
4157 | errors++; | |
a512bbf8 YZ |
4158 | } |
4159 | return errors < i ? 0 : -1; | |
4160 | } | |
4161 | ||
abbb3b8e DS |
4162 | /* |
4163 | * Wait for write completion of superblocks done by write_dev_supers, | |
4164 | * @max_mirrors same for write and wait phases. | |
4165 | * | |
314b6dd0 | 4166 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4167 | * date. |
4168 | */ | |
4169 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4170 | { | |
abbb3b8e DS |
4171 | int i; |
4172 | int errors = 0; | |
b6a535fa | 4173 | bool primary_failed = false; |
12659251 | 4174 | int ret; |
abbb3b8e DS |
4175 | u64 bytenr; |
4176 | ||
4177 | if (max_mirrors == 0) | |
4178 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4179 | ||
4180 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4181 | struct page *page; |
4182 | ||
12659251 NA |
4183 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4184 | if (ret == -ENOENT) { | |
4185 | break; | |
4186 | } else if (ret < 0) { | |
4187 | errors++; | |
4188 | if (i == 0) | |
4189 | primary_failed = true; | |
4190 | continue; | |
4191 | } | |
abbb3b8e DS |
4192 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4193 | device->commit_total_bytes) | |
4194 | break; | |
4195 | ||
314b6dd0 JT |
4196 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4197 | bytenr >> PAGE_SHIFT); | |
4198 | if (!page) { | |
abbb3b8e | 4199 | errors++; |
b6a535fa HM |
4200 | if (i == 0) |
4201 | primary_failed = true; | |
abbb3b8e DS |
4202 | continue; |
4203 | } | |
314b6dd0 JT |
4204 | /* Page is submitted locked and unlocked once the IO completes */ |
4205 | wait_on_page_locked(page); | |
4206 | if (PageError(page)) { | |
abbb3b8e | 4207 | errors++; |
b6a535fa HM |
4208 | if (i == 0) |
4209 | primary_failed = true; | |
4210 | } | |
abbb3b8e | 4211 | |
314b6dd0 JT |
4212 | /* Drop our reference */ |
4213 | put_page(page); | |
abbb3b8e | 4214 | |
314b6dd0 JT |
4215 | /* Drop the reference from the writing run */ |
4216 | put_page(page); | |
abbb3b8e DS |
4217 | } |
4218 | ||
b6a535fa HM |
4219 | /* log error, force error return */ |
4220 | if (primary_failed) { | |
4221 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4222 | device->devid); | |
4223 | return -1; | |
4224 | } | |
4225 | ||
abbb3b8e DS |
4226 | return errors < i ? 0 : -1; |
4227 | } | |
4228 | ||
387125fc CM |
4229 | /* |
4230 | * endio for the write_dev_flush, this will wake anyone waiting | |
4231 | * for the barrier when it is done | |
4232 | */ | |
4246a0b6 | 4233 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4234 | { |
f9e69aa9 | 4235 | bio_uninit(bio); |
e0ae9994 | 4236 | complete(bio->bi_private); |
387125fc CM |
4237 | } |
4238 | ||
4239 | /* | |
4fc6441a AJ |
4240 | * Submit a flush request to the device if it supports it. Error handling is |
4241 | * done in the waiting counterpart. | |
387125fc | 4242 | */ |
4fc6441a | 4243 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4244 | { |
f9e69aa9 | 4245 | struct bio *bio = &device->flush_bio; |
387125fc | 4246 | |
a91cf0ff WY |
4247 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4248 | /* | |
4249 | * When a disk has write caching disabled, we skip submission of a bio | |
4250 | * with flush and sync requests before writing the superblock, since | |
4251 | * it's not needed. However when the integrity checker is enabled, this | |
4252 | * results in reports that there are metadata blocks referred by a | |
4253 | * superblock that were not properly flushed. So don't skip the bio | |
4254 | * submission only when the integrity checker is enabled for the sake | |
4255 | * of simplicity, since this is a debug tool and not meant for use in | |
4256 | * non-debug builds. | |
4257 | */ | |
08e688fd | 4258 | if (!bdev_write_cache(device->bdev)) |
4fc6441a | 4259 | return; |
a91cf0ff | 4260 | #endif |
387125fc | 4261 | |
f9e69aa9 CH |
4262 | bio_init(bio, device->bdev, NULL, 0, |
4263 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 4264 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
4265 | init_completion(&device->flush_wait); |
4266 | bio->bi_private = &device->flush_wait; | |
387125fc | 4267 | |
58ff51f1 CH |
4268 | btrfsic_check_bio(bio); |
4269 | submit_bio(bio); | |
1c3063b6 | 4270 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4271 | } |
387125fc | 4272 | |
4fc6441a AJ |
4273 | /* |
4274 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4275 | */ | |
8c27cb35 | 4276 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4277 | { |
f9e69aa9 | 4278 | struct bio *bio = &device->flush_bio; |
387125fc | 4279 | |
1c3063b6 | 4280 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4281 | return BLK_STS_OK; |
387125fc | 4282 | |
1c3063b6 | 4283 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4284 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4285 | |
8c27cb35 | 4286 | return bio->bi_status; |
387125fc | 4287 | } |
387125fc | 4288 | |
d10b82fe | 4289 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4290 | { |
6528b99d | 4291 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4292 | return -EIO; |
387125fc CM |
4293 | return 0; |
4294 | } | |
4295 | ||
4296 | /* | |
4297 | * send an empty flush down to each device in parallel, | |
4298 | * then wait for them | |
4299 | */ | |
4300 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4301 | { | |
4302 | struct list_head *head; | |
4303 | struct btrfs_device *dev; | |
5af3e8cc | 4304 | int errors_wait = 0; |
4e4cbee9 | 4305 | blk_status_t ret; |
387125fc | 4306 | |
1538e6c5 | 4307 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4308 | /* send down all the barriers */ |
4309 | head = &info->fs_devices->devices; | |
1538e6c5 | 4310 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4311 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4312 | continue; |
cea7c8bf | 4313 | if (!dev->bdev) |
387125fc | 4314 | continue; |
e12c9621 | 4315 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4316 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4317 | continue; |
4318 | ||
4fc6441a | 4319 | write_dev_flush(dev); |
58efbc9f | 4320 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4321 | } |
4322 | ||
4323 | /* wait for all the barriers */ | |
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; |
387125fc | 4327 | if (!dev->bdev) { |
5af3e8cc | 4328 | errors_wait++; |
387125fc CM |
4329 | continue; |
4330 | } | |
e12c9621 | 4331 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4332 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4333 | continue; |
4334 | ||
4fc6441a | 4335 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4336 | if (ret) { |
4337 | dev->last_flush_error = ret; | |
66b4993e DS |
4338 | btrfs_dev_stat_inc_and_print(dev, |
4339 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4340 | errors_wait++; |
401b41e5 AJ |
4341 | } |
4342 | } | |
4343 | ||
cea7c8bf | 4344 | if (errors_wait) { |
401b41e5 AJ |
4345 | /* |
4346 | * At some point we need the status of all disks | |
4347 | * to arrive at the volume status. So error checking | |
4348 | * is being pushed to a separate loop. | |
4349 | */ | |
d10b82fe | 4350 | return check_barrier_error(info); |
387125fc | 4351 | } |
387125fc CM |
4352 | return 0; |
4353 | } | |
4354 | ||
943c6e99 ZL |
4355 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4356 | { | |
8789f4fe ZL |
4357 | int raid_type; |
4358 | int min_tolerated = INT_MAX; | |
943c6e99 | 4359 | |
8789f4fe ZL |
4360 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4361 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4362 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4363 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4364 | tolerated_failures); | |
943c6e99 | 4365 | |
8789f4fe ZL |
4366 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4367 | if (raid_type == BTRFS_RAID_SINGLE) | |
4368 | continue; | |
41a6e891 | 4369 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4370 | continue; |
8c3e3582 | 4371 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4372 | btrfs_raid_array[raid_type]. |
4373 | tolerated_failures); | |
4374 | } | |
943c6e99 | 4375 | |
8789f4fe | 4376 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4377 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4378 | min_tolerated = 0; |
4379 | } | |
4380 | ||
4381 | return min_tolerated; | |
943c6e99 ZL |
4382 | } |
4383 | ||
eece6a9c | 4384 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4385 | { |
e5e9a520 | 4386 | struct list_head *head; |
f2984462 | 4387 | struct btrfs_device *dev; |
a061fc8d | 4388 | struct btrfs_super_block *sb; |
f2984462 | 4389 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4390 | int ret; |
4391 | int do_barriers; | |
a236aed1 CM |
4392 | int max_errors; |
4393 | int total_errors = 0; | |
a061fc8d | 4394 | u64 flags; |
f2984462 | 4395 | |
0b246afa | 4396 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4397 | |
4398 | /* | |
4399 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4400 | * not from fsync where the tree roots in fs_info have not | |
4401 | * been consistent on disk. | |
4402 | */ | |
4403 | if (max_mirrors == 0) | |
4404 | backup_super_roots(fs_info); | |
f2984462 | 4405 | |
0b246afa | 4406 | sb = fs_info->super_for_commit; |
a061fc8d | 4407 | dev_item = &sb->dev_item; |
e5e9a520 | 4408 | |
0b246afa JM |
4409 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4410 | head = &fs_info->fs_devices->devices; | |
4411 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4412 | |
5af3e8cc | 4413 | if (do_barriers) { |
0b246afa | 4414 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4415 | if (ret) { |
4416 | mutex_unlock( | |
0b246afa JM |
4417 | &fs_info->fs_devices->device_list_mutex); |
4418 | btrfs_handle_fs_error(fs_info, ret, | |
4419 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4420 | return ret; |
4421 | } | |
4422 | } | |
387125fc | 4423 | |
1538e6c5 | 4424 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4425 | if (!dev->bdev) { |
4426 | total_errors++; | |
4427 | continue; | |
4428 | } | |
e12c9621 | 4429 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4430 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4431 | continue; |
4432 | ||
2b82032c | 4433 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4434 | btrfs_set_stack_device_type(dev_item, dev->type); |
4435 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4436 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4437 | dev->commit_total_bytes); |
ce7213c7 MX |
4438 | btrfs_set_stack_device_bytes_used(dev_item, |
4439 | dev->commit_bytes_used); | |
a061fc8d CM |
4440 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4441 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4442 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4443 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4444 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4445 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4446 | |
a061fc8d CM |
4447 | flags = btrfs_super_flags(sb); |
4448 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4449 | ||
75cb857d QW |
4450 | ret = btrfs_validate_write_super(fs_info, sb); |
4451 | if (ret < 0) { | |
4452 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4453 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4454 | "unexpected superblock corruption detected"); | |
4455 | return -EUCLEAN; | |
4456 | } | |
4457 | ||
abbb3b8e | 4458 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4459 | if (ret) |
4460 | total_errors++; | |
f2984462 | 4461 | } |
a236aed1 | 4462 | if (total_errors > max_errors) { |
0b246afa JM |
4463 | btrfs_err(fs_info, "%d errors while writing supers", |
4464 | total_errors); | |
4465 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4466 | |
9d565ba4 | 4467 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4468 | btrfs_handle_fs_error(fs_info, -EIO, |
4469 | "%d errors while writing supers", | |
4470 | total_errors); | |
9d565ba4 | 4471 | return -EIO; |
a236aed1 | 4472 | } |
f2984462 | 4473 | |
a512bbf8 | 4474 | total_errors = 0; |
1538e6c5 | 4475 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4476 | if (!dev->bdev) |
4477 | continue; | |
e12c9621 | 4478 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4479 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4480 | continue; |
4481 | ||
abbb3b8e | 4482 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4483 | if (ret) |
4484 | total_errors++; | |
f2984462 | 4485 | } |
0b246afa | 4486 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4487 | if (total_errors > max_errors) { |
0b246afa JM |
4488 | btrfs_handle_fs_error(fs_info, -EIO, |
4489 | "%d errors while writing supers", | |
4490 | total_errors); | |
79787eaa | 4491 | return -EIO; |
a236aed1 | 4492 | } |
f2984462 CM |
4493 | return 0; |
4494 | } | |
4495 | ||
cb517eab MX |
4496 | /* Drop a fs root from the radix tree and free it. */ |
4497 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4498 | struct btrfs_root *root) | |
2619ba1f | 4499 | { |
4785e24f JB |
4500 | bool drop_ref = false; |
4501 | ||
48b36a60 GN |
4502 | spin_lock(&fs_info->fs_roots_lock); |
4503 | xa_erase(&fs_info->fs_roots, (unsigned long)root->root_key.objectid); | |
4504 | if (test_and_clear_bit(BTRFS_ROOT_REGISTERED, &root->state)) | |
4785e24f | 4505 | drop_ref = true; |
48b36a60 | 4506 | spin_unlock(&fs_info->fs_roots_lock); |
76dda93c | 4507 | |
84961539 | 4508 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4509 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4510 | if (root->reloc_root) { |
00246528 | 4511 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4512 | root->reloc_root = NULL; |
4513 | } | |
4514 | } | |
3321719e | 4515 | |
4785e24f JB |
4516 | if (drop_ref) |
4517 | btrfs_put_root(root); | |
2619ba1f CM |
4518 | } |
4519 | ||
c146afad | 4520 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4521 | { |
48b36a60 GN |
4522 | struct btrfs_root *roots[8]; |
4523 | unsigned long index = 0; | |
4524 | int i; | |
65d33fd7 | 4525 | int err = 0; |
48b36a60 | 4526 | int grabbed; |
e089f05c | 4527 | |
c146afad | 4528 | while (1) { |
48b36a60 GN |
4529 | struct btrfs_root *root; |
4530 | ||
4531 | spin_lock(&fs_info->fs_roots_lock); | |
4532 | if (!xa_find(&fs_info->fs_roots, &index, ULONG_MAX, XA_PRESENT)) { | |
4533 | spin_unlock(&fs_info->fs_roots_lock); | |
4534 | return err; | |
65d33fd7 | 4535 | } |
65d33fd7 | 4536 | |
48b36a60 GN |
4537 | grabbed = 0; |
4538 | xa_for_each_start(&fs_info->fs_roots, index, root, index) { | |
4539 | /* Avoid grabbing roots in dead_roots */ | |
4540 | if (btrfs_root_refs(&root->root_item) > 0) | |
4541 | roots[grabbed++] = btrfs_grab_root(root); | |
4542 | if (grabbed >= ARRAY_SIZE(roots)) | |
4543 | break; | |
65d33fd7 | 4544 | } |
48b36a60 | 4545 | spin_unlock(&fs_info->fs_roots_lock); |
66b4ffd1 | 4546 | |
48b36a60 GN |
4547 | for (i = 0; i < grabbed; i++) { |
4548 | if (!roots[i]) | |
65d33fd7 | 4549 | continue; |
48b36a60 GN |
4550 | index = roots[i]->root_key.objectid; |
4551 | err = btrfs_orphan_cleanup(roots[i]); | |
66b4ffd1 | 4552 | if (err) |
48b36a60 GN |
4553 | goto out; |
4554 | btrfs_put_root(roots[i]); | |
c146afad | 4555 | } |
48b36a60 | 4556 | index++; |
c146afad | 4557 | } |
65d33fd7 | 4558 | |
48b36a60 GN |
4559 | out: |
4560 | /* Release the roots that remain uncleaned due to error */ | |
4561 | for (; i < grabbed; i++) { | |
4562 | if (roots[i]) | |
4563 | btrfs_put_root(roots[i]); | |
65d33fd7 QW |
4564 | } |
4565 | return err; | |
c146afad | 4566 | } |
a2135011 | 4567 | |
6bccf3ab | 4568 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4569 | { |
6bccf3ab | 4570 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4571 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4572 | |
0b246afa | 4573 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4574 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4575 | mutex_unlock(&fs_info->cleaner_mutex); |
4576 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4577 | |
4578 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4579 | down_write(&fs_info->cleanup_work_sem); |
4580 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4581 | |
7a7eaa40 | 4582 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4583 | if (IS_ERR(trans)) |
4584 | return PTR_ERR(trans); | |
3a45bb20 | 4585 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4586 | } |
4587 | ||
36c86a9e QW |
4588 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4589 | { | |
4590 | struct btrfs_transaction *trans; | |
4591 | struct btrfs_transaction *tmp; | |
4592 | bool found = false; | |
4593 | ||
4594 | if (list_empty(&fs_info->trans_list)) | |
4595 | return; | |
4596 | ||
4597 | /* | |
4598 | * This function is only called at the very end of close_ctree(), | |
4599 | * thus no other running transaction, no need to take trans_lock. | |
4600 | */ | |
4601 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4602 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4603 | struct extent_state *cached = NULL; | |
4604 | u64 dirty_bytes = 0; | |
4605 | u64 cur = 0; | |
4606 | u64 found_start; | |
4607 | u64 found_end; | |
4608 | ||
4609 | found = true; | |
4610 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4611 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4612 | dirty_bytes += found_end + 1 - found_start; | |
4613 | cur = found_end + 1; | |
4614 | } | |
4615 | btrfs_warn(fs_info, | |
4616 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4617 | trans->transid, dirty_bytes); | |
4618 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4619 | ||
4620 | if (trans == fs_info->running_transaction) | |
4621 | fs_info->running_transaction = NULL; | |
4622 | list_del_init(&trans->list); | |
4623 | ||
4624 | btrfs_put_transaction(trans); | |
4625 | trace_btrfs_transaction_commit(fs_info); | |
4626 | } | |
4627 | ASSERT(!found); | |
4628 | } | |
4629 | ||
b105e927 | 4630 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4631 | { |
c146afad YZ |
4632 | int ret; |
4633 | ||
afcdd129 | 4634 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 FM |
4635 | |
4636 | /* | |
4637 | * We may have the reclaim task running and relocating a data block group, | |
4638 | * in which case it may create delayed iputs. So stop it before we park | |
4639 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4640 | * parking the cleaner, and that can make the async reclaim task to hang | |
4641 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4642 | * parked and can not run delayed iputs - this will make us hang when | |
4643 | * trying to stop the async reclaim task. | |
4644 | */ | |
4645 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4646 | /* |
4647 | * We don't want the cleaner to start new transactions, add more delayed | |
4648 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4649 | * because that frees the task_struct, and the transaction kthread might | |
4650 | * still try to wake up the cleaner. | |
4651 | */ | |
4652 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4653 | |
b4be6aef JB |
4654 | /* |
4655 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4656 | * clear that bit and wake up relocation so it can stop. | |
4657 | */ | |
4658 | btrfs_wake_unfinished_drop(fs_info); | |
4659 | ||
7343dd61 | 4660 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4661 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4662 | |
803b2f54 SB |
4663 | /* wait for the uuid_scan task to finish */ |
4664 | down(&fs_info->uuid_tree_rescan_sem); | |
4665 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4666 | up(&fs_info->uuid_tree_rescan_sem); | |
4667 | ||
837d5b6e | 4668 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4669 | btrfs_pause_balance(fs_info); |
837d5b6e | 4670 | |
8dabb742 SB |
4671 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4672 | ||
aa1b8cd4 | 4673 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4674 | |
4675 | /* wait for any defraggers to finish */ | |
4676 | wait_event(fs_info->transaction_wait, | |
4677 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4678 | ||
4679 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4680 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4681 | |
21c7e756 | 4682 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4683 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4684 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4685 | |
b0643e59 DZ |
4686 | /* Cancel or finish ongoing discard work */ |
4687 | btrfs_discard_cleanup(fs_info); | |
4688 | ||
bc98a42c | 4689 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4690 | /* |
d6fd0ae2 OS |
4691 | * The cleaner kthread is stopped, so do one final pass over |
4692 | * unused block groups. | |
e44163e1 | 4693 | */ |
0b246afa | 4694 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4695 | |
f0cc2cd7 FM |
4696 | /* |
4697 | * There might be existing delayed inode workers still running | |
4698 | * and holding an empty delayed inode item. We must wait for | |
4699 | * them to complete first because they can create a transaction. | |
4700 | * This happens when someone calls btrfs_balance_delayed_items() | |
4701 | * and then a transaction commit runs the same delayed nodes | |
4702 | * before any delayed worker has done something with the nodes. | |
4703 | * We must wait for any worker here and not at transaction | |
4704 | * commit time since that could cause a deadlock. | |
4705 | * This is a very rare case. | |
4706 | */ | |
4707 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4708 | ||
6bccf3ab | 4709 | ret = btrfs_commit_super(fs_info); |
acce952b | 4710 | if (ret) |
04892340 | 4711 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4712 | } |
4713 | ||
84961539 | 4714 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4715 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4716 | |
e3029d9f AV |
4717 | kthread_stop(fs_info->transaction_kthread); |
4718 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4719 | |
e187831e | 4720 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4721 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4722 | |
5958253c QW |
4723 | if (btrfs_check_quota_leak(fs_info)) { |
4724 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4725 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4726 | } | |
4727 | ||
04892340 | 4728 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4729 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4730 | |
963d678b | 4731 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4732 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4733 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4734 | } |
bcc63abb | 4735 | |
5deb17e1 | 4736 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4737 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4738 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4739 | |
6618a59b | 4740 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4741 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4742 | |
1a4319cc LB |
4743 | btrfs_put_block_group_cache(fs_info); |
4744 | ||
de348ee0 WS |
4745 | /* |
4746 | * we must make sure there is not any read request to | |
4747 | * submit after we stopping all workers. | |
4748 | */ | |
4749 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4750 | btrfs_stop_all_workers(fs_info); |
4751 | ||
0a31daa4 | 4752 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4753 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4754 | |
afcdd129 | 4755 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4756 | free_root_pointers(fs_info, true); |
8c38938c | 4757 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4758 | |
4e19443d JB |
4759 | /* |
4760 | * We must free the block groups after dropping the fs_roots as we could | |
4761 | * have had an IO error and have left over tree log blocks that aren't | |
4762 | * cleaned up until the fs roots are freed. This makes the block group | |
4763 | * accounting appear to be wrong because there's pending reserved bytes, | |
4764 | * so make sure we do the block group cleanup afterwards. | |
4765 | */ | |
4766 | btrfs_free_block_groups(fs_info); | |
4767 | ||
13e6c37b | 4768 | iput(fs_info->btree_inode); |
d6bfde87 | 4769 | |
21adbd5c | 4770 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4771 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4772 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4773 | #endif |
4774 | ||
0b86a832 | 4775 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4776 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4777 | } |
4778 | ||
b9fab919 CM |
4779 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4780 | int atomic) | |
5f39d397 | 4781 | { |
1259ab75 | 4782 | int ret; |
727011e0 | 4783 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4784 | |
0b32f4bb | 4785 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4786 | if (!ret) |
4787 | return ret; | |
4788 | ||
4789 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4790 | parent_transid, atomic); |
4791 | if (ret == -EAGAIN) | |
4792 | return ret; | |
1259ab75 | 4793 | return !ret; |
5f39d397 CM |
4794 | } |
4795 | ||
5f39d397 CM |
4796 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4797 | { | |
2f4d60df | 4798 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4799 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4800 | int was_dirty; |
b4ce94de | 4801 | |
06ea65a3 JB |
4802 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4803 | /* | |
4804 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4805 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4806 | * outside of the sanity tests. |
4807 | */ | |
b0132a3b | 4808 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4809 | return; |
4810 | #endif | |
49d0c642 | 4811 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4812 | if (transid != fs_info->generation) |
5d163e0e | 4813 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4814 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4815 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4816 | if (!was_dirty) |
104b4e51 NB |
4817 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4818 | buf->len, | |
4819 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4820 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4821 | /* |
4822 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4823 | * but item data not updated. | |
4824 | * So here we should only check item pointers, not item data. | |
4825 | */ | |
4826 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4827 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4828 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4829 | ASSERT(0); |
4830 | } | |
4831 | #endif | |
eb60ceac CM |
4832 | } |
4833 | ||
2ff7e61e | 4834 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4835 | int flush_delayed) |
16cdcec7 MX |
4836 | { |
4837 | /* | |
4838 | * looks as though older kernels can get into trouble with | |
4839 | * this code, they end up stuck in balance_dirty_pages forever | |
4840 | */ | |
e2d84521 | 4841 | int ret; |
16cdcec7 MX |
4842 | |
4843 | if (current->flags & PF_MEMALLOC) | |
4844 | return; | |
4845 | ||
b53d3f5d | 4846 | if (flush_delayed) |
2ff7e61e | 4847 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4848 | |
d814a491 EL |
4849 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4850 | BTRFS_DIRTY_METADATA_THRESH, | |
4851 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4852 | if (ret > 0) { |
0b246afa | 4853 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4854 | } |
16cdcec7 MX |
4855 | } |
4856 | ||
2ff7e61e | 4857 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4858 | { |
2ff7e61e | 4859 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4860 | } |
585ad2c3 | 4861 | |
2ff7e61e | 4862 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4863 | { |
2ff7e61e | 4864 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4865 | } |
6b80053d | 4866 | |
2ff7e61e | 4867 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4868 | { |
fe816d0f NB |
4869 | /* cleanup FS via transaction */ |
4870 | btrfs_cleanup_transaction(fs_info); | |
4871 | ||
0b246afa | 4872 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4873 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4874 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4875 | |
0b246afa JM |
4876 | down_write(&fs_info->cleanup_work_sem); |
4877 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4878 | } |
4879 | ||
ef67963d JB |
4880 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4881 | { | |
48b36a60 GN |
4882 | unsigned long index = 0; |
4883 | int grabbed = 0; | |
4884 | struct btrfs_root *roots[8]; | |
4885 | ||
4886 | spin_lock(&fs_info->fs_roots_lock); | |
4887 | while ((grabbed = xa_extract(&fs_info->fs_roots, (void **)roots, index, | |
4888 | ULONG_MAX, 8, XA_PRESENT))) { | |
4889 | for (int i = 0; i < grabbed; i++) | |
4890 | roots[i] = btrfs_grab_root(roots[i]); | |
4891 | spin_unlock(&fs_info->fs_roots_lock); | |
4892 | ||
4893 | for (int i = 0; i < grabbed; i++) { | |
4894 | if (!roots[i]) | |
ef67963d | 4895 | continue; |
48b36a60 GN |
4896 | index = roots[i]->root_key.objectid; |
4897 | btrfs_free_log(NULL, roots[i]); | |
4898 | btrfs_put_root(roots[i]); | |
ef67963d | 4899 | } |
48b36a60 GN |
4900 | index++; |
4901 | spin_lock(&fs_info->fs_roots_lock); | |
ef67963d | 4902 | } |
48b36a60 | 4903 | spin_unlock(&fs_info->fs_roots_lock); |
ef67963d JB |
4904 | btrfs_free_log_root_tree(NULL, fs_info); |
4905 | } | |
4906 | ||
143bede5 | 4907 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4908 | { |
acce952b | 4909 | struct btrfs_ordered_extent *ordered; |
acce952b | 4910 | |
199c2a9c | 4911 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4912 | /* |
4913 | * This will just short circuit the ordered completion stuff which will | |
4914 | * make sure the ordered extent gets properly cleaned up. | |
4915 | */ | |
199c2a9c | 4916 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4917 | root_extent_list) |
4918 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4919 | spin_unlock(&root->ordered_extent_lock); |
4920 | } | |
4921 | ||
4922 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4923 | { | |
4924 | struct btrfs_root *root; | |
4925 | struct list_head splice; | |
4926 | ||
4927 | INIT_LIST_HEAD(&splice); | |
4928 | ||
4929 | spin_lock(&fs_info->ordered_root_lock); | |
4930 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4931 | while (!list_empty(&splice)) { | |
4932 | root = list_first_entry(&splice, struct btrfs_root, | |
4933 | ordered_root); | |
1de2cfde JB |
4934 | list_move_tail(&root->ordered_root, |
4935 | &fs_info->ordered_roots); | |
199c2a9c | 4936 | |
2a85d9ca | 4937 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4938 | btrfs_destroy_ordered_extents(root); |
4939 | ||
2a85d9ca LB |
4940 | cond_resched(); |
4941 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4942 | } |
4943 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4944 | |
4945 | /* | |
4946 | * We need this here because if we've been flipped read-only we won't | |
4947 | * get sync() from the umount, so we need to make sure any ordered | |
4948 | * extents that haven't had their dirty pages IO start writeout yet | |
4949 | * actually get run and error out properly. | |
4950 | */ | |
4951 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4952 | } |
4953 | ||
35a3621b | 4954 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4955 | struct btrfs_fs_info *fs_info) |
acce952b | 4956 | { |
4957 | struct rb_node *node; | |
4958 | struct btrfs_delayed_ref_root *delayed_refs; | |
4959 | struct btrfs_delayed_ref_node *ref; | |
4960 | int ret = 0; | |
4961 | ||
4962 | delayed_refs = &trans->delayed_refs; | |
4963 | ||
4964 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4965 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4966 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4967 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4968 | return ret; |
4969 | } | |
4970 | ||
5c9d028b | 4971 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4972 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4973 | struct rb_node *n; |
e78417d1 | 4974 | bool pin_bytes = false; |
acce952b | 4975 | |
d7df2c79 JB |
4976 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4977 | href_node); | |
3069bd26 | 4978 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4979 | continue; |
3069bd26 | 4980 | |
d7df2c79 | 4981 | spin_lock(&head->lock); |
e3d03965 | 4982 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4983 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4984 | ref_node); | |
d7df2c79 | 4985 | ref->in_tree = 0; |
e3d03965 | 4986 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4987 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4988 | if (!list_empty(&ref->add_list)) |
4989 | list_del(&ref->add_list); | |
d7df2c79 JB |
4990 | atomic_dec(&delayed_refs->num_entries); |
4991 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4992 | } |
d7df2c79 JB |
4993 | if (head->must_insert_reserved) |
4994 | pin_bytes = true; | |
4995 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4996 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4997 | spin_unlock(&head->lock); |
4998 | spin_unlock(&delayed_refs->lock); | |
4999 | mutex_unlock(&head->mutex); | |
acce952b | 5000 | |
f603bb94 NB |
5001 | if (pin_bytes) { |
5002 | struct btrfs_block_group *cache; | |
5003 | ||
5004 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
5005 | BUG_ON(!cache); | |
5006 | ||
5007 | spin_lock(&cache->space_info->lock); | |
5008 | spin_lock(&cache->lock); | |
5009 | cache->pinned += head->num_bytes; | |
5010 | btrfs_space_info_update_bytes_pinned(fs_info, | |
5011 | cache->space_info, head->num_bytes); | |
5012 | cache->reserved -= head->num_bytes; | |
5013 | cache->space_info->bytes_reserved -= head->num_bytes; | |
5014 | spin_unlock(&cache->lock); | |
5015 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
5016 | |
5017 | btrfs_put_block_group(cache); | |
5018 | ||
5019 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
5020 | head->bytenr + head->num_bytes - 1); | |
5021 | } | |
31890da0 | 5022 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 5023 | btrfs_put_delayed_ref_head(head); |
acce952b | 5024 | cond_resched(); |
5025 | spin_lock(&delayed_refs->lock); | |
5026 | } | |
81f7eb00 | 5027 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 5028 | |
5029 | spin_unlock(&delayed_refs->lock); | |
5030 | ||
5031 | return ret; | |
5032 | } | |
5033 | ||
143bede5 | 5034 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 5035 | { |
5036 | struct btrfs_inode *btrfs_inode; | |
5037 | struct list_head splice; | |
5038 | ||
5039 | INIT_LIST_HEAD(&splice); | |
5040 | ||
eb73c1b7 MX |
5041 | spin_lock(&root->delalloc_lock); |
5042 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 5043 | |
5044 | while (!list_empty(&splice)) { | |
fe816d0f | 5045 | struct inode *inode = NULL; |
eb73c1b7 MX |
5046 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
5047 | delalloc_inodes); | |
fe816d0f | 5048 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 5049 | spin_unlock(&root->delalloc_lock); |
acce952b | 5050 | |
fe816d0f NB |
5051 | /* |
5052 | * Make sure we get a live inode and that it'll not disappear | |
5053 | * meanwhile. | |
5054 | */ | |
5055 | inode = igrab(&btrfs_inode->vfs_inode); | |
5056 | if (inode) { | |
5057 | invalidate_inode_pages2(inode->i_mapping); | |
5058 | iput(inode); | |
5059 | } | |
eb73c1b7 | 5060 | spin_lock(&root->delalloc_lock); |
acce952b | 5061 | } |
eb73c1b7 MX |
5062 | spin_unlock(&root->delalloc_lock); |
5063 | } | |
5064 | ||
5065 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
5066 | { | |
5067 | struct btrfs_root *root; | |
5068 | struct list_head splice; | |
5069 | ||
5070 | INIT_LIST_HEAD(&splice); | |
5071 | ||
5072 | spin_lock(&fs_info->delalloc_root_lock); | |
5073 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
5074 | while (!list_empty(&splice)) { | |
5075 | root = list_first_entry(&splice, struct btrfs_root, | |
5076 | delalloc_root); | |
00246528 | 5077 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
5078 | BUG_ON(!root); |
5079 | spin_unlock(&fs_info->delalloc_root_lock); | |
5080 | ||
5081 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 5082 | btrfs_put_root(root); |
eb73c1b7 MX |
5083 | |
5084 | spin_lock(&fs_info->delalloc_root_lock); | |
5085 | } | |
5086 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 5087 | } |
5088 | ||
2ff7e61e | 5089 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 5090 | struct extent_io_tree *dirty_pages, |
5091 | int mark) | |
5092 | { | |
5093 | int ret; | |
acce952b | 5094 | struct extent_buffer *eb; |
5095 | u64 start = 0; | |
5096 | u64 end; | |
acce952b | 5097 | |
5098 | while (1) { | |
5099 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5100 | mark, NULL); |
acce952b | 5101 | if (ret) |
5102 | break; | |
5103 | ||
91166212 | 5104 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5105 | while (start <= end) { |
0b246afa JM |
5106 | eb = find_extent_buffer(fs_info, start); |
5107 | start += fs_info->nodesize; | |
fd8b2b61 | 5108 | if (!eb) |
acce952b | 5109 | continue; |
fd8b2b61 | 5110 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5111 | |
fd8b2b61 JB |
5112 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5113 | &eb->bflags)) | |
5114 | clear_extent_buffer_dirty(eb); | |
5115 | free_extent_buffer_stale(eb); | |
acce952b | 5116 | } |
5117 | } | |
5118 | ||
5119 | return ret; | |
5120 | } | |
5121 | ||
2ff7e61e | 5122 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5123 | struct extent_io_tree *unpin) |
acce952b | 5124 | { |
acce952b | 5125 | u64 start; |
5126 | u64 end; | |
5127 | int ret; | |
5128 | ||
acce952b | 5129 | while (1) { |
0e6ec385 FM |
5130 | struct extent_state *cached_state = NULL; |
5131 | ||
fcd5e742 LF |
5132 | /* |
5133 | * The btrfs_finish_extent_commit() may get the same range as | |
5134 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5135 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5136 | * the same extent range. | |
5137 | */ | |
5138 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5139 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5140 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5141 | if (ret) { |
5142 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5143 | break; |
fcd5e742 | 5144 | } |
acce952b | 5145 | |
0e6ec385 FM |
5146 | clear_extent_dirty(unpin, start, end, &cached_state); |
5147 | free_extent_state(cached_state); | |
2ff7e61e | 5148 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5149 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5150 | cond_resched(); |
5151 | } | |
5152 | ||
5153 | return 0; | |
5154 | } | |
5155 | ||
32da5386 | 5156 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5157 | { |
5158 | struct inode *inode; | |
5159 | ||
5160 | inode = cache->io_ctl.inode; | |
5161 | if (inode) { | |
5162 | invalidate_inode_pages2(inode->i_mapping); | |
5163 | BTRFS_I(inode)->generation = 0; | |
5164 | cache->io_ctl.inode = NULL; | |
5165 | iput(inode); | |
5166 | } | |
bbc37d6e | 5167 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5168 | btrfs_put_block_group(cache); |
5169 | } | |
5170 | ||
5171 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5172 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5173 | { |
32da5386 | 5174 | struct btrfs_block_group *cache; |
c79a1751 LB |
5175 | |
5176 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5177 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5178 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5179 | struct btrfs_block_group, |
c79a1751 | 5180 | dirty_list); |
c79a1751 LB |
5181 | |
5182 | if (!list_empty(&cache->io_list)) { | |
5183 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5184 | list_del_init(&cache->io_list); | |
5185 | btrfs_cleanup_bg_io(cache); | |
5186 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5187 | } | |
5188 | ||
5189 | list_del_init(&cache->dirty_list); | |
5190 | spin_lock(&cache->lock); | |
5191 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5192 | spin_unlock(&cache->lock); | |
5193 | ||
5194 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5195 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5196 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5197 | spin_lock(&cur_trans->dirty_bgs_lock); |
5198 | } | |
5199 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5200 | ||
45ae2c18 NB |
5201 | /* |
5202 | * Refer to the definition of io_bgs member for details why it's safe | |
5203 | * to use it without any locking | |
5204 | */ | |
c79a1751 LB |
5205 | while (!list_empty(&cur_trans->io_bgs)) { |
5206 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5207 | struct btrfs_block_group, |
c79a1751 | 5208 | io_list); |
c79a1751 LB |
5209 | |
5210 | list_del_init(&cache->io_list); | |
5211 | spin_lock(&cache->lock); | |
5212 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5213 | spin_unlock(&cache->lock); | |
5214 | btrfs_cleanup_bg_io(cache); | |
5215 | } | |
5216 | } | |
5217 | ||
49b25e05 | 5218 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5219 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5220 | { |
bbbf7243 NB |
5221 | struct btrfs_device *dev, *tmp; |
5222 | ||
2ff7e61e | 5223 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5224 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5225 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5226 | ||
bbbf7243 NB |
5227 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5228 | post_commit_list) { | |
5229 | list_del_init(&dev->post_commit_list); | |
5230 | } | |
5231 | ||
2ff7e61e | 5232 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5233 | |
4a9d8bde | 5234 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5235 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5236 | |
4a9d8bde | 5237 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5238 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5239 | |
ccdf9b30 | 5240 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5241 | |
2ff7e61e | 5242 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5243 | EXTENT_DIRTY); |
fe119a6e | 5244 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5245 | |
d3575156 NA |
5246 | btrfs_free_redirty_list(cur_trans); |
5247 | ||
4a9d8bde MX |
5248 | cur_trans->state =TRANS_STATE_COMPLETED; |
5249 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5250 | } |
5251 | ||
2ff7e61e | 5252 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5253 | { |
5254 | struct btrfs_transaction *t; | |
acce952b | 5255 | |
0b246afa | 5256 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5257 | |
0b246afa JM |
5258 | spin_lock(&fs_info->trans_lock); |
5259 | while (!list_empty(&fs_info->trans_list)) { | |
5260 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5261 | struct btrfs_transaction, list); |
5262 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5263 | refcount_inc(&t->use_count); |
0b246afa | 5264 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5265 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5266 | btrfs_put_transaction(t); |
0b246afa | 5267 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5268 | continue; |
5269 | } | |
0b246afa | 5270 | if (t == fs_info->running_transaction) { |
724e2315 | 5271 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5272 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5273 | /* |
5274 | * We wait for 0 num_writers since we don't hold a trans | |
5275 | * handle open currently for this transaction. | |
5276 | */ | |
5277 | wait_event(t->writer_wait, | |
5278 | atomic_read(&t->num_writers) == 0); | |
5279 | } else { | |
0b246afa | 5280 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5281 | } |
2ff7e61e | 5282 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5283 | |
0b246afa JM |
5284 | spin_lock(&fs_info->trans_lock); |
5285 | if (t == fs_info->running_transaction) | |
5286 | fs_info->running_transaction = NULL; | |
acce952b | 5287 | list_del_init(&t->list); |
0b246afa | 5288 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5289 | |
724e2315 | 5290 | btrfs_put_transaction(t); |
2e4e97ab | 5291 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5292 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5293 | } |
0b246afa JM |
5294 | spin_unlock(&fs_info->trans_lock); |
5295 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5296 | btrfs_destroy_delayed_inodes(fs_info); |
5297 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5298 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5299 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5300 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5301 | |
5302 | return 0; | |
5303 | } | |
ec7d6dfd | 5304 | |
453e4873 | 5305 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5306 | { |
5307 | struct btrfs_path *path; | |
5308 | int ret; | |
5309 | struct extent_buffer *l; | |
5310 | struct btrfs_key search_key; | |
5311 | struct btrfs_key found_key; | |
5312 | int slot; | |
5313 | ||
5314 | path = btrfs_alloc_path(); | |
5315 | if (!path) | |
5316 | return -ENOMEM; | |
5317 | ||
5318 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5319 | search_key.type = -1; | |
5320 | search_key.offset = (u64)-1; | |
5321 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5322 | if (ret < 0) | |
5323 | goto error; | |
5324 | BUG_ON(ret == 0); /* Corruption */ | |
5325 | if (path->slots[0] > 0) { | |
5326 | slot = path->slots[0] - 1; | |
5327 | l = path->nodes[0]; | |
5328 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5329 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5330 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5331 | } else { |
23125104 | 5332 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5333 | } |
5334 | ret = 0; | |
5335 | error: | |
5336 | btrfs_free_path(path); | |
5337 | return ret; | |
5338 | } | |
5339 | ||
543068a2 | 5340 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5341 | { |
5342 | int ret; | |
5343 | mutex_lock(&root->objectid_mutex); | |
5344 | ||
6b8fad57 | 5345 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5346 | btrfs_warn(root->fs_info, |
5347 | "the objectid of root %llu reaches its highest value", | |
5348 | root->root_key.objectid); | |
5349 | ret = -ENOSPC; | |
5350 | goto out; | |
5351 | } | |
5352 | ||
23125104 | 5353 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5354 | ret = 0; |
5355 | out: | |
5356 | mutex_unlock(&root->objectid_mutex); | |
5357 | return ret; | |
5358 | } |