Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
e20d96d6 | 6 | #include <linux/fs.h> |
d98237b3 | 7 | #include <linux/blkdev.h> |
0f7d52f4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
0b86a832 | 26 | #include "volumes.h" |
db94535d | 27 | #include "print-tree.h" |
925baedd | 28 | #include "locking.h" |
e02119d5 | 29 | #include "tree-log.h" |
fa9c0d79 | 30 | #include "free-space-cache.h" |
70f6d82e | 31 | #include "free-space-tree.h" |
21adbd5c | 32 | #include "check-integrity.h" |
606686ee | 33 | #include "rcu-string.h" |
8dabb742 | 34 | #include "dev-replace.h" |
53b381b3 | 35 | #include "raid56.h" |
5ac1d209 | 36 | #include "sysfs.h" |
fcebe456 | 37 | #include "qgroup.h" |
ebb8765b | 38 | #include "compression.h" |
557ea5dd | 39 | #include "tree-checker.h" |
fd708b81 | 40 | #include "ref-verify.h" |
aac0023c | 41 | #include "block-group.h" |
b0643e59 | 42 | #include "discard.h" |
f603bb94 | 43 | #include "space-info.h" |
b70f5097 | 44 | #include "zoned.h" |
139e8cd3 | 45 | #include "subpage.h" |
eb60ceac | 46 | |
319e4d06 QW |
47 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
48 | BTRFS_HEADER_FLAG_RELOC |\ | |
49 | BTRFS_SUPER_FLAG_ERROR |\ | |
50 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
51 | BTRFS_SUPER_FLAG_METADUMP |\ |
52 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 53 | |
8b712842 | 54 | static void end_workqueue_fn(struct btrfs_work *work); |
143bede5 | 55 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 56 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 57 | struct btrfs_fs_info *fs_info); |
143bede5 | 58 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 59 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 60 | struct extent_io_tree *dirty_pages, |
61 | int mark); | |
2ff7e61e | 62 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 63 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
64 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
65 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 66 | |
d352ac68 | 67 | /* |
97eb6b69 DS |
68 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
69 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
70 | * by writes to insert metadata for new file extents after IO is complete. |
71 | */ | |
97eb6b69 | 72 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
73 | struct bio *bio; |
74 | bio_end_io_t *end_io; | |
75 | void *private; | |
76 | struct btrfs_fs_info *info; | |
4e4cbee9 | 77 | blk_status_t status; |
bfebd8b5 | 78 | enum btrfs_wq_endio_type metadata; |
8b712842 | 79 | struct btrfs_work work; |
ce9adaa5 | 80 | }; |
0da5468f | 81 | |
97eb6b69 DS |
82 | static struct kmem_cache *btrfs_end_io_wq_cache; |
83 | ||
84 | int __init btrfs_end_io_wq_init(void) | |
85 | { | |
86 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
87 | sizeof(struct btrfs_end_io_wq), | |
88 | 0, | |
fba4b697 | 89 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
90 | NULL); |
91 | if (!btrfs_end_io_wq_cache) | |
92 | return -ENOMEM; | |
93 | return 0; | |
94 | } | |
95 | ||
e67c718b | 96 | void __cold btrfs_end_io_wq_exit(void) |
97eb6b69 | 97 | { |
5598e900 | 98 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
99 | } |
100 | ||
141386e1 JB |
101 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
102 | { | |
103 | if (fs_info->csum_shash) | |
104 | crypto_free_shash(fs_info->csum_shash); | |
105 | } | |
106 | ||
d352ac68 CM |
107 | /* |
108 | * async submit bios are used to offload expensive checksumming | |
109 | * onto the worker threads. They checksum file and metadata bios | |
110 | * just before they are sent down the IO stack. | |
111 | */ | |
44b8bd7e | 112 | struct async_submit_bio { |
8896a08d | 113 | struct inode *inode; |
44b8bd7e | 114 | struct bio *bio; |
a758781d | 115 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 116 | int mirror_num; |
1941b64b QW |
117 | |
118 | /* Optional parameter for submit_bio_start used by direct io */ | |
119 | u64 dio_file_offset; | |
8b712842 | 120 | struct btrfs_work work; |
4e4cbee9 | 121 | blk_status_t status; |
44b8bd7e CM |
122 | }; |
123 | ||
85d4e461 CM |
124 | /* |
125 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
126 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
127 | * the level the eb occupies in the tree. | |
128 | * | |
129 | * Different roots are used for different purposes and may nest inside each | |
130 | * other and they require separate keysets. As lockdep keys should be | |
131 | * static, assign keysets according to the purpose of the root as indicated | |
4fd786e6 MT |
132 | * by btrfs_root->root_key.objectid. This ensures that all special purpose |
133 | * roots have separate keysets. | |
4008c04a | 134 | * |
85d4e461 CM |
135 | * Lock-nesting across peer nodes is always done with the immediate parent |
136 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
137 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 138 | * |
85d4e461 CM |
139 | * The key is set by the readpage_end_io_hook after the buffer has passed |
140 | * csum validation but before the pages are unlocked. It is also set by | |
141 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 142 | * |
85d4e461 CM |
143 | * We also add a check to make sure the highest level of the tree is the |
144 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
145 | * needs update as well. | |
4008c04a CM |
146 | */ |
147 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
148 | # if BTRFS_MAX_LEVEL != 8 | |
149 | # error | |
150 | # endif | |
85d4e461 | 151 | |
ab1405aa DS |
152 | #define DEFINE_LEVEL(stem, level) \ |
153 | .names[level] = "btrfs-" stem "-0" #level, | |
154 | ||
155 | #define DEFINE_NAME(stem) \ | |
156 | DEFINE_LEVEL(stem, 0) \ | |
157 | DEFINE_LEVEL(stem, 1) \ | |
158 | DEFINE_LEVEL(stem, 2) \ | |
159 | DEFINE_LEVEL(stem, 3) \ | |
160 | DEFINE_LEVEL(stem, 4) \ | |
161 | DEFINE_LEVEL(stem, 5) \ | |
162 | DEFINE_LEVEL(stem, 6) \ | |
163 | DEFINE_LEVEL(stem, 7) | |
164 | ||
85d4e461 CM |
165 | static struct btrfs_lockdep_keyset { |
166 | u64 id; /* root objectid */ | |
ab1405aa | 167 | /* Longest entry: btrfs-free-space-00 */ |
387824af DS |
168 | char names[BTRFS_MAX_LEVEL][20]; |
169 | struct lock_class_key keys[BTRFS_MAX_LEVEL]; | |
85d4e461 | 170 | } btrfs_lockdep_keysets[] = { |
ab1405aa DS |
171 | { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, |
172 | { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, | |
173 | { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, | |
174 | { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, | |
ab1405aa DS |
175 | { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, |
176 | { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, | |
177 | { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, | |
178 | { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, | |
179 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, | |
180 | { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, | |
181 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, | |
182 | { .id = 0, DEFINE_NAME("tree") }, | |
4008c04a | 183 | }; |
85d4e461 | 184 | |
ab1405aa DS |
185 | #undef DEFINE_LEVEL |
186 | #undef DEFINE_NAME | |
85d4e461 CM |
187 | |
188 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
189 | int level) | |
190 | { | |
191 | struct btrfs_lockdep_keyset *ks; | |
192 | ||
193 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
194 | ||
195 | /* find the matching keyset, id 0 is the default entry */ | |
196 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
197 | if (ks->id == objectid) | |
198 | break; | |
199 | ||
200 | lockdep_set_class_and_name(&eb->lock, | |
201 | &ks->keys[level], ks->names[level]); | |
202 | } | |
203 | ||
4008c04a CM |
204 | #endif |
205 | ||
d352ac68 | 206 | /* |
2996e1f8 | 207 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 208 | */ |
c67b3892 | 209 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 210 | { |
d5178578 | 211 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 212 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 213 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 214 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 215 | char *kaddr; |
e9be5a30 | 216 | int i; |
d5178578 JT |
217 | |
218 | shash->tfm = fs_info->csum_shash; | |
219 | crypto_shash_init(shash); | |
a26663e7 | 220 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 221 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 222 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 223 | |
e9be5a30 DS |
224 | for (i = 1; i < num_pages; i++) { |
225 | kaddr = page_address(buf->pages[i]); | |
226 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 227 | } |
71a63551 | 228 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 229 | crypto_shash_final(shash, result); |
19c00ddc CM |
230 | } |
231 | ||
d352ac68 CM |
232 | /* |
233 | * we can't consider a given block up to date unless the transid of the | |
234 | * block matches the transid in the parent node's pointer. This is how we | |
235 | * detect blocks that either didn't get written at all or got written | |
236 | * in the wrong place. | |
237 | */ | |
1259ab75 | 238 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
239 | struct extent_buffer *eb, u64 parent_transid, |
240 | int atomic) | |
1259ab75 | 241 | { |
2ac55d41 | 242 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
243 | int ret; |
244 | ||
245 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
246 | return 0; | |
247 | ||
b9fab919 CM |
248 | if (atomic) |
249 | return -EAGAIN; | |
250 | ||
2ac55d41 | 251 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 252 | &cached_state); |
0b32f4bb | 253 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
254 | btrfs_header_generation(eb) == parent_transid) { |
255 | ret = 0; | |
256 | goto out; | |
257 | } | |
94647322 DS |
258 | btrfs_err_rl(eb->fs_info, |
259 | "parent transid verify failed on %llu wanted %llu found %llu", | |
260 | eb->start, | |
29549aec | 261 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 262 | ret = 1; |
35b22c19 | 263 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 264 | out: |
2ac55d41 | 265 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 266 | &cached_state); |
1259ab75 | 267 | return ret; |
1259ab75 CM |
268 | } |
269 | ||
e7e16f48 JT |
270 | static bool btrfs_supported_super_csum(u16 csum_type) |
271 | { | |
272 | switch (csum_type) { | |
273 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 274 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 275 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 276 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
277 | return true; |
278 | default: | |
279 | return false; | |
280 | } | |
281 | } | |
282 | ||
1104a885 DS |
283 | /* |
284 | * Return 0 if the superblock checksum type matches the checksum value of that | |
285 | * algorithm. Pass the raw disk superblock data. | |
286 | */ | |
ab8d0fc4 JM |
287 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
288 | char *raw_disk_sb) | |
1104a885 DS |
289 | { |
290 | struct btrfs_super_block *disk_sb = | |
291 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 292 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
293 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
294 | ||
295 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 296 | |
51bce6c9 JT |
297 | /* |
298 | * The super_block structure does not span the whole | |
299 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
300 | * filled with zeros and is included in the checksum. | |
301 | */ | |
fd08001f EB |
302 | crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
303 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); | |
1104a885 | 304 | |
55fc29be | 305 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 306 | return 1; |
1104a885 | 307 | |
e7e16f48 | 308 | return 0; |
1104a885 DS |
309 | } |
310 | ||
e064d5e9 | 311 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 312 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 313 | { |
e064d5e9 | 314 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
315 | int found_level; |
316 | struct btrfs_key found_key; | |
317 | int ret; | |
318 | ||
319 | found_level = btrfs_header_level(eb); | |
320 | if (found_level != level) { | |
63489055 QW |
321 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
322 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
323 | btrfs_err(fs_info, |
324 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
325 | eb->start, level, found_level); | |
581c1760 QW |
326 | return -EIO; |
327 | } | |
328 | ||
329 | if (!first_key) | |
330 | return 0; | |
331 | ||
5d41be6f QW |
332 | /* |
333 | * For live tree block (new tree blocks in current transaction), | |
334 | * we need proper lock context to avoid race, which is impossible here. | |
335 | * So we only checks tree blocks which is read from disk, whose | |
336 | * generation <= fs_info->last_trans_committed. | |
337 | */ | |
338 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
339 | return 0; | |
62fdaa52 QW |
340 | |
341 | /* We have @first_key, so this @eb must have at least one item */ | |
342 | if (btrfs_header_nritems(eb) == 0) { | |
343 | btrfs_err(fs_info, | |
344 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
345 | eb->start); | |
346 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
347 | return -EUCLEAN; | |
348 | } | |
349 | ||
581c1760 QW |
350 | if (found_level) |
351 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
352 | else | |
353 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
354 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
355 | ||
581c1760 | 356 | if (ret) { |
63489055 QW |
357 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
358 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 359 | btrfs_err(fs_info, |
ff76a864 LB |
360 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
361 | eb->start, parent_transid, first_key->objectid, | |
362 | first_key->type, first_key->offset, | |
363 | found_key.objectid, found_key.type, | |
364 | found_key.offset); | |
581c1760 | 365 | } |
581c1760 QW |
366 | return ret; |
367 | } | |
368 | ||
d352ac68 CM |
369 | /* |
370 | * helper to read a given tree block, doing retries as required when | |
371 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
372 | * |
373 | * @parent_transid: expected transid, skip check if 0 | |
374 | * @level: expected level, mandatory check | |
375 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 376 | */ |
5ab12d1f | 377 | static int btree_read_extent_buffer_pages(struct extent_buffer *eb, |
581c1760 QW |
378 | u64 parent_transid, int level, |
379 | struct btrfs_key *first_key) | |
f188591e | 380 | { |
5ab12d1f | 381 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 382 | struct extent_io_tree *io_tree; |
ea466794 | 383 | int failed = 0; |
f188591e CM |
384 | int ret; |
385 | int num_copies = 0; | |
386 | int mirror_num = 0; | |
ea466794 | 387 | int failed_mirror = 0; |
f188591e | 388 | |
0b246afa | 389 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 390 | while (1) { |
f8397d69 | 391 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 392 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 393 | if (!ret) { |
581c1760 | 394 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 395 | parent_transid, 0)) |
256dd1bb | 396 | ret = -EIO; |
e064d5e9 | 397 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 398 | first_key, parent_transid)) |
581c1760 QW |
399 | ret = -EUCLEAN; |
400 | else | |
401 | break; | |
256dd1bb | 402 | } |
d397712b | 403 | |
0b246afa | 404 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 405 | eb->start, eb->len); |
4235298e | 406 | if (num_copies == 1) |
ea466794 | 407 | break; |
4235298e | 408 | |
5cf1ab56 JB |
409 | if (!failed_mirror) { |
410 | failed = 1; | |
411 | failed_mirror = eb->read_mirror; | |
412 | } | |
413 | ||
f188591e | 414 | mirror_num++; |
ea466794 JB |
415 | if (mirror_num == failed_mirror) |
416 | mirror_num++; | |
417 | ||
4235298e | 418 | if (mirror_num > num_copies) |
ea466794 | 419 | break; |
f188591e | 420 | } |
ea466794 | 421 | |
c0901581 | 422 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 423 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
424 | |
425 | return ret; | |
f188591e | 426 | } |
19c00ddc | 427 | |
eca0f6f6 QW |
428 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
429 | { | |
430 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
431 | u8 result[BTRFS_CSUM_SIZE]; | |
432 | int ret; | |
433 | ||
434 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
435 | offsetof(struct btrfs_header, fsid), | |
436 | BTRFS_FSID_SIZE) == 0); | |
437 | csum_tree_block(eb, result); | |
438 | ||
439 | if (btrfs_header_level(eb)) | |
440 | ret = btrfs_check_node(eb); | |
441 | else | |
442 | ret = btrfs_check_leaf_full(eb); | |
443 | ||
444 | if (ret < 0) { | |
445 | btrfs_print_tree(eb, 0); | |
446 | btrfs_err(fs_info, | |
447 | "block=%llu write time tree block corruption detected", | |
448 | eb->start); | |
449 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
450 | return ret; | |
451 | } | |
452 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
453 | ||
454 | return 0; | |
455 | } | |
456 | ||
457 | /* Checksum all dirty extent buffers in one bio_vec */ | |
458 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
459 | struct bio_vec *bvec) | |
460 | { | |
461 | struct page *page = bvec->bv_page; | |
462 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
463 | u64 cur; | |
464 | int ret = 0; | |
465 | ||
466 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
467 | cur += fs_info->nodesize) { | |
468 | struct extent_buffer *eb; | |
469 | bool uptodate; | |
470 | ||
471 | eb = find_extent_buffer(fs_info, cur); | |
472 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
473 | fs_info->nodesize); | |
474 | ||
475 | /* A dirty eb shouldn't disappear from buffer_radix */ | |
476 | if (WARN_ON(!eb)) | |
477 | return -EUCLEAN; | |
478 | ||
479 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
480 | free_extent_buffer(eb); | |
481 | return -EUCLEAN; | |
482 | } | |
483 | if (WARN_ON(!uptodate)) { | |
484 | free_extent_buffer(eb); | |
485 | return -EUCLEAN; | |
486 | } | |
487 | ||
488 | ret = csum_one_extent_buffer(eb); | |
489 | free_extent_buffer(eb); | |
490 | if (ret < 0) | |
491 | return ret; | |
492 | } | |
493 | return ret; | |
494 | } | |
495 | ||
d352ac68 | 496 | /* |
ac303b69 QW |
497 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
498 | * we only fill in the checksum field in the first page of a multi-page block. | |
499 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 500 | */ |
ac303b69 | 501 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 502 | { |
ac303b69 | 503 | struct page *page = bvec->bv_page; |
4eee4fa4 | 504 | u64 start = page_offset(page); |
19c00ddc | 505 | u64 found_start; |
19c00ddc | 506 | struct extent_buffer *eb; |
eca0f6f6 QW |
507 | |
508 | if (fs_info->sectorsize < PAGE_SIZE) | |
509 | return csum_dirty_subpage_buffers(fs_info, bvec); | |
f188591e | 510 | |
4f2de97a JB |
511 | eb = (struct extent_buffer *)page->private; |
512 | if (page != eb->pages[0]) | |
513 | return 0; | |
0f805531 | 514 | |
19c00ddc | 515 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
516 | |
517 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
518 | WARN_ON(found_start != 0); | |
519 | return 0; | |
520 | } | |
521 | ||
0f805531 AL |
522 | /* |
523 | * Please do not consolidate these warnings into a single if. | |
524 | * It is useful to know what went wrong. | |
525 | */ | |
526 | if (WARN_ON(found_start != start)) | |
527 | return -EUCLEAN; | |
528 | if (WARN_ON(!PageUptodate(page))) | |
529 | return -EUCLEAN; | |
530 | ||
eca0f6f6 | 531 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
532 | } |
533 | ||
b0c9b3b0 | 534 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 535 | { |
b0c9b3b0 | 536 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 537 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 538 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 539 | u8 *metadata_uuid; |
2b82032c | 540 | |
9a8658e3 DS |
541 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
542 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
543 | /* |
544 | * Checking the incompat flag is only valid for the current fs. For | |
545 | * seed devices it's forbidden to have their uuid changed so reading | |
546 | * ->fsid in this case is fine | |
547 | */ | |
548 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
549 | metadata_uuid = fs_devices->metadata_uuid; | |
550 | else | |
551 | metadata_uuid = fs_devices->fsid; | |
552 | ||
553 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
554 | return 0; | |
555 | ||
556 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
557 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
558 | return 0; | |
559 | ||
560 | return 1; | |
2b82032c YZ |
561 | } |
562 | ||
77bf40a2 QW |
563 | /* Do basic extent buffer checks at read time */ |
564 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 565 | { |
77bf40a2 | 566 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 567 | u64 found_start; |
77bf40a2 QW |
568 | const u32 csum_size = fs_info->csum_size; |
569 | u8 found_level; | |
2996e1f8 | 570 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 571 | const u8 *header_csum; |
77bf40a2 | 572 | int ret = 0; |
ea466794 | 573 | |
ce9adaa5 | 574 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 575 | if (found_start != eb->start) { |
893bf4b1 SY |
576 | btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", |
577 | eb->start, found_start); | |
f188591e | 578 | ret = -EIO; |
77bf40a2 | 579 | goto out; |
ce9adaa5 | 580 | } |
b0c9b3b0 | 581 | if (check_tree_block_fsid(eb)) { |
02873e43 ZL |
582 | btrfs_err_rl(fs_info, "bad fsid on block %llu", |
583 | eb->start); | |
1259ab75 | 584 | ret = -EIO; |
77bf40a2 | 585 | goto out; |
1259ab75 | 586 | } |
ce9adaa5 | 587 | found_level = btrfs_header_level(eb); |
1c24c3ce | 588 | if (found_level >= BTRFS_MAX_LEVEL) { |
893bf4b1 SY |
589 | btrfs_err(fs_info, "bad tree block level %d on %llu", |
590 | (int)btrfs_header_level(eb), eb->start); | |
1c24c3ce | 591 | ret = -EIO; |
77bf40a2 | 592 | goto out; |
1c24c3ce | 593 | } |
ce9adaa5 | 594 | |
c67b3892 | 595 | csum_tree_block(eb, result); |
dfd29eed DS |
596 | header_csum = page_address(eb->pages[0]) + |
597 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 598 | |
dfd29eed | 599 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 600 | btrfs_warn_rl(fs_info, |
ff14aa79 DS |
601 | "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
602 | eb->start, | |
dfd29eed | 603 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
604 | CSUM_FMT_VALUE(csum_size, result), |
605 | btrfs_header_level(eb)); | |
2996e1f8 | 606 | ret = -EUCLEAN; |
77bf40a2 | 607 | goto out; |
2996e1f8 JT |
608 | } |
609 | ||
a826d6dc JB |
610 | /* |
611 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
612 | * that we don't try and read the other copies of this block, just | |
613 | * return -EIO. | |
614 | */ | |
1c4360ee | 615 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
616 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
617 | ret = -EIO; | |
618 | } | |
ce9adaa5 | 619 | |
813fd1dc | 620 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
621 | ret = -EIO; |
622 | ||
0b32f4bb JB |
623 | if (!ret) |
624 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
625 | else |
626 | btrfs_err(fs_info, | |
627 | "block=%llu read time tree block corruption detected", | |
628 | eb->start); | |
77bf40a2 QW |
629 | out: |
630 | return ret; | |
631 | } | |
632 | ||
371cdc07 QW |
633 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
634 | int mirror) | |
635 | { | |
636 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
637 | struct extent_buffer *eb; | |
638 | bool reads_done; | |
639 | int ret = 0; | |
640 | ||
641 | /* | |
642 | * We don't allow bio merge for subpage metadata read, so we should | |
643 | * only get one eb for each endio hook. | |
644 | */ | |
645 | ASSERT(end == start + fs_info->nodesize - 1); | |
646 | ASSERT(PagePrivate(page)); | |
647 | ||
648 | eb = find_extent_buffer(fs_info, start); | |
649 | /* | |
650 | * When we are reading one tree block, eb must have been inserted into | |
651 | * the radix tree. If not, something is wrong. | |
652 | */ | |
653 | ASSERT(eb); | |
654 | ||
655 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
656 | /* Subpage read must finish in page read */ | |
657 | ASSERT(reads_done); | |
658 | ||
659 | eb->read_mirror = mirror; | |
660 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
661 | ret = -EIO; | |
662 | goto err; | |
663 | } | |
664 | ret = validate_extent_buffer(eb); | |
665 | if (ret < 0) | |
666 | goto err; | |
667 | ||
371cdc07 QW |
668 | set_extent_buffer_uptodate(eb); |
669 | ||
670 | free_extent_buffer(eb); | |
671 | return ret; | |
672 | err: | |
673 | /* | |
674 | * end_bio_extent_readpage decrements io_pages in case of error, | |
675 | * make sure it has something to decrement. | |
676 | */ | |
677 | atomic_inc(&eb->io_pages); | |
678 | clear_extent_buffer_uptodate(eb); | |
679 | free_extent_buffer(eb); | |
680 | return ret; | |
681 | } | |
682 | ||
c3a3b19b | 683 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
684 | struct page *page, u64 start, u64 end, |
685 | int mirror) | |
686 | { | |
687 | struct extent_buffer *eb; | |
688 | int ret = 0; | |
689 | int reads_done; | |
690 | ||
691 | ASSERT(page->private); | |
371cdc07 QW |
692 | |
693 | if (btrfs_sb(page->mapping->host->i_sb)->sectorsize < PAGE_SIZE) | |
694 | return validate_subpage_buffer(page, start, end, mirror); | |
695 | ||
77bf40a2 QW |
696 | eb = (struct extent_buffer *)page->private; |
697 | ||
698 | /* | |
699 | * The pending IO might have been the only thing that kept this buffer | |
700 | * in memory. Make sure we have a ref for all this other checks | |
701 | */ | |
702 | atomic_inc(&eb->refs); | |
703 | ||
704 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
705 | if (!reads_done) | |
706 | goto err; | |
707 | ||
708 | eb->read_mirror = mirror; | |
709 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
710 | ret = -EIO; | |
711 | goto err; | |
712 | } | |
713 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 714 | err: |
53b381b3 DW |
715 | if (ret) { |
716 | /* | |
717 | * our io error hook is going to dec the io pages | |
718 | * again, we have to make sure it has something | |
719 | * to decrement | |
720 | */ | |
721 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 722 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 723 | } |
0b32f4bb | 724 | free_extent_buffer(eb); |
77bf40a2 | 725 | |
f188591e | 726 | return ret; |
ce9adaa5 CM |
727 | } |
728 | ||
4246a0b6 | 729 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 730 | { |
97eb6b69 | 731 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 732 | struct btrfs_fs_info *fs_info; |
9e0af237 | 733 | struct btrfs_workqueue *wq; |
ce9adaa5 | 734 | |
ce9adaa5 | 735 | fs_info = end_io_wq->info; |
4e4cbee9 | 736 | end_io_wq->status = bio->bi_status; |
d20f7043 | 737 | |
cfe94440 | 738 | if (btrfs_op(bio) == BTRFS_MAP_WRITE) { |
a0cac0ec | 739 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
9e0af237 | 740 | wq = fs_info->endio_meta_write_workers; |
a0cac0ec | 741 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
9e0af237 | 742 | wq = fs_info->endio_freespace_worker; |
a0cac0ec | 743 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 744 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 745 | else |
9e0af237 | 746 | wq = fs_info->endio_write_workers; |
d20f7043 | 747 | } else { |
5c047a69 | 748 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
9e0af237 | 749 | wq = fs_info->endio_raid56_workers; |
a0cac0ec | 750 | else if (end_io_wq->metadata) |
9e0af237 | 751 | wq = fs_info->endio_meta_workers; |
a0cac0ec | 752 | else |
9e0af237 | 753 | wq = fs_info->endio_workers; |
d20f7043 | 754 | } |
9e0af237 | 755 | |
a0cac0ec | 756 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
9e0af237 | 757 | btrfs_queue_work(wq, &end_io_wq->work); |
ce9adaa5 CM |
758 | } |
759 | ||
4e4cbee9 | 760 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 761 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 762 | { |
97eb6b69 | 763 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 764 | |
97eb6b69 | 765 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 766 | if (!end_io_wq) |
4e4cbee9 | 767 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
768 | |
769 | end_io_wq->private = bio->bi_private; | |
770 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 771 | end_io_wq->info = info; |
4e4cbee9 | 772 | end_io_wq->status = 0; |
ce9adaa5 | 773 | end_io_wq->bio = bio; |
22c59948 | 774 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
775 | |
776 | bio->bi_private = end_io_wq; | |
777 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
778 | return 0; |
779 | } | |
780 | ||
4a69a410 CM |
781 | static void run_one_async_start(struct btrfs_work *work) |
782 | { | |
4a69a410 | 783 | struct async_submit_bio *async; |
4e4cbee9 | 784 | blk_status_t ret; |
4a69a410 CM |
785 | |
786 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
787 | ret = async->submit_bio_start(async->inode, async->bio, |
788 | async->dio_file_offset); | |
79787eaa | 789 | if (ret) |
4e4cbee9 | 790 | async->status = ret; |
4a69a410 CM |
791 | } |
792 | ||
06ea01b1 DS |
793 | /* |
794 | * In order to insert checksums into the metadata in large chunks, we wait | |
795 | * until bio submission time. All the pages in the bio are checksummed and | |
796 | * sums are attached onto the ordered extent record. | |
797 | * | |
798 | * At IO completion time the csums attached on the ordered extent record are | |
799 | * inserted into the tree. | |
800 | */ | |
4a69a410 | 801 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 802 | { |
8b712842 | 803 | struct async_submit_bio *async; |
06ea01b1 DS |
804 | struct inode *inode; |
805 | blk_status_t ret; | |
8b712842 CM |
806 | |
807 | async = container_of(work, struct async_submit_bio, work); | |
8896a08d | 808 | inode = async->inode; |
4854ddd0 | 809 | |
bb7ab3b9 | 810 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
811 | if (async->status) { |
812 | async->bio->bi_status = async->status; | |
4246a0b6 | 813 | bio_endio(async->bio); |
79787eaa JM |
814 | return; |
815 | } | |
816 | ||
ec39f769 CM |
817 | /* |
818 | * All of the bios that pass through here are from async helpers. | |
819 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
820 | * This changes nothing when cgroups aren't in use. | |
821 | */ | |
822 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
08635bae | 823 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
06ea01b1 DS |
824 | if (ret) { |
825 | async->bio->bi_status = ret; | |
826 | bio_endio(async->bio); | |
827 | } | |
4a69a410 CM |
828 | } |
829 | ||
830 | static void run_one_async_free(struct btrfs_work *work) | |
831 | { | |
832 | struct async_submit_bio *async; | |
833 | ||
834 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
835 | kfree(async); |
836 | } | |
837 | ||
8896a08d | 838 | blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, |
8c27cb35 | 839 | int mirror_num, unsigned long bio_flags, |
1941b64b | 840 | u64 dio_file_offset, |
e288c080 | 841 | extent_submit_bio_start_t *submit_bio_start) |
44b8bd7e | 842 | { |
8896a08d | 843 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
844 | struct async_submit_bio *async; |
845 | ||
846 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
847 | if (!async) | |
4e4cbee9 | 848 | return BLK_STS_RESOURCE; |
44b8bd7e | 849 | |
8896a08d | 850 | async->inode = inode; |
44b8bd7e CM |
851 | async->bio = bio; |
852 | async->mirror_num = mirror_num; | |
4a69a410 | 853 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 854 | |
a0cac0ec OS |
855 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
856 | run_one_async_free); | |
4a69a410 | 857 | |
1941b64b | 858 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 859 | |
4e4cbee9 | 860 | async->status = 0; |
79787eaa | 861 | |
67f055c7 | 862 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 863 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 864 | |
5cdc7ad3 | 865 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
866 | return 0; |
867 | } | |
868 | ||
4e4cbee9 | 869 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 870 | { |
2c30c71b | 871 | struct bio_vec *bvec; |
ce3ed71a | 872 | struct btrfs_root *root; |
2b070cfe | 873 | int ret = 0; |
6dc4f100 | 874 | struct bvec_iter_all iter_all; |
ce3ed71a | 875 | |
c09abff8 | 876 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 877 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 878 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 879 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
880 | if (ret) |
881 | break; | |
ce3ed71a | 882 | } |
2c30c71b | 883 | |
4e4cbee9 | 884 | return errno_to_blk_status(ret); |
ce3ed71a CM |
885 | } |
886 | ||
8896a08d | 887 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 888 | u64 dio_file_offset) |
22c59948 | 889 | { |
8b712842 CM |
890 | /* |
891 | * when we're called for a write, we're already in the async | |
5443be45 | 892 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 893 | */ |
79787eaa | 894 | return btree_csum_one_bio(bio); |
4a69a410 | 895 | } |
22c59948 | 896 | |
f4dcfb30 | 897 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 898 | struct btrfs_inode *bi) |
de0022b9 | 899 | { |
4eef29ef | 900 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 901 | return false; |
6300463b | 902 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 903 | return false; |
9b4e675a | 904 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
905 | return false; |
906 | return true; | |
de0022b9 JB |
907 | } |
908 | ||
1b36294a NB |
909 | blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, |
910 | int mirror_num, unsigned long bio_flags) | |
44b8bd7e | 911 | { |
0b246afa | 912 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
4e4cbee9 | 913 | blk_status_t ret; |
cad321ad | 914 | |
cfe94440 | 915 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
4a69a410 CM |
916 | /* |
917 | * called for a read, do the setup so that checksum validation | |
918 | * can happen in the async kernel threads | |
919 | */ | |
0b246afa JM |
920 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
921 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 922 | if (ret) |
61891923 | 923 | goto out_w_error; |
08635bae | 924 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
f4dcfb30 | 925 | } else if (!should_async_write(fs_info, BTRFS_I(inode))) { |
de0022b9 JB |
926 | ret = btree_csum_one_bio(bio); |
927 | if (ret) | |
61891923 | 928 | goto out_w_error; |
08635bae | 929 | ret = btrfs_map_bio(fs_info, bio, mirror_num); |
61891923 SB |
930 | } else { |
931 | /* | |
932 | * kthread helpers are used to submit writes so that | |
933 | * checksumming can happen in parallel across all CPUs | |
934 | */ | |
8896a08d QW |
935 | ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, |
936 | 0, btree_submit_bio_start); | |
44b8bd7e | 937 | } |
d313d7a3 | 938 | |
4246a0b6 CH |
939 | if (ret) |
940 | goto out_w_error; | |
941 | return 0; | |
942 | ||
61891923 | 943 | out_w_error: |
4e4cbee9 | 944 | bio->bi_status = ret; |
4246a0b6 | 945 | bio_endio(bio); |
61891923 | 946 | return ret; |
44b8bd7e CM |
947 | } |
948 | ||
3dd1462e | 949 | #ifdef CONFIG_MIGRATION |
784b4e29 | 950 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
951 | struct page *newpage, struct page *page, |
952 | enum migrate_mode mode) | |
784b4e29 CM |
953 | { |
954 | /* | |
955 | * we can't safely write a btree page from here, | |
956 | * we haven't done the locking hook | |
957 | */ | |
958 | if (PageDirty(page)) | |
959 | return -EAGAIN; | |
960 | /* | |
961 | * Buffers may be managed in a filesystem specific way. | |
962 | * We must have no buffers or drop them. | |
963 | */ | |
964 | if (page_has_private(page) && | |
965 | !try_to_release_page(page, GFP_KERNEL)) | |
966 | return -EAGAIN; | |
a6bc32b8 | 967 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 968 | } |
3dd1462e | 969 | #endif |
784b4e29 | 970 | |
0da5468f CM |
971 | |
972 | static int btree_writepages(struct address_space *mapping, | |
973 | struct writeback_control *wbc) | |
974 | { | |
e2d84521 MX |
975 | struct btrfs_fs_info *fs_info; |
976 | int ret; | |
977 | ||
d8d5f3e1 | 978 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
979 | |
980 | if (wbc->for_kupdate) | |
981 | return 0; | |
982 | ||
e2d84521 | 983 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 984 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
985 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
986 | BTRFS_DIRTY_METADATA_THRESH, | |
987 | fs_info->dirty_metadata_batch); | |
e2d84521 | 988 | if (ret < 0) |
793955bc | 989 | return 0; |
793955bc | 990 | } |
0b32f4bb | 991 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
992 | } |
993 | ||
70dec807 | 994 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 995 | { |
98509cfc | 996 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 997 | return 0; |
0c4e538b | 998 | |
f7a52a40 | 999 | return try_release_extent_buffer(page); |
d98237b3 CM |
1000 | } |
1001 | ||
d47992f8 LC |
1002 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1003 | unsigned int length) | |
d98237b3 | 1004 | { |
d1310b2e CM |
1005 | struct extent_io_tree *tree; |
1006 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1007 | extent_invalidatepage(tree, page, offset); |
1008 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1009 | if (PagePrivate(page)) { |
efe120a0 FH |
1010 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1011 | "page private not zero on page %llu", | |
1012 | (unsigned long long)page_offset(page)); | |
d1b89bc0 | 1013 | detach_page_private(page); |
9ad6b7bc | 1014 | } |
d98237b3 CM |
1015 | } |
1016 | ||
0b32f4bb JB |
1017 | static int btree_set_page_dirty(struct page *page) |
1018 | { | |
bb146eb2 | 1019 | #ifdef DEBUG |
139e8cd3 QW |
1020 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); |
1021 | struct btrfs_subpage *subpage; | |
0b32f4bb | 1022 | struct extent_buffer *eb; |
139e8cd3 QW |
1023 | int cur_bit = 0; |
1024 | u64 page_start = page_offset(page); | |
1025 | ||
1026 | if (fs_info->sectorsize == PAGE_SIZE) { | |
1027 | BUG_ON(!PagePrivate(page)); | |
1028 | eb = (struct extent_buffer *)page->private; | |
1029 | BUG_ON(!eb); | |
1030 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1031 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 1032 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
1033 | return __set_page_dirty_nobuffers(page); |
1034 | } | |
1035 | ASSERT(PagePrivate(page) && page->private); | |
1036 | subpage = (struct btrfs_subpage *)page->private; | |
1037 | ||
1038 | ASSERT(subpage->dirty_bitmap); | |
1039 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
1040 | unsigned long flags; | |
1041 | u64 cur; | |
1042 | u16 tmp = (1 << cur_bit); | |
1043 | ||
1044 | spin_lock_irqsave(&subpage->lock, flags); | |
1045 | if (!(tmp & subpage->dirty_bitmap)) { | |
1046 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1047 | cur_bit++; | |
1048 | continue; | |
1049 | } | |
1050 | spin_unlock_irqrestore(&subpage->lock, flags); | |
1051 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 1052 | |
139e8cd3 QW |
1053 | eb = find_extent_buffer(fs_info, cur); |
1054 | ASSERT(eb); | |
1055 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1056 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 1057 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
1058 | free_extent_buffer(eb); |
1059 | ||
1060 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
1061 | } | |
bb146eb2 | 1062 | #endif |
0b32f4bb JB |
1063 | return __set_page_dirty_nobuffers(page); |
1064 | } | |
1065 | ||
7f09410b | 1066 | static const struct address_space_operations btree_aops = { |
0da5468f | 1067 | .writepages = btree_writepages, |
5f39d397 CM |
1068 | .releasepage = btree_releasepage, |
1069 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1070 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1071 | .migratepage = btree_migratepage, |
5a92bc88 | 1072 | #endif |
0b32f4bb | 1073 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1074 | }; |
1075 | ||
2ff7e61e JM |
1076 | struct extent_buffer *btrfs_find_create_tree_block( |
1077 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
1078 | u64 bytenr, u64 owner_root, |
1079 | int level) | |
0999df54 | 1080 | { |
0b246afa JM |
1081 | if (btrfs_is_testing(fs_info)) |
1082 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 1083 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
1084 | } |
1085 | ||
581c1760 QW |
1086 | /* |
1087 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1088 | * verification. | |
1089 | * | |
1b7ec85e | 1090 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
1091 | * @parent_transid: expected transid of this tree block, skip check if 0 |
1092 | * @level: expected level, mandatory check | |
1093 | * @first_key: expected key in slot 0, skip check if NULL | |
1094 | */ | |
2ff7e61e | 1095 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
1096 | u64 owner_root, u64 parent_transid, |
1097 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
1098 | { |
1099 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1100 | int ret; |
1101 | ||
3fbaf258 | 1102 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
1103 | if (IS_ERR(buf)) |
1104 | return buf; | |
0999df54 | 1105 | |
5ab12d1f | 1106 | ret = btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 1107 | level, first_key); |
0f0fe8f7 | 1108 | if (ret) { |
537f38f0 | 1109 | free_extent_buffer_stale(buf); |
64c043de | 1110 | return ERR_PTR(ret); |
0f0fe8f7 | 1111 | } |
5f39d397 | 1112 | return buf; |
ce9adaa5 | 1113 | |
eb60ceac CM |
1114 | } |
1115 | ||
6a884d7d | 1116 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1117 | { |
6a884d7d | 1118 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1119 | if (btrfs_header_generation(buf) == |
e2d84521 | 1120 | fs_info->running_transaction->transid) { |
49d0c642 | 1121 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 1122 | |
b9473439 | 1123 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1124 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1125 | -buf->len, | |
1126 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1127 | clear_extent_buffer_dirty(buf); |
1128 | } | |
925baedd | 1129 | } |
5f39d397 CM |
1130 | } |
1131 | ||
da17066c | 1132 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1133 | u64 objectid) |
d97e63b6 | 1134 | { |
7c0260ee | 1135 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
1136 | |
1137 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
1138 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
1139 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 1140 | root->fs_info = fs_info; |
2e608bd1 | 1141 | root->root_key.objectid = objectid; |
cfaa7295 | 1142 | root->node = NULL; |
a28ec197 | 1143 | root->commit_root = NULL; |
27cdeb70 | 1144 | root->state = 0; |
abed4aaa | 1145 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 1146 | |
0f7d52f4 | 1147 | root->last_trans = 0; |
6b8fad57 | 1148 | root->free_objectid = 0; |
eb73c1b7 | 1149 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1150 | root->nr_ordered_extents = 0; |
6bef4d31 | 1151 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1152 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
2e608bd1 JB |
1153 | |
1154 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
1155 | |
1156 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1157 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1158 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1159 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1160 | INIT_LIST_HEAD(&root->ordered_extents); |
1161 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1162 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1163 | INIT_LIST_HEAD(&root->logged_list[0]); |
1164 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1165 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1166 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1167 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1168 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1169 | spin_lock_init(&root->log_extents_lock[0]); |
1170 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1171 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1172 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1173 | mutex_init(&root->log_mutex); |
31f3d255 | 1174 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1175 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1176 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1177 | init_waitqueue_head(&root->log_writer_wait); |
1178 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1179 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1180 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1181 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1182 | atomic_set(&root->log_commit[0], 0); |
1183 | atomic_set(&root->log_commit[1], 0); | |
1184 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1185 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1186 | refcount_set(&root->refs, 1); |
8ecebf4d | 1187 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1188 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1189 | root->log_transid = 0; |
d1433deb | 1190 | root->log_transid_committed = -1; |
257c62e1 | 1191 | root->last_log_commit = 0; |
2e608bd1 | 1192 | root->anon_dev = 0; |
e289f03e | 1193 | if (!dummy) { |
43eb5f29 QW |
1194 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1195 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1196 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1197 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1198 | } | |
017e5369 | 1199 | |
5f3ab90a | 1200 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1201 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1202 | #ifdef CONFIG_BTRFS_DEBUG |
1203 | INIT_LIST_HEAD(&root->leak_list); | |
1204 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1205 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); | |
1206 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1207 | #endif | |
3768f368 CM |
1208 | } |
1209 | ||
74e4d827 | 1210 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1211 | u64 objectid, gfp_t flags) |
6f07e42e | 1212 | { |
74e4d827 | 1213 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1214 | if (root) |
96dfcb46 | 1215 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1216 | return root; |
1217 | } | |
1218 | ||
06ea65a3 JB |
1219 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1220 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1221 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1222 | { |
1223 | struct btrfs_root *root; | |
1224 | ||
7c0260ee JM |
1225 | if (!fs_info) |
1226 | return ERR_PTR(-EINVAL); | |
1227 | ||
96dfcb46 | 1228 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1229 | if (!root) |
1230 | return ERR_PTR(-ENOMEM); | |
da17066c | 1231 | |
b9ef22de | 1232 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1233 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1234 | |
1235 | return root; | |
1236 | } | |
1237 | #endif | |
1238 | ||
abed4aaa JB |
1239 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1240 | { | |
1241 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1242 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1243 | ||
1244 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1245 | } | |
1246 | ||
1247 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1248 | { | |
1249 | const struct btrfs_key *key = k; | |
1250 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1251 | ||
1252 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1253 | } | |
1254 | ||
1255 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1256 | { | |
1257 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1258 | struct rb_node *tmp; | |
1259 | ||
1260 | write_lock(&fs_info->global_root_lock); | |
1261 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1262 | write_unlock(&fs_info->global_root_lock); | |
1263 | ASSERT(!tmp); | |
1264 | ||
1265 | return tmp ? -EEXIST : 0; | |
1266 | } | |
1267 | ||
1268 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1269 | { | |
1270 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1271 | ||
1272 | write_lock(&fs_info->global_root_lock); | |
1273 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1274 | write_unlock(&fs_info->global_root_lock); | |
1275 | } | |
1276 | ||
1277 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1278 | struct btrfs_key *key) | |
1279 | { | |
1280 | struct rb_node *node; | |
1281 | struct btrfs_root *root = NULL; | |
1282 | ||
1283 | read_lock(&fs_info->global_root_lock); | |
1284 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1285 | if (node) | |
1286 | root = container_of(node, struct btrfs_root, rb_node); | |
1287 | read_unlock(&fs_info->global_root_lock); | |
1288 | ||
1289 | return root; | |
1290 | } | |
1291 | ||
1292 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1293 | { | |
1294 | struct btrfs_key key = { | |
1295 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1296 | .type = BTRFS_ROOT_ITEM_KEY, | |
1297 | .offset = 0, | |
1298 | }; | |
1299 | ||
1300 | return btrfs_global_root(fs_info, &key); | |
1301 | } | |
1302 | ||
1303 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1304 | { | |
1305 | struct btrfs_key key = { | |
1306 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1307 | .type = BTRFS_ROOT_ITEM_KEY, | |
1308 | .offset = 0, | |
1309 | }; | |
1310 | ||
1311 | return btrfs_global_root(fs_info, &key); | |
1312 | } | |
1313 | ||
20897f5c | 1314 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1315 | u64 objectid) |
1316 | { | |
9b7a2440 | 1317 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1318 | struct extent_buffer *leaf; |
1319 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1320 | struct btrfs_root *root; | |
1321 | struct btrfs_key key; | |
b89f6d1f | 1322 | unsigned int nofs_flag; |
20897f5c | 1323 | int ret = 0; |
20897f5c | 1324 | |
b89f6d1f FM |
1325 | /* |
1326 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1327 | * context to avoid deadlock if reclaim happens. | |
1328 | */ | |
1329 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1330 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1331 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1332 | if (!root) |
1333 | return ERR_PTR(-ENOMEM); | |
1334 | ||
20897f5c AJ |
1335 | root->root_key.objectid = objectid; |
1336 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1337 | root->root_key.offset = 0; | |
1338 | ||
9631e4cc JB |
1339 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1340 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1341 | if (IS_ERR(leaf)) { |
1342 | ret = PTR_ERR(leaf); | |
1dd05682 | 1343 | leaf = NULL; |
8a6a87cd | 1344 | goto fail_unlock; |
20897f5c AJ |
1345 | } |
1346 | ||
20897f5c | 1347 | root->node = leaf; |
20897f5c AJ |
1348 | btrfs_mark_buffer_dirty(leaf); |
1349 | ||
1350 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1351 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1352 | |
f944d2cb DS |
1353 | btrfs_set_root_flags(&root->root_item, 0); |
1354 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1355 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1356 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1357 | btrfs_set_root_level(&root->root_item, 0); | |
1358 | btrfs_set_root_refs(&root->root_item, 1); | |
1359 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1360 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1361 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1362 | if (is_fstree(objectid)) |
807fc790 AS |
1363 | generate_random_guid(root->root_item.uuid); |
1364 | else | |
1365 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1366 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1367 | |
8a6a87cd BB |
1368 | btrfs_tree_unlock(leaf); |
1369 | ||
20897f5c AJ |
1370 | key.objectid = objectid; |
1371 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1372 | key.offset = 0; | |
1373 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1374 | if (ret) | |
1375 | goto fail; | |
1376 | ||
1dd05682 TI |
1377 | return root; |
1378 | ||
8a6a87cd | 1379 | fail_unlock: |
8c38938c | 1380 | if (leaf) |
1dd05682 | 1381 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1382 | fail: |
00246528 | 1383 | btrfs_put_root(root); |
20897f5c | 1384 | |
1dd05682 | 1385 | return ERR_PTR(ret); |
20897f5c AJ |
1386 | } |
1387 | ||
7237f183 YZ |
1388 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1389 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1390 | { |
1391 | struct btrfs_root *root; | |
e02119d5 | 1392 | |
96dfcb46 | 1393 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1394 | if (!root) |
7237f183 | 1395 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1396 | |
e02119d5 CM |
1397 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1398 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1399 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1400 | |
6ab6ebb7 NA |
1401 | return root; |
1402 | } | |
1403 | ||
1404 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1405 | struct btrfs_root *root) | |
1406 | { | |
1407 | struct extent_buffer *leaf; | |
1408 | ||
7237f183 | 1409 | /* |
92a7cc42 | 1410 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1411 | * |
92a7cc42 QW |
1412 | * Log trees are not exposed to user space thus can't be snapshotted, |
1413 | * and they go away before a real commit is actually done. | |
1414 | * | |
1415 | * They do store pointers to file data extents, and those reference | |
1416 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1417 | */ |
e02119d5 | 1418 | |
4d75f8a9 | 1419 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1420 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1421 | if (IS_ERR(leaf)) |
1422 | return PTR_ERR(leaf); | |
e02119d5 | 1423 | |
7237f183 | 1424 | root->node = leaf; |
e02119d5 | 1425 | |
e02119d5 CM |
1426 | btrfs_mark_buffer_dirty(root->node); |
1427 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1428 | |
1429 | return 0; | |
7237f183 YZ |
1430 | } |
1431 | ||
1432 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1433 | struct btrfs_fs_info *fs_info) | |
1434 | { | |
1435 | struct btrfs_root *log_root; | |
1436 | ||
1437 | log_root = alloc_log_tree(trans, fs_info); | |
1438 | if (IS_ERR(log_root)) | |
1439 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1440 | |
3ddebf27 NA |
1441 | if (!btrfs_is_zoned(fs_info)) { |
1442 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1443 | ||
1444 | if (ret) { | |
1445 | btrfs_put_root(log_root); | |
1446 | return ret; | |
1447 | } | |
6ab6ebb7 NA |
1448 | } |
1449 | ||
7237f183 YZ |
1450 | WARN_ON(fs_info->log_root_tree); |
1451 | fs_info->log_root_tree = log_root; | |
1452 | return 0; | |
1453 | } | |
1454 | ||
1455 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1456 | struct btrfs_root *root) | |
1457 | { | |
0b246afa | 1458 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1459 | struct btrfs_root *log_root; |
1460 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1461 | int ret; |
7237f183 | 1462 | |
0b246afa | 1463 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1464 | if (IS_ERR(log_root)) |
1465 | return PTR_ERR(log_root); | |
1466 | ||
6ab6ebb7 NA |
1467 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1468 | if (ret) { | |
1469 | btrfs_put_root(log_root); | |
1470 | return ret; | |
1471 | } | |
1472 | ||
7237f183 YZ |
1473 | log_root->last_trans = trans->transid; |
1474 | log_root->root_key.offset = root->root_key.objectid; | |
1475 | ||
1476 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1477 | btrfs_set_stack_inode_generation(inode_item, 1); |
1478 | btrfs_set_stack_inode_size(inode_item, 3); | |
1479 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1480 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1481 | fs_info->nodesize); |
3cae210f | 1482 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1483 | |
5d4f98a2 | 1484 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1485 | |
1486 | WARN_ON(root->log_root); | |
1487 | root->log_root = log_root; | |
1488 | root->log_transid = 0; | |
d1433deb | 1489 | root->log_transid_committed = -1; |
257c62e1 | 1490 | root->last_log_commit = 0; |
e02119d5 CM |
1491 | return 0; |
1492 | } | |
1493 | ||
49d11bea JB |
1494 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1495 | struct btrfs_path *path, | |
1496 | struct btrfs_key *key) | |
e02119d5 CM |
1497 | { |
1498 | struct btrfs_root *root; | |
1499 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1500 | u64 generation; |
cb517eab | 1501 | int ret; |
581c1760 | 1502 | int level; |
0f7d52f4 | 1503 | |
96dfcb46 | 1504 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1505 | if (!root) |
1506 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1507 | |
cb517eab MX |
1508 | ret = btrfs_find_root(tree_root, key, path, |
1509 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1510 | if (ret) { |
13a8a7c8 YZ |
1511 | if (ret > 0) |
1512 | ret = -ENOENT; | |
49d11bea | 1513 | goto fail; |
0f7d52f4 | 1514 | } |
13a8a7c8 | 1515 | |
84234f3a | 1516 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1517 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1518 | root->node = read_tree_block(fs_info, |
1519 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1520 | key->objectid, generation, level, NULL); |
64c043de LB |
1521 | if (IS_ERR(root->node)) { |
1522 | ret = PTR_ERR(root->node); | |
8c38938c | 1523 | root->node = NULL; |
49d11bea | 1524 | goto fail; |
cb517eab MX |
1525 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { |
1526 | ret = -EIO; | |
49d11bea | 1527 | goto fail; |
416bc658 | 1528 | } |
5d4f98a2 | 1529 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1530 | return root; |
49d11bea | 1531 | fail: |
00246528 | 1532 | btrfs_put_root(root); |
49d11bea JB |
1533 | return ERR_PTR(ret); |
1534 | } | |
1535 | ||
1536 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1537 | struct btrfs_key *key) | |
1538 | { | |
1539 | struct btrfs_root *root; | |
1540 | struct btrfs_path *path; | |
1541 | ||
1542 | path = btrfs_alloc_path(); | |
1543 | if (!path) | |
1544 | return ERR_PTR(-ENOMEM); | |
1545 | root = read_tree_root_path(tree_root, path, key); | |
1546 | btrfs_free_path(path); | |
1547 | ||
1548 | return root; | |
cb517eab MX |
1549 | } |
1550 | ||
2dfb1e43 QW |
1551 | /* |
1552 | * Initialize subvolume root in-memory structure | |
1553 | * | |
1554 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1555 | */ | |
1556 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1557 | { |
1558 | int ret; | |
dcc3eb96 | 1559 | unsigned int nofs_flag; |
cb517eab | 1560 | |
dcc3eb96 NB |
1561 | /* |
1562 | * We might be called under a transaction (e.g. indirect backref | |
1563 | * resolution) which could deadlock if it triggers memory reclaim | |
1564 | */ | |
1565 | nofs_flag = memalloc_nofs_save(); | |
1566 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1567 | memalloc_nofs_restore(nofs_flag); | |
1568 | if (ret) | |
8257b2dc | 1569 | goto fail; |
8257b2dc | 1570 | |
aeb935a4 | 1571 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1572 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1573 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1574 | btrfs_check_and_init_root_item(&root->root_item); |
1575 | } | |
1576 | ||
851fd730 QW |
1577 | /* |
1578 | * Don't assign anonymous block device to roots that are not exposed to | |
1579 | * userspace, the id pool is limited to 1M | |
1580 | */ | |
1581 | if (is_fstree(root->root_key.objectid) && | |
1582 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1583 | if (!anon_dev) { |
1584 | ret = get_anon_bdev(&root->anon_dev); | |
1585 | if (ret) | |
1586 | goto fail; | |
1587 | } else { | |
1588 | root->anon_dev = anon_dev; | |
1589 | } | |
851fd730 | 1590 | } |
f32e48e9 CR |
1591 | |
1592 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1593 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1594 | if (ret) { |
1595 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1596 | goto fail; |
f32e48e9 CR |
1597 | } |
1598 | ||
6b8fad57 | 1599 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1600 | |
1601 | mutex_unlock(&root->objectid_mutex); | |
1602 | ||
cb517eab MX |
1603 | return 0; |
1604 | fail: | |
84db5ccf | 1605 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1606 | return ret; |
1607 | } | |
1608 | ||
a98db0f3 JB |
1609 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1610 | u64 root_id) | |
cb517eab MX |
1611 | { |
1612 | struct btrfs_root *root; | |
1613 | ||
1614 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1615 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1616 | (unsigned long)root_id); | |
bc44d7c4 | 1617 | if (root) |
00246528 | 1618 | root = btrfs_grab_root(root); |
cb517eab MX |
1619 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1620 | return root; | |
1621 | } | |
1622 | ||
49d11bea JB |
1623 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1624 | u64 objectid) | |
1625 | { | |
abed4aaa JB |
1626 | struct btrfs_key key = { |
1627 | .objectid = objectid, | |
1628 | .type = BTRFS_ROOT_ITEM_KEY, | |
1629 | .offset = 0, | |
1630 | }; | |
1631 | ||
49d11bea JB |
1632 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1633 | return btrfs_grab_root(fs_info->tree_root); | |
1634 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1635 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1636 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1637 | return btrfs_grab_root(fs_info->chunk_root); | |
1638 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1639 | return btrfs_grab_root(fs_info->dev_root); | |
1640 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1641 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1642 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1643 | return btrfs_grab_root(fs_info->quota_root) ? | |
1644 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1645 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1646 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1647 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1648 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1649 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1650 | ||
1651 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1652 | } | |
49d11bea JB |
1653 | return NULL; |
1654 | } | |
1655 | ||
cb517eab MX |
1656 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1657 | struct btrfs_root *root) | |
1658 | { | |
1659 | int ret; | |
1660 | ||
e1860a77 | 1661 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1662 | if (ret) |
1663 | return ret; | |
1664 | ||
1665 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1666 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1667 | (unsigned long)root->root_key.objectid, | |
1668 | root); | |
af01d2e5 | 1669 | if (ret == 0) { |
00246528 | 1670 | btrfs_grab_root(root); |
27cdeb70 | 1671 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1672 | } |
cb517eab MX |
1673 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1674 | radix_tree_preload_end(); | |
1675 | ||
1676 | return ret; | |
1677 | } | |
1678 | ||
bd647ce3 JB |
1679 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1680 | { | |
1681 | #ifdef CONFIG_BTRFS_DEBUG | |
1682 | struct btrfs_root *root; | |
1683 | ||
1684 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1685 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1686 | ||
bd647ce3 JB |
1687 | root = list_first_entry(&fs_info->allocated_roots, |
1688 | struct btrfs_root, leak_list); | |
457f1864 | 1689 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1690 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1691 | refcount_read(&root->refs)); |
1692 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1693 | btrfs_put_root(root); |
1694 | btrfs_put_root(root); | |
bd647ce3 JB |
1695 | } |
1696 | #endif | |
1697 | } | |
1698 | ||
abed4aaa JB |
1699 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1700 | { | |
1701 | struct btrfs_root *root; | |
1702 | struct rb_node *node; | |
1703 | ||
1704 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1705 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1706 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1707 | btrfs_put_root(root); | |
1708 | } | |
1709 | } | |
1710 | ||
0d4b0463 JB |
1711 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1712 | { | |
141386e1 JB |
1713 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1714 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1715 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1716 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1717 | btrfs_free_csum_hash(fs_info); | |
1718 | btrfs_free_stripe_hash_table(fs_info); | |
1719 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1720 | kfree(fs_info->balance_ctl); |
1721 | kfree(fs_info->delayed_root); | |
abed4aaa | 1722 | free_global_roots(fs_info); |
00246528 JB |
1723 | btrfs_put_root(fs_info->tree_root); |
1724 | btrfs_put_root(fs_info->chunk_root); | |
1725 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1726 | btrfs_put_root(fs_info->quota_root); |
1727 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1728 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1729 | btrfs_put_root(fs_info->data_reloc_root); |
bd647ce3 | 1730 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1731 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1732 | kfree(fs_info->super_copy); |
1733 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1734 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1735 | kvfree(fs_info); |
1736 | } | |
1737 | ||
1738 | ||
2dfb1e43 QW |
1739 | /* |
1740 | * Get an in-memory reference of a root structure. | |
1741 | * | |
1742 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1743 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1744 | * found, then read it from disk and add it to cached fs roots. | |
1745 | * | |
1746 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1747 | * | |
1748 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1749 | * same root objectid. | |
1750 | * | |
1751 | * @objectid: root id | |
1752 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1753 | * pass 0 for new allocation. | |
1754 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1755 | * for orphan roots | |
1756 | */ | |
1757 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1758 | u64 objectid, dev_t anon_dev, | |
1759 | bool check_ref) | |
5eda7b5e CM |
1760 | { |
1761 | struct btrfs_root *root; | |
381cf658 | 1762 | struct btrfs_path *path; |
1d4c08e0 | 1763 | struct btrfs_key key; |
5eda7b5e CM |
1764 | int ret; |
1765 | ||
49d11bea JB |
1766 | root = btrfs_get_global_root(fs_info, objectid); |
1767 | if (root) | |
1768 | return root; | |
4df27c4d | 1769 | again: |
56e9357a | 1770 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1771 | if (root) { |
2dfb1e43 QW |
1772 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1773 | ASSERT(!anon_dev); | |
bc44d7c4 | 1774 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1775 | btrfs_put_root(root); |
48475471 | 1776 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1777 | } |
5eda7b5e | 1778 | return root; |
48475471 | 1779 | } |
5eda7b5e | 1780 | |
56e9357a DS |
1781 | key.objectid = objectid; |
1782 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1783 | key.offset = (u64)-1; | |
1784 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1785 | if (IS_ERR(root)) |
1786 | return root; | |
3394e160 | 1787 | |
c00869f1 | 1788 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1789 | ret = -ENOENT; |
581bb050 | 1790 | goto fail; |
35a30d7c | 1791 | } |
581bb050 | 1792 | |
2dfb1e43 | 1793 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1794 | if (ret) |
1795 | goto fail; | |
3394e160 | 1796 | |
381cf658 DS |
1797 | path = btrfs_alloc_path(); |
1798 | if (!path) { | |
1799 | ret = -ENOMEM; | |
1800 | goto fail; | |
1801 | } | |
1d4c08e0 DS |
1802 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1803 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1804 | key.offset = objectid; |
1d4c08e0 DS |
1805 | |
1806 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1807 | btrfs_free_path(path); |
d68fc57b YZ |
1808 | if (ret < 0) |
1809 | goto fail; | |
1810 | if (ret == 0) | |
27cdeb70 | 1811 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1812 | |
cb517eab | 1813 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1814 | if (ret) { |
00246528 | 1815 | btrfs_put_root(root); |
4785e24f | 1816 | if (ret == -EEXIST) |
4df27c4d | 1817 | goto again; |
4df27c4d | 1818 | goto fail; |
0f7d52f4 | 1819 | } |
edbd8d4e | 1820 | return root; |
4df27c4d | 1821 | fail: |
33fab972 FM |
1822 | /* |
1823 | * If our caller provided us an anonymous device, then it's his | |
1824 | * responsability to free it in case we fail. So we have to set our | |
1825 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() | |
1826 | * and once again by our caller. | |
1827 | */ | |
1828 | if (anon_dev) | |
1829 | root->anon_dev = 0; | |
8c38938c | 1830 | btrfs_put_root(root); |
4df27c4d | 1831 | return ERR_PTR(ret); |
edbd8d4e CM |
1832 | } |
1833 | ||
2dfb1e43 QW |
1834 | /* |
1835 | * Get in-memory reference of a root structure | |
1836 | * | |
1837 | * @objectid: tree objectid | |
1838 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1839 | * one reference | |
1840 | */ | |
1841 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1842 | u64 objectid, bool check_ref) | |
1843 | { | |
1844 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1845 | } | |
1846 | ||
1847 | /* | |
1848 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1849 | * the anonymous block device id | |
1850 | * | |
1851 | * @objectid: tree objectid | |
1852 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1853 | * parameter value | |
1854 | */ | |
1855 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1856 | u64 objectid, dev_t anon_dev) | |
1857 | { | |
1858 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1859 | } | |
1860 | ||
49d11bea JB |
1861 | /* |
1862 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1863 | * @fs_info: the fs_info | |
1864 | * @objectid: the objectid we need to lookup | |
1865 | * | |
1866 | * This is exclusively used for backref walking, and exists specifically because | |
1867 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1868 | * creation time, which means we may have to read the tree_root in order to look | |
1869 | * up a fs root that is not in memory. If the root is not in memory we will | |
1870 | * read the tree root commit root and look up the fs root from there. This is a | |
1871 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1872 | * have the most uptodate information, it'll simply be discarded once the | |
1873 | * backref code is finished using the root. | |
1874 | */ | |
1875 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1876 | struct btrfs_path *path, | |
1877 | u64 objectid) | |
1878 | { | |
1879 | struct btrfs_root *root; | |
1880 | struct btrfs_key key; | |
1881 | ||
1882 | ASSERT(path->search_commit_root && path->skip_locking); | |
1883 | ||
1884 | /* | |
1885 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1886 | * since this is called via the backref walking code we won't be looking | |
1887 | * up a root that doesn't exist, unless there's corruption. So if root | |
1888 | * != NULL just return it. | |
1889 | */ | |
1890 | root = btrfs_get_global_root(fs_info, objectid); | |
1891 | if (root) | |
1892 | return root; | |
1893 | ||
1894 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1895 | if (root) | |
1896 | return root; | |
1897 | ||
1898 | key.objectid = objectid; | |
1899 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1900 | key.offset = (u64)-1; | |
1901 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1902 | btrfs_release_path(path); | |
1903 | ||
1904 | return root; | |
1905 | } | |
1906 | ||
8b712842 CM |
1907 | /* |
1908 | * called by the kthread helper functions to finally call the bio end_io | |
1909 | * functions. This is where read checksum verification actually happens | |
1910 | */ | |
1911 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1912 | { |
ce9adaa5 | 1913 | struct bio *bio; |
97eb6b69 | 1914 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1915 | |
97eb6b69 | 1916 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1917 | bio = end_io_wq->bio; |
ce9adaa5 | 1918 | |
4e4cbee9 | 1919 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1920 | bio->bi_private = end_io_wq->private; |
1921 | bio->bi_end_io = end_io_wq->end_io; | |
4246a0b6 | 1922 | bio_endio(bio); |
9be490f1 | 1923 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
44b8bd7e CM |
1924 | } |
1925 | ||
a74a4b97 CM |
1926 | static int cleaner_kthread(void *arg) |
1927 | { | |
1928 | struct btrfs_root *root = arg; | |
0b246afa | 1929 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1930 | int again; |
a74a4b97 | 1931 | |
d6fd0ae2 | 1932 | while (1) { |
d0278245 | 1933 | again = 0; |
a74a4b97 | 1934 | |
fd340d0f JB |
1935 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1936 | ||
d0278245 | 1937 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1938 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1939 | goto sleep; |
1940 | ||
90c711ab ZB |
1941 | /* |
1942 | * Do not do anything if we might cause open_ctree() to block | |
1943 | * before we have finished mounting the filesystem. | |
1944 | */ | |
0b246afa | 1945 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1946 | goto sleep; |
1947 | ||
0b246afa | 1948 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1949 | goto sleep; |
1950 | ||
dc7f370c MX |
1951 | /* |
1952 | * Avoid the problem that we change the status of the fs | |
1953 | * during the above check and trylock. | |
1954 | */ | |
2ff7e61e | 1955 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1956 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1957 | goto sleep; |
76dda93c | 1958 | } |
a74a4b97 | 1959 | |
2ff7e61e | 1960 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1961 | |
d0278245 | 1962 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1963 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1964 | |
1965 | /* | |
05323cd1 MX |
1966 | * The defragger has dealt with the R/O remount and umount, |
1967 | * needn't do anything special here. | |
d0278245 | 1968 | */ |
0b246afa | 1969 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1970 | |
1971 | /* | |
f3372065 | 1972 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1973 | * with relocation (btrfs_relocate_chunk) and relocation |
1974 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1975 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1976 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1977 | * unused block groups. | |
1978 | */ | |
0b246afa | 1979 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1980 | |
1981 | /* | |
1982 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1983 | * all unused block_groups. This possibly gives us some more free | |
1984 | * space. | |
1985 | */ | |
1986 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1987 | sleep: |
a0a1db70 | 1988 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1989 | if (kthread_should_park()) |
1990 | kthread_parkme(); | |
1991 | if (kthread_should_stop()) | |
1992 | return 0; | |
838fe188 | 1993 | if (!again) { |
a74a4b97 | 1994 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1995 | schedule(); |
a74a4b97 CM |
1996 | __set_current_state(TASK_RUNNING); |
1997 | } | |
da288d28 | 1998 | } |
a74a4b97 CM |
1999 | } |
2000 | ||
2001 | static int transaction_kthread(void *arg) | |
2002 | { | |
2003 | struct btrfs_root *root = arg; | |
0b246afa | 2004 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
2005 | struct btrfs_trans_handle *trans; |
2006 | struct btrfs_transaction *cur; | |
8929ecfa | 2007 | u64 transid; |
643900be | 2008 | time64_t delta; |
a74a4b97 | 2009 | unsigned long delay; |
914b2007 | 2010 | bool cannot_commit; |
a74a4b97 CM |
2011 | |
2012 | do { | |
914b2007 | 2013 | cannot_commit = false; |
ba1bc00f | 2014 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 2015 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 2016 | |
0b246afa JM |
2017 | spin_lock(&fs_info->trans_lock); |
2018 | cur = fs_info->running_transaction; | |
a74a4b97 | 2019 | if (!cur) { |
0b246afa | 2020 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
2021 | goto sleep; |
2022 | } | |
31153d81 | 2023 | |
643900be | 2024 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
2025 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
2026 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 2027 | delta < fs_info->commit_interval) { |
0b246afa | 2028 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
2029 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
2030 | delay = min(delay, | |
2031 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
2032 | goto sleep; |
2033 | } | |
8929ecfa | 2034 | transid = cur->transid; |
0b246afa | 2035 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 2036 | |
79787eaa | 2037 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 2038 | trans = btrfs_attach_transaction(root); |
914b2007 | 2039 | if (IS_ERR(trans)) { |
354aa0fb MX |
2040 | if (PTR_ERR(trans) != -ENOENT) |
2041 | cannot_commit = true; | |
79787eaa | 2042 | goto sleep; |
914b2007 | 2043 | } |
8929ecfa | 2044 | if (transid == trans->transid) { |
3a45bb20 | 2045 | btrfs_commit_transaction(trans); |
8929ecfa | 2046 | } else { |
3a45bb20 | 2047 | btrfs_end_transaction(trans); |
8929ecfa | 2048 | } |
a74a4b97 | 2049 | sleep: |
0b246afa JM |
2050 | wake_up_process(fs_info->cleaner_kthread); |
2051 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 2052 | |
84961539 | 2053 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 2054 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 2055 | if (!kthread_should_stop() && |
0b246afa | 2056 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 2057 | cannot_commit)) |
bc5511d0 | 2058 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
2059 | } while (!kthread_should_stop()); |
2060 | return 0; | |
2061 | } | |
2062 | ||
af31f5e5 | 2063 | /* |
01f0f9da NB |
2064 | * This will find the highest generation in the array of root backups. The |
2065 | * index of the highest array is returned, or -EINVAL if we can't find | |
2066 | * anything. | |
af31f5e5 CM |
2067 | * |
2068 | * We check to make sure the array is valid by comparing the | |
2069 | * generation of the latest root in the array with the generation | |
2070 | * in the super block. If they don't match we pitch it. | |
2071 | */ | |
01f0f9da | 2072 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 2073 | { |
01f0f9da | 2074 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 2075 | u64 cur; |
af31f5e5 CM |
2076 | struct btrfs_root_backup *root_backup; |
2077 | int i; | |
2078 | ||
2079 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2080 | root_backup = info->super_copy->super_roots + i; | |
2081 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2082 | if (cur == newest_gen) | |
01f0f9da | 2083 | return i; |
af31f5e5 CM |
2084 | } |
2085 | ||
01f0f9da | 2086 | return -EINVAL; |
af31f5e5 CM |
2087 | } |
2088 | ||
af31f5e5 CM |
2089 | /* |
2090 | * copy all the root pointers into the super backup array. | |
2091 | * this will bump the backup pointer by one when it is | |
2092 | * done | |
2093 | */ | |
2094 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2095 | { | |
6ef108dd | 2096 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2097 | struct btrfs_root_backup *root_backup; |
29cbcf40 | 2098 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
fc28b25e | 2099 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
af31f5e5 CM |
2100 | |
2101 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2102 | ||
2103 | /* | |
2104 | * make sure all of our padding and empty slots get zero filled | |
2105 | * regardless of which ones we use today | |
2106 | */ | |
2107 | memset(root_backup, 0, sizeof(*root_backup)); | |
2108 | ||
2109 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2110 | ||
2111 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2112 | btrfs_set_backup_tree_root_gen(root_backup, | |
2113 | btrfs_header_generation(info->tree_root->node)); | |
2114 | ||
2115 | btrfs_set_backup_tree_root_level(root_backup, | |
2116 | btrfs_header_level(info->tree_root->node)); | |
2117 | ||
2118 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2119 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2120 | btrfs_header_generation(info->chunk_root->node)); | |
2121 | btrfs_set_backup_chunk_root_level(root_backup, | |
2122 | btrfs_header_level(info->chunk_root->node)); | |
2123 | ||
29cbcf40 | 2124 | btrfs_set_backup_extent_root(root_backup, extent_root->node->start); |
af31f5e5 | 2125 | btrfs_set_backup_extent_root_gen(root_backup, |
29cbcf40 | 2126 | btrfs_header_generation(extent_root->node)); |
af31f5e5 | 2127 | btrfs_set_backup_extent_root_level(root_backup, |
29cbcf40 | 2128 | btrfs_header_level(extent_root->node)); |
af31f5e5 | 2129 | |
7c7e82a7 CM |
2130 | /* |
2131 | * we might commit during log recovery, which happens before we set | |
2132 | * the fs_root. Make sure it is valid before we fill it in. | |
2133 | */ | |
2134 | if (info->fs_root && info->fs_root->node) { | |
2135 | btrfs_set_backup_fs_root(root_backup, | |
2136 | info->fs_root->node->start); | |
2137 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2138 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2139 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2140 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2141 | } |
af31f5e5 CM |
2142 | |
2143 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2144 | btrfs_set_backup_dev_root_gen(root_backup, | |
2145 | btrfs_header_generation(info->dev_root->node)); | |
2146 | btrfs_set_backup_dev_root_level(root_backup, | |
2147 | btrfs_header_level(info->dev_root->node)); | |
2148 | ||
fc28b25e | 2149 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); |
af31f5e5 | 2150 | btrfs_set_backup_csum_root_gen(root_backup, |
fc28b25e | 2151 | btrfs_header_generation(csum_root->node)); |
af31f5e5 | 2152 | btrfs_set_backup_csum_root_level(root_backup, |
fc28b25e | 2153 | btrfs_header_level(csum_root->node)); |
af31f5e5 CM |
2154 | |
2155 | btrfs_set_backup_total_bytes(root_backup, | |
2156 | btrfs_super_total_bytes(info->super_copy)); | |
2157 | btrfs_set_backup_bytes_used(root_backup, | |
2158 | btrfs_super_bytes_used(info->super_copy)); | |
2159 | btrfs_set_backup_num_devices(root_backup, | |
2160 | btrfs_super_num_devices(info->super_copy)); | |
2161 | ||
2162 | /* | |
2163 | * if we don't copy this out to the super_copy, it won't get remembered | |
2164 | * for the next commit | |
2165 | */ | |
2166 | memcpy(&info->super_copy->super_roots, | |
2167 | &info->super_for_commit->super_roots, | |
2168 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2169 | } | |
2170 | ||
bd2336b2 NB |
2171 | /* |
2172 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2173 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2174 | * | |
2175 | * fs_info - filesystem whose backup roots need to be read | |
2176 | * priority - priority of backup root required | |
2177 | * | |
2178 | * Returns backup root index on success and -EINVAL otherwise. | |
2179 | */ | |
2180 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2181 | { | |
2182 | int backup_index = find_newest_super_backup(fs_info); | |
2183 | struct btrfs_super_block *super = fs_info->super_copy; | |
2184 | struct btrfs_root_backup *root_backup; | |
2185 | ||
2186 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2187 | if (priority == 0) | |
2188 | return backup_index; | |
2189 | ||
2190 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2191 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2192 | } else { | |
2193 | return -EINVAL; | |
2194 | } | |
2195 | ||
2196 | root_backup = super->super_roots + backup_index; | |
2197 | ||
2198 | btrfs_set_super_generation(super, | |
2199 | btrfs_backup_tree_root_gen(root_backup)); | |
2200 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2201 | btrfs_set_super_root_level(super, | |
2202 | btrfs_backup_tree_root_level(root_backup)); | |
2203 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2204 | ||
2205 | /* | |
2206 | * Fixme: the total bytes and num_devices need to match or we should | |
2207 | * need a fsck | |
2208 | */ | |
2209 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2210 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2211 | ||
2212 | return backup_index; | |
2213 | } | |
2214 | ||
7abadb64 LB |
2215 | /* helper to cleanup workers */ |
2216 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2217 | { | |
dc6e3209 | 2218 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2219 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2220 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2221 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2222 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
d05a33ac | 2223 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2224 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2225 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2226 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2227 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2228 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2229 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2230 | if (fs_info->discard_ctl.discard_workers) |
2231 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2232 | /* |
2233 | * Now that all other work queues are destroyed, we can safely destroy | |
2234 | * the queues used for metadata I/O, since tasks from those other work | |
2235 | * queues can do metadata I/O operations. | |
2236 | */ | |
2237 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2238 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2239 | } |
2240 | ||
2e9f5954 R |
2241 | static void free_root_extent_buffers(struct btrfs_root *root) |
2242 | { | |
2243 | if (root) { | |
2244 | free_extent_buffer(root->node); | |
2245 | free_extent_buffer(root->commit_root); | |
2246 | root->node = NULL; | |
2247 | root->commit_root = NULL; | |
2248 | } | |
2249 | } | |
2250 | ||
abed4aaa JB |
2251 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2252 | { | |
2253 | struct btrfs_root *root, *tmp; | |
2254 | ||
2255 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2256 | &fs_info->global_root_tree, | |
2257 | rb_node) | |
2258 | free_root_extent_buffers(root); | |
2259 | } | |
2260 | ||
af31f5e5 | 2261 | /* helper to cleanup tree roots */ |
4273eaff | 2262 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2263 | { |
2e9f5954 | 2264 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2265 | |
abed4aaa | 2266 | free_global_root_pointers(info); |
2e9f5954 | 2267 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2268 | free_root_extent_buffers(info->quota_root); |
2269 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2270 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2271 | free_root_extent_buffers(info->data_reloc_root); |
4273eaff | 2272 | if (free_chunk_root) |
2e9f5954 | 2273 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2274 | } |
2275 | ||
8c38938c JB |
2276 | void btrfs_put_root(struct btrfs_root *root) |
2277 | { | |
2278 | if (!root) | |
2279 | return; | |
2280 | ||
2281 | if (refcount_dec_and_test(&root->refs)) { | |
2282 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2283 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2284 | if (root->anon_dev) |
2285 | free_anon_bdev(root->anon_dev); | |
2286 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2287 | free_root_extent_buffers(root); |
8c38938c JB |
2288 | #ifdef CONFIG_BTRFS_DEBUG |
2289 | spin_lock(&root->fs_info->fs_roots_radix_lock); | |
2290 | list_del_init(&root->leak_list); | |
2291 | spin_unlock(&root->fs_info->fs_roots_radix_lock); | |
2292 | #endif | |
2293 | kfree(root); | |
2294 | } | |
2295 | } | |
2296 | ||
faa2dbf0 | 2297 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2298 | { |
2299 | int ret; | |
2300 | struct btrfs_root *gang[8]; | |
2301 | int i; | |
2302 | ||
2303 | while (!list_empty(&fs_info->dead_roots)) { | |
2304 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2305 | struct btrfs_root, root_list); | |
2306 | list_del(&gang[0]->root_list); | |
2307 | ||
8c38938c | 2308 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
cb517eab | 2309 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
dc9492c1 | 2310 | btrfs_put_root(gang[0]); |
171f6537 JB |
2311 | } |
2312 | ||
2313 | while (1) { | |
2314 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2315 | (void **)gang, 0, | |
2316 | ARRAY_SIZE(gang)); | |
2317 | if (!ret) | |
2318 | break; | |
2319 | for (i = 0; i < ret; i++) | |
cb517eab | 2320 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 JB |
2321 | } |
2322 | } | |
af31f5e5 | 2323 | |
638aa7ed ES |
2324 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2325 | { | |
2326 | mutex_init(&fs_info->scrub_lock); | |
2327 | atomic_set(&fs_info->scrubs_running, 0); | |
2328 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2329 | atomic_set(&fs_info->scrubs_paused, 0); | |
2330 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2331 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2332 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2333 | } |
2334 | ||
779a65a4 ES |
2335 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2336 | { | |
2337 | spin_lock_init(&fs_info->balance_lock); | |
2338 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2339 | atomic_set(&fs_info->balance_pause_req, 0); |
2340 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2341 | fs_info->balance_ctl = NULL; | |
2342 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2343 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2344 | } |
2345 | ||
6bccf3ab | 2346 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2347 | { |
2ff7e61e JM |
2348 | struct inode *inode = fs_info->btree_inode; |
2349 | ||
2350 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2351 | set_nlink(inode, 1); | |
f37938e0 ES |
2352 | /* |
2353 | * we set the i_size on the btree inode to the max possible int. | |
2354 | * the real end of the address space is determined by all of | |
2355 | * the devices in the system | |
2356 | */ | |
2ff7e61e JM |
2357 | inode->i_size = OFFSET_MAX; |
2358 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2359 | |
2ff7e61e | 2360 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2361 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
2c53a14d | 2362 | IO_TREE_BTREE_INODE_IO, inode); |
7b439738 | 2363 | BTRFS_I(inode)->io_tree.track_uptodate = false; |
2ff7e61e | 2364 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2365 | |
5c8fd99f | 2366 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
2ff7e61e JM |
2367 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); |
2368 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2369 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2370 | } |
2371 | ||
ad618368 ES |
2372 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2373 | { | |
ad618368 | 2374 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2375 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2376 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2377 | } |
2378 | ||
f9e92e40 ES |
2379 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2380 | { | |
2381 | spin_lock_init(&fs_info->qgroup_lock); | |
2382 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2383 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2384 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2385 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2386 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2387 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2388 | mutex_init(&fs_info->qgroup_rescan_lock); |
2389 | } | |
2390 | ||
d21deec5 | 2391 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2392 | { |
f7b885be | 2393 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2394 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2395 | |
2396 | fs_info->workers = | |
cb001095 JM |
2397 | btrfs_alloc_workqueue(fs_info, "worker", |
2398 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2399 | |
2400 | fs_info->delalloc_workers = | |
cb001095 JM |
2401 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2402 | flags, max_active, 2); | |
2a458198 ES |
2403 | |
2404 | fs_info->flush_workers = | |
cb001095 JM |
2405 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2406 | flags, max_active, 0); | |
2a458198 ES |
2407 | |
2408 | fs_info->caching_workers = | |
cb001095 | 2409 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2410 | |
2a458198 | 2411 | fs_info->fixup_workers = |
cb001095 | 2412 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2413 | |
2414 | /* | |
2415 | * endios are largely parallel and should have a very | |
2416 | * low idle thresh | |
2417 | */ | |
2418 | fs_info->endio_workers = | |
cb001095 | 2419 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2420 | fs_info->endio_meta_workers = |
cb001095 JM |
2421 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2422 | max_active, 4); | |
2a458198 | 2423 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2424 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2425 | max_active, 2); | |
2a458198 | 2426 | fs_info->endio_raid56_workers = |
cb001095 JM |
2427 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2428 | max_active, 4); | |
2a458198 | 2429 | fs_info->rmw_workers = |
cb001095 | 2430 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2431 | fs_info->endio_write_workers = |
cb001095 JM |
2432 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2433 | max_active, 2); | |
2a458198 | 2434 | fs_info->endio_freespace_worker = |
cb001095 JM |
2435 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2436 | max_active, 0); | |
2a458198 | 2437 | fs_info->delayed_workers = |
cb001095 JM |
2438 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2439 | max_active, 0); | |
2a458198 | 2440 | fs_info->qgroup_rescan_workers = |
cb001095 | 2441 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2442 | fs_info->discard_ctl.discard_workers = |
2443 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 ES |
2444 | |
2445 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
ba8a9d07 | 2446 | fs_info->flush_workers && |
2a458198 ES |
2447 | fs_info->endio_workers && fs_info->endio_meta_workers && |
2448 | fs_info->endio_meta_write_workers && | |
2a458198 ES |
2449 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2450 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
f26c9238 QW |
2451 | fs_info->caching_workers && fs_info->fixup_workers && |
2452 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2453 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2454 | return -ENOMEM; |
2455 | } | |
2456 | ||
2457 | return 0; | |
2458 | } | |
2459 | ||
6d97c6e3 JT |
2460 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2461 | { | |
2462 | struct crypto_shash *csum_shash; | |
b4e967be | 2463 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2464 | |
b4e967be | 2465 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2466 | |
2467 | if (IS_ERR(csum_shash)) { | |
2468 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2469 | csum_driver); |
6d97c6e3 JT |
2470 | return PTR_ERR(csum_shash); |
2471 | } | |
2472 | ||
2473 | fs_info->csum_shash = csum_shash; | |
2474 | ||
2475 | return 0; | |
2476 | } | |
2477 | ||
63443bf5 ES |
2478 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2479 | struct btrfs_fs_devices *fs_devices) | |
2480 | { | |
2481 | int ret; | |
63443bf5 ES |
2482 | struct btrfs_root *log_tree_root; |
2483 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2484 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2485 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2486 | |
2487 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2488 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2489 | return -EIO; |
2490 | } | |
2491 | ||
96dfcb46 JB |
2492 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2493 | GFP_KERNEL); | |
63443bf5 ES |
2494 | if (!log_tree_root) |
2495 | return -ENOMEM; | |
2496 | ||
2ff7e61e | 2497 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2498 | BTRFS_TREE_LOG_OBJECTID, |
2499 | fs_info->generation + 1, level, | |
2500 | NULL); | |
64c043de | 2501 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2502 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2503 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2504 | log_tree_root->node = NULL; |
00246528 | 2505 | btrfs_put_root(log_tree_root); |
0eeff236 | 2506 | return ret; |
64c043de | 2507 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2508 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2509 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2510 | return -EIO; |
2511 | } | |
2512 | /* returns with log_tree_root freed on success */ | |
2513 | ret = btrfs_recover_log_trees(log_tree_root); | |
2514 | if (ret) { | |
0b246afa JM |
2515 | btrfs_handle_fs_error(fs_info, ret, |
2516 | "Failed to recover log tree"); | |
00246528 | 2517 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2518 | return ret; |
2519 | } | |
2520 | ||
bc98a42c | 2521 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2522 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2523 | if (ret) |
2524 | return ret; | |
2525 | } | |
2526 | ||
2527 | return 0; | |
2528 | } | |
2529 | ||
abed4aaa JB |
2530 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2531 | struct btrfs_path *path, u64 objectid, | |
2532 | const char *name) | |
2533 | { | |
2534 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2535 | struct btrfs_root *root; | |
2536 | int ret; | |
2537 | struct btrfs_key key = { | |
2538 | .objectid = objectid, | |
2539 | .type = BTRFS_ROOT_ITEM_KEY, | |
2540 | .offset = 0, | |
2541 | }; | |
2542 | bool found = false; | |
2543 | ||
2544 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2545 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2546 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2547 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2548 | return 0; | |
2549 | } | |
2550 | ||
2551 | while (1) { | |
2552 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2553 | if (ret < 0) | |
2554 | break; | |
2555 | ||
2556 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2557 | ret = btrfs_next_leaf(tree_root, path); | |
2558 | if (ret) { | |
2559 | if (ret > 0) | |
2560 | ret = 0; | |
2561 | break; | |
2562 | } | |
2563 | } | |
2564 | ret = 0; | |
2565 | ||
2566 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2567 | if (key.objectid != objectid) | |
2568 | break; | |
2569 | btrfs_release_path(path); | |
2570 | ||
2571 | found = true; | |
2572 | root = read_tree_root_path(tree_root, path, &key); | |
2573 | if (IS_ERR(root)) { | |
2574 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2575 | ret = PTR_ERR(root); | |
2576 | break; | |
2577 | } | |
2578 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2579 | ret = btrfs_global_root_insert(root); | |
2580 | if (ret) { | |
2581 | btrfs_put_root(root); | |
2582 | break; | |
2583 | } | |
2584 | key.offset++; | |
2585 | } | |
2586 | btrfs_release_path(path); | |
2587 | ||
2588 | if (!found || ret) { | |
2589 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2590 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2591 | ||
2592 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2593 | ret = ret ? ret : -ENOENT; | |
2594 | else | |
2595 | ret = 0; | |
2596 | btrfs_err(fs_info, "failed to load root %s", name); | |
2597 | } | |
2598 | return ret; | |
2599 | } | |
2600 | ||
2601 | static int load_global_roots(struct btrfs_root *tree_root) | |
2602 | { | |
2603 | struct btrfs_path *path; | |
2604 | int ret = 0; | |
2605 | ||
2606 | path = btrfs_alloc_path(); | |
2607 | if (!path) | |
2608 | return -ENOMEM; | |
2609 | ||
2610 | ret = load_global_roots_objectid(tree_root, path, | |
2611 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2612 | if (ret) | |
2613 | goto out; | |
2614 | ret = load_global_roots_objectid(tree_root, path, | |
2615 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2616 | if (ret) | |
2617 | goto out; | |
2618 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2619 | goto out; | |
2620 | ret = load_global_roots_objectid(tree_root, path, | |
2621 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2622 | "free space"); | |
2623 | out: | |
2624 | btrfs_free_path(path); | |
2625 | return ret; | |
2626 | } | |
2627 | ||
6bccf3ab | 2628 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2629 | { |
6bccf3ab | 2630 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2631 | struct btrfs_root *root; |
4bbcaa64 ES |
2632 | struct btrfs_key location; |
2633 | int ret; | |
2634 | ||
6bccf3ab JM |
2635 | BUG_ON(!fs_info->tree_root); |
2636 | ||
abed4aaa JB |
2637 | ret = load_global_roots(tree_root); |
2638 | if (ret) | |
2639 | return ret; | |
2640 | ||
2641 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
4bbcaa64 ES |
2642 | location.type = BTRFS_ROOT_ITEM_KEY; |
2643 | location.offset = 0; | |
2644 | ||
a4f3d2c4 | 2645 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2646 | if (IS_ERR(root)) { |
42437a63 JB |
2647 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2648 | ret = PTR_ERR(root); | |
2649 | goto out; | |
2650 | } | |
2651 | } else { | |
2652 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2653 | fs_info->dev_root = root; | |
f50f4353 | 2654 | } |
820a49da JB |
2655 | /* Initialize fs_info for all devices in any case */ |
2656 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2657 | |
aeb935a4 QW |
2658 | /* |
2659 | * This tree can share blocks with some other fs tree during relocation | |
2660 | * and we need a proper setup by btrfs_get_fs_root | |
2661 | */ | |
56e9357a DS |
2662 | root = btrfs_get_fs_root(tree_root->fs_info, |
2663 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2664 | if (IS_ERR(root)) { |
42437a63 JB |
2665 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2666 | ret = PTR_ERR(root); | |
2667 | goto out; | |
2668 | } | |
2669 | } else { | |
2670 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2671 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2672 | } |
aeb935a4 | 2673 | |
4bbcaa64 | 2674 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2675 | root = btrfs_read_tree_root(tree_root, &location); |
2676 | if (!IS_ERR(root)) { | |
2677 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2678 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2679 | fs_info->quota_root = root; |
4bbcaa64 ES |
2680 | } |
2681 | ||
2682 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2683 | root = btrfs_read_tree_root(tree_root, &location); |
2684 | if (IS_ERR(root)) { | |
42437a63 JB |
2685 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2686 | ret = PTR_ERR(root); | |
2687 | if (ret != -ENOENT) | |
2688 | goto out; | |
2689 | } | |
4bbcaa64 | 2690 | } else { |
a4f3d2c4 DS |
2691 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2692 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2693 | } |
2694 | ||
2695 | return 0; | |
f50f4353 LB |
2696 | out: |
2697 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2698 | location.objectid, ret); | |
2699 | return ret; | |
4bbcaa64 ES |
2700 | } |
2701 | ||
069ec957 QW |
2702 | /* |
2703 | * Real super block validation | |
2704 | * NOTE: super csum type and incompat features will not be checked here. | |
2705 | * | |
2706 | * @sb: super block to check | |
2707 | * @mirror_num: the super block number to check its bytenr: | |
2708 | * 0 the primary (1st) sb | |
2709 | * 1, 2 2nd and 3rd backup copy | |
2710 | * -1 skip bytenr check | |
2711 | */ | |
2712 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2713 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2714 | { |
21a852b0 QW |
2715 | u64 nodesize = btrfs_super_nodesize(sb); |
2716 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2717 | int ret = 0; | |
2718 | ||
2719 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2720 | btrfs_err(fs_info, "no valid FS found"); | |
2721 | ret = -EINVAL; | |
2722 | } | |
2723 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2724 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2725 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2726 | ret = -EINVAL; | |
2727 | } | |
2728 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2729 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2730 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2731 | ret = -EINVAL; | |
2732 | } | |
2733 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2734 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2735 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2736 | ret = -EINVAL; | |
2737 | } | |
2738 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2739 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2740 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2741 | ret = -EINVAL; | |
2742 | } | |
2743 | ||
2744 | /* | |
2745 | * Check sectorsize and nodesize first, other check will need it. | |
2746 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2747 | */ | |
2748 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2749 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2750 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2751 | ret = -EINVAL; | |
2752 | } | |
0bb3eb3e QW |
2753 | |
2754 | /* | |
2755 | * For 4K page size, we only support 4K sector size. | |
50780d9b | 2756 | * For 64K page size, we support 64K and 4K sector sizes. |
0bb3eb3e QW |
2757 | */ |
2758 | if ((PAGE_SIZE == SZ_4K && sectorsize != PAGE_SIZE) || | |
2759 | (PAGE_SIZE == SZ_64K && (sectorsize != SZ_4K && | |
2760 | sectorsize != SZ_64K))) { | |
21a852b0 | 2761 | btrfs_err(fs_info, |
0bb3eb3e | 2762 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2763 | sectorsize, PAGE_SIZE); |
2764 | ret = -EINVAL; | |
2765 | } | |
0bb3eb3e | 2766 | |
21a852b0 QW |
2767 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2768 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2769 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2770 | ret = -EINVAL; | |
2771 | } | |
2772 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2773 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2774 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2775 | ret = -EINVAL; | |
2776 | } | |
2777 | ||
2778 | /* Root alignment check */ | |
2779 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2780 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2781 | btrfs_super_root(sb)); | |
2782 | ret = -EINVAL; | |
2783 | } | |
2784 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2785 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2786 | btrfs_super_chunk_root(sb)); | |
2787 | ret = -EINVAL; | |
2788 | } | |
2789 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2790 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2791 | btrfs_super_log_root(sb)); | |
2792 | ret = -EINVAL; | |
2793 | } | |
2794 | ||
aefd7f70 NB |
2795 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2796 | BTRFS_FSID_SIZE)) { | |
2797 | btrfs_err(fs_info, | |
2798 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2799 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2800 | ret = -EINVAL; | |
2801 | } | |
2802 | ||
2803 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2804 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2805 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2806 | btrfs_err(fs_info, | |
2807 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2808 | fs_info->super_copy->metadata_uuid, | |
2809 | fs_info->fs_devices->metadata_uuid); | |
2810 | ret = -EINVAL; | |
2811 | } | |
2812 | ||
de37aa51 | 2813 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2814 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2815 | btrfs_err(fs_info, |
7239ff4b | 2816 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2817 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2818 | ret = -EINVAL; |
2819 | } | |
2820 | ||
2821 | /* | |
2822 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2823 | * done later | |
2824 | */ | |
2825 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2826 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2827 | btrfs_super_bytes_used(sb)); | |
2828 | ret = -EINVAL; | |
2829 | } | |
2830 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2831 | btrfs_err(fs_info, "invalid stripesize %u", | |
2832 | btrfs_super_stripesize(sb)); | |
2833 | ret = -EINVAL; | |
2834 | } | |
2835 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2836 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2837 | btrfs_super_num_devices(sb)); | |
2838 | if (btrfs_super_num_devices(sb) == 0) { | |
2839 | btrfs_err(fs_info, "number of devices is 0"); | |
2840 | ret = -EINVAL; | |
2841 | } | |
2842 | ||
069ec957 QW |
2843 | if (mirror_num >= 0 && |
2844 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2845 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2846 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2847 | ret = -EINVAL; | |
2848 | } | |
2849 | ||
2850 | /* | |
2851 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2852 | * and one chunk | |
2853 | */ | |
2854 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2855 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2856 | btrfs_super_sys_array_size(sb), | |
2857 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2858 | ret = -EINVAL; | |
2859 | } | |
2860 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2861 | + sizeof(struct btrfs_chunk)) { | |
2862 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2863 | btrfs_super_sys_array_size(sb), | |
2864 | sizeof(struct btrfs_disk_key) | |
2865 | + sizeof(struct btrfs_chunk)); | |
2866 | ret = -EINVAL; | |
2867 | } | |
2868 | ||
2869 | /* | |
2870 | * The generation is a global counter, we'll trust it more than the others | |
2871 | * but it's still possible that it's the one that's wrong. | |
2872 | */ | |
2873 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2874 | btrfs_warn(fs_info, | |
2875 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2876 | btrfs_super_generation(sb), | |
2877 | btrfs_super_chunk_root_generation(sb)); | |
2878 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2879 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2880 | btrfs_warn(fs_info, | |
2881 | "suspicious: generation < cache_generation: %llu < %llu", | |
2882 | btrfs_super_generation(sb), | |
2883 | btrfs_super_cache_generation(sb)); | |
2884 | ||
2885 | return ret; | |
2886 | } | |
2887 | ||
069ec957 QW |
2888 | /* |
2889 | * Validation of super block at mount time. | |
2890 | * Some checks already done early at mount time, like csum type and incompat | |
2891 | * flags will be skipped. | |
2892 | */ | |
2893 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2894 | { | |
2895 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2896 | } | |
2897 | ||
75cb857d QW |
2898 | /* |
2899 | * Validation of super block at write time. | |
2900 | * Some checks like bytenr check will be skipped as their values will be | |
2901 | * overwritten soon. | |
2902 | * Extra checks like csum type and incompat flags will be done here. | |
2903 | */ | |
2904 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2905 | struct btrfs_super_block *sb) | |
2906 | { | |
2907 | int ret; | |
2908 | ||
2909 | ret = validate_super(fs_info, sb, -1); | |
2910 | if (ret < 0) | |
2911 | goto out; | |
e7e16f48 | 2912 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2913 | ret = -EUCLEAN; |
2914 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2915 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2916 | goto out; | |
2917 | } | |
2918 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2919 | ret = -EUCLEAN; | |
2920 | btrfs_err(fs_info, | |
2921 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2922 | btrfs_super_incompat_flags(sb), | |
2923 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2924 | goto out; | |
2925 | } | |
2926 | out: | |
2927 | if (ret < 0) | |
2928 | btrfs_err(fs_info, | |
2929 | "super block corruption detected before writing it to disk"); | |
2930 | return ret; | |
2931 | } | |
2932 | ||
6ef108dd | 2933 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2934 | { |
6ef108dd | 2935 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2936 | struct btrfs_super_block *sb = fs_info->super_copy; |
2937 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2938 | bool handle_error = false; | |
2939 | int ret = 0; | |
2940 | int i; | |
2941 | ||
2942 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2943 | u64 generation; | |
2944 | int level; | |
2945 | ||
2946 | if (handle_error) { | |
2947 | if (!IS_ERR(tree_root->node)) | |
2948 | free_extent_buffer(tree_root->node); | |
2949 | tree_root->node = NULL; | |
2950 | ||
2951 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2952 | break; | |
2953 | ||
2954 | free_root_pointers(fs_info, 0); | |
2955 | ||
2956 | /* | |
2957 | * Don't use the log in recovery mode, it won't be | |
2958 | * valid | |
2959 | */ | |
2960 | btrfs_set_super_log_root(sb, 0); | |
2961 | ||
2962 | /* We can't trust the free space cache either */ | |
2963 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2964 | ||
2965 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2966 | backup_index = ret; |
b8522a1e NB |
2967 | if (ret < 0) |
2968 | return ret; | |
2969 | } | |
2970 | generation = btrfs_super_generation(sb); | |
2971 | level = btrfs_super_root_level(sb); | |
2972 | tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), | |
1b7ec85e | 2973 | BTRFS_ROOT_TREE_OBJECTID, |
b8522a1e | 2974 | generation, level, NULL); |
217f5004 | 2975 | if (IS_ERR(tree_root->node)) { |
b8522a1e | 2976 | handle_error = true; |
217f5004 NB |
2977 | ret = PTR_ERR(tree_root->node); |
2978 | tree_root->node = NULL; | |
2979 | btrfs_warn(fs_info, "couldn't read tree root"); | |
2980 | continue; | |
b8522a1e | 2981 | |
217f5004 NB |
2982 | } else if (!extent_buffer_uptodate(tree_root->node)) { |
2983 | handle_error = true; | |
2984 | ret = -EIO; | |
2985 | btrfs_warn(fs_info, "error while reading tree root"); | |
b8522a1e NB |
2986 | continue; |
2987 | } | |
2988 | ||
2989 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); | |
2990 | tree_root->commit_root = btrfs_root_node(tree_root); | |
2991 | btrfs_set_root_refs(&tree_root->root_item, 1); | |
2992 | ||
336a0d8d NB |
2993 | /* |
2994 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2995 | * hasn't been fully initialised and we are the only user | |
2996 | */ | |
453e4873 | 2997 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2998 | if (ret < 0) { |
b8522a1e NB |
2999 | handle_error = true; |
3000 | continue; | |
3001 | } | |
3002 | ||
6b8fad57 | 3003 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
3004 | |
3005 | ret = btrfs_read_roots(fs_info); | |
3006 | if (ret < 0) { | |
3007 | handle_error = true; | |
3008 | continue; | |
3009 | } | |
3010 | ||
3011 | /* All successful */ | |
3012 | fs_info->generation = generation; | |
3013 | fs_info->last_trans_committed = generation; | |
d96b3424 | 3014 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
3015 | |
3016 | /* Always begin writing backup roots after the one being used */ | |
3017 | if (backup_index < 0) { | |
3018 | fs_info->backup_root_index = 0; | |
3019 | } else { | |
3020 | fs_info->backup_root_index = backup_index + 1; | |
3021 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
3022 | } | |
b8522a1e NB |
3023 | break; |
3024 | } | |
3025 | ||
3026 | return ret; | |
3027 | } | |
3028 | ||
8260edba | 3029 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 3030 | { |
76dda93c | 3031 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 3032 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 3033 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 3034 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 3035 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 3036 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 3037 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 3038 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 3039 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 3040 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 3041 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 3042 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 3043 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 3044 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 3045 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 3046 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 3047 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 3048 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 3049 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 3050 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 3051 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 3052 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 3053 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 3054 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 3055 | mutex_init(&fs_info->zoned_meta_io_lock); |
de98ced9 | 3056 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 3057 | |
0b86a832 | 3058 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3059 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3060 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3061 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3062 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3063 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3064 | #ifdef CONFIG_BTRFS_DEBUG |
3065 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3066 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3067 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3068 | #endif |
c8bf1b67 | 3069 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3070 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3071 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3072 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3073 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3074 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3075 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3076 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3077 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3078 | BTRFS_BLOCK_RSV_DELREFS); | |
3079 | ||
771ed689 | 3080 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3081 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3082 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3083 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3084 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3085 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3086 | fs_info->metadata_ratio = 0; |
4cb5300b | 3087 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3088 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3089 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3090 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3091 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3092 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3093 | |
b34b086c CM |
3094 | fs_info->thread_pool_size = min_t(unsigned long, |
3095 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3096 | |
199c2a9c MX |
3097 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3098 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3099 | |
638aa7ed | 3100 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3101 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3102 | fs_info->check_integrity_print_mask = 0; | |
3103 | #endif | |
779a65a4 | 3104 | btrfs_init_balance(fs_info); |
57056740 | 3105 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3106 | |
0f9dd46c | 3107 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 3108 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 3109 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 3110 | |
fe119a6e NB |
3111 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
3112 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
39279cc3 | 3113 | |
5a3f23d5 | 3114 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3115 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3116 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3117 | mutex_init(&fs_info->transaction_kthread_mutex); |
3118 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3119 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3120 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3121 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3122 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3123 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3124 | |
ad618368 | 3125 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3126 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3127 | btrfs_discard_init(fs_info); |
416ac51d | 3128 | |
fa9c0d79 CM |
3129 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3130 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3131 | ||
e6dcd2dc | 3132 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3133 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3134 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3135 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3136 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3137 | |
da17066c JM |
3138 | /* Usable values until the real ones are cached from the superblock */ |
3139 | fs_info->nodesize = 4096; | |
3140 | fs_info->sectorsize = 4096; | |
ab108d99 | 3141 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3142 | fs_info->stripesize = 4096; |
3143 | ||
eede2bf3 OS |
3144 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3145 | fs_info->swapfile_pins = RB_ROOT; | |
3146 | ||
18bb8bbf JT |
3147 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3148 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3149 | } |
3150 | ||
3151 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3152 | { | |
3153 | int ret; | |
3154 | ||
3155 | fs_info->sb = sb; | |
3156 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3157 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3158 | |
5deb17e1 | 3159 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3160 | if (ret) |
c75e8394 | 3161 | return ret; |
ae18c37a JB |
3162 | |
3163 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3164 | if (ret) | |
c75e8394 | 3165 | return ret; |
ae18c37a JB |
3166 | |
3167 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3168 | (1 + ilog2(nr_cpu_ids)); | |
3169 | ||
3170 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3171 | if (ret) | |
c75e8394 | 3172 | return ret; |
ae18c37a JB |
3173 | |
3174 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3175 | GFP_KERNEL); | |
3176 | if (ret) | |
c75e8394 | 3177 | return ret; |
ae18c37a JB |
3178 | |
3179 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3180 | GFP_KERNEL); | |
c75e8394 JB |
3181 | if (!fs_info->delayed_root) |
3182 | return -ENOMEM; | |
ae18c37a JB |
3183 | btrfs_init_delayed_root(fs_info->delayed_root); |
3184 | ||
a0a1db70 FM |
3185 | if (sb_rdonly(sb)) |
3186 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3187 | ||
c75e8394 | 3188 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3189 | } |
3190 | ||
97f4dd09 NB |
3191 | static int btrfs_uuid_rescan_kthread(void *data) |
3192 | { | |
3193 | struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; | |
3194 | int ret; | |
3195 | ||
3196 | /* | |
3197 | * 1st step is to iterate through the existing UUID tree and | |
3198 | * to delete all entries that contain outdated data. | |
3199 | * 2nd step is to add all missing entries to the UUID tree. | |
3200 | */ | |
3201 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3202 | if (ret < 0) { | |
c94bec2c JB |
3203 | if (ret != -EINTR) |
3204 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3205 | ret); | |
97f4dd09 NB |
3206 | up(&fs_info->uuid_tree_rescan_sem); |
3207 | return ret; | |
3208 | } | |
3209 | return btrfs_uuid_scan_kthread(data); | |
3210 | } | |
3211 | ||
3212 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3213 | { | |
3214 | struct task_struct *task; | |
3215 | ||
3216 | down(&fs_info->uuid_tree_rescan_sem); | |
3217 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3218 | if (IS_ERR(task)) { | |
3219 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3220 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3221 | up(&fs_info->uuid_tree_rescan_sem); | |
3222 | return PTR_ERR(task); | |
3223 | } | |
3224 | ||
3225 | return 0; | |
3226 | } | |
3227 | ||
8cd29088 BB |
3228 | /* |
3229 | * Some options only have meaning at mount time and shouldn't persist across | |
3230 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3231 | * code paths. | |
3232 | */ | |
3233 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3234 | { | |
3235 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3236 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3237 | } |
3238 | ||
44c0ca21 BB |
3239 | /* |
3240 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3241 | * and btrfs_remount when remounting from read-only to read-write. | |
3242 | */ | |
3243 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3244 | { | |
3245 | int ret; | |
94846229 | 3246 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3247 | bool clear_free_space_tree = false; |
3248 | ||
3249 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3250 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3251 | clear_free_space_tree = true; | |
3252 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3253 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3254 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3255 | clear_free_space_tree = true; | |
3256 | } | |
3257 | ||
3258 | if (clear_free_space_tree) { | |
3259 | btrfs_info(fs_info, "clearing free space tree"); | |
3260 | ret = btrfs_clear_free_space_tree(fs_info); | |
3261 | if (ret) { | |
3262 | btrfs_warn(fs_info, | |
3263 | "failed to clear free space tree: %d", ret); | |
3264 | goto out; | |
3265 | } | |
3266 | } | |
44c0ca21 | 3267 | |
8d488a8c FM |
3268 | /* |
3269 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3270 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
3271 | * them into the fs_info->fs_roots_radix tree. This must be done before | |
3272 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it | |
3273 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3274 | * item before the root's tree is deleted - this means that if we unmount | |
3275 | * or crash before the deletion completes, on the next mount we will not | |
3276 | * delete what remains of the tree because the orphan item does not | |
3277 | * exists anymore, which is what tells us we have a pending deletion. | |
3278 | */ | |
3279 | ret = btrfs_find_orphan_roots(fs_info); | |
3280 | if (ret) | |
3281 | goto out; | |
3282 | ||
44c0ca21 BB |
3283 | ret = btrfs_cleanup_fs_roots(fs_info); |
3284 | if (ret) | |
3285 | goto out; | |
3286 | ||
8f1c21d7 BB |
3287 | down_read(&fs_info->cleanup_work_sem); |
3288 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3289 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3290 | up_read(&fs_info->cleanup_work_sem); | |
3291 | goto out; | |
3292 | } | |
3293 | up_read(&fs_info->cleanup_work_sem); | |
3294 | ||
44c0ca21 BB |
3295 | mutex_lock(&fs_info->cleaner_mutex); |
3296 | ret = btrfs_recover_relocation(fs_info->tree_root); | |
3297 | mutex_unlock(&fs_info->cleaner_mutex); | |
3298 | if (ret < 0) { | |
3299 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3300 | goto out; | |
3301 | } | |
3302 | ||
5011139a BB |
3303 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3304 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3305 | btrfs_info(fs_info, "creating free space tree"); | |
3306 | ret = btrfs_create_free_space_tree(fs_info); | |
3307 | if (ret) { | |
3308 | btrfs_warn(fs_info, | |
3309 | "failed to create free space tree: %d", ret); | |
3310 | goto out; | |
3311 | } | |
3312 | } | |
3313 | ||
94846229 BB |
3314 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3315 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3316 | if (ret) | |
3317 | goto out; | |
3318 | } | |
3319 | ||
44c0ca21 BB |
3320 | ret = btrfs_resume_balance_async(fs_info); |
3321 | if (ret) | |
3322 | goto out; | |
3323 | ||
3324 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3325 | if (ret) { | |
3326 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3327 | goto out; | |
3328 | } | |
3329 | ||
3330 | btrfs_qgroup_rescan_resume(fs_info); | |
3331 | ||
3332 | if (!fs_info->uuid_root) { | |
3333 | btrfs_info(fs_info, "creating UUID tree"); | |
3334 | ret = btrfs_create_uuid_tree(fs_info); | |
3335 | if (ret) { | |
3336 | btrfs_warn(fs_info, | |
3337 | "failed to create the UUID tree %d", ret); | |
3338 | goto out; | |
3339 | } | |
3340 | } | |
3341 | ||
3342 | out: | |
3343 | return ret; | |
3344 | } | |
3345 | ||
ae18c37a JB |
3346 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3347 | char *options) | |
3348 | { | |
3349 | u32 sectorsize; | |
3350 | u32 nodesize; | |
3351 | u32 stripesize; | |
3352 | u64 generation; | |
3353 | u64 features; | |
3354 | u16 csum_type; | |
ae18c37a JB |
3355 | struct btrfs_super_block *disk_super; |
3356 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3357 | struct btrfs_root *tree_root; | |
3358 | struct btrfs_root *chunk_root; | |
3359 | int ret; | |
3360 | int err = -EINVAL; | |
ae18c37a JB |
3361 | int level; |
3362 | ||
8260edba | 3363 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3364 | if (ret) { |
83c8266a | 3365 | err = ret; |
ae18c37a | 3366 | goto fail; |
53b381b3 DW |
3367 | } |
3368 | ||
ae18c37a JB |
3369 | /* These need to be init'ed before we start creating inodes and such. */ |
3370 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3371 | GFP_KERNEL); | |
3372 | fs_info->tree_root = tree_root; | |
3373 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3374 | GFP_KERNEL); | |
3375 | fs_info->chunk_root = chunk_root; | |
3376 | if (!tree_root || !chunk_root) { | |
3377 | err = -ENOMEM; | |
c75e8394 | 3378 | goto fail; |
ae18c37a JB |
3379 | } |
3380 | ||
3381 | fs_info->btree_inode = new_inode(sb); | |
3382 | if (!fs_info->btree_inode) { | |
3383 | err = -ENOMEM; | |
c75e8394 | 3384 | goto fail; |
ae18c37a JB |
3385 | } |
3386 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3387 | btrfs_init_btree_inode(fs_info); | |
3388 | ||
d24fa5c1 | 3389 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3390 | |
3391 | /* | |
3392 | * Read super block and check the signature bytes only | |
3393 | */ | |
d24fa5c1 | 3394 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3395 | if (IS_ERR(disk_super)) { |
3396 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3397 | goto fail_alloc; |
20b45077 | 3398 | } |
39279cc3 | 3399 | |
8dc3f22c | 3400 | /* |
260db43c | 3401 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3402 | * corrupted, we'll find out |
3403 | */ | |
8f32380d | 3404 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3405 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3406 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3407 | csum_type); |
8dc3f22c | 3408 | err = -EINVAL; |
8f32380d | 3409 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3410 | goto fail_alloc; |
3411 | } | |
3412 | ||
83c68bbc SY |
3413 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3414 | ||
6d97c6e3 JT |
3415 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3416 | if (ret) { | |
3417 | err = ret; | |
8f32380d | 3418 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3419 | goto fail_alloc; |
3420 | } | |
3421 | ||
1104a885 DS |
3422 | /* |
3423 | * We want to check superblock checksum, the type is stored inside. | |
3424 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3425 | */ | |
8f32380d | 3426 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3427 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3428 | err = -EINVAL; |
8f32380d | 3429 | btrfs_release_disk_super(disk_super); |
141386e1 | 3430 | goto fail_alloc; |
1104a885 DS |
3431 | } |
3432 | ||
3433 | /* | |
3434 | * super_copy is zeroed at allocation time and we never touch the | |
3435 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3436 | * the whole block of INFO_SIZE | |
3437 | */ | |
8f32380d JT |
3438 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3439 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3440 | |
fbc6feae NB |
3441 | disk_super = fs_info->super_copy; |
3442 | ||
0b86a832 | 3443 | |
fbc6feae NB |
3444 | features = btrfs_super_flags(disk_super); |
3445 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3446 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3447 | btrfs_set_super_flags(disk_super, features); | |
3448 | btrfs_info(fs_info, | |
3449 | "found metadata UUID change in progress flag, clearing"); | |
3450 | } | |
3451 | ||
3452 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3453 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3454 | |
069ec957 | 3455 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3456 | if (ret) { |
05135f59 | 3457 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3458 | err = -EINVAL; |
141386e1 | 3459 | goto fail_alloc; |
1104a885 DS |
3460 | } |
3461 | ||
0f7d52f4 | 3462 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3463 | goto fail_alloc; |
0f7d52f4 | 3464 | |
acce952b | 3465 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3466 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3467 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3468 | |
75e7cb7f LB |
3469 | /* |
3470 | * In the long term, we'll store the compression type in the super | |
3471 | * block, and it'll be used for per file compression control. | |
3472 | */ | |
3473 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3474 | ||
6f93e834 AJ |
3475 | /* |
3476 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3477 | * than the page size. | |
3478 | */ | |
3479 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { | |
3480 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
3481 | btrfs_info(fs_info, | |
3482 | "flagging fs with big metadata feature"); | |
3483 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3484 | } | |
3485 | ||
3486 | /* Set up fs_info before parsing mount options */ | |
3487 | nodesize = btrfs_super_nodesize(disk_super); | |
3488 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3489 | stripesize = sectorsize; | |
3490 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3491 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3492 | ||
3493 | fs_info->nodesize = nodesize; | |
3494 | fs_info->sectorsize = sectorsize; | |
3495 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3496 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3497 | fs_info->stripesize = stripesize; | |
3498 | ||
2ff7e61e | 3499 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3500 | if (ret) { |
3501 | err = ret; | |
141386e1 | 3502 | goto fail_alloc; |
2b82032c | 3503 | } |
dfe25020 | 3504 | |
f2b636e8 JB |
3505 | features = btrfs_super_incompat_flags(disk_super) & |
3506 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3507 | if (features) { | |
05135f59 DS |
3508 | btrfs_err(fs_info, |
3509 | "cannot mount because of unsupported optional features (%llx)", | |
3510 | features); | |
f2b636e8 | 3511 | err = -EINVAL; |
141386e1 | 3512 | goto fail_alloc; |
f2b636e8 JB |
3513 | } |
3514 | ||
5d4f98a2 | 3515 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3516 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3517 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3518 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3519 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3520 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3521 | |
3173a18f | 3522 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 3523 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 3524 | |
bc3f116f CM |
3525 | /* |
3526 | * mixed block groups end up with duplicate but slightly offset | |
3527 | * extent buffers for the same range. It leads to corruptions | |
3528 | */ | |
3529 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3530 | (sectorsize != nodesize)) { |
05135f59 DS |
3531 | btrfs_err(fs_info, |
3532 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3533 | nodesize, sectorsize); | |
141386e1 | 3534 | goto fail_alloc; |
bc3f116f CM |
3535 | } |
3536 | ||
ceda0864 MX |
3537 | /* |
3538 | * Needn't use the lock because there is no other task which will | |
3539 | * update the flag. | |
3540 | */ | |
a6fa6fae | 3541 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3542 | |
f2b636e8 JB |
3543 | features = btrfs_super_compat_ro_flags(disk_super) & |
3544 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3545 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
3546 | btrfs_err(fs_info, |
3547 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 3548 | features); |
f2b636e8 | 3549 | err = -EINVAL; |
141386e1 | 3550 | goto fail_alloc; |
f2b636e8 | 3551 | } |
61d92c32 | 3552 | |
8481dd80 QW |
3553 | if (sectorsize < PAGE_SIZE) { |
3554 | struct btrfs_subpage_info *subpage_info; | |
3555 | ||
95ea0486 QW |
3556 | btrfs_warn(fs_info, |
3557 | "read-write for sector size %u with page size %lu is experimental", | |
3558 | sectorsize, PAGE_SIZE); | |
c8050b3b QW |
3559 | if (btrfs_super_incompat_flags(fs_info->super_copy) & |
3560 | BTRFS_FEATURE_INCOMPAT_RAID56) { | |
3561 | btrfs_err(fs_info, | |
3562 | "RAID56 is not yet supported for sector size %u with page size %lu", | |
3563 | sectorsize, PAGE_SIZE); | |
3564 | err = -EINVAL; | |
3565 | goto fail_alloc; | |
3566 | } | |
8481dd80 QW |
3567 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3568 | if (!subpage_info) | |
3569 | goto fail_alloc; | |
3570 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3571 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3572 | } |
0bb3eb3e | 3573 | |
d21deec5 | 3574 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3575 | if (ret) { |
3576 | err = ret; | |
0dc3b84a JB |
3577 | goto fail_sb_buffer; |
3578 | } | |
4543df7e | 3579 | |
9e11ceee JK |
3580 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3581 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3582 | |
a061fc8d CM |
3583 | sb->s_blocksize = sectorsize; |
3584 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3585 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3586 | |
925baedd | 3587 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3588 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3589 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3590 | if (ret) { |
05135f59 | 3591 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3592 | goto fail_sb_buffer; |
84eed90f | 3593 | } |
0b86a832 | 3594 | |
84234f3a | 3595 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3596 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3597 | |
2ff7e61e | 3598 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3599 | btrfs_super_chunk_root(disk_super), |
1b7ec85e | 3600 | BTRFS_CHUNK_TREE_OBJECTID, |
581c1760 | 3601 | generation, level, NULL); |
64c043de LB |
3602 | if (IS_ERR(chunk_root->node) || |
3603 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3604 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3605 | if (!IS_ERR(chunk_root->node)) |
3606 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3607 | chunk_root->node = NULL; |
af31f5e5 | 3608 | goto fail_tree_roots; |
83121942 | 3609 | } |
5d4f98a2 YZ |
3610 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3611 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3612 | |
e17cade2 | 3613 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3614 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3615 | BTRFS_UUID_SIZE); | |
e17cade2 | 3616 | |
5b4aacef | 3617 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3618 | if (ret) { |
05135f59 | 3619 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3620 | goto fail_tree_roots; |
2b82032c | 3621 | } |
0b86a832 | 3622 | |
8dabb742 | 3623 | /* |
bacce86a AJ |
3624 | * At this point we know all the devices that make this filesystem, |
3625 | * including the seed devices but we don't know yet if the replace | |
3626 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3627 | * filesystem but skip the replace target device which is checked |
bacce86a | 3628 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3629 | */ |
bacce86a | 3630 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3631 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3632 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3633 | goto fail_tree_roots; |
3634 | } | |
3635 | ||
b8522a1e | 3636 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3637 | if (ret) |
b8522a1e | 3638 | goto fail_tree_roots; |
8929ecfa | 3639 | |
73651042 NA |
3640 | /* |
3641 | * Get zone type information of zoned block devices. This will also | |
3642 | * handle emulation of a zoned filesystem if a regular device has the | |
3643 | * zoned incompat feature flag set. | |
3644 | */ | |
3645 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3646 | if (ret) { | |
3647 | btrfs_err(fs_info, | |
3648 | "zoned: failed to read device zone info: %d", | |
3649 | ret); | |
3650 | goto fail_block_groups; | |
3651 | } | |
3652 | ||
75ec1db8 JB |
3653 | /* |
3654 | * If we have a uuid root and we're not being told to rescan we need to | |
3655 | * check the generation here so we can set the | |
3656 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3657 | * transaction during a balance or the log replay without updating the | |
3658 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3659 | * even though it was perfectly fine. | |
3660 | */ | |
3661 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3662 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3663 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3664 | ||
cf90d884 QW |
3665 | ret = btrfs_verify_dev_extents(fs_info); |
3666 | if (ret) { | |
3667 | btrfs_err(fs_info, | |
3668 | "failed to verify dev extents against chunks: %d", | |
3669 | ret); | |
3670 | goto fail_block_groups; | |
3671 | } | |
68310a5e ID |
3672 | ret = btrfs_recover_balance(fs_info); |
3673 | if (ret) { | |
05135f59 | 3674 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3675 | goto fail_block_groups; |
3676 | } | |
3677 | ||
733f4fbb SB |
3678 | ret = btrfs_init_dev_stats(fs_info); |
3679 | if (ret) { | |
05135f59 | 3680 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3681 | goto fail_block_groups; |
3682 | } | |
3683 | ||
8dabb742 SB |
3684 | ret = btrfs_init_dev_replace(fs_info); |
3685 | if (ret) { | |
05135f59 | 3686 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3687 | goto fail_block_groups; |
3688 | } | |
3689 | ||
b70f5097 NA |
3690 | ret = btrfs_check_zoned_mode(fs_info); |
3691 | if (ret) { | |
3692 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3693 | ret); | |
3694 | goto fail_block_groups; | |
3695 | } | |
3696 | ||
c6761a9e | 3697 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3698 | if (ret) { |
05135f59 DS |
3699 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3700 | ret); | |
b7c35e81 AJ |
3701 | goto fail_block_groups; |
3702 | } | |
3703 | ||
96f3136e | 3704 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3705 | if (ret) { |
05135f59 | 3706 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3707 | goto fail_fsdev_sysfs; |
c59021f8 | 3708 | } |
3709 | ||
c59021f8 | 3710 | ret = btrfs_init_space_info(fs_info); |
3711 | if (ret) { | |
05135f59 | 3712 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3713 | goto fail_sysfs; |
c59021f8 | 3714 | } |
3715 | ||
5b4aacef | 3716 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3717 | if (ret) { |
05135f59 | 3718 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3719 | goto fail_sysfs; |
1b1d1f66 | 3720 | } |
4330e183 | 3721 | |
16beac87 NA |
3722 | btrfs_free_zone_cache(fs_info); |
3723 | ||
5c78a5e7 AJ |
3724 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3725 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3726 | btrfs_warn(fs_info, |
52042d8e | 3727 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3728 | goto fail_sysfs; |
292fd7fc | 3729 | } |
9078a3e1 | 3730 | |
a74a4b97 CM |
3731 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3732 | "btrfs-cleaner"); | |
57506d50 | 3733 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3734 | goto fail_sysfs; |
a74a4b97 CM |
3735 | |
3736 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3737 | tree_root, | |
3738 | "btrfs-transaction"); | |
57506d50 | 3739 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3740 | goto fail_cleaner; |
a74a4b97 | 3741 | |
583b7231 | 3742 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3743 | !fs_info->fs_devices->rotating) { |
583b7231 | 3744 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3745 | } |
3746 | ||
572d9ab7 | 3747 | /* |
01327610 | 3748 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3749 | * not have to wait for transaction commit |
3750 | */ | |
3751 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3752 | |
21adbd5c | 3753 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3754 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3755 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3756 | btrfs_test_opt(fs_info, |
cbeaae4f | 3757 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3758 | fs_info->check_integrity_print_mask); |
3759 | if (ret) | |
05135f59 DS |
3760 | btrfs_warn(fs_info, |
3761 | "failed to initialize integrity check module: %d", | |
3762 | ret); | |
21adbd5c SB |
3763 | } |
3764 | #endif | |
bcef60f2 AJ |
3765 | ret = btrfs_read_qgroup_config(fs_info); |
3766 | if (ret) | |
3767 | goto fail_trans_kthread; | |
21adbd5c | 3768 | |
fd708b81 JB |
3769 | if (btrfs_build_ref_tree(fs_info)) |
3770 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3771 | ||
96da0919 QW |
3772 | /* do not make disk changes in broken FS or nologreplay is given */ |
3773 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3774 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3775 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3776 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3777 | if (ret) { |
63443bf5 | 3778 | err = ret; |
28c16cbb | 3779 | goto fail_qgroup; |
79787eaa | 3780 | } |
e02119d5 | 3781 | } |
1a40e23b | 3782 | |
56e9357a | 3783 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3784 | if (IS_ERR(fs_info->fs_root)) { |
3785 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3786 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3787 | fs_info->fs_root = NULL; |
bcef60f2 | 3788 | goto fail_qgroup; |
3140c9a3 | 3789 | } |
c289811c | 3790 | |
bc98a42c | 3791 | if (sb_rdonly(sb)) |
8cd29088 | 3792 | goto clear_oneshot; |
59641015 | 3793 | |
44c0ca21 | 3794 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3795 | if (ret) { |
6bccf3ab | 3796 | close_ctree(fs_info); |
2b6ba629 | 3797 | return ret; |
e3acc2a6 | 3798 | } |
b0643e59 | 3799 | btrfs_discard_resume(fs_info); |
b382a324 | 3800 | |
44c0ca21 BB |
3801 | if (fs_info->uuid_root && |
3802 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3803 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3804 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3805 | ret = btrfs_check_uuid_tree(fs_info); |
3806 | if (ret) { | |
05135f59 DS |
3807 | btrfs_warn(fs_info, |
3808 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3809 | close_ctree(fs_info); |
70f80175 SB |
3810 | return ret; |
3811 | } | |
f7a81ea4 | 3812 | } |
94846229 | 3813 | |
afcdd129 | 3814 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3815 | |
b4be6aef JB |
3816 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3817 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3818 | wake_up_process(fs_info->cleaner_kthread); | |
3819 | ||
8cd29088 BB |
3820 | clear_oneshot: |
3821 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3822 | return 0; |
39279cc3 | 3823 | |
bcef60f2 AJ |
3824 | fail_qgroup: |
3825 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3826 | fail_trans_kthread: |
3827 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3828 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3829 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3830 | fail_cleaner: |
a74a4b97 | 3831 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3832 | |
3833 | /* | |
3834 | * make sure we're done with the btree inode before we stop our | |
3835 | * kthreads | |
3836 | */ | |
3837 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3838 | |
2365dd3c | 3839 | fail_sysfs: |
6618a59b | 3840 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3841 | |
b7c35e81 AJ |
3842 | fail_fsdev_sysfs: |
3843 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3844 | ||
1b1d1f66 | 3845 | fail_block_groups: |
54067ae9 | 3846 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3847 | |
3848 | fail_tree_roots: | |
9e3aa805 JB |
3849 | if (fs_info->data_reloc_root) |
3850 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3851 | free_root_pointers(fs_info, true); |
2b8195bb | 3852 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3853 | |
39279cc3 | 3854 | fail_sb_buffer: |
7abadb64 | 3855 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3856 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3857 | fail_alloc: |
586e46e2 ID |
3858 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3859 | ||
4543df7e | 3860 | iput(fs_info->btree_inode); |
7e662854 | 3861 | fail: |
586e46e2 | 3862 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3863 | return err; |
eb60ceac | 3864 | } |
663faf9f | 3865 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3866 | |
314b6dd0 | 3867 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3868 | { |
314b6dd0 JT |
3869 | struct btrfs_device *device = bio->bi_private; |
3870 | struct bio_vec *bvec; | |
3871 | struct bvec_iter_all iter_all; | |
3872 | struct page *page; | |
3873 | ||
3874 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3875 | page = bvec->bv_page; | |
3876 | ||
3877 | if (bio->bi_status) { | |
3878 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3879 | "lost page write due to IO error on %s (%d)", | |
3880 | rcu_str_deref(device->name), | |
3881 | blk_status_to_errno(bio->bi_status)); | |
3882 | ClearPageUptodate(page); | |
3883 | SetPageError(page); | |
3884 | btrfs_dev_stat_inc_and_print(device, | |
3885 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3886 | } else { | |
3887 | SetPageUptodate(page); | |
3888 | } | |
3889 | ||
3890 | put_page(page); | |
3891 | unlock_page(page); | |
f2984462 | 3892 | } |
314b6dd0 JT |
3893 | |
3894 | bio_put(bio); | |
f2984462 CM |
3895 | } |
3896 | ||
8f32380d JT |
3897 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
3898 | int copy_num) | |
29c36d72 | 3899 | { |
29c36d72 | 3900 | struct btrfs_super_block *super; |
8f32380d | 3901 | struct page *page; |
12659251 | 3902 | u64 bytenr, bytenr_orig; |
8f32380d | 3903 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3904 | int ret; |
3905 | ||
3906 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3907 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3908 | if (ret == -ENOENT) | |
3909 | return ERR_PTR(-EINVAL); | |
3910 | else if (ret) | |
3911 | return ERR_PTR(ret); | |
29c36d72 | 3912 | |
cda00eba | 3913 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3914 | return ERR_PTR(-EINVAL); |
29c36d72 | 3915 | |
8f32380d JT |
3916 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3917 | if (IS_ERR(page)) | |
3918 | return ERR_CAST(page); | |
29c36d72 | 3919 | |
8f32380d | 3920 | super = page_address(page); |
96c2e067 AJ |
3921 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3922 | btrfs_release_disk_super(super); | |
3923 | return ERR_PTR(-ENODATA); | |
3924 | } | |
3925 | ||
12659251 | 3926 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3927 | btrfs_release_disk_super(super); |
3928 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3929 | } |
3930 | ||
8f32380d | 3931 | return super; |
29c36d72 AJ |
3932 | } |
3933 | ||
3934 | ||
8f32380d | 3935 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3936 | { |
8f32380d | 3937 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3938 | int i; |
3939 | u64 transid = 0; | |
a512bbf8 YZ |
3940 | |
3941 | /* we would like to check all the supers, but that would make | |
3942 | * a btrfs mount succeed after a mkfs from a different FS. | |
3943 | * So, we need to add a special mount option to scan for | |
3944 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3945 | */ | |
3946 | for (i = 0; i < 1; i++) { | |
8f32380d JT |
3947 | super = btrfs_read_dev_one_super(bdev, i); |
3948 | if (IS_ERR(super)) | |
a512bbf8 YZ |
3949 | continue; |
3950 | ||
a512bbf8 | 3951 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3952 | if (latest) |
3953 | btrfs_release_disk_super(super); | |
3954 | ||
3955 | latest = super; | |
a512bbf8 | 3956 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3957 | } |
3958 | } | |
92fc03fb | 3959 | |
8f32380d | 3960 | return super; |
a512bbf8 YZ |
3961 | } |
3962 | ||
4eedeb75 | 3963 | /* |
abbb3b8e | 3964 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3965 | * pages we use for writing are locked. |
4eedeb75 | 3966 | * |
abbb3b8e DS |
3967 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3968 | * the expected device size at commit time. Note that max_mirrors must be | |
3969 | * same for write and wait phases. | |
4eedeb75 | 3970 | * |
314b6dd0 | 3971 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3972 | */ |
a512bbf8 | 3973 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3974 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3975 | { |
d5178578 | 3976 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3977 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3978 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3979 | int i; |
a512bbf8 | 3980 | int errors = 0; |
12659251 NA |
3981 | int ret; |
3982 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3983 | |
3984 | if (max_mirrors == 0) | |
3985 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3986 | ||
d5178578 JT |
3987 | shash->tfm = fs_info->csum_shash; |
3988 | ||
a512bbf8 | 3989 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3990 | struct page *page; |
3991 | struct bio *bio; | |
3992 | struct btrfs_super_block *disk_super; | |
3993 | ||
12659251 NA |
3994 | bytenr_orig = btrfs_sb_offset(i); |
3995 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3996 | if (ret == -ENOENT) { | |
3997 | continue; | |
3998 | } else if (ret < 0) { | |
3999 | btrfs_err(device->fs_info, | |
4000 | "couldn't get super block location for mirror %d", | |
4001 | i); | |
4002 | errors++; | |
4003 | continue; | |
4004 | } | |
935e5cc9 MX |
4005 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4006 | device->commit_total_bytes) | |
a512bbf8 YZ |
4007 | break; |
4008 | ||
12659251 | 4009 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 4010 | |
fd08001f EB |
4011 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
4012 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
4013 | sb->csum); | |
4eedeb75 | 4014 | |
314b6dd0 JT |
4015 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
4016 | GFP_NOFS); | |
4017 | if (!page) { | |
abbb3b8e | 4018 | btrfs_err(device->fs_info, |
314b6dd0 | 4019 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
4020 | bytenr); |
4021 | errors++; | |
4eedeb75 | 4022 | continue; |
abbb3b8e | 4023 | } |
634554dc | 4024 | |
314b6dd0 JT |
4025 | /* Bump the refcount for wait_dev_supers() */ |
4026 | get_page(page); | |
a512bbf8 | 4027 | |
314b6dd0 JT |
4028 | disk_super = page_address(page); |
4029 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 4030 | |
314b6dd0 JT |
4031 | /* |
4032 | * Directly use bios here instead of relying on the page cache | |
4033 | * to do I/O, so we don't lose the ability to do integrity | |
4034 | * checking. | |
4035 | */ | |
4036 | bio = bio_alloc(GFP_NOFS, 1); | |
4037 | bio_set_dev(bio, device->bdev); | |
4038 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; | |
4039 | bio->bi_private = device; | |
4040 | bio->bi_end_io = btrfs_end_super_write; | |
4041 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4042 | offset_in_page(bytenr)); | |
a512bbf8 | 4043 | |
387125fc | 4044 | /* |
314b6dd0 JT |
4045 | * We FUA only the first super block. The others we allow to |
4046 | * go down lazy and there's a short window where the on-disk | |
4047 | * copies might still contain the older version. | |
387125fc | 4048 | */ |
314b6dd0 | 4049 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; |
1b9e619c | 4050 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4051 | bio->bi_opf |= REQ_FUA; |
4052 | ||
4053 | btrfsic_submit_bio(bio); | |
8376d9e1 NA |
4054 | |
4055 | if (btrfs_advance_sb_log(device, i)) | |
4056 | errors++; | |
a512bbf8 YZ |
4057 | } |
4058 | return errors < i ? 0 : -1; | |
4059 | } | |
4060 | ||
abbb3b8e DS |
4061 | /* |
4062 | * Wait for write completion of superblocks done by write_dev_supers, | |
4063 | * @max_mirrors same for write and wait phases. | |
4064 | * | |
314b6dd0 | 4065 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4066 | * date. |
4067 | */ | |
4068 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4069 | { | |
abbb3b8e DS |
4070 | int i; |
4071 | int errors = 0; | |
b6a535fa | 4072 | bool primary_failed = false; |
12659251 | 4073 | int ret; |
abbb3b8e DS |
4074 | u64 bytenr; |
4075 | ||
4076 | if (max_mirrors == 0) | |
4077 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4078 | ||
4079 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4080 | struct page *page; |
4081 | ||
12659251 NA |
4082 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4083 | if (ret == -ENOENT) { | |
4084 | break; | |
4085 | } else if (ret < 0) { | |
4086 | errors++; | |
4087 | if (i == 0) | |
4088 | primary_failed = true; | |
4089 | continue; | |
4090 | } | |
abbb3b8e DS |
4091 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4092 | device->commit_total_bytes) | |
4093 | break; | |
4094 | ||
314b6dd0 JT |
4095 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4096 | bytenr >> PAGE_SHIFT); | |
4097 | if (!page) { | |
abbb3b8e | 4098 | errors++; |
b6a535fa HM |
4099 | if (i == 0) |
4100 | primary_failed = true; | |
abbb3b8e DS |
4101 | continue; |
4102 | } | |
314b6dd0 JT |
4103 | /* Page is submitted locked and unlocked once the IO completes */ |
4104 | wait_on_page_locked(page); | |
4105 | if (PageError(page)) { | |
abbb3b8e | 4106 | errors++; |
b6a535fa HM |
4107 | if (i == 0) |
4108 | primary_failed = true; | |
4109 | } | |
abbb3b8e | 4110 | |
314b6dd0 JT |
4111 | /* Drop our reference */ |
4112 | put_page(page); | |
abbb3b8e | 4113 | |
314b6dd0 JT |
4114 | /* Drop the reference from the writing run */ |
4115 | put_page(page); | |
abbb3b8e DS |
4116 | } |
4117 | ||
b6a535fa HM |
4118 | /* log error, force error return */ |
4119 | if (primary_failed) { | |
4120 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4121 | device->devid); | |
4122 | return -1; | |
4123 | } | |
4124 | ||
abbb3b8e DS |
4125 | return errors < i ? 0 : -1; |
4126 | } | |
4127 | ||
387125fc CM |
4128 | /* |
4129 | * endio for the write_dev_flush, this will wake anyone waiting | |
4130 | * for the barrier when it is done | |
4131 | */ | |
4246a0b6 | 4132 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4133 | { |
e0ae9994 | 4134 | complete(bio->bi_private); |
387125fc CM |
4135 | } |
4136 | ||
4137 | /* | |
4fc6441a AJ |
4138 | * Submit a flush request to the device if it supports it. Error handling is |
4139 | * done in the waiting counterpart. | |
387125fc | 4140 | */ |
4fc6441a | 4141 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4142 | { |
e0ae9994 | 4143 | struct bio *bio = device->flush_bio; |
387125fc | 4144 | |
a91cf0ff WY |
4145 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4146 | /* | |
4147 | * When a disk has write caching disabled, we skip submission of a bio | |
4148 | * with flush and sync requests before writing the superblock, since | |
4149 | * it's not needed. However when the integrity checker is enabled, this | |
4150 | * results in reports that there are metadata blocks referred by a | |
4151 | * superblock that were not properly flushed. So don't skip the bio | |
4152 | * submission only when the integrity checker is enabled for the sake | |
4153 | * of simplicity, since this is a debug tool and not meant for use in | |
4154 | * non-debug builds. | |
4155 | */ | |
4156 | struct request_queue *q = bdev_get_queue(device->bdev); | |
c2a9c7ab | 4157 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 4158 | return; |
a91cf0ff | 4159 | #endif |
387125fc | 4160 | |
e0ae9994 | 4161 | bio_reset(bio); |
387125fc | 4162 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 4163 | bio_set_dev(bio, device->bdev); |
8d910125 | 4164 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
4165 | init_completion(&device->flush_wait); |
4166 | bio->bi_private = &device->flush_wait; | |
387125fc | 4167 | |
43a01111 | 4168 | btrfsic_submit_bio(bio); |
1c3063b6 | 4169 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4170 | } |
387125fc | 4171 | |
4fc6441a AJ |
4172 | /* |
4173 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4174 | */ | |
8c27cb35 | 4175 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4176 | { |
4fc6441a | 4177 | struct bio *bio = device->flush_bio; |
387125fc | 4178 | |
1c3063b6 | 4179 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4180 | return BLK_STS_OK; |
387125fc | 4181 | |
1c3063b6 | 4182 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4183 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4184 | |
8c27cb35 | 4185 | return bio->bi_status; |
387125fc | 4186 | } |
387125fc | 4187 | |
d10b82fe | 4188 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4189 | { |
6528b99d | 4190 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4191 | return -EIO; |
387125fc CM |
4192 | return 0; |
4193 | } | |
4194 | ||
4195 | /* | |
4196 | * send an empty flush down to each device in parallel, | |
4197 | * then wait for them | |
4198 | */ | |
4199 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4200 | { | |
4201 | struct list_head *head; | |
4202 | struct btrfs_device *dev; | |
5af3e8cc | 4203 | int errors_wait = 0; |
4e4cbee9 | 4204 | blk_status_t ret; |
387125fc | 4205 | |
1538e6c5 | 4206 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4207 | /* send down all the barriers */ |
4208 | head = &info->fs_devices->devices; | |
1538e6c5 | 4209 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4210 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4211 | continue; |
cea7c8bf | 4212 | if (!dev->bdev) |
387125fc | 4213 | continue; |
e12c9621 | 4214 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4215 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4216 | continue; |
4217 | ||
4fc6441a | 4218 | write_dev_flush(dev); |
58efbc9f | 4219 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4220 | } |
4221 | ||
4222 | /* wait for all the barriers */ | |
1538e6c5 | 4223 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4224 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4225 | continue; |
387125fc | 4226 | if (!dev->bdev) { |
5af3e8cc | 4227 | errors_wait++; |
387125fc CM |
4228 | continue; |
4229 | } | |
e12c9621 | 4230 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4231 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4232 | continue; |
4233 | ||
4fc6441a | 4234 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4235 | if (ret) { |
4236 | dev->last_flush_error = ret; | |
66b4993e DS |
4237 | btrfs_dev_stat_inc_and_print(dev, |
4238 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4239 | errors_wait++; |
401b41e5 AJ |
4240 | } |
4241 | } | |
4242 | ||
cea7c8bf | 4243 | if (errors_wait) { |
401b41e5 AJ |
4244 | /* |
4245 | * At some point we need the status of all disks | |
4246 | * to arrive at the volume status. So error checking | |
4247 | * is being pushed to a separate loop. | |
4248 | */ | |
d10b82fe | 4249 | return check_barrier_error(info); |
387125fc | 4250 | } |
387125fc CM |
4251 | return 0; |
4252 | } | |
4253 | ||
943c6e99 ZL |
4254 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4255 | { | |
8789f4fe ZL |
4256 | int raid_type; |
4257 | int min_tolerated = INT_MAX; | |
943c6e99 | 4258 | |
8789f4fe ZL |
4259 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4260 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4261 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4262 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4263 | tolerated_failures); | |
943c6e99 | 4264 | |
8789f4fe ZL |
4265 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4266 | if (raid_type == BTRFS_RAID_SINGLE) | |
4267 | continue; | |
41a6e891 | 4268 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4269 | continue; |
8c3e3582 | 4270 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4271 | btrfs_raid_array[raid_type]. |
4272 | tolerated_failures); | |
4273 | } | |
943c6e99 | 4274 | |
8789f4fe | 4275 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4276 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4277 | min_tolerated = 0; |
4278 | } | |
4279 | ||
4280 | return min_tolerated; | |
943c6e99 ZL |
4281 | } |
4282 | ||
eece6a9c | 4283 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4284 | { |
e5e9a520 | 4285 | struct list_head *head; |
f2984462 | 4286 | struct btrfs_device *dev; |
a061fc8d | 4287 | struct btrfs_super_block *sb; |
f2984462 | 4288 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4289 | int ret; |
4290 | int do_barriers; | |
a236aed1 CM |
4291 | int max_errors; |
4292 | int total_errors = 0; | |
a061fc8d | 4293 | u64 flags; |
f2984462 | 4294 | |
0b246afa | 4295 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4296 | |
4297 | /* | |
4298 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4299 | * not from fsync where the tree roots in fs_info have not | |
4300 | * been consistent on disk. | |
4301 | */ | |
4302 | if (max_mirrors == 0) | |
4303 | backup_super_roots(fs_info); | |
f2984462 | 4304 | |
0b246afa | 4305 | sb = fs_info->super_for_commit; |
a061fc8d | 4306 | dev_item = &sb->dev_item; |
e5e9a520 | 4307 | |
0b246afa JM |
4308 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4309 | head = &fs_info->fs_devices->devices; | |
4310 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4311 | |
5af3e8cc | 4312 | if (do_barriers) { |
0b246afa | 4313 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4314 | if (ret) { |
4315 | mutex_unlock( | |
0b246afa JM |
4316 | &fs_info->fs_devices->device_list_mutex); |
4317 | btrfs_handle_fs_error(fs_info, ret, | |
4318 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4319 | return ret; |
4320 | } | |
4321 | } | |
387125fc | 4322 | |
1538e6c5 | 4323 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4324 | if (!dev->bdev) { |
4325 | total_errors++; | |
4326 | continue; | |
4327 | } | |
e12c9621 | 4328 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4329 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4330 | continue; |
4331 | ||
2b82032c | 4332 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4333 | btrfs_set_stack_device_type(dev_item, dev->type); |
4334 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4335 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4336 | dev->commit_total_bytes); |
ce7213c7 MX |
4337 | btrfs_set_stack_device_bytes_used(dev_item, |
4338 | dev->commit_bytes_used); | |
a061fc8d CM |
4339 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4340 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4341 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4342 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4343 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4344 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4345 | |
a061fc8d CM |
4346 | flags = btrfs_super_flags(sb); |
4347 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4348 | ||
75cb857d QW |
4349 | ret = btrfs_validate_write_super(fs_info, sb); |
4350 | if (ret < 0) { | |
4351 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4352 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4353 | "unexpected superblock corruption detected"); | |
4354 | return -EUCLEAN; | |
4355 | } | |
4356 | ||
abbb3b8e | 4357 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4358 | if (ret) |
4359 | total_errors++; | |
f2984462 | 4360 | } |
a236aed1 | 4361 | if (total_errors > max_errors) { |
0b246afa JM |
4362 | btrfs_err(fs_info, "%d errors while writing supers", |
4363 | total_errors); | |
4364 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4365 | |
9d565ba4 | 4366 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4367 | btrfs_handle_fs_error(fs_info, -EIO, |
4368 | "%d errors while writing supers", | |
4369 | total_errors); | |
9d565ba4 | 4370 | return -EIO; |
a236aed1 | 4371 | } |
f2984462 | 4372 | |
a512bbf8 | 4373 | total_errors = 0; |
1538e6c5 | 4374 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4375 | if (!dev->bdev) |
4376 | continue; | |
e12c9621 | 4377 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4378 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4379 | continue; |
4380 | ||
abbb3b8e | 4381 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4382 | if (ret) |
4383 | total_errors++; | |
f2984462 | 4384 | } |
0b246afa | 4385 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4386 | if (total_errors > max_errors) { |
0b246afa JM |
4387 | btrfs_handle_fs_error(fs_info, -EIO, |
4388 | "%d errors while writing supers", | |
4389 | total_errors); | |
79787eaa | 4390 | return -EIO; |
a236aed1 | 4391 | } |
f2984462 CM |
4392 | return 0; |
4393 | } | |
4394 | ||
cb517eab MX |
4395 | /* Drop a fs root from the radix tree and free it. */ |
4396 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4397 | struct btrfs_root *root) | |
2619ba1f | 4398 | { |
4785e24f JB |
4399 | bool drop_ref = false; |
4400 | ||
4df27c4d | 4401 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
4402 | radix_tree_delete(&fs_info->fs_roots_radix, |
4403 | (unsigned long)root->root_key.objectid); | |
af01d2e5 | 4404 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) |
4785e24f | 4405 | drop_ref = true; |
4df27c4d | 4406 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4407 | |
84961539 | 4408 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4409 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4410 | if (root->reloc_root) { |
00246528 | 4411 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4412 | root->reloc_root = NULL; |
4413 | } | |
4414 | } | |
3321719e | 4415 | |
4785e24f JB |
4416 | if (drop_ref) |
4417 | btrfs_put_root(root); | |
2619ba1f CM |
4418 | } |
4419 | ||
c146afad | 4420 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4421 | { |
c146afad YZ |
4422 | u64 root_objectid = 0; |
4423 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
4424 | int i = 0; |
4425 | int err = 0; | |
4426 | unsigned int ret = 0; | |
e089f05c | 4427 | |
c146afad | 4428 | while (1) { |
efc34534 | 4429 | spin_lock(&fs_info->fs_roots_radix_lock); |
c146afad YZ |
4430 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
4431 | (void **)gang, root_objectid, | |
4432 | ARRAY_SIZE(gang)); | |
65d33fd7 | 4433 | if (!ret) { |
efc34534 | 4434 | spin_unlock(&fs_info->fs_roots_radix_lock); |
c146afad | 4435 | break; |
65d33fd7 | 4436 | } |
5d4f98a2 | 4437 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4438 | |
c146afad | 4439 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
4440 | /* Avoid to grab roots in dead_roots */ |
4441 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4442 | gang[i] = NULL; | |
4443 | continue; | |
4444 | } | |
4445 | /* grab all the search result for later use */ | |
00246528 | 4446 | gang[i] = btrfs_grab_root(gang[i]); |
65d33fd7 | 4447 | } |
efc34534 | 4448 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4449 | |
65d33fd7 QW |
4450 | for (i = 0; i < ret; i++) { |
4451 | if (!gang[i]) | |
4452 | continue; | |
c146afad | 4453 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
4454 | err = btrfs_orphan_cleanup(gang[i]); |
4455 | if (err) | |
65d33fd7 | 4456 | break; |
00246528 | 4457 | btrfs_put_root(gang[i]); |
c146afad YZ |
4458 | } |
4459 | root_objectid++; | |
4460 | } | |
65d33fd7 QW |
4461 | |
4462 | /* release the uncleaned roots due to error */ | |
4463 | for (; i < ret; i++) { | |
4464 | if (gang[i]) | |
00246528 | 4465 | btrfs_put_root(gang[i]); |
65d33fd7 QW |
4466 | } |
4467 | return err; | |
c146afad | 4468 | } |
a2135011 | 4469 | |
6bccf3ab | 4470 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4471 | { |
6bccf3ab | 4472 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4473 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4474 | |
0b246afa | 4475 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4476 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4477 | mutex_unlock(&fs_info->cleaner_mutex); |
4478 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4479 | |
4480 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4481 | down_write(&fs_info->cleanup_work_sem); |
4482 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4483 | |
7a7eaa40 | 4484 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4485 | if (IS_ERR(trans)) |
4486 | return PTR_ERR(trans); | |
3a45bb20 | 4487 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4488 | } |
4489 | ||
36c86a9e QW |
4490 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4491 | { | |
4492 | struct btrfs_transaction *trans; | |
4493 | struct btrfs_transaction *tmp; | |
4494 | bool found = false; | |
4495 | ||
4496 | if (list_empty(&fs_info->trans_list)) | |
4497 | return; | |
4498 | ||
4499 | /* | |
4500 | * This function is only called at the very end of close_ctree(), | |
4501 | * thus no other running transaction, no need to take trans_lock. | |
4502 | */ | |
4503 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4504 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4505 | struct extent_state *cached = NULL; | |
4506 | u64 dirty_bytes = 0; | |
4507 | u64 cur = 0; | |
4508 | u64 found_start; | |
4509 | u64 found_end; | |
4510 | ||
4511 | found = true; | |
4512 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4513 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4514 | dirty_bytes += found_end + 1 - found_start; | |
4515 | cur = found_end + 1; | |
4516 | } | |
4517 | btrfs_warn(fs_info, | |
4518 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4519 | trans->transid, dirty_bytes); | |
4520 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4521 | ||
4522 | if (trans == fs_info->running_transaction) | |
4523 | fs_info->running_transaction = NULL; | |
4524 | list_del_init(&trans->list); | |
4525 | ||
4526 | btrfs_put_transaction(trans); | |
4527 | trace_btrfs_transaction_commit(fs_info); | |
4528 | } | |
4529 | ASSERT(!found); | |
4530 | } | |
4531 | ||
b105e927 | 4532 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4533 | { |
c146afad YZ |
4534 | int ret; |
4535 | ||
afcdd129 | 4536 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
d6fd0ae2 OS |
4537 | /* |
4538 | * We don't want the cleaner to start new transactions, add more delayed | |
4539 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4540 | * because that frees the task_struct, and the transaction kthread might | |
4541 | * still try to wake up the cleaner. | |
4542 | */ | |
4543 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4544 | |
b4be6aef JB |
4545 | /* |
4546 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4547 | * clear that bit and wake up relocation so it can stop. | |
4548 | */ | |
4549 | btrfs_wake_unfinished_drop(fs_info); | |
4550 | ||
7343dd61 | 4551 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4552 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4553 | |
803b2f54 SB |
4554 | /* wait for the uuid_scan task to finish */ |
4555 | down(&fs_info->uuid_tree_rescan_sem); | |
4556 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4557 | up(&fs_info->uuid_tree_rescan_sem); | |
4558 | ||
837d5b6e | 4559 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4560 | btrfs_pause_balance(fs_info); |
837d5b6e | 4561 | |
8dabb742 SB |
4562 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4563 | ||
aa1b8cd4 | 4564 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4565 | |
4566 | /* wait for any defraggers to finish */ | |
4567 | wait_event(fs_info->transaction_wait, | |
4568 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4569 | ||
4570 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4571 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4572 | |
21c7e756 | 4573 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4574 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4575 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4576 | |
18bb8bbf JT |
4577 | cancel_work_sync(&fs_info->reclaim_bgs_work); |
4578 | ||
b0643e59 DZ |
4579 | /* Cancel or finish ongoing discard work */ |
4580 | btrfs_discard_cleanup(fs_info); | |
4581 | ||
bc98a42c | 4582 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4583 | /* |
d6fd0ae2 OS |
4584 | * The cleaner kthread is stopped, so do one final pass over |
4585 | * unused block groups. | |
e44163e1 | 4586 | */ |
0b246afa | 4587 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4588 | |
f0cc2cd7 FM |
4589 | /* |
4590 | * There might be existing delayed inode workers still running | |
4591 | * and holding an empty delayed inode item. We must wait for | |
4592 | * them to complete first because they can create a transaction. | |
4593 | * This happens when someone calls btrfs_balance_delayed_items() | |
4594 | * and then a transaction commit runs the same delayed nodes | |
4595 | * before any delayed worker has done something with the nodes. | |
4596 | * We must wait for any worker here and not at transaction | |
4597 | * commit time since that could cause a deadlock. | |
4598 | * This is a very rare case. | |
4599 | */ | |
4600 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4601 | ||
6bccf3ab | 4602 | ret = btrfs_commit_super(fs_info); |
acce952b | 4603 | if (ret) |
04892340 | 4604 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4605 | } |
4606 | ||
84961539 | 4607 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4608 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4609 | |
e3029d9f AV |
4610 | kthread_stop(fs_info->transaction_kthread); |
4611 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4612 | |
e187831e | 4613 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4614 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4615 | |
5958253c QW |
4616 | if (btrfs_check_quota_leak(fs_info)) { |
4617 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4618 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4619 | } | |
4620 | ||
04892340 | 4621 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4622 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4623 | |
963d678b | 4624 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4625 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4626 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4627 | } |
bcc63abb | 4628 | |
5deb17e1 | 4629 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4630 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4631 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4632 | |
6618a59b | 4633 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4634 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4635 | |
1a4319cc LB |
4636 | btrfs_put_block_group_cache(fs_info); |
4637 | ||
de348ee0 WS |
4638 | /* |
4639 | * we must make sure there is not any read request to | |
4640 | * submit after we stopping all workers. | |
4641 | */ | |
4642 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4643 | btrfs_stop_all_workers(fs_info); |
4644 | ||
0a31daa4 | 4645 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4646 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4647 | |
afcdd129 | 4648 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4649 | free_root_pointers(fs_info, true); |
8c38938c | 4650 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4651 | |
4e19443d JB |
4652 | /* |
4653 | * We must free the block groups after dropping the fs_roots as we could | |
4654 | * have had an IO error and have left over tree log blocks that aren't | |
4655 | * cleaned up until the fs roots are freed. This makes the block group | |
4656 | * accounting appear to be wrong because there's pending reserved bytes, | |
4657 | * so make sure we do the block group cleanup afterwards. | |
4658 | */ | |
4659 | btrfs_free_block_groups(fs_info); | |
4660 | ||
13e6c37b | 4661 | iput(fs_info->btree_inode); |
d6bfde87 | 4662 | |
21adbd5c | 4663 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4664 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4665 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4666 | #endif |
4667 | ||
0b86a832 | 4668 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4669 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4670 | } |
4671 | ||
b9fab919 CM |
4672 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4673 | int atomic) | |
5f39d397 | 4674 | { |
1259ab75 | 4675 | int ret; |
727011e0 | 4676 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4677 | |
0b32f4bb | 4678 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4679 | if (!ret) |
4680 | return ret; | |
4681 | ||
4682 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4683 | parent_transid, atomic); |
4684 | if (ret == -EAGAIN) | |
4685 | return ret; | |
1259ab75 | 4686 | return !ret; |
5f39d397 CM |
4687 | } |
4688 | ||
5f39d397 CM |
4689 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4690 | { | |
2f4d60df | 4691 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4692 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4693 | int was_dirty; |
b4ce94de | 4694 | |
06ea65a3 JB |
4695 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4696 | /* | |
4697 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4698 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4699 | * outside of the sanity tests. |
4700 | */ | |
b0132a3b | 4701 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4702 | return; |
4703 | #endif | |
49d0c642 | 4704 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4705 | if (transid != fs_info->generation) |
5d163e0e | 4706 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4707 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4708 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4709 | if (!was_dirty) |
104b4e51 NB |
4710 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4711 | buf->len, | |
4712 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4713 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4714 | /* |
4715 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4716 | * but item data not updated. | |
4717 | * So here we should only check item pointers, not item data. | |
4718 | */ | |
4719 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4720 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4721 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4722 | ASSERT(0); |
4723 | } | |
4724 | #endif | |
eb60ceac CM |
4725 | } |
4726 | ||
2ff7e61e | 4727 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4728 | int flush_delayed) |
16cdcec7 MX |
4729 | { |
4730 | /* | |
4731 | * looks as though older kernels can get into trouble with | |
4732 | * this code, they end up stuck in balance_dirty_pages forever | |
4733 | */ | |
e2d84521 | 4734 | int ret; |
16cdcec7 MX |
4735 | |
4736 | if (current->flags & PF_MEMALLOC) | |
4737 | return; | |
4738 | ||
b53d3f5d | 4739 | if (flush_delayed) |
2ff7e61e | 4740 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4741 | |
d814a491 EL |
4742 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4743 | BTRFS_DIRTY_METADATA_THRESH, | |
4744 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4745 | if (ret > 0) { |
0b246afa | 4746 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4747 | } |
16cdcec7 MX |
4748 | } |
4749 | ||
2ff7e61e | 4750 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4751 | { |
2ff7e61e | 4752 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4753 | } |
585ad2c3 | 4754 | |
2ff7e61e | 4755 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4756 | { |
2ff7e61e | 4757 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4758 | } |
6b80053d | 4759 | |
581c1760 QW |
4760 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4761 | struct btrfs_key *first_key) | |
6b80053d | 4762 | { |
5ab12d1f | 4763 | return btree_read_extent_buffer_pages(buf, parent_transid, |
581c1760 | 4764 | level, first_key); |
6b80053d | 4765 | } |
0da5468f | 4766 | |
2ff7e61e | 4767 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4768 | { |
fe816d0f NB |
4769 | /* cleanup FS via transaction */ |
4770 | btrfs_cleanup_transaction(fs_info); | |
4771 | ||
0b246afa | 4772 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4773 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4774 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4775 | |
0b246afa JM |
4776 | down_write(&fs_info->cleanup_work_sem); |
4777 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4778 | } |
4779 | ||
ef67963d JB |
4780 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4781 | { | |
4782 | struct btrfs_root *gang[8]; | |
4783 | u64 root_objectid = 0; | |
4784 | int ret; | |
4785 | ||
4786 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4787 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4788 | (void **)gang, root_objectid, | |
4789 | ARRAY_SIZE(gang))) != 0) { | |
4790 | int i; | |
4791 | ||
4792 | for (i = 0; i < ret; i++) | |
4793 | gang[i] = btrfs_grab_root(gang[i]); | |
4794 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4795 | ||
4796 | for (i = 0; i < ret; i++) { | |
4797 | if (!gang[i]) | |
4798 | continue; | |
4799 | root_objectid = gang[i]->root_key.objectid; | |
4800 | btrfs_free_log(NULL, gang[i]); | |
4801 | btrfs_put_root(gang[i]); | |
4802 | } | |
4803 | root_objectid++; | |
4804 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4805 | } | |
4806 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4807 | btrfs_free_log_root_tree(NULL, fs_info); | |
4808 | } | |
4809 | ||
143bede5 | 4810 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4811 | { |
acce952b | 4812 | struct btrfs_ordered_extent *ordered; |
acce952b | 4813 | |
199c2a9c | 4814 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4815 | /* |
4816 | * This will just short circuit the ordered completion stuff which will | |
4817 | * make sure the ordered extent gets properly cleaned up. | |
4818 | */ | |
199c2a9c | 4819 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4820 | root_extent_list) |
4821 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4822 | spin_unlock(&root->ordered_extent_lock); |
4823 | } | |
4824 | ||
4825 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4826 | { | |
4827 | struct btrfs_root *root; | |
4828 | struct list_head splice; | |
4829 | ||
4830 | INIT_LIST_HEAD(&splice); | |
4831 | ||
4832 | spin_lock(&fs_info->ordered_root_lock); | |
4833 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4834 | while (!list_empty(&splice)) { | |
4835 | root = list_first_entry(&splice, struct btrfs_root, | |
4836 | ordered_root); | |
1de2cfde JB |
4837 | list_move_tail(&root->ordered_root, |
4838 | &fs_info->ordered_roots); | |
199c2a9c | 4839 | |
2a85d9ca | 4840 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4841 | btrfs_destroy_ordered_extents(root); |
4842 | ||
2a85d9ca LB |
4843 | cond_resched(); |
4844 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4845 | } |
4846 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4847 | |
4848 | /* | |
4849 | * We need this here because if we've been flipped read-only we won't | |
4850 | * get sync() from the umount, so we need to make sure any ordered | |
4851 | * extents that haven't had their dirty pages IO start writeout yet | |
4852 | * actually get run and error out properly. | |
4853 | */ | |
4854 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4855 | } |
4856 | ||
35a3621b | 4857 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4858 | struct btrfs_fs_info *fs_info) |
acce952b | 4859 | { |
4860 | struct rb_node *node; | |
4861 | struct btrfs_delayed_ref_root *delayed_refs; | |
4862 | struct btrfs_delayed_ref_node *ref; | |
4863 | int ret = 0; | |
4864 | ||
4865 | delayed_refs = &trans->delayed_refs; | |
4866 | ||
4867 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4868 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4869 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4870 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4871 | return ret; |
4872 | } | |
4873 | ||
5c9d028b | 4874 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4875 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4876 | struct rb_node *n; |
e78417d1 | 4877 | bool pin_bytes = false; |
acce952b | 4878 | |
d7df2c79 JB |
4879 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4880 | href_node); | |
3069bd26 | 4881 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4882 | continue; |
3069bd26 | 4883 | |
d7df2c79 | 4884 | spin_lock(&head->lock); |
e3d03965 | 4885 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4886 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4887 | ref_node); | |
d7df2c79 | 4888 | ref->in_tree = 0; |
e3d03965 | 4889 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4890 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4891 | if (!list_empty(&ref->add_list)) |
4892 | list_del(&ref->add_list); | |
d7df2c79 JB |
4893 | atomic_dec(&delayed_refs->num_entries); |
4894 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4895 | } |
d7df2c79 JB |
4896 | if (head->must_insert_reserved) |
4897 | pin_bytes = true; | |
4898 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4899 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4900 | spin_unlock(&head->lock); |
4901 | spin_unlock(&delayed_refs->lock); | |
4902 | mutex_unlock(&head->mutex); | |
acce952b | 4903 | |
f603bb94 NB |
4904 | if (pin_bytes) { |
4905 | struct btrfs_block_group *cache; | |
4906 | ||
4907 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4908 | BUG_ON(!cache); | |
4909 | ||
4910 | spin_lock(&cache->space_info->lock); | |
4911 | spin_lock(&cache->lock); | |
4912 | cache->pinned += head->num_bytes; | |
4913 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4914 | cache->space_info, head->num_bytes); | |
4915 | cache->reserved -= head->num_bytes; | |
4916 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4917 | spin_unlock(&cache->lock); | |
4918 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4919 | |
4920 | btrfs_put_block_group(cache); | |
4921 | ||
4922 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4923 | head->bytenr + head->num_bytes - 1); | |
4924 | } | |
31890da0 | 4925 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4926 | btrfs_put_delayed_ref_head(head); |
acce952b | 4927 | cond_resched(); |
4928 | spin_lock(&delayed_refs->lock); | |
4929 | } | |
81f7eb00 | 4930 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4931 | |
4932 | spin_unlock(&delayed_refs->lock); | |
4933 | ||
4934 | return ret; | |
4935 | } | |
4936 | ||
143bede5 | 4937 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4938 | { |
4939 | struct btrfs_inode *btrfs_inode; | |
4940 | struct list_head splice; | |
4941 | ||
4942 | INIT_LIST_HEAD(&splice); | |
4943 | ||
eb73c1b7 MX |
4944 | spin_lock(&root->delalloc_lock); |
4945 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4946 | |
4947 | while (!list_empty(&splice)) { | |
fe816d0f | 4948 | struct inode *inode = NULL; |
eb73c1b7 MX |
4949 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4950 | delalloc_inodes); | |
fe816d0f | 4951 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4952 | spin_unlock(&root->delalloc_lock); |
acce952b | 4953 | |
fe816d0f NB |
4954 | /* |
4955 | * Make sure we get a live inode and that it'll not disappear | |
4956 | * meanwhile. | |
4957 | */ | |
4958 | inode = igrab(&btrfs_inode->vfs_inode); | |
4959 | if (inode) { | |
4960 | invalidate_inode_pages2(inode->i_mapping); | |
4961 | iput(inode); | |
4962 | } | |
eb73c1b7 | 4963 | spin_lock(&root->delalloc_lock); |
acce952b | 4964 | } |
eb73c1b7 MX |
4965 | spin_unlock(&root->delalloc_lock); |
4966 | } | |
4967 | ||
4968 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4969 | { | |
4970 | struct btrfs_root *root; | |
4971 | struct list_head splice; | |
4972 | ||
4973 | INIT_LIST_HEAD(&splice); | |
4974 | ||
4975 | spin_lock(&fs_info->delalloc_root_lock); | |
4976 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4977 | while (!list_empty(&splice)) { | |
4978 | root = list_first_entry(&splice, struct btrfs_root, | |
4979 | delalloc_root); | |
00246528 | 4980 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4981 | BUG_ON(!root); |
4982 | spin_unlock(&fs_info->delalloc_root_lock); | |
4983 | ||
4984 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4985 | btrfs_put_root(root); |
eb73c1b7 MX |
4986 | |
4987 | spin_lock(&fs_info->delalloc_root_lock); | |
4988 | } | |
4989 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4990 | } |
4991 | ||
2ff7e61e | 4992 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4993 | struct extent_io_tree *dirty_pages, |
4994 | int mark) | |
4995 | { | |
4996 | int ret; | |
acce952b | 4997 | struct extent_buffer *eb; |
4998 | u64 start = 0; | |
4999 | u64 end; | |
acce952b | 5000 | |
5001 | while (1) { | |
5002 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5003 | mark, NULL); |
acce952b | 5004 | if (ret) |
5005 | break; | |
5006 | ||
91166212 | 5007 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5008 | while (start <= end) { |
0b246afa JM |
5009 | eb = find_extent_buffer(fs_info, start); |
5010 | start += fs_info->nodesize; | |
fd8b2b61 | 5011 | if (!eb) |
acce952b | 5012 | continue; |
fd8b2b61 | 5013 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5014 | |
fd8b2b61 JB |
5015 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5016 | &eb->bflags)) | |
5017 | clear_extent_buffer_dirty(eb); | |
5018 | free_extent_buffer_stale(eb); | |
acce952b | 5019 | } |
5020 | } | |
5021 | ||
5022 | return ret; | |
5023 | } | |
5024 | ||
2ff7e61e | 5025 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5026 | struct extent_io_tree *unpin) |
acce952b | 5027 | { |
acce952b | 5028 | u64 start; |
5029 | u64 end; | |
5030 | int ret; | |
5031 | ||
acce952b | 5032 | while (1) { |
0e6ec385 FM |
5033 | struct extent_state *cached_state = NULL; |
5034 | ||
fcd5e742 LF |
5035 | /* |
5036 | * The btrfs_finish_extent_commit() may get the same range as | |
5037 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5038 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5039 | * the same extent range. | |
5040 | */ | |
5041 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5042 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5043 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5044 | if (ret) { |
5045 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5046 | break; |
fcd5e742 | 5047 | } |
acce952b | 5048 | |
0e6ec385 FM |
5049 | clear_extent_dirty(unpin, start, end, &cached_state); |
5050 | free_extent_state(cached_state); | |
2ff7e61e | 5051 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5052 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5053 | cond_resched(); |
5054 | } | |
5055 | ||
5056 | return 0; | |
5057 | } | |
5058 | ||
32da5386 | 5059 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5060 | { |
5061 | struct inode *inode; | |
5062 | ||
5063 | inode = cache->io_ctl.inode; | |
5064 | if (inode) { | |
5065 | invalidate_inode_pages2(inode->i_mapping); | |
5066 | BTRFS_I(inode)->generation = 0; | |
5067 | cache->io_ctl.inode = NULL; | |
5068 | iput(inode); | |
5069 | } | |
bbc37d6e | 5070 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5071 | btrfs_put_block_group(cache); |
5072 | } | |
5073 | ||
5074 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5075 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5076 | { |
32da5386 | 5077 | struct btrfs_block_group *cache; |
c79a1751 LB |
5078 | |
5079 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5080 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5081 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5082 | struct btrfs_block_group, |
c79a1751 | 5083 | dirty_list); |
c79a1751 LB |
5084 | |
5085 | if (!list_empty(&cache->io_list)) { | |
5086 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5087 | list_del_init(&cache->io_list); | |
5088 | btrfs_cleanup_bg_io(cache); | |
5089 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5090 | } | |
5091 | ||
5092 | list_del_init(&cache->dirty_list); | |
5093 | spin_lock(&cache->lock); | |
5094 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5095 | spin_unlock(&cache->lock); | |
5096 | ||
5097 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5098 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5099 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5100 | spin_lock(&cur_trans->dirty_bgs_lock); |
5101 | } | |
5102 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5103 | ||
45ae2c18 NB |
5104 | /* |
5105 | * Refer to the definition of io_bgs member for details why it's safe | |
5106 | * to use it without any locking | |
5107 | */ | |
c79a1751 LB |
5108 | while (!list_empty(&cur_trans->io_bgs)) { |
5109 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5110 | struct btrfs_block_group, |
c79a1751 | 5111 | io_list); |
c79a1751 LB |
5112 | |
5113 | list_del_init(&cache->io_list); | |
5114 | spin_lock(&cache->lock); | |
5115 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5116 | spin_unlock(&cache->lock); | |
5117 | btrfs_cleanup_bg_io(cache); | |
5118 | } | |
5119 | } | |
5120 | ||
49b25e05 | 5121 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5122 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5123 | { |
bbbf7243 NB |
5124 | struct btrfs_device *dev, *tmp; |
5125 | ||
2ff7e61e | 5126 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5127 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5128 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5129 | ||
bbbf7243 NB |
5130 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5131 | post_commit_list) { | |
5132 | list_del_init(&dev->post_commit_list); | |
5133 | } | |
5134 | ||
2ff7e61e | 5135 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5136 | |
4a9d8bde | 5137 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5138 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5139 | |
4a9d8bde | 5140 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5141 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5142 | |
ccdf9b30 | 5143 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5144 | |
2ff7e61e | 5145 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5146 | EXTENT_DIRTY); |
fe119a6e | 5147 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5148 | |
d3575156 NA |
5149 | btrfs_free_redirty_list(cur_trans); |
5150 | ||
4a9d8bde MX |
5151 | cur_trans->state =TRANS_STATE_COMPLETED; |
5152 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5153 | } |
5154 | ||
2ff7e61e | 5155 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5156 | { |
5157 | struct btrfs_transaction *t; | |
acce952b | 5158 | |
0b246afa | 5159 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5160 | |
0b246afa JM |
5161 | spin_lock(&fs_info->trans_lock); |
5162 | while (!list_empty(&fs_info->trans_list)) { | |
5163 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5164 | struct btrfs_transaction, list); |
5165 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5166 | refcount_inc(&t->use_count); |
0b246afa | 5167 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5168 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5169 | btrfs_put_transaction(t); |
0b246afa | 5170 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5171 | continue; |
5172 | } | |
0b246afa | 5173 | if (t == fs_info->running_transaction) { |
724e2315 | 5174 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5175 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5176 | /* |
5177 | * We wait for 0 num_writers since we don't hold a trans | |
5178 | * handle open currently for this transaction. | |
5179 | */ | |
5180 | wait_event(t->writer_wait, | |
5181 | atomic_read(&t->num_writers) == 0); | |
5182 | } else { | |
0b246afa | 5183 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5184 | } |
2ff7e61e | 5185 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5186 | |
0b246afa JM |
5187 | spin_lock(&fs_info->trans_lock); |
5188 | if (t == fs_info->running_transaction) | |
5189 | fs_info->running_transaction = NULL; | |
acce952b | 5190 | list_del_init(&t->list); |
0b246afa | 5191 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5192 | |
724e2315 | 5193 | btrfs_put_transaction(t); |
2e4e97ab | 5194 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5195 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5196 | } |
0b246afa JM |
5197 | spin_unlock(&fs_info->trans_lock); |
5198 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5199 | btrfs_destroy_delayed_inodes(fs_info); |
5200 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5201 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5202 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5203 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5204 | |
5205 | return 0; | |
5206 | } | |
ec7d6dfd | 5207 | |
453e4873 | 5208 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5209 | { |
5210 | struct btrfs_path *path; | |
5211 | int ret; | |
5212 | struct extent_buffer *l; | |
5213 | struct btrfs_key search_key; | |
5214 | struct btrfs_key found_key; | |
5215 | int slot; | |
5216 | ||
5217 | path = btrfs_alloc_path(); | |
5218 | if (!path) | |
5219 | return -ENOMEM; | |
5220 | ||
5221 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5222 | search_key.type = -1; | |
5223 | search_key.offset = (u64)-1; | |
5224 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5225 | if (ret < 0) | |
5226 | goto error; | |
5227 | BUG_ON(ret == 0); /* Corruption */ | |
5228 | if (path->slots[0] > 0) { | |
5229 | slot = path->slots[0] - 1; | |
5230 | l = path->nodes[0]; | |
5231 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5232 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5233 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5234 | } else { |
23125104 | 5235 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5236 | } |
5237 | ret = 0; | |
5238 | error: | |
5239 | btrfs_free_path(path); | |
5240 | return ret; | |
5241 | } | |
5242 | ||
543068a2 | 5243 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5244 | { |
5245 | int ret; | |
5246 | mutex_lock(&root->objectid_mutex); | |
5247 | ||
6b8fad57 | 5248 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5249 | btrfs_warn(root->fs_info, |
5250 | "the objectid of root %llu reaches its highest value", | |
5251 | root->root_key.objectid); | |
5252 | ret = -ENOSPC; | |
5253 | goto out; | |
5254 | } | |
5255 | ||
23125104 | 5256 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5257 | ret = 0; |
5258 | out: | |
5259 | mutex_unlock(&root->objectid_mutex); | |
5260 | return ret; | |
5261 | } |