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> |
87cbda5c | 8 | #include <linux/scatterlist.h> |
22b0ebda | 9 | #include <linux/swap.h> |
0f7d52f4 | 10 | #include <linux/radix-tree.h> |
35b7e476 | 11 | #include <linux/writeback.h> |
d397712b | 12 | #include <linux/buffer_head.h> |
ce9adaa5 | 13 | #include <linux/workqueue.h> |
a74a4b97 | 14 | #include <linux/kthread.h> |
5a0e3ad6 | 15 | #include <linux/slab.h> |
784b4e29 | 16 | #include <linux/migrate.h> |
7a36ddec | 17 | #include <linux/ratelimit.h> |
6463fe58 | 18 | #include <linux/uuid.h> |
803b2f54 | 19 | #include <linux/semaphore.h> |
540adea3 | 20 | #include <linux/error-injection.h> |
9678c543 | 21 | #include <linux/crc32c.h> |
7e75bf3f | 22 | #include <asm/unaligned.h> |
eb60ceac CM |
23 | #include "ctree.h" |
24 | #include "disk-io.h" | |
e089f05c | 25 | #include "transaction.h" |
0f7d52f4 | 26 | #include "btrfs_inode.h" |
0b86a832 | 27 | #include "volumes.h" |
db94535d | 28 | #include "print-tree.h" |
925baedd | 29 | #include "locking.h" |
e02119d5 | 30 | #include "tree-log.h" |
fa9c0d79 | 31 | #include "free-space-cache.h" |
70f6d82e | 32 | #include "free-space-tree.h" |
581bb050 | 33 | #include "inode-map.h" |
21adbd5c | 34 | #include "check-integrity.h" |
606686ee | 35 | #include "rcu-string.h" |
8dabb742 | 36 | #include "dev-replace.h" |
53b381b3 | 37 | #include "raid56.h" |
5ac1d209 | 38 | #include "sysfs.h" |
fcebe456 | 39 | #include "qgroup.h" |
ebb8765b | 40 | #include "compression.h" |
557ea5dd | 41 | #include "tree-checker.h" |
fd708b81 | 42 | #include "ref-verify.h" |
eb60ceac | 43 | |
de0022b9 JB |
44 | #ifdef CONFIG_X86 |
45 | #include <asm/cpufeature.h> | |
46 | #endif | |
47 | ||
319e4d06 QW |
48 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
49 | BTRFS_HEADER_FLAG_RELOC |\ | |
50 | BTRFS_SUPER_FLAG_ERROR |\ | |
51 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
52 | BTRFS_SUPER_FLAG_METADUMP |\ |
53 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 54 | |
e8c9f186 | 55 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 56 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 57 | static void free_fs_root(struct btrfs_root *root); |
143bede5 | 58 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 59 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 60 | struct btrfs_fs_info *fs_info); |
143bede5 | 61 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 62 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 63 | struct extent_io_tree *dirty_pages, |
64 | int mark); | |
2ff7e61e | 65 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 66 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
67 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
68 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 69 | |
d352ac68 | 70 | /* |
97eb6b69 DS |
71 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
72 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
73 | * by writes to insert metadata for new file extents after IO is complete. |
74 | */ | |
97eb6b69 | 75 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
76 | struct bio *bio; |
77 | bio_end_io_t *end_io; | |
78 | void *private; | |
79 | struct btrfs_fs_info *info; | |
4e4cbee9 | 80 | blk_status_t status; |
bfebd8b5 | 81 | enum btrfs_wq_endio_type metadata; |
8b712842 | 82 | struct btrfs_work work; |
ce9adaa5 | 83 | }; |
0da5468f | 84 | |
97eb6b69 DS |
85 | static struct kmem_cache *btrfs_end_io_wq_cache; |
86 | ||
87 | int __init btrfs_end_io_wq_init(void) | |
88 | { | |
89 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
90 | sizeof(struct btrfs_end_io_wq), | |
91 | 0, | |
fba4b697 | 92 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
93 | NULL); |
94 | if (!btrfs_end_io_wq_cache) | |
95 | return -ENOMEM; | |
96 | return 0; | |
97 | } | |
98 | ||
e67c718b | 99 | void __cold btrfs_end_io_wq_exit(void) |
97eb6b69 | 100 | { |
5598e900 | 101 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
102 | } |
103 | ||
d352ac68 CM |
104 | /* |
105 | * async submit bios are used to offload expensive checksumming | |
106 | * onto the worker threads. They checksum file and metadata bios | |
107 | * just before they are sent down the IO stack. | |
108 | */ | |
44b8bd7e | 109 | struct async_submit_bio { |
c6100a4b JB |
110 | void *private_data; |
111 | struct btrfs_fs_info *fs_info; | |
44b8bd7e | 112 | struct bio *bio; |
a758781d DS |
113 | extent_submit_bio_start_t *submit_bio_start; |
114 | extent_submit_bio_done_t *submit_bio_done; | |
44b8bd7e | 115 | int mirror_num; |
c8b97818 | 116 | unsigned long bio_flags; |
eaf25d93 CM |
117 | /* |
118 | * bio_offset is optional, can be used if the pages in the bio | |
119 | * can't tell us where in the file the bio should go | |
120 | */ | |
121 | u64 bio_offset; | |
8b712842 | 122 | struct btrfs_work work; |
4e4cbee9 | 123 | blk_status_t status; |
44b8bd7e CM |
124 | }; |
125 | ||
85d4e461 CM |
126 | /* |
127 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
128 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
129 | * the level the eb occupies in the tree. | |
130 | * | |
131 | * Different roots are used for different purposes and may nest inside each | |
132 | * other and they require separate keysets. As lockdep keys should be | |
133 | * static, assign keysets according to the purpose of the root as indicated | |
134 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
135 | * have separate keysets. | |
4008c04a | 136 | * |
85d4e461 CM |
137 | * Lock-nesting across peer nodes is always done with the immediate parent |
138 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
139 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 140 | * |
85d4e461 CM |
141 | * The key is set by the readpage_end_io_hook after the buffer has passed |
142 | * csum validation but before the pages are unlocked. It is also set by | |
143 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 144 | * |
85d4e461 CM |
145 | * We also add a check to make sure the highest level of the tree is the |
146 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
147 | * needs update as well. | |
4008c04a CM |
148 | */ |
149 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
150 | # if BTRFS_MAX_LEVEL != 8 | |
151 | # error | |
152 | # endif | |
85d4e461 CM |
153 | |
154 | static struct btrfs_lockdep_keyset { | |
155 | u64 id; /* root objectid */ | |
156 | const char *name_stem; /* lock name stem */ | |
157 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
158 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
159 | } btrfs_lockdep_keysets[] = { | |
160 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
161 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
162 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
163 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
164 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
165 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 166 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
167 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
168 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
169 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 170 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 171 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 172 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 173 | }; |
85d4e461 CM |
174 | |
175 | void __init btrfs_init_lockdep(void) | |
176 | { | |
177 | int i, j; | |
178 | ||
179 | /* initialize lockdep class names */ | |
180 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
181 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
182 | ||
183 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
184 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
185 | "btrfs-%s-%02d", ks->name_stem, j); | |
186 | } | |
187 | } | |
188 | ||
189 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
190 | int level) | |
191 | { | |
192 | struct btrfs_lockdep_keyset *ks; | |
193 | ||
194 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
195 | ||
196 | /* find the matching keyset, id 0 is the default entry */ | |
197 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
198 | if (ks->id == objectid) | |
199 | break; | |
200 | ||
201 | lockdep_set_class_and_name(&eb->lock, | |
202 | &ks->keys[level], ks->names[level]); | |
203 | } | |
204 | ||
4008c04a CM |
205 | #endif |
206 | ||
d352ac68 CM |
207 | /* |
208 | * extents on the btree inode are pretty simple, there's one extent | |
209 | * that covers the entire device | |
210 | */ | |
6af49dbd | 211 | struct extent_map *btree_get_extent(struct btrfs_inode *inode, |
306e16ce | 212 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 213 | int create) |
7eccb903 | 214 | { |
fc4f21b1 NB |
215 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb); |
216 | struct extent_map_tree *em_tree = &inode->extent_tree; | |
5f39d397 CM |
217 | struct extent_map *em; |
218 | int ret; | |
219 | ||
890871be | 220 | read_lock(&em_tree->lock); |
d1310b2e | 221 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d | 222 | if (em) { |
0b246afa | 223 | em->bdev = fs_info->fs_devices->latest_bdev; |
890871be | 224 | read_unlock(&em_tree->lock); |
5f39d397 | 225 | goto out; |
a061fc8d | 226 | } |
890871be | 227 | read_unlock(&em_tree->lock); |
7b13b7b1 | 228 | |
172ddd60 | 229 | em = alloc_extent_map(); |
5f39d397 CM |
230 | if (!em) { |
231 | em = ERR_PTR(-ENOMEM); | |
232 | goto out; | |
233 | } | |
234 | em->start = 0; | |
0afbaf8c | 235 | em->len = (u64)-1; |
c8b97818 | 236 | em->block_len = (u64)-1; |
5f39d397 | 237 | em->block_start = 0; |
0b246afa | 238 | em->bdev = fs_info->fs_devices->latest_bdev; |
d1310b2e | 239 | |
890871be | 240 | write_lock(&em_tree->lock); |
09a2a8f9 | 241 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
242 | if (ret == -EEXIST) { |
243 | free_extent_map(em); | |
7b13b7b1 | 244 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 245 | if (!em) |
0433f20d | 246 | em = ERR_PTR(-EIO); |
5f39d397 | 247 | } else if (ret) { |
7b13b7b1 | 248 | free_extent_map(em); |
0433f20d | 249 | em = ERR_PTR(ret); |
5f39d397 | 250 | } |
890871be | 251 | write_unlock(&em_tree->lock); |
7b13b7b1 | 252 | |
5f39d397 CM |
253 | out: |
254 | return em; | |
7eccb903 CM |
255 | } |
256 | ||
9ed57367 | 257 | u32 btrfs_csum_data(const char *data, u32 seed, size_t len) |
19c00ddc | 258 | { |
9678c543 | 259 | return crc32c(seed, data, len); |
19c00ddc CM |
260 | } |
261 | ||
0b5e3daf | 262 | void btrfs_csum_final(u32 crc, u8 *result) |
19c00ddc | 263 | { |
7e75bf3f | 264 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
265 | } |
266 | ||
d352ac68 CM |
267 | /* |
268 | * compute the csum for a btree block, and either verify it or write it | |
269 | * into the csum field of the block. | |
270 | */ | |
01d58472 DD |
271 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
272 | struct extent_buffer *buf, | |
19c00ddc CM |
273 | int verify) |
274 | { | |
01d58472 | 275 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
71a63551 | 276 | char result[BTRFS_CSUM_SIZE]; |
19c00ddc CM |
277 | unsigned long len; |
278 | unsigned long cur_len; | |
279 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
280 | char *kaddr; |
281 | unsigned long map_start; | |
282 | unsigned long map_len; | |
283 | int err; | |
284 | u32 crc = ~(u32)0; | |
285 | ||
286 | len = buf->len - offset; | |
d397712b | 287 | while (len > 0) { |
19c00ddc | 288 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 289 | &kaddr, &map_start, &map_len); |
d397712b | 290 | if (err) |
8bd98f0e | 291 | return err; |
19c00ddc | 292 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 293 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
294 | crc, cur_len); |
295 | len -= cur_len; | |
296 | offset += cur_len; | |
19c00ddc | 297 | } |
71a63551 | 298 | memset(result, 0, BTRFS_CSUM_SIZE); |
607d432d | 299 | |
19c00ddc CM |
300 | btrfs_csum_final(crc, result); |
301 | ||
302 | if (verify) { | |
607d432d | 303 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
304 | u32 val; |
305 | u32 found = 0; | |
607d432d | 306 | memcpy(&found, result, csum_size); |
e4204ded | 307 | |
607d432d | 308 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 | 309 | btrfs_warn_rl(fs_info, |
5d163e0e | 310 | "%s checksum verify failed on %llu wanted %X found %X level %d", |
01d58472 | 311 | fs_info->sb->s_id, buf->start, |
efe120a0 | 312 | val, found, btrfs_header_level(buf)); |
8bd98f0e | 313 | return -EUCLEAN; |
19c00ddc CM |
314 | } |
315 | } else { | |
607d432d | 316 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 317 | } |
71a63551 | 318 | |
19c00ddc CM |
319 | return 0; |
320 | } | |
321 | ||
d352ac68 CM |
322 | /* |
323 | * we can't consider a given block up to date unless the transid of the | |
324 | * block matches the transid in the parent node's pointer. This is how we | |
325 | * detect blocks that either didn't get written at all or got written | |
326 | * in the wrong place. | |
327 | */ | |
1259ab75 | 328 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
329 | struct extent_buffer *eb, u64 parent_transid, |
330 | int atomic) | |
1259ab75 | 331 | { |
2ac55d41 | 332 | struct extent_state *cached_state = NULL; |
1259ab75 | 333 | int ret; |
2755a0de | 334 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
335 | |
336 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
337 | return 0; | |
338 | ||
b9fab919 CM |
339 | if (atomic) |
340 | return -EAGAIN; | |
341 | ||
a26e8c9f JB |
342 | if (need_lock) { |
343 | btrfs_tree_read_lock(eb); | |
344 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
345 | } | |
346 | ||
2ac55d41 | 347 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 348 | &cached_state); |
0b32f4bb | 349 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
350 | btrfs_header_generation(eb) == parent_transid) { |
351 | ret = 0; | |
352 | goto out; | |
353 | } | |
94647322 DS |
354 | btrfs_err_rl(eb->fs_info, |
355 | "parent transid verify failed on %llu wanted %llu found %llu", | |
356 | eb->start, | |
29549aec | 357 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 358 | ret = 1; |
a26e8c9f JB |
359 | |
360 | /* | |
361 | * Things reading via commit roots that don't have normal protection, | |
362 | * like send, can have a really old block in cache that may point at a | |
01327610 | 363 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
364 | * if we find an eb that is under IO (dirty/writeback) because we could |
365 | * end up reading in the stale data and then writing it back out and | |
366 | * making everybody very sad. | |
367 | */ | |
368 | if (!extent_buffer_under_io(eb)) | |
369 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 370 | out: |
2ac55d41 | 371 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
e43bbe5e | 372 | &cached_state); |
472b909f JB |
373 | if (need_lock) |
374 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 375 | return ret; |
1259ab75 CM |
376 | } |
377 | ||
1104a885 DS |
378 | /* |
379 | * Return 0 if the superblock checksum type matches the checksum value of that | |
380 | * algorithm. Pass the raw disk superblock data. | |
381 | */ | |
ab8d0fc4 JM |
382 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
383 | char *raw_disk_sb) | |
1104a885 DS |
384 | { |
385 | struct btrfs_super_block *disk_sb = | |
386 | (struct btrfs_super_block *)raw_disk_sb; | |
387 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
388 | int ret = 0; | |
389 | ||
390 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
391 | u32 crc = ~(u32)0; | |
776c4a7c | 392 | char result[sizeof(crc)]; |
1104a885 DS |
393 | |
394 | /* | |
395 | * The super_block structure does not span the whole | |
396 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
01327610 | 397 | * is filled with zeros and is included in the checksum. |
1104a885 DS |
398 | */ |
399 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
400 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
401 | btrfs_csum_final(crc, result); | |
402 | ||
776c4a7c | 403 | if (memcmp(raw_disk_sb, result, sizeof(result))) |
1104a885 DS |
404 | ret = 1; |
405 | } | |
406 | ||
407 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
ab8d0fc4 | 408 | btrfs_err(fs_info, "unsupported checksum algorithm %u", |
1104a885 DS |
409 | csum_type); |
410 | ret = 1; | |
411 | } | |
412 | ||
413 | return ret; | |
414 | } | |
415 | ||
581c1760 QW |
416 | static int verify_level_key(struct btrfs_fs_info *fs_info, |
417 | struct extent_buffer *eb, int level, | |
ff76a864 | 418 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 QW |
419 | { |
420 | int found_level; | |
421 | struct btrfs_key found_key; | |
422 | int ret; | |
423 | ||
424 | found_level = btrfs_header_level(eb); | |
425 | if (found_level != level) { | |
426 | #ifdef CONFIG_BTRFS_DEBUG | |
427 | WARN_ON(1); | |
428 | btrfs_err(fs_info, | |
429 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
430 | eb->start, level, found_level); | |
431 | #endif | |
432 | return -EIO; | |
433 | } | |
434 | ||
435 | if (!first_key) | |
436 | return 0; | |
437 | ||
5d41be6f QW |
438 | /* |
439 | * For live tree block (new tree blocks in current transaction), | |
440 | * we need proper lock context to avoid race, which is impossible here. | |
441 | * So we only checks tree blocks which is read from disk, whose | |
442 | * generation <= fs_info->last_trans_committed. | |
443 | */ | |
444 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
445 | return 0; | |
581c1760 QW |
446 | if (found_level) |
447 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
448 | else | |
449 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
450 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
451 | ||
452 | #ifdef CONFIG_BTRFS_DEBUG | |
453 | if (ret) { | |
454 | WARN_ON(1); | |
455 | btrfs_err(fs_info, | |
ff76a864 LB |
456 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
457 | eb->start, parent_transid, first_key->objectid, | |
458 | first_key->type, first_key->offset, | |
459 | found_key.objectid, found_key.type, | |
460 | found_key.offset); | |
581c1760 QW |
461 | } |
462 | #endif | |
463 | return ret; | |
464 | } | |
465 | ||
d352ac68 CM |
466 | /* |
467 | * helper to read a given tree block, doing retries as required when | |
468 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
469 | * |
470 | * @parent_transid: expected transid, skip check if 0 | |
471 | * @level: expected level, mandatory check | |
472 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 473 | */ |
2ff7e61e | 474 | static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, |
f188591e | 475 | struct extent_buffer *eb, |
581c1760 QW |
476 | u64 parent_transid, int level, |
477 | struct btrfs_key *first_key) | |
f188591e CM |
478 | { |
479 | struct extent_io_tree *io_tree; | |
ea466794 | 480 | int failed = 0; |
f188591e CM |
481 | int ret; |
482 | int num_copies = 0; | |
483 | int mirror_num = 0; | |
ea466794 | 484 | int failed_mirror = 0; |
f188591e | 485 | |
a826d6dc | 486 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
0b246afa | 487 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 488 | while (1) { |
8436ea91 | 489 | ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, |
6af49dbd | 490 | mirror_num); |
256dd1bb | 491 | if (!ret) { |
581c1760 | 492 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 493 | parent_transid, 0)) |
256dd1bb | 494 | ret = -EIO; |
581c1760 | 495 | else if (verify_level_key(fs_info, eb, level, |
ff76a864 | 496 | first_key, parent_transid)) |
581c1760 QW |
497 | ret = -EUCLEAN; |
498 | else | |
499 | break; | |
256dd1bb | 500 | } |
d397712b | 501 | |
a826d6dc JB |
502 | /* |
503 | * This buffer's crc is fine, but its contents are corrupted, so | |
504 | * there is no reason to read the other copies, they won't be | |
505 | * any less wrong. | |
506 | */ | |
581c1760 QW |
507 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags) || |
508 | ret == -EUCLEAN) | |
ea466794 JB |
509 | break; |
510 | ||
0b246afa | 511 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 512 | eb->start, eb->len); |
4235298e | 513 | if (num_copies == 1) |
ea466794 | 514 | break; |
4235298e | 515 | |
5cf1ab56 JB |
516 | if (!failed_mirror) { |
517 | failed = 1; | |
518 | failed_mirror = eb->read_mirror; | |
519 | } | |
520 | ||
f188591e | 521 | mirror_num++; |
ea466794 JB |
522 | if (mirror_num == failed_mirror) |
523 | mirror_num++; | |
524 | ||
4235298e | 525 | if (mirror_num > num_copies) |
ea466794 | 526 | break; |
f188591e | 527 | } |
ea466794 | 528 | |
c0901581 | 529 | if (failed && !ret && failed_mirror) |
2ff7e61e | 530 | repair_eb_io_failure(fs_info, eb, failed_mirror); |
ea466794 JB |
531 | |
532 | return ret; | |
f188591e | 533 | } |
19c00ddc | 534 | |
d352ac68 | 535 | /* |
d397712b CM |
536 | * checksum a dirty tree block before IO. This has extra checks to make sure |
537 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 538 | */ |
d397712b | 539 | |
01d58472 | 540 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 541 | { |
4eee4fa4 | 542 | u64 start = page_offset(page); |
19c00ddc | 543 | u64 found_start; |
19c00ddc | 544 | struct extent_buffer *eb; |
f188591e | 545 | |
4f2de97a JB |
546 | eb = (struct extent_buffer *)page->private; |
547 | if (page != eb->pages[0]) | |
548 | return 0; | |
0f805531 | 549 | |
19c00ddc | 550 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
551 | /* |
552 | * Please do not consolidate these warnings into a single if. | |
553 | * It is useful to know what went wrong. | |
554 | */ | |
555 | if (WARN_ON(found_start != start)) | |
556 | return -EUCLEAN; | |
557 | if (WARN_ON(!PageUptodate(page))) | |
558 | return -EUCLEAN; | |
559 | ||
560 | ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, | |
561 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); | |
562 | ||
8bd98f0e | 563 | return csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
564 | } |
565 | ||
01d58472 | 566 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
567 | struct extent_buffer *eb) |
568 | { | |
01d58472 | 569 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
44880fdc | 570 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c YZ |
571 | int ret = 1; |
572 | ||
0a4e5586 | 573 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
574 | while (fs_devices) { |
575 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
576 | ret = 0; | |
577 | break; | |
578 | } | |
579 | fs_devices = fs_devices->seed; | |
580 | } | |
581 | return ret; | |
582 | } | |
583 | ||
facc8a22 MX |
584 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
585 | u64 phy_offset, struct page *page, | |
586 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 587 | { |
ce9adaa5 CM |
588 | u64 found_start; |
589 | int found_level; | |
ce9adaa5 CM |
590 | struct extent_buffer *eb; |
591 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 592 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 593 | int ret = 0; |
727011e0 | 594 | int reads_done; |
ce9adaa5 | 595 | |
ce9adaa5 CM |
596 | if (!page->private) |
597 | goto out; | |
d397712b | 598 | |
4f2de97a | 599 | eb = (struct extent_buffer *)page->private; |
d397712b | 600 | |
0b32f4bb JB |
601 | /* the pending IO might have been the only thing that kept this buffer |
602 | * in memory. Make sure we have a ref for all this other checks | |
603 | */ | |
604 | extent_buffer_get(eb); | |
605 | ||
606 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
607 | if (!reads_done) |
608 | goto err; | |
f188591e | 609 | |
5cf1ab56 | 610 | eb->read_mirror = mirror; |
656f30db | 611 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
612 | ret = -EIO; |
613 | goto err; | |
614 | } | |
615 | ||
ce9adaa5 | 616 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 617 | if (found_start != eb->start) { |
02873e43 ZL |
618 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
619 | found_start, eb->start); | |
f188591e | 620 | ret = -EIO; |
ce9adaa5 CM |
621 | goto err; |
622 | } | |
02873e43 ZL |
623 | if (check_tree_block_fsid(fs_info, eb)) { |
624 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
625 | eb->start); | |
1259ab75 CM |
626 | ret = -EIO; |
627 | goto err; | |
628 | } | |
ce9adaa5 | 629 | found_level = btrfs_header_level(eb); |
1c24c3ce | 630 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
631 | btrfs_err(fs_info, "bad tree block level %d", |
632 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
633 | ret = -EIO; |
634 | goto err; | |
635 | } | |
ce9adaa5 | 636 | |
85d4e461 CM |
637 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
638 | eb, found_level); | |
4008c04a | 639 | |
02873e43 | 640 | ret = csum_tree_block(fs_info, eb, 1); |
8bd98f0e | 641 | if (ret) |
a826d6dc | 642 | goto err; |
a826d6dc JB |
643 | |
644 | /* | |
645 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
646 | * that we don't try and read the other copies of this block, just | |
647 | * return -EIO. | |
648 | */ | |
2f659546 | 649 | if (found_level == 0 && btrfs_check_leaf_full(fs_info, eb)) { |
a826d6dc JB |
650 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
651 | ret = -EIO; | |
652 | } | |
ce9adaa5 | 653 | |
2f659546 | 654 | if (found_level > 0 && btrfs_check_node(fs_info, eb)) |
053ab70f LB |
655 | ret = -EIO; |
656 | ||
0b32f4bb JB |
657 | if (!ret) |
658 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 659 | err: |
79fb65a1 JB |
660 | if (reads_done && |
661 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 662 | btree_readahead_hook(eb, ret); |
4bb31e92 | 663 | |
53b381b3 DW |
664 | if (ret) { |
665 | /* | |
666 | * our io error hook is going to dec the io pages | |
667 | * again, we have to make sure it has something | |
668 | * to decrement | |
669 | */ | |
670 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 671 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 672 | } |
0b32f4bb | 673 | free_extent_buffer(eb); |
ce9adaa5 | 674 | out: |
f188591e | 675 | return ret; |
ce9adaa5 CM |
676 | } |
677 | ||
ea466794 | 678 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 679 | { |
4bb31e92 | 680 | struct extent_buffer *eb; |
4bb31e92 | 681 | |
4f2de97a | 682 | eb = (struct extent_buffer *)page->private; |
656f30db | 683 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 684 | eb->read_mirror = failed_mirror; |
53b381b3 | 685 | atomic_dec(&eb->io_pages); |
ea466794 | 686 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
d48d71aa | 687 | btree_readahead_hook(eb, -EIO); |
4bb31e92 AJ |
688 | return -EIO; /* we fixed nothing */ |
689 | } | |
690 | ||
4246a0b6 | 691 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 692 | { |
97eb6b69 | 693 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 694 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
695 | struct btrfs_workqueue *wq; |
696 | btrfs_work_func_t func; | |
ce9adaa5 | 697 | |
ce9adaa5 | 698 | fs_info = end_io_wq->info; |
4e4cbee9 | 699 | end_io_wq->status = bio->bi_status; |
d20f7043 | 700 | |
37226b21 | 701 | if (bio_op(bio) == REQ_OP_WRITE) { |
9e0af237 LB |
702 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
703 | wq = fs_info->endio_meta_write_workers; | |
704 | func = btrfs_endio_meta_write_helper; | |
705 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
706 | wq = fs_info->endio_freespace_worker; | |
707 | func = btrfs_freespace_write_helper; | |
708 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
709 | wq = fs_info->endio_raid56_workers; | |
710 | func = btrfs_endio_raid56_helper; | |
711 | } else { | |
712 | wq = fs_info->endio_write_workers; | |
713 | func = btrfs_endio_write_helper; | |
714 | } | |
d20f7043 | 715 | } else { |
8b110e39 MX |
716 | if (unlikely(end_io_wq->metadata == |
717 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
718 | wq = fs_info->endio_repair_workers; | |
719 | func = btrfs_endio_repair_helper; | |
720 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
721 | wq = fs_info->endio_raid56_workers; |
722 | func = btrfs_endio_raid56_helper; | |
723 | } else if (end_io_wq->metadata) { | |
724 | wq = fs_info->endio_meta_workers; | |
725 | func = btrfs_endio_meta_helper; | |
726 | } else { | |
727 | wq = fs_info->endio_workers; | |
728 | func = btrfs_endio_helper; | |
729 | } | |
d20f7043 | 730 | } |
9e0af237 LB |
731 | |
732 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
733 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
734 | } |
735 | ||
4e4cbee9 | 736 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 737 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 738 | { |
97eb6b69 | 739 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 740 | |
97eb6b69 | 741 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 742 | if (!end_io_wq) |
4e4cbee9 | 743 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
744 | |
745 | end_io_wq->private = bio->bi_private; | |
746 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 747 | end_io_wq->info = info; |
4e4cbee9 | 748 | end_io_wq->status = 0; |
ce9adaa5 | 749 | end_io_wq->bio = bio; |
22c59948 | 750 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
751 | |
752 | bio->bi_private = end_io_wq; | |
753 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
754 | return 0; |
755 | } | |
756 | ||
4a69a410 CM |
757 | static void run_one_async_start(struct btrfs_work *work) |
758 | { | |
4a69a410 | 759 | struct async_submit_bio *async; |
4e4cbee9 | 760 | blk_status_t ret; |
4a69a410 CM |
761 | |
762 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 763 | ret = async->submit_bio_start(async->private_data, async->bio, |
79787eaa JM |
764 | async->bio_offset); |
765 | if (ret) | |
4e4cbee9 | 766 | async->status = ret; |
4a69a410 CM |
767 | } |
768 | ||
769 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 | 770 | { |
8b712842 CM |
771 | struct async_submit_bio *async; |
772 | ||
773 | async = container_of(work, struct async_submit_bio, work); | |
4854ddd0 | 774 | |
bb7ab3b9 | 775 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
776 | if (async->status) { |
777 | async->bio->bi_status = async->status; | |
4246a0b6 | 778 | bio_endio(async->bio); |
79787eaa JM |
779 | return; |
780 | } | |
781 | ||
6c553435 | 782 | async->submit_bio_done(async->private_data, async->bio, async->mirror_num); |
4a69a410 CM |
783 | } |
784 | ||
785 | static void run_one_async_free(struct btrfs_work *work) | |
786 | { | |
787 | struct async_submit_bio *async; | |
788 | ||
789 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
790 | kfree(async); |
791 | } | |
792 | ||
8c27cb35 LT |
793 | blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, |
794 | int mirror_num, unsigned long bio_flags, | |
795 | u64 bio_offset, void *private_data, | |
a758781d DS |
796 | extent_submit_bio_start_t *submit_bio_start, |
797 | extent_submit_bio_done_t *submit_bio_done) | |
44b8bd7e CM |
798 | { |
799 | struct async_submit_bio *async; | |
800 | ||
801 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
802 | if (!async) | |
4e4cbee9 | 803 | return BLK_STS_RESOURCE; |
44b8bd7e | 804 | |
c6100a4b JB |
805 | async->private_data = private_data; |
806 | async->fs_info = fs_info; | |
44b8bd7e CM |
807 | async->bio = bio; |
808 | async->mirror_num = mirror_num; | |
4a69a410 CM |
809 | async->submit_bio_start = submit_bio_start; |
810 | async->submit_bio_done = submit_bio_done; | |
811 | ||
9e0af237 | 812 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 813 | run_one_async_done, run_one_async_free); |
4a69a410 | 814 | |
c8b97818 | 815 | async->bio_flags = bio_flags; |
eaf25d93 | 816 | async->bio_offset = bio_offset; |
8c8bee1d | 817 | |
4e4cbee9 | 818 | async->status = 0; |
79787eaa | 819 | |
67f055c7 | 820 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 821 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 822 | |
5cdc7ad3 | 823 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
824 | return 0; |
825 | } | |
826 | ||
4e4cbee9 | 827 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 828 | { |
2c30c71b | 829 | struct bio_vec *bvec; |
ce3ed71a | 830 | struct btrfs_root *root; |
2c30c71b | 831 | int i, ret = 0; |
ce3ed71a | 832 | |
c09abff8 | 833 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2c30c71b | 834 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 835 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 836 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
837 | if (ret) |
838 | break; | |
ce3ed71a | 839 | } |
2c30c71b | 840 | |
4e4cbee9 | 841 | return errno_to_blk_status(ret); |
ce3ed71a CM |
842 | } |
843 | ||
d0ee3934 | 844 | static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, |
8c27cb35 | 845 | u64 bio_offset) |
22c59948 | 846 | { |
8b712842 CM |
847 | /* |
848 | * when we're called for a write, we're already in the async | |
5443be45 | 849 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 850 | */ |
79787eaa | 851 | return btree_csum_one_bio(bio); |
4a69a410 | 852 | } |
22c59948 | 853 | |
d0ee3934 | 854 | static blk_status_t btree_submit_bio_done(void *private_data, struct bio *bio, |
6c553435 | 855 | int mirror_num) |
4a69a410 | 856 | { |
c6100a4b | 857 | struct inode *inode = private_data; |
4e4cbee9 | 858 | blk_status_t ret; |
61891923 | 859 | |
8b712842 | 860 | /* |
4a69a410 CM |
861 | * when we're called for a write, we're already in the async |
862 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 863 | */ |
2ff7e61e | 864 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1); |
4246a0b6 | 865 | if (ret) { |
4e4cbee9 | 866 | bio->bi_status = ret; |
4246a0b6 CH |
867 | bio_endio(bio); |
868 | } | |
61891923 | 869 | return ret; |
0b86a832 CM |
870 | } |
871 | ||
18fdc679 | 872 | static int check_async_write(struct btrfs_inode *bi) |
de0022b9 | 873 | { |
6300463b LB |
874 | if (atomic_read(&bi->sync_writers)) |
875 | return 0; | |
de0022b9 | 876 | #ifdef CONFIG_X86 |
bc696ca0 | 877 | if (static_cpu_has(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
878 | return 0; |
879 | #endif | |
880 | return 1; | |
881 | } | |
882 | ||
8c27cb35 LT |
883 | static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, |
884 | int mirror_num, unsigned long bio_flags, | |
885 | u64 bio_offset) | |
44b8bd7e | 886 | { |
c6100a4b | 887 | struct inode *inode = private_data; |
0b246afa | 888 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
18fdc679 | 889 | int async = check_async_write(BTRFS_I(inode)); |
4e4cbee9 | 890 | blk_status_t ret; |
cad321ad | 891 | |
37226b21 | 892 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
893 | /* |
894 | * called for a read, do the setup so that checksum validation | |
895 | * can happen in the async kernel threads | |
896 | */ | |
0b246afa JM |
897 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
898 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 899 | if (ret) |
61891923 | 900 | goto out_w_error; |
2ff7e61e | 901 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
de0022b9 JB |
902 | } else if (!async) { |
903 | ret = btree_csum_one_bio(bio); | |
904 | if (ret) | |
61891923 | 905 | goto out_w_error; |
2ff7e61e | 906 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
61891923 SB |
907 | } else { |
908 | /* | |
909 | * kthread helpers are used to submit writes so that | |
910 | * checksumming can happen in parallel across all CPUs | |
911 | */ | |
c6100a4b JB |
912 | ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, |
913 | bio_offset, private_data, | |
d0ee3934 DS |
914 | btree_submit_bio_start, |
915 | btree_submit_bio_done); | |
44b8bd7e | 916 | } |
d313d7a3 | 917 | |
4246a0b6 CH |
918 | if (ret) |
919 | goto out_w_error; | |
920 | return 0; | |
921 | ||
61891923 | 922 | out_w_error: |
4e4cbee9 | 923 | bio->bi_status = ret; |
4246a0b6 | 924 | bio_endio(bio); |
61891923 | 925 | return ret; |
44b8bd7e CM |
926 | } |
927 | ||
3dd1462e | 928 | #ifdef CONFIG_MIGRATION |
784b4e29 | 929 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
930 | struct page *newpage, struct page *page, |
931 | enum migrate_mode mode) | |
784b4e29 CM |
932 | { |
933 | /* | |
934 | * we can't safely write a btree page from here, | |
935 | * we haven't done the locking hook | |
936 | */ | |
937 | if (PageDirty(page)) | |
938 | return -EAGAIN; | |
939 | /* | |
940 | * Buffers may be managed in a filesystem specific way. | |
941 | * We must have no buffers or drop them. | |
942 | */ | |
943 | if (page_has_private(page) && | |
944 | !try_to_release_page(page, GFP_KERNEL)) | |
945 | return -EAGAIN; | |
a6bc32b8 | 946 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 947 | } |
3dd1462e | 948 | #endif |
784b4e29 | 949 | |
0da5468f CM |
950 | |
951 | static int btree_writepages(struct address_space *mapping, | |
952 | struct writeback_control *wbc) | |
953 | { | |
e2d84521 MX |
954 | struct btrfs_fs_info *fs_info; |
955 | int ret; | |
956 | ||
d8d5f3e1 | 957 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
958 | |
959 | if (wbc->for_kupdate) | |
960 | return 0; | |
961 | ||
e2d84521 | 962 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 963 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
964 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
965 | BTRFS_DIRTY_METADATA_THRESH); | |
966 | if (ret < 0) | |
793955bc | 967 | return 0; |
793955bc | 968 | } |
0b32f4bb | 969 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
970 | } |
971 | ||
b2950863 | 972 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 973 | { |
d1310b2e CM |
974 | struct extent_io_tree *tree; |
975 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 976 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 977 | } |
22b0ebda | 978 | |
70dec807 | 979 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 980 | { |
98509cfc | 981 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 982 | return 0; |
0c4e538b | 983 | |
f7a52a40 | 984 | return try_release_extent_buffer(page); |
d98237b3 CM |
985 | } |
986 | ||
d47992f8 LC |
987 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
988 | unsigned int length) | |
d98237b3 | 989 | { |
d1310b2e CM |
990 | struct extent_io_tree *tree; |
991 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
992 | extent_invalidatepage(tree, page, offset); |
993 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 994 | if (PagePrivate(page)) { |
efe120a0 FH |
995 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
996 | "page private not zero on page %llu", | |
997 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
998 | ClearPagePrivate(page); |
999 | set_page_private(page, 0); | |
09cbfeaf | 1000 | put_page(page); |
9ad6b7bc | 1001 | } |
d98237b3 CM |
1002 | } |
1003 | ||
0b32f4bb JB |
1004 | static int btree_set_page_dirty(struct page *page) |
1005 | { | |
bb146eb2 | 1006 | #ifdef DEBUG |
0b32f4bb JB |
1007 | struct extent_buffer *eb; |
1008 | ||
1009 | BUG_ON(!PagePrivate(page)); | |
1010 | eb = (struct extent_buffer *)page->private; | |
1011 | BUG_ON(!eb); | |
1012 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1013 | BUG_ON(!atomic_read(&eb->refs)); | |
1014 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1015 | #endif |
0b32f4bb JB |
1016 | return __set_page_dirty_nobuffers(page); |
1017 | } | |
1018 | ||
7f09410b | 1019 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1020 | .readpage = btree_readpage, |
0da5468f | 1021 | .writepages = btree_writepages, |
5f39d397 CM |
1022 | .releasepage = btree_releasepage, |
1023 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1024 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1025 | .migratepage = btree_migratepage, |
5a92bc88 | 1026 | #endif |
0b32f4bb | 1027 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1028 | }; |
1029 | ||
2ff7e61e | 1030 | void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) |
090d1875 | 1031 | { |
5f39d397 | 1032 | struct extent_buffer *buf = NULL; |
2ff7e61e | 1033 | struct inode *btree_inode = fs_info->btree_inode; |
090d1875 | 1034 | |
2ff7e61e | 1035 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1036 | if (IS_ERR(buf)) |
6197d86e | 1037 | return; |
d1310b2e | 1038 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
6af49dbd | 1039 | buf, WAIT_NONE, 0); |
5f39d397 | 1040 | free_extent_buffer(buf); |
090d1875 CM |
1041 | } |
1042 | ||
2ff7e61e | 1043 | int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, |
ab0fff03 AJ |
1044 | int mirror_num, struct extent_buffer **eb) |
1045 | { | |
1046 | struct extent_buffer *buf = NULL; | |
2ff7e61e | 1047 | struct inode *btree_inode = fs_info->btree_inode; |
ab0fff03 AJ |
1048 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; |
1049 | int ret; | |
1050 | ||
2ff7e61e | 1051 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1052 | if (IS_ERR(buf)) |
ab0fff03 AJ |
1053 | return 0; |
1054 | ||
1055 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1056 | ||
8436ea91 | 1057 | ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, |
6af49dbd | 1058 | mirror_num); |
ab0fff03 AJ |
1059 | if (ret) { |
1060 | free_extent_buffer(buf); | |
1061 | return ret; | |
1062 | } | |
1063 | ||
1064 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1065 | free_extent_buffer(buf); | |
1066 | return -EIO; | |
0b32f4bb | 1067 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1068 | *eb = buf; |
1069 | } else { | |
1070 | free_extent_buffer(buf); | |
1071 | } | |
1072 | return 0; | |
1073 | } | |
1074 | ||
2ff7e61e JM |
1075 | struct extent_buffer *btrfs_find_create_tree_block( |
1076 | struct btrfs_fs_info *fs_info, | |
1077 | u64 bytenr) | |
0999df54 | 1078 | { |
0b246afa JM |
1079 | if (btrfs_is_testing(fs_info)) |
1080 | return alloc_test_extent_buffer(fs_info, bytenr); | |
1081 | return alloc_extent_buffer(fs_info, bytenr); | |
0999df54 CM |
1082 | } |
1083 | ||
1084 | ||
e02119d5 CM |
1085 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1086 | { | |
727011e0 | 1087 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1088 | buf->start + buf->len - 1); |
e02119d5 CM |
1089 | } |
1090 | ||
3189ff77 | 1091 | void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) |
e02119d5 | 1092 | { |
3189ff77 JL |
1093 | filemap_fdatawait_range(buf->pages[0]->mapping, |
1094 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
1095 | } |
1096 | ||
581c1760 QW |
1097 | /* |
1098 | * Read tree block at logical address @bytenr and do variant basic but critical | |
1099 | * verification. | |
1100 | * | |
1101 | * @parent_transid: expected transid of this tree block, skip check if 0 | |
1102 | * @level: expected level, mandatory check | |
1103 | * @first_key: expected key in slot 0, skip check if NULL | |
1104 | */ | |
2ff7e61e | 1105 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
581c1760 QW |
1106 | u64 parent_transid, int level, |
1107 | struct btrfs_key *first_key) | |
0999df54 CM |
1108 | { |
1109 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1110 | int ret; |
1111 | ||
2ff7e61e | 1112 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 LB |
1113 | if (IS_ERR(buf)) |
1114 | return buf; | |
0999df54 | 1115 | |
581c1760 QW |
1116 | ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid, |
1117 | level, first_key); | |
0f0fe8f7 FDBM |
1118 | if (ret) { |
1119 | free_extent_buffer(buf); | |
64c043de | 1120 | return ERR_PTR(ret); |
0f0fe8f7 | 1121 | } |
5f39d397 | 1122 | return buf; |
ce9adaa5 | 1123 | |
eb60ceac CM |
1124 | } |
1125 | ||
7c302b49 | 1126 | void clean_tree_block(struct btrfs_fs_info *fs_info, |
d5c13f92 | 1127 | struct extent_buffer *buf) |
ed2ff2cb | 1128 | { |
55c69072 | 1129 | if (btrfs_header_generation(buf) == |
e2d84521 | 1130 | fs_info->running_transaction->transid) { |
b9447ef8 | 1131 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1132 | |
b9473439 | 1133 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1134 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1135 | -buf->len, | |
1136 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1137 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1138 | btrfs_set_lock_blocking(buf); | |
1139 | clear_extent_buffer_dirty(buf); | |
1140 | } | |
925baedd | 1141 | } |
5f39d397 CM |
1142 | } |
1143 | ||
8257b2dc MX |
1144 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1145 | { | |
1146 | struct btrfs_subvolume_writers *writers; | |
1147 | int ret; | |
1148 | ||
1149 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1150 | if (!writers) | |
1151 | return ERR_PTR(-ENOMEM); | |
1152 | ||
8a5a916d | 1153 | ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS); |
8257b2dc MX |
1154 | if (ret < 0) { |
1155 | kfree(writers); | |
1156 | return ERR_PTR(ret); | |
1157 | } | |
1158 | ||
1159 | init_waitqueue_head(&writers->wait); | |
1160 | return writers; | |
1161 | } | |
1162 | ||
1163 | static void | |
1164 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1165 | { | |
1166 | percpu_counter_destroy(&writers->counter); | |
1167 | kfree(writers); | |
1168 | } | |
1169 | ||
da17066c | 1170 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1171 | u64 objectid) |
d97e63b6 | 1172 | { |
7c0260ee | 1173 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
cfaa7295 | 1174 | root->node = NULL; |
a28ec197 | 1175 | root->commit_root = NULL; |
27cdeb70 | 1176 | root->state = 0; |
d68fc57b | 1177 | root->orphan_cleanup_state = 0; |
0b86a832 | 1178 | |
0f7d52f4 CM |
1179 | root->objectid = objectid; |
1180 | root->last_trans = 0; | |
13a8a7c8 | 1181 | root->highest_objectid = 0; |
eb73c1b7 | 1182 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1183 | root->nr_ordered_extents = 0; |
58176a96 | 1184 | root->name = NULL; |
6bef4d31 | 1185 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1186 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1187 | root->block_rsv = NULL; |
0b86a832 CM |
1188 | |
1189 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1190 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1191 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1192 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1193 | INIT_LIST_HEAD(&root->ordered_extents); |
1194 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1195 | INIT_LIST_HEAD(&root->logged_list[0]); |
1196 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1197 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1198 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1199 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1200 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1201 | spin_lock_init(&root->log_extents_lock[0]); |
1202 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1203 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1204 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1205 | mutex_init(&root->log_mutex); |
31f3d255 | 1206 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1207 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1208 | init_waitqueue_head(&root->log_writer_wait); |
1209 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1210 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1211 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1212 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1213 | atomic_set(&root->log_commit[0], 0); |
1214 | atomic_set(&root->log_commit[1], 0); | |
1215 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1216 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1217 | refcount_set(&root->refs, 1); |
ea14b57f | 1218 | atomic_set(&root->will_be_snapshotted, 0); |
7237f183 | 1219 | root->log_transid = 0; |
d1433deb | 1220 | root->log_transid_committed = -1; |
257c62e1 | 1221 | root->last_log_commit = 0; |
7c0260ee | 1222 | if (!dummy) |
c6100a4b | 1223 | extent_io_tree_init(&root->dirty_log_pages, NULL); |
017e5369 | 1224 | |
3768f368 CM |
1225 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1226 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1227 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1228 | if (!dummy) |
06ea65a3 JB |
1229 | root->defrag_trans_start = fs_info->generation; |
1230 | else | |
1231 | root->defrag_trans_start = 0; | |
4d775673 | 1232 | root->root_key.objectid = objectid; |
0ee5dc67 | 1233 | root->anon_dev = 0; |
8ea05e3a | 1234 | |
5f3ab90a | 1235 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1236 | } |
1237 | ||
74e4d827 DS |
1238 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1239 | gfp_t flags) | |
6f07e42e | 1240 | { |
74e4d827 | 1241 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1242 | if (root) |
1243 | root->fs_info = fs_info; | |
1244 | return root; | |
1245 | } | |
1246 | ||
06ea65a3 JB |
1247 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1248 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1249 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1250 | { |
1251 | struct btrfs_root *root; | |
1252 | ||
7c0260ee JM |
1253 | if (!fs_info) |
1254 | return ERR_PTR(-EINVAL); | |
1255 | ||
1256 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
06ea65a3 JB |
1257 | if (!root) |
1258 | return ERR_PTR(-ENOMEM); | |
da17066c | 1259 | |
b9ef22de | 1260 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
da17066c | 1261 | __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
faa2dbf0 | 1262 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1263 | |
1264 | return root; | |
1265 | } | |
1266 | #endif | |
1267 | ||
20897f5c AJ |
1268 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1269 | struct btrfs_fs_info *fs_info, | |
1270 | u64 objectid) | |
1271 | { | |
1272 | struct extent_buffer *leaf; | |
1273 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1274 | struct btrfs_root *root; | |
1275 | struct btrfs_key key; | |
1276 | int ret = 0; | |
33d85fda | 1277 | uuid_le uuid = NULL_UUID_LE; |
20897f5c | 1278 | |
74e4d827 | 1279 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1280 | if (!root) |
1281 | return ERR_PTR(-ENOMEM); | |
1282 | ||
da17066c | 1283 | __setup_root(root, fs_info, objectid); |
20897f5c AJ |
1284 | root->root_key.objectid = objectid; |
1285 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1286 | root->root_key.offset = 0; | |
1287 | ||
4d75f8a9 | 1288 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1289 | if (IS_ERR(leaf)) { |
1290 | ret = PTR_ERR(leaf); | |
1dd05682 | 1291 | leaf = NULL; |
20897f5c AJ |
1292 | goto fail; |
1293 | } | |
1294 | ||
b159fa28 | 1295 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
20897f5c AJ |
1296 | btrfs_set_header_bytenr(leaf, leaf->start); |
1297 | btrfs_set_header_generation(leaf, trans->transid); | |
1298 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1299 | btrfs_set_header_owner(leaf, objectid); | |
1300 | root->node = leaf; | |
1301 | ||
d24ee97b DS |
1302 | write_extent_buffer_fsid(leaf, fs_info->fsid); |
1303 | write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); | |
20897f5c AJ |
1304 | btrfs_mark_buffer_dirty(leaf); |
1305 | ||
1306 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1307 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1308 | |
1309 | root->root_item.flags = 0; | |
1310 | root->root_item.byte_limit = 0; | |
1311 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1312 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1313 | btrfs_set_root_level(&root->root_item, 0); | |
1314 | btrfs_set_root_refs(&root->root_item, 1); | |
1315 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1316 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1317 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda QW |
1318 | if (is_fstree(objectid)) |
1319 | uuid_le_gen(&uuid); | |
6463fe58 | 1320 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); |
20897f5c AJ |
1321 | root->root_item.drop_level = 0; |
1322 | ||
1323 | key.objectid = objectid; | |
1324 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1325 | key.offset = 0; | |
1326 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1327 | if (ret) | |
1328 | goto fail; | |
1329 | ||
1330 | btrfs_tree_unlock(leaf); | |
1331 | ||
1dd05682 TI |
1332 | return root; |
1333 | ||
20897f5c | 1334 | fail: |
1dd05682 TI |
1335 | if (leaf) { |
1336 | btrfs_tree_unlock(leaf); | |
59885b39 | 1337 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1338 | free_extent_buffer(leaf); |
1339 | } | |
1340 | kfree(root); | |
20897f5c | 1341 | |
1dd05682 | 1342 | return ERR_PTR(ret); |
20897f5c AJ |
1343 | } |
1344 | ||
7237f183 YZ |
1345 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1346 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1347 | { |
1348 | struct btrfs_root *root; | |
7237f183 | 1349 | struct extent_buffer *leaf; |
e02119d5 | 1350 | |
74e4d827 | 1351 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1352 | if (!root) |
7237f183 | 1353 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1354 | |
da17066c | 1355 | __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
e02119d5 CM |
1356 | |
1357 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1358 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1359 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1360 | |
7237f183 | 1361 | /* |
27cdeb70 MX |
1362 | * DON'T set REF_COWS for log trees |
1363 | * | |
7237f183 YZ |
1364 | * log trees do not get reference counted because they go away |
1365 | * before a real commit is actually done. They do store pointers | |
1366 | * to file data extents, and those reference counts still get | |
1367 | * updated (along with back refs to the log tree). | |
1368 | */ | |
e02119d5 | 1369 | |
4d75f8a9 DS |
1370 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1371 | NULL, 0, 0, 0); | |
7237f183 YZ |
1372 | if (IS_ERR(leaf)) { |
1373 | kfree(root); | |
1374 | return ERR_CAST(leaf); | |
1375 | } | |
e02119d5 | 1376 | |
b159fa28 | 1377 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
5d4f98a2 YZ |
1378 | btrfs_set_header_bytenr(leaf, leaf->start); |
1379 | btrfs_set_header_generation(leaf, trans->transid); | |
1380 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1381 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1382 | root->node = leaf; |
e02119d5 | 1383 | |
0b246afa | 1384 | write_extent_buffer_fsid(root->node, fs_info->fsid); |
e02119d5 CM |
1385 | btrfs_mark_buffer_dirty(root->node); |
1386 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1387 | return root; |
1388 | } | |
1389 | ||
1390 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1391 | struct btrfs_fs_info *fs_info) | |
1392 | { | |
1393 | struct btrfs_root *log_root; | |
1394 | ||
1395 | log_root = alloc_log_tree(trans, fs_info); | |
1396 | if (IS_ERR(log_root)) | |
1397 | return PTR_ERR(log_root); | |
1398 | WARN_ON(fs_info->log_root_tree); | |
1399 | fs_info->log_root_tree = log_root; | |
1400 | return 0; | |
1401 | } | |
1402 | ||
1403 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1404 | struct btrfs_root *root) | |
1405 | { | |
0b246afa | 1406 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1407 | struct btrfs_root *log_root; |
1408 | struct btrfs_inode_item *inode_item; | |
1409 | ||
0b246afa | 1410 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1411 | if (IS_ERR(log_root)) |
1412 | return PTR_ERR(log_root); | |
1413 | ||
1414 | log_root->last_trans = trans->transid; | |
1415 | log_root->root_key.offset = root->root_key.objectid; | |
1416 | ||
1417 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1418 | btrfs_set_stack_inode_generation(inode_item, 1); |
1419 | btrfs_set_stack_inode_size(inode_item, 3); | |
1420 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1421 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1422 | fs_info->nodesize); |
3cae210f | 1423 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1424 | |
5d4f98a2 | 1425 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1426 | |
1427 | WARN_ON(root->log_root); | |
1428 | root->log_root = log_root; | |
1429 | root->log_transid = 0; | |
d1433deb | 1430 | root->log_transid_committed = -1; |
257c62e1 | 1431 | root->last_log_commit = 0; |
e02119d5 CM |
1432 | return 0; |
1433 | } | |
1434 | ||
35a3621b SB |
1435 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1436 | struct btrfs_key *key) | |
e02119d5 CM |
1437 | { |
1438 | struct btrfs_root *root; | |
1439 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1440 | struct btrfs_path *path; |
84234f3a | 1441 | u64 generation; |
cb517eab | 1442 | int ret; |
581c1760 | 1443 | int level; |
0f7d52f4 | 1444 | |
cb517eab MX |
1445 | path = btrfs_alloc_path(); |
1446 | if (!path) | |
0f7d52f4 | 1447 | return ERR_PTR(-ENOMEM); |
cb517eab | 1448 | |
74e4d827 | 1449 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1450 | if (!root) { |
1451 | ret = -ENOMEM; | |
1452 | goto alloc_fail; | |
0f7d52f4 CM |
1453 | } |
1454 | ||
da17066c | 1455 | __setup_root(root, fs_info, key->objectid); |
0f7d52f4 | 1456 | |
cb517eab MX |
1457 | ret = btrfs_find_root(tree_root, key, path, |
1458 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1459 | if (ret) { |
13a8a7c8 YZ |
1460 | if (ret > 0) |
1461 | ret = -ENOENT; | |
cb517eab | 1462 | goto find_fail; |
0f7d52f4 | 1463 | } |
13a8a7c8 | 1464 | |
84234f3a | 1465 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1466 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1467 | root->node = read_tree_block(fs_info, |
1468 | btrfs_root_bytenr(&root->root_item), | |
581c1760 | 1469 | generation, level, NULL); |
64c043de LB |
1470 | if (IS_ERR(root->node)) { |
1471 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1472 | goto find_fail; |
1473 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1474 | ret = -EIO; | |
64c043de LB |
1475 | free_extent_buffer(root->node); |
1476 | goto find_fail; | |
416bc658 | 1477 | } |
5d4f98a2 | 1478 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1479 | out: |
cb517eab MX |
1480 | btrfs_free_path(path); |
1481 | return root; | |
1482 | ||
cb517eab MX |
1483 | find_fail: |
1484 | kfree(root); | |
1485 | alloc_fail: | |
1486 | root = ERR_PTR(ret); | |
1487 | goto out; | |
1488 | } | |
1489 | ||
1490 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1491 | struct btrfs_key *location) | |
1492 | { | |
1493 | struct btrfs_root *root; | |
1494 | ||
1495 | root = btrfs_read_tree_root(tree_root, location); | |
1496 | if (IS_ERR(root)) | |
1497 | return root; | |
1498 | ||
1499 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1500 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1501 | btrfs_check_and_init_root_item(&root->root_item); |
1502 | } | |
13a8a7c8 | 1503 | |
5eda7b5e CM |
1504 | return root; |
1505 | } | |
1506 | ||
cb517eab MX |
1507 | int btrfs_init_fs_root(struct btrfs_root *root) |
1508 | { | |
1509 | int ret; | |
8257b2dc | 1510 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1511 | |
1512 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1513 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1514 | GFP_NOFS); | |
1515 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1516 | ret = -ENOMEM; | |
1517 | goto fail; | |
1518 | } | |
1519 | ||
8257b2dc MX |
1520 | writers = btrfs_alloc_subvolume_writers(); |
1521 | if (IS_ERR(writers)) { | |
1522 | ret = PTR_ERR(writers); | |
1523 | goto fail; | |
1524 | } | |
1525 | root->subv_writers = writers; | |
1526 | ||
cb517eab | 1527 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1528 | spin_lock_init(&root->ino_cache_lock); |
1529 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1530 | |
1531 | ret = get_anon_bdev(&root->anon_dev); | |
1532 | if (ret) | |
876d2cf1 | 1533 | goto fail; |
f32e48e9 CR |
1534 | |
1535 | mutex_lock(&root->objectid_mutex); | |
1536 | ret = btrfs_find_highest_objectid(root, | |
1537 | &root->highest_objectid); | |
1538 | if (ret) { | |
1539 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1540 | goto fail; |
f32e48e9 CR |
1541 | } |
1542 | ||
1543 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1544 | ||
1545 | mutex_unlock(&root->objectid_mutex); | |
1546 | ||
cb517eab MX |
1547 | return 0; |
1548 | fail: | |
876d2cf1 | 1549 | /* the caller is responsible to call free_fs_root */ |
cb517eab MX |
1550 | return ret; |
1551 | } | |
1552 | ||
35bbb97f JM |
1553 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1554 | u64 root_id) | |
cb517eab MX |
1555 | { |
1556 | struct btrfs_root *root; | |
1557 | ||
1558 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1559 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1560 | (unsigned long)root_id); | |
1561 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1562 | return root; | |
1563 | } | |
1564 | ||
1565 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1566 | struct btrfs_root *root) | |
1567 | { | |
1568 | int ret; | |
1569 | ||
e1860a77 | 1570 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1571 | if (ret) |
1572 | return ret; | |
1573 | ||
1574 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1575 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1576 | (unsigned long)root->root_key.objectid, | |
1577 | root); | |
1578 | if (ret == 0) | |
27cdeb70 | 1579 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1580 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1581 | radix_tree_preload_end(); | |
1582 | ||
1583 | return ret; | |
1584 | } | |
1585 | ||
c00869f1 MX |
1586 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1587 | struct btrfs_key *location, | |
1588 | bool check_ref) | |
5eda7b5e CM |
1589 | { |
1590 | struct btrfs_root *root; | |
381cf658 | 1591 | struct btrfs_path *path; |
1d4c08e0 | 1592 | struct btrfs_key key; |
5eda7b5e CM |
1593 | int ret; |
1594 | ||
edbd8d4e CM |
1595 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1596 | return fs_info->tree_root; | |
1597 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1598 | return fs_info->extent_root; | |
8f18cf13 CM |
1599 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1600 | return fs_info->chunk_root; | |
1601 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1602 | return fs_info->dev_root; | |
0403e47e YZ |
1603 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1604 | return fs_info->csum_root; | |
bcef60f2 AJ |
1605 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1606 | return fs_info->quota_root ? fs_info->quota_root : | |
1607 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1608 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1609 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1610 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1611 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1612 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1613 | ERR_PTR(-ENOENT); | |
4df27c4d | 1614 | again: |
cb517eab | 1615 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1616 | if (root) { |
c00869f1 | 1617 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1618 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1619 | return root; |
48475471 | 1620 | } |
5eda7b5e | 1621 | |
cb517eab | 1622 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1623 | if (IS_ERR(root)) |
1624 | return root; | |
3394e160 | 1625 | |
c00869f1 | 1626 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1627 | ret = -ENOENT; |
581bb050 | 1628 | goto fail; |
35a30d7c | 1629 | } |
581bb050 | 1630 | |
cb517eab | 1631 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1632 | if (ret) |
1633 | goto fail; | |
3394e160 | 1634 | |
381cf658 DS |
1635 | path = btrfs_alloc_path(); |
1636 | if (!path) { | |
1637 | ret = -ENOMEM; | |
1638 | goto fail; | |
1639 | } | |
1d4c08e0 DS |
1640 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1641 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1642 | key.offset = location->objectid; | |
1643 | ||
1644 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1645 | btrfs_free_path(path); |
d68fc57b YZ |
1646 | if (ret < 0) |
1647 | goto fail; | |
1648 | if (ret == 0) | |
27cdeb70 | 1649 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1650 | |
cb517eab | 1651 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1652 | if (ret) { |
4df27c4d YZ |
1653 | if (ret == -EEXIST) { |
1654 | free_fs_root(root); | |
1655 | goto again; | |
1656 | } | |
1657 | goto fail; | |
0f7d52f4 | 1658 | } |
edbd8d4e | 1659 | return root; |
4df27c4d YZ |
1660 | fail: |
1661 | free_fs_root(root); | |
1662 | return ERR_PTR(ret); | |
edbd8d4e CM |
1663 | } |
1664 | ||
04160088 CM |
1665 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1666 | { | |
1667 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1668 | int ret = 0; | |
04160088 CM |
1669 | struct btrfs_device *device; |
1670 | struct backing_dev_info *bdi; | |
b7967db7 | 1671 | |
1f78160c XG |
1672 | rcu_read_lock(); |
1673 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1674 | if (!device->bdev) |
1675 | continue; | |
efa7c9f9 | 1676 | bdi = device->bdev->bd_bdi; |
ff9ea323 | 1677 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1678 | ret = 1; |
1679 | break; | |
1680 | } | |
1681 | } | |
1f78160c | 1682 | rcu_read_unlock(); |
04160088 CM |
1683 | return ret; |
1684 | } | |
1685 | ||
8b712842 CM |
1686 | /* |
1687 | * called by the kthread helper functions to finally call the bio end_io | |
1688 | * functions. This is where read checksum verification actually happens | |
1689 | */ | |
1690 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1691 | { |
ce9adaa5 | 1692 | struct bio *bio; |
97eb6b69 | 1693 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1694 | |
97eb6b69 | 1695 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1696 | bio = end_io_wq->bio; |
ce9adaa5 | 1697 | |
4e4cbee9 | 1698 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1699 | bio->bi_private = end_io_wq->private; |
1700 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1701 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1702 | bio_endio(bio); |
44b8bd7e CM |
1703 | } |
1704 | ||
a74a4b97 CM |
1705 | static int cleaner_kthread(void *arg) |
1706 | { | |
1707 | struct btrfs_root *root = arg; | |
0b246afa | 1708 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1709 | int again; |
da288d28 | 1710 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1711 | |
1712 | do { | |
d0278245 | 1713 | again = 0; |
a74a4b97 | 1714 | |
d0278245 | 1715 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1716 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1717 | goto sleep; |
1718 | ||
90c711ab ZB |
1719 | /* |
1720 | * Do not do anything if we might cause open_ctree() to block | |
1721 | * before we have finished mounting the filesystem. | |
1722 | */ | |
0b246afa | 1723 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1724 | goto sleep; |
1725 | ||
0b246afa | 1726 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1727 | goto sleep; |
1728 | ||
dc7f370c MX |
1729 | /* |
1730 | * Avoid the problem that we change the status of the fs | |
1731 | * during the above check and trylock. | |
1732 | */ | |
2ff7e61e | 1733 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1734 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1735 | goto sleep; |
76dda93c | 1736 | } |
a74a4b97 | 1737 | |
0b246afa | 1738 | mutex_lock(&fs_info->cleaner_delayed_iput_mutex); |
2ff7e61e | 1739 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 1740 | mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); |
c2d6cb16 | 1741 | |
d0278245 | 1742 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1743 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1744 | |
1745 | /* | |
05323cd1 MX |
1746 | * The defragger has dealt with the R/O remount and umount, |
1747 | * needn't do anything special here. | |
d0278245 | 1748 | */ |
0b246afa | 1749 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1750 | |
1751 | /* | |
1752 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1753 | * with relocation (btrfs_relocate_chunk) and relocation | |
1754 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1755 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1756 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1757 | * unused block groups. | |
1758 | */ | |
0b246afa | 1759 | btrfs_delete_unused_bgs(fs_info); |
d0278245 | 1760 | sleep: |
838fe188 | 1761 | if (!again) { |
a74a4b97 | 1762 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1763 | if (!kthread_should_stop()) |
1764 | schedule(); | |
a74a4b97 CM |
1765 | __set_current_state(TASK_RUNNING); |
1766 | } | |
1767 | } while (!kthread_should_stop()); | |
da288d28 FM |
1768 | |
1769 | /* | |
1770 | * Transaction kthread is stopped before us and wakes us up. | |
1771 | * However we might have started a new transaction and COWed some | |
1772 | * tree blocks when deleting unused block groups for example. So | |
1773 | * make sure we commit the transaction we started to have a clean | |
1774 | * shutdown when evicting the btree inode - if it has dirty pages | |
1775 | * when we do the final iput() on it, eviction will trigger a | |
1776 | * writeback for it which will fail with null pointer dereferences | |
1777 | * since work queues and other resources were already released and | |
1778 | * destroyed by the time the iput/eviction/writeback is made. | |
1779 | */ | |
1780 | trans = btrfs_attach_transaction(root); | |
1781 | if (IS_ERR(trans)) { | |
1782 | if (PTR_ERR(trans) != -ENOENT) | |
0b246afa | 1783 | btrfs_err(fs_info, |
da288d28 FM |
1784 | "cleaner transaction attach returned %ld", |
1785 | PTR_ERR(trans)); | |
1786 | } else { | |
1787 | int ret; | |
1788 | ||
3a45bb20 | 1789 | ret = btrfs_commit_transaction(trans); |
da288d28 | 1790 | if (ret) |
0b246afa | 1791 | btrfs_err(fs_info, |
da288d28 FM |
1792 | "cleaner open transaction commit returned %d", |
1793 | ret); | |
1794 | } | |
1795 | ||
a74a4b97 CM |
1796 | return 0; |
1797 | } | |
1798 | ||
1799 | static int transaction_kthread(void *arg) | |
1800 | { | |
1801 | struct btrfs_root *root = arg; | |
0b246afa | 1802 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1803 | struct btrfs_trans_handle *trans; |
1804 | struct btrfs_transaction *cur; | |
8929ecfa | 1805 | u64 transid; |
a74a4b97 CM |
1806 | unsigned long now; |
1807 | unsigned long delay; | |
914b2007 | 1808 | bool cannot_commit; |
a74a4b97 CM |
1809 | |
1810 | do { | |
914b2007 | 1811 | cannot_commit = false; |
0b246afa JM |
1812 | delay = HZ * fs_info->commit_interval; |
1813 | mutex_lock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1814 | |
0b246afa JM |
1815 | spin_lock(&fs_info->trans_lock); |
1816 | cur = fs_info->running_transaction; | |
a74a4b97 | 1817 | if (!cur) { |
0b246afa | 1818 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1819 | goto sleep; |
1820 | } | |
31153d81 | 1821 | |
a74a4b97 | 1822 | now = get_seconds(); |
4a9d8bde | 1823 | if (cur->state < TRANS_STATE_BLOCKED && |
a514d638 | 1824 | !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) && |
8b87dc17 | 1825 | (now < cur->start_time || |
0b246afa JM |
1826 | now - cur->start_time < fs_info->commit_interval)) { |
1827 | spin_unlock(&fs_info->trans_lock); | |
a74a4b97 CM |
1828 | delay = HZ * 5; |
1829 | goto sleep; | |
1830 | } | |
8929ecfa | 1831 | transid = cur->transid; |
0b246afa | 1832 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1833 | |
79787eaa | 1834 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1835 | trans = btrfs_attach_transaction(root); |
914b2007 | 1836 | if (IS_ERR(trans)) { |
354aa0fb MX |
1837 | if (PTR_ERR(trans) != -ENOENT) |
1838 | cannot_commit = true; | |
79787eaa | 1839 | goto sleep; |
914b2007 | 1840 | } |
8929ecfa | 1841 | if (transid == trans->transid) { |
3a45bb20 | 1842 | btrfs_commit_transaction(trans); |
8929ecfa | 1843 | } else { |
3a45bb20 | 1844 | btrfs_end_transaction(trans); |
8929ecfa | 1845 | } |
a74a4b97 | 1846 | sleep: |
0b246afa JM |
1847 | wake_up_process(fs_info->cleaner_kthread); |
1848 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1849 | |
4e121c06 | 1850 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1851 | &fs_info->fs_state))) |
2ff7e61e | 1852 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1853 | if (!kthread_should_stop() && |
0b246afa | 1854 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1855 | cannot_commit)) |
bc5511d0 | 1856 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1857 | } while (!kthread_should_stop()); |
1858 | return 0; | |
1859 | } | |
1860 | ||
af31f5e5 CM |
1861 | /* |
1862 | * this will find the highest generation in the array of | |
1863 | * root backups. The index of the highest array is returned, | |
1864 | * or -1 if we can't find anything. | |
1865 | * | |
1866 | * We check to make sure the array is valid by comparing the | |
1867 | * generation of the latest root in the array with the generation | |
1868 | * in the super block. If they don't match we pitch it. | |
1869 | */ | |
1870 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1871 | { | |
1872 | u64 cur; | |
1873 | int newest_index = -1; | |
1874 | struct btrfs_root_backup *root_backup; | |
1875 | int i; | |
1876 | ||
1877 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1878 | root_backup = info->super_copy->super_roots + i; | |
1879 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1880 | if (cur == newest_gen) | |
1881 | newest_index = i; | |
1882 | } | |
1883 | ||
1884 | /* check to see if we actually wrapped around */ | |
1885 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1886 | root_backup = info->super_copy->super_roots; | |
1887 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1888 | if (cur == newest_gen) | |
1889 | newest_index = 0; | |
1890 | } | |
1891 | return newest_index; | |
1892 | } | |
1893 | ||
1894 | ||
1895 | /* | |
1896 | * find the oldest backup so we know where to store new entries | |
1897 | * in the backup array. This will set the backup_root_index | |
1898 | * field in the fs_info struct | |
1899 | */ | |
1900 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1901 | u64 newest_gen) | |
1902 | { | |
1903 | int newest_index = -1; | |
1904 | ||
1905 | newest_index = find_newest_super_backup(info, newest_gen); | |
1906 | /* if there was garbage in there, just move along */ | |
1907 | if (newest_index == -1) { | |
1908 | info->backup_root_index = 0; | |
1909 | } else { | |
1910 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1911 | } | |
1912 | } | |
1913 | ||
1914 | /* | |
1915 | * copy all the root pointers into the super backup array. | |
1916 | * this will bump the backup pointer by one when it is | |
1917 | * done | |
1918 | */ | |
1919 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1920 | { | |
1921 | int next_backup; | |
1922 | struct btrfs_root_backup *root_backup; | |
1923 | int last_backup; | |
1924 | ||
1925 | next_backup = info->backup_root_index; | |
1926 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1927 | BTRFS_NUM_BACKUP_ROOTS; | |
1928 | ||
1929 | /* | |
1930 | * just overwrite the last backup if we're at the same generation | |
1931 | * this happens only at umount | |
1932 | */ | |
1933 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1934 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1935 | btrfs_header_generation(info->tree_root->node)) | |
1936 | next_backup = last_backup; | |
1937 | ||
1938 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1939 | ||
1940 | /* | |
1941 | * make sure all of our padding and empty slots get zero filled | |
1942 | * regardless of which ones we use today | |
1943 | */ | |
1944 | memset(root_backup, 0, sizeof(*root_backup)); | |
1945 | ||
1946 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1947 | ||
1948 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1949 | btrfs_set_backup_tree_root_gen(root_backup, | |
1950 | btrfs_header_generation(info->tree_root->node)); | |
1951 | ||
1952 | btrfs_set_backup_tree_root_level(root_backup, | |
1953 | btrfs_header_level(info->tree_root->node)); | |
1954 | ||
1955 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1956 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1957 | btrfs_header_generation(info->chunk_root->node)); | |
1958 | btrfs_set_backup_chunk_root_level(root_backup, | |
1959 | btrfs_header_level(info->chunk_root->node)); | |
1960 | ||
1961 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1962 | btrfs_set_backup_extent_root_gen(root_backup, | |
1963 | btrfs_header_generation(info->extent_root->node)); | |
1964 | btrfs_set_backup_extent_root_level(root_backup, | |
1965 | btrfs_header_level(info->extent_root->node)); | |
1966 | ||
7c7e82a7 CM |
1967 | /* |
1968 | * we might commit during log recovery, which happens before we set | |
1969 | * the fs_root. Make sure it is valid before we fill it in. | |
1970 | */ | |
1971 | if (info->fs_root && info->fs_root->node) { | |
1972 | btrfs_set_backup_fs_root(root_backup, | |
1973 | info->fs_root->node->start); | |
1974 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1975 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1976 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1977 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1978 | } |
af31f5e5 CM |
1979 | |
1980 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1981 | btrfs_set_backup_dev_root_gen(root_backup, | |
1982 | btrfs_header_generation(info->dev_root->node)); | |
1983 | btrfs_set_backup_dev_root_level(root_backup, | |
1984 | btrfs_header_level(info->dev_root->node)); | |
1985 | ||
1986 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1987 | btrfs_set_backup_csum_root_gen(root_backup, | |
1988 | btrfs_header_generation(info->csum_root->node)); | |
1989 | btrfs_set_backup_csum_root_level(root_backup, | |
1990 | btrfs_header_level(info->csum_root->node)); | |
1991 | ||
1992 | btrfs_set_backup_total_bytes(root_backup, | |
1993 | btrfs_super_total_bytes(info->super_copy)); | |
1994 | btrfs_set_backup_bytes_used(root_backup, | |
1995 | btrfs_super_bytes_used(info->super_copy)); | |
1996 | btrfs_set_backup_num_devices(root_backup, | |
1997 | btrfs_super_num_devices(info->super_copy)); | |
1998 | ||
1999 | /* | |
2000 | * if we don't copy this out to the super_copy, it won't get remembered | |
2001 | * for the next commit | |
2002 | */ | |
2003 | memcpy(&info->super_copy->super_roots, | |
2004 | &info->super_for_commit->super_roots, | |
2005 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2006 | } | |
2007 | ||
2008 | /* | |
2009 | * this copies info out of the root backup array and back into | |
2010 | * the in-memory super block. It is meant to help iterate through | |
2011 | * the array, so you send it the number of backups you've already | |
2012 | * tried and the last backup index you used. | |
2013 | * | |
2014 | * this returns -1 when it has tried all the backups | |
2015 | */ | |
2016 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2017 | struct btrfs_super_block *super, | |
2018 | int *num_backups_tried, int *backup_index) | |
2019 | { | |
2020 | struct btrfs_root_backup *root_backup; | |
2021 | int newest = *backup_index; | |
2022 | ||
2023 | if (*num_backups_tried == 0) { | |
2024 | u64 gen = btrfs_super_generation(super); | |
2025 | ||
2026 | newest = find_newest_super_backup(info, gen); | |
2027 | if (newest == -1) | |
2028 | return -1; | |
2029 | ||
2030 | *backup_index = newest; | |
2031 | *num_backups_tried = 1; | |
2032 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2033 | /* we've tried all the backups, all done */ | |
2034 | return -1; | |
2035 | } else { | |
2036 | /* jump to the next oldest backup */ | |
2037 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2038 | BTRFS_NUM_BACKUP_ROOTS; | |
2039 | *backup_index = newest; | |
2040 | *num_backups_tried += 1; | |
2041 | } | |
2042 | root_backup = super->super_roots + newest; | |
2043 | ||
2044 | btrfs_set_super_generation(super, | |
2045 | btrfs_backup_tree_root_gen(root_backup)); | |
2046 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2047 | btrfs_set_super_root_level(super, | |
2048 | btrfs_backup_tree_root_level(root_backup)); | |
2049 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2050 | ||
2051 | /* | |
2052 | * fixme: the total bytes and num_devices need to match or we should | |
2053 | * need a fsck | |
2054 | */ | |
2055 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2056 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2057 | return 0; | |
2058 | } | |
2059 | ||
7abadb64 LB |
2060 | /* helper to cleanup workers */ |
2061 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2062 | { | |
dc6e3209 | 2063 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2064 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2065 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2066 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2067 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
8b110e39 | 2068 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2069 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2070 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2071 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2072 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2073 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2074 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2075 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2076 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2077 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2078 | btrfs_destroy_workqueue(fs_info->extent_workers); |
a9b9477d FM |
2079 | /* |
2080 | * Now that all other work queues are destroyed, we can safely destroy | |
2081 | * the queues used for metadata I/O, since tasks from those other work | |
2082 | * queues can do metadata I/O operations. | |
2083 | */ | |
2084 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2085 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2086 | } |
2087 | ||
2e9f5954 R |
2088 | static void free_root_extent_buffers(struct btrfs_root *root) |
2089 | { | |
2090 | if (root) { | |
2091 | free_extent_buffer(root->node); | |
2092 | free_extent_buffer(root->commit_root); | |
2093 | root->node = NULL; | |
2094 | root->commit_root = NULL; | |
2095 | } | |
2096 | } | |
2097 | ||
af31f5e5 CM |
2098 | /* helper to cleanup tree roots */ |
2099 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2100 | { | |
2e9f5954 | 2101 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2102 | |
2e9f5954 R |
2103 | free_root_extent_buffers(info->dev_root); |
2104 | free_root_extent_buffers(info->extent_root); | |
2105 | free_root_extent_buffers(info->csum_root); | |
2106 | free_root_extent_buffers(info->quota_root); | |
2107 | free_root_extent_buffers(info->uuid_root); | |
2108 | if (chunk_root) | |
2109 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2110 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2111 | } |
2112 | ||
faa2dbf0 | 2113 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2114 | { |
2115 | int ret; | |
2116 | struct btrfs_root *gang[8]; | |
2117 | int i; | |
2118 | ||
2119 | while (!list_empty(&fs_info->dead_roots)) { | |
2120 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2121 | struct btrfs_root, root_list); | |
2122 | list_del(&gang[0]->root_list); | |
2123 | ||
27cdeb70 | 2124 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2125 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2126 | } else { |
2127 | free_extent_buffer(gang[0]->node); | |
2128 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2129 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2130 | } |
2131 | } | |
2132 | ||
2133 | while (1) { | |
2134 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2135 | (void **)gang, 0, | |
2136 | ARRAY_SIZE(gang)); | |
2137 | if (!ret) | |
2138 | break; | |
2139 | for (i = 0; i < ret; i++) | |
cb517eab | 2140 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2141 | } |
1a4319cc LB |
2142 | |
2143 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2144 | btrfs_free_log_root_tree(NULL, fs_info); | |
2ff7e61e | 2145 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
1a4319cc | 2146 | } |
171f6537 | 2147 | } |
af31f5e5 | 2148 | |
638aa7ed ES |
2149 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2150 | { | |
2151 | mutex_init(&fs_info->scrub_lock); | |
2152 | atomic_set(&fs_info->scrubs_running, 0); | |
2153 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2154 | atomic_set(&fs_info->scrubs_paused, 0); | |
2155 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2156 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2157 | fs_info->scrub_workers_refcnt = 0; | |
2158 | } | |
2159 | ||
779a65a4 ES |
2160 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2161 | { | |
2162 | spin_lock_init(&fs_info->balance_lock); | |
2163 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2164 | atomic_set(&fs_info->balance_pause_req, 0); |
2165 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2166 | fs_info->balance_ctl = NULL; | |
2167 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2168 | } | |
2169 | ||
6bccf3ab | 2170 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2171 | { |
2ff7e61e JM |
2172 | struct inode *inode = fs_info->btree_inode; |
2173 | ||
2174 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2175 | set_nlink(inode, 1); | |
f37938e0 ES |
2176 | /* |
2177 | * we set the i_size on the btree inode to the max possible int. | |
2178 | * the real end of the address space is determined by all of | |
2179 | * the devices in the system | |
2180 | */ | |
2ff7e61e JM |
2181 | inode->i_size = OFFSET_MAX; |
2182 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2183 | |
2ff7e61e | 2184 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
c6100a4b | 2185 | extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); |
2ff7e61e JM |
2186 | BTRFS_I(inode)->io_tree.track_uptodate = 0; |
2187 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); | |
f37938e0 | 2188 | |
2ff7e61e | 2189 | BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; |
f37938e0 | 2190 | |
2ff7e61e JM |
2191 | BTRFS_I(inode)->root = fs_info->tree_root; |
2192 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); | |
2193 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2194 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2195 | } |
2196 | ||
ad618368 ES |
2197 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2198 | { | |
2199 | fs_info->dev_replace.lock_owner = 0; | |
2200 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2201 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2202 | rwlock_init(&fs_info->dev_replace.lock); |
2203 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2204 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2205 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2206 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2207 | } |
2208 | ||
f9e92e40 ES |
2209 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2210 | { | |
2211 | spin_lock_init(&fs_info->qgroup_lock); | |
2212 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2213 | fs_info->qgroup_tree = RB_ROOT; | |
2214 | fs_info->qgroup_op_tree = RB_ROOT; | |
2215 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2216 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2217 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2218 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2219 | mutex_init(&fs_info->qgroup_rescan_lock); |
2220 | } | |
2221 | ||
2a458198 ES |
2222 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2223 | struct btrfs_fs_devices *fs_devices) | |
2224 | { | |
f7b885be | 2225 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2226 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2227 | |
2228 | fs_info->workers = | |
cb001095 JM |
2229 | btrfs_alloc_workqueue(fs_info, "worker", |
2230 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2231 | |
2232 | fs_info->delalloc_workers = | |
cb001095 JM |
2233 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2234 | flags, max_active, 2); | |
2a458198 ES |
2235 | |
2236 | fs_info->flush_workers = | |
cb001095 JM |
2237 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2238 | flags, max_active, 0); | |
2a458198 ES |
2239 | |
2240 | fs_info->caching_workers = | |
cb001095 | 2241 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 ES |
2242 | |
2243 | /* | |
2244 | * a higher idle thresh on the submit workers makes it much more | |
2245 | * likely that bios will be send down in a sane order to the | |
2246 | * devices | |
2247 | */ | |
2248 | fs_info->submit_workers = | |
cb001095 | 2249 | btrfs_alloc_workqueue(fs_info, "submit", flags, |
2a458198 ES |
2250 | min_t(u64, fs_devices->num_devices, |
2251 | max_active), 64); | |
2252 | ||
2253 | fs_info->fixup_workers = | |
cb001095 | 2254 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2255 | |
2256 | /* | |
2257 | * endios are largely parallel and should have a very | |
2258 | * low idle thresh | |
2259 | */ | |
2260 | fs_info->endio_workers = | |
cb001095 | 2261 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2262 | fs_info->endio_meta_workers = |
cb001095 JM |
2263 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2264 | max_active, 4); | |
2a458198 | 2265 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2266 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2267 | max_active, 2); | |
2a458198 | 2268 | fs_info->endio_raid56_workers = |
cb001095 JM |
2269 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2270 | max_active, 4); | |
2a458198 | 2271 | fs_info->endio_repair_workers = |
cb001095 | 2272 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2273 | fs_info->rmw_workers = |
cb001095 | 2274 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2275 | fs_info->endio_write_workers = |
cb001095 JM |
2276 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2277 | max_active, 2); | |
2a458198 | 2278 | fs_info->endio_freespace_worker = |
cb001095 JM |
2279 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2280 | max_active, 0); | |
2a458198 | 2281 | fs_info->delayed_workers = |
cb001095 JM |
2282 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2283 | max_active, 0); | |
2a458198 | 2284 | fs_info->readahead_workers = |
cb001095 JM |
2285 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2286 | max_active, 2); | |
2a458198 | 2287 | fs_info->qgroup_rescan_workers = |
cb001095 | 2288 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
2a458198 | 2289 | fs_info->extent_workers = |
cb001095 | 2290 | btrfs_alloc_workqueue(fs_info, "extent-refs", flags, |
2a458198 ES |
2291 | min_t(u64, fs_devices->num_devices, |
2292 | max_active), 8); | |
2293 | ||
2294 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2295 | fs_info->submit_workers && fs_info->flush_workers && | |
2296 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2297 | fs_info->endio_meta_write_workers && | |
2298 | fs_info->endio_repair_workers && | |
2299 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2300 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2301 | fs_info->caching_workers && fs_info->readahead_workers && | |
2302 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2303 | fs_info->extent_workers && | |
2304 | fs_info->qgroup_rescan_workers)) { | |
2305 | return -ENOMEM; | |
2306 | } | |
2307 | ||
2308 | return 0; | |
2309 | } | |
2310 | ||
63443bf5 ES |
2311 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2312 | struct btrfs_fs_devices *fs_devices) | |
2313 | { | |
2314 | int ret; | |
63443bf5 ES |
2315 | struct btrfs_root *log_tree_root; |
2316 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2317 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2318 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2319 | |
2320 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2321 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2322 | return -EIO; |
2323 | } | |
2324 | ||
74e4d827 | 2325 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2326 | if (!log_tree_root) |
2327 | return -ENOMEM; | |
2328 | ||
da17066c | 2329 | __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
63443bf5 | 2330 | |
2ff7e61e | 2331 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
581c1760 QW |
2332 | fs_info->generation + 1, |
2333 | level, NULL); | |
64c043de | 2334 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2335 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2336 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2337 | kfree(log_tree_root); |
0eeff236 | 2338 | return ret; |
64c043de | 2339 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2340 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2341 | free_extent_buffer(log_tree_root->node); |
2342 | kfree(log_tree_root); | |
2343 | return -EIO; | |
2344 | } | |
2345 | /* returns with log_tree_root freed on success */ | |
2346 | ret = btrfs_recover_log_trees(log_tree_root); | |
2347 | if (ret) { | |
0b246afa JM |
2348 | btrfs_handle_fs_error(fs_info, ret, |
2349 | "Failed to recover log tree"); | |
63443bf5 ES |
2350 | free_extent_buffer(log_tree_root->node); |
2351 | kfree(log_tree_root); | |
2352 | return ret; | |
2353 | } | |
2354 | ||
bc98a42c | 2355 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2356 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2357 | if (ret) |
2358 | return ret; | |
2359 | } | |
2360 | ||
2361 | return 0; | |
2362 | } | |
2363 | ||
6bccf3ab | 2364 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2365 | { |
6bccf3ab | 2366 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2367 | struct btrfs_root *root; |
4bbcaa64 ES |
2368 | struct btrfs_key location; |
2369 | int ret; | |
2370 | ||
6bccf3ab JM |
2371 | BUG_ON(!fs_info->tree_root); |
2372 | ||
4bbcaa64 ES |
2373 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2374 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2375 | location.offset = 0; | |
2376 | ||
a4f3d2c4 | 2377 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2378 | if (IS_ERR(root)) { |
2379 | ret = PTR_ERR(root); | |
2380 | goto out; | |
2381 | } | |
a4f3d2c4 DS |
2382 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2383 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2384 | |
2385 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2386 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2387 | if (IS_ERR(root)) { |
2388 | ret = PTR_ERR(root); | |
2389 | goto out; | |
2390 | } | |
a4f3d2c4 DS |
2391 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2392 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2393 | btrfs_init_devices_late(fs_info); |
2394 | ||
2395 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 | 2396 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 LB |
2397 | if (IS_ERR(root)) { |
2398 | ret = PTR_ERR(root); | |
2399 | goto out; | |
2400 | } | |
a4f3d2c4 DS |
2401 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2402 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2403 | |
2404 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2405 | root = btrfs_read_tree_root(tree_root, &location); |
2406 | if (!IS_ERR(root)) { | |
2407 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2408 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2409 | fs_info->quota_root = root; |
4bbcaa64 ES |
2410 | } |
2411 | ||
2412 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2413 | root = btrfs_read_tree_root(tree_root, &location); |
2414 | if (IS_ERR(root)) { | |
2415 | ret = PTR_ERR(root); | |
4bbcaa64 | 2416 | if (ret != -ENOENT) |
f50f4353 | 2417 | goto out; |
4bbcaa64 | 2418 | } else { |
a4f3d2c4 DS |
2419 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2420 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2421 | } |
2422 | ||
70f6d82e OS |
2423 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2424 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2425 | root = btrfs_read_tree_root(tree_root, &location); | |
f50f4353 LB |
2426 | if (IS_ERR(root)) { |
2427 | ret = PTR_ERR(root); | |
2428 | goto out; | |
2429 | } | |
70f6d82e OS |
2430 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2431 | fs_info->free_space_root = root; | |
2432 | } | |
2433 | ||
4bbcaa64 | 2434 | return 0; |
f50f4353 LB |
2435 | out: |
2436 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2437 | location.objectid, ret); | |
2438 | return ret; | |
4bbcaa64 ES |
2439 | } |
2440 | ||
069ec957 QW |
2441 | /* |
2442 | * Real super block validation | |
2443 | * NOTE: super csum type and incompat features will not be checked here. | |
2444 | * | |
2445 | * @sb: super block to check | |
2446 | * @mirror_num: the super block number to check its bytenr: | |
2447 | * 0 the primary (1st) sb | |
2448 | * 1, 2 2nd and 3rd backup copy | |
2449 | * -1 skip bytenr check | |
2450 | */ | |
2451 | static int validate_super(struct btrfs_fs_info *fs_info, | |
2452 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2453 | { |
21a852b0 QW |
2454 | u64 nodesize = btrfs_super_nodesize(sb); |
2455 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2456 | int ret = 0; | |
2457 | ||
2458 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2459 | btrfs_err(fs_info, "no valid FS found"); | |
2460 | ret = -EINVAL; | |
2461 | } | |
2462 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2463 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2464 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2465 | ret = -EINVAL; | |
2466 | } | |
2467 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2468 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2469 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2470 | ret = -EINVAL; | |
2471 | } | |
2472 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2473 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2474 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2475 | ret = -EINVAL; | |
2476 | } | |
2477 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2478 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2479 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2480 | ret = -EINVAL; | |
2481 | } | |
2482 | ||
2483 | /* | |
2484 | * Check sectorsize and nodesize first, other check will need it. | |
2485 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2486 | */ | |
2487 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2488 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2489 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2490 | ret = -EINVAL; | |
2491 | } | |
2492 | /* Only PAGE SIZE is supported yet */ | |
2493 | if (sectorsize != PAGE_SIZE) { | |
2494 | btrfs_err(fs_info, | |
2495 | "sectorsize %llu not supported yet, only support %lu", | |
2496 | sectorsize, PAGE_SIZE); | |
2497 | ret = -EINVAL; | |
2498 | } | |
2499 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
2500 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2501 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2502 | ret = -EINVAL; | |
2503 | } | |
2504 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2505 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2506 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2507 | ret = -EINVAL; | |
2508 | } | |
2509 | ||
2510 | /* Root alignment check */ | |
2511 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2512 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2513 | btrfs_super_root(sb)); | |
2514 | ret = -EINVAL; | |
2515 | } | |
2516 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2517 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2518 | btrfs_super_chunk_root(sb)); | |
2519 | ret = -EINVAL; | |
2520 | } | |
2521 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2522 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2523 | btrfs_super_log_root(sb)); | |
2524 | ret = -EINVAL; | |
2525 | } | |
2526 | ||
2527 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { | |
2528 | btrfs_err(fs_info, | |
2529 | "dev_item UUID does not match fsid: %pU != %pU", | |
2530 | fs_info->fsid, sb->dev_item.fsid); | |
2531 | ret = -EINVAL; | |
2532 | } | |
2533 | ||
2534 | /* | |
2535 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2536 | * done later | |
2537 | */ | |
2538 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2539 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2540 | btrfs_super_bytes_used(sb)); | |
2541 | ret = -EINVAL; | |
2542 | } | |
2543 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2544 | btrfs_err(fs_info, "invalid stripesize %u", | |
2545 | btrfs_super_stripesize(sb)); | |
2546 | ret = -EINVAL; | |
2547 | } | |
2548 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2549 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2550 | btrfs_super_num_devices(sb)); | |
2551 | if (btrfs_super_num_devices(sb) == 0) { | |
2552 | btrfs_err(fs_info, "number of devices is 0"); | |
2553 | ret = -EINVAL; | |
2554 | } | |
2555 | ||
069ec957 QW |
2556 | if (mirror_num >= 0 && |
2557 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2558 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2559 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2560 | ret = -EINVAL; | |
2561 | } | |
2562 | ||
2563 | /* | |
2564 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2565 | * and one chunk | |
2566 | */ | |
2567 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2568 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2569 | btrfs_super_sys_array_size(sb), | |
2570 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2571 | ret = -EINVAL; | |
2572 | } | |
2573 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2574 | + sizeof(struct btrfs_chunk)) { | |
2575 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2576 | btrfs_super_sys_array_size(sb), | |
2577 | sizeof(struct btrfs_disk_key) | |
2578 | + sizeof(struct btrfs_chunk)); | |
2579 | ret = -EINVAL; | |
2580 | } | |
2581 | ||
2582 | /* | |
2583 | * The generation is a global counter, we'll trust it more than the others | |
2584 | * but it's still possible that it's the one that's wrong. | |
2585 | */ | |
2586 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2587 | btrfs_warn(fs_info, | |
2588 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2589 | btrfs_super_generation(sb), | |
2590 | btrfs_super_chunk_root_generation(sb)); | |
2591 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2592 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2593 | btrfs_warn(fs_info, | |
2594 | "suspicious: generation < cache_generation: %llu < %llu", | |
2595 | btrfs_super_generation(sb), | |
2596 | btrfs_super_cache_generation(sb)); | |
2597 | ||
2598 | return ret; | |
2599 | } | |
2600 | ||
069ec957 QW |
2601 | /* |
2602 | * Validation of super block at mount time. | |
2603 | * Some checks already done early at mount time, like csum type and incompat | |
2604 | * flags will be skipped. | |
2605 | */ | |
2606 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2607 | { | |
2608 | return validate_super(fs_info, fs_info->super_copy, 0); | |
2609 | } | |
2610 | ||
75cb857d QW |
2611 | /* |
2612 | * Validation of super block at write time. | |
2613 | * Some checks like bytenr check will be skipped as their values will be | |
2614 | * overwritten soon. | |
2615 | * Extra checks like csum type and incompat flags will be done here. | |
2616 | */ | |
2617 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2618 | struct btrfs_super_block *sb) | |
2619 | { | |
2620 | int ret; | |
2621 | ||
2622 | ret = validate_super(fs_info, sb, -1); | |
2623 | if (ret < 0) | |
2624 | goto out; | |
2625 | if (btrfs_super_csum_type(sb) != BTRFS_CSUM_TYPE_CRC32) { | |
2626 | ret = -EUCLEAN; | |
2627 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2628 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2629 | goto out; | |
2630 | } | |
2631 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2632 | ret = -EUCLEAN; | |
2633 | btrfs_err(fs_info, | |
2634 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2635 | btrfs_super_incompat_flags(sb), | |
2636 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2637 | goto out; | |
2638 | } | |
2639 | out: | |
2640 | if (ret < 0) | |
2641 | btrfs_err(fs_info, | |
2642 | "super block corruption detected before writing it to disk"); | |
2643 | return ret; | |
2644 | } | |
2645 | ||
ad2b2c80 AV |
2646 | int open_ctree(struct super_block *sb, |
2647 | struct btrfs_fs_devices *fs_devices, | |
2648 | char *options) | |
2e635a27 | 2649 | { |
db94535d CM |
2650 | u32 sectorsize; |
2651 | u32 nodesize; | |
87ee04eb | 2652 | u32 stripesize; |
84234f3a | 2653 | u64 generation; |
f2b636e8 | 2654 | u64 features; |
3de4586c | 2655 | struct btrfs_key location; |
a061fc8d | 2656 | struct buffer_head *bh; |
4d34b278 | 2657 | struct btrfs_super_block *disk_super; |
815745cf | 2658 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2659 | struct btrfs_root *tree_root; |
4d34b278 | 2660 | struct btrfs_root *chunk_root; |
eb60ceac | 2661 | int ret; |
e58ca020 | 2662 | int err = -EINVAL; |
af31f5e5 CM |
2663 | int num_backups_tried = 0; |
2664 | int backup_index = 0; | |
6675df31 | 2665 | int clear_free_space_tree = 0; |
581c1760 | 2666 | int level; |
4543df7e | 2667 | |
74e4d827 DS |
2668 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2669 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2670 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2671 | err = -ENOMEM; |
2672 | goto fail; | |
2673 | } | |
76dda93c YZ |
2674 | |
2675 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2676 | if (ret) { | |
2677 | err = ret; | |
2678 | goto fail; | |
2679 | } | |
2680 | ||
908c7f19 | 2681 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2682 | if (ret) { |
2683 | err = ret; | |
9e11ceee | 2684 | goto fail_srcu; |
e2d84521 | 2685 | } |
09cbfeaf | 2686 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2687 | (1 + ilog2(nr_cpu_ids)); |
2688 | ||
908c7f19 | 2689 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2690 | if (ret) { |
2691 | err = ret; | |
2692 | goto fail_dirty_metadata_bytes; | |
2693 | } | |
2694 | ||
908c7f19 | 2695 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2696 | if (ret) { |
2697 | err = ret; | |
2698 | goto fail_delalloc_bytes; | |
2699 | } | |
2700 | ||
76dda93c | 2701 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2702 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2703 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2704 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2705 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2706 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2707 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
75cb379d JM |
2708 | INIT_LIST_HEAD(&fs_info->pending_raid_kobjs); |
2709 | spin_lock_init(&fs_info->pending_raid_kobjs_lock); | |
eb73c1b7 | 2710 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2711 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2712 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2713 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2714 | spin_lock_init(&fs_info->defrag_inodes_lock); |
f29021b2 | 2715 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2716 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2717 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2718 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2719 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2720 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2721 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2722 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2723 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2724 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2725 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2726 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2727 | |
0b86a832 | 2728 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2729 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2730 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2731 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2732 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2733 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2734 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2735 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2736 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2737 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2738 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2739 | BTRFS_BLOCK_RSV_DELOPS); | |
771ed689 | 2740 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2741 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2742 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2743 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2744 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2745 | fs_info->sb = sb; |
95ac567a | 2746 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2747 | fs_info->metadata_ratio = 0; |
4cb5300b | 2748 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2749 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2750 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2751 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2752 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2753 | /* readahead state */ |
d0164adc | 2754 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2755 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2756 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2757 | |
b34b086c CM |
2758 | fs_info->thread_pool_size = min_t(unsigned long, |
2759 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2760 | |
199c2a9c MX |
2761 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2762 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 JB |
2763 | |
2764 | fs_info->btree_inode = new_inode(sb); | |
2765 | if (!fs_info->btree_inode) { | |
2766 | err = -ENOMEM; | |
2767 | goto fail_bio_counter; | |
2768 | } | |
2769 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
2770 | ||
16cdcec7 | 2771 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2772 | GFP_KERNEL); |
16cdcec7 MX |
2773 | if (!fs_info->delayed_root) { |
2774 | err = -ENOMEM; | |
2775 | goto fail_iput; | |
2776 | } | |
2777 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2778 | |
638aa7ed | 2779 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2780 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2781 | fs_info->check_integrity_print_mask = 0; | |
2782 | #endif | |
779a65a4 | 2783 | btrfs_init_balance(fs_info); |
21c7e756 | 2784 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2785 | |
9f6d2510 DS |
2786 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2787 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
a061fc8d | 2788 | |
6bccf3ab | 2789 | btrfs_init_btree_inode(fs_info); |
76dda93c | 2790 | |
0f9dd46c | 2791 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2792 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2793 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2794 | |
c6100a4b JB |
2795 | extent_io_tree_init(&fs_info->freed_extents[0], NULL); |
2796 | extent_io_tree_init(&fs_info->freed_extents[1], NULL); | |
11833d66 | 2797 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
afcdd129 | 2798 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2799 | |
5a3f23d5 | 2800 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2801 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2802 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2803 | mutex_init(&fs_info->transaction_kthread_mutex); |
2804 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2805 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2806 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2807 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2808 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2809 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2810 | |
ad618368 | 2811 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2812 | btrfs_init_qgroup(fs_info); |
416ac51d | 2813 | |
fa9c0d79 CM |
2814 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2815 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2816 | ||
e6dcd2dc | 2817 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2818 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2819 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2820 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2821 | |
04216820 FM |
2822 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2823 | ||
da17066c JM |
2824 | /* Usable values until the real ones are cached from the superblock */ |
2825 | fs_info->nodesize = 4096; | |
2826 | fs_info->sectorsize = 4096; | |
2827 | fs_info->stripesize = 4096; | |
2828 | ||
53b381b3 DW |
2829 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2830 | if (ret) { | |
83c8266a | 2831 | err = ret; |
53b381b3 DW |
2832 | goto fail_alloc; |
2833 | } | |
2834 | ||
da17066c | 2835 | __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2836 | |
3c4bb26b | 2837 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2838 | |
2839 | /* | |
2840 | * Read super block and check the signature bytes only | |
2841 | */ | |
a512bbf8 | 2842 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2843 | if (IS_ERR(bh)) { |
2844 | err = PTR_ERR(bh); | |
16cdcec7 | 2845 | goto fail_alloc; |
20b45077 | 2846 | } |
39279cc3 | 2847 | |
1104a885 DS |
2848 | /* |
2849 | * We want to check superblock checksum, the type is stored inside. | |
2850 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2851 | */ | |
ab8d0fc4 | 2852 | if (btrfs_check_super_csum(fs_info, bh->b_data)) { |
05135f59 | 2853 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2854 | err = -EINVAL; |
b2acdddf | 2855 | brelse(bh); |
1104a885 DS |
2856 | goto fail_alloc; |
2857 | } | |
2858 | ||
2859 | /* | |
2860 | * super_copy is zeroed at allocation time and we never touch the | |
2861 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2862 | * the whole block of INFO_SIZE | |
2863 | */ | |
6c41761f DS |
2864 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2865 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2866 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2867 | brelse(bh); |
5f39d397 | 2868 | |
6c41761f | 2869 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2870 | |
069ec957 | 2871 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 2872 | if (ret) { |
05135f59 | 2873 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 DS |
2874 | err = -EINVAL; |
2875 | goto fail_alloc; | |
2876 | } | |
2877 | ||
6c41761f | 2878 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2879 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2880 | goto fail_alloc; |
0f7d52f4 | 2881 | |
acce952b | 2882 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2883 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2884 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2885 | |
af31f5e5 CM |
2886 | /* |
2887 | * run through our array of backup supers and setup | |
2888 | * our ring pointer to the oldest one | |
2889 | */ | |
2890 | generation = btrfs_super_generation(disk_super); | |
2891 | find_oldest_super_backup(fs_info, generation); | |
2892 | ||
75e7cb7f LB |
2893 | /* |
2894 | * In the long term, we'll store the compression type in the super | |
2895 | * block, and it'll be used for per file compression control. | |
2896 | */ | |
2897 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2898 | ||
2ff7e61e | 2899 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
2900 | if (ret) { |
2901 | err = ret; | |
16cdcec7 | 2902 | goto fail_alloc; |
2b82032c | 2903 | } |
dfe25020 | 2904 | |
f2b636e8 JB |
2905 | features = btrfs_super_incompat_flags(disk_super) & |
2906 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2907 | if (features) { | |
05135f59 DS |
2908 | btrfs_err(fs_info, |
2909 | "cannot mount because of unsupported optional features (%llx)", | |
2910 | features); | |
f2b636e8 | 2911 | err = -EINVAL; |
16cdcec7 | 2912 | goto fail_alloc; |
f2b636e8 JB |
2913 | } |
2914 | ||
5d4f98a2 | 2915 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2916 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 2917 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2918 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
2919 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
2920 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 2921 | |
3173a18f | 2922 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2923 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2924 | |
727011e0 CM |
2925 | /* |
2926 | * flag our filesystem as having big metadata blocks if | |
2927 | * they are bigger than the page size | |
2928 | */ | |
09cbfeaf | 2929 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2930 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2931 | btrfs_info(fs_info, |
2932 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2933 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2934 | } | |
2935 | ||
bc3f116f | 2936 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2937 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2938 | stripesize = sectorsize; |
707e8a07 | 2939 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2940 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f | 2941 | |
da17066c JM |
2942 | /* Cache block sizes */ |
2943 | fs_info->nodesize = nodesize; | |
2944 | fs_info->sectorsize = sectorsize; | |
2945 | fs_info->stripesize = stripesize; | |
2946 | ||
bc3f116f CM |
2947 | /* |
2948 | * mixed block groups end up with duplicate but slightly offset | |
2949 | * extent buffers for the same range. It leads to corruptions | |
2950 | */ | |
2951 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2952 | (sectorsize != nodesize)) { |
05135f59 DS |
2953 | btrfs_err(fs_info, |
2954 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2955 | nodesize, sectorsize); | |
bc3f116f CM |
2956 | goto fail_alloc; |
2957 | } | |
2958 | ||
ceda0864 MX |
2959 | /* |
2960 | * Needn't use the lock because there is no other task which will | |
2961 | * update the flag. | |
2962 | */ | |
a6fa6fae | 2963 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2964 | |
f2b636e8 JB |
2965 | features = btrfs_super_compat_ro_flags(disk_super) & |
2966 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 2967 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
2968 | btrfs_err(fs_info, |
2969 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2970 | features); |
f2b636e8 | 2971 | err = -EINVAL; |
16cdcec7 | 2972 | goto fail_alloc; |
f2b636e8 | 2973 | } |
61d92c32 | 2974 | |
2a458198 ES |
2975 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2976 | if (ret) { | |
2977 | err = ret; | |
0dc3b84a JB |
2978 | goto fail_sb_buffer; |
2979 | } | |
4543df7e | 2980 | |
9e11ceee JK |
2981 | sb->s_bdi->congested_fn = btrfs_congested_fn; |
2982 | sb->s_bdi->congested_data = fs_info; | |
2983 | sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; | |
d4417e22 | 2984 | sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; |
9e11ceee JK |
2985 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
2986 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 2987 | |
a061fc8d CM |
2988 | sb->s_blocksize = sectorsize; |
2989 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
ee87cf5e | 2990 | memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); |
db94535d | 2991 | |
925baedd | 2992 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 2993 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 2994 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2995 | if (ret) { |
05135f59 | 2996 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 2997 | goto fail_sb_buffer; |
84eed90f | 2998 | } |
0b86a832 | 2999 | |
84234f3a | 3000 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3001 | level = btrfs_super_chunk_root_level(disk_super); |
0b86a832 | 3002 | |
da17066c | 3003 | __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); |
0b86a832 | 3004 | |
2ff7e61e | 3005 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 3006 | btrfs_super_chunk_root(disk_super), |
581c1760 | 3007 | generation, level, NULL); |
64c043de LB |
3008 | if (IS_ERR(chunk_root->node) || |
3009 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 3010 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 3011 | if (!IS_ERR(chunk_root->node)) |
3012 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 3013 | chunk_root->node = NULL; |
af31f5e5 | 3014 | goto fail_tree_roots; |
83121942 | 3015 | } |
5d4f98a2 YZ |
3016 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
3017 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 3018 | |
e17cade2 | 3019 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 3020 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 3021 | |
5b4aacef | 3022 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3023 | if (ret) { |
05135f59 | 3024 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3025 | goto fail_tree_roots; |
2b82032c | 3026 | } |
0b86a832 | 3027 | |
8dabb742 | 3028 | /* |
9b99b115 AJ |
3029 | * Keep the devid that is marked to be the target device for the |
3030 | * device replace procedure | |
8dabb742 | 3031 | */ |
9b99b115 | 3032 | btrfs_free_extra_devids(fs_devices, 0); |
dfe25020 | 3033 | |
a6b0d5c8 | 3034 | if (!fs_devices->latest_bdev) { |
05135f59 | 3035 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3036 | goto fail_tree_roots; |
3037 | } | |
3038 | ||
af31f5e5 | 3039 | retry_root_backup: |
84234f3a | 3040 | generation = btrfs_super_generation(disk_super); |
581c1760 | 3041 | level = btrfs_super_root_level(disk_super); |
0b86a832 | 3042 | |
2ff7e61e | 3043 | tree_root->node = read_tree_block(fs_info, |
db94535d | 3044 | btrfs_super_root(disk_super), |
581c1760 | 3045 | generation, level, NULL); |
64c043de LB |
3046 | if (IS_ERR(tree_root->node) || |
3047 | !extent_buffer_uptodate(tree_root->node)) { | |
05135f59 | 3048 | btrfs_warn(fs_info, "failed to read tree root"); |
e5fffbac | 3049 | if (!IS_ERR(tree_root->node)) |
3050 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 3051 | tree_root->node = NULL; |
af31f5e5 | 3052 | goto recovery_tree_root; |
83121942 | 3053 | } |
af31f5e5 | 3054 | |
5d4f98a2 YZ |
3055 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
3056 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 3057 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 3058 | |
f32e48e9 CR |
3059 | mutex_lock(&tree_root->objectid_mutex); |
3060 | ret = btrfs_find_highest_objectid(tree_root, | |
3061 | &tree_root->highest_objectid); | |
3062 | if (ret) { | |
3063 | mutex_unlock(&tree_root->objectid_mutex); | |
3064 | goto recovery_tree_root; | |
3065 | } | |
3066 | ||
3067 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
3068 | ||
3069 | mutex_unlock(&tree_root->objectid_mutex); | |
3070 | ||
6bccf3ab | 3071 | ret = btrfs_read_roots(fs_info); |
4bbcaa64 | 3072 | if (ret) |
af31f5e5 | 3073 | goto recovery_tree_root; |
f7a81ea4 | 3074 | |
8929ecfa YZ |
3075 | fs_info->generation = generation; |
3076 | fs_info->last_trans_committed = generation; | |
8929ecfa | 3077 | |
68310a5e ID |
3078 | ret = btrfs_recover_balance(fs_info); |
3079 | if (ret) { | |
05135f59 | 3080 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3081 | goto fail_block_groups; |
3082 | } | |
3083 | ||
733f4fbb SB |
3084 | ret = btrfs_init_dev_stats(fs_info); |
3085 | if (ret) { | |
05135f59 | 3086 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3087 | goto fail_block_groups; |
3088 | } | |
3089 | ||
8dabb742 SB |
3090 | ret = btrfs_init_dev_replace(fs_info); |
3091 | if (ret) { | |
05135f59 | 3092 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3093 | goto fail_block_groups; |
3094 | } | |
3095 | ||
9b99b115 | 3096 | btrfs_free_extra_devids(fs_devices, 1); |
8dabb742 | 3097 | |
b7c35e81 AJ |
3098 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
3099 | if (ret) { | |
05135f59 DS |
3100 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3101 | ret); | |
b7c35e81 AJ |
3102 | goto fail_block_groups; |
3103 | } | |
3104 | ||
3105 | ret = btrfs_sysfs_add_device(fs_devices); | |
3106 | if (ret) { | |
05135f59 DS |
3107 | btrfs_err(fs_info, "failed to init sysfs device interface: %d", |
3108 | ret); | |
b7c35e81 AJ |
3109 | goto fail_fsdev_sysfs; |
3110 | } | |
3111 | ||
96f3136e | 3112 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3113 | if (ret) { |
05135f59 | 3114 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3115 | goto fail_fsdev_sysfs; |
c59021f8 | 3116 | } |
3117 | ||
c59021f8 | 3118 | ret = btrfs_init_space_info(fs_info); |
3119 | if (ret) { | |
05135f59 | 3120 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3121 | goto fail_sysfs; |
c59021f8 | 3122 | } |
3123 | ||
5b4aacef | 3124 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3125 | if (ret) { |
05135f59 | 3126 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3127 | goto fail_sysfs; |
1b1d1f66 | 3128 | } |
4330e183 | 3129 | |
6528b99d | 3130 | if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { |
05135f59 | 3131 | btrfs_warn(fs_info, |
4330e183 | 3132 | "writeable mount is not allowed due to too many missing devices"); |
2365dd3c | 3133 | goto fail_sysfs; |
292fd7fc | 3134 | } |
9078a3e1 | 3135 | |
a74a4b97 CM |
3136 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3137 | "btrfs-cleaner"); | |
57506d50 | 3138 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3139 | goto fail_sysfs; |
a74a4b97 CM |
3140 | |
3141 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3142 | tree_root, | |
3143 | "btrfs-transaction"); | |
57506d50 | 3144 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3145 | goto fail_cleaner; |
a74a4b97 | 3146 | |
583b7231 | 3147 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3148 | !fs_info->fs_devices->rotating) { |
583b7231 | 3149 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3150 | } |
3151 | ||
572d9ab7 | 3152 | /* |
01327610 | 3153 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3154 | * not have to wait for transaction commit |
3155 | */ | |
3156 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3157 | |
21adbd5c | 3158 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3159 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3160 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3161 | btrfs_test_opt(fs_info, |
21adbd5c SB |
3162 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3163 | 1 : 0, | |
3164 | fs_info->check_integrity_print_mask); | |
3165 | if (ret) | |
05135f59 DS |
3166 | btrfs_warn(fs_info, |
3167 | "failed to initialize integrity check module: %d", | |
3168 | ret); | |
21adbd5c SB |
3169 | } |
3170 | #endif | |
bcef60f2 AJ |
3171 | ret = btrfs_read_qgroup_config(fs_info); |
3172 | if (ret) | |
3173 | goto fail_trans_kthread; | |
21adbd5c | 3174 | |
fd708b81 JB |
3175 | if (btrfs_build_ref_tree(fs_info)) |
3176 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3177 | ||
96da0919 QW |
3178 | /* do not make disk changes in broken FS or nologreplay is given */ |
3179 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3180 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
63443bf5 | 3181 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3182 | if (ret) { |
63443bf5 | 3183 | err = ret; |
28c16cbb | 3184 | goto fail_qgroup; |
79787eaa | 3185 | } |
e02119d5 | 3186 | } |
1a40e23b | 3187 | |
6bccf3ab | 3188 | ret = btrfs_find_orphan_roots(fs_info); |
79787eaa | 3189 | if (ret) |
28c16cbb | 3190 | goto fail_qgroup; |
76dda93c | 3191 | |
bc98a42c | 3192 | if (!sb_rdonly(sb)) { |
d68fc57b | 3193 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3194 | if (ret) |
28c16cbb | 3195 | goto fail_qgroup; |
90c711ab ZB |
3196 | |
3197 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3198 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3199 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3200 | if (ret < 0) { |
05135f59 DS |
3201 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3202 | ret); | |
d7ce5843 | 3203 | err = -EINVAL; |
bcef60f2 | 3204 | goto fail_qgroup; |
d7ce5843 | 3205 | } |
7c2ca468 | 3206 | } |
1a40e23b | 3207 | |
3de4586c CM |
3208 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3209 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3210 | location.offset = 0; |
3de4586c | 3211 | |
3de4586c | 3212 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3213 | if (IS_ERR(fs_info->fs_root)) { |
3214 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3215 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
bcef60f2 | 3216 | goto fail_qgroup; |
3140c9a3 | 3217 | } |
c289811c | 3218 | |
bc98a42c | 3219 | if (sb_rdonly(sb)) |
2b6ba629 | 3220 | return 0; |
59641015 | 3221 | |
f8d468a1 OS |
3222 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && |
3223 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
6675df31 OS |
3224 | clear_free_space_tree = 1; |
3225 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3226 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3227 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3228 | clear_free_space_tree = 1; | |
3229 | } | |
3230 | ||
3231 | if (clear_free_space_tree) { | |
f8d468a1 OS |
3232 | btrfs_info(fs_info, "clearing free space tree"); |
3233 | ret = btrfs_clear_free_space_tree(fs_info); | |
3234 | if (ret) { | |
3235 | btrfs_warn(fs_info, | |
3236 | "failed to clear free space tree: %d", ret); | |
6bccf3ab | 3237 | close_ctree(fs_info); |
f8d468a1 OS |
3238 | return ret; |
3239 | } | |
3240 | } | |
3241 | ||
0b246afa | 3242 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
511711af | 3243 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3244 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3245 | ret = btrfs_create_free_space_tree(fs_info); |
3246 | if (ret) { | |
05135f59 DS |
3247 | btrfs_warn(fs_info, |
3248 | "failed to create free space tree: %d", ret); | |
6bccf3ab | 3249 | close_ctree(fs_info); |
511711af CM |
3250 | return ret; |
3251 | } | |
3252 | } | |
3253 | ||
2b6ba629 ID |
3254 | down_read(&fs_info->cleanup_work_sem); |
3255 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3256 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3257 | up_read(&fs_info->cleanup_work_sem); |
6bccf3ab | 3258 | close_ctree(fs_info); |
2b6ba629 ID |
3259 | return ret; |
3260 | } | |
3261 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3262 | |
2b6ba629 ID |
3263 | ret = btrfs_resume_balance_async(fs_info); |
3264 | if (ret) { | |
05135f59 | 3265 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
6bccf3ab | 3266 | close_ctree(fs_info); |
2b6ba629 | 3267 | return ret; |
e3acc2a6 JB |
3268 | } |
3269 | ||
8dabb742 SB |
3270 | ret = btrfs_resume_dev_replace_async(fs_info); |
3271 | if (ret) { | |
05135f59 | 3272 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
6bccf3ab | 3273 | close_ctree(fs_info); |
8dabb742 SB |
3274 | return ret; |
3275 | } | |
3276 | ||
b382a324 JS |
3277 | btrfs_qgroup_rescan_resume(fs_info); |
3278 | ||
4bbcaa64 | 3279 | if (!fs_info->uuid_root) { |
05135f59 | 3280 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3281 | ret = btrfs_create_uuid_tree(fs_info); |
3282 | if (ret) { | |
05135f59 DS |
3283 | btrfs_warn(fs_info, |
3284 | "failed to create the UUID tree: %d", ret); | |
6bccf3ab | 3285 | close_ctree(fs_info); |
f7a81ea4 SB |
3286 | return ret; |
3287 | } | |
0b246afa | 3288 | } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3289 | fs_info->generation != |
3290 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3291 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3292 | ret = btrfs_check_uuid_tree(fs_info); |
3293 | if (ret) { | |
05135f59 DS |
3294 | btrfs_warn(fs_info, |
3295 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3296 | close_ctree(fs_info); |
70f80175 SB |
3297 | return ret; |
3298 | } | |
3299 | } else { | |
afcdd129 | 3300 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); |
f7a81ea4 | 3301 | } |
afcdd129 | 3302 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3303 | |
8dcddfa0 QW |
3304 | /* |
3305 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3306 | * no need to keep the flag | |
3307 | */ | |
3308 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3309 | ||
ad2b2c80 | 3310 | return 0; |
39279cc3 | 3311 | |
bcef60f2 AJ |
3312 | fail_qgroup: |
3313 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3314 | fail_trans_kthread: |
3315 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3316 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3317 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3318 | fail_cleaner: |
a74a4b97 | 3319 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3320 | |
3321 | /* | |
3322 | * make sure we're done with the btree inode before we stop our | |
3323 | * kthreads | |
3324 | */ | |
3325 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3326 | |
2365dd3c | 3327 | fail_sysfs: |
6618a59b | 3328 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3329 | |
b7c35e81 AJ |
3330 | fail_fsdev_sysfs: |
3331 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3332 | ||
1b1d1f66 | 3333 | fail_block_groups: |
54067ae9 | 3334 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3335 | |
3336 | fail_tree_roots: | |
3337 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3338 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3339 | |
39279cc3 | 3340 | fail_sb_buffer: |
7abadb64 | 3341 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3342 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3343 | fail_alloc: |
4543df7e | 3344 | fail_iput: |
586e46e2 ID |
3345 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3346 | ||
4543df7e | 3347 | iput(fs_info->btree_inode); |
c404e0dc MX |
3348 | fail_bio_counter: |
3349 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3350 | fail_delalloc_bytes: |
3351 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3352 | fail_dirty_metadata_bytes: |
3353 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
76dda93c YZ |
3354 | fail_srcu: |
3355 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3356 | fail: |
53b381b3 | 3357 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3358 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3359 | return err; |
af31f5e5 CM |
3360 | |
3361 | recovery_tree_root: | |
0b246afa | 3362 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) |
af31f5e5 CM |
3363 | goto fail_tree_roots; |
3364 | ||
3365 | free_root_pointers(fs_info, 0); | |
3366 | ||
3367 | /* don't use the log in recovery mode, it won't be valid */ | |
3368 | btrfs_set_super_log_root(disk_super, 0); | |
3369 | ||
3370 | /* we can't trust the free space cache either */ | |
3371 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3372 | ||
3373 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3374 | &num_backups_tried, &backup_index); | |
3375 | if (ret == -1) | |
3376 | goto fail_block_groups; | |
3377 | goto retry_root_backup; | |
eb60ceac | 3378 | } |
663faf9f | 3379 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3380 | |
f2984462 CM |
3381 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3382 | { | |
f2984462 CM |
3383 | if (uptodate) { |
3384 | set_buffer_uptodate(bh); | |
3385 | } else { | |
442a4f63 SB |
3386 | struct btrfs_device *device = (struct btrfs_device *) |
3387 | bh->b_private; | |
3388 | ||
fb456252 | 3389 | btrfs_warn_rl_in_rcu(device->fs_info, |
b14af3b4 | 3390 | "lost page write due to IO error on %s", |
606686ee | 3391 | rcu_str_deref(device->name)); |
01327610 | 3392 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3393 | * our own ways of dealing with the IO errors |
3394 | */ | |
f2984462 | 3395 | clear_buffer_uptodate(bh); |
442a4f63 | 3396 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3397 | } |
3398 | unlock_buffer(bh); | |
3399 | put_bh(bh); | |
3400 | } | |
3401 | ||
29c36d72 AJ |
3402 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3403 | struct buffer_head **bh_ret) | |
3404 | { | |
3405 | struct buffer_head *bh; | |
3406 | struct btrfs_super_block *super; | |
3407 | u64 bytenr; | |
3408 | ||
3409 | bytenr = btrfs_sb_offset(copy_num); | |
3410 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3411 | return -EINVAL; | |
3412 | ||
9f6d2510 | 3413 | bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); |
29c36d72 AJ |
3414 | /* |
3415 | * If we fail to read from the underlying devices, as of now | |
3416 | * the best option we have is to mark it EIO. | |
3417 | */ | |
3418 | if (!bh) | |
3419 | return -EIO; | |
3420 | ||
3421 | super = (struct btrfs_super_block *)bh->b_data; | |
3422 | if (btrfs_super_bytenr(super) != bytenr || | |
3423 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3424 | brelse(bh); | |
3425 | return -EINVAL; | |
3426 | } | |
3427 | ||
3428 | *bh_ret = bh; | |
3429 | return 0; | |
3430 | } | |
3431 | ||
3432 | ||
a512bbf8 YZ |
3433 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3434 | { | |
3435 | struct buffer_head *bh; | |
3436 | struct buffer_head *latest = NULL; | |
3437 | struct btrfs_super_block *super; | |
3438 | int i; | |
3439 | u64 transid = 0; | |
92fc03fb | 3440 | int ret = -EINVAL; |
a512bbf8 YZ |
3441 | |
3442 | /* we would like to check all the supers, but that would make | |
3443 | * a btrfs mount succeed after a mkfs from a different FS. | |
3444 | * So, we need to add a special mount option to scan for | |
3445 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3446 | */ | |
3447 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3448 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3449 | if (ret) | |
a512bbf8 YZ |
3450 | continue; |
3451 | ||
3452 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3453 | |
3454 | if (!latest || btrfs_super_generation(super) > transid) { | |
3455 | brelse(latest); | |
3456 | latest = bh; | |
3457 | transid = btrfs_super_generation(super); | |
3458 | } else { | |
3459 | brelse(bh); | |
3460 | } | |
3461 | } | |
92fc03fb AJ |
3462 | |
3463 | if (!latest) | |
3464 | return ERR_PTR(ret); | |
3465 | ||
a512bbf8 YZ |
3466 | return latest; |
3467 | } | |
3468 | ||
4eedeb75 | 3469 | /* |
abbb3b8e DS |
3470 | * Write superblock @sb to the @device. Do not wait for completion, all the |
3471 | * buffer heads we write are pinned. | |
4eedeb75 | 3472 | * |
abbb3b8e DS |
3473 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3474 | * the expected device size at commit time. Note that max_mirrors must be | |
3475 | * same for write and wait phases. | |
4eedeb75 | 3476 | * |
abbb3b8e | 3477 | * Return number of errors when buffer head is not found or submission fails. |
4eedeb75 | 3478 | */ |
a512bbf8 | 3479 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3480 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 YZ |
3481 | { |
3482 | struct buffer_head *bh; | |
3483 | int i; | |
3484 | int ret; | |
3485 | int errors = 0; | |
3486 | u32 crc; | |
3487 | u64 bytenr; | |
1b9e619c | 3488 | int op_flags; |
a512bbf8 YZ |
3489 | |
3490 | if (max_mirrors == 0) | |
3491 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3492 | ||
a512bbf8 YZ |
3493 | for (i = 0; i < max_mirrors; i++) { |
3494 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3495 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3496 | device->commit_total_bytes) | |
a512bbf8 YZ |
3497 | break; |
3498 | ||
abbb3b8e | 3499 | btrfs_set_super_bytenr(sb, bytenr); |
4eedeb75 | 3500 | |
abbb3b8e DS |
3501 | crc = ~(u32)0; |
3502 | crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, | |
3503 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
3504 | btrfs_csum_final(crc, sb->csum); | |
4eedeb75 | 3505 | |
abbb3b8e | 3506 | /* One reference for us, and we leave it for the caller */ |
9f6d2510 | 3507 | bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, |
abbb3b8e DS |
3508 | BTRFS_SUPER_INFO_SIZE); |
3509 | if (!bh) { | |
3510 | btrfs_err(device->fs_info, | |
3511 | "couldn't get super buffer head for bytenr %llu", | |
3512 | bytenr); | |
3513 | errors++; | |
4eedeb75 | 3514 | continue; |
abbb3b8e | 3515 | } |
634554dc | 3516 | |
abbb3b8e | 3517 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
a512bbf8 | 3518 | |
abbb3b8e DS |
3519 | /* one reference for submit_bh */ |
3520 | get_bh(bh); | |
4eedeb75 | 3521 | |
abbb3b8e DS |
3522 | set_buffer_uptodate(bh); |
3523 | lock_buffer(bh); | |
3524 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
3525 | bh->b_private = device; | |
a512bbf8 | 3526 | |
387125fc CM |
3527 | /* |
3528 | * we fua the first super. The others we allow | |
3529 | * to go down lazy. | |
3530 | */ | |
1b9e619c OS |
3531 | op_flags = REQ_SYNC | REQ_META | REQ_PRIO; |
3532 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) | |
3533 | op_flags |= REQ_FUA; | |
3534 | ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); | |
4eedeb75 | 3535 | if (ret) |
a512bbf8 | 3536 | errors++; |
a512bbf8 YZ |
3537 | } |
3538 | return errors < i ? 0 : -1; | |
3539 | } | |
3540 | ||
abbb3b8e DS |
3541 | /* |
3542 | * Wait for write completion of superblocks done by write_dev_supers, | |
3543 | * @max_mirrors same for write and wait phases. | |
3544 | * | |
3545 | * Return number of errors when buffer head is not found or not marked up to | |
3546 | * date. | |
3547 | */ | |
3548 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3549 | { | |
3550 | struct buffer_head *bh; | |
3551 | int i; | |
3552 | int errors = 0; | |
b6a535fa | 3553 | bool primary_failed = false; |
abbb3b8e DS |
3554 | u64 bytenr; |
3555 | ||
3556 | if (max_mirrors == 0) | |
3557 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3558 | ||
3559 | for (i = 0; i < max_mirrors; i++) { | |
3560 | bytenr = btrfs_sb_offset(i); | |
3561 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
3562 | device->commit_total_bytes) | |
3563 | break; | |
3564 | ||
9f6d2510 DS |
3565 | bh = __find_get_block(device->bdev, |
3566 | bytenr / BTRFS_BDEV_BLOCKSIZE, | |
abbb3b8e DS |
3567 | BTRFS_SUPER_INFO_SIZE); |
3568 | if (!bh) { | |
3569 | errors++; | |
b6a535fa HM |
3570 | if (i == 0) |
3571 | primary_failed = true; | |
abbb3b8e DS |
3572 | continue; |
3573 | } | |
3574 | wait_on_buffer(bh); | |
b6a535fa | 3575 | if (!buffer_uptodate(bh)) { |
abbb3b8e | 3576 | errors++; |
b6a535fa HM |
3577 | if (i == 0) |
3578 | primary_failed = true; | |
3579 | } | |
abbb3b8e DS |
3580 | |
3581 | /* drop our reference */ | |
3582 | brelse(bh); | |
3583 | ||
3584 | /* drop the reference from the writing run */ | |
3585 | brelse(bh); | |
3586 | } | |
3587 | ||
b6a535fa HM |
3588 | /* log error, force error return */ |
3589 | if (primary_failed) { | |
3590 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3591 | device->devid); | |
3592 | return -1; | |
3593 | } | |
3594 | ||
abbb3b8e DS |
3595 | return errors < i ? 0 : -1; |
3596 | } | |
3597 | ||
387125fc CM |
3598 | /* |
3599 | * endio for the write_dev_flush, this will wake anyone waiting | |
3600 | * for the barrier when it is done | |
3601 | */ | |
4246a0b6 | 3602 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3603 | { |
e0ae9994 | 3604 | complete(bio->bi_private); |
387125fc CM |
3605 | } |
3606 | ||
3607 | /* | |
4fc6441a AJ |
3608 | * Submit a flush request to the device if it supports it. Error handling is |
3609 | * done in the waiting counterpart. | |
387125fc | 3610 | */ |
4fc6441a | 3611 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3612 | { |
c2a9c7ab | 3613 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3614 | struct bio *bio = device->flush_bio; |
387125fc | 3615 | |
c2a9c7ab | 3616 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3617 | return; |
387125fc | 3618 | |
e0ae9994 | 3619 | bio_reset(bio); |
387125fc | 3620 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 3621 | bio_set_dev(bio, device->bdev); |
8d910125 | 3622 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3623 | init_completion(&device->flush_wait); |
3624 | bio->bi_private = &device->flush_wait; | |
387125fc | 3625 | |
43a01111 | 3626 | btrfsic_submit_bio(bio); |
1c3063b6 | 3627 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3628 | } |
387125fc | 3629 | |
4fc6441a AJ |
3630 | /* |
3631 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
3632 | */ | |
8c27cb35 | 3633 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3634 | { |
4fc6441a | 3635 | struct bio *bio = device->flush_bio; |
387125fc | 3636 | |
1c3063b6 | 3637 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 3638 | return BLK_STS_OK; |
387125fc | 3639 | |
1c3063b6 | 3640 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 3641 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3642 | |
8c27cb35 | 3643 | return bio->bi_status; |
387125fc | 3644 | } |
387125fc | 3645 | |
d10b82fe | 3646 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 3647 | { |
6528b99d | 3648 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 3649 | return -EIO; |
387125fc CM |
3650 | return 0; |
3651 | } | |
3652 | ||
3653 | /* | |
3654 | * send an empty flush down to each device in parallel, | |
3655 | * then wait for them | |
3656 | */ | |
3657 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3658 | { | |
3659 | struct list_head *head; | |
3660 | struct btrfs_device *dev; | |
5af3e8cc | 3661 | int errors_wait = 0; |
4e4cbee9 | 3662 | blk_status_t ret; |
387125fc | 3663 | |
1538e6c5 | 3664 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3665 | /* send down all the barriers */ |
3666 | head = &info->fs_devices->devices; | |
1538e6c5 | 3667 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3668 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3669 | continue; |
cea7c8bf | 3670 | if (!dev->bdev) |
387125fc | 3671 | continue; |
e12c9621 | 3672 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3673 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3674 | continue; |
3675 | ||
4fc6441a | 3676 | write_dev_flush(dev); |
58efbc9f | 3677 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
3678 | } |
3679 | ||
3680 | /* wait for all the barriers */ | |
1538e6c5 | 3681 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3682 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3683 | continue; |
387125fc | 3684 | if (!dev->bdev) { |
5af3e8cc | 3685 | errors_wait++; |
387125fc CM |
3686 | continue; |
3687 | } | |
e12c9621 | 3688 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3689 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3690 | continue; |
3691 | ||
4fc6441a | 3692 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
3693 | if (ret) { |
3694 | dev->last_flush_error = ret; | |
66b4993e DS |
3695 | btrfs_dev_stat_inc_and_print(dev, |
3696 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 3697 | errors_wait++; |
401b41e5 AJ |
3698 | } |
3699 | } | |
3700 | ||
cea7c8bf | 3701 | if (errors_wait) { |
401b41e5 AJ |
3702 | /* |
3703 | * At some point we need the status of all disks | |
3704 | * to arrive at the volume status. So error checking | |
3705 | * is being pushed to a separate loop. | |
3706 | */ | |
d10b82fe | 3707 | return check_barrier_error(info); |
387125fc | 3708 | } |
387125fc CM |
3709 | return 0; |
3710 | } | |
3711 | ||
943c6e99 ZL |
3712 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3713 | { | |
8789f4fe ZL |
3714 | int raid_type; |
3715 | int min_tolerated = INT_MAX; | |
943c6e99 | 3716 | |
8789f4fe ZL |
3717 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3718 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3719 | min_tolerated = min(min_tolerated, | |
3720 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3721 | tolerated_failures); | |
943c6e99 | 3722 | |
8789f4fe ZL |
3723 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3724 | if (raid_type == BTRFS_RAID_SINGLE) | |
3725 | continue; | |
41a6e891 | 3726 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe ZL |
3727 | continue; |
3728 | min_tolerated = min(min_tolerated, | |
3729 | btrfs_raid_array[raid_type]. | |
3730 | tolerated_failures); | |
3731 | } | |
943c6e99 | 3732 | |
8789f4fe | 3733 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 3734 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
3735 | min_tolerated = 0; |
3736 | } | |
3737 | ||
3738 | return min_tolerated; | |
943c6e99 ZL |
3739 | } |
3740 | ||
eece6a9c | 3741 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 3742 | { |
e5e9a520 | 3743 | struct list_head *head; |
f2984462 | 3744 | struct btrfs_device *dev; |
a061fc8d | 3745 | struct btrfs_super_block *sb; |
f2984462 | 3746 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3747 | int ret; |
3748 | int do_barriers; | |
a236aed1 CM |
3749 | int max_errors; |
3750 | int total_errors = 0; | |
a061fc8d | 3751 | u64 flags; |
f2984462 | 3752 | |
0b246afa | 3753 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
3754 | |
3755 | /* | |
3756 | * max_mirrors == 0 indicates we're from commit_transaction, | |
3757 | * not from fsync where the tree roots in fs_info have not | |
3758 | * been consistent on disk. | |
3759 | */ | |
3760 | if (max_mirrors == 0) | |
3761 | backup_super_roots(fs_info); | |
f2984462 | 3762 | |
0b246afa | 3763 | sb = fs_info->super_for_commit; |
a061fc8d | 3764 | dev_item = &sb->dev_item; |
e5e9a520 | 3765 | |
0b246afa JM |
3766 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
3767 | head = &fs_info->fs_devices->devices; | |
3768 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 3769 | |
5af3e8cc | 3770 | if (do_barriers) { |
0b246afa | 3771 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
3772 | if (ret) { |
3773 | mutex_unlock( | |
0b246afa JM |
3774 | &fs_info->fs_devices->device_list_mutex); |
3775 | btrfs_handle_fs_error(fs_info, ret, | |
3776 | "errors while submitting device barriers."); | |
5af3e8cc SB |
3777 | return ret; |
3778 | } | |
3779 | } | |
387125fc | 3780 | |
1538e6c5 | 3781 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3782 | if (!dev->bdev) { |
3783 | total_errors++; | |
3784 | continue; | |
3785 | } | |
e12c9621 | 3786 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3787 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3788 | continue; |
3789 | ||
2b82032c | 3790 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3791 | btrfs_set_stack_device_type(dev_item, dev->type); |
3792 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3793 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3794 | dev->commit_total_bytes); |
ce7213c7 MX |
3795 | btrfs_set_stack_device_bytes_used(dev_item, |
3796 | dev->commit_bytes_used); | |
a061fc8d CM |
3797 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3798 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3799 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3800 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
44880fdc | 3801 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); |
a512bbf8 | 3802 | |
a061fc8d CM |
3803 | flags = btrfs_super_flags(sb); |
3804 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3805 | ||
75cb857d QW |
3806 | ret = btrfs_validate_write_super(fs_info, sb); |
3807 | if (ret < 0) { | |
3808 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
3809 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
3810 | "unexpected superblock corruption detected"); | |
3811 | return -EUCLEAN; | |
3812 | } | |
3813 | ||
abbb3b8e | 3814 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
3815 | if (ret) |
3816 | total_errors++; | |
f2984462 | 3817 | } |
a236aed1 | 3818 | if (total_errors > max_errors) { |
0b246afa JM |
3819 | btrfs_err(fs_info, "%d errors while writing supers", |
3820 | total_errors); | |
3821 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 3822 | |
9d565ba4 | 3823 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
3824 | btrfs_handle_fs_error(fs_info, -EIO, |
3825 | "%d errors while writing supers", | |
3826 | total_errors); | |
9d565ba4 | 3827 | return -EIO; |
a236aed1 | 3828 | } |
f2984462 | 3829 | |
a512bbf8 | 3830 | total_errors = 0; |
1538e6c5 | 3831 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
3832 | if (!dev->bdev) |
3833 | continue; | |
e12c9621 | 3834 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3835 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
3836 | continue; |
3837 | ||
abbb3b8e | 3838 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
3839 | if (ret) |
3840 | total_errors++; | |
f2984462 | 3841 | } |
0b246afa | 3842 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3843 | if (total_errors > max_errors) { |
0b246afa JM |
3844 | btrfs_handle_fs_error(fs_info, -EIO, |
3845 | "%d errors while writing supers", | |
3846 | total_errors); | |
79787eaa | 3847 | return -EIO; |
a236aed1 | 3848 | } |
f2984462 CM |
3849 | return 0; |
3850 | } | |
3851 | ||
cb517eab MX |
3852 | /* Drop a fs root from the radix tree and free it. */ |
3853 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3854 | struct btrfs_root *root) | |
2619ba1f | 3855 | { |
4df27c4d | 3856 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3857 | radix_tree_delete(&fs_info->fs_roots_radix, |
3858 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3859 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3860 | |
3861 | if (btrfs_root_refs(&root->root_item) == 0) | |
3862 | synchronize_srcu(&fs_info->subvol_srcu); | |
3863 | ||
1c1ea4f7 | 3864 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3865 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3866 | if (root->reloc_root) { |
3867 | free_extent_buffer(root->reloc_root->node); | |
3868 | free_extent_buffer(root->reloc_root->commit_root); | |
3869 | btrfs_put_fs_root(root->reloc_root); | |
3870 | root->reloc_root = NULL; | |
3871 | } | |
3872 | } | |
3321719e | 3873 | |
faa2dbf0 JB |
3874 | if (root->free_ino_pinned) |
3875 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3876 | if (root->free_ino_ctl) | |
3877 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3878 | free_fs_root(root); |
4df27c4d YZ |
3879 | } |
3880 | ||
3881 | static void free_fs_root(struct btrfs_root *root) | |
3882 | { | |
57cdc8db | 3883 | iput(root->ino_cache_inode); |
4df27c4d | 3884 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
3885 | if (root->anon_dev) |
3886 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3887 | if (root->subv_writers) |
3888 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3889 | free_extent_buffer(root->node); |
3890 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3891 | kfree(root->free_ino_ctl); |
3892 | kfree(root->free_ino_pinned); | |
d397712b | 3893 | kfree(root->name); |
b0feb9d9 | 3894 | btrfs_put_fs_root(root); |
2619ba1f CM |
3895 | } |
3896 | ||
cb517eab MX |
3897 | void btrfs_free_fs_root(struct btrfs_root *root) |
3898 | { | |
3899 | free_fs_root(root); | |
2619ba1f CM |
3900 | } |
3901 | ||
c146afad | 3902 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3903 | { |
c146afad YZ |
3904 | u64 root_objectid = 0; |
3905 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3906 | int i = 0; |
3907 | int err = 0; | |
3908 | unsigned int ret = 0; | |
3909 | int index; | |
e089f05c | 3910 | |
c146afad | 3911 | while (1) { |
65d33fd7 | 3912 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3913 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3914 | (void **)gang, root_objectid, | |
3915 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3916 | if (!ret) { |
3917 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3918 | break; |
65d33fd7 | 3919 | } |
5d4f98a2 | 3920 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3921 | |
c146afad | 3922 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3923 | /* Avoid to grab roots in dead_roots */ |
3924 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3925 | gang[i] = NULL; | |
3926 | continue; | |
3927 | } | |
3928 | /* grab all the search result for later use */ | |
3929 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3930 | } | |
3931 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3932 | |
65d33fd7 QW |
3933 | for (i = 0; i < ret; i++) { |
3934 | if (!gang[i]) | |
3935 | continue; | |
c146afad | 3936 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3937 | err = btrfs_orphan_cleanup(gang[i]); |
3938 | if (err) | |
65d33fd7 QW |
3939 | break; |
3940 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3941 | } |
3942 | root_objectid++; | |
3943 | } | |
65d33fd7 QW |
3944 | |
3945 | /* release the uncleaned roots due to error */ | |
3946 | for (; i < ret; i++) { | |
3947 | if (gang[i]) | |
3948 | btrfs_put_fs_root(gang[i]); | |
3949 | } | |
3950 | return err; | |
c146afad | 3951 | } |
a2135011 | 3952 | |
6bccf3ab | 3953 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 3954 | { |
6bccf3ab | 3955 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 3956 | struct btrfs_trans_handle *trans; |
a74a4b97 | 3957 | |
0b246afa | 3958 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 3959 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
3960 | mutex_unlock(&fs_info->cleaner_mutex); |
3961 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
3962 | |
3963 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
3964 | down_write(&fs_info->cleanup_work_sem); |
3965 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 3966 | |
7a7eaa40 | 3967 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3968 | if (IS_ERR(trans)) |
3969 | return PTR_ERR(trans); | |
3a45bb20 | 3970 | return btrfs_commit_transaction(trans); |
c146afad YZ |
3971 | } |
3972 | ||
6bccf3ab | 3973 | void close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 3974 | { |
c146afad YZ |
3975 | int ret; |
3976 | ||
afcdd129 | 3977 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
c146afad | 3978 | |
7343dd61 | 3979 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3980 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3981 | |
803b2f54 SB |
3982 | /* wait for the uuid_scan task to finish */ |
3983 | down(&fs_info->uuid_tree_rescan_sem); | |
3984 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3985 | up(&fs_info->uuid_tree_rescan_sem); | |
3986 | ||
837d5b6e | 3987 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3988 | btrfs_pause_balance(fs_info); |
837d5b6e | 3989 | |
8dabb742 SB |
3990 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3991 | ||
aa1b8cd4 | 3992 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3993 | |
3994 | /* wait for any defraggers to finish */ | |
3995 | wait_event(fs_info->transaction_wait, | |
3996 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3997 | ||
3998 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3999 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4000 | |
21c7e756 MX |
4001 | cancel_work_sync(&fs_info->async_reclaim_work); |
4002 | ||
bc98a42c | 4003 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 JM |
4004 | /* |
4005 | * If the cleaner thread is stopped and there are | |
4006 | * block groups queued for removal, the deletion will be | |
4007 | * skipped when we quit the cleaner thread. | |
4008 | */ | |
0b246afa | 4009 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4010 | |
6bccf3ab | 4011 | ret = btrfs_commit_super(fs_info); |
acce952b | 4012 | if (ret) |
04892340 | 4013 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4014 | } |
4015 | ||
af722733 LB |
4016 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || |
4017 | test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) | |
2ff7e61e | 4018 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4019 | |
e3029d9f AV |
4020 | kthread_stop(fs_info->transaction_kthread); |
4021 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4022 | |
afcdd129 | 4023 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4024 | |
04892340 | 4025 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4026 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4027 | |
963d678b | 4028 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4029 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4030 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4031 | } |
bcc63abb | 4032 | |
6618a59b | 4033 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4034 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4035 | |
faa2dbf0 | 4036 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 4037 | |
1a4319cc LB |
4038 | btrfs_put_block_group_cache(fs_info); |
4039 | ||
de348ee0 WS |
4040 | /* |
4041 | * we must make sure there is not any read request to | |
4042 | * submit after we stopping all workers. | |
4043 | */ | |
4044 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4045 | btrfs_stop_all_workers(fs_info); |
4046 | ||
5cdd7db6 FM |
4047 | btrfs_free_block_groups(fs_info); |
4048 | ||
afcdd129 | 4049 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
13e6c37b | 4050 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 4051 | |
13e6c37b | 4052 | iput(fs_info->btree_inode); |
d6bfde87 | 4053 | |
21adbd5c | 4054 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4055 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4056 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4057 | #endif |
4058 | ||
dfe25020 | 4059 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 4060 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 4061 | |
e2d84521 | 4062 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 4063 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 4064 | percpu_counter_destroy(&fs_info->bio_counter); |
76dda93c | 4065 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 4066 | |
53b381b3 | 4067 | btrfs_free_stripe_hash_table(fs_info); |
fd708b81 | 4068 | btrfs_free_ref_cache(fs_info); |
53b381b3 | 4069 | |
04216820 FM |
4070 | while (!list_empty(&fs_info->pinned_chunks)) { |
4071 | struct extent_map *em; | |
4072 | ||
4073 | em = list_first_entry(&fs_info->pinned_chunks, | |
4074 | struct extent_map, list); | |
4075 | list_del_init(&em->list); | |
4076 | free_extent_map(em); | |
4077 | } | |
eb60ceac CM |
4078 | } |
4079 | ||
b9fab919 CM |
4080 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4081 | int atomic) | |
5f39d397 | 4082 | { |
1259ab75 | 4083 | int ret; |
727011e0 | 4084 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4085 | |
0b32f4bb | 4086 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4087 | if (!ret) |
4088 | return ret; | |
4089 | ||
4090 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4091 | parent_transid, atomic); |
4092 | if (ret == -EAGAIN) | |
4093 | return ret; | |
1259ab75 | 4094 | return !ret; |
5f39d397 CM |
4095 | } |
4096 | ||
5f39d397 CM |
4097 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4098 | { | |
0b246afa | 4099 | struct btrfs_fs_info *fs_info; |
06ea65a3 | 4100 | struct btrfs_root *root; |
5f39d397 | 4101 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4102 | int was_dirty; |
b4ce94de | 4103 | |
06ea65a3 JB |
4104 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4105 | /* | |
4106 | * This is a fast path so only do this check if we have sanity tests | |
4107 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
4108 | * outside of the sanity tests. | |
4109 | */ | |
4110 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
4111 | return; | |
4112 | #endif | |
4113 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
0b246afa | 4114 | fs_info = root->fs_info; |
b9447ef8 | 4115 | btrfs_assert_tree_locked(buf); |
0b246afa | 4116 | if (transid != fs_info->generation) |
5d163e0e | 4117 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4118 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4119 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4120 | if (!was_dirty) |
104b4e51 NB |
4121 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4122 | buf->len, | |
4123 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4124 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4125 | /* |
4126 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4127 | * but item data not updated. | |
4128 | * So here we should only check item pointers, not item data. | |
4129 | */ | |
4130 | if (btrfs_header_level(buf) == 0 && | |
2f659546 | 4131 | btrfs_check_leaf_relaxed(fs_info, buf)) { |
a4f78750 | 4132 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4133 | ASSERT(0); |
4134 | } | |
4135 | #endif | |
eb60ceac CM |
4136 | } |
4137 | ||
2ff7e61e | 4138 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4139 | int flush_delayed) |
16cdcec7 MX |
4140 | { |
4141 | /* | |
4142 | * looks as though older kernels can get into trouble with | |
4143 | * this code, they end up stuck in balance_dirty_pages forever | |
4144 | */ | |
e2d84521 | 4145 | int ret; |
16cdcec7 MX |
4146 | |
4147 | if (current->flags & PF_MEMALLOC) | |
4148 | return; | |
4149 | ||
b53d3f5d | 4150 | if (flush_delayed) |
2ff7e61e | 4151 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4152 | |
0b246afa | 4153 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
e2d84521 MX |
4154 | BTRFS_DIRTY_METADATA_THRESH); |
4155 | if (ret > 0) { | |
0b246afa | 4156 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4157 | } |
16cdcec7 MX |
4158 | } |
4159 | ||
2ff7e61e | 4160 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4161 | { |
2ff7e61e | 4162 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4163 | } |
585ad2c3 | 4164 | |
2ff7e61e | 4165 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4166 | { |
2ff7e61e | 4167 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4168 | } |
6b80053d | 4169 | |
581c1760 QW |
4170 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, |
4171 | struct btrfs_key *first_key) | |
6b80053d | 4172 | { |
727011e0 | 4173 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
2ff7e61e JM |
4174 | struct btrfs_fs_info *fs_info = root->fs_info; |
4175 | ||
581c1760 QW |
4176 | return btree_read_extent_buffer_pages(fs_info, buf, parent_transid, |
4177 | level, first_key); | |
6b80053d | 4178 | } |
0da5468f | 4179 | |
2ff7e61e | 4180 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4181 | { |
fe816d0f NB |
4182 | /* cleanup FS via transaction */ |
4183 | btrfs_cleanup_transaction(fs_info); | |
4184 | ||
0b246afa | 4185 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4186 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4187 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4188 | |
0b246afa JM |
4189 | down_write(&fs_info->cleanup_work_sem); |
4190 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4191 | } |
4192 | ||
143bede5 | 4193 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4194 | { |
acce952b | 4195 | struct btrfs_ordered_extent *ordered; |
acce952b | 4196 | |
199c2a9c | 4197 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4198 | /* |
4199 | * This will just short circuit the ordered completion stuff which will | |
4200 | * make sure the ordered extent gets properly cleaned up. | |
4201 | */ | |
199c2a9c | 4202 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4203 | root_extent_list) |
4204 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4205 | spin_unlock(&root->ordered_extent_lock); |
4206 | } | |
4207 | ||
4208 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4209 | { | |
4210 | struct btrfs_root *root; | |
4211 | struct list_head splice; | |
4212 | ||
4213 | INIT_LIST_HEAD(&splice); | |
4214 | ||
4215 | spin_lock(&fs_info->ordered_root_lock); | |
4216 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4217 | while (!list_empty(&splice)) { | |
4218 | root = list_first_entry(&splice, struct btrfs_root, | |
4219 | ordered_root); | |
1de2cfde JB |
4220 | list_move_tail(&root->ordered_root, |
4221 | &fs_info->ordered_roots); | |
199c2a9c | 4222 | |
2a85d9ca | 4223 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4224 | btrfs_destroy_ordered_extents(root); |
4225 | ||
2a85d9ca LB |
4226 | cond_resched(); |
4227 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4228 | } |
4229 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4230 | } |
4231 | ||
35a3621b | 4232 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4233 | struct btrfs_fs_info *fs_info) |
acce952b | 4234 | { |
4235 | struct rb_node *node; | |
4236 | struct btrfs_delayed_ref_root *delayed_refs; | |
4237 | struct btrfs_delayed_ref_node *ref; | |
4238 | int ret = 0; | |
4239 | ||
4240 | delayed_refs = &trans->delayed_refs; | |
4241 | ||
4242 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4243 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4244 | spin_unlock(&delayed_refs->lock); |
0b246afa | 4245 | btrfs_info(fs_info, "delayed_refs has NO entry"); |
acce952b | 4246 | return ret; |
4247 | } | |
4248 | ||
d7df2c79 JB |
4249 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4250 | struct btrfs_delayed_ref_head *head; | |
0e0adbcf | 4251 | struct rb_node *n; |
e78417d1 | 4252 | bool pin_bytes = false; |
acce952b | 4253 | |
d7df2c79 JB |
4254 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4255 | href_node); | |
4256 | if (!mutex_trylock(&head->mutex)) { | |
d278850e | 4257 | refcount_inc(&head->refs); |
d7df2c79 | 4258 | spin_unlock(&delayed_refs->lock); |
eb12db69 | 4259 | |
d7df2c79 | 4260 | mutex_lock(&head->mutex); |
e78417d1 | 4261 | mutex_unlock(&head->mutex); |
d278850e | 4262 | btrfs_put_delayed_ref_head(head); |
d7df2c79 JB |
4263 | spin_lock(&delayed_refs->lock); |
4264 | continue; | |
4265 | } | |
4266 | spin_lock(&head->lock); | |
0e0adbcf JB |
4267 | while ((n = rb_first(&head->ref_tree)) != NULL) { |
4268 | ref = rb_entry(n, struct btrfs_delayed_ref_node, | |
4269 | ref_node); | |
d7df2c79 | 4270 | ref->in_tree = 0; |
0e0adbcf JB |
4271 | rb_erase(&ref->ref_node, &head->ref_tree); |
4272 | RB_CLEAR_NODE(&ref->ref_node); | |
1d57ee94 WX |
4273 | if (!list_empty(&ref->add_list)) |
4274 | list_del(&ref->add_list); | |
d7df2c79 JB |
4275 | atomic_dec(&delayed_refs->num_entries); |
4276 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4277 | } |
d7df2c79 JB |
4278 | if (head->must_insert_reserved) |
4279 | pin_bytes = true; | |
4280 | btrfs_free_delayed_extent_op(head->extent_op); | |
4281 | delayed_refs->num_heads--; | |
4282 | if (head->processing == 0) | |
4283 | delayed_refs->num_heads_ready--; | |
4284 | atomic_dec(&delayed_refs->num_entries); | |
d7df2c79 | 4285 | rb_erase(&head->href_node, &delayed_refs->href_root); |
d278850e | 4286 | RB_CLEAR_NODE(&head->href_node); |
d7df2c79 JB |
4287 | spin_unlock(&head->lock); |
4288 | spin_unlock(&delayed_refs->lock); | |
4289 | mutex_unlock(&head->mutex); | |
acce952b | 4290 | |
d7df2c79 | 4291 | if (pin_bytes) |
d278850e JB |
4292 | btrfs_pin_extent(fs_info, head->bytenr, |
4293 | head->num_bytes, 1); | |
4294 | btrfs_put_delayed_ref_head(head); | |
acce952b | 4295 | cond_resched(); |
4296 | spin_lock(&delayed_refs->lock); | |
4297 | } | |
4298 | ||
4299 | spin_unlock(&delayed_refs->lock); | |
4300 | ||
4301 | return ret; | |
4302 | } | |
4303 | ||
143bede5 | 4304 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4305 | { |
4306 | struct btrfs_inode *btrfs_inode; | |
4307 | struct list_head splice; | |
4308 | ||
4309 | INIT_LIST_HEAD(&splice); | |
4310 | ||
eb73c1b7 MX |
4311 | spin_lock(&root->delalloc_lock); |
4312 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4313 | |
4314 | while (!list_empty(&splice)) { | |
fe816d0f | 4315 | struct inode *inode = NULL; |
eb73c1b7 MX |
4316 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4317 | delalloc_inodes); | |
fe816d0f | 4318 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4319 | spin_unlock(&root->delalloc_lock); |
acce952b | 4320 | |
fe816d0f NB |
4321 | /* |
4322 | * Make sure we get a live inode and that it'll not disappear | |
4323 | * meanwhile. | |
4324 | */ | |
4325 | inode = igrab(&btrfs_inode->vfs_inode); | |
4326 | if (inode) { | |
4327 | invalidate_inode_pages2(inode->i_mapping); | |
4328 | iput(inode); | |
4329 | } | |
eb73c1b7 | 4330 | spin_lock(&root->delalloc_lock); |
acce952b | 4331 | } |
eb73c1b7 MX |
4332 | spin_unlock(&root->delalloc_lock); |
4333 | } | |
4334 | ||
4335 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4336 | { | |
4337 | struct btrfs_root *root; | |
4338 | struct list_head splice; | |
4339 | ||
4340 | INIT_LIST_HEAD(&splice); | |
4341 | ||
4342 | spin_lock(&fs_info->delalloc_root_lock); | |
4343 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4344 | while (!list_empty(&splice)) { | |
4345 | root = list_first_entry(&splice, struct btrfs_root, | |
4346 | delalloc_root); | |
eb73c1b7 MX |
4347 | root = btrfs_grab_fs_root(root); |
4348 | BUG_ON(!root); | |
4349 | spin_unlock(&fs_info->delalloc_root_lock); | |
4350 | ||
4351 | btrfs_destroy_delalloc_inodes(root); | |
4352 | btrfs_put_fs_root(root); | |
4353 | ||
4354 | spin_lock(&fs_info->delalloc_root_lock); | |
4355 | } | |
4356 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4357 | } |
4358 | ||
2ff7e61e | 4359 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4360 | struct extent_io_tree *dirty_pages, |
4361 | int mark) | |
4362 | { | |
4363 | int ret; | |
acce952b | 4364 | struct extent_buffer *eb; |
4365 | u64 start = 0; | |
4366 | u64 end; | |
acce952b | 4367 | |
4368 | while (1) { | |
4369 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4370 | mark, NULL); |
acce952b | 4371 | if (ret) |
4372 | break; | |
4373 | ||
91166212 | 4374 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4375 | while (start <= end) { |
0b246afa JM |
4376 | eb = find_extent_buffer(fs_info, start); |
4377 | start += fs_info->nodesize; | |
fd8b2b61 | 4378 | if (!eb) |
acce952b | 4379 | continue; |
fd8b2b61 | 4380 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4381 | |
fd8b2b61 JB |
4382 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4383 | &eb->bflags)) | |
4384 | clear_extent_buffer_dirty(eb); | |
4385 | free_extent_buffer_stale(eb); | |
acce952b | 4386 | } |
4387 | } | |
4388 | ||
4389 | return ret; | |
4390 | } | |
4391 | ||
2ff7e61e | 4392 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 4393 | struct extent_io_tree *pinned_extents) |
4394 | { | |
4395 | struct extent_io_tree *unpin; | |
4396 | u64 start; | |
4397 | u64 end; | |
4398 | int ret; | |
ed0eaa14 | 4399 | bool loop = true; |
acce952b | 4400 | |
4401 | unpin = pinned_extents; | |
ed0eaa14 | 4402 | again: |
acce952b | 4403 | while (1) { |
4404 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4405 | EXTENT_DIRTY, NULL); |
acce952b | 4406 | if (ret) |
4407 | break; | |
4408 | ||
af6f8f60 | 4409 | clear_extent_dirty(unpin, start, end); |
2ff7e61e | 4410 | btrfs_error_unpin_extent_range(fs_info, start, end); |
acce952b | 4411 | cond_resched(); |
4412 | } | |
4413 | ||
ed0eaa14 | 4414 | if (loop) { |
0b246afa JM |
4415 | if (unpin == &fs_info->freed_extents[0]) |
4416 | unpin = &fs_info->freed_extents[1]; | |
ed0eaa14 | 4417 | else |
0b246afa | 4418 | unpin = &fs_info->freed_extents[0]; |
ed0eaa14 LB |
4419 | loop = false; |
4420 | goto again; | |
4421 | } | |
4422 | ||
acce952b | 4423 | return 0; |
4424 | } | |
4425 | ||
c79a1751 LB |
4426 | static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) |
4427 | { | |
4428 | struct inode *inode; | |
4429 | ||
4430 | inode = cache->io_ctl.inode; | |
4431 | if (inode) { | |
4432 | invalidate_inode_pages2(inode->i_mapping); | |
4433 | BTRFS_I(inode)->generation = 0; | |
4434 | cache->io_ctl.inode = NULL; | |
4435 | iput(inode); | |
4436 | } | |
4437 | btrfs_put_block_group(cache); | |
4438 | } | |
4439 | ||
4440 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4441 | struct btrfs_fs_info *fs_info) |
c79a1751 LB |
4442 | { |
4443 | struct btrfs_block_group_cache *cache; | |
4444 | ||
4445 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4446 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4447 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
4448 | struct btrfs_block_group_cache, | |
4449 | dirty_list); | |
c79a1751 LB |
4450 | |
4451 | if (!list_empty(&cache->io_list)) { | |
4452 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4453 | list_del_init(&cache->io_list); | |
4454 | btrfs_cleanup_bg_io(cache); | |
4455 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4456 | } | |
4457 | ||
4458 | list_del_init(&cache->dirty_list); | |
4459 | spin_lock(&cache->lock); | |
4460 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4461 | spin_unlock(&cache->lock); | |
4462 | ||
4463 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4464 | btrfs_put_block_group(cache); | |
4465 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4466 | } | |
4467 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4468 | ||
45ae2c18 NB |
4469 | /* |
4470 | * Refer to the definition of io_bgs member for details why it's safe | |
4471 | * to use it without any locking | |
4472 | */ | |
c79a1751 LB |
4473 | while (!list_empty(&cur_trans->io_bgs)) { |
4474 | cache = list_first_entry(&cur_trans->io_bgs, | |
4475 | struct btrfs_block_group_cache, | |
4476 | io_list); | |
c79a1751 LB |
4477 | |
4478 | list_del_init(&cache->io_list); | |
4479 | spin_lock(&cache->lock); | |
4480 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4481 | spin_unlock(&cache->lock); | |
4482 | btrfs_cleanup_bg_io(cache); | |
4483 | } | |
4484 | } | |
4485 | ||
49b25e05 | 4486 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4487 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4488 | { |
2ff7e61e | 4489 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4490 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4491 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4492 | ||
2ff7e61e | 4493 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4494 | |
4a9d8bde | 4495 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4496 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4497 | |
4a9d8bde | 4498 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4499 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4500 | |
ccdf9b30 JM |
4501 | btrfs_destroy_delayed_inodes(fs_info); |
4502 | btrfs_assert_delayed_root_empty(fs_info); | |
49b25e05 | 4503 | |
2ff7e61e | 4504 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4505 | EXTENT_DIRTY); |
2ff7e61e | 4506 | btrfs_destroy_pinned_extent(fs_info, |
0b246afa | 4507 | fs_info->pinned_extents); |
49b25e05 | 4508 | |
4a9d8bde MX |
4509 | cur_trans->state =TRANS_STATE_COMPLETED; |
4510 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4511 | } |
4512 | ||
2ff7e61e | 4513 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4514 | { |
4515 | struct btrfs_transaction *t; | |
acce952b | 4516 | |
0b246afa | 4517 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4518 | |
0b246afa JM |
4519 | spin_lock(&fs_info->trans_lock); |
4520 | while (!list_empty(&fs_info->trans_list)) { | |
4521 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4522 | struct btrfs_transaction, list); |
4523 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4524 | refcount_inc(&t->use_count); |
0b246afa | 4525 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4526 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4527 | btrfs_put_transaction(t); |
0b246afa | 4528 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4529 | continue; |
4530 | } | |
0b246afa | 4531 | if (t == fs_info->running_transaction) { |
724e2315 | 4532 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4533 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4534 | /* |
4535 | * We wait for 0 num_writers since we don't hold a trans | |
4536 | * handle open currently for this transaction. | |
4537 | */ | |
4538 | wait_event(t->writer_wait, | |
4539 | atomic_read(&t->num_writers) == 0); | |
4540 | } else { | |
0b246afa | 4541 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4542 | } |
2ff7e61e | 4543 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4544 | |
0b246afa JM |
4545 | spin_lock(&fs_info->trans_lock); |
4546 | if (t == fs_info->running_transaction) | |
4547 | fs_info->running_transaction = NULL; | |
acce952b | 4548 | list_del_init(&t->list); |
0b246afa | 4549 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4550 | |
724e2315 | 4551 | btrfs_put_transaction(t); |
2ff7e61e | 4552 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4553 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4554 | } |
0b246afa JM |
4555 | spin_unlock(&fs_info->trans_lock); |
4556 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4557 | btrfs_destroy_delayed_inodes(fs_info); |
4558 | btrfs_assert_delayed_root_empty(fs_info); | |
2ff7e61e | 4559 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
0b246afa JM |
4560 | btrfs_destroy_all_delalloc_inodes(fs_info); |
4561 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
acce952b | 4562 | |
4563 | return 0; | |
4564 | } | |
4565 | ||
c6100a4b JB |
4566 | static struct btrfs_fs_info *btree_fs_info(void *private_data) |
4567 | { | |
4568 | struct inode *inode = private_data; | |
4569 | return btrfs_sb(inode->i_sb); | |
4570 | } | |
4571 | ||
e8c9f186 | 4572 | static const struct extent_io_ops btree_extent_io_ops = { |
4d53dddb | 4573 | /* mandatory callbacks */ |
0b86a832 | 4574 | .submit_bio_hook = btree_submit_bio_hook, |
4d53dddb | 4575 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
239b14b3 CM |
4576 | /* note we're sharing with inode.c for the merge bio hook */ |
4577 | .merge_bio_hook = btrfs_merge_bio_hook, | |
20a7db8a | 4578 | .readpage_io_failed_hook = btree_io_failed_hook, |
c6100a4b JB |
4579 | .set_range_writeback = btrfs_set_range_writeback, |
4580 | .tree_fs_info = btree_fs_info, | |
4d53dddb DS |
4581 | |
4582 | /* optional callbacks */ | |
0da5468f | 4583 | }; |