]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
5a0e3ad6 | 28 | #include <linux/slab.h> |
784b4e29 | 29 | #include <linux/migrate.h> |
7a36ddec | 30 | #include <linux/ratelimit.h> |
6463fe58 | 31 | #include <linux/uuid.h> |
803b2f54 | 32 | #include <linux/semaphore.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
eb60ceac CM |
34 | #include "ctree.h" |
35 | #include "disk-io.h" | |
0b947aff | 36 | #include "hash.h" |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
925baedd | 41 | #include "locking.h" |
e02119d5 | 42 | #include "tree-log.h" |
fa9c0d79 | 43 | #include "free-space-cache.h" |
70f6d82e | 44 | #include "free-space-tree.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
606686ee | 47 | #include "rcu-string.h" |
8dabb742 | 48 | #include "dev-replace.h" |
53b381b3 | 49 | #include "raid56.h" |
5ac1d209 | 50 | #include "sysfs.h" |
fcebe456 | 51 | #include "qgroup.h" |
ebb8765b | 52 | #include "compression.h" |
eb60ceac | 53 | |
de0022b9 JB |
54 | #ifdef CONFIG_X86 |
55 | #include <asm/cpufeature.h> | |
56 | #endif | |
57 | ||
319e4d06 QW |
58 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
59 | BTRFS_HEADER_FLAG_RELOC |\ | |
60 | BTRFS_SUPER_FLAG_ERROR |\ | |
61 | BTRFS_SUPER_FLAG_SEEDING |\ | |
62 | BTRFS_SUPER_FLAG_METADUMP) | |
63 | ||
e8c9f186 | 64 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 65 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 66 | static void free_fs_root(struct btrfs_root *root); |
3d3a126a | 67 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info); |
143bede5 | 68 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 69 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 70 | struct btrfs_fs_info *fs_info); |
143bede5 | 71 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 72 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 73 | struct extent_io_tree *dirty_pages, |
74 | int mark); | |
2ff7e61e | 75 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 76 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
77 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
78 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 79 | |
d352ac68 | 80 | /* |
97eb6b69 DS |
81 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
82 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
83 | * by writes to insert metadata for new file extents after IO is complete. |
84 | */ | |
97eb6b69 | 85 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
86 | struct bio *bio; |
87 | bio_end_io_t *end_io; | |
88 | void *private; | |
89 | struct btrfs_fs_info *info; | |
4e4cbee9 | 90 | blk_status_t status; |
bfebd8b5 | 91 | enum btrfs_wq_endio_type metadata; |
8b712842 | 92 | struct btrfs_work work; |
ce9adaa5 | 93 | }; |
0da5468f | 94 | |
97eb6b69 DS |
95 | static struct kmem_cache *btrfs_end_io_wq_cache; |
96 | ||
97 | int __init btrfs_end_io_wq_init(void) | |
98 | { | |
99 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
100 | sizeof(struct btrfs_end_io_wq), | |
101 | 0, | |
fba4b697 | 102 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
103 | NULL); |
104 | if (!btrfs_end_io_wq_cache) | |
105 | return -ENOMEM; | |
106 | return 0; | |
107 | } | |
108 | ||
109 | void btrfs_end_io_wq_exit(void) | |
110 | { | |
5598e900 | 111 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
112 | } |
113 | ||
d352ac68 CM |
114 | /* |
115 | * async submit bios are used to offload expensive checksumming | |
116 | * onto the worker threads. They checksum file and metadata bios | |
117 | * just before they are sent down the IO stack. | |
118 | */ | |
44b8bd7e | 119 | struct async_submit_bio { |
c6100a4b JB |
120 | void *private_data; |
121 | struct btrfs_fs_info *fs_info; | |
44b8bd7e | 122 | struct bio *bio; |
4a69a410 CM |
123 | extent_submit_bio_hook_t *submit_bio_start; |
124 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e | 125 | int mirror_num; |
c8b97818 | 126 | unsigned long bio_flags; |
eaf25d93 CM |
127 | /* |
128 | * bio_offset is optional, can be used if the pages in the bio | |
129 | * can't tell us where in the file the bio should go | |
130 | */ | |
131 | u64 bio_offset; | |
8b712842 | 132 | struct btrfs_work work; |
4e4cbee9 | 133 | blk_status_t status; |
44b8bd7e CM |
134 | }; |
135 | ||
85d4e461 CM |
136 | /* |
137 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
138 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
139 | * the level the eb occupies in the tree. | |
140 | * | |
141 | * Different roots are used for different purposes and may nest inside each | |
142 | * other and they require separate keysets. As lockdep keys should be | |
143 | * static, assign keysets according to the purpose of the root as indicated | |
144 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
145 | * have separate keysets. | |
4008c04a | 146 | * |
85d4e461 CM |
147 | * Lock-nesting across peer nodes is always done with the immediate parent |
148 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
149 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 150 | * |
85d4e461 CM |
151 | * The key is set by the readpage_end_io_hook after the buffer has passed |
152 | * csum validation but before the pages are unlocked. It is also set by | |
153 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 154 | * |
85d4e461 CM |
155 | * We also add a check to make sure the highest level of the tree is the |
156 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
157 | * needs update as well. | |
4008c04a CM |
158 | */ |
159 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
160 | # if BTRFS_MAX_LEVEL != 8 | |
161 | # error | |
162 | # endif | |
85d4e461 CM |
163 | |
164 | static struct btrfs_lockdep_keyset { | |
165 | u64 id; /* root objectid */ | |
166 | const char *name_stem; /* lock name stem */ | |
167 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
168 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
169 | } btrfs_lockdep_keysets[] = { | |
170 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
171 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
172 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
173 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
174 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
175 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 176 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
177 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
178 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
179 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 180 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 181 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 182 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 183 | }; |
85d4e461 CM |
184 | |
185 | void __init btrfs_init_lockdep(void) | |
186 | { | |
187 | int i, j; | |
188 | ||
189 | /* initialize lockdep class names */ | |
190 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
191 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
192 | ||
193 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
194 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
195 | "btrfs-%s-%02d", ks->name_stem, j); | |
196 | } | |
197 | } | |
198 | ||
199 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
200 | int level) | |
201 | { | |
202 | struct btrfs_lockdep_keyset *ks; | |
203 | ||
204 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
205 | ||
206 | /* find the matching keyset, id 0 is the default entry */ | |
207 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
208 | if (ks->id == objectid) | |
209 | break; | |
210 | ||
211 | lockdep_set_class_and_name(&eb->lock, | |
212 | &ks->keys[level], ks->names[level]); | |
213 | } | |
214 | ||
4008c04a CM |
215 | #endif |
216 | ||
d352ac68 CM |
217 | /* |
218 | * extents on the btree inode are pretty simple, there's one extent | |
219 | * that covers the entire device | |
220 | */ | |
fc4f21b1 | 221 | static struct extent_map *btree_get_extent(struct btrfs_inode *inode, |
306e16ce | 222 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 223 | int create) |
7eccb903 | 224 | { |
fc4f21b1 NB |
225 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb); |
226 | struct extent_map_tree *em_tree = &inode->extent_tree; | |
5f39d397 CM |
227 | struct extent_map *em; |
228 | int ret; | |
229 | ||
890871be | 230 | read_lock(&em_tree->lock); |
d1310b2e | 231 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d | 232 | if (em) { |
0b246afa | 233 | em->bdev = fs_info->fs_devices->latest_bdev; |
890871be | 234 | read_unlock(&em_tree->lock); |
5f39d397 | 235 | goto out; |
a061fc8d | 236 | } |
890871be | 237 | read_unlock(&em_tree->lock); |
7b13b7b1 | 238 | |
172ddd60 | 239 | em = alloc_extent_map(); |
5f39d397 CM |
240 | if (!em) { |
241 | em = ERR_PTR(-ENOMEM); | |
242 | goto out; | |
243 | } | |
244 | em->start = 0; | |
0afbaf8c | 245 | em->len = (u64)-1; |
c8b97818 | 246 | em->block_len = (u64)-1; |
5f39d397 | 247 | em->block_start = 0; |
0b246afa | 248 | em->bdev = fs_info->fs_devices->latest_bdev; |
d1310b2e | 249 | |
890871be | 250 | write_lock(&em_tree->lock); |
09a2a8f9 | 251 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
252 | if (ret == -EEXIST) { |
253 | free_extent_map(em); | |
7b13b7b1 | 254 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 255 | if (!em) |
0433f20d | 256 | em = ERR_PTR(-EIO); |
5f39d397 | 257 | } else if (ret) { |
7b13b7b1 | 258 | free_extent_map(em); |
0433f20d | 259 | em = ERR_PTR(ret); |
5f39d397 | 260 | } |
890871be | 261 | write_unlock(&em_tree->lock); |
7b13b7b1 | 262 | |
5f39d397 CM |
263 | out: |
264 | return em; | |
7eccb903 CM |
265 | } |
266 | ||
9ed57367 | 267 | u32 btrfs_csum_data(const char *data, u32 seed, size_t len) |
19c00ddc | 268 | { |
0b947aff | 269 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
270 | } |
271 | ||
0b5e3daf | 272 | void btrfs_csum_final(u32 crc, u8 *result) |
19c00ddc | 273 | { |
7e75bf3f | 274 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
275 | } |
276 | ||
d352ac68 CM |
277 | /* |
278 | * compute the csum for a btree block, and either verify it or write it | |
279 | * into the csum field of the block. | |
280 | */ | |
01d58472 DD |
281 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
282 | struct extent_buffer *buf, | |
19c00ddc CM |
283 | int verify) |
284 | { | |
01d58472 | 285 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
607d432d | 286 | char *result = NULL; |
19c00ddc CM |
287 | unsigned long len; |
288 | unsigned long cur_len; | |
289 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
290 | char *kaddr; |
291 | unsigned long map_start; | |
292 | unsigned long map_len; | |
293 | int err; | |
294 | u32 crc = ~(u32)0; | |
607d432d | 295 | unsigned long inline_result; |
19c00ddc CM |
296 | |
297 | len = buf->len - offset; | |
d397712b | 298 | while (len > 0) { |
19c00ddc | 299 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 300 | &kaddr, &map_start, &map_len); |
d397712b | 301 | if (err) |
8bd98f0e | 302 | return err; |
19c00ddc | 303 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 304 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
305 | crc, cur_len); |
306 | len -= cur_len; | |
307 | offset += cur_len; | |
19c00ddc | 308 | } |
607d432d | 309 | if (csum_size > sizeof(inline_result)) { |
31e818fe | 310 | result = kzalloc(csum_size, GFP_NOFS); |
607d432d | 311 | if (!result) |
8bd98f0e | 312 | return -ENOMEM; |
607d432d JB |
313 | } else { |
314 | result = (char *)&inline_result; | |
315 | } | |
316 | ||
19c00ddc CM |
317 | btrfs_csum_final(crc, result); |
318 | ||
319 | if (verify) { | |
607d432d | 320 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
321 | u32 val; |
322 | u32 found = 0; | |
607d432d | 323 | memcpy(&found, result, csum_size); |
e4204ded | 324 | |
607d432d | 325 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 | 326 | btrfs_warn_rl(fs_info, |
5d163e0e | 327 | "%s checksum verify failed on %llu wanted %X found %X level %d", |
01d58472 | 328 | fs_info->sb->s_id, buf->start, |
efe120a0 | 329 | val, found, btrfs_header_level(buf)); |
607d432d JB |
330 | if (result != (char *)&inline_result) |
331 | kfree(result); | |
8bd98f0e | 332 | return -EUCLEAN; |
19c00ddc CM |
333 | } |
334 | } else { | |
607d432d | 335 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 336 | } |
607d432d JB |
337 | if (result != (char *)&inline_result) |
338 | kfree(result); | |
19c00ddc CM |
339 | return 0; |
340 | } | |
341 | ||
d352ac68 CM |
342 | /* |
343 | * we can't consider a given block up to date unless the transid of the | |
344 | * block matches the transid in the parent node's pointer. This is how we | |
345 | * detect blocks that either didn't get written at all or got written | |
346 | * in the wrong place. | |
347 | */ | |
1259ab75 | 348 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
349 | struct extent_buffer *eb, u64 parent_transid, |
350 | int atomic) | |
1259ab75 | 351 | { |
2ac55d41 | 352 | struct extent_state *cached_state = NULL; |
1259ab75 | 353 | int ret; |
2755a0de | 354 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
355 | |
356 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
357 | return 0; | |
358 | ||
b9fab919 CM |
359 | if (atomic) |
360 | return -EAGAIN; | |
361 | ||
a26e8c9f JB |
362 | if (need_lock) { |
363 | btrfs_tree_read_lock(eb); | |
364 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
365 | } | |
366 | ||
2ac55d41 | 367 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 368 | &cached_state); |
0b32f4bb | 369 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
370 | btrfs_header_generation(eb) == parent_transid) { |
371 | ret = 0; | |
372 | goto out; | |
373 | } | |
94647322 DS |
374 | btrfs_err_rl(eb->fs_info, |
375 | "parent transid verify failed on %llu wanted %llu found %llu", | |
376 | eb->start, | |
29549aec | 377 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 378 | ret = 1; |
a26e8c9f JB |
379 | |
380 | /* | |
381 | * Things reading via commit roots that don't have normal protection, | |
382 | * like send, can have a really old block in cache that may point at a | |
01327610 | 383 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
384 | * if we find an eb that is under IO (dirty/writeback) because we could |
385 | * end up reading in the stale data and then writing it back out and | |
386 | * making everybody very sad. | |
387 | */ | |
388 | if (!extent_buffer_under_io(eb)) | |
389 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 390 | out: |
2ac55d41 JB |
391 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
392 | &cached_state, GFP_NOFS); | |
472b909f JB |
393 | if (need_lock) |
394 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 395 | return ret; |
1259ab75 CM |
396 | } |
397 | ||
1104a885 DS |
398 | /* |
399 | * Return 0 if the superblock checksum type matches the checksum value of that | |
400 | * algorithm. Pass the raw disk superblock data. | |
401 | */ | |
ab8d0fc4 JM |
402 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
403 | char *raw_disk_sb) | |
1104a885 DS |
404 | { |
405 | struct btrfs_super_block *disk_sb = | |
406 | (struct btrfs_super_block *)raw_disk_sb; | |
407 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
408 | int ret = 0; | |
409 | ||
410 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
411 | u32 crc = ~(u32)0; | |
412 | const int csum_size = sizeof(crc); | |
413 | char result[csum_size]; | |
414 | ||
415 | /* | |
416 | * The super_block structure does not span the whole | |
417 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
01327610 | 418 | * is filled with zeros and is included in the checksum. |
1104a885 DS |
419 | */ |
420 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
421 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
422 | btrfs_csum_final(crc, result); | |
423 | ||
424 | if (memcmp(raw_disk_sb, result, csum_size)) | |
425 | ret = 1; | |
426 | } | |
427 | ||
428 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
ab8d0fc4 | 429 | btrfs_err(fs_info, "unsupported checksum algorithm %u", |
1104a885 DS |
430 | csum_type); |
431 | ret = 1; | |
432 | } | |
433 | ||
434 | return ret; | |
435 | } | |
436 | ||
d352ac68 CM |
437 | /* |
438 | * helper to read a given tree block, doing retries as required when | |
439 | * the checksums don't match and we have alternate mirrors to try. | |
440 | */ | |
2ff7e61e | 441 | static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, |
f188591e | 442 | struct extent_buffer *eb, |
8436ea91 | 443 | u64 parent_transid) |
f188591e CM |
444 | { |
445 | struct extent_io_tree *io_tree; | |
ea466794 | 446 | int failed = 0; |
f188591e CM |
447 | int ret; |
448 | int num_copies = 0; | |
449 | int mirror_num = 0; | |
ea466794 | 450 | int failed_mirror = 0; |
f188591e | 451 | |
a826d6dc | 452 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
0b246afa | 453 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 454 | while (1) { |
8436ea91 | 455 | ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, |
f188591e | 456 | btree_get_extent, mirror_num); |
256dd1bb SB |
457 | if (!ret) { |
458 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 459 | parent_transid, 0)) |
256dd1bb SB |
460 | break; |
461 | else | |
462 | ret = -EIO; | |
463 | } | |
d397712b | 464 | |
a826d6dc JB |
465 | /* |
466 | * This buffer's crc is fine, but its contents are corrupted, so | |
467 | * there is no reason to read the other copies, they won't be | |
468 | * any less wrong. | |
469 | */ | |
470 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
471 | break; |
472 | ||
0b246afa | 473 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 474 | eb->start, eb->len); |
4235298e | 475 | if (num_copies == 1) |
ea466794 | 476 | break; |
4235298e | 477 | |
5cf1ab56 JB |
478 | if (!failed_mirror) { |
479 | failed = 1; | |
480 | failed_mirror = eb->read_mirror; | |
481 | } | |
482 | ||
f188591e | 483 | mirror_num++; |
ea466794 JB |
484 | if (mirror_num == failed_mirror) |
485 | mirror_num++; | |
486 | ||
4235298e | 487 | if (mirror_num > num_copies) |
ea466794 | 488 | break; |
f188591e | 489 | } |
ea466794 | 490 | |
c0901581 | 491 | if (failed && !ret && failed_mirror) |
2ff7e61e | 492 | repair_eb_io_failure(fs_info, eb, failed_mirror); |
ea466794 JB |
493 | |
494 | return ret; | |
f188591e | 495 | } |
19c00ddc | 496 | |
d352ac68 | 497 | /* |
d397712b CM |
498 | * checksum a dirty tree block before IO. This has extra checks to make sure |
499 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 500 | */ |
d397712b | 501 | |
01d58472 | 502 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 503 | { |
4eee4fa4 | 504 | u64 start = page_offset(page); |
19c00ddc | 505 | u64 found_start; |
19c00ddc | 506 | struct extent_buffer *eb; |
f188591e | 507 | |
4f2de97a JB |
508 | eb = (struct extent_buffer *)page->private; |
509 | if (page != eb->pages[0]) | |
510 | return 0; | |
0f805531 | 511 | |
19c00ddc | 512 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
513 | /* |
514 | * Please do not consolidate these warnings into a single if. | |
515 | * It is useful to know what went wrong. | |
516 | */ | |
517 | if (WARN_ON(found_start != start)) | |
518 | return -EUCLEAN; | |
519 | if (WARN_ON(!PageUptodate(page))) | |
520 | return -EUCLEAN; | |
521 | ||
522 | ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, | |
523 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); | |
524 | ||
8bd98f0e | 525 | return csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
526 | } |
527 | ||
01d58472 | 528 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
529 | struct extent_buffer *eb) |
530 | { | |
01d58472 | 531 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
44880fdc | 532 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c YZ |
533 | int ret = 1; |
534 | ||
0a4e5586 | 535 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
536 | while (fs_devices) { |
537 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
538 | ret = 0; | |
539 | break; | |
540 | } | |
541 | fs_devices = fs_devices->seed; | |
542 | } | |
543 | return ret; | |
544 | } | |
545 | ||
0b246afa JM |
546 | #define CORRUPT(reason, eb, root, slot) \ |
547 | btrfs_crit(root->fs_info, \ | |
548 | "corrupt %s, %s: block=%llu, root=%llu, slot=%d", \ | |
549 | btrfs_header_level(eb) == 0 ? "leaf" : "node", \ | |
6b722c17 | 550 | reason, btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
551 | |
552 | static noinline int check_leaf(struct btrfs_root *root, | |
553 | struct extent_buffer *leaf) | |
554 | { | |
0b246afa | 555 | struct btrfs_fs_info *fs_info = root->fs_info; |
a826d6dc JB |
556 | struct btrfs_key key; |
557 | struct btrfs_key leaf_key; | |
558 | u32 nritems = btrfs_header_nritems(leaf); | |
559 | int slot; | |
560 | ||
f177d739 FM |
561 | /* |
562 | * Extent buffers from a relocation tree have a owner field that | |
563 | * corresponds to the subvolume tree they are based on. So just from an | |
564 | * extent buffer alone we can not find out what is the id of the | |
565 | * corresponding subvolume tree, so we can not figure out if the extent | |
566 | * buffer corresponds to the root of the relocation tree or not. So skip | |
567 | * this check for relocation trees. | |
568 | */ | |
569 | if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) { | |
1ba98d08 LB |
570 | struct btrfs_root *check_root; |
571 | ||
572 | key.objectid = btrfs_header_owner(leaf); | |
573 | key.type = BTRFS_ROOT_ITEM_KEY; | |
574 | key.offset = (u64)-1; | |
575 | ||
0b246afa | 576 | check_root = btrfs_get_fs_root(fs_info, &key, false); |
1ba98d08 LB |
577 | /* |
578 | * The only reason we also check NULL here is that during | |
579 | * open_ctree() some roots has not yet been set up. | |
580 | */ | |
581 | if (!IS_ERR_OR_NULL(check_root)) { | |
ef85b25e LB |
582 | struct extent_buffer *eb; |
583 | ||
584 | eb = btrfs_root_node(check_root); | |
1ba98d08 | 585 | /* if leaf is the root, then it's fine */ |
ef85b25e | 586 | if (leaf != eb) { |
1ba98d08 | 587 | CORRUPT("non-root leaf's nritems is 0", |
ef85b25e LB |
588 | leaf, check_root, 0); |
589 | free_extent_buffer(eb); | |
1ba98d08 LB |
590 | return -EIO; |
591 | } | |
ef85b25e | 592 | free_extent_buffer(eb); |
1ba98d08 | 593 | } |
a826d6dc | 594 | return 0; |
1ba98d08 | 595 | } |
a826d6dc | 596 | |
f177d739 FM |
597 | if (nritems == 0) |
598 | return 0; | |
599 | ||
a826d6dc JB |
600 | /* Check the 0 item */ |
601 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
0b246afa | 602 | BTRFS_LEAF_DATA_SIZE(fs_info)) { |
a826d6dc JB |
603 | CORRUPT("invalid item offset size pair", leaf, root, 0); |
604 | return -EIO; | |
605 | } | |
606 | ||
607 | /* | |
608 | * Check to make sure each items keys are in the correct order and their | |
609 | * offsets make sense. We only have to loop through nritems-1 because | |
610 | * we check the current slot against the next slot, which verifies the | |
611 | * next slot's offset+size makes sense and that the current's slot | |
612 | * offset is correct. | |
613 | */ | |
614 | for (slot = 0; slot < nritems - 1; slot++) { | |
615 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
616 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
617 | ||
618 | /* Make sure the keys are in the right order */ | |
619 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
620 | CORRUPT("bad key order", leaf, root, slot); | |
621 | return -EIO; | |
622 | } | |
623 | ||
624 | /* | |
625 | * Make sure the offset and ends are right, remember that the | |
626 | * item data starts at the end of the leaf and grows towards the | |
627 | * front. | |
628 | */ | |
629 | if (btrfs_item_offset_nr(leaf, slot) != | |
630 | btrfs_item_end_nr(leaf, slot + 1)) { | |
631 | CORRUPT("slot offset bad", leaf, root, slot); | |
632 | return -EIO; | |
633 | } | |
634 | ||
635 | /* | |
636 | * Check to make sure that we don't point outside of the leaf, | |
01327610 | 637 | * just in case all the items are consistent to each other, but |
a826d6dc JB |
638 | * all point outside of the leaf. |
639 | */ | |
640 | if (btrfs_item_end_nr(leaf, slot) > | |
0b246afa | 641 | BTRFS_LEAF_DATA_SIZE(fs_info)) { |
a826d6dc JB |
642 | CORRUPT("slot end outside of leaf", leaf, root, slot); |
643 | return -EIO; | |
644 | } | |
645 | } | |
646 | ||
647 | return 0; | |
648 | } | |
649 | ||
053ab70f LB |
650 | static int check_node(struct btrfs_root *root, struct extent_buffer *node) |
651 | { | |
652 | unsigned long nr = btrfs_header_nritems(node); | |
6b722c17 LB |
653 | struct btrfs_key key, next_key; |
654 | int slot; | |
655 | u64 bytenr; | |
656 | int ret = 0; | |
053ab70f | 657 | |
da17066c | 658 | if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root->fs_info)) { |
053ab70f LB |
659 | btrfs_crit(root->fs_info, |
660 | "corrupt node: block %llu root %llu nritems %lu", | |
661 | node->start, root->objectid, nr); | |
662 | return -EIO; | |
663 | } | |
6b722c17 LB |
664 | |
665 | for (slot = 0; slot < nr - 1; slot++) { | |
666 | bytenr = btrfs_node_blockptr(node, slot); | |
667 | btrfs_node_key_to_cpu(node, &key, slot); | |
668 | btrfs_node_key_to_cpu(node, &next_key, slot + 1); | |
669 | ||
670 | if (!bytenr) { | |
671 | CORRUPT("invalid item slot", node, root, slot); | |
672 | ret = -EIO; | |
673 | goto out; | |
674 | } | |
675 | ||
676 | if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) { | |
677 | CORRUPT("bad key order", node, root, slot); | |
678 | ret = -EIO; | |
679 | goto out; | |
680 | } | |
681 | } | |
682 | out: | |
683 | return ret; | |
053ab70f LB |
684 | } |
685 | ||
facc8a22 MX |
686 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
687 | u64 phy_offset, struct page *page, | |
688 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 689 | { |
ce9adaa5 CM |
690 | u64 found_start; |
691 | int found_level; | |
ce9adaa5 CM |
692 | struct extent_buffer *eb; |
693 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 694 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 695 | int ret = 0; |
727011e0 | 696 | int reads_done; |
ce9adaa5 | 697 | |
ce9adaa5 CM |
698 | if (!page->private) |
699 | goto out; | |
d397712b | 700 | |
4f2de97a | 701 | eb = (struct extent_buffer *)page->private; |
d397712b | 702 | |
0b32f4bb JB |
703 | /* the pending IO might have been the only thing that kept this buffer |
704 | * in memory. Make sure we have a ref for all this other checks | |
705 | */ | |
706 | extent_buffer_get(eb); | |
707 | ||
708 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
709 | if (!reads_done) |
710 | goto err; | |
f188591e | 711 | |
5cf1ab56 | 712 | eb->read_mirror = mirror; |
656f30db | 713 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
714 | ret = -EIO; |
715 | goto err; | |
716 | } | |
717 | ||
ce9adaa5 | 718 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 719 | if (found_start != eb->start) { |
02873e43 ZL |
720 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
721 | found_start, eb->start); | |
f188591e | 722 | ret = -EIO; |
ce9adaa5 CM |
723 | goto err; |
724 | } | |
02873e43 ZL |
725 | if (check_tree_block_fsid(fs_info, eb)) { |
726 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
727 | eb->start); | |
1259ab75 CM |
728 | ret = -EIO; |
729 | goto err; | |
730 | } | |
ce9adaa5 | 731 | found_level = btrfs_header_level(eb); |
1c24c3ce | 732 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
733 | btrfs_err(fs_info, "bad tree block level %d", |
734 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
735 | ret = -EIO; |
736 | goto err; | |
737 | } | |
ce9adaa5 | 738 | |
85d4e461 CM |
739 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
740 | eb, found_level); | |
4008c04a | 741 | |
02873e43 | 742 | ret = csum_tree_block(fs_info, eb, 1); |
8bd98f0e | 743 | if (ret) |
a826d6dc | 744 | goto err; |
a826d6dc JB |
745 | |
746 | /* | |
747 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
748 | * that we don't try and read the other copies of this block, just | |
749 | * return -EIO. | |
750 | */ | |
751 | if (found_level == 0 && check_leaf(root, eb)) { | |
752 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
753 | ret = -EIO; | |
754 | } | |
ce9adaa5 | 755 | |
053ab70f LB |
756 | if (found_level > 0 && check_node(root, eb)) |
757 | ret = -EIO; | |
758 | ||
0b32f4bb JB |
759 | if (!ret) |
760 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 761 | err: |
79fb65a1 JB |
762 | if (reads_done && |
763 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 764 | btree_readahead_hook(eb, ret); |
4bb31e92 | 765 | |
53b381b3 DW |
766 | if (ret) { |
767 | /* | |
768 | * our io error hook is going to dec the io pages | |
769 | * again, we have to make sure it has something | |
770 | * to decrement | |
771 | */ | |
772 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 773 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 774 | } |
0b32f4bb | 775 | free_extent_buffer(eb); |
ce9adaa5 | 776 | out: |
f188591e | 777 | return ret; |
ce9adaa5 CM |
778 | } |
779 | ||
ea466794 | 780 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 781 | { |
4bb31e92 | 782 | struct extent_buffer *eb; |
4bb31e92 | 783 | |
4f2de97a | 784 | eb = (struct extent_buffer *)page->private; |
656f30db | 785 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 786 | eb->read_mirror = failed_mirror; |
53b381b3 | 787 | atomic_dec(&eb->io_pages); |
ea466794 | 788 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
d48d71aa | 789 | btree_readahead_hook(eb, -EIO); |
4bb31e92 AJ |
790 | return -EIO; /* we fixed nothing */ |
791 | } | |
792 | ||
4246a0b6 | 793 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 794 | { |
97eb6b69 | 795 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 796 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
797 | struct btrfs_workqueue *wq; |
798 | btrfs_work_func_t func; | |
ce9adaa5 | 799 | |
ce9adaa5 | 800 | fs_info = end_io_wq->info; |
4e4cbee9 | 801 | end_io_wq->status = bio->bi_status; |
d20f7043 | 802 | |
37226b21 | 803 | if (bio_op(bio) == REQ_OP_WRITE) { |
9e0af237 LB |
804 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
805 | wq = fs_info->endio_meta_write_workers; | |
806 | func = btrfs_endio_meta_write_helper; | |
807 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
808 | wq = fs_info->endio_freespace_worker; | |
809 | func = btrfs_freespace_write_helper; | |
810 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
811 | wq = fs_info->endio_raid56_workers; | |
812 | func = btrfs_endio_raid56_helper; | |
813 | } else { | |
814 | wq = fs_info->endio_write_workers; | |
815 | func = btrfs_endio_write_helper; | |
816 | } | |
d20f7043 | 817 | } else { |
8b110e39 MX |
818 | if (unlikely(end_io_wq->metadata == |
819 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
820 | wq = fs_info->endio_repair_workers; | |
821 | func = btrfs_endio_repair_helper; | |
822 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
823 | wq = fs_info->endio_raid56_workers; |
824 | func = btrfs_endio_raid56_helper; | |
825 | } else if (end_io_wq->metadata) { | |
826 | wq = fs_info->endio_meta_workers; | |
827 | func = btrfs_endio_meta_helper; | |
828 | } else { | |
829 | wq = fs_info->endio_workers; | |
830 | func = btrfs_endio_helper; | |
831 | } | |
d20f7043 | 832 | } |
9e0af237 LB |
833 | |
834 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
835 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
836 | } |
837 | ||
4e4cbee9 | 838 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 839 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 840 | { |
97eb6b69 | 841 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 842 | |
97eb6b69 | 843 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 844 | if (!end_io_wq) |
4e4cbee9 | 845 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
846 | |
847 | end_io_wq->private = bio->bi_private; | |
848 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 849 | end_io_wq->info = info; |
4e4cbee9 | 850 | end_io_wq->status = 0; |
ce9adaa5 | 851 | end_io_wq->bio = bio; |
22c59948 | 852 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
853 | |
854 | bio->bi_private = end_io_wq; | |
855 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
856 | return 0; |
857 | } | |
858 | ||
b64a2851 | 859 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 860 | { |
4854ddd0 | 861 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 862 | info->thread_pool_size, |
4854ddd0 CM |
863 | info->fs_devices->open_devices); |
864 | return 256 * limit; | |
865 | } | |
0986fe9e | 866 | |
4a69a410 CM |
867 | static void run_one_async_start(struct btrfs_work *work) |
868 | { | |
4a69a410 | 869 | struct async_submit_bio *async; |
4e4cbee9 | 870 | blk_status_t ret; |
4a69a410 CM |
871 | |
872 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 873 | ret = async->submit_bio_start(async->private_data, async->bio, |
79787eaa JM |
874 | async->mirror_num, async->bio_flags, |
875 | async->bio_offset); | |
876 | if (ret) | |
4e4cbee9 | 877 | async->status = ret; |
4a69a410 CM |
878 | } |
879 | ||
880 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
881 | { |
882 | struct btrfs_fs_info *fs_info; | |
883 | struct async_submit_bio *async; | |
4854ddd0 | 884 | int limit; |
8b712842 CM |
885 | |
886 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 887 | fs_info = async->fs_info; |
4854ddd0 | 888 | |
b64a2851 | 889 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
890 | limit = limit * 2 / 3; |
891 | ||
ee863954 DS |
892 | /* |
893 | * atomic_dec_return implies a barrier for waitqueue_active | |
894 | */ | |
66657b31 | 895 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 896 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
897 | wake_up(&fs_info->async_submit_wait); |
898 | ||
bb7ab3b9 | 899 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
900 | if (async->status) { |
901 | async->bio->bi_status = async->status; | |
4246a0b6 | 902 | bio_endio(async->bio); |
79787eaa JM |
903 | return; |
904 | } | |
905 | ||
c6100a4b | 906 | async->submit_bio_done(async->private_data, async->bio, async->mirror_num, |
81a75f67 | 907 | async->bio_flags, async->bio_offset); |
4a69a410 CM |
908 | } |
909 | ||
910 | static void run_one_async_free(struct btrfs_work *work) | |
911 | { | |
912 | struct async_submit_bio *async; | |
913 | ||
914 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
915 | kfree(async); |
916 | } | |
917 | ||
8c27cb35 LT |
918 | blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, |
919 | int mirror_num, unsigned long bio_flags, | |
920 | u64 bio_offset, void *private_data, | |
921 | extent_submit_bio_hook_t *submit_bio_start, | |
922 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
923 | { |
924 | struct async_submit_bio *async; | |
925 | ||
926 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
927 | if (!async) | |
4e4cbee9 | 928 | return BLK_STS_RESOURCE; |
44b8bd7e | 929 | |
c6100a4b JB |
930 | async->private_data = private_data; |
931 | async->fs_info = fs_info; | |
44b8bd7e CM |
932 | async->bio = bio; |
933 | async->mirror_num = mirror_num; | |
4a69a410 CM |
934 | async->submit_bio_start = submit_bio_start; |
935 | async->submit_bio_done = submit_bio_done; | |
936 | ||
9e0af237 | 937 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 938 | run_one_async_done, run_one_async_free); |
4a69a410 | 939 | |
c8b97818 | 940 | async->bio_flags = bio_flags; |
eaf25d93 | 941 | async->bio_offset = bio_offset; |
8c8bee1d | 942 | |
4e4cbee9 | 943 | async->status = 0; |
79787eaa | 944 | |
cb03c743 | 945 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 946 | |
67f055c7 | 947 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 948 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 949 | |
5cdc7ad3 | 950 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 951 | |
d397712b | 952 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
953 | atomic_read(&fs_info->nr_async_submits)) { |
954 | wait_event(fs_info->async_submit_wait, | |
955 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
956 | } | |
957 | ||
44b8bd7e CM |
958 | return 0; |
959 | } | |
960 | ||
4e4cbee9 | 961 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 962 | { |
2c30c71b | 963 | struct bio_vec *bvec; |
ce3ed71a | 964 | struct btrfs_root *root; |
2c30c71b | 965 | int i, ret = 0; |
ce3ed71a | 966 | |
c09abff8 | 967 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2c30c71b | 968 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 969 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 970 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
971 | if (ret) |
972 | break; | |
ce3ed71a | 973 | } |
2c30c71b | 974 | |
4e4cbee9 | 975 | return errno_to_blk_status(ret); |
ce3ed71a CM |
976 | } |
977 | ||
8c27cb35 LT |
978 | static blk_status_t __btree_submit_bio_start(void *private_data, struct bio *bio, |
979 | int mirror_num, unsigned long bio_flags, | |
980 | u64 bio_offset) | |
22c59948 | 981 | { |
8b712842 CM |
982 | /* |
983 | * when we're called for a write, we're already in the async | |
5443be45 | 984 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 985 | */ |
79787eaa | 986 | return btree_csum_one_bio(bio); |
4a69a410 | 987 | } |
22c59948 | 988 | |
8c27cb35 LT |
989 | static blk_status_t __btree_submit_bio_done(void *private_data, struct bio *bio, |
990 | int mirror_num, unsigned long bio_flags, | |
991 | u64 bio_offset) | |
4a69a410 | 992 | { |
c6100a4b | 993 | struct inode *inode = private_data; |
4e4cbee9 | 994 | blk_status_t ret; |
61891923 | 995 | |
8b712842 | 996 | /* |
4a69a410 CM |
997 | * when we're called for a write, we're already in the async |
998 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 999 | */ |
2ff7e61e | 1000 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1); |
4246a0b6 | 1001 | if (ret) { |
4e4cbee9 | 1002 | bio->bi_status = ret; |
4246a0b6 CH |
1003 | bio_endio(bio); |
1004 | } | |
61891923 | 1005 | return ret; |
0b86a832 CM |
1006 | } |
1007 | ||
e27f6265 | 1008 | static int check_async_write(unsigned long bio_flags) |
de0022b9 JB |
1009 | { |
1010 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
1011 | return 0; | |
1012 | #ifdef CONFIG_X86 | |
bc696ca0 | 1013 | if (static_cpu_has(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
1014 | return 0; |
1015 | #endif | |
1016 | return 1; | |
1017 | } | |
1018 | ||
8c27cb35 LT |
1019 | static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, |
1020 | int mirror_num, unsigned long bio_flags, | |
1021 | u64 bio_offset) | |
44b8bd7e | 1022 | { |
c6100a4b | 1023 | struct inode *inode = private_data; |
0b246afa | 1024 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
e27f6265 | 1025 | int async = check_async_write(bio_flags); |
4e4cbee9 | 1026 | blk_status_t ret; |
cad321ad | 1027 | |
37226b21 | 1028 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
1029 | /* |
1030 | * called for a read, do the setup so that checksum validation | |
1031 | * can happen in the async kernel threads | |
1032 | */ | |
0b246afa JM |
1033 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
1034 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 1035 | if (ret) |
61891923 | 1036 | goto out_w_error; |
2ff7e61e | 1037 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
de0022b9 JB |
1038 | } else if (!async) { |
1039 | ret = btree_csum_one_bio(bio); | |
1040 | if (ret) | |
61891923 | 1041 | goto out_w_error; |
2ff7e61e | 1042 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
61891923 SB |
1043 | } else { |
1044 | /* | |
1045 | * kthread helpers are used to submit writes so that | |
1046 | * checksumming can happen in parallel across all CPUs | |
1047 | */ | |
c6100a4b JB |
1048 | ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, |
1049 | bio_offset, private_data, | |
61891923 SB |
1050 | __btree_submit_bio_start, |
1051 | __btree_submit_bio_done); | |
44b8bd7e | 1052 | } |
d313d7a3 | 1053 | |
4246a0b6 CH |
1054 | if (ret) |
1055 | goto out_w_error; | |
1056 | return 0; | |
1057 | ||
61891923 | 1058 | out_w_error: |
4e4cbee9 | 1059 | bio->bi_status = ret; |
4246a0b6 | 1060 | bio_endio(bio); |
61891923 | 1061 | return ret; |
44b8bd7e CM |
1062 | } |
1063 | ||
3dd1462e | 1064 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1065 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
1066 | struct page *newpage, struct page *page, |
1067 | enum migrate_mode mode) | |
784b4e29 CM |
1068 | { |
1069 | /* | |
1070 | * we can't safely write a btree page from here, | |
1071 | * we haven't done the locking hook | |
1072 | */ | |
1073 | if (PageDirty(page)) | |
1074 | return -EAGAIN; | |
1075 | /* | |
1076 | * Buffers may be managed in a filesystem specific way. | |
1077 | * We must have no buffers or drop them. | |
1078 | */ | |
1079 | if (page_has_private(page) && | |
1080 | !try_to_release_page(page, GFP_KERNEL)) | |
1081 | return -EAGAIN; | |
a6bc32b8 | 1082 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 1083 | } |
3dd1462e | 1084 | #endif |
784b4e29 | 1085 | |
0da5468f CM |
1086 | |
1087 | static int btree_writepages(struct address_space *mapping, | |
1088 | struct writeback_control *wbc) | |
1089 | { | |
e2d84521 MX |
1090 | struct btrfs_fs_info *fs_info; |
1091 | int ret; | |
1092 | ||
d8d5f3e1 | 1093 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1094 | |
1095 | if (wbc->for_kupdate) | |
1096 | return 0; | |
1097 | ||
e2d84521 | 1098 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1099 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
1100 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1101 | BTRFS_DIRTY_METADATA_THRESH); | |
1102 | if (ret < 0) | |
793955bc | 1103 | return 0; |
793955bc | 1104 | } |
0b32f4bb | 1105 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1106 | } |
1107 | ||
b2950863 | 1108 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 1109 | { |
d1310b2e CM |
1110 | struct extent_io_tree *tree; |
1111 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 1112 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 1113 | } |
22b0ebda | 1114 | |
70dec807 | 1115 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1116 | { |
98509cfc | 1117 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1118 | return 0; |
0c4e538b | 1119 | |
f7a52a40 | 1120 | return try_release_extent_buffer(page); |
d98237b3 CM |
1121 | } |
1122 | ||
d47992f8 LC |
1123 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1124 | unsigned int length) | |
d98237b3 | 1125 | { |
d1310b2e CM |
1126 | struct extent_io_tree *tree; |
1127 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1128 | extent_invalidatepage(tree, page, offset); |
1129 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1130 | if (PagePrivate(page)) { |
efe120a0 FH |
1131 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1132 | "page private not zero on page %llu", | |
1133 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1134 | ClearPagePrivate(page); |
1135 | set_page_private(page, 0); | |
09cbfeaf | 1136 | put_page(page); |
9ad6b7bc | 1137 | } |
d98237b3 CM |
1138 | } |
1139 | ||
0b32f4bb JB |
1140 | static int btree_set_page_dirty(struct page *page) |
1141 | { | |
bb146eb2 | 1142 | #ifdef DEBUG |
0b32f4bb JB |
1143 | struct extent_buffer *eb; |
1144 | ||
1145 | BUG_ON(!PagePrivate(page)); | |
1146 | eb = (struct extent_buffer *)page->private; | |
1147 | BUG_ON(!eb); | |
1148 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1149 | BUG_ON(!atomic_read(&eb->refs)); | |
1150 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1151 | #endif |
0b32f4bb JB |
1152 | return __set_page_dirty_nobuffers(page); |
1153 | } | |
1154 | ||
7f09410b | 1155 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1156 | .readpage = btree_readpage, |
0da5468f | 1157 | .writepages = btree_writepages, |
5f39d397 CM |
1158 | .releasepage = btree_releasepage, |
1159 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1160 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1161 | .migratepage = btree_migratepage, |
5a92bc88 | 1162 | #endif |
0b32f4bb | 1163 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1164 | }; |
1165 | ||
2ff7e61e | 1166 | void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) |
090d1875 | 1167 | { |
5f39d397 | 1168 | struct extent_buffer *buf = NULL; |
2ff7e61e | 1169 | struct inode *btree_inode = fs_info->btree_inode; |
090d1875 | 1170 | |
2ff7e61e | 1171 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1172 | if (IS_ERR(buf)) |
6197d86e | 1173 | return; |
d1310b2e | 1174 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
8436ea91 | 1175 | buf, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1176 | free_extent_buffer(buf); |
090d1875 CM |
1177 | } |
1178 | ||
2ff7e61e | 1179 | int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, |
ab0fff03 AJ |
1180 | int mirror_num, struct extent_buffer **eb) |
1181 | { | |
1182 | struct extent_buffer *buf = NULL; | |
2ff7e61e | 1183 | struct inode *btree_inode = fs_info->btree_inode; |
ab0fff03 AJ |
1184 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; |
1185 | int ret; | |
1186 | ||
2ff7e61e | 1187 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1188 | if (IS_ERR(buf)) |
ab0fff03 AJ |
1189 | return 0; |
1190 | ||
1191 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1192 | ||
8436ea91 | 1193 | ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, |
ab0fff03 AJ |
1194 | btree_get_extent, mirror_num); |
1195 | if (ret) { | |
1196 | free_extent_buffer(buf); | |
1197 | return ret; | |
1198 | } | |
1199 | ||
1200 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1201 | free_extent_buffer(buf); | |
1202 | return -EIO; | |
0b32f4bb | 1203 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1204 | *eb = buf; |
1205 | } else { | |
1206 | free_extent_buffer(buf); | |
1207 | } | |
1208 | return 0; | |
1209 | } | |
1210 | ||
2ff7e61e JM |
1211 | struct extent_buffer *btrfs_find_create_tree_block( |
1212 | struct btrfs_fs_info *fs_info, | |
1213 | u64 bytenr) | |
0999df54 | 1214 | { |
0b246afa JM |
1215 | if (btrfs_is_testing(fs_info)) |
1216 | return alloc_test_extent_buffer(fs_info, bytenr); | |
1217 | return alloc_extent_buffer(fs_info, bytenr); | |
0999df54 CM |
1218 | } |
1219 | ||
1220 | ||
e02119d5 CM |
1221 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1222 | { | |
727011e0 | 1223 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1224 | buf->start + buf->len - 1); |
e02119d5 CM |
1225 | } |
1226 | ||
3189ff77 | 1227 | void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) |
e02119d5 | 1228 | { |
3189ff77 JL |
1229 | filemap_fdatawait_range(buf->pages[0]->mapping, |
1230 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
1231 | } |
1232 | ||
2ff7e61e | 1233 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
ce86cd59 | 1234 | u64 parent_transid) |
0999df54 CM |
1235 | { |
1236 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1237 | int ret; |
1238 | ||
2ff7e61e | 1239 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 LB |
1240 | if (IS_ERR(buf)) |
1241 | return buf; | |
0999df54 | 1242 | |
2ff7e61e | 1243 | ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid); |
0f0fe8f7 FDBM |
1244 | if (ret) { |
1245 | free_extent_buffer(buf); | |
64c043de | 1246 | return ERR_PTR(ret); |
0f0fe8f7 | 1247 | } |
5f39d397 | 1248 | return buf; |
ce9adaa5 | 1249 | |
eb60ceac CM |
1250 | } |
1251 | ||
7c302b49 | 1252 | void clean_tree_block(struct btrfs_fs_info *fs_info, |
d5c13f92 | 1253 | struct extent_buffer *buf) |
ed2ff2cb | 1254 | { |
55c69072 | 1255 | if (btrfs_header_generation(buf) == |
e2d84521 | 1256 | fs_info->running_transaction->transid) { |
b9447ef8 | 1257 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1258 | |
b9473439 | 1259 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1260 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1261 | -buf->len, | |
1262 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1263 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1264 | btrfs_set_lock_blocking(buf); | |
1265 | clear_extent_buffer_dirty(buf); | |
1266 | } | |
925baedd | 1267 | } |
5f39d397 CM |
1268 | } |
1269 | ||
8257b2dc MX |
1270 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1271 | { | |
1272 | struct btrfs_subvolume_writers *writers; | |
1273 | int ret; | |
1274 | ||
1275 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1276 | if (!writers) | |
1277 | return ERR_PTR(-ENOMEM); | |
1278 | ||
908c7f19 | 1279 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1280 | if (ret < 0) { |
1281 | kfree(writers); | |
1282 | return ERR_PTR(ret); | |
1283 | } | |
1284 | ||
1285 | init_waitqueue_head(&writers->wait); | |
1286 | return writers; | |
1287 | } | |
1288 | ||
1289 | static void | |
1290 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1291 | { | |
1292 | percpu_counter_destroy(&writers->counter); | |
1293 | kfree(writers); | |
1294 | } | |
1295 | ||
da17066c | 1296 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1297 | u64 objectid) |
d97e63b6 | 1298 | { |
7c0260ee | 1299 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
cfaa7295 | 1300 | root->node = NULL; |
a28ec197 | 1301 | root->commit_root = NULL; |
27cdeb70 | 1302 | root->state = 0; |
d68fc57b | 1303 | root->orphan_cleanup_state = 0; |
0b86a832 | 1304 | |
0f7d52f4 CM |
1305 | root->objectid = objectid; |
1306 | root->last_trans = 0; | |
13a8a7c8 | 1307 | root->highest_objectid = 0; |
eb73c1b7 | 1308 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1309 | root->nr_ordered_extents = 0; |
58176a96 | 1310 | root->name = NULL; |
6bef4d31 | 1311 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1312 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1313 | root->block_rsv = NULL; |
d68fc57b | 1314 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1315 | |
1316 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1317 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1318 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1319 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1320 | INIT_LIST_HEAD(&root->ordered_extents); |
1321 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1322 | INIT_LIST_HEAD(&root->logged_list[0]); |
1323 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1324 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1325 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1326 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1327 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1328 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1329 | spin_lock_init(&root->log_extents_lock[0]); |
1330 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1331 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1332 | mutex_init(&root->log_mutex); |
31f3d255 | 1333 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1334 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1335 | init_waitqueue_head(&root->log_writer_wait); |
1336 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1337 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1338 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1339 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1340 | atomic_set(&root->log_commit[0], 0); |
1341 | atomic_set(&root->log_commit[1], 0); | |
1342 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1343 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1344 | atomic_set(&root->orphan_inodes, 0); |
0700cea7 | 1345 | refcount_set(&root->refs, 1); |
ea14b57f | 1346 | atomic_set(&root->will_be_snapshotted, 0); |
ce0dcee6 | 1347 | atomic64_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1348 | root->log_transid = 0; |
d1433deb | 1349 | root->log_transid_committed = -1; |
257c62e1 | 1350 | root->last_log_commit = 0; |
7c0260ee | 1351 | if (!dummy) |
c6100a4b | 1352 | extent_io_tree_init(&root->dirty_log_pages, NULL); |
017e5369 | 1353 | |
3768f368 CM |
1354 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1355 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1356 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1357 | if (!dummy) |
06ea65a3 JB |
1358 | root->defrag_trans_start = fs_info->generation; |
1359 | else | |
1360 | root->defrag_trans_start = 0; | |
4d775673 | 1361 | root->root_key.objectid = objectid; |
0ee5dc67 | 1362 | root->anon_dev = 0; |
8ea05e3a | 1363 | |
5f3ab90a | 1364 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1365 | } |
1366 | ||
74e4d827 DS |
1367 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1368 | gfp_t flags) | |
6f07e42e | 1369 | { |
74e4d827 | 1370 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1371 | if (root) |
1372 | root->fs_info = fs_info; | |
1373 | return root; | |
1374 | } | |
1375 | ||
06ea65a3 JB |
1376 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1377 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1378 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1379 | { |
1380 | struct btrfs_root *root; | |
1381 | ||
7c0260ee JM |
1382 | if (!fs_info) |
1383 | return ERR_PTR(-EINVAL); | |
1384 | ||
1385 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
06ea65a3 JB |
1386 | if (!root) |
1387 | return ERR_PTR(-ENOMEM); | |
da17066c | 1388 | |
b9ef22de | 1389 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
da17066c | 1390 | __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
faa2dbf0 | 1391 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1392 | |
1393 | return root; | |
1394 | } | |
1395 | #endif | |
1396 | ||
20897f5c AJ |
1397 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1398 | struct btrfs_fs_info *fs_info, | |
1399 | u64 objectid) | |
1400 | { | |
1401 | struct extent_buffer *leaf; | |
1402 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1403 | struct btrfs_root *root; | |
1404 | struct btrfs_key key; | |
1405 | int ret = 0; | |
6463fe58 | 1406 | uuid_le uuid; |
20897f5c | 1407 | |
74e4d827 | 1408 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1409 | if (!root) |
1410 | return ERR_PTR(-ENOMEM); | |
1411 | ||
da17066c | 1412 | __setup_root(root, fs_info, objectid); |
20897f5c AJ |
1413 | root->root_key.objectid = objectid; |
1414 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1415 | root->root_key.offset = 0; | |
1416 | ||
4d75f8a9 | 1417 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1418 | if (IS_ERR(leaf)) { |
1419 | ret = PTR_ERR(leaf); | |
1dd05682 | 1420 | leaf = NULL; |
20897f5c AJ |
1421 | goto fail; |
1422 | } | |
1423 | ||
b159fa28 | 1424 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
20897f5c AJ |
1425 | btrfs_set_header_bytenr(leaf, leaf->start); |
1426 | btrfs_set_header_generation(leaf, trans->transid); | |
1427 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1428 | btrfs_set_header_owner(leaf, objectid); | |
1429 | root->node = leaf; | |
1430 | ||
d24ee97b DS |
1431 | write_extent_buffer_fsid(leaf, fs_info->fsid); |
1432 | write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); | |
20897f5c AJ |
1433 | btrfs_mark_buffer_dirty(leaf); |
1434 | ||
1435 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1436 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1437 | |
1438 | root->root_item.flags = 0; | |
1439 | root->root_item.byte_limit = 0; | |
1440 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1441 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1442 | btrfs_set_root_level(&root->root_item, 0); | |
1443 | btrfs_set_root_refs(&root->root_item, 1); | |
1444 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1445 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1446 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1447 | uuid_le_gen(&uuid); |
1448 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1449 | root->root_item.drop_level = 0; |
1450 | ||
1451 | key.objectid = objectid; | |
1452 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1453 | key.offset = 0; | |
1454 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1455 | if (ret) | |
1456 | goto fail; | |
1457 | ||
1458 | btrfs_tree_unlock(leaf); | |
1459 | ||
1dd05682 TI |
1460 | return root; |
1461 | ||
20897f5c | 1462 | fail: |
1dd05682 TI |
1463 | if (leaf) { |
1464 | btrfs_tree_unlock(leaf); | |
59885b39 | 1465 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1466 | free_extent_buffer(leaf); |
1467 | } | |
1468 | kfree(root); | |
20897f5c | 1469 | |
1dd05682 | 1470 | return ERR_PTR(ret); |
20897f5c AJ |
1471 | } |
1472 | ||
7237f183 YZ |
1473 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1474 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1475 | { |
1476 | struct btrfs_root *root; | |
7237f183 | 1477 | struct extent_buffer *leaf; |
e02119d5 | 1478 | |
74e4d827 | 1479 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1480 | if (!root) |
7237f183 | 1481 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1482 | |
da17066c | 1483 | __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
e02119d5 CM |
1484 | |
1485 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1486 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1487 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1488 | |
7237f183 | 1489 | /* |
27cdeb70 MX |
1490 | * DON'T set REF_COWS for log trees |
1491 | * | |
7237f183 YZ |
1492 | * log trees do not get reference counted because they go away |
1493 | * before a real commit is actually done. They do store pointers | |
1494 | * to file data extents, and those reference counts still get | |
1495 | * updated (along with back refs to the log tree). | |
1496 | */ | |
e02119d5 | 1497 | |
4d75f8a9 DS |
1498 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1499 | NULL, 0, 0, 0); | |
7237f183 YZ |
1500 | if (IS_ERR(leaf)) { |
1501 | kfree(root); | |
1502 | return ERR_CAST(leaf); | |
1503 | } | |
e02119d5 | 1504 | |
b159fa28 | 1505 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
5d4f98a2 YZ |
1506 | btrfs_set_header_bytenr(leaf, leaf->start); |
1507 | btrfs_set_header_generation(leaf, trans->transid); | |
1508 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1509 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1510 | root->node = leaf; |
e02119d5 | 1511 | |
0b246afa | 1512 | write_extent_buffer_fsid(root->node, fs_info->fsid); |
e02119d5 CM |
1513 | btrfs_mark_buffer_dirty(root->node); |
1514 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1515 | return root; |
1516 | } | |
1517 | ||
1518 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1519 | struct btrfs_fs_info *fs_info) | |
1520 | { | |
1521 | struct btrfs_root *log_root; | |
1522 | ||
1523 | log_root = alloc_log_tree(trans, fs_info); | |
1524 | if (IS_ERR(log_root)) | |
1525 | return PTR_ERR(log_root); | |
1526 | WARN_ON(fs_info->log_root_tree); | |
1527 | fs_info->log_root_tree = log_root; | |
1528 | return 0; | |
1529 | } | |
1530 | ||
1531 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1532 | struct btrfs_root *root) | |
1533 | { | |
0b246afa | 1534 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1535 | struct btrfs_root *log_root; |
1536 | struct btrfs_inode_item *inode_item; | |
1537 | ||
0b246afa | 1538 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1539 | if (IS_ERR(log_root)) |
1540 | return PTR_ERR(log_root); | |
1541 | ||
1542 | log_root->last_trans = trans->transid; | |
1543 | log_root->root_key.offset = root->root_key.objectid; | |
1544 | ||
1545 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1546 | btrfs_set_stack_inode_generation(inode_item, 1); |
1547 | btrfs_set_stack_inode_size(inode_item, 3); | |
1548 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1549 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1550 | fs_info->nodesize); |
3cae210f | 1551 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1552 | |
5d4f98a2 | 1553 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1554 | |
1555 | WARN_ON(root->log_root); | |
1556 | root->log_root = log_root; | |
1557 | root->log_transid = 0; | |
d1433deb | 1558 | root->log_transid_committed = -1; |
257c62e1 | 1559 | root->last_log_commit = 0; |
e02119d5 CM |
1560 | return 0; |
1561 | } | |
1562 | ||
35a3621b SB |
1563 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1564 | struct btrfs_key *key) | |
e02119d5 CM |
1565 | { |
1566 | struct btrfs_root *root; | |
1567 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1568 | struct btrfs_path *path; |
84234f3a | 1569 | u64 generation; |
cb517eab | 1570 | int ret; |
0f7d52f4 | 1571 | |
cb517eab MX |
1572 | path = btrfs_alloc_path(); |
1573 | if (!path) | |
0f7d52f4 | 1574 | return ERR_PTR(-ENOMEM); |
cb517eab | 1575 | |
74e4d827 | 1576 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1577 | if (!root) { |
1578 | ret = -ENOMEM; | |
1579 | goto alloc_fail; | |
0f7d52f4 CM |
1580 | } |
1581 | ||
da17066c | 1582 | __setup_root(root, fs_info, key->objectid); |
0f7d52f4 | 1583 | |
cb517eab MX |
1584 | ret = btrfs_find_root(tree_root, key, path, |
1585 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1586 | if (ret) { |
13a8a7c8 YZ |
1587 | if (ret > 0) |
1588 | ret = -ENOENT; | |
cb517eab | 1589 | goto find_fail; |
0f7d52f4 | 1590 | } |
13a8a7c8 | 1591 | |
84234f3a | 1592 | generation = btrfs_root_generation(&root->root_item); |
2ff7e61e JM |
1593 | root->node = read_tree_block(fs_info, |
1594 | btrfs_root_bytenr(&root->root_item), | |
ce86cd59 | 1595 | generation); |
64c043de LB |
1596 | if (IS_ERR(root->node)) { |
1597 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1598 | goto find_fail; |
1599 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1600 | ret = -EIO; | |
64c043de LB |
1601 | free_extent_buffer(root->node); |
1602 | goto find_fail; | |
416bc658 | 1603 | } |
5d4f98a2 | 1604 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1605 | out: |
cb517eab MX |
1606 | btrfs_free_path(path); |
1607 | return root; | |
1608 | ||
cb517eab MX |
1609 | find_fail: |
1610 | kfree(root); | |
1611 | alloc_fail: | |
1612 | root = ERR_PTR(ret); | |
1613 | goto out; | |
1614 | } | |
1615 | ||
1616 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1617 | struct btrfs_key *location) | |
1618 | { | |
1619 | struct btrfs_root *root; | |
1620 | ||
1621 | root = btrfs_read_tree_root(tree_root, location); | |
1622 | if (IS_ERR(root)) | |
1623 | return root; | |
1624 | ||
1625 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1626 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1627 | btrfs_check_and_init_root_item(&root->root_item); |
1628 | } | |
13a8a7c8 | 1629 | |
5eda7b5e CM |
1630 | return root; |
1631 | } | |
1632 | ||
cb517eab MX |
1633 | int btrfs_init_fs_root(struct btrfs_root *root) |
1634 | { | |
1635 | int ret; | |
8257b2dc | 1636 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1637 | |
1638 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1639 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1640 | GFP_NOFS); | |
1641 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1642 | ret = -ENOMEM; | |
1643 | goto fail; | |
1644 | } | |
1645 | ||
8257b2dc MX |
1646 | writers = btrfs_alloc_subvolume_writers(); |
1647 | if (IS_ERR(writers)) { | |
1648 | ret = PTR_ERR(writers); | |
1649 | goto fail; | |
1650 | } | |
1651 | root->subv_writers = writers; | |
1652 | ||
cb517eab | 1653 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1654 | spin_lock_init(&root->ino_cache_lock); |
1655 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1656 | |
1657 | ret = get_anon_bdev(&root->anon_dev); | |
1658 | if (ret) | |
876d2cf1 | 1659 | goto fail; |
f32e48e9 CR |
1660 | |
1661 | mutex_lock(&root->objectid_mutex); | |
1662 | ret = btrfs_find_highest_objectid(root, | |
1663 | &root->highest_objectid); | |
1664 | if (ret) { | |
1665 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1666 | goto fail; |
f32e48e9 CR |
1667 | } |
1668 | ||
1669 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1670 | ||
1671 | mutex_unlock(&root->objectid_mutex); | |
1672 | ||
cb517eab MX |
1673 | return 0; |
1674 | fail: | |
876d2cf1 | 1675 | /* the caller is responsible to call free_fs_root */ |
cb517eab MX |
1676 | return ret; |
1677 | } | |
1678 | ||
35bbb97f JM |
1679 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1680 | u64 root_id) | |
cb517eab MX |
1681 | { |
1682 | struct btrfs_root *root; | |
1683 | ||
1684 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1685 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1686 | (unsigned long)root_id); | |
1687 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1688 | return root; | |
1689 | } | |
1690 | ||
1691 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1692 | struct btrfs_root *root) | |
1693 | { | |
1694 | int ret; | |
1695 | ||
e1860a77 | 1696 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1697 | if (ret) |
1698 | return ret; | |
1699 | ||
1700 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1701 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1702 | (unsigned long)root->root_key.objectid, | |
1703 | root); | |
1704 | if (ret == 0) | |
27cdeb70 | 1705 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1706 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1707 | radix_tree_preload_end(); | |
1708 | ||
1709 | return ret; | |
1710 | } | |
1711 | ||
c00869f1 MX |
1712 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1713 | struct btrfs_key *location, | |
1714 | bool check_ref) | |
5eda7b5e CM |
1715 | { |
1716 | struct btrfs_root *root; | |
381cf658 | 1717 | struct btrfs_path *path; |
1d4c08e0 | 1718 | struct btrfs_key key; |
5eda7b5e CM |
1719 | int ret; |
1720 | ||
edbd8d4e CM |
1721 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1722 | return fs_info->tree_root; | |
1723 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1724 | return fs_info->extent_root; | |
8f18cf13 CM |
1725 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1726 | return fs_info->chunk_root; | |
1727 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1728 | return fs_info->dev_root; | |
0403e47e YZ |
1729 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1730 | return fs_info->csum_root; | |
bcef60f2 AJ |
1731 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1732 | return fs_info->quota_root ? fs_info->quota_root : | |
1733 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1734 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1735 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1736 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1737 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1738 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1739 | ERR_PTR(-ENOENT); | |
4df27c4d | 1740 | again: |
cb517eab | 1741 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1742 | if (root) { |
c00869f1 | 1743 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1744 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1745 | return root; |
48475471 | 1746 | } |
5eda7b5e | 1747 | |
cb517eab | 1748 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1749 | if (IS_ERR(root)) |
1750 | return root; | |
3394e160 | 1751 | |
c00869f1 | 1752 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1753 | ret = -ENOENT; |
581bb050 | 1754 | goto fail; |
35a30d7c | 1755 | } |
581bb050 | 1756 | |
cb517eab | 1757 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1758 | if (ret) |
1759 | goto fail; | |
3394e160 | 1760 | |
381cf658 DS |
1761 | path = btrfs_alloc_path(); |
1762 | if (!path) { | |
1763 | ret = -ENOMEM; | |
1764 | goto fail; | |
1765 | } | |
1d4c08e0 DS |
1766 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1767 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1768 | key.offset = location->objectid; | |
1769 | ||
1770 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1771 | btrfs_free_path(path); |
d68fc57b YZ |
1772 | if (ret < 0) |
1773 | goto fail; | |
1774 | if (ret == 0) | |
27cdeb70 | 1775 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1776 | |
cb517eab | 1777 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1778 | if (ret) { |
4df27c4d YZ |
1779 | if (ret == -EEXIST) { |
1780 | free_fs_root(root); | |
1781 | goto again; | |
1782 | } | |
1783 | goto fail; | |
0f7d52f4 | 1784 | } |
edbd8d4e | 1785 | return root; |
4df27c4d YZ |
1786 | fail: |
1787 | free_fs_root(root); | |
1788 | return ERR_PTR(ret); | |
edbd8d4e CM |
1789 | } |
1790 | ||
04160088 CM |
1791 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1792 | { | |
1793 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1794 | int ret = 0; | |
04160088 CM |
1795 | struct btrfs_device *device; |
1796 | struct backing_dev_info *bdi; | |
b7967db7 | 1797 | |
1f78160c XG |
1798 | rcu_read_lock(); |
1799 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1800 | if (!device->bdev) |
1801 | continue; | |
efa7c9f9 | 1802 | bdi = device->bdev->bd_bdi; |
ff9ea323 | 1803 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1804 | ret = 1; |
1805 | break; | |
1806 | } | |
1807 | } | |
1f78160c | 1808 | rcu_read_unlock(); |
04160088 CM |
1809 | return ret; |
1810 | } | |
1811 | ||
8b712842 CM |
1812 | /* |
1813 | * called by the kthread helper functions to finally call the bio end_io | |
1814 | * functions. This is where read checksum verification actually happens | |
1815 | */ | |
1816 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1817 | { |
ce9adaa5 | 1818 | struct bio *bio; |
97eb6b69 | 1819 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1820 | |
97eb6b69 | 1821 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1822 | bio = end_io_wq->bio; |
ce9adaa5 | 1823 | |
4e4cbee9 | 1824 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1825 | bio->bi_private = end_io_wq->private; |
1826 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1827 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1828 | bio_endio(bio); |
44b8bd7e CM |
1829 | } |
1830 | ||
a74a4b97 CM |
1831 | static int cleaner_kthread(void *arg) |
1832 | { | |
1833 | struct btrfs_root *root = arg; | |
0b246afa | 1834 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1835 | int again; |
da288d28 | 1836 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1837 | |
1838 | do { | |
d0278245 | 1839 | again = 0; |
a74a4b97 | 1840 | |
d0278245 | 1841 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1842 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1843 | goto sleep; |
1844 | ||
90c711ab ZB |
1845 | /* |
1846 | * Do not do anything if we might cause open_ctree() to block | |
1847 | * before we have finished mounting the filesystem. | |
1848 | */ | |
0b246afa | 1849 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1850 | goto sleep; |
1851 | ||
0b246afa | 1852 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1853 | goto sleep; |
1854 | ||
dc7f370c MX |
1855 | /* |
1856 | * Avoid the problem that we change the status of the fs | |
1857 | * during the above check and trylock. | |
1858 | */ | |
2ff7e61e | 1859 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1860 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1861 | goto sleep; |
76dda93c | 1862 | } |
a74a4b97 | 1863 | |
0b246afa | 1864 | mutex_lock(&fs_info->cleaner_delayed_iput_mutex); |
2ff7e61e | 1865 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 1866 | mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); |
c2d6cb16 | 1867 | |
d0278245 | 1868 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1869 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1870 | |
1871 | /* | |
05323cd1 MX |
1872 | * The defragger has dealt with the R/O remount and umount, |
1873 | * needn't do anything special here. | |
d0278245 | 1874 | */ |
0b246afa | 1875 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1876 | |
1877 | /* | |
1878 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1879 | * with relocation (btrfs_relocate_chunk) and relocation | |
1880 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1881 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1882 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1883 | * unused block groups. | |
1884 | */ | |
0b246afa | 1885 | btrfs_delete_unused_bgs(fs_info); |
d0278245 | 1886 | sleep: |
838fe188 | 1887 | if (!again) { |
a74a4b97 | 1888 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1889 | if (!kthread_should_stop()) |
1890 | schedule(); | |
a74a4b97 CM |
1891 | __set_current_state(TASK_RUNNING); |
1892 | } | |
1893 | } while (!kthread_should_stop()); | |
da288d28 FM |
1894 | |
1895 | /* | |
1896 | * Transaction kthread is stopped before us and wakes us up. | |
1897 | * However we might have started a new transaction and COWed some | |
1898 | * tree blocks when deleting unused block groups for example. So | |
1899 | * make sure we commit the transaction we started to have a clean | |
1900 | * shutdown when evicting the btree inode - if it has dirty pages | |
1901 | * when we do the final iput() on it, eviction will trigger a | |
1902 | * writeback for it which will fail with null pointer dereferences | |
1903 | * since work queues and other resources were already released and | |
1904 | * destroyed by the time the iput/eviction/writeback is made. | |
1905 | */ | |
1906 | trans = btrfs_attach_transaction(root); | |
1907 | if (IS_ERR(trans)) { | |
1908 | if (PTR_ERR(trans) != -ENOENT) | |
0b246afa | 1909 | btrfs_err(fs_info, |
da288d28 FM |
1910 | "cleaner transaction attach returned %ld", |
1911 | PTR_ERR(trans)); | |
1912 | } else { | |
1913 | int ret; | |
1914 | ||
3a45bb20 | 1915 | ret = btrfs_commit_transaction(trans); |
da288d28 | 1916 | if (ret) |
0b246afa | 1917 | btrfs_err(fs_info, |
da288d28 FM |
1918 | "cleaner open transaction commit returned %d", |
1919 | ret); | |
1920 | } | |
1921 | ||
a74a4b97 CM |
1922 | return 0; |
1923 | } | |
1924 | ||
1925 | static int transaction_kthread(void *arg) | |
1926 | { | |
1927 | struct btrfs_root *root = arg; | |
0b246afa | 1928 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1929 | struct btrfs_trans_handle *trans; |
1930 | struct btrfs_transaction *cur; | |
8929ecfa | 1931 | u64 transid; |
a74a4b97 CM |
1932 | unsigned long now; |
1933 | unsigned long delay; | |
914b2007 | 1934 | bool cannot_commit; |
a74a4b97 CM |
1935 | |
1936 | do { | |
914b2007 | 1937 | cannot_commit = false; |
0b246afa JM |
1938 | delay = HZ * fs_info->commit_interval; |
1939 | mutex_lock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1940 | |
0b246afa JM |
1941 | spin_lock(&fs_info->trans_lock); |
1942 | cur = fs_info->running_transaction; | |
a74a4b97 | 1943 | if (!cur) { |
0b246afa | 1944 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1945 | goto sleep; |
1946 | } | |
31153d81 | 1947 | |
a74a4b97 | 1948 | now = get_seconds(); |
4a9d8bde | 1949 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 | 1950 | (now < cur->start_time || |
0b246afa JM |
1951 | now - cur->start_time < fs_info->commit_interval)) { |
1952 | spin_unlock(&fs_info->trans_lock); | |
a74a4b97 CM |
1953 | delay = HZ * 5; |
1954 | goto sleep; | |
1955 | } | |
8929ecfa | 1956 | transid = cur->transid; |
0b246afa | 1957 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1958 | |
79787eaa | 1959 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1960 | trans = btrfs_attach_transaction(root); |
914b2007 | 1961 | if (IS_ERR(trans)) { |
354aa0fb MX |
1962 | if (PTR_ERR(trans) != -ENOENT) |
1963 | cannot_commit = true; | |
79787eaa | 1964 | goto sleep; |
914b2007 | 1965 | } |
8929ecfa | 1966 | if (transid == trans->transid) { |
3a45bb20 | 1967 | btrfs_commit_transaction(trans); |
8929ecfa | 1968 | } else { |
3a45bb20 | 1969 | btrfs_end_transaction(trans); |
8929ecfa | 1970 | } |
a74a4b97 | 1971 | sleep: |
0b246afa JM |
1972 | wake_up_process(fs_info->cleaner_kthread); |
1973 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1974 | |
4e121c06 | 1975 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1976 | &fs_info->fs_state))) |
2ff7e61e | 1977 | btrfs_cleanup_transaction(fs_info); |
ce63f891 JK |
1978 | set_current_state(TASK_INTERRUPTIBLE); |
1979 | if (!kthread_should_stop() && | |
0b246afa | 1980 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 JK |
1981 | cannot_commit)) |
1982 | schedule_timeout(delay); | |
1983 | __set_current_state(TASK_RUNNING); | |
a74a4b97 CM |
1984 | } while (!kthread_should_stop()); |
1985 | return 0; | |
1986 | } | |
1987 | ||
af31f5e5 CM |
1988 | /* |
1989 | * this will find the highest generation in the array of | |
1990 | * root backups. The index of the highest array is returned, | |
1991 | * or -1 if we can't find anything. | |
1992 | * | |
1993 | * We check to make sure the array is valid by comparing the | |
1994 | * generation of the latest root in the array with the generation | |
1995 | * in the super block. If they don't match we pitch it. | |
1996 | */ | |
1997 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1998 | { | |
1999 | u64 cur; | |
2000 | int newest_index = -1; | |
2001 | struct btrfs_root_backup *root_backup; | |
2002 | int i; | |
2003 | ||
2004 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
2005 | root_backup = info->super_copy->super_roots + i; | |
2006 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2007 | if (cur == newest_gen) | |
2008 | newest_index = i; | |
2009 | } | |
2010 | ||
2011 | /* check to see if we actually wrapped around */ | |
2012 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
2013 | root_backup = info->super_copy->super_roots; | |
2014 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2015 | if (cur == newest_gen) | |
2016 | newest_index = 0; | |
2017 | } | |
2018 | return newest_index; | |
2019 | } | |
2020 | ||
2021 | ||
2022 | /* | |
2023 | * find the oldest backup so we know where to store new entries | |
2024 | * in the backup array. This will set the backup_root_index | |
2025 | * field in the fs_info struct | |
2026 | */ | |
2027 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
2028 | u64 newest_gen) | |
2029 | { | |
2030 | int newest_index = -1; | |
2031 | ||
2032 | newest_index = find_newest_super_backup(info, newest_gen); | |
2033 | /* if there was garbage in there, just move along */ | |
2034 | if (newest_index == -1) { | |
2035 | info->backup_root_index = 0; | |
2036 | } else { | |
2037 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2038 | } | |
2039 | } | |
2040 | ||
2041 | /* | |
2042 | * copy all the root pointers into the super backup array. | |
2043 | * this will bump the backup pointer by one when it is | |
2044 | * done | |
2045 | */ | |
2046 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2047 | { | |
2048 | int next_backup; | |
2049 | struct btrfs_root_backup *root_backup; | |
2050 | int last_backup; | |
2051 | ||
2052 | next_backup = info->backup_root_index; | |
2053 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2054 | BTRFS_NUM_BACKUP_ROOTS; | |
2055 | ||
2056 | /* | |
2057 | * just overwrite the last backup if we're at the same generation | |
2058 | * this happens only at umount | |
2059 | */ | |
2060 | root_backup = info->super_for_commit->super_roots + last_backup; | |
2061 | if (btrfs_backup_tree_root_gen(root_backup) == | |
2062 | btrfs_header_generation(info->tree_root->node)) | |
2063 | next_backup = last_backup; | |
2064 | ||
2065 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2066 | ||
2067 | /* | |
2068 | * make sure all of our padding and empty slots get zero filled | |
2069 | * regardless of which ones we use today | |
2070 | */ | |
2071 | memset(root_backup, 0, sizeof(*root_backup)); | |
2072 | ||
2073 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2074 | ||
2075 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2076 | btrfs_set_backup_tree_root_gen(root_backup, | |
2077 | btrfs_header_generation(info->tree_root->node)); | |
2078 | ||
2079 | btrfs_set_backup_tree_root_level(root_backup, | |
2080 | btrfs_header_level(info->tree_root->node)); | |
2081 | ||
2082 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2083 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2084 | btrfs_header_generation(info->chunk_root->node)); | |
2085 | btrfs_set_backup_chunk_root_level(root_backup, | |
2086 | btrfs_header_level(info->chunk_root->node)); | |
2087 | ||
2088 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2089 | btrfs_set_backup_extent_root_gen(root_backup, | |
2090 | btrfs_header_generation(info->extent_root->node)); | |
2091 | btrfs_set_backup_extent_root_level(root_backup, | |
2092 | btrfs_header_level(info->extent_root->node)); | |
2093 | ||
7c7e82a7 CM |
2094 | /* |
2095 | * we might commit during log recovery, which happens before we set | |
2096 | * the fs_root. Make sure it is valid before we fill it in. | |
2097 | */ | |
2098 | if (info->fs_root && info->fs_root->node) { | |
2099 | btrfs_set_backup_fs_root(root_backup, | |
2100 | info->fs_root->node->start); | |
2101 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2102 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2103 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2104 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2105 | } |
af31f5e5 CM |
2106 | |
2107 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2108 | btrfs_set_backup_dev_root_gen(root_backup, | |
2109 | btrfs_header_generation(info->dev_root->node)); | |
2110 | btrfs_set_backup_dev_root_level(root_backup, | |
2111 | btrfs_header_level(info->dev_root->node)); | |
2112 | ||
2113 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2114 | btrfs_set_backup_csum_root_gen(root_backup, | |
2115 | btrfs_header_generation(info->csum_root->node)); | |
2116 | btrfs_set_backup_csum_root_level(root_backup, | |
2117 | btrfs_header_level(info->csum_root->node)); | |
2118 | ||
2119 | btrfs_set_backup_total_bytes(root_backup, | |
2120 | btrfs_super_total_bytes(info->super_copy)); | |
2121 | btrfs_set_backup_bytes_used(root_backup, | |
2122 | btrfs_super_bytes_used(info->super_copy)); | |
2123 | btrfs_set_backup_num_devices(root_backup, | |
2124 | btrfs_super_num_devices(info->super_copy)); | |
2125 | ||
2126 | /* | |
2127 | * if we don't copy this out to the super_copy, it won't get remembered | |
2128 | * for the next commit | |
2129 | */ | |
2130 | memcpy(&info->super_copy->super_roots, | |
2131 | &info->super_for_commit->super_roots, | |
2132 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2133 | } | |
2134 | ||
2135 | /* | |
2136 | * this copies info out of the root backup array and back into | |
2137 | * the in-memory super block. It is meant to help iterate through | |
2138 | * the array, so you send it the number of backups you've already | |
2139 | * tried and the last backup index you used. | |
2140 | * | |
2141 | * this returns -1 when it has tried all the backups | |
2142 | */ | |
2143 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2144 | struct btrfs_super_block *super, | |
2145 | int *num_backups_tried, int *backup_index) | |
2146 | { | |
2147 | struct btrfs_root_backup *root_backup; | |
2148 | int newest = *backup_index; | |
2149 | ||
2150 | if (*num_backups_tried == 0) { | |
2151 | u64 gen = btrfs_super_generation(super); | |
2152 | ||
2153 | newest = find_newest_super_backup(info, gen); | |
2154 | if (newest == -1) | |
2155 | return -1; | |
2156 | ||
2157 | *backup_index = newest; | |
2158 | *num_backups_tried = 1; | |
2159 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2160 | /* we've tried all the backups, all done */ | |
2161 | return -1; | |
2162 | } else { | |
2163 | /* jump to the next oldest backup */ | |
2164 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2165 | BTRFS_NUM_BACKUP_ROOTS; | |
2166 | *backup_index = newest; | |
2167 | *num_backups_tried += 1; | |
2168 | } | |
2169 | root_backup = super->super_roots + newest; | |
2170 | ||
2171 | btrfs_set_super_generation(super, | |
2172 | btrfs_backup_tree_root_gen(root_backup)); | |
2173 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2174 | btrfs_set_super_root_level(super, | |
2175 | btrfs_backup_tree_root_level(root_backup)); | |
2176 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2177 | ||
2178 | /* | |
2179 | * fixme: the total bytes and num_devices need to match or we should | |
2180 | * need a fsck | |
2181 | */ | |
2182 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2183 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2184 | return 0; | |
2185 | } | |
2186 | ||
7abadb64 LB |
2187 | /* helper to cleanup workers */ |
2188 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2189 | { | |
dc6e3209 | 2190 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2191 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2192 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2193 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2194 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
8b110e39 | 2195 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2196 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2197 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2198 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2199 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2200 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2201 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2202 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2203 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2204 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2205 | btrfs_destroy_workqueue(fs_info->extent_workers); |
a9b9477d FM |
2206 | /* |
2207 | * Now that all other work queues are destroyed, we can safely destroy | |
2208 | * the queues used for metadata I/O, since tasks from those other work | |
2209 | * queues can do metadata I/O operations. | |
2210 | */ | |
2211 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2212 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2213 | } |
2214 | ||
2e9f5954 R |
2215 | static void free_root_extent_buffers(struct btrfs_root *root) |
2216 | { | |
2217 | if (root) { | |
2218 | free_extent_buffer(root->node); | |
2219 | free_extent_buffer(root->commit_root); | |
2220 | root->node = NULL; | |
2221 | root->commit_root = NULL; | |
2222 | } | |
2223 | } | |
2224 | ||
af31f5e5 CM |
2225 | /* helper to cleanup tree roots */ |
2226 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2227 | { | |
2e9f5954 | 2228 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2229 | |
2e9f5954 R |
2230 | free_root_extent_buffers(info->dev_root); |
2231 | free_root_extent_buffers(info->extent_root); | |
2232 | free_root_extent_buffers(info->csum_root); | |
2233 | free_root_extent_buffers(info->quota_root); | |
2234 | free_root_extent_buffers(info->uuid_root); | |
2235 | if (chunk_root) | |
2236 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2237 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2238 | } |
2239 | ||
faa2dbf0 | 2240 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2241 | { |
2242 | int ret; | |
2243 | struct btrfs_root *gang[8]; | |
2244 | int i; | |
2245 | ||
2246 | while (!list_empty(&fs_info->dead_roots)) { | |
2247 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2248 | struct btrfs_root, root_list); | |
2249 | list_del(&gang[0]->root_list); | |
2250 | ||
27cdeb70 | 2251 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2252 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2253 | } else { |
2254 | free_extent_buffer(gang[0]->node); | |
2255 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2256 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2257 | } |
2258 | } | |
2259 | ||
2260 | while (1) { | |
2261 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2262 | (void **)gang, 0, | |
2263 | ARRAY_SIZE(gang)); | |
2264 | if (!ret) | |
2265 | break; | |
2266 | for (i = 0; i < ret; i++) | |
cb517eab | 2267 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2268 | } |
1a4319cc LB |
2269 | |
2270 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2271 | btrfs_free_log_root_tree(NULL, fs_info); | |
2ff7e61e | 2272 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
1a4319cc | 2273 | } |
171f6537 | 2274 | } |
af31f5e5 | 2275 | |
638aa7ed ES |
2276 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2277 | { | |
2278 | mutex_init(&fs_info->scrub_lock); | |
2279 | atomic_set(&fs_info->scrubs_running, 0); | |
2280 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2281 | atomic_set(&fs_info->scrubs_paused, 0); | |
2282 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2283 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2284 | fs_info->scrub_workers_refcnt = 0; | |
2285 | } | |
2286 | ||
779a65a4 ES |
2287 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2288 | { | |
2289 | spin_lock_init(&fs_info->balance_lock); | |
2290 | mutex_init(&fs_info->balance_mutex); | |
2291 | atomic_set(&fs_info->balance_running, 0); | |
2292 | atomic_set(&fs_info->balance_pause_req, 0); | |
2293 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2294 | fs_info->balance_ctl = NULL; | |
2295 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2296 | } | |
2297 | ||
6bccf3ab | 2298 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2299 | { |
2ff7e61e JM |
2300 | struct inode *inode = fs_info->btree_inode; |
2301 | ||
2302 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2303 | set_nlink(inode, 1); | |
f37938e0 ES |
2304 | /* |
2305 | * we set the i_size on the btree inode to the max possible int. | |
2306 | * the real end of the address space is determined by all of | |
2307 | * the devices in the system | |
2308 | */ | |
2ff7e61e JM |
2309 | inode->i_size = OFFSET_MAX; |
2310 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2311 | |
2ff7e61e | 2312 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
c6100a4b | 2313 | extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); |
2ff7e61e JM |
2314 | BTRFS_I(inode)->io_tree.track_uptodate = 0; |
2315 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); | |
f37938e0 | 2316 | |
2ff7e61e | 2317 | BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; |
f37938e0 | 2318 | |
2ff7e61e JM |
2319 | BTRFS_I(inode)->root = fs_info->tree_root; |
2320 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); | |
2321 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2322 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2323 | } |
2324 | ||
ad618368 ES |
2325 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2326 | { | |
2327 | fs_info->dev_replace.lock_owner = 0; | |
2328 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2329 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2330 | rwlock_init(&fs_info->dev_replace.lock); |
2331 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2332 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2333 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2334 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2335 | } |
2336 | ||
f9e92e40 ES |
2337 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2338 | { | |
2339 | spin_lock_init(&fs_info->qgroup_lock); | |
2340 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2341 | fs_info->qgroup_tree = RB_ROOT; | |
2342 | fs_info->qgroup_op_tree = RB_ROOT; | |
2343 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2344 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2345 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2346 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2347 | mutex_init(&fs_info->qgroup_rescan_lock); |
2348 | } | |
2349 | ||
2a458198 ES |
2350 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2351 | struct btrfs_fs_devices *fs_devices) | |
2352 | { | |
2353 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2354 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2355 | |
2356 | fs_info->workers = | |
cb001095 JM |
2357 | btrfs_alloc_workqueue(fs_info, "worker", |
2358 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2359 | |
2360 | fs_info->delalloc_workers = | |
cb001095 JM |
2361 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2362 | flags, max_active, 2); | |
2a458198 ES |
2363 | |
2364 | fs_info->flush_workers = | |
cb001095 JM |
2365 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2366 | flags, max_active, 0); | |
2a458198 ES |
2367 | |
2368 | fs_info->caching_workers = | |
cb001095 | 2369 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 ES |
2370 | |
2371 | /* | |
2372 | * a higher idle thresh on the submit workers makes it much more | |
2373 | * likely that bios will be send down in a sane order to the | |
2374 | * devices | |
2375 | */ | |
2376 | fs_info->submit_workers = | |
cb001095 | 2377 | btrfs_alloc_workqueue(fs_info, "submit", flags, |
2a458198 ES |
2378 | min_t(u64, fs_devices->num_devices, |
2379 | max_active), 64); | |
2380 | ||
2381 | fs_info->fixup_workers = | |
cb001095 | 2382 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2383 | |
2384 | /* | |
2385 | * endios are largely parallel and should have a very | |
2386 | * low idle thresh | |
2387 | */ | |
2388 | fs_info->endio_workers = | |
cb001095 | 2389 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2390 | fs_info->endio_meta_workers = |
cb001095 JM |
2391 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2392 | max_active, 4); | |
2a458198 | 2393 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2394 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2395 | max_active, 2); | |
2a458198 | 2396 | fs_info->endio_raid56_workers = |
cb001095 JM |
2397 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2398 | max_active, 4); | |
2a458198 | 2399 | fs_info->endio_repair_workers = |
cb001095 | 2400 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2401 | fs_info->rmw_workers = |
cb001095 | 2402 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2403 | fs_info->endio_write_workers = |
cb001095 JM |
2404 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2405 | max_active, 2); | |
2a458198 | 2406 | fs_info->endio_freespace_worker = |
cb001095 JM |
2407 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2408 | max_active, 0); | |
2a458198 | 2409 | fs_info->delayed_workers = |
cb001095 JM |
2410 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2411 | max_active, 0); | |
2a458198 | 2412 | fs_info->readahead_workers = |
cb001095 JM |
2413 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2414 | max_active, 2); | |
2a458198 | 2415 | fs_info->qgroup_rescan_workers = |
cb001095 | 2416 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
2a458198 | 2417 | fs_info->extent_workers = |
cb001095 | 2418 | btrfs_alloc_workqueue(fs_info, "extent-refs", flags, |
2a458198 ES |
2419 | min_t(u64, fs_devices->num_devices, |
2420 | max_active), 8); | |
2421 | ||
2422 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2423 | fs_info->submit_workers && fs_info->flush_workers && | |
2424 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2425 | fs_info->endio_meta_write_workers && | |
2426 | fs_info->endio_repair_workers && | |
2427 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2428 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2429 | fs_info->caching_workers && fs_info->readahead_workers && | |
2430 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2431 | fs_info->extent_workers && | |
2432 | fs_info->qgroup_rescan_workers)) { | |
2433 | return -ENOMEM; | |
2434 | } | |
2435 | ||
2436 | return 0; | |
2437 | } | |
2438 | ||
63443bf5 ES |
2439 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2440 | struct btrfs_fs_devices *fs_devices) | |
2441 | { | |
2442 | int ret; | |
63443bf5 ES |
2443 | struct btrfs_root *log_tree_root; |
2444 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2445 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2446 | ||
2447 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2448 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2449 | return -EIO; |
2450 | } | |
2451 | ||
74e4d827 | 2452 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2453 | if (!log_tree_root) |
2454 | return -ENOMEM; | |
2455 | ||
da17066c | 2456 | __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
63443bf5 | 2457 | |
2ff7e61e JM |
2458 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
2459 | fs_info->generation + 1); | |
64c043de | 2460 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2461 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2462 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2463 | kfree(log_tree_root); |
0eeff236 | 2464 | return ret; |
64c043de | 2465 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2466 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2467 | free_extent_buffer(log_tree_root->node); |
2468 | kfree(log_tree_root); | |
2469 | return -EIO; | |
2470 | } | |
2471 | /* returns with log_tree_root freed on success */ | |
2472 | ret = btrfs_recover_log_trees(log_tree_root); | |
2473 | if (ret) { | |
0b246afa JM |
2474 | btrfs_handle_fs_error(fs_info, ret, |
2475 | "Failed to recover log tree"); | |
63443bf5 ES |
2476 | free_extent_buffer(log_tree_root->node); |
2477 | kfree(log_tree_root); | |
2478 | return ret; | |
2479 | } | |
2480 | ||
2481 | if (fs_info->sb->s_flags & MS_RDONLY) { | |
6bccf3ab | 2482 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2483 | if (ret) |
2484 | return ret; | |
2485 | } | |
2486 | ||
2487 | return 0; | |
2488 | } | |
2489 | ||
6bccf3ab | 2490 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2491 | { |
6bccf3ab | 2492 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2493 | struct btrfs_root *root; |
4bbcaa64 ES |
2494 | struct btrfs_key location; |
2495 | int ret; | |
2496 | ||
6bccf3ab JM |
2497 | BUG_ON(!fs_info->tree_root); |
2498 | ||
4bbcaa64 ES |
2499 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2500 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2501 | location.offset = 0; | |
2502 | ||
a4f3d2c4 DS |
2503 | root = btrfs_read_tree_root(tree_root, &location); |
2504 | if (IS_ERR(root)) | |
2505 | return PTR_ERR(root); | |
2506 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2507 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2508 | |
2509 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2510 | root = btrfs_read_tree_root(tree_root, &location); |
2511 | if (IS_ERR(root)) | |
2512 | return PTR_ERR(root); | |
2513 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2514 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2515 | btrfs_init_devices_late(fs_info); |
2516 | ||
2517 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2518 | root = btrfs_read_tree_root(tree_root, &location); |
2519 | if (IS_ERR(root)) | |
2520 | return PTR_ERR(root); | |
2521 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2522 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2523 | |
2524 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2525 | root = btrfs_read_tree_root(tree_root, &location); |
2526 | if (!IS_ERR(root)) { | |
2527 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2528 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2529 | fs_info->quota_root = root; |
4bbcaa64 ES |
2530 | } |
2531 | ||
2532 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2533 | root = btrfs_read_tree_root(tree_root, &location); |
2534 | if (IS_ERR(root)) { | |
2535 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2536 | if (ret != -ENOENT) |
2537 | return ret; | |
2538 | } else { | |
a4f3d2c4 DS |
2539 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2540 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2541 | } |
2542 | ||
70f6d82e OS |
2543 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2544 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2545 | root = btrfs_read_tree_root(tree_root, &location); | |
2546 | if (IS_ERR(root)) | |
2547 | return PTR_ERR(root); | |
2548 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2549 | fs_info->free_space_root = root; | |
2550 | } | |
2551 | ||
4bbcaa64 ES |
2552 | return 0; |
2553 | } | |
2554 | ||
ad2b2c80 AV |
2555 | int open_ctree(struct super_block *sb, |
2556 | struct btrfs_fs_devices *fs_devices, | |
2557 | char *options) | |
2e635a27 | 2558 | { |
db94535d CM |
2559 | u32 sectorsize; |
2560 | u32 nodesize; | |
87ee04eb | 2561 | u32 stripesize; |
84234f3a | 2562 | u64 generation; |
f2b636e8 | 2563 | u64 features; |
3de4586c | 2564 | struct btrfs_key location; |
a061fc8d | 2565 | struct buffer_head *bh; |
4d34b278 | 2566 | struct btrfs_super_block *disk_super; |
815745cf | 2567 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2568 | struct btrfs_root *tree_root; |
4d34b278 | 2569 | struct btrfs_root *chunk_root; |
eb60ceac | 2570 | int ret; |
e58ca020 | 2571 | int err = -EINVAL; |
af31f5e5 CM |
2572 | int num_backups_tried = 0; |
2573 | int backup_index = 0; | |
5cdc7ad3 | 2574 | int max_active; |
6675df31 | 2575 | int clear_free_space_tree = 0; |
4543df7e | 2576 | |
74e4d827 DS |
2577 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2578 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2579 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2580 | err = -ENOMEM; |
2581 | goto fail; | |
2582 | } | |
76dda93c YZ |
2583 | |
2584 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2585 | if (ret) { | |
2586 | err = ret; | |
2587 | goto fail; | |
2588 | } | |
2589 | ||
908c7f19 | 2590 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2591 | if (ret) { |
2592 | err = ret; | |
9e11ceee | 2593 | goto fail_srcu; |
e2d84521 | 2594 | } |
09cbfeaf | 2595 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2596 | (1 + ilog2(nr_cpu_ids)); |
2597 | ||
908c7f19 | 2598 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2599 | if (ret) { |
2600 | err = ret; | |
2601 | goto fail_dirty_metadata_bytes; | |
2602 | } | |
2603 | ||
908c7f19 | 2604 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2605 | if (ret) { |
2606 | err = ret; | |
2607 | goto fail_delalloc_bytes; | |
2608 | } | |
2609 | ||
76dda93c YZ |
2610 | fs_info->btree_inode = new_inode(sb); |
2611 | if (!fs_info->btree_inode) { | |
2612 | err = -ENOMEM; | |
c404e0dc | 2613 | goto fail_bio_counter; |
76dda93c YZ |
2614 | } |
2615 | ||
a6591715 | 2616 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2617 | |
76dda93c | 2618 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2619 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2620 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2621 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2622 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2623 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2624 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2625 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2626 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2627 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2628 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2629 | spin_lock_init(&fs_info->defrag_inodes_lock); |
f29021b2 | 2630 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2631 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2632 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2633 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2634 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2635 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2636 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2637 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2638 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2639 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2640 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2641 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2642 | |
0b86a832 | 2643 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2644 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2645 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2646 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2647 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2648 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2649 | BTRFS_BLOCK_RSV_GLOBAL); | |
2650 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2651 | BTRFS_BLOCK_RSV_DELALLOC); | |
2652 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2653 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2654 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2655 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2656 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2657 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2658 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2659 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2660 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2661 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2662 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2663 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2664 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2665 | fs_info->sb = sb; |
95ac567a | 2666 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2667 | fs_info->metadata_ratio = 0; |
4cb5300b | 2668 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2669 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2670 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2671 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2672 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2673 | /* readahead state */ |
d0164adc | 2674 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2675 | spin_lock_init(&fs_info->reada_lock); |
c8b97818 | 2676 | |
b34b086c CM |
2677 | fs_info->thread_pool_size = min_t(unsigned long, |
2678 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2679 | |
199c2a9c MX |
2680 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2681 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 | 2682 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2683 | GFP_KERNEL); |
16cdcec7 MX |
2684 | if (!fs_info->delayed_root) { |
2685 | err = -ENOMEM; | |
2686 | goto fail_iput; | |
2687 | } | |
2688 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2689 | |
638aa7ed | 2690 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2691 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2692 | fs_info->check_integrity_print_mask = 0; | |
2693 | #endif | |
779a65a4 | 2694 | btrfs_init_balance(fs_info); |
21c7e756 | 2695 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2696 | |
9f6d2510 DS |
2697 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2698 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
a061fc8d | 2699 | |
6bccf3ab | 2700 | btrfs_init_btree_inode(fs_info); |
76dda93c | 2701 | |
0f9dd46c | 2702 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2703 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2704 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2705 | |
c6100a4b JB |
2706 | extent_io_tree_init(&fs_info->freed_extents[0], NULL); |
2707 | extent_io_tree_init(&fs_info->freed_extents[1], NULL); | |
11833d66 | 2708 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
afcdd129 | 2709 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2710 | |
5a3f23d5 | 2711 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2712 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2713 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2714 | mutex_init(&fs_info->transaction_kthread_mutex); |
2715 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2716 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2717 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2718 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2719 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2720 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2721 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2722 | |
ad618368 | 2723 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2724 | btrfs_init_qgroup(fs_info); |
416ac51d | 2725 | |
fa9c0d79 CM |
2726 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2727 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2728 | ||
e6dcd2dc | 2729 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2730 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2731 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2732 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2733 | |
04216820 FM |
2734 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2735 | ||
da17066c JM |
2736 | /* Usable values until the real ones are cached from the superblock */ |
2737 | fs_info->nodesize = 4096; | |
2738 | fs_info->sectorsize = 4096; | |
2739 | fs_info->stripesize = 4096; | |
2740 | ||
53b381b3 DW |
2741 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2742 | if (ret) { | |
83c8266a | 2743 | err = ret; |
53b381b3 DW |
2744 | goto fail_alloc; |
2745 | } | |
2746 | ||
da17066c | 2747 | __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2748 | |
3c4bb26b | 2749 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2750 | |
2751 | /* | |
2752 | * Read super block and check the signature bytes only | |
2753 | */ | |
a512bbf8 | 2754 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2755 | if (IS_ERR(bh)) { |
2756 | err = PTR_ERR(bh); | |
16cdcec7 | 2757 | goto fail_alloc; |
20b45077 | 2758 | } |
39279cc3 | 2759 | |
1104a885 DS |
2760 | /* |
2761 | * We want to check superblock checksum, the type is stored inside. | |
2762 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2763 | */ | |
ab8d0fc4 | 2764 | if (btrfs_check_super_csum(fs_info, bh->b_data)) { |
05135f59 | 2765 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2766 | err = -EINVAL; |
b2acdddf | 2767 | brelse(bh); |
1104a885 DS |
2768 | goto fail_alloc; |
2769 | } | |
2770 | ||
2771 | /* | |
2772 | * super_copy is zeroed at allocation time and we never touch the | |
2773 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2774 | * the whole block of INFO_SIZE | |
2775 | */ | |
6c41761f DS |
2776 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2777 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2778 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2779 | brelse(bh); |
5f39d397 | 2780 | |
6c41761f | 2781 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2782 | |
3d3a126a | 2783 | ret = btrfs_check_super_valid(fs_info); |
1104a885 | 2784 | if (ret) { |
05135f59 | 2785 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 DS |
2786 | err = -EINVAL; |
2787 | goto fail_alloc; | |
2788 | } | |
2789 | ||
6c41761f | 2790 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2791 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2792 | goto fail_alloc; |
0f7d52f4 | 2793 | |
acce952b | 2794 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2795 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2796 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2797 | |
af31f5e5 CM |
2798 | /* |
2799 | * run through our array of backup supers and setup | |
2800 | * our ring pointer to the oldest one | |
2801 | */ | |
2802 | generation = btrfs_super_generation(disk_super); | |
2803 | find_oldest_super_backup(fs_info, generation); | |
2804 | ||
75e7cb7f LB |
2805 | /* |
2806 | * In the long term, we'll store the compression type in the super | |
2807 | * block, and it'll be used for per file compression control. | |
2808 | */ | |
2809 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2810 | ||
2ff7e61e | 2811 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
2812 | if (ret) { |
2813 | err = ret; | |
16cdcec7 | 2814 | goto fail_alloc; |
2b82032c | 2815 | } |
dfe25020 | 2816 | |
f2b636e8 JB |
2817 | features = btrfs_super_incompat_flags(disk_super) & |
2818 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2819 | if (features) { | |
05135f59 DS |
2820 | btrfs_err(fs_info, |
2821 | "cannot mount because of unsupported optional features (%llx)", | |
2822 | features); | |
f2b636e8 | 2823 | err = -EINVAL; |
16cdcec7 | 2824 | goto fail_alloc; |
f2b636e8 JB |
2825 | } |
2826 | ||
5d4f98a2 | 2827 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2828 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 2829 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2830 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2831 | |
3173a18f | 2832 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2833 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2834 | |
727011e0 CM |
2835 | /* |
2836 | * flag our filesystem as having big metadata blocks if | |
2837 | * they are bigger than the page size | |
2838 | */ | |
09cbfeaf | 2839 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2840 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2841 | btrfs_info(fs_info, |
2842 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2843 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2844 | } | |
2845 | ||
bc3f116f | 2846 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2847 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2848 | stripesize = sectorsize; |
707e8a07 | 2849 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2850 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f | 2851 | |
da17066c JM |
2852 | /* Cache block sizes */ |
2853 | fs_info->nodesize = nodesize; | |
2854 | fs_info->sectorsize = sectorsize; | |
2855 | fs_info->stripesize = stripesize; | |
2856 | ||
bc3f116f CM |
2857 | /* |
2858 | * mixed block groups end up with duplicate but slightly offset | |
2859 | * extent buffers for the same range. It leads to corruptions | |
2860 | */ | |
2861 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2862 | (sectorsize != nodesize)) { |
05135f59 DS |
2863 | btrfs_err(fs_info, |
2864 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2865 | nodesize, sectorsize); | |
bc3f116f CM |
2866 | goto fail_alloc; |
2867 | } | |
2868 | ||
ceda0864 MX |
2869 | /* |
2870 | * Needn't use the lock because there is no other task which will | |
2871 | * update the flag. | |
2872 | */ | |
a6fa6fae | 2873 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2874 | |
f2b636e8 JB |
2875 | features = btrfs_super_compat_ro_flags(disk_super) & |
2876 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2877 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
05135f59 DS |
2878 | btrfs_err(fs_info, |
2879 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2880 | features); |
f2b636e8 | 2881 | err = -EINVAL; |
16cdcec7 | 2882 | goto fail_alloc; |
f2b636e8 | 2883 | } |
61d92c32 | 2884 | |
5cdc7ad3 | 2885 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2886 | |
2a458198 ES |
2887 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2888 | if (ret) { | |
2889 | err = ret; | |
0dc3b84a JB |
2890 | goto fail_sb_buffer; |
2891 | } | |
4543df7e | 2892 | |
9e11ceee JK |
2893 | sb->s_bdi->congested_fn = btrfs_congested_fn; |
2894 | sb->s_bdi->congested_data = fs_info; | |
2895 | sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; | |
2896 | sb->s_bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE; | |
2897 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); | |
2898 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 2899 | |
a061fc8d CM |
2900 | sb->s_blocksize = sectorsize; |
2901 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2902 | |
925baedd | 2903 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 2904 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 2905 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2906 | if (ret) { |
05135f59 | 2907 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 2908 | goto fail_sb_buffer; |
84eed90f | 2909 | } |
0b86a832 | 2910 | |
84234f3a | 2911 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2912 | |
da17066c | 2913 | __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); |
0b86a832 | 2914 | |
2ff7e61e | 2915 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 2916 | btrfs_super_chunk_root(disk_super), |
ce86cd59 | 2917 | generation); |
64c043de LB |
2918 | if (IS_ERR(chunk_root->node) || |
2919 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 2920 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 2921 | if (!IS_ERR(chunk_root->node)) |
2922 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2923 | chunk_root->node = NULL; |
af31f5e5 | 2924 | goto fail_tree_roots; |
83121942 | 2925 | } |
5d4f98a2 YZ |
2926 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2927 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2928 | |
e17cade2 | 2929 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2930 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2931 | |
5b4aacef | 2932 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 2933 | if (ret) { |
05135f59 | 2934 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 2935 | goto fail_tree_roots; |
2b82032c | 2936 | } |
0b86a832 | 2937 | |
8dabb742 SB |
2938 | /* |
2939 | * keep the device that is marked to be the target device for the | |
2940 | * dev_replace procedure | |
2941 | */ | |
9eaed21e | 2942 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2943 | |
a6b0d5c8 | 2944 | if (!fs_devices->latest_bdev) { |
05135f59 | 2945 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
2946 | goto fail_tree_roots; |
2947 | } | |
2948 | ||
af31f5e5 | 2949 | retry_root_backup: |
84234f3a | 2950 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2951 | |
2ff7e61e | 2952 | tree_root->node = read_tree_block(fs_info, |
db94535d | 2953 | btrfs_super_root(disk_super), |
ce86cd59 | 2954 | generation); |
64c043de LB |
2955 | if (IS_ERR(tree_root->node) || |
2956 | !extent_buffer_uptodate(tree_root->node)) { | |
05135f59 | 2957 | btrfs_warn(fs_info, "failed to read tree root"); |
e5fffbac | 2958 | if (!IS_ERR(tree_root->node)) |
2959 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2960 | tree_root->node = NULL; |
af31f5e5 | 2961 | goto recovery_tree_root; |
83121942 | 2962 | } |
af31f5e5 | 2963 | |
5d4f98a2 YZ |
2964 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2965 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2966 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2967 | |
f32e48e9 CR |
2968 | mutex_lock(&tree_root->objectid_mutex); |
2969 | ret = btrfs_find_highest_objectid(tree_root, | |
2970 | &tree_root->highest_objectid); | |
2971 | if (ret) { | |
2972 | mutex_unlock(&tree_root->objectid_mutex); | |
2973 | goto recovery_tree_root; | |
2974 | } | |
2975 | ||
2976 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2977 | ||
2978 | mutex_unlock(&tree_root->objectid_mutex); | |
2979 | ||
6bccf3ab | 2980 | ret = btrfs_read_roots(fs_info); |
4bbcaa64 | 2981 | if (ret) |
af31f5e5 | 2982 | goto recovery_tree_root; |
f7a81ea4 | 2983 | |
8929ecfa YZ |
2984 | fs_info->generation = generation; |
2985 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2986 | |
68310a5e ID |
2987 | ret = btrfs_recover_balance(fs_info); |
2988 | if (ret) { | |
05135f59 | 2989 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
2990 | goto fail_block_groups; |
2991 | } | |
2992 | ||
733f4fbb SB |
2993 | ret = btrfs_init_dev_stats(fs_info); |
2994 | if (ret) { | |
05135f59 | 2995 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
2996 | goto fail_block_groups; |
2997 | } | |
2998 | ||
8dabb742 SB |
2999 | ret = btrfs_init_dev_replace(fs_info); |
3000 | if (ret) { | |
05135f59 | 3001 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3002 | goto fail_block_groups; |
3003 | } | |
3004 | ||
9eaed21e | 3005 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 3006 | |
b7c35e81 AJ |
3007 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
3008 | if (ret) { | |
05135f59 DS |
3009 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3010 | ret); | |
b7c35e81 AJ |
3011 | goto fail_block_groups; |
3012 | } | |
3013 | ||
3014 | ret = btrfs_sysfs_add_device(fs_devices); | |
3015 | if (ret) { | |
05135f59 DS |
3016 | btrfs_err(fs_info, "failed to init sysfs device interface: %d", |
3017 | ret); | |
b7c35e81 AJ |
3018 | goto fail_fsdev_sysfs; |
3019 | } | |
3020 | ||
96f3136e | 3021 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3022 | if (ret) { |
05135f59 | 3023 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3024 | goto fail_fsdev_sysfs; |
c59021f8 | 3025 | } |
3026 | ||
c59021f8 | 3027 | ret = btrfs_init_space_info(fs_info); |
3028 | if (ret) { | |
05135f59 | 3029 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3030 | goto fail_sysfs; |
c59021f8 | 3031 | } |
3032 | ||
5b4aacef | 3033 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3034 | if (ret) { |
05135f59 | 3035 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3036 | goto fail_sysfs; |
1b1d1f66 | 3037 | } |
4330e183 QW |
3038 | |
3039 | if (!(sb->s_flags & MS_RDONLY) && !btrfs_check_rw_degradable(fs_info)) { | |
05135f59 | 3040 | btrfs_warn(fs_info, |
4330e183 | 3041 | "writeable mount is not allowed due to too many missing devices"); |
2365dd3c | 3042 | goto fail_sysfs; |
292fd7fc | 3043 | } |
9078a3e1 | 3044 | |
a74a4b97 CM |
3045 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3046 | "btrfs-cleaner"); | |
57506d50 | 3047 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3048 | goto fail_sysfs; |
a74a4b97 CM |
3049 | |
3050 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3051 | tree_root, | |
3052 | "btrfs-transaction"); | |
57506d50 | 3053 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3054 | goto fail_cleaner; |
a74a4b97 | 3055 | |
583b7231 | 3056 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3057 | !fs_info->fs_devices->rotating) { |
583b7231 | 3058 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3059 | } |
3060 | ||
572d9ab7 | 3061 | /* |
01327610 | 3062 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3063 | * not have to wait for transaction commit |
3064 | */ | |
3065 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3066 | |
21adbd5c | 3067 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3068 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3069 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3070 | btrfs_test_opt(fs_info, |
21adbd5c SB |
3071 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3072 | 1 : 0, | |
3073 | fs_info->check_integrity_print_mask); | |
3074 | if (ret) | |
05135f59 DS |
3075 | btrfs_warn(fs_info, |
3076 | "failed to initialize integrity check module: %d", | |
3077 | ret); | |
21adbd5c SB |
3078 | } |
3079 | #endif | |
bcef60f2 AJ |
3080 | ret = btrfs_read_qgroup_config(fs_info); |
3081 | if (ret) | |
3082 | goto fail_trans_kthread; | |
21adbd5c | 3083 | |
96da0919 QW |
3084 | /* do not make disk changes in broken FS or nologreplay is given */ |
3085 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3086 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
63443bf5 | 3087 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3088 | if (ret) { |
63443bf5 | 3089 | err = ret; |
28c16cbb | 3090 | goto fail_qgroup; |
79787eaa | 3091 | } |
e02119d5 | 3092 | } |
1a40e23b | 3093 | |
6bccf3ab | 3094 | ret = btrfs_find_orphan_roots(fs_info); |
79787eaa | 3095 | if (ret) |
28c16cbb | 3096 | goto fail_qgroup; |
76dda93c | 3097 | |
7c2ca468 | 3098 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 3099 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3100 | if (ret) |
28c16cbb | 3101 | goto fail_qgroup; |
90c711ab ZB |
3102 | |
3103 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3104 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3105 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3106 | if (ret < 0) { |
05135f59 DS |
3107 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3108 | ret); | |
d7ce5843 | 3109 | err = -EINVAL; |
bcef60f2 | 3110 | goto fail_qgroup; |
d7ce5843 | 3111 | } |
7c2ca468 | 3112 | } |
1a40e23b | 3113 | |
3de4586c CM |
3114 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3115 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3116 | location.offset = 0; |
3de4586c | 3117 | |
3de4586c | 3118 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3119 | if (IS_ERR(fs_info->fs_root)) { |
3120 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 3121 | goto fail_qgroup; |
3140c9a3 | 3122 | } |
c289811c | 3123 | |
2b6ba629 ID |
3124 | if (sb->s_flags & MS_RDONLY) |
3125 | return 0; | |
59641015 | 3126 | |
f8d468a1 OS |
3127 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && |
3128 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
6675df31 OS |
3129 | clear_free_space_tree = 1; |
3130 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3131 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3132 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3133 | clear_free_space_tree = 1; | |
3134 | } | |
3135 | ||
3136 | if (clear_free_space_tree) { | |
f8d468a1 OS |
3137 | btrfs_info(fs_info, "clearing free space tree"); |
3138 | ret = btrfs_clear_free_space_tree(fs_info); | |
3139 | if (ret) { | |
3140 | btrfs_warn(fs_info, | |
3141 | "failed to clear free space tree: %d", ret); | |
6bccf3ab | 3142 | close_ctree(fs_info); |
f8d468a1 OS |
3143 | return ret; |
3144 | } | |
3145 | } | |
3146 | ||
0b246afa | 3147 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
511711af | 3148 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3149 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3150 | ret = btrfs_create_free_space_tree(fs_info); |
3151 | if (ret) { | |
05135f59 DS |
3152 | btrfs_warn(fs_info, |
3153 | "failed to create free space tree: %d", ret); | |
6bccf3ab | 3154 | close_ctree(fs_info); |
511711af CM |
3155 | return ret; |
3156 | } | |
3157 | } | |
3158 | ||
2b6ba629 ID |
3159 | down_read(&fs_info->cleanup_work_sem); |
3160 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3161 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3162 | up_read(&fs_info->cleanup_work_sem); |
6bccf3ab | 3163 | close_ctree(fs_info); |
2b6ba629 ID |
3164 | return ret; |
3165 | } | |
3166 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3167 | |
2b6ba629 ID |
3168 | ret = btrfs_resume_balance_async(fs_info); |
3169 | if (ret) { | |
05135f59 | 3170 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
6bccf3ab | 3171 | close_ctree(fs_info); |
2b6ba629 | 3172 | return ret; |
e3acc2a6 JB |
3173 | } |
3174 | ||
8dabb742 SB |
3175 | ret = btrfs_resume_dev_replace_async(fs_info); |
3176 | if (ret) { | |
05135f59 | 3177 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
6bccf3ab | 3178 | close_ctree(fs_info); |
8dabb742 SB |
3179 | return ret; |
3180 | } | |
3181 | ||
b382a324 JS |
3182 | btrfs_qgroup_rescan_resume(fs_info); |
3183 | ||
4bbcaa64 | 3184 | if (!fs_info->uuid_root) { |
05135f59 | 3185 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3186 | ret = btrfs_create_uuid_tree(fs_info); |
3187 | if (ret) { | |
05135f59 DS |
3188 | btrfs_warn(fs_info, |
3189 | "failed to create the UUID tree: %d", ret); | |
6bccf3ab | 3190 | close_ctree(fs_info); |
f7a81ea4 SB |
3191 | return ret; |
3192 | } | |
0b246afa | 3193 | } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3194 | fs_info->generation != |
3195 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3196 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3197 | ret = btrfs_check_uuid_tree(fs_info); |
3198 | if (ret) { | |
05135f59 DS |
3199 | btrfs_warn(fs_info, |
3200 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3201 | close_ctree(fs_info); |
70f80175 SB |
3202 | return ret; |
3203 | } | |
3204 | } else { | |
afcdd129 | 3205 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); |
f7a81ea4 | 3206 | } |
afcdd129 | 3207 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3208 | |
8dcddfa0 QW |
3209 | /* |
3210 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3211 | * no need to keep the flag | |
3212 | */ | |
3213 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3214 | ||
ad2b2c80 | 3215 | return 0; |
39279cc3 | 3216 | |
bcef60f2 AJ |
3217 | fail_qgroup: |
3218 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3219 | fail_trans_kthread: |
3220 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3221 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3222 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3223 | fail_cleaner: |
a74a4b97 | 3224 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3225 | |
3226 | /* | |
3227 | * make sure we're done with the btree inode before we stop our | |
3228 | * kthreads | |
3229 | */ | |
3230 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3231 | |
2365dd3c | 3232 | fail_sysfs: |
6618a59b | 3233 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3234 | |
b7c35e81 AJ |
3235 | fail_fsdev_sysfs: |
3236 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3237 | ||
1b1d1f66 | 3238 | fail_block_groups: |
54067ae9 | 3239 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3240 | |
3241 | fail_tree_roots: | |
3242 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3243 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3244 | |
39279cc3 | 3245 | fail_sb_buffer: |
7abadb64 | 3246 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3247 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3248 | fail_alloc: |
4543df7e | 3249 | fail_iput: |
586e46e2 ID |
3250 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3251 | ||
4543df7e | 3252 | iput(fs_info->btree_inode); |
c404e0dc MX |
3253 | fail_bio_counter: |
3254 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3255 | fail_delalloc_bytes: |
3256 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3257 | fail_dirty_metadata_bytes: |
3258 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
76dda93c YZ |
3259 | fail_srcu: |
3260 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3261 | fail: |
53b381b3 | 3262 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3263 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3264 | return err; |
af31f5e5 CM |
3265 | |
3266 | recovery_tree_root: | |
0b246afa | 3267 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) |
af31f5e5 CM |
3268 | goto fail_tree_roots; |
3269 | ||
3270 | free_root_pointers(fs_info, 0); | |
3271 | ||
3272 | /* don't use the log in recovery mode, it won't be valid */ | |
3273 | btrfs_set_super_log_root(disk_super, 0); | |
3274 | ||
3275 | /* we can't trust the free space cache either */ | |
3276 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3277 | ||
3278 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3279 | &num_backups_tried, &backup_index); | |
3280 | if (ret == -1) | |
3281 | goto fail_block_groups; | |
3282 | goto retry_root_backup; | |
eb60ceac CM |
3283 | } |
3284 | ||
f2984462 CM |
3285 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3286 | { | |
f2984462 CM |
3287 | if (uptodate) { |
3288 | set_buffer_uptodate(bh); | |
3289 | } else { | |
442a4f63 SB |
3290 | struct btrfs_device *device = (struct btrfs_device *) |
3291 | bh->b_private; | |
3292 | ||
fb456252 | 3293 | btrfs_warn_rl_in_rcu(device->fs_info, |
b14af3b4 | 3294 | "lost page write due to IO error on %s", |
606686ee | 3295 | rcu_str_deref(device->name)); |
01327610 | 3296 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3297 | * our own ways of dealing with the IO errors |
3298 | */ | |
f2984462 | 3299 | clear_buffer_uptodate(bh); |
442a4f63 | 3300 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3301 | } |
3302 | unlock_buffer(bh); | |
3303 | put_bh(bh); | |
3304 | } | |
3305 | ||
29c36d72 AJ |
3306 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3307 | struct buffer_head **bh_ret) | |
3308 | { | |
3309 | struct buffer_head *bh; | |
3310 | struct btrfs_super_block *super; | |
3311 | u64 bytenr; | |
3312 | ||
3313 | bytenr = btrfs_sb_offset(copy_num); | |
3314 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3315 | return -EINVAL; | |
3316 | ||
9f6d2510 | 3317 | bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); |
29c36d72 AJ |
3318 | /* |
3319 | * If we fail to read from the underlying devices, as of now | |
3320 | * the best option we have is to mark it EIO. | |
3321 | */ | |
3322 | if (!bh) | |
3323 | return -EIO; | |
3324 | ||
3325 | super = (struct btrfs_super_block *)bh->b_data; | |
3326 | if (btrfs_super_bytenr(super) != bytenr || | |
3327 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3328 | brelse(bh); | |
3329 | return -EINVAL; | |
3330 | } | |
3331 | ||
3332 | *bh_ret = bh; | |
3333 | return 0; | |
3334 | } | |
3335 | ||
3336 | ||
a512bbf8 YZ |
3337 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3338 | { | |
3339 | struct buffer_head *bh; | |
3340 | struct buffer_head *latest = NULL; | |
3341 | struct btrfs_super_block *super; | |
3342 | int i; | |
3343 | u64 transid = 0; | |
92fc03fb | 3344 | int ret = -EINVAL; |
a512bbf8 YZ |
3345 | |
3346 | /* we would like to check all the supers, but that would make | |
3347 | * a btrfs mount succeed after a mkfs from a different FS. | |
3348 | * So, we need to add a special mount option to scan for | |
3349 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3350 | */ | |
3351 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3352 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3353 | if (ret) | |
a512bbf8 YZ |
3354 | continue; |
3355 | ||
3356 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3357 | |
3358 | if (!latest || btrfs_super_generation(super) > transid) { | |
3359 | brelse(latest); | |
3360 | latest = bh; | |
3361 | transid = btrfs_super_generation(super); | |
3362 | } else { | |
3363 | brelse(bh); | |
3364 | } | |
3365 | } | |
92fc03fb AJ |
3366 | |
3367 | if (!latest) | |
3368 | return ERR_PTR(ret); | |
3369 | ||
a512bbf8 YZ |
3370 | return latest; |
3371 | } | |
3372 | ||
4eedeb75 | 3373 | /* |
abbb3b8e DS |
3374 | * Write superblock @sb to the @device. Do not wait for completion, all the |
3375 | * buffer heads we write are pinned. | |
4eedeb75 | 3376 | * |
abbb3b8e DS |
3377 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3378 | * the expected device size at commit time. Note that max_mirrors must be | |
3379 | * same for write and wait phases. | |
4eedeb75 | 3380 | * |
abbb3b8e | 3381 | * Return number of errors when buffer head is not found or submission fails. |
4eedeb75 | 3382 | */ |
a512bbf8 | 3383 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3384 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 YZ |
3385 | { |
3386 | struct buffer_head *bh; | |
3387 | int i; | |
3388 | int ret; | |
3389 | int errors = 0; | |
3390 | u32 crc; | |
3391 | u64 bytenr; | |
a512bbf8 YZ |
3392 | |
3393 | if (max_mirrors == 0) | |
3394 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3395 | ||
a512bbf8 YZ |
3396 | for (i = 0; i < max_mirrors; i++) { |
3397 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3398 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3399 | device->commit_total_bytes) | |
a512bbf8 YZ |
3400 | break; |
3401 | ||
abbb3b8e | 3402 | btrfs_set_super_bytenr(sb, bytenr); |
4eedeb75 | 3403 | |
abbb3b8e DS |
3404 | crc = ~(u32)0; |
3405 | crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, | |
3406 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
3407 | btrfs_csum_final(crc, sb->csum); | |
4eedeb75 | 3408 | |
abbb3b8e | 3409 | /* One reference for us, and we leave it for the caller */ |
9f6d2510 | 3410 | bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, |
abbb3b8e DS |
3411 | BTRFS_SUPER_INFO_SIZE); |
3412 | if (!bh) { | |
3413 | btrfs_err(device->fs_info, | |
3414 | "couldn't get super buffer head for bytenr %llu", | |
3415 | bytenr); | |
3416 | errors++; | |
4eedeb75 | 3417 | continue; |
abbb3b8e | 3418 | } |
634554dc | 3419 | |
abbb3b8e | 3420 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
a512bbf8 | 3421 | |
abbb3b8e DS |
3422 | /* one reference for submit_bh */ |
3423 | get_bh(bh); | |
4eedeb75 | 3424 | |
abbb3b8e DS |
3425 | set_buffer_uptodate(bh); |
3426 | lock_buffer(bh); | |
3427 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
3428 | bh->b_private = device; | |
a512bbf8 | 3429 | |
387125fc CM |
3430 | /* |
3431 | * we fua the first super. The others we allow | |
3432 | * to go down lazy. | |
3433 | */ | |
8d910125 JK |
3434 | if (i == 0) { |
3435 | ret = btrfsic_submit_bh(REQ_OP_WRITE, | |
db95c876 | 3436 | REQ_SYNC | REQ_FUA | REQ_META | REQ_PRIO, bh); |
8d910125 | 3437 | } else { |
db95c876 DS |
3438 | ret = btrfsic_submit_bh(REQ_OP_WRITE, |
3439 | REQ_SYNC | REQ_META | REQ_PRIO, bh); | |
8d910125 | 3440 | } |
4eedeb75 | 3441 | if (ret) |
a512bbf8 | 3442 | errors++; |
a512bbf8 YZ |
3443 | } |
3444 | return errors < i ? 0 : -1; | |
3445 | } | |
3446 | ||
abbb3b8e DS |
3447 | /* |
3448 | * Wait for write completion of superblocks done by write_dev_supers, | |
3449 | * @max_mirrors same for write and wait phases. | |
3450 | * | |
3451 | * Return number of errors when buffer head is not found or not marked up to | |
3452 | * date. | |
3453 | */ | |
3454 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3455 | { | |
3456 | struct buffer_head *bh; | |
3457 | int i; | |
3458 | int errors = 0; | |
3459 | u64 bytenr; | |
3460 | ||
3461 | if (max_mirrors == 0) | |
3462 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3463 | ||
3464 | for (i = 0; i < max_mirrors; i++) { | |
3465 | bytenr = btrfs_sb_offset(i); | |
3466 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
3467 | device->commit_total_bytes) | |
3468 | break; | |
3469 | ||
9f6d2510 DS |
3470 | bh = __find_get_block(device->bdev, |
3471 | bytenr / BTRFS_BDEV_BLOCKSIZE, | |
abbb3b8e DS |
3472 | BTRFS_SUPER_INFO_SIZE); |
3473 | if (!bh) { | |
3474 | errors++; | |
3475 | continue; | |
3476 | } | |
3477 | wait_on_buffer(bh); | |
3478 | if (!buffer_uptodate(bh)) | |
3479 | errors++; | |
3480 | ||
3481 | /* drop our reference */ | |
3482 | brelse(bh); | |
3483 | ||
3484 | /* drop the reference from the writing run */ | |
3485 | brelse(bh); | |
3486 | } | |
3487 | ||
3488 | return errors < i ? 0 : -1; | |
3489 | } | |
3490 | ||
387125fc CM |
3491 | /* |
3492 | * endio for the write_dev_flush, this will wake anyone waiting | |
3493 | * for the barrier when it is done | |
3494 | */ | |
4246a0b6 | 3495 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3496 | { |
e0ae9994 | 3497 | complete(bio->bi_private); |
387125fc CM |
3498 | } |
3499 | ||
3500 | /* | |
4fc6441a AJ |
3501 | * Submit a flush request to the device if it supports it. Error handling is |
3502 | * done in the waiting counterpart. | |
387125fc | 3503 | */ |
4fc6441a | 3504 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3505 | { |
c2a9c7ab | 3506 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3507 | struct bio *bio = device->flush_bio; |
387125fc | 3508 | |
c2a9c7ab | 3509 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3510 | return; |
387125fc | 3511 | |
e0ae9994 | 3512 | bio_reset(bio); |
387125fc CM |
3513 | bio->bi_end_io = btrfs_end_empty_barrier; |
3514 | bio->bi_bdev = device->bdev; | |
8d910125 | 3515 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3516 | init_completion(&device->flush_wait); |
3517 | bio->bi_private = &device->flush_wait; | |
387125fc | 3518 | |
43a01111 | 3519 | btrfsic_submit_bio(bio); |
e0ae9994 | 3520 | device->flush_bio_sent = 1; |
4fc6441a | 3521 | } |
387125fc | 3522 | |
4fc6441a AJ |
3523 | /* |
3524 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
3525 | */ | |
8c27cb35 | 3526 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3527 | { |
4fc6441a | 3528 | struct bio *bio = device->flush_bio; |
387125fc | 3529 | |
e0ae9994 | 3530 | if (!device->flush_bio_sent) |
4fc6441a | 3531 | return 0; |
387125fc | 3532 | |
e0ae9994 | 3533 | device->flush_bio_sent = 0; |
2980d574 | 3534 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3535 | |
8c27cb35 | 3536 | return bio->bi_status; |
387125fc | 3537 | } |
387125fc | 3538 | |
d10b82fe | 3539 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 3540 | { |
d10b82fe | 3541 | if (!btrfs_check_rw_degradable(fs_info)) |
401b41e5 | 3542 | return -EIO; |
387125fc CM |
3543 | return 0; |
3544 | } | |
3545 | ||
3546 | /* | |
3547 | * send an empty flush down to each device in parallel, | |
3548 | * then wait for them | |
3549 | */ | |
3550 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3551 | { | |
3552 | struct list_head *head; | |
3553 | struct btrfs_device *dev; | |
5af3e8cc | 3554 | int errors_wait = 0; |
4e4cbee9 | 3555 | blk_status_t ret; |
387125fc CM |
3556 | |
3557 | /* send down all the barriers */ | |
3558 | head = &info->fs_devices->devices; | |
3559 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3560 | if (dev->missing) |
3561 | continue; | |
cea7c8bf | 3562 | if (!dev->bdev) |
387125fc | 3563 | continue; |
387125fc CM |
3564 | if (!dev->in_fs_metadata || !dev->writeable) |
3565 | continue; | |
3566 | ||
4fc6441a | 3567 | write_dev_flush(dev); |
12b9bf0b | 3568 | dev->last_flush_error = 0; |
387125fc CM |
3569 | } |
3570 | ||
3571 | /* wait for all the barriers */ | |
3572 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3573 | if (dev->missing) |
3574 | continue; | |
387125fc | 3575 | if (!dev->bdev) { |
5af3e8cc | 3576 | errors_wait++; |
387125fc CM |
3577 | continue; |
3578 | } | |
3579 | if (!dev->in_fs_metadata || !dev->writeable) | |
3580 | continue; | |
3581 | ||
4fc6441a | 3582 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
3583 | if (ret) { |
3584 | dev->last_flush_error = ret; | |
66b4993e DS |
3585 | btrfs_dev_stat_inc_and_print(dev, |
3586 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 3587 | errors_wait++; |
401b41e5 AJ |
3588 | } |
3589 | } | |
3590 | ||
cea7c8bf | 3591 | if (errors_wait) { |
401b41e5 AJ |
3592 | /* |
3593 | * At some point we need the status of all disks | |
3594 | * to arrive at the volume status. So error checking | |
3595 | * is being pushed to a separate loop. | |
3596 | */ | |
d10b82fe | 3597 | return check_barrier_error(info); |
387125fc | 3598 | } |
387125fc CM |
3599 | return 0; |
3600 | } | |
3601 | ||
943c6e99 ZL |
3602 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3603 | { | |
8789f4fe ZL |
3604 | int raid_type; |
3605 | int min_tolerated = INT_MAX; | |
943c6e99 | 3606 | |
8789f4fe ZL |
3607 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3608 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3609 | min_tolerated = min(min_tolerated, | |
3610 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3611 | tolerated_failures); | |
943c6e99 | 3612 | |
8789f4fe ZL |
3613 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3614 | if (raid_type == BTRFS_RAID_SINGLE) | |
3615 | continue; | |
3616 | if (!(flags & btrfs_raid_group[raid_type])) | |
3617 | continue; | |
3618 | min_tolerated = min(min_tolerated, | |
3619 | btrfs_raid_array[raid_type]. | |
3620 | tolerated_failures); | |
3621 | } | |
943c6e99 | 3622 | |
8789f4fe | 3623 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 3624 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
3625 | min_tolerated = 0; |
3626 | } | |
3627 | ||
3628 | return min_tolerated; | |
943c6e99 ZL |
3629 | } |
3630 | ||
eece6a9c | 3631 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 3632 | { |
e5e9a520 | 3633 | struct list_head *head; |
f2984462 | 3634 | struct btrfs_device *dev; |
a061fc8d | 3635 | struct btrfs_super_block *sb; |
f2984462 | 3636 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3637 | int ret; |
3638 | int do_barriers; | |
a236aed1 CM |
3639 | int max_errors; |
3640 | int total_errors = 0; | |
a061fc8d | 3641 | u64 flags; |
f2984462 | 3642 | |
0b246afa JM |
3643 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
3644 | backup_super_roots(fs_info); | |
f2984462 | 3645 | |
0b246afa | 3646 | sb = fs_info->super_for_commit; |
a061fc8d | 3647 | dev_item = &sb->dev_item; |
e5e9a520 | 3648 | |
0b246afa JM |
3649 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
3650 | head = &fs_info->fs_devices->devices; | |
3651 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 3652 | |
5af3e8cc | 3653 | if (do_barriers) { |
0b246afa | 3654 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
3655 | if (ret) { |
3656 | mutex_unlock( | |
0b246afa JM |
3657 | &fs_info->fs_devices->device_list_mutex); |
3658 | btrfs_handle_fs_error(fs_info, ret, | |
3659 | "errors while submitting device barriers."); | |
5af3e8cc SB |
3660 | return ret; |
3661 | } | |
3662 | } | |
387125fc | 3663 | |
1f78160c | 3664 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3665 | if (!dev->bdev) { |
3666 | total_errors++; | |
3667 | continue; | |
3668 | } | |
2b82032c | 3669 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3670 | continue; |
3671 | ||
2b82032c | 3672 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3673 | btrfs_set_stack_device_type(dev_item, dev->type); |
3674 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3675 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3676 | dev->commit_total_bytes); |
ce7213c7 MX |
3677 | btrfs_set_stack_device_bytes_used(dev_item, |
3678 | dev->commit_bytes_used); | |
a061fc8d CM |
3679 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3680 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3681 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3682 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
44880fdc | 3683 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); |
a512bbf8 | 3684 | |
a061fc8d CM |
3685 | flags = btrfs_super_flags(sb); |
3686 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3687 | ||
abbb3b8e | 3688 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
3689 | if (ret) |
3690 | total_errors++; | |
f2984462 | 3691 | } |
a236aed1 | 3692 | if (total_errors > max_errors) { |
0b246afa JM |
3693 | btrfs_err(fs_info, "%d errors while writing supers", |
3694 | total_errors); | |
3695 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 3696 | |
9d565ba4 | 3697 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
3698 | btrfs_handle_fs_error(fs_info, -EIO, |
3699 | "%d errors while writing supers", | |
3700 | total_errors); | |
9d565ba4 | 3701 | return -EIO; |
a236aed1 | 3702 | } |
f2984462 | 3703 | |
a512bbf8 | 3704 | total_errors = 0; |
1f78160c | 3705 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3706 | if (!dev->bdev) |
3707 | continue; | |
2b82032c | 3708 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3709 | continue; |
3710 | ||
abbb3b8e | 3711 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
3712 | if (ret) |
3713 | total_errors++; | |
f2984462 | 3714 | } |
0b246afa | 3715 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3716 | if (total_errors > max_errors) { |
0b246afa JM |
3717 | btrfs_handle_fs_error(fs_info, -EIO, |
3718 | "%d errors while writing supers", | |
3719 | total_errors); | |
79787eaa | 3720 | return -EIO; |
a236aed1 | 3721 | } |
f2984462 CM |
3722 | return 0; |
3723 | } | |
3724 | ||
cb517eab MX |
3725 | /* Drop a fs root from the radix tree and free it. */ |
3726 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3727 | struct btrfs_root *root) | |
2619ba1f | 3728 | { |
4df27c4d | 3729 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3730 | radix_tree_delete(&fs_info->fs_roots_radix, |
3731 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3732 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3733 | |
3734 | if (btrfs_root_refs(&root->root_item) == 0) | |
3735 | synchronize_srcu(&fs_info->subvol_srcu); | |
3736 | ||
1c1ea4f7 | 3737 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3738 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3739 | if (root->reloc_root) { |
3740 | free_extent_buffer(root->reloc_root->node); | |
3741 | free_extent_buffer(root->reloc_root->commit_root); | |
3742 | btrfs_put_fs_root(root->reloc_root); | |
3743 | root->reloc_root = NULL; | |
3744 | } | |
3745 | } | |
3321719e | 3746 | |
faa2dbf0 JB |
3747 | if (root->free_ino_pinned) |
3748 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3749 | if (root->free_ino_ctl) | |
3750 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3751 | free_fs_root(root); |
4df27c4d YZ |
3752 | } |
3753 | ||
3754 | static void free_fs_root(struct btrfs_root *root) | |
3755 | { | |
57cdc8db | 3756 | iput(root->ino_cache_inode); |
4df27c4d | 3757 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
2ff7e61e | 3758 | btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv); |
1cb048f5 | 3759 | root->orphan_block_rsv = NULL; |
0ee5dc67 AV |
3760 | if (root->anon_dev) |
3761 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3762 | if (root->subv_writers) |
3763 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3764 | free_extent_buffer(root->node); |
3765 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3766 | kfree(root->free_ino_ctl); |
3767 | kfree(root->free_ino_pinned); | |
d397712b | 3768 | kfree(root->name); |
b0feb9d9 | 3769 | btrfs_put_fs_root(root); |
2619ba1f CM |
3770 | } |
3771 | ||
cb517eab MX |
3772 | void btrfs_free_fs_root(struct btrfs_root *root) |
3773 | { | |
3774 | free_fs_root(root); | |
2619ba1f CM |
3775 | } |
3776 | ||
c146afad | 3777 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3778 | { |
c146afad YZ |
3779 | u64 root_objectid = 0; |
3780 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3781 | int i = 0; |
3782 | int err = 0; | |
3783 | unsigned int ret = 0; | |
3784 | int index; | |
e089f05c | 3785 | |
c146afad | 3786 | while (1) { |
65d33fd7 | 3787 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3788 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3789 | (void **)gang, root_objectid, | |
3790 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3791 | if (!ret) { |
3792 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3793 | break; |
65d33fd7 | 3794 | } |
5d4f98a2 | 3795 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3796 | |
c146afad | 3797 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3798 | /* Avoid to grab roots in dead_roots */ |
3799 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3800 | gang[i] = NULL; | |
3801 | continue; | |
3802 | } | |
3803 | /* grab all the search result for later use */ | |
3804 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3805 | } | |
3806 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3807 | |
65d33fd7 QW |
3808 | for (i = 0; i < ret; i++) { |
3809 | if (!gang[i]) | |
3810 | continue; | |
c146afad | 3811 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3812 | err = btrfs_orphan_cleanup(gang[i]); |
3813 | if (err) | |
65d33fd7 QW |
3814 | break; |
3815 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3816 | } |
3817 | root_objectid++; | |
3818 | } | |
65d33fd7 QW |
3819 | |
3820 | /* release the uncleaned roots due to error */ | |
3821 | for (; i < ret; i++) { | |
3822 | if (gang[i]) | |
3823 | btrfs_put_fs_root(gang[i]); | |
3824 | } | |
3825 | return err; | |
c146afad | 3826 | } |
a2135011 | 3827 | |
6bccf3ab | 3828 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 3829 | { |
6bccf3ab | 3830 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 3831 | struct btrfs_trans_handle *trans; |
a74a4b97 | 3832 | |
0b246afa | 3833 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 3834 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
3835 | mutex_unlock(&fs_info->cleaner_mutex); |
3836 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
3837 | |
3838 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
3839 | down_write(&fs_info->cleanup_work_sem); |
3840 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 3841 | |
7a7eaa40 | 3842 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3843 | if (IS_ERR(trans)) |
3844 | return PTR_ERR(trans); | |
3a45bb20 | 3845 | return btrfs_commit_transaction(trans); |
c146afad YZ |
3846 | } |
3847 | ||
6bccf3ab | 3848 | void close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 3849 | { |
6bccf3ab | 3850 | struct btrfs_root *root = fs_info->tree_root; |
c146afad YZ |
3851 | int ret; |
3852 | ||
afcdd129 | 3853 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
c146afad | 3854 | |
7343dd61 | 3855 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3856 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3857 | |
803b2f54 SB |
3858 | /* wait for the uuid_scan task to finish */ |
3859 | down(&fs_info->uuid_tree_rescan_sem); | |
3860 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3861 | up(&fs_info->uuid_tree_rescan_sem); | |
3862 | ||
837d5b6e | 3863 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3864 | btrfs_pause_balance(fs_info); |
837d5b6e | 3865 | |
8dabb742 SB |
3866 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3867 | ||
aa1b8cd4 | 3868 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3869 | |
3870 | /* wait for any defraggers to finish */ | |
3871 | wait_event(fs_info->transaction_wait, | |
3872 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3873 | ||
3874 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3875 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3876 | |
21c7e756 MX |
3877 | cancel_work_sync(&fs_info->async_reclaim_work); |
3878 | ||
c146afad | 3879 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
e44163e1 JM |
3880 | /* |
3881 | * If the cleaner thread is stopped and there are | |
3882 | * block groups queued for removal, the deletion will be | |
3883 | * skipped when we quit the cleaner thread. | |
3884 | */ | |
0b246afa | 3885 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 3886 | |
6bccf3ab | 3887 | ret = btrfs_commit_super(fs_info); |
acce952b | 3888 | if (ret) |
04892340 | 3889 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3890 | } |
3891 | ||
87533c47 | 3892 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
2ff7e61e | 3893 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 3894 | |
e3029d9f AV |
3895 | kthread_stop(fs_info->transaction_kthread); |
3896 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3897 | |
afcdd129 | 3898 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 3899 | |
04892340 | 3900 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3901 | |
963d678b | 3902 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3903 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3904 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3905 | } |
bcc63abb | 3906 | |
6618a59b | 3907 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3908 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3909 | |
faa2dbf0 | 3910 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3911 | |
1a4319cc LB |
3912 | btrfs_put_block_group_cache(fs_info); |
3913 | ||
de348ee0 WS |
3914 | /* |
3915 | * we must make sure there is not any read request to | |
3916 | * submit after we stopping all workers. | |
3917 | */ | |
3918 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3919 | btrfs_stop_all_workers(fs_info); |
3920 | ||
5cdd7db6 FM |
3921 | btrfs_free_block_groups(fs_info); |
3922 | ||
afcdd129 | 3923 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
13e6c37b | 3924 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3925 | |
13e6c37b | 3926 | iput(fs_info->btree_inode); |
d6bfde87 | 3927 | |
21adbd5c | 3928 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3929 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 3930 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
3931 | #endif |
3932 | ||
dfe25020 | 3933 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3934 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3935 | |
e2d84521 | 3936 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3937 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3938 | percpu_counter_destroy(&fs_info->bio_counter); |
76dda93c | 3939 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3940 | |
53b381b3 DW |
3941 | btrfs_free_stripe_hash_table(fs_info); |
3942 | ||
cdfb080e | 3943 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3944 | root->orphan_block_rsv = NULL; |
04216820 | 3945 | |
04216820 FM |
3946 | while (!list_empty(&fs_info->pinned_chunks)) { |
3947 | struct extent_map *em; | |
3948 | ||
3949 | em = list_first_entry(&fs_info->pinned_chunks, | |
3950 | struct extent_map, list); | |
3951 | list_del_init(&em->list); | |
3952 | free_extent_map(em); | |
3953 | } | |
eb60ceac CM |
3954 | } |
3955 | ||
b9fab919 CM |
3956 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3957 | int atomic) | |
5f39d397 | 3958 | { |
1259ab75 | 3959 | int ret; |
727011e0 | 3960 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3961 | |
0b32f4bb | 3962 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3963 | if (!ret) |
3964 | return ret; | |
3965 | ||
3966 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3967 | parent_transid, atomic); |
3968 | if (ret == -EAGAIN) | |
3969 | return ret; | |
1259ab75 | 3970 | return !ret; |
5f39d397 CM |
3971 | } |
3972 | ||
5f39d397 CM |
3973 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3974 | { | |
0b246afa | 3975 | struct btrfs_fs_info *fs_info; |
06ea65a3 | 3976 | struct btrfs_root *root; |
5f39d397 | 3977 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3978 | int was_dirty; |
b4ce94de | 3979 | |
06ea65a3 JB |
3980 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3981 | /* | |
3982 | * This is a fast path so only do this check if we have sanity tests | |
3983 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3984 | * outside of the sanity tests. | |
3985 | */ | |
3986 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3987 | return; | |
3988 | #endif | |
3989 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
0b246afa | 3990 | fs_info = root->fs_info; |
b9447ef8 | 3991 | btrfs_assert_tree_locked(buf); |
0b246afa | 3992 | if (transid != fs_info->generation) |
5d163e0e | 3993 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 3994 | buf->start, transid, fs_info->generation); |
0b32f4bb | 3995 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 3996 | if (!was_dirty) |
104b4e51 NB |
3997 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
3998 | buf->len, | |
3999 | fs_info->dirty_metadata_batch); | |
1f21ef0a FM |
4000 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4001 | if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { | |
a4f78750 | 4002 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4003 | ASSERT(0); |
4004 | } | |
4005 | #endif | |
eb60ceac CM |
4006 | } |
4007 | ||
2ff7e61e | 4008 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4009 | int flush_delayed) |
16cdcec7 MX |
4010 | { |
4011 | /* | |
4012 | * looks as though older kernels can get into trouble with | |
4013 | * this code, they end up stuck in balance_dirty_pages forever | |
4014 | */ | |
e2d84521 | 4015 | int ret; |
16cdcec7 MX |
4016 | |
4017 | if (current->flags & PF_MEMALLOC) | |
4018 | return; | |
4019 | ||
b53d3f5d | 4020 | if (flush_delayed) |
2ff7e61e | 4021 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4022 | |
0b246afa | 4023 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
e2d84521 MX |
4024 | BTRFS_DIRTY_METADATA_THRESH); |
4025 | if (ret > 0) { | |
0b246afa | 4026 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4027 | } |
16cdcec7 MX |
4028 | } |
4029 | ||
2ff7e61e | 4030 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4031 | { |
2ff7e61e | 4032 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4033 | } |
585ad2c3 | 4034 | |
2ff7e61e | 4035 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4036 | { |
2ff7e61e | 4037 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4038 | } |
6b80053d | 4039 | |
ca7a79ad | 4040 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 4041 | { |
727011e0 | 4042 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
2ff7e61e JM |
4043 | struct btrfs_fs_info *fs_info = root->fs_info; |
4044 | ||
4045 | return btree_read_extent_buffer_pages(fs_info, buf, parent_transid); | |
6b80053d | 4046 | } |
0da5468f | 4047 | |
3d3a126a | 4048 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info) |
acce952b | 4049 | { |
c926093e | 4050 | struct btrfs_super_block *sb = fs_info->super_copy; |
319e4d06 QW |
4051 | u64 nodesize = btrfs_super_nodesize(sb); |
4052 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
c926093e DS |
4053 | int ret = 0; |
4054 | ||
319e4d06 | 4055 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { |
ab8d0fc4 | 4056 | btrfs_err(fs_info, "no valid FS found"); |
319e4d06 QW |
4057 | ret = -EINVAL; |
4058 | } | |
4059 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) | |
ab8d0fc4 | 4060 | btrfs_warn(fs_info, "unrecognized super flag: %llu", |
319e4d06 | 4061 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); |
21e7626b | 4062 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 4063 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", |
21e7626b | 4064 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
4065 | ret = -EINVAL; |
4066 | } | |
21e7626b | 4067 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 4068 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", |
21e7626b | 4069 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
4070 | ret = -EINVAL; |
4071 | } | |
21e7626b | 4072 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 4073 | btrfs_err(fs_info, "log_root level too big: %d >= %d", |
21e7626b | 4074 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
4075 | ret = -EINVAL; |
4076 | } | |
4077 | ||
1104a885 | 4078 | /* |
319e4d06 QW |
4079 | * Check sectorsize and nodesize first, other check will need it. |
4080 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
1104a885 | 4081 | */ |
319e4d06 QW |
4082 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || |
4083 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
ab8d0fc4 | 4084 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); |
319e4d06 QW |
4085 | ret = -EINVAL; |
4086 | } | |
4087 | /* Only PAGE SIZE is supported yet */ | |
09cbfeaf | 4088 | if (sectorsize != PAGE_SIZE) { |
ab8d0fc4 JM |
4089 | btrfs_err(fs_info, |
4090 | "sectorsize %llu not supported yet, only support %lu", | |
4091 | sectorsize, PAGE_SIZE); | |
319e4d06 QW |
4092 | ret = -EINVAL; |
4093 | } | |
4094 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
4095 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
ab8d0fc4 | 4096 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); |
319e4d06 QW |
4097 | ret = -EINVAL; |
4098 | } | |
4099 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
ab8d0fc4 JM |
4100 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", |
4101 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
319e4d06 QW |
4102 | ret = -EINVAL; |
4103 | } | |
4104 | ||
4105 | /* Root alignment check */ | |
4106 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
ab8d0fc4 JM |
4107 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", |
4108 | btrfs_super_root(sb)); | |
319e4d06 QW |
4109 | ret = -EINVAL; |
4110 | } | |
4111 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
ab8d0fc4 JM |
4112 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", |
4113 | btrfs_super_chunk_root(sb)); | |
75d6ad38 DS |
4114 | ret = -EINVAL; |
4115 | } | |
319e4d06 | 4116 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { |
ab8d0fc4 JM |
4117 | btrfs_warn(fs_info, "log_root block unaligned: %llu", |
4118 | btrfs_super_log_root(sb)); | |
75d6ad38 DS |
4119 | ret = -EINVAL; |
4120 | } | |
4121 | ||
44880fdc | 4122 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { |
ab8d0fc4 JM |
4123 | btrfs_err(fs_info, |
4124 | "dev_item UUID does not match fsid: %pU != %pU", | |
4125 | fs_info->fsid, sb->dev_item.fsid); | |
c926093e DS |
4126 | ret = -EINVAL; |
4127 | } | |
4128 | ||
4129 | /* | |
4130 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
4131 | * done later | |
4132 | */ | |
99e3ecfc LB |
4133 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { |
4134 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
ab8d0fc4 | 4135 | btrfs_super_bytes_used(sb)); |
99e3ecfc LB |
4136 | ret = -EINVAL; |
4137 | } | |
b7f67055 | 4138 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { |
99e3ecfc | 4139 | btrfs_err(fs_info, "invalid stripesize %u", |
ab8d0fc4 | 4140 | btrfs_super_stripesize(sb)); |
99e3ecfc LB |
4141 | ret = -EINVAL; |
4142 | } | |
21e7626b | 4143 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
ab8d0fc4 JM |
4144 | btrfs_warn(fs_info, "suspicious number of devices: %llu", |
4145 | btrfs_super_num_devices(sb)); | |
75d6ad38 | 4146 | if (btrfs_super_num_devices(sb) == 0) { |
ab8d0fc4 | 4147 | btrfs_err(fs_info, "number of devices is 0"); |
75d6ad38 DS |
4148 | ret = -EINVAL; |
4149 | } | |
c926093e | 4150 | |
21e7626b | 4151 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
ab8d0fc4 JM |
4152 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
4153 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
c926093e DS |
4154 | ret = -EINVAL; |
4155 | } | |
4156 | ||
ce7fca5f DS |
4157 | /* |
4158 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4159 | * and one chunk | |
4160 | */ | |
4161 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
ab8d0fc4 JM |
4162 | btrfs_err(fs_info, "system chunk array too big %u > %u", |
4163 | btrfs_super_sys_array_size(sb), | |
4164 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
ce7fca5f DS |
4165 | ret = -EINVAL; |
4166 | } | |
4167 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4168 | + sizeof(struct btrfs_chunk)) { | |
ab8d0fc4 JM |
4169 | btrfs_err(fs_info, "system chunk array too small %u < %zu", |
4170 | btrfs_super_sys_array_size(sb), | |
4171 | sizeof(struct btrfs_disk_key) | |
4172 | + sizeof(struct btrfs_chunk)); | |
ce7fca5f DS |
4173 | ret = -EINVAL; |
4174 | } | |
4175 | ||
c926093e DS |
4176 | /* |
4177 | * The generation is a global counter, we'll trust it more than the others | |
4178 | * but it's still possible that it's the one that's wrong. | |
4179 | */ | |
21e7626b | 4180 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
ab8d0fc4 JM |
4181 | btrfs_warn(fs_info, |
4182 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
4183 | btrfs_super_generation(sb), | |
4184 | btrfs_super_chunk_root_generation(sb)); | |
21e7626b DS |
4185 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) |
4186 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
ab8d0fc4 JM |
4187 | btrfs_warn(fs_info, |
4188 | "suspicious: generation < cache_generation: %llu < %llu", | |
4189 | btrfs_super_generation(sb), | |
4190 | btrfs_super_cache_generation(sb)); | |
c926093e DS |
4191 | |
4192 | return ret; | |
acce952b | 4193 | } |
4194 | ||
2ff7e61e | 4195 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4196 | { |
0b246afa | 4197 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4198 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4199 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4200 | |
0b246afa JM |
4201 | down_write(&fs_info->cleanup_work_sem); |
4202 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4203 | |
4204 | /* cleanup FS via transaction */ | |
2ff7e61e | 4205 | btrfs_cleanup_transaction(fs_info); |
acce952b | 4206 | } |
4207 | ||
143bede5 | 4208 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4209 | { |
acce952b | 4210 | struct btrfs_ordered_extent *ordered; |
acce952b | 4211 | |
199c2a9c | 4212 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4213 | /* |
4214 | * This will just short circuit the ordered completion stuff which will | |
4215 | * make sure the ordered extent gets properly cleaned up. | |
4216 | */ | |
199c2a9c | 4217 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4218 | root_extent_list) |
4219 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4220 | spin_unlock(&root->ordered_extent_lock); |
4221 | } | |
4222 | ||
4223 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4224 | { | |
4225 | struct btrfs_root *root; | |
4226 | struct list_head splice; | |
4227 | ||
4228 | INIT_LIST_HEAD(&splice); | |
4229 | ||
4230 | spin_lock(&fs_info->ordered_root_lock); | |
4231 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4232 | while (!list_empty(&splice)) { | |
4233 | root = list_first_entry(&splice, struct btrfs_root, | |
4234 | ordered_root); | |
1de2cfde JB |
4235 | list_move_tail(&root->ordered_root, |
4236 | &fs_info->ordered_roots); | |
199c2a9c | 4237 | |
2a85d9ca | 4238 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4239 | btrfs_destroy_ordered_extents(root); |
4240 | ||
2a85d9ca LB |
4241 | cond_resched(); |
4242 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4243 | } |
4244 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4245 | } |
4246 | ||
35a3621b | 4247 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4248 | struct btrfs_fs_info *fs_info) |
acce952b | 4249 | { |
4250 | struct rb_node *node; | |
4251 | struct btrfs_delayed_ref_root *delayed_refs; | |
4252 | struct btrfs_delayed_ref_node *ref; | |
4253 | int ret = 0; | |
4254 | ||
4255 | delayed_refs = &trans->delayed_refs; | |
4256 | ||
4257 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4258 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4259 | spin_unlock(&delayed_refs->lock); |
0b246afa | 4260 | btrfs_info(fs_info, "delayed_refs has NO entry"); |
acce952b | 4261 | return ret; |
4262 | } | |
4263 | ||
d7df2c79 JB |
4264 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4265 | struct btrfs_delayed_ref_head *head; | |
c6fc2454 | 4266 | struct btrfs_delayed_ref_node *tmp; |
e78417d1 | 4267 | bool pin_bytes = false; |
acce952b | 4268 | |
d7df2c79 JB |
4269 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4270 | href_node); | |
4271 | if (!mutex_trylock(&head->mutex)) { | |
6df8cdf5 | 4272 | refcount_inc(&head->node.refs); |
d7df2c79 | 4273 | spin_unlock(&delayed_refs->lock); |
eb12db69 | 4274 | |
d7df2c79 | 4275 | mutex_lock(&head->mutex); |
e78417d1 | 4276 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
4277 | btrfs_put_delayed_ref(&head->node); |
4278 | spin_lock(&delayed_refs->lock); | |
4279 | continue; | |
4280 | } | |
4281 | spin_lock(&head->lock); | |
c6fc2454 QW |
4282 | list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, |
4283 | list) { | |
d7df2c79 | 4284 | ref->in_tree = 0; |
c6fc2454 | 4285 | list_del(&ref->list); |
1d57ee94 WX |
4286 | if (!list_empty(&ref->add_list)) |
4287 | list_del(&ref->add_list); | |
d7df2c79 JB |
4288 | atomic_dec(&delayed_refs->num_entries); |
4289 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4290 | } |
d7df2c79 JB |
4291 | if (head->must_insert_reserved) |
4292 | pin_bytes = true; | |
4293 | btrfs_free_delayed_extent_op(head->extent_op); | |
4294 | delayed_refs->num_heads--; | |
4295 | if (head->processing == 0) | |
4296 | delayed_refs->num_heads_ready--; | |
4297 | atomic_dec(&delayed_refs->num_entries); | |
4298 | head->node.in_tree = 0; | |
4299 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
4300 | spin_unlock(&head->lock); | |
4301 | spin_unlock(&delayed_refs->lock); | |
4302 | mutex_unlock(&head->mutex); | |
acce952b | 4303 | |
d7df2c79 | 4304 | if (pin_bytes) |
2ff7e61e | 4305 | btrfs_pin_extent(fs_info, head->node.bytenr, |
d7df2c79 JB |
4306 | head->node.num_bytes, 1); |
4307 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 4308 | cond_resched(); |
4309 | spin_lock(&delayed_refs->lock); | |
4310 | } | |
4311 | ||
4312 | spin_unlock(&delayed_refs->lock); | |
4313 | ||
4314 | return ret; | |
4315 | } | |
4316 | ||
143bede5 | 4317 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4318 | { |
4319 | struct btrfs_inode *btrfs_inode; | |
4320 | struct list_head splice; | |
4321 | ||
4322 | INIT_LIST_HEAD(&splice); | |
4323 | ||
eb73c1b7 MX |
4324 | spin_lock(&root->delalloc_lock); |
4325 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4326 | |
4327 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4328 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4329 | delalloc_inodes); | |
acce952b | 4330 | |
4331 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4332 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4333 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4334 | spin_unlock(&root->delalloc_lock); |
acce952b | 4335 | |
4336 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4337 | |
eb73c1b7 | 4338 | spin_lock(&root->delalloc_lock); |
acce952b | 4339 | } |
4340 | ||
eb73c1b7 MX |
4341 | spin_unlock(&root->delalloc_lock); |
4342 | } | |
4343 | ||
4344 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4345 | { | |
4346 | struct btrfs_root *root; | |
4347 | struct list_head splice; | |
4348 | ||
4349 | INIT_LIST_HEAD(&splice); | |
4350 | ||
4351 | spin_lock(&fs_info->delalloc_root_lock); | |
4352 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4353 | while (!list_empty(&splice)) { | |
4354 | root = list_first_entry(&splice, struct btrfs_root, | |
4355 | delalloc_root); | |
4356 | list_del_init(&root->delalloc_root); | |
4357 | root = btrfs_grab_fs_root(root); | |
4358 | BUG_ON(!root); | |
4359 | spin_unlock(&fs_info->delalloc_root_lock); | |
4360 | ||
4361 | btrfs_destroy_delalloc_inodes(root); | |
4362 | btrfs_put_fs_root(root); | |
4363 | ||
4364 | spin_lock(&fs_info->delalloc_root_lock); | |
4365 | } | |
4366 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4367 | } |
4368 | ||
2ff7e61e | 4369 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4370 | struct extent_io_tree *dirty_pages, |
4371 | int mark) | |
4372 | { | |
4373 | int ret; | |
acce952b | 4374 | struct extent_buffer *eb; |
4375 | u64 start = 0; | |
4376 | u64 end; | |
acce952b | 4377 | |
4378 | while (1) { | |
4379 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4380 | mark, NULL); |
acce952b | 4381 | if (ret) |
4382 | break; | |
4383 | ||
91166212 | 4384 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4385 | while (start <= end) { |
0b246afa JM |
4386 | eb = find_extent_buffer(fs_info, start); |
4387 | start += fs_info->nodesize; | |
fd8b2b61 | 4388 | if (!eb) |
acce952b | 4389 | continue; |
fd8b2b61 | 4390 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4391 | |
fd8b2b61 JB |
4392 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4393 | &eb->bflags)) | |
4394 | clear_extent_buffer_dirty(eb); | |
4395 | free_extent_buffer_stale(eb); | |
acce952b | 4396 | } |
4397 | } | |
4398 | ||
4399 | return ret; | |
4400 | } | |
4401 | ||
2ff7e61e | 4402 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 4403 | struct extent_io_tree *pinned_extents) |
4404 | { | |
4405 | struct extent_io_tree *unpin; | |
4406 | u64 start; | |
4407 | u64 end; | |
4408 | int ret; | |
ed0eaa14 | 4409 | bool loop = true; |
acce952b | 4410 | |
4411 | unpin = pinned_extents; | |
ed0eaa14 | 4412 | again: |
acce952b | 4413 | while (1) { |
4414 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4415 | EXTENT_DIRTY, NULL); |
acce952b | 4416 | if (ret) |
4417 | break; | |
4418 | ||
af6f8f60 | 4419 | clear_extent_dirty(unpin, start, end); |
2ff7e61e | 4420 | btrfs_error_unpin_extent_range(fs_info, start, end); |
acce952b | 4421 | cond_resched(); |
4422 | } | |
4423 | ||
ed0eaa14 | 4424 | if (loop) { |
0b246afa JM |
4425 | if (unpin == &fs_info->freed_extents[0]) |
4426 | unpin = &fs_info->freed_extents[1]; | |
ed0eaa14 | 4427 | else |
0b246afa | 4428 | unpin = &fs_info->freed_extents[0]; |
ed0eaa14 LB |
4429 | loop = false; |
4430 | goto again; | |
4431 | } | |
4432 | ||
acce952b | 4433 | return 0; |
4434 | } | |
4435 | ||
c79a1751 LB |
4436 | static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) |
4437 | { | |
4438 | struct inode *inode; | |
4439 | ||
4440 | inode = cache->io_ctl.inode; | |
4441 | if (inode) { | |
4442 | invalidate_inode_pages2(inode->i_mapping); | |
4443 | BTRFS_I(inode)->generation = 0; | |
4444 | cache->io_ctl.inode = NULL; | |
4445 | iput(inode); | |
4446 | } | |
4447 | btrfs_put_block_group(cache); | |
4448 | } | |
4449 | ||
4450 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4451 | struct btrfs_fs_info *fs_info) |
c79a1751 LB |
4452 | { |
4453 | struct btrfs_block_group_cache *cache; | |
4454 | ||
4455 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4456 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4457 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
4458 | struct btrfs_block_group_cache, | |
4459 | dirty_list); | |
4460 | if (!cache) { | |
0b246afa | 4461 | btrfs_err(fs_info, "orphan block group dirty_bgs list"); |
c79a1751 LB |
4462 | spin_unlock(&cur_trans->dirty_bgs_lock); |
4463 | return; | |
4464 | } | |
4465 | ||
4466 | if (!list_empty(&cache->io_list)) { | |
4467 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4468 | list_del_init(&cache->io_list); | |
4469 | btrfs_cleanup_bg_io(cache); | |
4470 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4471 | } | |
4472 | ||
4473 | list_del_init(&cache->dirty_list); | |
4474 | spin_lock(&cache->lock); | |
4475 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4476 | spin_unlock(&cache->lock); | |
4477 | ||
4478 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4479 | btrfs_put_block_group(cache); | |
4480 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4481 | } | |
4482 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4483 | ||
4484 | while (!list_empty(&cur_trans->io_bgs)) { | |
4485 | cache = list_first_entry(&cur_trans->io_bgs, | |
4486 | struct btrfs_block_group_cache, | |
4487 | io_list); | |
4488 | if (!cache) { | |
0b246afa | 4489 | btrfs_err(fs_info, "orphan block group on io_bgs list"); |
c79a1751 LB |
4490 | return; |
4491 | } | |
4492 | ||
4493 | list_del_init(&cache->io_list); | |
4494 | spin_lock(&cache->lock); | |
4495 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4496 | spin_unlock(&cache->lock); | |
4497 | btrfs_cleanup_bg_io(cache); | |
4498 | } | |
4499 | } | |
4500 | ||
49b25e05 | 4501 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4502 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4503 | { |
2ff7e61e | 4504 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4505 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4506 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4507 | ||
2ff7e61e | 4508 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4509 | |
4a9d8bde | 4510 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4511 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4512 | |
4a9d8bde | 4513 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4514 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4515 | |
ccdf9b30 JM |
4516 | btrfs_destroy_delayed_inodes(fs_info); |
4517 | btrfs_assert_delayed_root_empty(fs_info); | |
49b25e05 | 4518 | |
2ff7e61e | 4519 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4520 | EXTENT_DIRTY); |
2ff7e61e | 4521 | btrfs_destroy_pinned_extent(fs_info, |
0b246afa | 4522 | fs_info->pinned_extents); |
49b25e05 | 4523 | |
4a9d8bde MX |
4524 | cur_trans->state =TRANS_STATE_COMPLETED; |
4525 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4526 | } |
4527 | ||
2ff7e61e | 4528 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4529 | { |
4530 | struct btrfs_transaction *t; | |
acce952b | 4531 | |
0b246afa | 4532 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4533 | |
0b246afa JM |
4534 | spin_lock(&fs_info->trans_lock); |
4535 | while (!list_empty(&fs_info->trans_list)) { | |
4536 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4537 | struct btrfs_transaction, list); |
4538 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4539 | refcount_inc(&t->use_count); |
0b246afa | 4540 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4541 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4542 | btrfs_put_transaction(t); |
0b246afa | 4543 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4544 | continue; |
4545 | } | |
0b246afa | 4546 | if (t == fs_info->running_transaction) { |
724e2315 | 4547 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4548 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4549 | /* |
4550 | * We wait for 0 num_writers since we don't hold a trans | |
4551 | * handle open currently for this transaction. | |
4552 | */ | |
4553 | wait_event(t->writer_wait, | |
4554 | atomic_read(&t->num_writers) == 0); | |
4555 | } else { | |
0b246afa | 4556 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4557 | } |
2ff7e61e | 4558 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4559 | |
0b246afa JM |
4560 | spin_lock(&fs_info->trans_lock); |
4561 | if (t == fs_info->running_transaction) | |
4562 | fs_info->running_transaction = NULL; | |
acce952b | 4563 | list_del_init(&t->list); |
0b246afa | 4564 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4565 | |
724e2315 | 4566 | btrfs_put_transaction(t); |
2ff7e61e | 4567 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4568 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4569 | } |
0b246afa JM |
4570 | spin_unlock(&fs_info->trans_lock); |
4571 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4572 | btrfs_destroy_delayed_inodes(fs_info); |
4573 | btrfs_assert_delayed_root_empty(fs_info); | |
2ff7e61e | 4574 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
0b246afa JM |
4575 | btrfs_destroy_all_delalloc_inodes(fs_info); |
4576 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
acce952b | 4577 | |
4578 | return 0; | |
4579 | } | |
4580 | ||
c6100a4b JB |
4581 | static struct btrfs_fs_info *btree_fs_info(void *private_data) |
4582 | { | |
4583 | struct inode *inode = private_data; | |
4584 | return btrfs_sb(inode->i_sb); | |
4585 | } | |
4586 | ||
e8c9f186 | 4587 | static const struct extent_io_ops btree_extent_io_ops = { |
4d53dddb | 4588 | /* mandatory callbacks */ |
0b86a832 | 4589 | .submit_bio_hook = btree_submit_bio_hook, |
4d53dddb | 4590 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
239b14b3 CM |
4591 | /* note we're sharing with inode.c for the merge bio hook */ |
4592 | .merge_bio_hook = btrfs_merge_bio_hook, | |
20a7db8a | 4593 | .readpage_io_failed_hook = btree_io_failed_hook, |
c6100a4b JB |
4594 | .set_range_writeback = btrfs_set_range_writeback, |
4595 | .tree_fs_info = btree_fs_info, | |
4d53dddb DS |
4596 | |
4597 | /* optional callbacks */ | |
0da5468f | 4598 | }; |