]>
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> |
4b4e25f2 | 28 | #include <linux/freezer.h> |
5a0e3ad6 | 29 | #include <linux/slab.h> |
784b4e29 | 30 | #include <linux/migrate.h> |
7a36ddec | 31 | #include <linux/ratelimit.h> |
6463fe58 | 32 | #include <linux/uuid.h> |
803b2f54 | 33 | #include <linux/semaphore.h> |
7e75bf3f | 34 | #include <asm/unaligned.h> |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
0b947aff | 37 | #include "hash.h" |
e089f05c | 38 | #include "transaction.h" |
0f7d52f4 | 39 | #include "btrfs_inode.h" |
0b86a832 | 40 | #include "volumes.h" |
db94535d | 41 | #include "print-tree.h" |
925baedd | 42 | #include "locking.h" |
e02119d5 | 43 | #include "tree-log.h" |
fa9c0d79 | 44 | #include "free-space-cache.h" |
70f6d82e | 45 | #include "free-space-tree.h" |
581bb050 | 46 | #include "inode-map.h" |
21adbd5c | 47 | #include "check-integrity.h" |
606686ee | 48 | #include "rcu-string.h" |
8dabb742 | 49 | #include "dev-replace.h" |
53b381b3 | 50 | #include "raid56.h" |
5ac1d209 | 51 | #include "sysfs.h" |
fcebe456 | 52 | #include "qgroup.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); |
fcd1f065 | 67 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 68 | int read_only); |
143bede5 | 69 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 70 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
71 | struct btrfs_root *root); | |
143bede5 | 72 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
acce952b | 73 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
74 | struct extent_io_tree *dirty_pages, | |
75 | int mark); | |
76 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
77 | struct extent_io_tree *pinned_extents); | |
48a3b636 ES |
78 | static int btrfs_cleanup_transaction(struct btrfs_root *root); |
79 | static void btrfs_error_commit_super(struct btrfs_root *root); | |
ce9adaa5 | 80 | |
d352ac68 | 81 | /* |
97eb6b69 DS |
82 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
83 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
84 | * by writes to insert metadata for new file extents after IO is complete. |
85 | */ | |
97eb6b69 | 86 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
87 | struct bio *bio; |
88 | bio_end_io_t *end_io; | |
89 | void *private; | |
90 | struct btrfs_fs_info *info; | |
91 | int error; | |
bfebd8b5 | 92 | enum btrfs_wq_endio_type metadata; |
ce9adaa5 | 93 | struct list_head list; |
8b712842 | 94 | struct btrfs_work work; |
ce9adaa5 | 95 | }; |
0da5468f | 96 | |
97eb6b69 DS |
97 | static struct kmem_cache *btrfs_end_io_wq_cache; |
98 | ||
99 | int __init btrfs_end_io_wq_init(void) | |
100 | { | |
101 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
102 | sizeof(struct btrfs_end_io_wq), | |
103 | 0, | |
104 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, | |
105 | NULL); | |
106 | if (!btrfs_end_io_wq_cache) | |
107 | return -ENOMEM; | |
108 | return 0; | |
109 | } | |
110 | ||
111 | void btrfs_end_io_wq_exit(void) | |
112 | { | |
5598e900 | 113 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
114 | } |
115 | ||
d352ac68 CM |
116 | /* |
117 | * async submit bios are used to offload expensive checksumming | |
118 | * onto the worker threads. They checksum file and metadata bios | |
119 | * just before they are sent down the IO stack. | |
120 | */ | |
44b8bd7e CM |
121 | struct async_submit_bio { |
122 | struct inode *inode; | |
123 | struct bio *bio; | |
124 | struct list_head list; | |
4a69a410 CM |
125 | extent_submit_bio_hook_t *submit_bio_start; |
126 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
127 | int rw; |
128 | int mirror_num; | |
c8b97818 | 129 | unsigned long bio_flags; |
eaf25d93 CM |
130 | /* |
131 | * bio_offset is optional, can be used if the pages in the bio | |
132 | * can't tell us where in the file the bio should go | |
133 | */ | |
134 | u64 bio_offset; | |
8b712842 | 135 | struct btrfs_work work; |
79787eaa | 136 | int error; |
44b8bd7e CM |
137 | }; |
138 | ||
85d4e461 CM |
139 | /* |
140 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
141 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
142 | * the level the eb occupies in the tree. | |
143 | * | |
144 | * Different roots are used for different purposes and may nest inside each | |
145 | * other and they require separate keysets. As lockdep keys should be | |
146 | * static, assign keysets according to the purpose of the root as indicated | |
147 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
148 | * have separate keysets. | |
4008c04a | 149 | * |
85d4e461 CM |
150 | * Lock-nesting across peer nodes is always done with the immediate parent |
151 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
152 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 153 | * |
85d4e461 CM |
154 | * The key is set by the readpage_end_io_hook after the buffer has passed |
155 | * csum validation but before the pages are unlocked. It is also set by | |
156 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 157 | * |
85d4e461 CM |
158 | * We also add a check to make sure the highest level of the tree is the |
159 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
160 | * needs update as well. | |
4008c04a CM |
161 | */ |
162 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
163 | # if BTRFS_MAX_LEVEL != 8 | |
164 | # error | |
165 | # endif | |
85d4e461 CM |
166 | |
167 | static struct btrfs_lockdep_keyset { | |
168 | u64 id; /* root objectid */ | |
169 | const char *name_stem; /* lock name stem */ | |
170 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
171 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
172 | } btrfs_lockdep_keysets[] = { | |
173 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
174 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
175 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
176 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
177 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
178 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 179 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
180 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
181 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
182 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 183 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 184 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 185 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 186 | }; |
85d4e461 CM |
187 | |
188 | void __init btrfs_init_lockdep(void) | |
189 | { | |
190 | int i, j; | |
191 | ||
192 | /* initialize lockdep class names */ | |
193 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
194 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
195 | ||
196 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
197 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
198 | "btrfs-%s-%02d", ks->name_stem, j); | |
199 | } | |
200 | } | |
201 | ||
202 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
203 | int level) | |
204 | { | |
205 | struct btrfs_lockdep_keyset *ks; | |
206 | ||
207 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
208 | ||
209 | /* find the matching keyset, id 0 is the default entry */ | |
210 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
211 | if (ks->id == objectid) | |
212 | break; | |
213 | ||
214 | lockdep_set_class_and_name(&eb->lock, | |
215 | &ks->keys[level], ks->names[level]); | |
216 | } | |
217 | ||
4008c04a CM |
218 | #endif |
219 | ||
d352ac68 CM |
220 | /* |
221 | * extents on the btree inode are pretty simple, there's one extent | |
222 | * that covers the entire device | |
223 | */ | |
b2950863 | 224 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 225 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 226 | int create) |
7eccb903 | 227 | { |
5f39d397 CM |
228 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
229 | struct extent_map *em; | |
230 | int ret; | |
231 | ||
890871be | 232 | read_lock(&em_tree->lock); |
d1310b2e | 233 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
234 | if (em) { |
235 | em->bdev = | |
236 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 237 | read_unlock(&em_tree->lock); |
5f39d397 | 238 | goto out; |
a061fc8d | 239 | } |
890871be | 240 | read_unlock(&em_tree->lock); |
7b13b7b1 | 241 | |
172ddd60 | 242 | em = alloc_extent_map(); |
5f39d397 CM |
243 | if (!em) { |
244 | em = ERR_PTR(-ENOMEM); | |
245 | goto out; | |
246 | } | |
247 | em->start = 0; | |
0afbaf8c | 248 | em->len = (u64)-1; |
c8b97818 | 249 | em->block_len = (u64)-1; |
5f39d397 | 250 | em->block_start = 0; |
a061fc8d | 251 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 252 | |
890871be | 253 | write_lock(&em_tree->lock); |
09a2a8f9 | 254 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
255 | if (ret == -EEXIST) { |
256 | free_extent_map(em); | |
7b13b7b1 | 257 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 258 | if (!em) |
0433f20d | 259 | em = ERR_PTR(-EIO); |
5f39d397 | 260 | } else if (ret) { |
7b13b7b1 | 261 | free_extent_map(em); |
0433f20d | 262 | em = ERR_PTR(ret); |
5f39d397 | 263 | } |
890871be | 264 | write_unlock(&em_tree->lock); |
7b13b7b1 | 265 | |
5f39d397 CM |
266 | out: |
267 | return em; | |
7eccb903 CM |
268 | } |
269 | ||
b0496686 | 270 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 271 | { |
0b947aff | 272 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
273 | } |
274 | ||
275 | void btrfs_csum_final(u32 crc, char *result) | |
276 | { | |
7e75bf3f | 277 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
278 | } |
279 | ||
d352ac68 CM |
280 | /* |
281 | * compute the csum for a btree block, and either verify it or write it | |
282 | * into the csum field of the block. | |
283 | */ | |
01d58472 DD |
284 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
285 | struct extent_buffer *buf, | |
19c00ddc CM |
286 | int verify) |
287 | { | |
01d58472 | 288 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
607d432d | 289 | char *result = NULL; |
19c00ddc CM |
290 | unsigned long len; |
291 | unsigned long cur_len; | |
292 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
293 | char *kaddr; |
294 | unsigned long map_start; | |
295 | unsigned long map_len; | |
296 | int err; | |
297 | u32 crc = ~(u32)0; | |
607d432d | 298 | unsigned long inline_result; |
19c00ddc CM |
299 | |
300 | len = buf->len - offset; | |
d397712b | 301 | while (len > 0) { |
19c00ddc | 302 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 303 | &kaddr, &map_start, &map_len); |
d397712b | 304 | if (err) |
19c00ddc | 305 | return 1; |
19c00ddc | 306 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 307 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
308 | crc, cur_len); |
309 | len -= cur_len; | |
310 | offset += cur_len; | |
19c00ddc | 311 | } |
607d432d | 312 | if (csum_size > sizeof(inline_result)) { |
31e818fe | 313 | result = kzalloc(csum_size, GFP_NOFS); |
607d432d JB |
314 | if (!result) |
315 | return 1; | |
316 | } else { | |
317 | result = (char *)&inline_result; | |
318 | } | |
319 | ||
19c00ddc CM |
320 | btrfs_csum_final(crc, result); |
321 | ||
322 | if (verify) { | |
607d432d | 323 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
324 | u32 val; |
325 | u32 found = 0; | |
607d432d | 326 | memcpy(&found, result, csum_size); |
e4204ded | 327 | |
607d432d | 328 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 DS |
329 | btrfs_warn_rl(fs_info, |
330 | "%s checksum verify failed on %llu wanted %X found %X " | |
331 | "level %d", | |
01d58472 | 332 | fs_info->sb->s_id, buf->start, |
efe120a0 | 333 | val, found, btrfs_header_level(buf)); |
607d432d JB |
334 | if (result != (char *)&inline_result) |
335 | kfree(result); | |
19c00ddc CM |
336 | return 1; |
337 | } | |
338 | } else { | |
607d432d | 339 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 340 | } |
607d432d JB |
341 | if (result != (char *)&inline_result) |
342 | kfree(result); | |
19c00ddc CM |
343 | return 0; |
344 | } | |
345 | ||
d352ac68 CM |
346 | /* |
347 | * we can't consider a given block up to date unless the transid of the | |
348 | * block matches the transid in the parent node's pointer. This is how we | |
349 | * detect blocks that either didn't get written at all or got written | |
350 | * in the wrong place. | |
351 | */ | |
1259ab75 | 352 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
353 | struct extent_buffer *eb, u64 parent_transid, |
354 | int atomic) | |
1259ab75 | 355 | { |
2ac55d41 | 356 | struct extent_state *cached_state = NULL; |
1259ab75 | 357 | int ret; |
2755a0de | 358 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
359 | |
360 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
361 | return 0; | |
362 | ||
b9fab919 CM |
363 | if (atomic) |
364 | return -EAGAIN; | |
365 | ||
a26e8c9f JB |
366 | if (need_lock) { |
367 | btrfs_tree_read_lock(eb); | |
368 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
369 | } | |
370 | ||
2ac55d41 | 371 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 372 | &cached_state); |
0b32f4bb | 373 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
374 | btrfs_header_generation(eb) == parent_transid) { |
375 | ret = 0; | |
376 | goto out; | |
377 | } | |
94647322 DS |
378 | btrfs_err_rl(eb->fs_info, |
379 | "parent transid verify failed on %llu wanted %llu found %llu", | |
380 | eb->start, | |
29549aec | 381 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 382 | ret = 1; |
a26e8c9f JB |
383 | |
384 | /* | |
385 | * Things reading via commit roots that don't have normal protection, | |
386 | * like send, can have a really old block in cache that may point at a | |
387 | * block that has been free'd and re-allocated. So don't clear uptodate | |
388 | * if we find an eb that is under IO (dirty/writeback) because we could | |
389 | * end up reading in the stale data and then writing it back out and | |
390 | * making everybody very sad. | |
391 | */ | |
392 | if (!extent_buffer_under_io(eb)) | |
393 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 394 | out: |
2ac55d41 JB |
395 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
396 | &cached_state, GFP_NOFS); | |
472b909f JB |
397 | if (need_lock) |
398 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 399 | return ret; |
1259ab75 CM |
400 | } |
401 | ||
1104a885 DS |
402 | /* |
403 | * Return 0 if the superblock checksum type matches the checksum value of that | |
404 | * algorithm. Pass the raw disk superblock data. | |
405 | */ | |
406 | static int btrfs_check_super_csum(char *raw_disk_sb) | |
407 | { | |
408 | struct btrfs_super_block *disk_sb = | |
409 | (struct btrfs_super_block *)raw_disk_sb; | |
410 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
411 | int ret = 0; | |
412 | ||
413 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
414 | u32 crc = ~(u32)0; | |
415 | const int csum_size = sizeof(crc); | |
416 | char result[csum_size]; | |
417 | ||
418 | /* | |
419 | * The super_block structure does not span the whole | |
420 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
421 | * is filled with zeros and is included in the checkum. | |
422 | */ | |
423 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
424 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
425 | btrfs_csum_final(crc, result); | |
426 | ||
427 | if (memcmp(raw_disk_sb, result, csum_size)) | |
428 | ret = 1; | |
429 | } | |
430 | ||
431 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
efe120a0 | 432 | printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", |
1104a885 DS |
433 | csum_type); |
434 | ret = 1; | |
435 | } | |
436 | ||
437 | return ret; | |
438 | } | |
439 | ||
d352ac68 CM |
440 | /* |
441 | * helper to read a given tree block, doing retries as required when | |
442 | * the checksums don't match and we have alternate mirrors to try. | |
443 | */ | |
f188591e CM |
444 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
445 | struct extent_buffer *eb, | |
ca7a79ad | 446 | u64 start, u64 parent_transid) |
f188591e CM |
447 | { |
448 | struct extent_io_tree *io_tree; | |
ea466794 | 449 | int failed = 0; |
f188591e CM |
450 | int ret; |
451 | int num_copies = 0; | |
452 | int mirror_num = 0; | |
ea466794 | 453 | int failed_mirror = 0; |
f188591e | 454 | |
a826d6dc | 455 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
456 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
457 | while (1) { | |
bb82ab88 AJ |
458 | ret = read_extent_buffer_pages(io_tree, eb, start, |
459 | WAIT_COMPLETE, | |
f188591e | 460 | btree_get_extent, mirror_num); |
256dd1bb SB |
461 | if (!ret) { |
462 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 463 | parent_transid, 0)) |
256dd1bb SB |
464 | break; |
465 | else | |
466 | ret = -EIO; | |
467 | } | |
d397712b | 468 | |
a826d6dc JB |
469 | /* |
470 | * This buffer's crc is fine, but its contents are corrupted, so | |
471 | * there is no reason to read the other copies, they won't be | |
472 | * any less wrong. | |
473 | */ | |
474 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
475 | break; |
476 | ||
5d964051 | 477 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 478 | eb->start, eb->len); |
4235298e | 479 | if (num_copies == 1) |
ea466794 | 480 | break; |
4235298e | 481 | |
5cf1ab56 JB |
482 | if (!failed_mirror) { |
483 | failed = 1; | |
484 | failed_mirror = eb->read_mirror; | |
485 | } | |
486 | ||
f188591e | 487 | mirror_num++; |
ea466794 JB |
488 | if (mirror_num == failed_mirror) |
489 | mirror_num++; | |
490 | ||
4235298e | 491 | if (mirror_num > num_copies) |
ea466794 | 492 | break; |
f188591e | 493 | } |
ea466794 | 494 | |
c0901581 | 495 | if (failed && !ret && failed_mirror) |
ea466794 JB |
496 | repair_eb_io_failure(root, eb, failed_mirror); |
497 | ||
498 | return ret; | |
f188591e | 499 | } |
19c00ddc | 500 | |
d352ac68 | 501 | /* |
d397712b CM |
502 | * checksum a dirty tree block before IO. This has extra checks to make sure |
503 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 504 | */ |
d397712b | 505 | |
01d58472 | 506 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 507 | { |
4eee4fa4 | 508 | u64 start = page_offset(page); |
19c00ddc | 509 | u64 found_start; |
19c00ddc | 510 | struct extent_buffer *eb; |
f188591e | 511 | |
4f2de97a JB |
512 | eb = (struct extent_buffer *)page->private; |
513 | if (page != eb->pages[0]) | |
514 | return 0; | |
19c00ddc | 515 | found_start = btrfs_header_bytenr(eb); |
fae7f21c | 516 | if (WARN_ON(found_start != start || !PageUptodate(page))) |
4f2de97a | 517 | return 0; |
01d58472 | 518 | csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
519 | return 0; |
520 | } | |
521 | ||
01d58472 | 522 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
523 | struct extent_buffer *eb) |
524 | { | |
01d58472 | 525 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
2b82032c YZ |
526 | u8 fsid[BTRFS_UUID_SIZE]; |
527 | int ret = 1; | |
528 | ||
0a4e5586 | 529 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
530 | while (fs_devices) { |
531 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
532 | ret = 0; | |
533 | break; | |
534 | } | |
535 | fs_devices = fs_devices->seed; | |
536 | } | |
537 | return ret; | |
538 | } | |
539 | ||
a826d6dc | 540 | #define CORRUPT(reason, eb, root, slot) \ |
efe120a0 FH |
541 | btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ |
542 | "root=%llu, slot=%d", reason, \ | |
c1c9ff7c | 543 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
544 | |
545 | static noinline int check_leaf(struct btrfs_root *root, | |
546 | struct extent_buffer *leaf) | |
547 | { | |
548 | struct btrfs_key key; | |
549 | struct btrfs_key leaf_key; | |
550 | u32 nritems = btrfs_header_nritems(leaf); | |
551 | int slot; | |
552 | ||
553 | if (nritems == 0) | |
554 | return 0; | |
555 | ||
556 | /* Check the 0 item */ | |
557 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
558 | BTRFS_LEAF_DATA_SIZE(root)) { | |
559 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
560 | return -EIO; | |
561 | } | |
562 | ||
563 | /* | |
564 | * Check to make sure each items keys are in the correct order and their | |
565 | * offsets make sense. We only have to loop through nritems-1 because | |
566 | * we check the current slot against the next slot, which verifies the | |
567 | * next slot's offset+size makes sense and that the current's slot | |
568 | * offset is correct. | |
569 | */ | |
570 | for (slot = 0; slot < nritems - 1; slot++) { | |
571 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
572 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
573 | ||
574 | /* Make sure the keys are in the right order */ | |
575 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
576 | CORRUPT("bad key order", leaf, root, slot); | |
577 | return -EIO; | |
578 | } | |
579 | ||
580 | /* | |
581 | * Make sure the offset and ends are right, remember that the | |
582 | * item data starts at the end of the leaf and grows towards the | |
583 | * front. | |
584 | */ | |
585 | if (btrfs_item_offset_nr(leaf, slot) != | |
586 | btrfs_item_end_nr(leaf, slot + 1)) { | |
587 | CORRUPT("slot offset bad", leaf, root, slot); | |
588 | return -EIO; | |
589 | } | |
590 | ||
591 | /* | |
592 | * Check to make sure that we don't point outside of the leaf, | |
593 | * just incase all the items are consistent to eachother, but | |
594 | * all point outside of the leaf. | |
595 | */ | |
596 | if (btrfs_item_end_nr(leaf, slot) > | |
597 | BTRFS_LEAF_DATA_SIZE(root)) { | |
598 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
599 | return -EIO; | |
600 | } | |
601 | } | |
602 | ||
603 | return 0; | |
604 | } | |
605 | ||
facc8a22 MX |
606 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
607 | u64 phy_offset, struct page *page, | |
608 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 609 | { |
ce9adaa5 CM |
610 | u64 found_start; |
611 | int found_level; | |
ce9adaa5 CM |
612 | struct extent_buffer *eb; |
613 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 614 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 615 | int ret = 0; |
727011e0 | 616 | int reads_done; |
ce9adaa5 | 617 | |
ce9adaa5 CM |
618 | if (!page->private) |
619 | goto out; | |
d397712b | 620 | |
4f2de97a | 621 | eb = (struct extent_buffer *)page->private; |
d397712b | 622 | |
0b32f4bb JB |
623 | /* the pending IO might have been the only thing that kept this buffer |
624 | * in memory. Make sure we have a ref for all this other checks | |
625 | */ | |
626 | extent_buffer_get(eb); | |
627 | ||
628 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
629 | if (!reads_done) |
630 | goto err; | |
f188591e | 631 | |
5cf1ab56 | 632 | eb->read_mirror = mirror; |
656f30db | 633 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
634 | ret = -EIO; |
635 | goto err; | |
636 | } | |
637 | ||
ce9adaa5 | 638 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 639 | if (found_start != eb->start) { |
02873e43 ZL |
640 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
641 | found_start, eb->start); | |
f188591e | 642 | ret = -EIO; |
ce9adaa5 CM |
643 | goto err; |
644 | } | |
02873e43 ZL |
645 | if (check_tree_block_fsid(fs_info, eb)) { |
646 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
647 | eb->start); | |
1259ab75 CM |
648 | ret = -EIO; |
649 | goto err; | |
650 | } | |
ce9adaa5 | 651 | found_level = btrfs_header_level(eb); |
1c24c3ce | 652 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
653 | btrfs_err(fs_info, "bad tree block level %d", |
654 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
655 | ret = -EIO; |
656 | goto err; | |
657 | } | |
ce9adaa5 | 658 | |
85d4e461 CM |
659 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
660 | eb, found_level); | |
4008c04a | 661 | |
02873e43 | 662 | ret = csum_tree_block(fs_info, eb, 1); |
a826d6dc | 663 | if (ret) { |
f188591e | 664 | ret = -EIO; |
a826d6dc JB |
665 | goto err; |
666 | } | |
667 | ||
668 | /* | |
669 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
670 | * that we don't try and read the other copies of this block, just | |
671 | * return -EIO. | |
672 | */ | |
673 | if (found_level == 0 && check_leaf(root, eb)) { | |
674 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
675 | ret = -EIO; | |
676 | } | |
ce9adaa5 | 677 | |
0b32f4bb JB |
678 | if (!ret) |
679 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 680 | err: |
79fb65a1 JB |
681 | if (reads_done && |
682 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
02873e43 | 683 | btree_readahead_hook(fs_info, eb, eb->start, ret); |
4bb31e92 | 684 | |
53b381b3 DW |
685 | if (ret) { |
686 | /* | |
687 | * our io error hook is going to dec the io pages | |
688 | * again, we have to make sure it has something | |
689 | * to decrement | |
690 | */ | |
691 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 692 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 693 | } |
0b32f4bb | 694 | free_extent_buffer(eb); |
ce9adaa5 | 695 | out: |
f188591e | 696 | return ret; |
ce9adaa5 CM |
697 | } |
698 | ||
ea466794 | 699 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 700 | { |
4bb31e92 | 701 | struct extent_buffer *eb; |
4bb31e92 | 702 | |
4f2de97a | 703 | eb = (struct extent_buffer *)page->private; |
656f30db | 704 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 705 | eb->read_mirror = failed_mirror; |
53b381b3 | 706 | atomic_dec(&eb->io_pages); |
ea466794 | 707 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
02873e43 | 708 | btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO); |
4bb31e92 AJ |
709 | return -EIO; /* we fixed nothing */ |
710 | } | |
711 | ||
4246a0b6 | 712 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 713 | { |
97eb6b69 | 714 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 715 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
716 | struct btrfs_workqueue *wq; |
717 | btrfs_work_func_t func; | |
ce9adaa5 | 718 | |
ce9adaa5 | 719 | fs_info = end_io_wq->info; |
4246a0b6 | 720 | end_io_wq->error = bio->bi_error; |
d20f7043 | 721 | |
7b6d91da | 722 | if (bio->bi_rw & REQ_WRITE) { |
9e0af237 LB |
723 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
724 | wq = fs_info->endio_meta_write_workers; | |
725 | func = btrfs_endio_meta_write_helper; | |
726 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
727 | wq = fs_info->endio_freespace_worker; | |
728 | func = btrfs_freespace_write_helper; | |
729 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
730 | wq = fs_info->endio_raid56_workers; | |
731 | func = btrfs_endio_raid56_helper; | |
732 | } else { | |
733 | wq = fs_info->endio_write_workers; | |
734 | func = btrfs_endio_write_helper; | |
735 | } | |
d20f7043 | 736 | } else { |
8b110e39 MX |
737 | if (unlikely(end_io_wq->metadata == |
738 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
739 | wq = fs_info->endio_repair_workers; | |
740 | func = btrfs_endio_repair_helper; | |
741 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
742 | wq = fs_info->endio_raid56_workers; |
743 | func = btrfs_endio_raid56_helper; | |
744 | } else if (end_io_wq->metadata) { | |
745 | wq = fs_info->endio_meta_workers; | |
746 | func = btrfs_endio_meta_helper; | |
747 | } else { | |
748 | wq = fs_info->endio_workers; | |
749 | func = btrfs_endio_helper; | |
750 | } | |
d20f7043 | 751 | } |
9e0af237 LB |
752 | |
753 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
754 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
755 | } |
756 | ||
22c59948 | 757 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 758 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 759 | { |
97eb6b69 | 760 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 761 | |
97eb6b69 | 762 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 CM |
763 | if (!end_io_wq) |
764 | return -ENOMEM; | |
765 | ||
766 | end_io_wq->private = bio->bi_private; | |
767 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 768 | end_io_wq->info = info; |
ce9adaa5 CM |
769 | end_io_wq->error = 0; |
770 | end_io_wq->bio = bio; | |
22c59948 | 771 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
772 | |
773 | bio->bi_private = end_io_wq; | |
774 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
775 | return 0; |
776 | } | |
777 | ||
b64a2851 | 778 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 779 | { |
4854ddd0 | 780 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 781 | info->thread_pool_size, |
4854ddd0 CM |
782 | info->fs_devices->open_devices); |
783 | return 256 * limit; | |
784 | } | |
0986fe9e | 785 | |
4a69a410 CM |
786 | static void run_one_async_start(struct btrfs_work *work) |
787 | { | |
4a69a410 | 788 | struct async_submit_bio *async; |
79787eaa | 789 | int ret; |
4a69a410 CM |
790 | |
791 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
792 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
793 | async->mirror_num, async->bio_flags, | |
794 | async->bio_offset); | |
795 | if (ret) | |
796 | async->error = ret; | |
4a69a410 CM |
797 | } |
798 | ||
799 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
800 | { |
801 | struct btrfs_fs_info *fs_info; | |
802 | struct async_submit_bio *async; | |
4854ddd0 | 803 | int limit; |
8b712842 CM |
804 | |
805 | async = container_of(work, struct async_submit_bio, work); | |
806 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 807 | |
b64a2851 | 808 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
809 | limit = limit * 2 / 3; |
810 | ||
ee863954 DS |
811 | /* |
812 | * atomic_dec_return implies a barrier for waitqueue_active | |
813 | */ | |
66657b31 | 814 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 815 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
816 | wake_up(&fs_info->async_submit_wait); |
817 | ||
79787eaa JM |
818 | /* If an error occured we just want to clean up the bio and move on */ |
819 | if (async->error) { | |
4246a0b6 CH |
820 | async->bio->bi_error = async->error; |
821 | bio_endio(async->bio); | |
79787eaa JM |
822 | return; |
823 | } | |
824 | ||
4a69a410 | 825 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
826 | async->mirror_num, async->bio_flags, |
827 | async->bio_offset); | |
4a69a410 CM |
828 | } |
829 | ||
830 | static void run_one_async_free(struct btrfs_work *work) | |
831 | { | |
832 | struct async_submit_bio *async; | |
833 | ||
834 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
835 | kfree(async); |
836 | } | |
837 | ||
44b8bd7e CM |
838 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
839 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 840 | unsigned long bio_flags, |
eaf25d93 | 841 | u64 bio_offset, |
4a69a410 CM |
842 | extent_submit_bio_hook_t *submit_bio_start, |
843 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
844 | { |
845 | struct async_submit_bio *async; | |
846 | ||
847 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
848 | if (!async) | |
849 | return -ENOMEM; | |
850 | ||
851 | async->inode = inode; | |
852 | async->rw = rw; | |
853 | async->bio = bio; | |
854 | async->mirror_num = mirror_num; | |
4a69a410 CM |
855 | async->submit_bio_start = submit_bio_start; |
856 | async->submit_bio_done = submit_bio_done; | |
857 | ||
9e0af237 | 858 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 859 | run_one_async_done, run_one_async_free); |
4a69a410 | 860 | |
c8b97818 | 861 | async->bio_flags = bio_flags; |
eaf25d93 | 862 | async->bio_offset = bio_offset; |
8c8bee1d | 863 | |
79787eaa JM |
864 | async->error = 0; |
865 | ||
cb03c743 | 866 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 867 | |
7b6d91da | 868 | if (rw & REQ_SYNC) |
5cdc7ad3 | 869 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 870 | |
5cdc7ad3 | 871 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 872 | |
d397712b | 873 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
874 | atomic_read(&fs_info->nr_async_submits)) { |
875 | wait_event(fs_info->async_submit_wait, | |
876 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
877 | } | |
878 | ||
44b8bd7e CM |
879 | return 0; |
880 | } | |
881 | ||
ce3ed71a CM |
882 | static int btree_csum_one_bio(struct bio *bio) |
883 | { | |
2c30c71b | 884 | struct bio_vec *bvec; |
ce3ed71a | 885 | struct btrfs_root *root; |
2c30c71b | 886 | int i, ret = 0; |
ce3ed71a | 887 | |
2c30c71b | 888 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 889 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 890 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
891 | if (ret) |
892 | break; | |
ce3ed71a | 893 | } |
2c30c71b | 894 | |
79787eaa | 895 | return ret; |
ce3ed71a CM |
896 | } |
897 | ||
4a69a410 CM |
898 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
899 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
900 | unsigned long bio_flags, |
901 | u64 bio_offset) | |
22c59948 | 902 | { |
8b712842 CM |
903 | /* |
904 | * when we're called for a write, we're already in the async | |
5443be45 | 905 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 906 | */ |
79787eaa | 907 | return btree_csum_one_bio(bio); |
4a69a410 | 908 | } |
22c59948 | 909 | |
4a69a410 | 910 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
911 | int mirror_num, unsigned long bio_flags, |
912 | u64 bio_offset) | |
4a69a410 | 913 | { |
61891923 SB |
914 | int ret; |
915 | ||
8b712842 | 916 | /* |
4a69a410 CM |
917 | * when we're called for a write, we're already in the async |
918 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 919 | */ |
61891923 | 920 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
4246a0b6 CH |
921 | if (ret) { |
922 | bio->bi_error = ret; | |
923 | bio_endio(bio); | |
924 | } | |
61891923 | 925 | return ret; |
0b86a832 CM |
926 | } |
927 | ||
de0022b9 JB |
928 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
929 | { | |
930 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
931 | return 0; | |
932 | #ifdef CONFIG_X86 | |
362f924b | 933 | if (static_cpu_has_safe(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
934 | return 0; |
935 | #endif | |
936 | return 1; | |
937 | } | |
938 | ||
44b8bd7e | 939 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
940 | int mirror_num, unsigned long bio_flags, |
941 | u64 bio_offset) | |
44b8bd7e | 942 | { |
de0022b9 | 943 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
944 | int ret; |
945 | ||
7b6d91da | 946 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
947 | /* |
948 | * called for a read, do the setup so that checksum validation | |
949 | * can happen in the async kernel threads | |
950 | */ | |
f3f266ab | 951 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
bfebd8b5 | 952 | bio, BTRFS_WQ_ENDIO_METADATA); |
1d4284bd | 953 | if (ret) |
61891923 SB |
954 | goto out_w_error; |
955 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
956 | mirror_num, 0); | |
de0022b9 JB |
957 | } else if (!async) { |
958 | ret = btree_csum_one_bio(bio); | |
959 | if (ret) | |
61891923 SB |
960 | goto out_w_error; |
961 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
962 | mirror_num, 0); | |
963 | } else { | |
964 | /* | |
965 | * kthread helpers are used to submit writes so that | |
966 | * checksumming can happen in parallel across all CPUs | |
967 | */ | |
968 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
969 | inode, rw, bio, mirror_num, 0, | |
970 | bio_offset, | |
971 | __btree_submit_bio_start, | |
972 | __btree_submit_bio_done); | |
44b8bd7e | 973 | } |
d313d7a3 | 974 | |
4246a0b6 CH |
975 | if (ret) |
976 | goto out_w_error; | |
977 | return 0; | |
978 | ||
61891923 | 979 | out_w_error: |
4246a0b6 CH |
980 | bio->bi_error = ret; |
981 | bio_endio(bio); | |
61891923 | 982 | return ret; |
44b8bd7e CM |
983 | } |
984 | ||
3dd1462e | 985 | #ifdef CONFIG_MIGRATION |
784b4e29 | 986 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
987 | struct page *newpage, struct page *page, |
988 | enum migrate_mode mode) | |
784b4e29 CM |
989 | { |
990 | /* | |
991 | * we can't safely write a btree page from here, | |
992 | * we haven't done the locking hook | |
993 | */ | |
994 | if (PageDirty(page)) | |
995 | return -EAGAIN; | |
996 | /* | |
997 | * Buffers may be managed in a filesystem specific way. | |
998 | * We must have no buffers or drop them. | |
999 | */ | |
1000 | if (page_has_private(page) && | |
1001 | !try_to_release_page(page, GFP_KERNEL)) | |
1002 | return -EAGAIN; | |
a6bc32b8 | 1003 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 1004 | } |
3dd1462e | 1005 | #endif |
784b4e29 | 1006 | |
0da5468f CM |
1007 | |
1008 | static int btree_writepages(struct address_space *mapping, | |
1009 | struct writeback_control *wbc) | |
1010 | { | |
e2d84521 MX |
1011 | struct btrfs_fs_info *fs_info; |
1012 | int ret; | |
1013 | ||
d8d5f3e1 | 1014 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1015 | |
1016 | if (wbc->for_kupdate) | |
1017 | return 0; | |
1018 | ||
e2d84521 | 1019 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1020 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
1021 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1022 | BTRFS_DIRTY_METADATA_THRESH); | |
1023 | if (ret < 0) | |
793955bc | 1024 | return 0; |
793955bc | 1025 | } |
0b32f4bb | 1026 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1027 | } |
1028 | ||
b2950863 | 1029 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 1030 | { |
d1310b2e CM |
1031 | struct extent_io_tree *tree; |
1032 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 1033 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 1034 | } |
22b0ebda | 1035 | |
70dec807 | 1036 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1037 | { |
98509cfc | 1038 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1039 | return 0; |
0c4e538b | 1040 | |
f7a52a40 | 1041 | return try_release_extent_buffer(page); |
d98237b3 CM |
1042 | } |
1043 | ||
d47992f8 LC |
1044 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1045 | unsigned int length) | |
d98237b3 | 1046 | { |
d1310b2e CM |
1047 | struct extent_io_tree *tree; |
1048 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1049 | extent_invalidatepage(tree, page, offset); |
1050 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1051 | if (PagePrivate(page)) { |
efe120a0 FH |
1052 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1053 | "page private not zero on page %llu", | |
1054 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1055 | ClearPagePrivate(page); |
1056 | set_page_private(page, 0); | |
1057 | page_cache_release(page); | |
1058 | } | |
d98237b3 CM |
1059 | } |
1060 | ||
0b32f4bb JB |
1061 | static int btree_set_page_dirty(struct page *page) |
1062 | { | |
bb146eb2 | 1063 | #ifdef DEBUG |
0b32f4bb JB |
1064 | struct extent_buffer *eb; |
1065 | ||
1066 | BUG_ON(!PagePrivate(page)); | |
1067 | eb = (struct extent_buffer *)page->private; | |
1068 | BUG_ON(!eb); | |
1069 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1070 | BUG_ON(!atomic_read(&eb->refs)); | |
1071 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1072 | #endif |
0b32f4bb JB |
1073 | return __set_page_dirty_nobuffers(page); |
1074 | } | |
1075 | ||
7f09410b | 1076 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1077 | .readpage = btree_readpage, |
0da5468f | 1078 | .writepages = btree_writepages, |
5f39d397 CM |
1079 | .releasepage = btree_releasepage, |
1080 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1081 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1082 | .migratepage = btree_migratepage, |
5a92bc88 | 1083 | #endif |
0b32f4bb | 1084 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1085 | }; |
1086 | ||
d3e46fea | 1087 | void readahead_tree_block(struct btrfs_root *root, u64 bytenr) |
090d1875 | 1088 | { |
5f39d397 CM |
1089 | struct extent_buffer *buf = NULL; |
1090 | struct inode *btree_inode = root->fs_info->btree_inode; | |
090d1875 | 1091 | |
a83fffb7 | 1092 | buf = btrfs_find_create_tree_block(root, bytenr); |
5f39d397 | 1093 | if (!buf) |
6197d86e | 1094 | return; |
d1310b2e | 1095 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1096 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1097 | free_extent_buffer(buf); |
090d1875 CM |
1098 | } |
1099 | ||
c0dcaa4d | 1100 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, |
ab0fff03 AJ |
1101 | int mirror_num, struct extent_buffer **eb) |
1102 | { | |
1103 | struct extent_buffer *buf = NULL; | |
1104 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1105 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1106 | int ret; | |
1107 | ||
a83fffb7 | 1108 | buf = btrfs_find_create_tree_block(root, bytenr); |
ab0fff03 AJ |
1109 | if (!buf) |
1110 | return 0; | |
1111 | ||
1112 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1113 | ||
1114 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1115 | btree_get_extent, mirror_num); | |
1116 | if (ret) { | |
1117 | free_extent_buffer(buf); | |
1118 | return ret; | |
1119 | } | |
1120 | ||
1121 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1122 | free_extent_buffer(buf); | |
1123 | return -EIO; | |
0b32f4bb | 1124 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1125 | *eb = buf; |
1126 | } else { | |
1127 | free_extent_buffer(buf); | |
1128 | } | |
1129 | return 0; | |
1130 | } | |
1131 | ||
01d58472 | 1132 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, |
0308af44 | 1133 | u64 bytenr) |
0999df54 | 1134 | { |
01d58472 | 1135 | return find_extent_buffer(fs_info, bytenr); |
0999df54 CM |
1136 | } |
1137 | ||
1138 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
a83fffb7 | 1139 | u64 bytenr) |
0999df54 | 1140 | { |
fccb84c9 | 1141 | if (btrfs_test_is_dummy_root(root)) |
ce3e6984 DS |
1142 | return alloc_test_extent_buffer(root->fs_info, bytenr); |
1143 | return alloc_extent_buffer(root->fs_info, bytenr); | |
0999df54 CM |
1144 | } |
1145 | ||
1146 | ||
e02119d5 CM |
1147 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1148 | { | |
727011e0 | 1149 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1150 | buf->start + buf->len - 1); |
e02119d5 CM |
1151 | } |
1152 | ||
1153 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1154 | { | |
727011e0 | 1155 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1156 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1157 | } |
1158 | ||
0999df54 | 1159 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ce86cd59 | 1160 | u64 parent_transid) |
0999df54 CM |
1161 | { |
1162 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1163 | int ret; |
1164 | ||
a83fffb7 | 1165 | buf = btrfs_find_create_tree_block(root, bytenr); |
0999df54 | 1166 | if (!buf) |
64c043de | 1167 | return ERR_PTR(-ENOMEM); |
0999df54 | 1168 | |
ca7a79ad | 1169 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1170 | if (ret) { |
1171 | free_extent_buffer(buf); | |
64c043de | 1172 | return ERR_PTR(ret); |
0f0fe8f7 | 1173 | } |
5f39d397 | 1174 | return buf; |
ce9adaa5 | 1175 | |
eb60ceac CM |
1176 | } |
1177 | ||
01d58472 DD |
1178 | void clean_tree_block(struct btrfs_trans_handle *trans, |
1179 | struct btrfs_fs_info *fs_info, | |
d5c13f92 | 1180 | struct extent_buffer *buf) |
ed2ff2cb | 1181 | { |
55c69072 | 1182 | if (btrfs_header_generation(buf) == |
e2d84521 | 1183 | fs_info->running_transaction->transid) { |
b9447ef8 | 1184 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1185 | |
b9473439 | 1186 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1187 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1188 | -buf->len, | |
1189 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1190 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1191 | btrfs_set_lock_blocking(buf); | |
1192 | clear_extent_buffer_dirty(buf); | |
1193 | } | |
925baedd | 1194 | } |
5f39d397 CM |
1195 | } |
1196 | ||
8257b2dc MX |
1197 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1198 | { | |
1199 | struct btrfs_subvolume_writers *writers; | |
1200 | int ret; | |
1201 | ||
1202 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1203 | if (!writers) | |
1204 | return ERR_PTR(-ENOMEM); | |
1205 | ||
908c7f19 | 1206 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1207 | if (ret < 0) { |
1208 | kfree(writers); | |
1209 | return ERR_PTR(ret); | |
1210 | } | |
1211 | ||
1212 | init_waitqueue_head(&writers->wait); | |
1213 | return writers; | |
1214 | } | |
1215 | ||
1216 | static void | |
1217 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1218 | { | |
1219 | percpu_counter_destroy(&writers->counter); | |
1220 | kfree(writers); | |
1221 | } | |
1222 | ||
707e8a07 DS |
1223 | static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, |
1224 | struct btrfs_root *root, struct btrfs_fs_info *fs_info, | |
143bede5 | 1225 | u64 objectid) |
d97e63b6 | 1226 | { |
cfaa7295 | 1227 | root->node = NULL; |
a28ec197 | 1228 | root->commit_root = NULL; |
db94535d CM |
1229 | root->sectorsize = sectorsize; |
1230 | root->nodesize = nodesize; | |
87ee04eb | 1231 | root->stripesize = stripesize; |
27cdeb70 | 1232 | root->state = 0; |
d68fc57b | 1233 | root->orphan_cleanup_state = 0; |
0b86a832 | 1234 | |
0f7d52f4 CM |
1235 | root->objectid = objectid; |
1236 | root->last_trans = 0; | |
13a8a7c8 | 1237 | root->highest_objectid = 0; |
eb73c1b7 | 1238 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1239 | root->nr_ordered_extents = 0; |
58176a96 | 1240 | root->name = NULL; |
6bef4d31 | 1241 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1242 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1243 | root->block_rsv = NULL; |
d68fc57b | 1244 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1245 | |
1246 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1247 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1248 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1249 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1250 | INIT_LIST_HEAD(&root->ordered_extents); |
1251 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1252 | INIT_LIST_HEAD(&root->logged_list[0]); |
1253 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1254 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1255 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1256 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1257 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1258 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1259 | spin_lock_init(&root->log_extents_lock[0]); |
1260 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1261 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1262 | mutex_init(&root->log_mutex); |
31f3d255 | 1263 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1264 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1265 | init_waitqueue_head(&root->log_writer_wait); |
1266 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1267 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1268 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1269 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1270 | atomic_set(&root->log_commit[0], 0); |
1271 | atomic_set(&root->log_commit[1], 0); | |
1272 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1273 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1274 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1275 | atomic_set(&root->refs, 1); |
8257b2dc | 1276 | atomic_set(&root->will_be_snapshoted, 0); |
55eeaf05 | 1277 | atomic_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1278 | root->log_transid = 0; |
d1433deb | 1279 | root->log_transid_committed = -1; |
257c62e1 | 1280 | root->last_log_commit = 0; |
06ea65a3 JB |
1281 | if (fs_info) |
1282 | extent_io_tree_init(&root->dirty_log_pages, | |
1283 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1284 | |
3768f368 CM |
1285 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1286 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1287 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
06ea65a3 JB |
1288 | if (fs_info) |
1289 | root->defrag_trans_start = fs_info->generation; | |
1290 | else | |
1291 | root->defrag_trans_start = 0; | |
4d775673 | 1292 | root->root_key.objectid = objectid; |
0ee5dc67 | 1293 | root->anon_dev = 0; |
8ea05e3a | 1294 | |
5f3ab90a | 1295 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1296 | } |
1297 | ||
74e4d827 DS |
1298 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1299 | gfp_t flags) | |
6f07e42e | 1300 | { |
74e4d827 | 1301 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1302 | if (root) |
1303 | root->fs_info = fs_info; | |
1304 | return root; | |
1305 | } | |
1306 | ||
06ea65a3 JB |
1307 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1308 | /* Should only be used by the testing infrastructure */ | |
1309 | struct btrfs_root *btrfs_alloc_dummy_root(void) | |
1310 | { | |
1311 | struct btrfs_root *root; | |
1312 | ||
74e4d827 | 1313 | root = btrfs_alloc_root(NULL, GFP_KERNEL); |
06ea65a3 JB |
1314 | if (!root) |
1315 | return ERR_PTR(-ENOMEM); | |
707e8a07 | 1316 | __setup_root(4096, 4096, 4096, root, NULL, 1); |
27cdeb70 | 1317 | set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); |
faa2dbf0 | 1318 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1319 | |
1320 | return root; | |
1321 | } | |
1322 | #endif | |
1323 | ||
20897f5c AJ |
1324 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1325 | struct btrfs_fs_info *fs_info, | |
1326 | u64 objectid) | |
1327 | { | |
1328 | struct extent_buffer *leaf; | |
1329 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1330 | struct btrfs_root *root; | |
1331 | struct btrfs_key key; | |
1332 | int ret = 0; | |
6463fe58 | 1333 | uuid_le uuid; |
20897f5c | 1334 | |
74e4d827 | 1335 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1336 | if (!root) |
1337 | return ERR_PTR(-ENOMEM); | |
1338 | ||
707e8a07 DS |
1339 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1340 | tree_root->stripesize, root, fs_info, objectid); | |
20897f5c AJ |
1341 | root->root_key.objectid = objectid; |
1342 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1343 | root->root_key.offset = 0; | |
1344 | ||
4d75f8a9 | 1345 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1346 | if (IS_ERR(leaf)) { |
1347 | ret = PTR_ERR(leaf); | |
1dd05682 | 1348 | leaf = NULL; |
20897f5c AJ |
1349 | goto fail; |
1350 | } | |
1351 | ||
20897f5c AJ |
1352 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1353 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1354 | btrfs_set_header_generation(leaf, trans->transid); | |
1355 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1356 | btrfs_set_header_owner(leaf, objectid); | |
1357 | root->node = leaf; | |
1358 | ||
0a4e5586 | 1359 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1360 | BTRFS_FSID_SIZE); |
1361 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1362 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1363 | BTRFS_UUID_SIZE); |
1364 | btrfs_mark_buffer_dirty(leaf); | |
1365 | ||
1366 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1367 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1368 | |
1369 | root->root_item.flags = 0; | |
1370 | root->root_item.byte_limit = 0; | |
1371 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1372 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1373 | btrfs_set_root_level(&root->root_item, 0); | |
1374 | btrfs_set_root_refs(&root->root_item, 1); | |
1375 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1376 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1377 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1378 | uuid_le_gen(&uuid); |
1379 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1380 | root->root_item.drop_level = 0; |
1381 | ||
1382 | key.objectid = objectid; | |
1383 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1384 | key.offset = 0; | |
1385 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1386 | if (ret) | |
1387 | goto fail; | |
1388 | ||
1389 | btrfs_tree_unlock(leaf); | |
1390 | ||
1dd05682 TI |
1391 | return root; |
1392 | ||
20897f5c | 1393 | fail: |
1dd05682 TI |
1394 | if (leaf) { |
1395 | btrfs_tree_unlock(leaf); | |
59885b39 | 1396 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1397 | free_extent_buffer(leaf); |
1398 | } | |
1399 | kfree(root); | |
20897f5c | 1400 | |
1dd05682 | 1401 | return ERR_PTR(ret); |
20897f5c AJ |
1402 | } |
1403 | ||
7237f183 YZ |
1404 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1405 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1406 | { |
1407 | struct btrfs_root *root; | |
1408 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1409 | struct extent_buffer *leaf; |
e02119d5 | 1410 | |
74e4d827 | 1411 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1412 | if (!root) |
7237f183 | 1413 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1414 | |
707e8a07 DS |
1415 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1416 | tree_root->stripesize, root, fs_info, | |
1417 | BTRFS_TREE_LOG_OBJECTID); | |
e02119d5 CM |
1418 | |
1419 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1420 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1421 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1422 | |
7237f183 | 1423 | /* |
27cdeb70 MX |
1424 | * DON'T set REF_COWS for log trees |
1425 | * | |
7237f183 YZ |
1426 | * log trees do not get reference counted because they go away |
1427 | * before a real commit is actually done. They do store pointers | |
1428 | * to file data extents, and those reference counts still get | |
1429 | * updated (along with back refs to the log tree). | |
1430 | */ | |
e02119d5 | 1431 | |
4d75f8a9 DS |
1432 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1433 | NULL, 0, 0, 0); | |
7237f183 YZ |
1434 | if (IS_ERR(leaf)) { |
1435 | kfree(root); | |
1436 | return ERR_CAST(leaf); | |
1437 | } | |
e02119d5 | 1438 | |
5d4f98a2 YZ |
1439 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1440 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1441 | btrfs_set_header_generation(leaf, trans->transid); | |
1442 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1443 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1444 | root->node = leaf; |
e02119d5 CM |
1445 | |
1446 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1447 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1448 | btrfs_mark_buffer_dirty(root->node); |
1449 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1450 | return root; |
1451 | } | |
1452 | ||
1453 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1454 | struct btrfs_fs_info *fs_info) | |
1455 | { | |
1456 | struct btrfs_root *log_root; | |
1457 | ||
1458 | log_root = alloc_log_tree(trans, fs_info); | |
1459 | if (IS_ERR(log_root)) | |
1460 | return PTR_ERR(log_root); | |
1461 | WARN_ON(fs_info->log_root_tree); | |
1462 | fs_info->log_root_tree = log_root; | |
1463 | return 0; | |
1464 | } | |
1465 | ||
1466 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1467 | struct btrfs_root *root) | |
1468 | { | |
1469 | struct btrfs_root *log_root; | |
1470 | struct btrfs_inode_item *inode_item; | |
1471 | ||
1472 | log_root = alloc_log_tree(trans, root->fs_info); | |
1473 | if (IS_ERR(log_root)) | |
1474 | return PTR_ERR(log_root); | |
1475 | ||
1476 | log_root->last_trans = trans->transid; | |
1477 | log_root->root_key.offset = root->root_key.objectid; | |
1478 | ||
1479 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1480 | btrfs_set_stack_inode_generation(inode_item, 1); |
1481 | btrfs_set_stack_inode_size(inode_item, 3); | |
1482 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 1483 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 1484 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1485 | |
5d4f98a2 | 1486 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1487 | |
1488 | WARN_ON(root->log_root); | |
1489 | root->log_root = log_root; | |
1490 | root->log_transid = 0; | |
d1433deb | 1491 | root->log_transid_committed = -1; |
257c62e1 | 1492 | root->last_log_commit = 0; |
e02119d5 CM |
1493 | return 0; |
1494 | } | |
1495 | ||
35a3621b SB |
1496 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1497 | struct btrfs_key *key) | |
e02119d5 CM |
1498 | { |
1499 | struct btrfs_root *root; | |
1500 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1501 | struct btrfs_path *path; |
84234f3a | 1502 | u64 generation; |
cb517eab | 1503 | int ret; |
0f7d52f4 | 1504 | |
cb517eab MX |
1505 | path = btrfs_alloc_path(); |
1506 | if (!path) | |
0f7d52f4 | 1507 | return ERR_PTR(-ENOMEM); |
cb517eab | 1508 | |
74e4d827 | 1509 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1510 | if (!root) { |
1511 | ret = -ENOMEM; | |
1512 | goto alloc_fail; | |
0f7d52f4 CM |
1513 | } |
1514 | ||
707e8a07 DS |
1515 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1516 | tree_root->stripesize, root, fs_info, key->objectid); | |
0f7d52f4 | 1517 | |
cb517eab MX |
1518 | ret = btrfs_find_root(tree_root, key, path, |
1519 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1520 | if (ret) { |
13a8a7c8 YZ |
1521 | if (ret > 0) |
1522 | ret = -ENOENT; | |
cb517eab | 1523 | goto find_fail; |
0f7d52f4 | 1524 | } |
13a8a7c8 | 1525 | |
84234f3a | 1526 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1527 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
ce86cd59 | 1528 | generation); |
64c043de LB |
1529 | if (IS_ERR(root->node)) { |
1530 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1531 | goto find_fail; |
1532 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1533 | ret = -EIO; | |
64c043de LB |
1534 | free_extent_buffer(root->node); |
1535 | goto find_fail; | |
416bc658 | 1536 | } |
5d4f98a2 | 1537 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1538 | out: |
cb517eab MX |
1539 | btrfs_free_path(path); |
1540 | return root; | |
1541 | ||
cb517eab MX |
1542 | find_fail: |
1543 | kfree(root); | |
1544 | alloc_fail: | |
1545 | root = ERR_PTR(ret); | |
1546 | goto out; | |
1547 | } | |
1548 | ||
1549 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1550 | struct btrfs_key *location) | |
1551 | { | |
1552 | struct btrfs_root *root; | |
1553 | ||
1554 | root = btrfs_read_tree_root(tree_root, location); | |
1555 | if (IS_ERR(root)) | |
1556 | return root; | |
1557 | ||
1558 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1559 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1560 | btrfs_check_and_init_root_item(&root->root_item); |
1561 | } | |
13a8a7c8 | 1562 | |
5eda7b5e CM |
1563 | return root; |
1564 | } | |
1565 | ||
cb517eab MX |
1566 | int btrfs_init_fs_root(struct btrfs_root *root) |
1567 | { | |
1568 | int ret; | |
8257b2dc | 1569 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1570 | |
1571 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1572 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1573 | GFP_NOFS); | |
1574 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1575 | ret = -ENOMEM; | |
1576 | goto fail; | |
1577 | } | |
1578 | ||
8257b2dc MX |
1579 | writers = btrfs_alloc_subvolume_writers(); |
1580 | if (IS_ERR(writers)) { | |
1581 | ret = PTR_ERR(writers); | |
1582 | goto fail; | |
1583 | } | |
1584 | root->subv_writers = writers; | |
1585 | ||
cb517eab | 1586 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1587 | spin_lock_init(&root->ino_cache_lock); |
1588 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1589 | |
1590 | ret = get_anon_bdev(&root->anon_dev); | |
1591 | if (ret) | |
8257b2dc | 1592 | goto free_writers; |
f32e48e9 CR |
1593 | |
1594 | mutex_lock(&root->objectid_mutex); | |
1595 | ret = btrfs_find_highest_objectid(root, | |
1596 | &root->highest_objectid); | |
1597 | if (ret) { | |
1598 | mutex_unlock(&root->objectid_mutex); | |
1599 | goto free_root_dev; | |
1600 | } | |
1601 | ||
1602 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1603 | ||
1604 | mutex_unlock(&root->objectid_mutex); | |
1605 | ||
cb517eab | 1606 | return 0; |
8257b2dc | 1607 | |
f32e48e9 CR |
1608 | free_root_dev: |
1609 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
1610 | free_writers: |
1611 | btrfs_free_subvolume_writers(root->subv_writers); | |
cb517eab MX |
1612 | fail: |
1613 | kfree(root->free_ino_ctl); | |
1614 | kfree(root->free_ino_pinned); | |
1615 | return ret; | |
1616 | } | |
1617 | ||
171170c1 ST |
1618 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1619 | u64 root_id) | |
cb517eab MX |
1620 | { |
1621 | struct btrfs_root *root; | |
1622 | ||
1623 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1624 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1625 | (unsigned long)root_id); | |
1626 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1627 | return root; | |
1628 | } | |
1629 | ||
1630 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1631 | struct btrfs_root *root) | |
1632 | { | |
1633 | int ret; | |
1634 | ||
1635 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); | |
1636 | if (ret) | |
1637 | return ret; | |
1638 | ||
1639 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1640 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1641 | (unsigned long)root->root_key.objectid, | |
1642 | root); | |
1643 | if (ret == 0) | |
27cdeb70 | 1644 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1645 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1646 | radix_tree_preload_end(); | |
1647 | ||
1648 | return ret; | |
1649 | } | |
1650 | ||
c00869f1 MX |
1651 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1652 | struct btrfs_key *location, | |
1653 | bool check_ref) | |
5eda7b5e CM |
1654 | { |
1655 | struct btrfs_root *root; | |
381cf658 | 1656 | struct btrfs_path *path; |
1d4c08e0 | 1657 | struct btrfs_key key; |
5eda7b5e CM |
1658 | int ret; |
1659 | ||
edbd8d4e CM |
1660 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1661 | return fs_info->tree_root; | |
1662 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1663 | return fs_info->extent_root; | |
8f18cf13 CM |
1664 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1665 | return fs_info->chunk_root; | |
1666 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1667 | return fs_info->dev_root; | |
0403e47e YZ |
1668 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1669 | return fs_info->csum_root; | |
bcef60f2 AJ |
1670 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1671 | return fs_info->quota_root ? fs_info->quota_root : | |
1672 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1673 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1674 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1675 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1676 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1677 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1678 | ERR_PTR(-ENOENT); | |
4df27c4d | 1679 | again: |
cb517eab | 1680 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1681 | if (root) { |
c00869f1 | 1682 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1683 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1684 | return root; |
48475471 | 1685 | } |
5eda7b5e | 1686 | |
cb517eab | 1687 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1688 | if (IS_ERR(root)) |
1689 | return root; | |
3394e160 | 1690 | |
c00869f1 | 1691 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1692 | ret = -ENOENT; |
581bb050 | 1693 | goto fail; |
35a30d7c | 1694 | } |
581bb050 | 1695 | |
cb517eab | 1696 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1697 | if (ret) |
1698 | goto fail; | |
3394e160 | 1699 | |
381cf658 DS |
1700 | path = btrfs_alloc_path(); |
1701 | if (!path) { | |
1702 | ret = -ENOMEM; | |
1703 | goto fail; | |
1704 | } | |
1d4c08e0 DS |
1705 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1706 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1707 | key.offset = location->objectid; | |
1708 | ||
1709 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1710 | btrfs_free_path(path); |
d68fc57b YZ |
1711 | if (ret < 0) |
1712 | goto fail; | |
1713 | if (ret == 0) | |
27cdeb70 | 1714 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1715 | |
cb517eab | 1716 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1717 | if (ret) { |
4df27c4d YZ |
1718 | if (ret == -EEXIST) { |
1719 | free_fs_root(root); | |
1720 | goto again; | |
1721 | } | |
1722 | goto fail; | |
0f7d52f4 | 1723 | } |
edbd8d4e | 1724 | return root; |
4df27c4d YZ |
1725 | fail: |
1726 | free_fs_root(root); | |
1727 | return ERR_PTR(ret); | |
edbd8d4e CM |
1728 | } |
1729 | ||
04160088 CM |
1730 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1731 | { | |
1732 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1733 | int ret = 0; | |
04160088 CM |
1734 | struct btrfs_device *device; |
1735 | struct backing_dev_info *bdi; | |
b7967db7 | 1736 | |
1f78160c XG |
1737 | rcu_read_lock(); |
1738 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1739 | if (!device->bdev) |
1740 | continue; | |
04160088 | 1741 | bdi = blk_get_backing_dev_info(device->bdev); |
ff9ea323 | 1742 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1743 | ret = 1; |
1744 | break; | |
1745 | } | |
1746 | } | |
1f78160c | 1747 | rcu_read_unlock(); |
04160088 CM |
1748 | return ret; |
1749 | } | |
1750 | ||
04160088 CM |
1751 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1752 | { | |
ad081f14 JA |
1753 | int err; |
1754 | ||
b4caecd4 | 1755 | err = bdi_setup_and_register(bdi, "btrfs"); |
ad081f14 JA |
1756 | if (err) |
1757 | return err; | |
1758 | ||
df0ce26c | 1759 | bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE; |
04160088 CM |
1760 | bdi->congested_fn = btrfs_congested_fn; |
1761 | bdi->congested_data = info; | |
da2f0f74 | 1762 | bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; |
04160088 CM |
1763 | return 0; |
1764 | } | |
1765 | ||
8b712842 CM |
1766 | /* |
1767 | * called by the kthread helper functions to finally call the bio end_io | |
1768 | * functions. This is where read checksum verification actually happens | |
1769 | */ | |
1770 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1771 | { |
ce9adaa5 | 1772 | struct bio *bio; |
97eb6b69 | 1773 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1774 | |
97eb6b69 | 1775 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1776 | bio = end_io_wq->bio; |
ce9adaa5 | 1777 | |
4246a0b6 | 1778 | bio->bi_error = end_io_wq->error; |
8b712842 CM |
1779 | bio->bi_private = end_io_wq->private; |
1780 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1781 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1782 | bio_endio(bio); |
44b8bd7e CM |
1783 | } |
1784 | ||
a74a4b97 CM |
1785 | static int cleaner_kthread(void *arg) |
1786 | { | |
1787 | struct btrfs_root *root = arg; | |
d0278245 | 1788 | int again; |
da288d28 | 1789 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1790 | |
1791 | do { | |
d0278245 | 1792 | again = 0; |
a74a4b97 | 1793 | |
d0278245 | 1794 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1795 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1796 | goto sleep; |
1797 | ||
1798 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) | |
1799 | goto sleep; | |
1800 | ||
dc7f370c MX |
1801 | /* |
1802 | * Avoid the problem that we change the status of the fs | |
1803 | * during the above check and trylock. | |
1804 | */ | |
babbf170 | 1805 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1806 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1807 | goto sleep; | |
76dda93c | 1808 | } |
a74a4b97 | 1809 | |
c2d6cb16 | 1810 | mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); |
d0278245 | 1811 | btrfs_run_delayed_iputs(root); |
c2d6cb16 FM |
1812 | mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); |
1813 | ||
d0278245 MX |
1814 | again = btrfs_clean_one_deleted_snapshot(root); |
1815 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1816 | ||
1817 | /* | |
05323cd1 MX |
1818 | * The defragger has dealt with the R/O remount and umount, |
1819 | * needn't do anything special here. | |
d0278245 MX |
1820 | */ |
1821 | btrfs_run_defrag_inodes(root->fs_info); | |
67c5e7d4 FM |
1822 | |
1823 | /* | |
1824 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1825 | * with relocation (btrfs_relocate_chunk) and relocation | |
1826 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1827 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1828 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1829 | * unused block groups. | |
1830 | */ | |
1831 | btrfs_delete_unused_bgs(root->fs_info); | |
d0278245 | 1832 | sleep: |
9d1a2a3a | 1833 | if (!try_to_freeze() && !again) { |
a74a4b97 | 1834 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1835 | if (!kthread_should_stop()) |
1836 | schedule(); | |
a74a4b97 CM |
1837 | __set_current_state(TASK_RUNNING); |
1838 | } | |
1839 | } while (!kthread_should_stop()); | |
da288d28 FM |
1840 | |
1841 | /* | |
1842 | * Transaction kthread is stopped before us and wakes us up. | |
1843 | * However we might have started a new transaction and COWed some | |
1844 | * tree blocks when deleting unused block groups for example. So | |
1845 | * make sure we commit the transaction we started to have a clean | |
1846 | * shutdown when evicting the btree inode - if it has dirty pages | |
1847 | * when we do the final iput() on it, eviction will trigger a | |
1848 | * writeback for it which will fail with null pointer dereferences | |
1849 | * since work queues and other resources were already released and | |
1850 | * destroyed by the time the iput/eviction/writeback is made. | |
1851 | */ | |
1852 | trans = btrfs_attach_transaction(root); | |
1853 | if (IS_ERR(trans)) { | |
1854 | if (PTR_ERR(trans) != -ENOENT) | |
1855 | btrfs_err(root->fs_info, | |
1856 | "cleaner transaction attach returned %ld", | |
1857 | PTR_ERR(trans)); | |
1858 | } else { | |
1859 | int ret; | |
1860 | ||
1861 | ret = btrfs_commit_transaction(trans, root); | |
1862 | if (ret) | |
1863 | btrfs_err(root->fs_info, | |
1864 | "cleaner open transaction commit returned %d", | |
1865 | ret); | |
1866 | } | |
1867 | ||
a74a4b97 CM |
1868 | return 0; |
1869 | } | |
1870 | ||
1871 | static int transaction_kthread(void *arg) | |
1872 | { | |
1873 | struct btrfs_root *root = arg; | |
1874 | struct btrfs_trans_handle *trans; | |
1875 | struct btrfs_transaction *cur; | |
8929ecfa | 1876 | u64 transid; |
a74a4b97 CM |
1877 | unsigned long now; |
1878 | unsigned long delay; | |
914b2007 | 1879 | bool cannot_commit; |
a74a4b97 CM |
1880 | |
1881 | do { | |
914b2007 | 1882 | cannot_commit = false; |
8b87dc17 | 1883 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1884 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1885 | ||
a4abeea4 | 1886 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1887 | cur = root->fs_info->running_transaction; |
1888 | if (!cur) { | |
a4abeea4 | 1889 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1890 | goto sleep; |
1891 | } | |
31153d81 | 1892 | |
a74a4b97 | 1893 | now = get_seconds(); |
4a9d8bde | 1894 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1895 | (now < cur->start_time || |
1896 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1897 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1898 | delay = HZ * 5; |
1899 | goto sleep; | |
1900 | } | |
8929ecfa | 1901 | transid = cur->transid; |
a4abeea4 | 1902 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1903 | |
79787eaa | 1904 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1905 | trans = btrfs_attach_transaction(root); |
914b2007 | 1906 | if (IS_ERR(trans)) { |
354aa0fb MX |
1907 | if (PTR_ERR(trans) != -ENOENT) |
1908 | cannot_commit = true; | |
79787eaa | 1909 | goto sleep; |
914b2007 | 1910 | } |
8929ecfa | 1911 | if (transid == trans->transid) { |
79787eaa | 1912 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1913 | } else { |
1914 | btrfs_end_transaction(trans, root); | |
1915 | } | |
a74a4b97 CM |
1916 | sleep: |
1917 | wake_up_process(root->fs_info->cleaner_kthread); | |
1918 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1919 | ||
4e121c06 JB |
1920 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1921 | &root->fs_info->fs_state))) | |
1922 | btrfs_cleanup_transaction(root); | |
a0acae0e | 1923 | if (!try_to_freeze()) { |
a74a4b97 | 1924 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1925 | if (!kthread_should_stop() && |
914b2007 JK |
1926 | (!btrfs_transaction_blocked(root->fs_info) || |
1927 | cannot_commit)) | |
8929ecfa | 1928 | schedule_timeout(delay); |
a74a4b97 CM |
1929 | __set_current_state(TASK_RUNNING); |
1930 | } | |
1931 | } while (!kthread_should_stop()); | |
1932 | return 0; | |
1933 | } | |
1934 | ||
af31f5e5 CM |
1935 | /* |
1936 | * this will find the highest generation in the array of | |
1937 | * root backups. The index of the highest array is returned, | |
1938 | * or -1 if we can't find anything. | |
1939 | * | |
1940 | * We check to make sure the array is valid by comparing the | |
1941 | * generation of the latest root in the array with the generation | |
1942 | * in the super block. If they don't match we pitch it. | |
1943 | */ | |
1944 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1945 | { | |
1946 | u64 cur; | |
1947 | int newest_index = -1; | |
1948 | struct btrfs_root_backup *root_backup; | |
1949 | int i; | |
1950 | ||
1951 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1952 | root_backup = info->super_copy->super_roots + i; | |
1953 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1954 | if (cur == newest_gen) | |
1955 | newest_index = i; | |
1956 | } | |
1957 | ||
1958 | /* check to see if we actually wrapped around */ | |
1959 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1960 | root_backup = info->super_copy->super_roots; | |
1961 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1962 | if (cur == newest_gen) | |
1963 | newest_index = 0; | |
1964 | } | |
1965 | return newest_index; | |
1966 | } | |
1967 | ||
1968 | ||
1969 | /* | |
1970 | * find the oldest backup so we know where to store new entries | |
1971 | * in the backup array. This will set the backup_root_index | |
1972 | * field in the fs_info struct | |
1973 | */ | |
1974 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1975 | u64 newest_gen) | |
1976 | { | |
1977 | int newest_index = -1; | |
1978 | ||
1979 | newest_index = find_newest_super_backup(info, newest_gen); | |
1980 | /* if there was garbage in there, just move along */ | |
1981 | if (newest_index == -1) { | |
1982 | info->backup_root_index = 0; | |
1983 | } else { | |
1984 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1985 | } | |
1986 | } | |
1987 | ||
1988 | /* | |
1989 | * copy all the root pointers into the super backup array. | |
1990 | * this will bump the backup pointer by one when it is | |
1991 | * done | |
1992 | */ | |
1993 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1994 | { | |
1995 | int next_backup; | |
1996 | struct btrfs_root_backup *root_backup; | |
1997 | int last_backup; | |
1998 | ||
1999 | next_backup = info->backup_root_index; | |
2000 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2001 | BTRFS_NUM_BACKUP_ROOTS; | |
2002 | ||
2003 | /* | |
2004 | * just overwrite the last backup if we're at the same generation | |
2005 | * this happens only at umount | |
2006 | */ | |
2007 | root_backup = info->super_for_commit->super_roots + last_backup; | |
2008 | if (btrfs_backup_tree_root_gen(root_backup) == | |
2009 | btrfs_header_generation(info->tree_root->node)) | |
2010 | next_backup = last_backup; | |
2011 | ||
2012 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2013 | ||
2014 | /* | |
2015 | * make sure all of our padding and empty slots get zero filled | |
2016 | * regardless of which ones we use today | |
2017 | */ | |
2018 | memset(root_backup, 0, sizeof(*root_backup)); | |
2019 | ||
2020 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2021 | ||
2022 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2023 | btrfs_set_backup_tree_root_gen(root_backup, | |
2024 | btrfs_header_generation(info->tree_root->node)); | |
2025 | ||
2026 | btrfs_set_backup_tree_root_level(root_backup, | |
2027 | btrfs_header_level(info->tree_root->node)); | |
2028 | ||
2029 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2030 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2031 | btrfs_header_generation(info->chunk_root->node)); | |
2032 | btrfs_set_backup_chunk_root_level(root_backup, | |
2033 | btrfs_header_level(info->chunk_root->node)); | |
2034 | ||
2035 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2036 | btrfs_set_backup_extent_root_gen(root_backup, | |
2037 | btrfs_header_generation(info->extent_root->node)); | |
2038 | btrfs_set_backup_extent_root_level(root_backup, | |
2039 | btrfs_header_level(info->extent_root->node)); | |
2040 | ||
7c7e82a7 CM |
2041 | /* |
2042 | * we might commit during log recovery, which happens before we set | |
2043 | * the fs_root. Make sure it is valid before we fill it in. | |
2044 | */ | |
2045 | if (info->fs_root && info->fs_root->node) { | |
2046 | btrfs_set_backup_fs_root(root_backup, | |
2047 | info->fs_root->node->start); | |
2048 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2049 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2050 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2051 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2052 | } |
af31f5e5 CM |
2053 | |
2054 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2055 | btrfs_set_backup_dev_root_gen(root_backup, | |
2056 | btrfs_header_generation(info->dev_root->node)); | |
2057 | btrfs_set_backup_dev_root_level(root_backup, | |
2058 | btrfs_header_level(info->dev_root->node)); | |
2059 | ||
2060 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2061 | btrfs_set_backup_csum_root_gen(root_backup, | |
2062 | btrfs_header_generation(info->csum_root->node)); | |
2063 | btrfs_set_backup_csum_root_level(root_backup, | |
2064 | btrfs_header_level(info->csum_root->node)); | |
2065 | ||
2066 | btrfs_set_backup_total_bytes(root_backup, | |
2067 | btrfs_super_total_bytes(info->super_copy)); | |
2068 | btrfs_set_backup_bytes_used(root_backup, | |
2069 | btrfs_super_bytes_used(info->super_copy)); | |
2070 | btrfs_set_backup_num_devices(root_backup, | |
2071 | btrfs_super_num_devices(info->super_copy)); | |
2072 | ||
2073 | /* | |
2074 | * if we don't copy this out to the super_copy, it won't get remembered | |
2075 | * for the next commit | |
2076 | */ | |
2077 | memcpy(&info->super_copy->super_roots, | |
2078 | &info->super_for_commit->super_roots, | |
2079 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2080 | } | |
2081 | ||
2082 | /* | |
2083 | * this copies info out of the root backup array and back into | |
2084 | * the in-memory super block. It is meant to help iterate through | |
2085 | * the array, so you send it the number of backups you've already | |
2086 | * tried and the last backup index you used. | |
2087 | * | |
2088 | * this returns -1 when it has tried all the backups | |
2089 | */ | |
2090 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2091 | struct btrfs_super_block *super, | |
2092 | int *num_backups_tried, int *backup_index) | |
2093 | { | |
2094 | struct btrfs_root_backup *root_backup; | |
2095 | int newest = *backup_index; | |
2096 | ||
2097 | if (*num_backups_tried == 0) { | |
2098 | u64 gen = btrfs_super_generation(super); | |
2099 | ||
2100 | newest = find_newest_super_backup(info, gen); | |
2101 | if (newest == -1) | |
2102 | return -1; | |
2103 | ||
2104 | *backup_index = newest; | |
2105 | *num_backups_tried = 1; | |
2106 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2107 | /* we've tried all the backups, all done */ | |
2108 | return -1; | |
2109 | } else { | |
2110 | /* jump to the next oldest backup */ | |
2111 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2112 | BTRFS_NUM_BACKUP_ROOTS; | |
2113 | *backup_index = newest; | |
2114 | *num_backups_tried += 1; | |
2115 | } | |
2116 | root_backup = super->super_roots + newest; | |
2117 | ||
2118 | btrfs_set_super_generation(super, | |
2119 | btrfs_backup_tree_root_gen(root_backup)); | |
2120 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2121 | btrfs_set_super_root_level(super, | |
2122 | btrfs_backup_tree_root_level(root_backup)); | |
2123 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2124 | ||
2125 | /* | |
2126 | * fixme: the total bytes and num_devices need to match or we should | |
2127 | * need a fsck | |
2128 | */ | |
2129 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2130 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2131 | return 0; | |
2132 | } | |
2133 | ||
7abadb64 LB |
2134 | /* helper to cleanup workers */ |
2135 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2136 | { | |
dc6e3209 | 2137 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2138 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2139 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 QW |
2140 | btrfs_destroy_workqueue(fs_info->endio_workers); |
2141 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2142 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); | |
8b110e39 | 2143 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2144 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2145 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); |
2146 | btrfs_destroy_workqueue(fs_info->endio_write_workers); | |
2147 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2148 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2149 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2150 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2151 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2152 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2153 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2154 | btrfs_destroy_workqueue(fs_info->extent_workers); |
7abadb64 LB |
2155 | } |
2156 | ||
2e9f5954 R |
2157 | static void free_root_extent_buffers(struct btrfs_root *root) |
2158 | { | |
2159 | if (root) { | |
2160 | free_extent_buffer(root->node); | |
2161 | free_extent_buffer(root->commit_root); | |
2162 | root->node = NULL; | |
2163 | root->commit_root = NULL; | |
2164 | } | |
2165 | } | |
2166 | ||
af31f5e5 CM |
2167 | /* helper to cleanup tree roots */ |
2168 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2169 | { | |
2e9f5954 | 2170 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2171 | |
2e9f5954 R |
2172 | free_root_extent_buffers(info->dev_root); |
2173 | free_root_extent_buffers(info->extent_root); | |
2174 | free_root_extent_buffers(info->csum_root); | |
2175 | free_root_extent_buffers(info->quota_root); | |
2176 | free_root_extent_buffers(info->uuid_root); | |
2177 | if (chunk_root) | |
2178 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2179 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2180 | } |
2181 | ||
faa2dbf0 | 2182 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2183 | { |
2184 | int ret; | |
2185 | struct btrfs_root *gang[8]; | |
2186 | int i; | |
2187 | ||
2188 | while (!list_empty(&fs_info->dead_roots)) { | |
2189 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2190 | struct btrfs_root, root_list); | |
2191 | list_del(&gang[0]->root_list); | |
2192 | ||
27cdeb70 | 2193 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2194 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2195 | } else { |
2196 | free_extent_buffer(gang[0]->node); | |
2197 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2198 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2199 | } |
2200 | } | |
2201 | ||
2202 | while (1) { | |
2203 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2204 | (void **)gang, 0, | |
2205 | ARRAY_SIZE(gang)); | |
2206 | if (!ret) | |
2207 | break; | |
2208 | for (i = 0; i < ret; i++) | |
cb517eab | 2209 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2210 | } |
1a4319cc LB |
2211 | |
2212 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2213 | btrfs_free_log_root_tree(NULL, fs_info); | |
2214 | btrfs_destroy_pinned_extent(fs_info->tree_root, | |
2215 | fs_info->pinned_extents); | |
2216 | } | |
171f6537 | 2217 | } |
af31f5e5 | 2218 | |
638aa7ed ES |
2219 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2220 | { | |
2221 | mutex_init(&fs_info->scrub_lock); | |
2222 | atomic_set(&fs_info->scrubs_running, 0); | |
2223 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2224 | atomic_set(&fs_info->scrubs_paused, 0); | |
2225 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2226 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2227 | fs_info->scrub_workers_refcnt = 0; | |
2228 | } | |
2229 | ||
779a65a4 ES |
2230 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2231 | { | |
2232 | spin_lock_init(&fs_info->balance_lock); | |
2233 | mutex_init(&fs_info->balance_mutex); | |
2234 | atomic_set(&fs_info->balance_running, 0); | |
2235 | atomic_set(&fs_info->balance_pause_req, 0); | |
2236 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2237 | fs_info->balance_ctl = NULL; | |
2238 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2239 | } | |
2240 | ||
f37938e0 ES |
2241 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, |
2242 | struct btrfs_root *tree_root) | |
2243 | { | |
2244 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2245 | set_nlink(fs_info->btree_inode, 1); | |
2246 | /* | |
2247 | * we set the i_size on the btree inode to the max possible int. | |
2248 | * the real end of the address space is determined by all of | |
2249 | * the devices in the system | |
2250 | */ | |
2251 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
2252 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 ES |
2253 | |
2254 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); | |
2255 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, | |
2256 | fs_info->btree_inode->i_mapping); | |
2257 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; | |
2258 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); | |
2259 | ||
2260 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
2261 | ||
2262 | BTRFS_I(fs_info->btree_inode)->root = tree_root; | |
2263 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2264 | sizeof(struct btrfs_key)); | |
2265 | set_bit(BTRFS_INODE_DUMMY, | |
2266 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
2267 | btrfs_insert_inode_hash(fs_info->btree_inode); | |
2268 | } | |
2269 | ||
ad618368 ES |
2270 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2271 | { | |
2272 | fs_info->dev_replace.lock_owner = 0; | |
2273 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2274 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2275 | rwlock_init(&fs_info->dev_replace.lock); |
2276 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2277 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2278 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2279 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2280 | } |
2281 | ||
f9e92e40 ES |
2282 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2283 | { | |
2284 | spin_lock_init(&fs_info->qgroup_lock); | |
2285 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2286 | fs_info->qgroup_tree = RB_ROOT; | |
2287 | fs_info->qgroup_op_tree = RB_ROOT; | |
2288 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2289 | fs_info->qgroup_seq = 1; | |
2290 | fs_info->quota_enabled = 0; | |
2291 | fs_info->pending_quota_state = 0; | |
2292 | fs_info->qgroup_ulist = NULL; | |
2293 | mutex_init(&fs_info->qgroup_rescan_lock); | |
2294 | } | |
2295 | ||
2a458198 ES |
2296 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2297 | struct btrfs_fs_devices *fs_devices) | |
2298 | { | |
2299 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2300 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2301 | |
2302 | fs_info->workers = | |
2303 | btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, | |
2304 | max_active, 16); | |
2305 | ||
2306 | fs_info->delalloc_workers = | |
2307 | btrfs_alloc_workqueue("delalloc", flags, max_active, 2); | |
2308 | ||
2309 | fs_info->flush_workers = | |
2310 | btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); | |
2311 | ||
2312 | fs_info->caching_workers = | |
2313 | btrfs_alloc_workqueue("cache", flags, max_active, 0); | |
2314 | ||
2315 | /* | |
2316 | * a higher idle thresh on the submit workers makes it much more | |
2317 | * likely that bios will be send down in a sane order to the | |
2318 | * devices | |
2319 | */ | |
2320 | fs_info->submit_workers = | |
2321 | btrfs_alloc_workqueue("submit", flags, | |
2322 | min_t(u64, fs_devices->num_devices, | |
2323 | max_active), 64); | |
2324 | ||
2325 | fs_info->fixup_workers = | |
2326 | btrfs_alloc_workqueue("fixup", flags, 1, 0); | |
2327 | ||
2328 | /* | |
2329 | * endios are largely parallel and should have a very | |
2330 | * low idle thresh | |
2331 | */ | |
2332 | fs_info->endio_workers = | |
2333 | btrfs_alloc_workqueue("endio", flags, max_active, 4); | |
2334 | fs_info->endio_meta_workers = | |
2335 | btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); | |
2336 | fs_info->endio_meta_write_workers = | |
2337 | btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); | |
2338 | fs_info->endio_raid56_workers = | |
2339 | btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); | |
2340 | fs_info->endio_repair_workers = | |
2341 | btrfs_alloc_workqueue("endio-repair", flags, 1, 0); | |
2342 | fs_info->rmw_workers = | |
2343 | btrfs_alloc_workqueue("rmw", flags, max_active, 2); | |
2344 | fs_info->endio_write_workers = | |
2345 | btrfs_alloc_workqueue("endio-write", flags, max_active, 2); | |
2346 | fs_info->endio_freespace_worker = | |
2347 | btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); | |
2348 | fs_info->delayed_workers = | |
2349 | btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); | |
2350 | fs_info->readahead_workers = | |
2351 | btrfs_alloc_workqueue("readahead", flags, max_active, 2); | |
2352 | fs_info->qgroup_rescan_workers = | |
2353 | btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); | |
2354 | fs_info->extent_workers = | |
2355 | btrfs_alloc_workqueue("extent-refs", flags, | |
2356 | min_t(u64, fs_devices->num_devices, | |
2357 | max_active), 8); | |
2358 | ||
2359 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2360 | fs_info->submit_workers && fs_info->flush_workers && | |
2361 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2362 | fs_info->endio_meta_write_workers && | |
2363 | fs_info->endio_repair_workers && | |
2364 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2365 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2366 | fs_info->caching_workers && fs_info->readahead_workers && | |
2367 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2368 | fs_info->extent_workers && | |
2369 | fs_info->qgroup_rescan_workers)) { | |
2370 | return -ENOMEM; | |
2371 | } | |
2372 | ||
2373 | return 0; | |
2374 | } | |
2375 | ||
63443bf5 ES |
2376 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2377 | struct btrfs_fs_devices *fs_devices) | |
2378 | { | |
2379 | int ret; | |
2380 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2381 | struct btrfs_root *log_tree_root; | |
2382 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2383 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2384 | ||
2385 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2386 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2387 | return -EIO; |
2388 | } | |
2389 | ||
74e4d827 | 2390 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2391 | if (!log_tree_root) |
2392 | return -ENOMEM; | |
2393 | ||
2394 | __setup_root(tree_root->nodesize, tree_root->sectorsize, | |
2395 | tree_root->stripesize, log_tree_root, fs_info, | |
2396 | BTRFS_TREE_LOG_OBJECTID); | |
2397 | ||
2398 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
2399 | fs_info->generation + 1); | |
64c043de | 2400 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2401 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2402 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2403 | kfree(log_tree_root); |
0eeff236 | 2404 | return ret; |
64c043de | 2405 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2406 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2407 | free_extent_buffer(log_tree_root->node); |
2408 | kfree(log_tree_root); | |
2409 | return -EIO; | |
2410 | } | |
2411 | /* returns with log_tree_root freed on success */ | |
2412 | ret = btrfs_recover_log_trees(log_tree_root); | |
2413 | if (ret) { | |
a4553fef | 2414 | btrfs_std_error(tree_root->fs_info, ret, |
63443bf5 ES |
2415 | "Failed to recover log tree"); |
2416 | free_extent_buffer(log_tree_root->node); | |
2417 | kfree(log_tree_root); | |
2418 | return ret; | |
2419 | } | |
2420 | ||
2421 | if (fs_info->sb->s_flags & MS_RDONLY) { | |
2422 | ret = btrfs_commit_super(tree_root); | |
2423 | if (ret) | |
2424 | return ret; | |
2425 | } | |
2426 | ||
2427 | return 0; | |
2428 | } | |
2429 | ||
4bbcaa64 ES |
2430 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info, |
2431 | struct btrfs_root *tree_root) | |
2432 | { | |
a4f3d2c4 | 2433 | struct btrfs_root *root; |
4bbcaa64 ES |
2434 | struct btrfs_key location; |
2435 | int ret; | |
2436 | ||
2437 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; | |
2438 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2439 | location.offset = 0; | |
2440 | ||
a4f3d2c4 DS |
2441 | root = btrfs_read_tree_root(tree_root, &location); |
2442 | if (IS_ERR(root)) | |
2443 | return PTR_ERR(root); | |
2444 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2445 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2446 | |
2447 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2448 | root = btrfs_read_tree_root(tree_root, &location); |
2449 | if (IS_ERR(root)) | |
2450 | return PTR_ERR(root); | |
2451 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2452 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2453 | btrfs_init_devices_late(fs_info); |
2454 | ||
2455 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2456 | root = btrfs_read_tree_root(tree_root, &location); |
2457 | if (IS_ERR(root)) | |
2458 | return PTR_ERR(root); | |
2459 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2460 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2461 | |
2462 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2463 | root = btrfs_read_tree_root(tree_root, &location); |
2464 | if (!IS_ERR(root)) { | |
2465 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
4bbcaa64 ES |
2466 | fs_info->quota_enabled = 1; |
2467 | fs_info->pending_quota_state = 1; | |
a4f3d2c4 | 2468 | fs_info->quota_root = root; |
4bbcaa64 ES |
2469 | } |
2470 | ||
2471 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2472 | root = btrfs_read_tree_root(tree_root, &location); |
2473 | if (IS_ERR(root)) { | |
2474 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2475 | if (ret != -ENOENT) |
2476 | return ret; | |
2477 | } else { | |
a4f3d2c4 DS |
2478 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2479 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2480 | } |
2481 | ||
70f6d82e OS |
2482 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2483 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2484 | root = btrfs_read_tree_root(tree_root, &location); | |
2485 | if (IS_ERR(root)) | |
2486 | return PTR_ERR(root); | |
2487 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2488 | fs_info->free_space_root = root; | |
2489 | } | |
2490 | ||
4bbcaa64 ES |
2491 | return 0; |
2492 | } | |
2493 | ||
ad2b2c80 AV |
2494 | int open_ctree(struct super_block *sb, |
2495 | struct btrfs_fs_devices *fs_devices, | |
2496 | char *options) | |
2e635a27 | 2497 | { |
db94535d CM |
2498 | u32 sectorsize; |
2499 | u32 nodesize; | |
87ee04eb | 2500 | u32 stripesize; |
84234f3a | 2501 | u64 generation; |
f2b636e8 | 2502 | u64 features; |
3de4586c | 2503 | struct btrfs_key location; |
a061fc8d | 2504 | struct buffer_head *bh; |
4d34b278 | 2505 | struct btrfs_super_block *disk_super; |
815745cf | 2506 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2507 | struct btrfs_root *tree_root; |
4d34b278 | 2508 | struct btrfs_root *chunk_root; |
eb60ceac | 2509 | int ret; |
e58ca020 | 2510 | int err = -EINVAL; |
af31f5e5 CM |
2511 | int num_backups_tried = 0; |
2512 | int backup_index = 0; | |
5cdc7ad3 | 2513 | int max_active; |
4543df7e | 2514 | |
74e4d827 DS |
2515 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2516 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2517 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2518 | err = -ENOMEM; |
2519 | goto fail; | |
2520 | } | |
76dda93c YZ |
2521 | |
2522 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2523 | if (ret) { | |
2524 | err = ret; | |
2525 | goto fail; | |
2526 | } | |
2527 | ||
2528 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2529 | if (ret) { | |
2530 | err = ret; | |
2531 | goto fail_srcu; | |
2532 | } | |
2533 | ||
908c7f19 | 2534 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2535 | if (ret) { |
2536 | err = ret; | |
2537 | goto fail_bdi; | |
2538 | } | |
2539 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2540 | (1 + ilog2(nr_cpu_ids)); | |
2541 | ||
908c7f19 | 2542 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2543 | if (ret) { |
2544 | err = ret; | |
2545 | goto fail_dirty_metadata_bytes; | |
2546 | } | |
2547 | ||
908c7f19 | 2548 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2549 | if (ret) { |
2550 | err = ret; | |
2551 | goto fail_delalloc_bytes; | |
2552 | } | |
2553 | ||
76dda93c YZ |
2554 | fs_info->btree_inode = new_inode(sb); |
2555 | if (!fs_info->btree_inode) { | |
2556 | err = -ENOMEM; | |
c404e0dc | 2557 | goto fail_bio_counter; |
76dda93c YZ |
2558 | } |
2559 | ||
a6591715 | 2560 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2561 | |
76dda93c | 2562 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2563 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2564 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2565 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2566 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2567 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2568 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2569 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2570 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2571 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2572 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2573 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2574 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2575 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2576 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2577 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2578 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2579 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2580 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2581 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2582 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2583 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2584 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2585 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2586 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2587 | |
0b86a832 | 2588 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2589 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2590 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2591 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2592 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2593 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2594 | BTRFS_BLOCK_RSV_GLOBAL); | |
2595 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2596 | BTRFS_BLOCK_RSV_DELALLOC); | |
2597 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2598 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2599 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2600 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2601 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2602 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2603 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2604 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2605 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2606 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2607 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2608 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2609 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2610 | fs_info->sb = sb; |
95ac567a | 2611 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2612 | fs_info->metadata_ratio = 0; |
4cb5300b | 2613 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2614 | fs_info->free_chunk_space = 0; |
f29021b2 | 2615 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2616 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2617 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2618 | /* readahead state */ |
d0164adc | 2619 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2620 | spin_lock_init(&fs_info->reada_lock); |
c8b97818 | 2621 | |
b34b086c CM |
2622 | fs_info->thread_pool_size = min_t(unsigned long, |
2623 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2624 | |
199c2a9c MX |
2625 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2626 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 | 2627 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2628 | GFP_KERNEL); |
16cdcec7 MX |
2629 | if (!fs_info->delayed_root) { |
2630 | err = -ENOMEM; | |
2631 | goto fail_iput; | |
2632 | } | |
2633 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2634 | |
638aa7ed | 2635 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2636 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2637 | fs_info->check_integrity_print_mask = 0; | |
2638 | #endif | |
779a65a4 | 2639 | btrfs_init_balance(fs_info); |
21c7e756 | 2640 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2641 | |
a061fc8d CM |
2642 | sb->s_blocksize = 4096; |
2643 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2644 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2645 | |
f37938e0 | 2646 | btrfs_init_btree_inode(fs_info, tree_root); |
76dda93c | 2647 | |
0f9dd46c | 2648 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2649 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2650 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2651 | |
11833d66 | 2652 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2653 | fs_info->btree_inode->i_mapping); |
11833d66 | 2654 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2655 | fs_info->btree_inode->i_mapping); |
11833d66 | 2656 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2657 | fs_info->do_barriers = 1; |
e18e4809 | 2658 | |
39279cc3 | 2659 | |
5a3f23d5 | 2660 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2661 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2662 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2663 | mutex_init(&fs_info->transaction_kthread_mutex); |
2664 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2665 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2666 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2667 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2668 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2669 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2670 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2671 | |
ad618368 | 2672 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2673 | btrfs_init_qgroup(fs_info); |
416ac51d | 2674 | |
fa9c0d79 CM |
2675 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2676 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2677 | ||
e6dcd2dc | 2678 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2679 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2680 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2681 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2682 | |
04216820 FM |
2683 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2684 | ||
53b381b3 DW |
2685 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2686 | if (ret) { | |
83c8266a | 2687 | err = ret; |
53b381b3 DW |
2688 | goto fail_alloc; |
2689 | } | |
2690 | ||
707e8a07 | 2691 | __setup_root(4096, 4096, 4096, tree_root, |
2c90e5d6 | 2692 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2693 | |
3c4bb26b | 2694 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2695 | |
2696 | /* | |
2697 | * Read super block and check the signature bytes only | |
2698 | */ | |
a512bbf8 | 2699 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2700 | if (IS_ERR(bh)) { |
2701 | err = PTR_ERR(bh); | |
16cdcec7 | 2702 | goto fail_alloc; |
20b45077 | 2703 | } |
39279cc3 | 2704 | |
1104a885 DS |
2705 | /* |
2706 | * We want to check superblock checksum, the type is stored inside. | |
2707 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2708 | */ | |
2709 | if (btrfs_check_super_csum(bh->b_data)) { | |
efe120a0 | 2710 | printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); |
1104a885 | 2711 | err = -EINVAL; |
b2acdddf | 2712 | brelse(bh); |
1104a885 DS |
2713 | goto fail_alloc; |
2714 | } | |
2715 | ||
2716 | /* | |
2717 | * super_copy is zeroed at allocation time and we never touch the | |
2718 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2719 | * the whole block of INFO_SIZE | |
2720 | */ | |
6c41761f DS |
2721 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2722 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2723 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2724 | brelse(bh); |
5f39d397 | 2725 | |
6c41761f | 2726 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2727 | |
1104a885 DS |
2728 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2729 | if (ret) { | |
efe120a0 | 2730 | printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); |
1104a885 DS |
2731 | err = -EINVAL; |
2732 | goto fail_alloc; | |
2733 | } | |
2734 | ||
6c41761f | 2735 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2736 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2737 | goto fail_alloc; |
0f7d52f4 | 2738 | |
acce952b | 2739 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2740 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2741 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2742 | |
af31f5e5 CM |
2743 | /* |
2744 | * run through our array of backup supers and setup | |
2745 | * our ring pointer to the oldest one | |
2746 | */ | |
2747 | generation = btrfs_super_generation(disk_super); | |
2748 | find_oldest_super_backup(fs_info, generation); | |
2749 | ||
75e7cb7f LB |
2750 | /* |
2751 | * In the long term, we'll store the compression type in the super | |
2752 | * block, and it'll be used for per file compression control. | |
2753 | */ | |
2754 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2755 | ||
96da0919 | 2756 | ret = btrfs_parse_options(tree_root, options, sb->s_flags); |
2b82032c YZ |
2757 | if (ret) { |
2758 | err = ret; | |
16cdcec7 | 2759 | goto fail_alloc; |
2b82032c | 2760 | } |
dfe25020 | 2761 | |
f2b636e8 JB |
2762 | features = btrfs_super_incompat_flags(disk_super) & |
2763 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2764 | if (features) { | |
2765 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2766 | "unsupported optional features (%Lx).\n", | |
c1c9ff7c | 2767 | features); |
f2b636e8 | 2768 | err = -EINVAL; |
16cdcec7 | 2769 | goto fail_alloc; |
f2b636e8 JB |
2770 | } |
2771 | ||
5d4f98a2 | 2772 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2773 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2774 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2775 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2776 | |
3173a18f | 2777 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
5efa0490 | 2778 | printk(KERN_INFO "BTRFS: has skinny extents\n"); |
3173a18f | 2779 | |
727011e0 CM |
2780 | /* |
2781 | * flag our filesystem as having big metadata blocks if | |
2782 | * they are bigger than the page size | |
2783 | */ | |
707e8a07 | 2784 | if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) { |
727011e0 | 2785 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
efe120a0 | 2786 | printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); |
727011e0 CM |
2787 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2788 | } | |
2789 | ||
bc3f116f | 2790 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f CM |
2791 | sectorsize = btrfs_super_sectorsize(disk_super); |
2792 | stripesize = btrfs_super_stripesize(disk_super); | |
707e8a07 | 2793 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2794 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2795 | |
2796 | /* | |
2797 | * mixed block groups end up with duplicate but slightly offset | |
2798 | * extent buffers for the same range. It leads to corruptions | |
2799 | */ | |
2800 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2801 | (sectorsize != nodesize)) { |
aa8ee312 | 2802 | printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes " |
bc3f116f CM |
2803 | "are not allowed for mixed block groups on %s\n", |
2804 | sb->s_id); | |
2805 | goto fail_alloc; | |
2806 | } | |
2807 | ||
ceda0864 MX |
2808 | /* |
2809 | * Needn't use the lock because there is no other task which will | |
2810 | * update the flag. | |
2811 | */ | |
a6fa6fae | 2812 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2813 | |
f2b636e8 JB |
2814 | features = btrfs_super_compat_ro_flags(disk_super) & |
2815 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2816 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2817 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2818 | "unsupported option features (%Lx).\n", | |
c1c9ff7c | 2819 | features); |
f2b636e8 | 2820 | err = -EINVAL; |
16cdcec7 | 2821 | goto fail_alloc; |
f2b636e8 | 2822 | } |
61d92c32 | 2823 | |
5cdc7ad3 | 2824 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2825 | |
2a458198 ES |
2826 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2827 | if (ret) { | |
2828 | err = ret; | |
0dc3b84a JB |
2829 | goto fail_sb_buffer; |
2830 | } | |
4543df7e | 2831 | |
4575c9cc | 2832 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 | 2833 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
ee22184b | 2834 | SZ_4M / PAGE_CACHE_SIZE); |
4575c9cc | 2835 | |
db94535d | 2836 | tree_root->nodesize = nodesize; |
db94535d | 2837 | tree_root->sectorsize = sectorsize; |
87ee04eb | 2838 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2839 | |
2840 | sb->s_blocksize = sectorsize; | |
2841 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2842 | |
925baedd | 2843 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2844 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2845 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2846 | if (ret) { |
aa8ee312 | 2847 | printk(KERN_ERR "BTRFS: failed to read the system " |
d397712b | 2848 | "array on %s\n", sb->s_id); |
5d4f98a2 | 2849 | goto fail_sb_buffer; |
84eed90f | 2850 | } |
0b86a832 | 2851 | |
84234f3a | 2852 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2853 | |
707e8a07 DS |
2854 | __setup_root(nodesize, sectorsize, stripesize, chunk_root, |
2855 | fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
0b86a832 CM |
2856 | |
2857 | chunk_root->node = read_tree_block(chunk_root, | |
2858 | btrfs_super_chunk_root(disk_super), | |
ce86cd59 | 2859 | generation); |
64c043de LB |
2860 | if (IS_ERR(chunk_root->node) || |
2861 | !extent_buffer_uptodate(chunk_root->node)) { | |
aa8ee312 | 2862 | printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n", |
83121942 | 2863 | sb->s_id); |
e5fffbac | 2864 | if (!IS_ERR(chunk_root->node)) |
2865 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2866 | chunk_root->node = NULL; |
af31f5e5 | 2867 | goto fail_tree_roots; |
83121942 | 2868 | } |
5d4f98a2 YZ |
2869 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2870 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2871 | |
e17cade2 | 2872 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2873 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2874 | |
0b86a832 | 2875 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2876 | if (ret) { |
aa8ee312 | 2877 | printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n", |
d397712b | 2878 | sb->s_id); |
af31f5e5 | 2879 | goto fail_tree_roots; |
2b82032c | 2880 | } |
0b86a832 | 2881 | |
8dabb742 SB |
2882 | /* |
2883 | * keep the device that is marked to be the target device for the | |
2884 | * dev_replace procedure | |
2885 | */ | |
9eaed21e | 2886 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2887 | |
a6b0d5c8 | 2888 | if (!fs_devices->latest_bdev) { |
aa8ee312 | 2889 | printk(KERN_ERR "BTRFS: failed to read devices on %s\n", |
a6b0d5c8 CM |
2890 | sb->s_id); |
2891 | goto fail_tree_roots; | |
2892 | } | |
2893 | ||
af31f5e5 | 2894 | retry_root_backup: |
84234f3a | 2895 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2896 | |
e20d96d6 | 2897 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2898 | btrfs_super_root(disk_super), |
ce86cd59 | 2899 | generation); |
64c043de LB |
2900 | if (IS_ERR(tree_root->node) || |
2901 | !extent_buffer_uptodate(tree_root->node)) { | |
efe120a0 | 2902 | printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", |
83121942 | 2903 | sb->s_id); |
e5fffbac | 2904 | if (!IS_ERR(tree_root->node)) |
2905 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2906 | tree_root->node = NULL; |
af31f5e5 | 2907 | goto recovery_tree_root; |
83121942 | 2908 | } |
af31f5e5 | 2909 | |
5d4f98a2 YZ |
2910 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2911 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2912 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2913 | |
f32e48e9 CR |
2914 | mutex_lock(&tree_root->objectid_mutex); |
2915 | ret = btrfs_find_highest_objectid(tree_root, | |
2916 | &tree_root->highest_objectid); | |
2917 | if (ret) { | |
2918 | mutex_unlock(&tree_root->objectid_mutex); | |
2919 | goto recovery_tree_root; | |
2920 | } | |
2921 | ||
2922 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2923 | ||
2924 | mutex_unlock(&tree_root->objectid_mutex); | |
2925 | ||
4bbcaa64 ES |
2926 | ret = btrfs_read_roots(fs_info, tree_root); |
2927 | if (ret) | |
af31f5e5 | 2928 | goto recovery_tree_root; |
f7a81ea4 | 2929 | |
8929ecfa YZ |
2930 | fs_info->generation = generation; |
2931 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2932 | |
68310a5e ID |
2933 | ret = btrfs_recover_balance(fs_info); |
2934 | if (ret) { | |
aa8ee312 | 2935 | printk(KERN_ERR "BTRFS: failed to recover balance\n"); |
68310a5e ID |
2936 | goto fail_block_groups; |
2937 | } | |
2938 | ||
733f4fbb SB |
2939 | ret = btrfs_init_dev_stats(fs_info); |
2940 | if (ret) { | |
efe120a0 | 2941 | printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", |
733f4fbb SB |
2942 | ret); |
2943 | goto fail_block_groups; | |
2944 | } | |
2945 | ||
8dabb742 SB |
2946 | ret = btrfs_init_dev_replace(fs_info); |
2947 | if (ret) { | |
efe120a0 | 2948 | pr_err("BTRFS: failed to init dev_replace: %d\n", ret); |
8dabb742 SB |
2949 | goto fail_block_groups; |
2950 | } | |
2951 | ||
9eaed21e | 2952 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 2953 | |
b7c35e81 AJ |
2954 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
2955 | if (ret) { | |
2956 | pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret); | |
2957 | goto fail_block_groups; | |
2958 | } | |
2959 | ||
2960 | ret = btrfs_sysfs_add_device(fs_devices); | |
2961 | if (ret) { | |
2962 | pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret); | |
2963 | goto fail_fsdev_sysfs; | |
2964 | } | |
2965 | ||
96f3136e | 2966 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 2967 | if (ret) { |
efe120a0 | 2968 | pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); |
b7c35e81 | 2969 | goto fail_fsdev_sysfs; |
c59021f8 | 2970 | } |
2971 | ||
c59021f8 | 2972 | ret = btrfs_init_space_info(fs_info); |
2973 | if (ret) { | |
efe120a0 | 2974 | printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); |
2365dd3c | 2975 | goto fail_sysfs; |
c59021f8 | 2976 | } |
2977 | ||
4bbcaa64 | 2978 | ret = btrfs_read_block_groups(fs_info->extent_root); |
1b1d1f66 | 2979 | if (ret) { |
efe120a0 | 2980 | printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); |
2365dd3c | 2981 | goto fail_sysfs; |
1b1d1f66 | 2982 | } |
5af3e8cc SB |
2983 | fs_info->num_tolerated_disk_barrier_failures = |
2984 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
2985 | if (fs_info->fs_devices->missing_devices > |
2986 | fs_info->num_tolerated_disk_barrier_failures && | |
2987 | !(sb->s_flags & MS_RDONLY)) { | |
78fa1770 ZL |
2988 | pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n", |
2989 | fs_info->fs_devices->missing_devices, | |
2990 | fs_info->num_tolerated_disk_barrier_failures); | |
2365dd3c | 2991 | goto fail_sysfs; |
292fd7fc | 2992 | } |
9078a3e1 | 2993 | |
a74a4b97 CM |
2994 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2995 | "btrfs-cleaner"); | |
57506d50 | 2996 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 2997 | goto fail_sysfs; |
a74a4b97 CM |
2998 | |
2999 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3000 | tree_root, | |
3001 | "btrfs-transaction"); | |
57506d50 | 3002 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3003 | goto fail_cleaner; |
a74a4b97 | 3004 | |
c289811c CM |
3005 | if (!btrfs_test_opt(tree_root, SSD) && |
3006 | !btrfs_test_opt(tree_root, NOSSD) && | |
3007 | !fs_info->fs_devices->rotating) { | |
efe120a0 | 3008 | printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " |
c289811c CM |
3009 | "mode\n"); |
3010 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
3011 | } | |
3012 | ||
572d9ab7 DS |
3013 | /* |
3014 | * Mount does not set all options immediatelly, we can do it now and do | |
3015 | * not have to wait for transaction commit | |
3016 | */ | |
3017 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3018 | |
21adbd5c SB |
3019 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3020 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
3021 | ret = btrfsic_mount(tree_root, fs_devices, | |
3022 | btrfs_test_opt(tree_root, | |
3023 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
3024 | 1 : 0, | |
3025 | fs_info->check_integrity_print_mask); | |
3026 | if (ret) | |
efe120a0 | 3027 | printk(KERN_WARNING "BTRFS: failed to initialize" |
21adbd5c SB |
3028 | " integrity check module %s\n", sb->s_id); |
3029 | } | |
3030 | #endif | |
bcef60f2 AJ |
3031 | ret = btrfs_read_qgroup_config(fs_info); |
3032 | if (ret) | |
3033 | goto fail_trans_kthread; | |
21adbd5c | 3034 | |
96da0919 QW |
3035 | /* do not make disk changes in broken FS or nologreplay is given */ |
3036 | if (btrfs_super_log_root(disk_super) != 0 && | |
3037 | !btrfs_test_opt(tree_root, NOLOGREPLAY)) { | |
63443bf5 | 3038 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3039 | if (ret) { |
63443bf5 | 3040 | err = ret; |
28c16cbb | 3041 | goto fail_qgroup; |
79787eaa | 3042 | } |
e02119d5 | 3043 | } |
1a40e23b | 3044 | |
76dda93c | 3045 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa | 3046 | if (ret) |
28c16cbb | 3047 | goto fail_qgroup; |
76dda93c | 3048 | |
7c2ca468 | 3049 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 3050 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3051 | if (ret) |
28c16cbb | 3052 | goto fail_qgroup; |
d68fc57b | 3053 | |
5f316481 | 3054 | mutex_lock(&fs_info->cleaner_mutex); |
5d4f98a2 | 3055 | ret = btrfs_recover_relocation(tree_root); |
5f316481 | 3056 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 MX |
3057 | if (ret < 0) { |
3058 | printk(KERN_WARNING | |
efe120a0 | 3059 | "BTRFS: failed to recover relocation\n"); |
d7ce5843 | 3060 | err = -EINVAL; |
bcef60f2 | 3061 | goto fail_qgroup; |
d7ce5843 | 3062 | } |
7c2ca468 | 3063 | } |
1a40e23b | 3064 | |
3de4586c CM |
3065 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3066 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3067 | location.offset = 0; |
3de4586c | 3068 | |
3de4586c | 3069 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3070 | if (IS_ERR(fs_info->fs_root)) { |
3071 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 3072 | goto fail_qgroup; |
3140c9a3 | 3073 | } |
c289811c | 3074 | |
2b6ba629 ID |
3075 | if (sb->s_flags & MS_RDONLY) |
3076 | return 0; | |
59641015 | 3077 | |
511711af CM |
3078 | if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) && |
3079 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3080 | pr_info("BTRFS: creating free space tree\n"); | |
3081 | ret = btrfs_create_free_space_tree(fs_info); | |
3082 | if (ret) { | |
3083 | pr_warn("BTRFS: failed to create free space tree %d\n", | |
3084 | ret); | |
3085 | close_ctree(tree_root); | |
3086 | return ret; | |
3087 | } | |
3088 | } | |
3089 | ||
2b6ba629 ID |
3090 | down_read(&fs_info->cleanup_work_sem); |
3091 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3092 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3093 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
3094 | close_ctree(tree_root); |
3095 | return ret; | |
3096 | } | |
3097 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3098 | |
2b6ba629 ID |
3099 | ret = btrfs_resume_balance_async(fs_info); |
3100 | if (ret) { | |
efe120a0 | 3101 | printk(KERN_WARNING "BTRFS: failed to resume balance\n"); |
2b6ba629 ID |
3102 | close_ctree(tree_root); |
3103 | return ret; | |
e3acc2a6 JB |
3104 | } |
3105 | ||
8dabb742 SB |
3106 | ret = btrfs_resume_dev_replace_async(fs_info); |
3107 | if (ret) { | |
efe120a0 | 3108 | pr_warn("BTRFS: failed to resume dev_replace\n"); |
8dabb742 SB |
3109 | close_ctree(tree_root); |
3110 | return ret; | |
3111 | } | |
3112 | ||
b382a324 JS |
3113 | btrfs_qgroup_rescan_resume(fs_info); |
3114 | ||
70f6d82e OS |
3115 | if (btrfs_test_opt(tree_root, CLEAR_CACHE) && |
3116 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3117 | pr_info("BTRFS: clearing free space tree\n"); | |
3118 | ret = btrfs_clear_free_space_tree(fs_info); | |
3119 | if (ret) { | |
3120 | pr_warn("BTRFS: failed to clear free space tree %d\n", | |
3121 | ret); | |
3122 | close_ctree(tree_root); | |
3123 | return ret; | |
3124 | } | |
3125 | } | |
3126 | ||
4bbcaa64 | 3127 | if (!fs_info->uuid_root) { |
efe120a0 | 3128 | pr_info("BTRFS: creating UUID tree\n"); |
f7a81ea4 SB |
3129 | ret = btrfs_create_uuid_tree(fs_info); |
3130 | if (ret) { | |
efe120a0 | 3131 | pr_warn("BTRFS: failed to create the UUID tree %d\n", |
f7a81ea4 SB |
3132 | ret); |
3133 | close_ctree(tree_root); | |
3134 | return ret; | |
3135 | } | |
4bbcaa64 ES |
3136 | } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) || |
3137 | fs_info->generation != | |
3138 | btrfs_super_uuid_tree_generation(disk_super)) { | |
efe120a0 | 3139 | pr_info("BTRFS: checking UUID tree\n"); |
70f80175 SB |
3140 | ret = btrfs_check_uuid_tree(fs_info); |
3141 | if (ret) { | |
efe120a0 | 3142 | pr_warn("BTRFS: failed to check the UUID tree %d\n", |
70f80175 SB |
3143 | ret); |
3144 | close_ctree(tree_root); | |
3145 | return ret; | |
3146 | } | |
3147 | } else { | |
3148 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
3149 | } |
3150 | ||
47ab2a6c JB |
3151 | fs_info->open = 1; |
3152 | ||
8dcddfa0 QW |
3153 | /* |
3154 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3155 | * no need to keep the flag | |
3156 | */ | |
3157 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3158 | ||
ad2b2c80 | 3159 | return 0; |
39279cc3 | 3160 | |
bcef60f2 AJ |
3161 | fail_qgroup: |
3162 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3163 | fail_trans_kthread: |
3164 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 3165 | btrfs_cleanup_transaction(fs_info->tree_root); |
faa2dbf0 | 3166 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3167 | fail_cleaner: |
a74a4b97 | 3168 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3169 | |
3170 | /* | |
3171 | * make sure we're done with the btree inode before we stop our | |
3172 | * kthreads | |
3173 | */ | |
3174 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3175 | |
2365dd3c | 3176 | fail_sysfs: |
6618a59b | 3177 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3178 | |
b7c35e81 AJ |
3179 | fail_fsdev_sysfs: |
3180 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3181 | ||
1b1d1f66 | 3182 | fail_block_groups: |
54067ae9 | 3183 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 3184 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
3185 | |
3186 | fail_tree_roots: | |
3187 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3188 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3189 | |
39279cc3 | 3190 | fail_sb_buffer: |
7abadb64 | 3191 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 3192 | fail_alloc: |
4543df7e | 3193 | fail_iput: |
586e46e2 ID |
3194 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3195 | ||
4543df7e | 3196 | iput(fs_info->btree_inode); |
c404e0dc MX |
3197 | fail_bio_counter: |
3198 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3199 | fail_delalloc_bytes: |
3200 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3201 | fail_dirty_metadata_bytes: |
3202 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 3203 | fail_bdi: |
7e662854 | 3204 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
3205 | fail_srcu: |
3206 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3207 | fail: |
53b381b3 | 3208 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3209 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3210 | return err; |
af31f5e5 CM |
3211 | |
3212 | recovery_tree_root: | |
8dcddfa0 | 3213 | if (!btrfs_test_opt(tree_root, USEBACKUPROOT)) |
af31f5e5 CM |
3214 | goto fail_tree_roots; |
3215 | ||
3216 | free_root_pointers(fs_info, 0); | |
3217 | ||
3218 | /* don't use the log in recovery mode, it won't be valid */ | |
3219 | btrfs_set_super_log_root(disk_super, 0); | |
3220 | ||
3221 | /* we can't trust the free space cache either */ | |
3222 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3223 | ||
3224 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3225 | &num_backups_tried, &backup_index); | |
3226 | if (ret == -1) | |
3227 | goto fail_block_groups; | |
3228 | goto retry_root_backup; | |
eb60ceac CM |
3229 | } |
3230 | ||
f2984462 CM |
3231 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3232 | { | |
f2984462 CM |
3233 | if (uptodate) { |
3234 | set_buffer_uptodate(bh); | |
3235 | } else { | |
442a4f63 SB |
3236 | struct btrfs_device *device = (struct btrfs_device *) |
3237 | bh->b_private; | |
3238 | ||
b14af3b4 DS |
3239 | btrfs_warn_rl_in_rcu(device->dev_root->fs_info, |
3240 | "lost page write due to IO error on %s", | |
606686ee | 3241 | rcu_str_deref(device->name)); |
1259ab75 CM |
3242 | /* note, we dont' set_buffer_write_io_error because we have |
3243 | * our own ways of dealing with the IO errors | |
3244 | */ | |
f2984462 | 3245 | clear_buffer_uptodate(bh); |
442a4f63 | 3246 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3247 | } |
3248 | unlock_buffer(bh); | |
3249 | put_bh(bh); | |
3250 | } | |
3251 | ||
29c36d72 AJ |
3252 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3253 | struct buffer_head **bh_ret) | |
3254 | { | |
3255 | struct buffer_head *bh; | |
3256 | struct btrfs_super_block *super; | |
3257 | u64 bytenr; | |
3258 | ||
3259 | bytenr = btrfs_sb_offset(copy_num); | |
3260 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3261 | return -EINVAL; | |
3262 | ||
3263 | bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); | |
3264 | /* | |
3265 | * If we fail to read from the underlying devices, as of now | |
3266 | * the best option we have is to mark it EIO. | |
3267 | */ | |
3268 | if (!bh) | |
3269 | return -EIO; | |
3270 | ||
3271 | super = (struct btrfs_super_block *)bh->b_data; | |
3272 | if (btrfs_super_bytenr(super) != bytenr || | |
3273 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3274 | brelse(bh); | |
3275 | return -EINVAL; | |
3276 | } | |
3277 | ||
3278 | *bh_ret = bh; | |
3279 | return 0; | |
3280 | } | |
3281 | ||
3282 | ||
a512bbf8 YZ |
3283 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3284 | { | |
3285 | struct buffer_head *bh; | |
3286 | struct buffer_head *latest = NULL; | |
3287 | struct btrfs_super_block *super; | |
3288 | int i; | |
3289 | u64 transid = 0; | |
92fc03fb | 3290 | int ret = -EINVAL; |
a512bbf8 YZ |
3291 | |
3292 | /* we would like to check all the supers, but that would make | |
3293 | * a btrfs mount succeed after a mkfs from a different FS. | |
3294 | * So, we need to add a special mount option to scan for | |
3295 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3296 | */ | |
3297 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3298 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3299 | if (ret) | |
a512bbf8 YZ |
3300 | continue; |
3301 | ||
3302 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3303 | |
3304 | if (!latest || btrfs_super_generation(super) > transid) { | |
3305 | brelse(latest); | |
3306 | latest = bh; | |
3307 | transid = btrfs_super_generation(super); | |
3308 | } else { | |
3309 | brelse(bh); | |
3310 | } | |
3311 | } | |
92fc03fb AJ |
3312 | |
3313 | if (!latest) | |
3314 | return ERR_PTR(ret); | |
3315 | ||
a512bbf8 YZ |
3316 | return latest; |
3317 | } | |
3318 | ||
4eedeb75 HH |
3319 | /* |
3320 | * this should be called twice, once with wait == 0 and | |
3321 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3322 | * we write are pinned. | |
3323 | * | |
3324 | * They are released when wait == 1 is done. | |
3325 | * max_mirrors must be the same for both runs, and it indicates how | |
3326 | * many supers on this one device should be written. | |
3327 | * | |
3328 | * max_mirrors == 0 means to write them all. | |
3329 | */ | |
a512bbf8 YZ |
3330 | static int write_dev_supers(struct btrfs_device *device, |
3331 | struct btrfs_super_block *sb, | |
3332 | int do_barriers, int wait, int max_mirrors) | |
3333 | { | |
3334 | struct buffer_head *bh; | |
3335 | int i; | |
3336 | int ret; | |
3337 | int errors = 0; | |
3338 | u32 crc; | |
3339 | u64 bytenr; | |
a512bbf8 YZ |
3340 | |
3341 | if (max_mirrors == 0) | |
3342 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3343 | ||
a512bbf8 YZ |
3344 | for (i = 0; i < max_mirrors; i++) { |
3345 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3346 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3347 | device->commit_total_bytes) | |
a512bbf8 YZ |
3348 | break; |
3349 | ||
3350 | if (wait) { | |
3351 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3352 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3353 | if (!bh) { |
3354 | errors++; | |
3355 | continue; | |
3356 | } | |
a512bbf8 | 3357 | wait_on_buffer(bh); |
4eedeb75 HH |
3358 | if (!buffer_uptodate(bh)) |
3359 | errors++; | |
3360 | ||
3361 | /* drop our reference */ | |
3362 | brelse(bh); | |
3363 | ||
3364 | /* drop the reference from the wait == 0 run */ | |
3365 | brelse(bh); | |
3366 | continue; | |
a512bbf8 YZ |
3367 | } else { |
3368 | btrfs_set_super_bytenr(sb, bytenr); | |
3369 | ||
3370 | crc = ~(u32)0; | |
b0496686 | 3371 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3372 | BTRFS_CSUM_SIZE, crc, |
3373 | BTRFS_SUPER_INFO_SIZE - | |
3374 | BTRFS_CSUM_SIZE); | |
3375 | btrfs_csum_final(crc, sb->csum); | |
3376 | ||
4eedeb75 HH |
3377 | /* |
3378 | * one reference for us, and we leave it for the | |
3379 | * caller | |
3380 | */ | |
a512bbf8 YZ |
3381 | bh = __getblk(device->bdev, bytenr / 4096, |
3382 | BTRFS_SUPER_INFO_SIZE); | |
634554dc | 3383 | if (!bh) { |
f14d104d DS |
3384 | btrfs_err(device->dev_root->fs_info, |
3385 | "couldn't get super buffer head for bytenr %llu", | |
3386 | bytenr); | |
634554dc JB |
3387 | errors++; |
3388 | continue; | |
3389 | } | |
3390 | ||
a512bbf8 YZ |
3391 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3392 | ||
4eedeb75 | 3393 | /* one reference for submit_bh */ |
a512bbf8 | 3394 | get_bh(bh); |
4eedeb75 HH |
3395 | |
3396 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3397 | lock_buffer(bh); |
3398 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3399 | bh->b_private = device; |
a512bbf8 YZ |
3400 | } |
3401 | ||
387125fc CM |
3402 | /* |
3403 | * we fua the first super. The others we allow | |
3404 | * to go down lazy. | |
3405 | */ | |
e8117c26 WS |
3406 | if (i == 0) |
3407 | ret = btrfsic_submit_bh(WRITE_FUA, bh); | |
3408 | else | |
3409 | ret = btrfsic_submit_bh(WRITE_SYNC, bh); | |
4eedeb75 | 3410 | if (ret) |
a512bbf8 | 3411 | errors++; |
a512bbf8 YZ |
3412 | } |
3413 | return errors < i ? 0 : -1; | |
3414 | } | |
3415 | ||
387125fc CM |
3416 | /* |
3417 | * endio for the write_dev_flush, this will wake anyone waiting | |
3418 | * for the barrier when it is done | |
3419 | */ | |
4246a0b6 | 3420 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3421 | { |
387125fc CM |
3422 | if (bio->bi_private) |
3423 | complete(bio->bi_private); | |
3424 | bio_put(bio); | |
3425 | } | |
3426 | ||
3427 | /* | |
3428 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3429 | * sent down. With wait == 1, it waits for the previous flush. | |
3430 | * | |
3431 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3432 | * capable | |
3433 | */ | |
3434 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3435 | { | |
3436 | struct bio *bio; | |
3437 | int ret = 0; | |
3438 | ||
3439 | if (device->nobarriers) | |
3440 | return 0; | |
3441 | ||
3442 | if (wait) { | |
3443 | bio = device->flush_bio; | |
3444 | if (!bio) | |
3445 | return 0; | |
3446 | ||
3447 | wait_for_completion(&device->flush_wait); | |
3448 | ||
4246a0b6 CH |
3449 | if (bio->bi_error) { |
3450 | ret = bio->bi_error; | |
5af3e8cc SB |
3451 | btrfs_dev_stat_inc_and_print(device, |
3452 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3453 | } |
3454 | ||
3455 | /* drop the reference from the wait == 0 run */ | |
3456 | bio_put(bio); | |
3457 | device->flush_bio = NULL; | |
3458 | ||
3459 | return ret; | |
3460 | } | |
3461 | ||
3462 | /* | |
3463 | * one reference for us, and we leave it for the | |
3464 | * caller | |
3465 | */ | |
9c017abc | 3466 | device->flush_bio = NULL; |
9be3395b | 3467 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3468 | if (!bio) |
3469 | return -ENOMEM; | |
3470 | ||
3471 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3472 | bio->bi_bdev = device->bdev; | |
3473 | init_completion(&device->flush_wait); | |
3474 | bio->bi_private = &device->flush_wait; | |
3475 | device->flush_bio = bio; | |
3476 | ||
3477 | bio_get(bio); | |
21adbd5c | 3478 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3479 | |
3480 | return 0; | |
3481 | } | |
3482 | ||
3483 | /* | |
3484 | * send an empty flush down to each device in parallel, | |
3485 | * then wait for them | |
3486 | */ | |
3487 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3488 | { | |
3489 | struct list_head *head; | |
3490 | struct btrfs_device *dev; | |
5af3e8cc SB |
3491 | int errors_send = 0; |
3492 | int errors_wait = 0; | |
387125fc CM |
3493 | int ret; |
3494 | ||
3495 | /* send down all the barriers */ | |
3496 | head = &info->fs_devices->devices; | |
3497 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3498 | if (dev->missing) |
3499 | continue; | |
387125fc | 3500 | if (!dev->bdev) { |
5af3e8cc | 3501 | errors_send++; |
387125fc CM |
3502 | continue; |
3503 | } | |
3504 | if (!dev->in_fs_metadata || !dev->writeable) | |
3505 | continue; | |
3506 | ||
3507 | ret = write_dev_flush(dev, 0); | |
3508 | if (ret) | |
5af3e8cc | 3509 | errors_send++; |
387125fc CM |
3510 | } |
3511 | ||
3512 | /* wait for all the barriers */ | |
3513 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3514 | if (dev->missing) |
3515 | continue; | |
387125fc | 3516 | if (!dev->bdev) { |
5af3e8cc | 3517 | errors_wait++; |
387125fc CM |
3518 | continue; |
3519 | } | |
3520 | if (!dev->in_fs_metadata || !dev->writeable) | |
3521 | continue; | |
3522 | ||
3523 | ret = write_dev_flush(dev, 1); | |
3524 | if (ret) | |
5af3e8cc | 3525 | errors_wait++; |
387125fc | 3526 | } |
5af3e8cc SB |
3527 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3528 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3529 | return -EIO; |
3530 | return 0; | |
3531 | } | |
3532 | ||
943c6e99 ZL |
3533 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3534 | { | |
8789f4fe ZL |
3535 | int raid_type; |
3536 | int min_tolerated = INT_MAX; | |
943c6e99 | 3537 | |
8789f4fe ZL |
3538 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3539 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3540 | min_tolerated = min(min_tolerated, | |
3541 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3542 | tolerated_failures); | |
943c6e99 | 3543 | |
8789f4fe ZL |
3544 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3545 | if (raid_type == BTRFS_RAID_SINGLE) | |
3546 | continue; | |
3547 | if (!(flags & btrfs_raid_group[raid_type])) | |
3548 | continue; | |
3549 | min_tolerated = min(min_tolerated, | |
3550 | btrfs_raid_array[raid_type]. | |
3551 | tolerated_failures); | |
3552 | } | |
943c6e99 | 3553 | |
8789f4fe ZL |
3554 | if (min_tolerated == INT_MAX) { |
3555 | pr_warn("BTRFS: unknown raid flag: %llu\n", flags); | |
3556 | min_tolerated = 0; | |
3557 | } | |
3558 | ||
3559 | return min_tolerated; | |
943c6e99 ZL |
3560 | } |
3561 | ||
5af3e8cc SB |
3562 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3563 | struct btrfs_fs_info *fs_info) | |
3564 | { | |
3565 | struct btrfs_ioctl_space_info space; | |
3566 | struct btrfs_space_info *sinfo; | |
3567 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3568 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3569 | BTRFS_BLOCK_GROUP_METADATA, | |
3570 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
5af3e8cc SB |
3571 | int i; |
3572 | int c; | |
3573 | int num_tolerated_disk_barrier_failures = | |
3574 | (int)fs_info->fs_devices->num_devices; | |
3575 | ||
2c458045 | 3576 | for (i = 0; i < ARRAY_SIZE(types); i++) { |
5af3e8cc SB |
3577 | struct btrfs_space_info *tmp; |
3578 | ||
3579 | sinfo = NULL; | |
3580 | rcu_read_lock(); | |
3581 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3582 | if (tmp->flags == types[i]) { | |
3583 | sinfo = tmp; | |
3584 | break; | |
3585 | } | |
3586 | } | |
3587 | rcu_read_unlock(); | |
3588 | ||
3589 | if (!sinfo) | |
3590 | continue; | |
3591 | ||
3592 | down_read(&sinfo->groups_sem); | |
3593 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2c458045 ZL |
3594 | u64 flags; |
3595 | ||
3596 | if (list_empty(&sinfo->block_groups[c])) | |
3597 | continue; | |
3598 | ||
3599 | btrfs_get_block_group_info(&sinfo->block_groups[c], | |
3600 | &space); | |
3601 | if (space.total_bytes == 0 || space.used_bytes == 0) | |
3602 | continue; | |
3603 | flags = space.flags; | |
943c6e99 ZL |
3604 | |
3605 | num_tolerated_disk_barrier_failures = min( | |
3606 | num_tolerated_disk_barrier_failures, | |
3607 | btrfs_get_num_tolerated_disk_barrier_failures( | |
3608 | flags)); | |
5af3e8cc SB |
3609 | } |
3610 | up_read(&sinfo->groups_sem); | |
3611 | } | |
3612 | ||
3613 | return num_tolerated_disk_barrier_failures; | |
3614 | } | |
3615 | ||
48a3b636 | 3616 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3617 | { |
e5e9a520 | 3618 | struct list_head *head; |
f2984462 | 3619 | struct btrfs_device *dev; |
a061fc8d | 3620 | struct btrfs_super_block *sb; |
f2984462 | 3621 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3622 | int ret; |
3623 | int do_barriers; | |
a236aed1 CM |
3624 | int max_errors; |
3625 | int total_errors = 0; | |
a061fc8d | 3626 | u64 flags; |
f2984462 CM |
3627 | |
3628 | do_barriers = !btrfs_test_opt(root, NOBARRIER); | |
af31f5e5 | 3629 | backup_super_roots(root->fs_info); |
f2984462 | 3630 | |
6c41761f | 3631 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3632 | dev_item = &sb->dev_item; |
e5e9a520 | 3633 | |
174ba509 | 3634 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3635 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3636 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3637 | |
5af3e8cc SB |
3638 | if (do_barriers) { |
3639 | ret = barrier_all_devices(root->fs_info); | |
3640 | if (ret) { | |
3641 | mutex_unlock( | |
3642 | &root->fs_info->fs_devices->device_list_mutex); | |
a4553fef | 3643 | btrfs_std_error(root->fs_info, ret, |
5af3e8cc SB |
3644 | "errors while submitting device barriers."); |
3645 | return ret; | |
3646 | } | |
3647 | } | |
387125fc | 3648 | |
1f78160c | 3649 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3650 | if (!dev->bdev) { |
3651 | total_errors++; | |
3652 | continue; | |
3653 | } | |
2b82032c | 3654 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3655 | continue; |
3656 | ||
2b82032c | 3657 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3658 | btrfs_set_stack_device_type(dev_item, dev->type); |
3659 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3660 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3661 | dev->commit_total_bytes); |
ce7213c7 MX |
3662 | btrfs_set_stack_device_bytes_used(dev_item, |
3663 | dev->commit_bytes_used); | |
a061fc8d CM |
3664 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3665 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3666 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3667 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3668 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3669 | |
a061fc8d CM |
3670 | flags = btrfs_super_flags(sb); |
3671 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3672 | ||
a512bbf8 | 3673 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3674 | if (ret) |
3675 | total_errors++; | |
f2984462 | 3676 | } |
a236aed1 | 3677 | if (total_errors > max_errors) { |
efe120a0 | 3678 | btrfs_err(root->fs_info, "%d errors while writing supers", |
d397712b | 3679 | total_errors); |
a724b436 | 3680 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3681 | |
9d565ba4 | 3682 | /* FUA is masked off if unsupported and can't be the reason */ |
a4553fef | 3683 | btrfs_std_error(root->fs_info, -EIO, |
9d565ba4 SB |
3684 | "%d errors while writing supers", total_errors); |
3685 | return -EIO; | |
a236aed1 | 3686 | } |
f2984462 | 3687 | |
a512bbf8 | 3688 | total_errors = 0; |
1f78160c | 3689 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3690 | if (!dev->bdev) |
3691 | continue; | |
2b82032c | 3692 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3693 | continue; |
3694 | ||
a512bbf8 YZ |
3695 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3696 | if (ret) | |
3697 | total_errors++; | |
f2984462 | 3698 | } |
174ba509 | 3699 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3700 | if (total_errors > max_errors) { |
a4553fef | 3701 | btrfs_std_error(root->fs_info, -EIO, |
79787eaa JM |
3702 | "%d errors while writing supers", total_errors); |
3703 | return -EIO; | |
a236aed1 | 3704 | } |
f2984462 CM |
3705 | return 0; |
3706 | } | |
3707 | ||
a512bbf8 YZ |
3708 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3709 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3710 | { |
f570e757 | 3711 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3712 | } |
3713 | ||
cb517eab MX |
3714 | /* Drop a fs root from the radix tree and free it. */ |
3715 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3716 | struct btrfs_root *root) | |
2619ba1f | 3717 | { |
4df27c4d | 3718 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3719 | radix_tree_delete(&fs_info->fs_roots_radix, |
3720 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3721 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3722 | |
3723 | if (btrfs_root_refs(&root->root_item) == 0) | |
3724 | synchronize_srcu(&fs_info->subvol_srcu); | |
3725 | ||
1a4319cc | 3726 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
3321719e | 3727 | btrfs_free_log(NULL, root); |
3321719e | 3728 | |
faa2dbf0 JB |
3729 | if (root->free_ino_pinned) |
3730 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3731 | if (root->free_ino_ctl) | |
3732 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3733 | free_fs_root(root); |
4df27c4d YZ |
3734 | } |
3735 | ||
3736 | static void free_fs_root(struct btrfs_root *root) | |
3737 | { | |
57cdc8db | 3738 | iput(root->ino_cache_inode); |
4df27c4d | 3739 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3740 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3741 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3742 | if (root->anon_dev) |
3743 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3744 | if (root->subv_writers) |
3745 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3746 | free_extent_buffer(root->node); |
3747 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3748 | kfree(root->free_ino_ctl); |
3749 | kfree(root->free_ino_pinned); | |
d397712b | 3750 | kfree(root->name); |
b0feb9d9 | 3751 | btrfs_put_fs_root(root); |
2619ba1f CM |
3752 | } |
3753 | ||
cb517eab MX |
3754 | void btrfs_free_fs_root(struct btrfs_root *root) |
3755 | { | |
3756 | free_fs_root(root); | |
2619ba1f CM |
3757 | } |
3758 | ||
c146afad | 3759 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3760 | { |
c146afad YZ |
3761 | u64 root_objectid = 0; |
3762 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3763 | int i = 0; |
3764 | int err = 0; | |
3765 | unsigned int ret = 0; | |
3766 | int index; | |
e089f05c | 3767 | |
c146afad | 3768 | while (1) { |
65d33fd7 | 3769 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3770 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3771 | (void **)gang, root_objectid, | |
3772 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3773 | if (!ret) { |
3774 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3775 | break; |
65d33fd7 | 3776 | } |
5d4f98a2 | 3777 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3778 | |
c146afad | 3779 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3780 | /* Avoid to grab roots in dead_roots */ |
3781 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3782 | gang[i] = NULL; | |
3783 | continue; | |
3784 | } | |
3785 | /* grab all the search result for later use */ | |
3786 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3787 | } | |
3788 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3789 | |
65d33fd7 QW |
3790 | for (i = 0; i < ret; i++) { |
3791 | if (!gang[i]) | |
3792 | continue; | |
c146afad | 3793 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3794 | err = btrfs_orphan_cleanup(gang[i]); |
3795 | if (err) | |
65d33fd7 QW |
3796 | break; |
3797 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3798 | } |
3799 | root_objectid++; | |
3800 | } | |
65d33fd7 QW |
3801 | |
3802 | /* release the uncleaned roots due to error */ | |
3803 | for (; i < ret; i++) { | |
3804 | if (gang[i]) | |
3805 | btrfs_put_fs_root(gang[i]); | |
3806 | } | |
3807 | return err; | |
c146afad | 3808 | } |
a2135011 | 3809 | |
c146afad YZ |
3810 | int btrfs_commit_super(struct btrfs_root *root) |
3811 | { | |
3812 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3813 | |
c146afad | 3814 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3815 | btrfs_run_delayed_iputs(root); |
c146afad | 3816 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3817 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3818 | |
3819 | /* wait until ongoing cleanup work done */ | |
3820 | down_write(&root->fs_info->cleanup_work_sem); | |
3821 | up_write(&root->fs_info->cleanup_work_sem); | |
3822 | ||
7a7eaa40 | 3823 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3824 | if (IS_ERR(trans)) |
3825 | return PTR_ERR(trans); | |
d52c1bcc | 3826 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3827 | } |
3828 | ||
3abdbd78 | 3829 | void close_ctree(struct btrfs_root *root) |
c146afad YZ |
3830 | { |
3831 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3832 | int ret; | |
3833 | ||
3834 | fs_info->closing = 1; | |
3835 | smp_mb(); | |
3836 | ||
7343dd61 JM |
3837 | /* wait for the qgroup rescan worker to stop */ |
3838 | btrfs_qgroup_wait_for_completion(fs_info); | |
3839 | ||
803b2f54 SB |
3840 | /* wait for the uuid_scan task to finish */ |
3841 | down(&fs_info->uuid_tree_rescan_sem); | |
3842 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3843 | up(&fs_info->uuid_tree_rescan_sem); | |
3844 | ||
837d5b6e | 3845 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3846 | btrfs_pause_balance(fs_info); |
837d5b6e | 3847 | |
8dabb742 SB |
3848 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3849 | ||
aa1b8cd4 | 3850 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3851 | |
3852 | /* wait for any defraggers to finish */ | |
3853 | wait_event(fs_info->transaction_wait, | |
3854 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3855 | ||
3856 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3857 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3858 | |
21c7e756 MX |
3859 | cancel_work_sync(&fs_info->async_reclaim_work); |
3860 | ||
c146afad | 3861 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
e44163e1 JM |
3862 | /* |
3863 | * If the cleaner thread is stopped and there are | |
3864 | * block groups queued for removal, the deletion will be | |
3865 | * skipped when we quit the cleaner thread. | |
3866 | */ | |
e44163e1 | 3867 | btrfs_delete_unused_bgs(root->fs_info); |
e44163e1 | 3868 | |
acce952b | 3869 | ret = btrfs_commit_super(root); |
3870 | if (ret) | |
04892340 | 3871 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3872 | } |
3873 | ||
87533c47 | 3874 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3875 | btrfs_error_commit_super(root); |
0f7d52f4 | 3876 | |
e3029d9f AV |
3877 | kthread_stop(fs_info->transaction_kthread); |
3878 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3879 | |
f25784b3 YZ |
3880 | fs_info->closing = 2; |
3881 | smp_mb(); | |
3882 | ||
04892340 | 3883 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3884 | |
963d678b | 3885 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3886 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3887 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3888 | } |
bcc63abb | 3889 | |
6618a59b | 3890 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3891 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3892 | |
faa2dbf0 | 3893 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3894 | |
1a4319cc LB |
3895 | btrfs_put_block_group_cache(fs_info); |
3896 | ||
2b1360da JB |
3897 | btrfs_free_block_groups(fs_info); |
3898 | ||
de348ee0 WS |
3899 | /* |
3900 | * we must make sure there is not any read request to | |
3901 | * submit after we stopping all workers. | |
3902 | */ | |
3903 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3904 | btrfs_stop_all_workers(fs_info); |
3905 | ||
47ab2a6c | 3906 | fs_info->open = 0; |
13e6c37b | 3907 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3908 | |
13e6c37b | 3909 | iput(fs_info->btree_inode); |
d6bfde87 | 3910 | |
21adbd5c SB |
3911 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3912 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3913 | btrfsic_unmount(root, fs_info->fs_devices); | |
3914 | #endif | |
3915 | ||
dfe25020 | 3916 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3917 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3918 | |
e2d84521 | 3919 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3920 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3921 | percpu_counter_destroy(&fs_info->bio_counter); |
04160088 | 3922 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3923 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3924 | |
53b381b3 DW |
3925 | btrfs_free_stripe_hash_table(fs_info); |
3926 | ||
cdfb080e | 3927 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3928 | root->orphan_block_rsv = NULL; |
04216820 FM |
3929 | |
3930 | lock_chunks(root); | |
3931 | while (!list_empty(&fs_info->pinned_chunks)) { | |
3932 | struct extent_map *em; | |
3933 | ||
3934 | em = list_first_entry(&fs_info->pinned_chunks, | |
3935 | struct extent_map, list); | |
3936 | list_del_init(&em->list); | |
3937 | free_extent_map(em); | |
3938 | } | |
3939 | unlock_chunks(root); | |
eb60ceac CM |
3940 | } |
3941 | ||
b9fab919 CM |
3942 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3943 | int atomic) | |
5f39d397 | 3944 | { |
1259ab75 | 3945 | int ret; |
727011e0 | 3946 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3947 | |
0b32f4bb | 3948 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3949 | if (!ret) |
3950 | return ret; | |
3951 | ||
3952 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3953 | parent_transid, atomic); |
3954 | if (ret == -EAGAIN) | |
3955 | return ret; | |
1259ab75 | 3956 | return !ret; |
5f39d397 CM |
3957 | } |
3958 | ||
5f39d397 CM |
3959 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3960 | { | |
06ea65a3 | 3961 | struct btrfs_root *root; |
5f39d397 | 3962 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3963 | int was_dirty; |
b4ce94de | 3964 | |
06ea65a3 JB |
3965 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3966 | /* | |
3967 | * This is a fast path so only do this check if we have sanity tests | |
3968 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3969 | * outside of the sanity tests. | |
3970 | */ | |
3971 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3972 | return; | |
3973 | #endif | |
3974 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 3975 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3976 | if (transid != root->fs_info->generation) |
3977 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3978 | "found %llu running %llu\n", |
c1c9ff7c | 3979 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 3980 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3981 | if (!was_dirty) |
3982 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3983 | buf->len, | |
3984 | root->fs_info->dirty_metadata_batch); | |
1f21ef0a FM |
3985 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3986 | if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { | |
3987 | btrfs_print_leaf(root, buf); | |
3988 | ASSERT(0); | |
3989 | } | |
3990 | #endif | |
eb60ceac CM |
3991 | } |
3992 | ||
b53d3f5d LB |
3993 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
3994 | int flush_delayed) | |
16cdcec7 MX |
3995 | { |
3996 | /* | |
3997 | * looks as though older kernels can get into trouble with | |
3998 | * this code, they end up stuck in balance_dirty_pages forever | |
3999 | */ | |
e2d84521 | 4000 | int ret; |
16cdcec7 MX |
4001 | |
4002 | if (current->flags & PF_MEMALLOC) | |
4003 | return; | |
4004 | ||
b53d3f5d LB |
4005 | if (flush_delayed) |
4006 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 4007 | |
e2d84521 MX |
4008 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
4009 | BTRFS_DIRTY_METADATA_THRESH); | |
4010 | if (ret > 0) { | |
d0e1d66b NJ |
4011 | balance_dirty_pages_ratelimited( |
4012 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 | 4013 | } |
16cdcec7 MX |
4014 | } |
4015 | ||
b53d3f5d | 4016 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 4017 | { |
b53d3f5d LB |
4018 | __btrfs_btree_balance_dirty(root, 1); |
4019 | } | |
585ad2c3 | 4020 | |
b53d3f5d LB |
4021 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
4022 | { | |
4023 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 4024 | } |
6b80053d | 4025 | |
ca7a79ad | 4026 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 4027 | { |
727011e0 | 4028 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 4029 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 4030 | } |
0da5468f | 4031 | |
fcd1f065 | 4032 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 4033 | int read_only) |
4034 | { | |
c926093e | 4035 | struct btrfs_super_block *sb = fs_info->super_copy; |
319e4d06 QW |
4036 | u64 nodesize = btrfs_super_nodesize(sb); |
4037 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
c926093e DS |
4038 | int ret = 0; |
4039 | ||
319e4d06 QW |
4040 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { |
4041 | printk(KERN_ERR "BTRFS: no valid FS found\n"); | |
4042 | ret = -EINVAL; | |
4043 | } | |
4044 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) | |
4045 | printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n", | |
4046 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
21e7626b DS |
4047 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4048 | printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", | |
4049 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4050 | ret = -EINVAL; |
4051 | } | |
21e7626b DS |
4052 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4053 | printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", | |
4054 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4055 | ret = -EINVAL; |
4056 | } | |
21e7626b DS |
4057 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4058 | printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", | |
4059 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4060 | ret = -EINVAL; |
4061 | } | |
4062 | ||
1104a885 | 4063 | /* |
319e4d06 QW |
4064 | * Check sectorsize and nodesize first, other check will need it. |
4065 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
1104a885 | 4066 | */ |
319e4d06 QW |
4067 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || |
4068 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4069 | printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize); | |
4070 | ret = -EINVAL; | |
4071 | } | |
4072 | /* Only PAGE SIZE is supported yet */ | |
4073 | if (sectorsize != PAGE_CACHE_SIZE) { | |
4074 | printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n", | |
4075 | sectorsize, PAGE_CACHE_SIZE); | |
4076 | ret = -EINVAL; | |
4077 | } | |
4078 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
4079 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4080 | printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize); | |
4081 | ret = -EINVAL; | |
4082 | } | |
4083 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
4084 | printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n", | |
4085 | le32_to_cpu(sb->__unused_leafsize), | |
4086 | nodesize); | |
4087 | ret = -EINVAL; | |
4088 | } | |
4089 | ||
4090 | /* Root alignment check */ | |
4091 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
c926093e | 4092 | printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", |
cd743fac | 4093 | btrfs_super_root(sb)); |
319e4d06 QW |
4094 | ret = -EINVAL; |
4095 | } | |
4096 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
cd743fac DS |
4097 | printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", |
4098 | btrfs_super_chunk_root(sb)); | |
75d6ad38 DS |
4099 | ret = -EINVAL; |
4100 | } | |
319e4d06 QW |
4101 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { |
4102 | printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", | |
4103 | btrfs_super_log_root(sb)); | |
75d6ad38 DS |
4104 | ret = -EINVAL; |
4105 | } | |
4106 | ||
c926093e DS |
4107 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { |
4108 | printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", | |
4109 | fs_info->fsid, sb->dev_item.fsid); | |
4110 | ret = -EINVAL; | |
4111 | } | |
4112 | ||
4113 | /* | |
4114 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
4115 | * done later | |
4116 | */ | |
21e7626b | 4117 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
c926093e | 4118 | printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", |
21e7626b | 4119 | btrfs_super_num_devices(sb)); |
75d6ad38 DS |
4120 | if (btrfs_super_num_devices(sb) == 0) { |
4121 | printk(KERN_ERR "BTRFS: number of devices is 0\n"); | |
4122 | ret = -EINVAL; | |
4123 | } | |
c926093e | 4124 | |
21e7626b | 4125 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
c926093e | 4126 | printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", |
21e7626b | 4127 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); |
c926093e DS |
4128 | ret = -EINVAL; |
4129 | } | |
4130 | ||
ce7fca5f DS |
4131 | /* |
4132 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4133 | * and one chunk | |
4134 | */ | |
4135 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
4136 | printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", | |
4137 | btrfs_super_sys_array_size(sb), | |
4138 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
4139 | ret = -EINVAL; | |
4140 | } | |
4141 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4142 | + sizeof(struct btrfs_chunk)) { | |
d2207129 | 4143 | printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", |
ce7fca5f DS |
4144 | btrfs_super_sys_array_size(sb), |
4145 | sizeof(struct btrfs_disk_key) | |
4146 | + sizeof(struct btrfs_chunk)); | |
4147 | ret = -EINVAL; | |
4148 | } | |
4149 | ||
c926093e DS |
4150 | /* |
4151 | * The generation is a global counter, we'll trust it more than the others | |
4152 | * but it's still possible that it's the one that's wrong. | |
4153 | */ | |
21e7626b | 4154 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
c926093e DS |
4155 | printk(KERN_WARNING |
4156 | "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", | |
21e7626b DS |
4157 | btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); |
4158 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
4159 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
c926093e DS |
4160 | printk(KERN_WARNING |
4161 | "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", | |
21e7626b | 4162 | btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); |
c926093e DS |
4163 | |
4164 | return ret; | |
acce952b | 4165 | } |
4166 | ||
48a3b636 | 4167 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 4168 | { |
acce952b | 4169 | mutex_lock(&root->fs_info->cleaner_mutex); |
4170 | btrfs_run_delayed_iputs(root); | |
4171 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4172 | ||
4173 | down_write(&root->fs_info->cleanup_work_sem); | |
4174 | up_write(&root->fs_info->cleanup_work_sem); | |
4175 | ||
4176 | /* cleanup FS via transaction */ | |
4177 | btrfs_cleanup_transaction(root); | |
acce952b | 4178 | } |
4179 | ||
143bede5 | 4180 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4181 | { |
acce952b | 4182 | struct btrfs_ordered_extent *ordered; |
acce952b | 4183 | |
199c2a9c | 4184 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4185 | /* |
4186 | * This will just short circuit the ordered completion stuff which will | |
4187 | * make sure the ordered extent gets properly cleaned up. | |
4188 | */ | |
199c2a9c | 4189 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4190 | root_extent_list) |
4191 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4192 | spin_unlock(&root->ordered_extent_lock); |
4193 | } | |
4194 | ||
4195 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4196 | { | |
4197 | struct btrfs_root *root; | |
4198 | struct list_head splice; | |
4199 | ||
4200 | INIT_LIST_HEAD(&splice); | |
4201 | ||
4202 | spin_lock(&fs_info->ordered_root_lock); | |
4203 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4204 | while (!list_empty(&splice)) { | |
4205 | root = list_first_entry(&splice, struct btrfs_root, | |
4206 | ordered_root); | |
1de2cfde JB |
4207 | list_move_tail(&root->ordered_root, |
4208 | &fs_info->ordered_roots); | |
199c2a9c | 4209 | |
2a85d9ca | 4210 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4211 | btrfs_destroy_ordered_extents(root); |
4212 | ||
2a85d9ca LB |
4213 | cond_resched(); |
4214 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4215 | } |
4216 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4217 | } |
4218 | ||
35a3621b SB |
4219 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4220 | struct btrfs_root *root) | |
acce952b | 4221 | { |
4222 | struct rb_node *node; | |
4223 | struct btrfs_delayed_ref_root *delayed_refs; | |
4224 | struct btrfs_delayed_ref_node *ref; | |
4225 | int ret = 0; | |
4226 | ||
4227 | delayed_refs = &trans->delayed_refs; | |
4228 | ||
4229 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4230 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4231 | spin_unlock(&delayed_refs->lock); |
efe120a0 | 4232 | btrfs_info(root->fs_info, "delayed_refs has NO entry"); |
acce952b | 4233 | return ret; |
4234 | } | |
4235 | ||
d7df2c79 JB |
4236 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4237 | struct btrfs_delayed_ref_head *head; | |
c6fc2454 | 4238 | struct btrfs_delayed_ref_node *tmp; |
e78417d1 | 4239 | bool pin_bytes = false; |
acce952b | 4240 | |
d7df2c79 JB |
4241 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4242 | href_node); | |
4243 | if (!mutex_trylock(&head->mutex)) { | |
4244 | atomic_inc(&head->node.refs); | |
4245 | spin_unlock(&delayed_refs->lock); | |
eb12db69 | 4246 | |
d7df2c79 | 4247 | mutex_lock(&head->mutex); |
e78417d1 | 4248 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
4249 | btrfs_put_delayed_ref(&head->node); |
4250 | spin_lock(&delayed_refs->lock); | |
4251 | continue; | |
4252 | } | |
4253 | spin_lock(&head->lock); | |
c6fc2454 QW |
4254 | list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, |
4255 | list) { | |
d7df2c79 | 4256 | ref->in_tree = 0; |
c6fc2454 | 4257 | list_del(&ref->list); |
d7df2c79 JB |
4258 | atomic_dec(&delayed_refs->num_entries); |
4259 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4260 | } |
d7df2c79 JB |
4261 | if (head->must_insert_reserved) |
4262 | pin_bytes = true; | |
4263 | btrfs_free_delayed_extent_op(head->extent_op); | |
4264 | delayed_refs->num_heads--; | |
4265 | if (head->processing == 0) | |
4266 | delayed_refs->num_heads_ready--; | |
4267 | atomic_dec(&delayed_refs->num_entries); | |
4268 | head->node.in_tree = 0; | |
4269 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
4270 | spin_unlock(&head->lock); | |
4271 | spin_unlock(&delayed_refs->lock); | |
4272 | mutex_unlock(&head->mutex); | |
acce952b | 4273 | |
d7df2c79 JB |
4274 | if (pin_bytes) |
4275 | btrfs_pin_extent(root, head->node.bytenr, | |
4276 | head->node.num_bytes, 1); | |
4277 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 4278 | cond_resched(); |
4279 | spin_lock(&delayed_refs->lock); | |
4280 | } | |
4281 | ||
4282 | spin_unlock(&delayed_refs->lock); | |
4283 | ||
4284 | return ret; | |
4285 | } | |
4286 | ||
143bede5 | 4287 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4288 | { |
4289 | struct btrfs_inode *btrfs_inode; | |
4290 | struct list_head splice; | |
4291 | ||
4292 | INIT_LIST_HEAD(&splice); | |
4293 | ||
eb73c1b7 MX |
4294 | spin_lock(&root->delalloc_lock); |
4295 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4296 | |
4297 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4298 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4299 | delalloc_inodes); | |
acce952b | 4300 | |
4301 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4302 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4303 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4304 | spin_unlock(&root->delalloc_lock); |
acce952b | 4305 | |
4306 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4307 | |
eb73c1b7 | 4308 | spin_lock(&root->delalloc_lock); |
acce952b | 4309 | } |
4310 | ||
eb73c1b7 MX |
4311 | spin_unlock(&root->delalloc_lock); |
4312 | } | |
4313 | ||
4314 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4315 | { | |
4316 | struct btrfs_root *root; | |
4317 | struct list_head splice; | |
4318 | ||
4319 | INIT_LIST_HEAD(&splice); | |
4320 | ||
4321 | spin_lock(&fs_info->delalloc_root_lock); | |
4322 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4323 | while (!list_empty(&splice)) { | |
4324 | root = list_first_entry(&splice, struct btrfs_root, | |
4325 | delalloc_root); | |
4326 | list_del_init(&root->delalloc_root); | |
4327 | root = btrfs_grab_fs_root(root); | |
4328 | BUG_ON(!root); | |
4329 | spin_unlock(&fs_info->delalloc_root_lock); | |
4330 | ||
4331 | btrfs_destroy_delalloc_inodes(root); | |
4332 | btrfs_put_fs_root(root); | |
4333 | ||
4334 | spin_lock(&fs_info->delalloc_root_lock); | |
4335 | } | |
4336 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4337 | } |
4338 | ||
4339 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
4340 | struct extent_io_tree *dirty_pages, | |
4341 | int mark) | |
4342 | { | |
4343 | int ret; | |
acce952b | 4344 | struct extent_buffer *eb; |
4345 | u64 start = 0; | |
4346 | u64 end; | |
acce952b | 4347 | |
4348 | while (1) { | |
4349 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4350 | mark, NULL); |
acce952b | 4351 | if (ret) |
4352 | break; | |
4353 | ||
4354 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
4355 | while (start <= end) { | |
01d58472 | 4356 | eb = btrfs_find_tree_block(root->fs_info, start); |
707e8a07 | 4357 | start += root->nodesize; |
fd8b2b61 | 4358 | if (!eb) |
acce952b | 4359 | continue; |
fd8b2b61 | 4360 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4361 | |
fd8b2b61 JB |
4362 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4363 | &eb->bflags)) | |
4364 | clear_extent_buffer_dirty(eb); | |
4365 | free_extent_buffer_stale(eb); | |
acce952b | 4366 | } |
4367 | } | |
4368 | ||
4369 | return ret; | |
4370 | } | |
4371 | ||
4372 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
4373 | struct extent_io_tree *pinned_extents) | |
4374 | { | |
4375 | struct extent_io_tree *unpin; | |
4376 | u64 start; | |
4377 | u64 end; | |
4378 | int ret; | |
ed0eaa14 | 4379 | bool loop = true; |
acce952b | 4380 | |
4381 | unpin = pinned_extents; | |
ed0eaa14 | 4382 | again: |
acce952b | 4383 | while (1) { |
4384 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4385 | EXTENT_DIRTY, NULL); |
acce952b | 4386 | if (ret) |
4387 | break; | |
4388 | ||
acce952b | 4389 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
4390 | btrfs_error_unpin_extent_range(root, start, end); | |
4391 | cond_resched(); | |
4392 | } | |
4393 | ||
ed0eaa14 LB |
4394 | if (loop) { |
4395 | if (unpin == &root->fs_info->freed_extents[0]) | |
4396 | unpin = &root->fs_info->freed_extents[1]; | |
4397 | else | |
4398 | unpin = &root->fs_info->freed_extents[0]; | |
4399 | loop = false; | |
4400 | goto again; | |
4401 | } | |
4402 | ||
acce952b | 4403 | return 0; |
4404 | } | |
4405 | ||
49b25e05 JM |
4406 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
4407 | struct btrfs_root *root) | |
4408 | { | |
4409 | btrfs_destroy_delayed_refs(cur_trans, root); | |
49b25e05 | 4410 | |
4a9d8bde | 4411 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 4412 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 4413 | |
4a9d8bde | 4414 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 4415 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 4416 | |
67cde344 MX |
4417 | btrfs_destroy_delayed_inodes(root); |
4418 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4419 | |
49b25e05 JM |
4420 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4421 | EXTENT_DIRTY); | |
6e841e32 LB |
4422 | btrfs_destroy_pinned_extent(root, |
4423 | root->fs_info->pinned_extents); | |
49b25e05 | 4424 | |
4a9d8bde MX |
4425 | cur_trans->state =TRANS_STATE_COMPLETED; |
4426 | wake_up(&cur_trans->commit_wait); | |
4427 | ||
49b25e05 JM |
4428 | /* |
4429 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4430 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4431 | */ | |
4432 | } | |
4433 | ||
48a3b636 | 4434 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4435 | { |
4436 | struct btrfs_transaction *t; | |
acce952b | 4437 | |
acce952b | 4438 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4439 | ||
a4abeea4 | 4440 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4441 | while (!list_empty(&root->fs_info->trans_list)) { |
4442 | t = list_first_entry(&root->fs_info->trans_list, | |
4443 | struct btrfs_transaction, list); | |
4444 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4445 | atomic_inc(&t->use_count); | |
4446 | spin_unlock(&root->fs_info->trans_lock); | |
4447 | btrfs_wait_for_commit(root, t->transid); | |
4448 | btrfs_put_transaction(t); | |
4449 | spin_lock(&root->fs_info->trans_lock); | |
4450 | continue; | |
4451 | } | |
4452 | if (t == root->fs_info->running_transaction) { | |
4453 | t->state = TRANS_STATE_COMMIT_DOING; | |
4454 | spin_unlock(&root->fs_info->trans_lock); | |
4455 | /* | |
4456 | * We wait for 0 num_writers since we don't hold a trans | |
4457 | * handle open currently for this transaction. | |
4458 | */ | |
4459 | wait_event(t->writer_wait, | |
4460 | atomic_read(&t->num_writers) == 0); | |
4461 | } else { | |
4462 | spin_unlock(&root->fs_info->trans_lock); | |
4463 | } | |
4464 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4465 | |
724e2315 JB |
4466 | spin_lock(&root->fs_info->trans_lock); |
4467 | if (t == root->fs_info->running_transaction) | |
4468 | root->fs_info->running_transaction = NULL; | |
acce952b | 4469 | list_del_init(&t->list); |
724e2315 | 4470 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4471 | |
724e2315 JB |
4472 | btrfs_put_transaction(t); |
4473 | trace_btrfs_transaction_commit(root); | |
4474 | spin_lock(&root->fs_info->trans_lock); | |
4475 | } | |
4476 | spin_unlock(&root->fs_info->trans_lock); | |
4477 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4478 | btrfs_destroy_delayed_inodes(root); | |
4479 | btrfs_assert_delayed_root_empty(root); | |
4480 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4481 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4482 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4483 | |
4484 | return 0; | |
4485 | } | |
4486 | ||
e8c9f186 | 4487 | static const struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4488 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4489 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4490 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4491 | /* note we're sharing with inode.c for the merge bio hook */ |
4492 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4493 | }; |