Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
5a0e3ad6 | 28 | #include <linux/slab.h> |
784b4e29 | 29 | #include <linux/migrate.h> |
7a36ddec | 30 | #include <linux/ratelimit.h> |
6463fe58 | 31 | #include <linux/uuid.h> |
803b2f54 | 32 | #include <linux/semaphore.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
eb60ceac CM |
34 | #include "ctree.h" |
35 | #include "disk-io.h" | |
0b947aff | 36 | #include "hash.h" |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
925baedd | 41 | #include "locking.h" |
e02119d5 | 42 | #include "tree-log.h" |
fa9c0d79 | 43 | #include "free-space-cache.h" |
70f6d82e | 44 | #include "free-space-tree.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
606686ee | 47 | #include "rcu-string.h" |
8dabb742 | 48 | #include "dev-replace.h" |
53b381b3 | 49 | #include "raid56.h" |
5ac1d209 | 50 | #include "sysfs.h" |
fcebe456 | 51 | #include "qgroup.h" |
ebb8765b | 52 | #include "compression.h" |
eb60ceac | 53 | |
de0022b9 JB |
54 | #ifdef CONFIG_X86 |
55 | #include <asm/cpufeature.h> | |
56 | #endif | |
57 | ||
319e4d06 QW |
58 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
59 | BTRFS_HEADER_FLAG_RELOC |\ | |
60 | BTRFS_SUPER_FLAG_ERROR |\ | |
61 | BTRFS_SUPER_FLAG_SEEDING |\ | |
62 | BTRFS_SUPER_FLAG_METADUMP) | |
63 | ||
e8c9f186 | 64 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 65 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 66 | static void free_fs_root(struct btrfs_root *root); |
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, | |
fba4b697 | 104 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
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) |
8bd98f0e | 305 | return err; |
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 | 314 | if (!result) |
8bd98f0e | 315 | return -ENOMEM; |
607d432d JB |
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); | |
8bd98f0e | 336 | return -EUCLEAN; |
19c00ddc CM |
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 | |
01327610 | 387 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
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 | |
01327610 | 421 | * is filled with zeros and is included in the checksum. |
1104a885 DS |
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; | |
0f805531 | 515 | |
19c00ddc | 516 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
517 | /* |
518 | * Please do not consolidate these warnings into a single if. | |
519 | * It is useful to know what went wrong. | |
520 | */ | |
521 | if (WARN_ON(found_start != start)) | |
522 | return -EUCLEAN; | |
523 | if (WARN_ON(!PageUptodate(page))) | |
524 | return -EUCLEAN; | |
525 | ||
526 | ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, | |
527 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); | |
528 | ||
8bd98f0e | 529 | return csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
530 | } |
531 | ||
01d58472 | 532 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
533 | struct extent_buffer *eb) |
534 | { | |
01d58472 | 535 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
2b82032c YZ |
536 | u8 fsid[BTRFS_UUID_SIZE]; |
537 | int ret = 1; | |
538 | ||
0a4e5586 | 539 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
540 | while (fs_devices) { |
541 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
542 | ret = 0; | |
543 | break; | |
544 | } | |
545 | fs_devices = fs_devices->seed; | |
546 | } | |
547 | return ret; | |
548 | } | |
549 | ||
a826d6dc | 550 | #define CORRUPT(reason, eb, root, slot) \ |
efe120a0 FH |
551 | btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ |
552 | "root=%llu, slot=%d", reason, \ | |
c1c9ff7c | 553 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
554 | |
555 | static noinline int check_leaf(struct btrfs_root *root, | |
556 | struct extent_buffer *leaf) | |
557 | { | |
558 | struct btrfs_key key; | |
559 | struct btrfs_key leaf_key; | |
560 | u32 nritems = btrfs_header_nritems(leaf); | |
561 | int slot; | |
562 | ||
563 | if (nritems == 0) | |
564 | return 0; | |
565 | ||
566 | /* Check the 0 item */ | |
567 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
568 | BTRFS_LEAF_DATA_SIZE(root)) { | |
569 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
570 | return -EIO; | |
571 | } | |
572 | ||
573 | /* | |
574 | * Check to make sure each items keys are in the correct order and their | |
575 | * offsets make sense. We only have to loop through nritems-1 because | |
576 | * we check the current slot against the next slot, which verifies the | |
577 | * next slot's offset+size makes sense and that the current's slot | |
578 | * offset is correct. | |
579 | */ | |
580 | for (slot = 0; slot < nritems - 1; slot++) { | |
581 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
582 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
583 | ||
584 | /* Make sure the keys are in the right order */ | |
585 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
586 | CORRUPT("bad key order", leaf, root, slot); | |
587 | return -EIO; | |
588 | } | |
589 | ||
590 | /* | |
591 | * Make sure the offset and ends are right, remember that the | |
592 | * item data starts at the end of the leaf and grows towards the | |
593 | * front. | |
594 | */ | |
595 | if (btrfs_item_offset_nr(leaf, slot) != | |
596 | btrfs_item_end_nr(leaf, slot + 1)) { | |
597 | CORRUPT("slot offset bad", leaf, root, slot); | |
598 | return -EIO; | |
599 | } | |
600 | ||
601 | /* | |
602 | * Check to make sure that we don't point outside of the leaf, | |
01327610 | 603 | * just in case all the items are consistent to each other, but |
a826d6dc JB |
604 | * all point outside of the leaf. |
605 | */ | |
606 | if (btrfs_item_end_nr(leaf, slot) > | |
607 | BTRFS_LEAF_DATA_SIZE(root)) { | |
608 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
609 | return -EIO; | |
610 | } | |
611 | } | |
612 | ||
613 | return 0; | |
614 | } | |
615 | ||
053ab70f LB |
616 | static int check_node(struct btrfs_root *root, struct extent_buffer *node) |
617 | { | |
618 | unsigned long nr = btrfs_header_nritems(node); | |
619 | ||
620 | if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) { | |
621 | btrfs_crit(root->fs_info, | |
622 | "corrupt node: block %llu root %llu nritems %lu", | |
623 | node->start, root->objectid, nr); | |
624 | return -EIO; | |
625 | } | |
626 | return 0; | |
627 | } | |
628 | ||
facc8a22 MX |
629 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
630 | u64 phy_offset, struct page *page, | |
631 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 632 | { |
ce9adaa5 CM |
633 | u64 found_start; |
634 | int found_level; | |
ce9adaa5 CM |
635 | struct extent_buffer *eb; |
636 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 637 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 638 | int ret = 0; |
727011e0 | 639 | int reads_done; |
ce9adaa5 | 640 | |
ce9adaa5 CM |
641 | if (!page->private) |
642 | goto out; | |
d397712b | 643 | |
4f2de97a | 644 | eb = (struct extent_buffer *)page->private; |
d397712b | 645 | |
0b32f4bb JB |
646 | /* the pending IO might have been the only thing that kept this buffer |
647 | * in memory. Make sure we have a ref for all this other checks | |
648 | */ | |
649 | extent_buffer_get(eb); | |
650 | ||
651 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
652 | if (!reads_done) |
653 | goto err; | |
f188591e | 654 | |
5cf1ab56 | 655 | eb->read_mirror = mirror; |
656f30db | 656 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
657 | ret = -EIO; |
658 | goto err; | |
659 | } | |
660 | ||
ce9adaa5 | 661 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 662 | if (found_start != eb->start) { |
02873e43 ZL |
663 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
664 | found_start, eb->start); | |
f188591e | 665 | ret = -EIO; |
ce9adaa5 CM |
666 | goto err; |
667 | } | |
02873e43 ZL |
668 | if (check_tree_block_fsid(fs_info, eb)) { |
669 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
670 | eb->start); | |
1259ab75 CM |
671 | ret = -EIO; |
672 | goto err; | |
673 | } | |
ce9adaa5 | 674 | found_level = btrfs_header_level(eb); |
1c24c3ce | 675 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
676 | btrfs_err(fs_info, "bad tree block level %d", |
677 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
678 | ret = -EIO; |
679 | goto err; | |
680 | } | |
ce9adaa5 | 681 | |
85d4e461 CM |
682 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
683 | eb, found_level); | |
4008c04a | 684 | |
02873e43 | 685 | ret = csum_tree_block(fs_info, eb, 1); |
8bd98f0e | 686 | if (ret) |
a826d6dc | 687 | goto err; |
a826d6dc JB |
688 | |
689 | /* | |
690 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
691 | * that we don't try and read the other copies of this block, just | |
692 | * return -EIO. | |
693 | */ | |
694 | if (found_level == 0 && check_leaf(root, eb)) { | |
695 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
696 | ret = -EIO; | |
697 | } | |
ce9adaa5 | 698 | |
053ab70f LB |
699 | if (found_level > 0 && check_node(root, eb)) |
700 | ret = -EIO; | |
701 | ||
0b32f4bb JB |
702 | if (!ret) |
703 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 704 | err: |
79fb65a1 JB |
705 | if (reads_done && |
706 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
02873e43 | 707 | btree_readahead_hook(fs_info, eb, eb->start, ret); |
4bb31e92 | 708 | |
53b381b3 DW |
709 | if (ret) { |
710 | /* | |
711 | * our io error hook is going to dec the io pages | |
712 | * again, we have to make sure it has something | |
713 | * to decrement | |
714 | */ | |
715 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 716 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 717 | } |
0b32f4bb | 718 | free_extent_buffer(eb); |
ce9adaa5 | 719 | out: |
f188591e | 720 | return ret; |
ce9adaa5 CM |
721 | } |
722 | ||
ea466794 | 723 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 724 | { |
4bb31e92 | 725 | struct extent_buffer *eb; |
4bb31e92 | 726 | |
4f2de97a | 727 | eb = (struct extent_buffer *)page->private; |
656f30db | 728 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 729 | eb->read_mirror = failed_mirror; |
53b381b3 | 730 | atomic_dec(&eb->io_pages); |
ea466794 | 731 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
02873e43 | 732 | btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO); |
4bb31e92 AJ |
733 | return -EIO; /* we fixed nothing */ |
734 | } | |
735 | ||
4246a0b6 | 736 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 737 | { |
97eb6b69 | 738 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 739 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
740 | struct btrfs_workqueue *wq; |
741 | btrfs_work_func_t func; | |
ce9adaa5 | 742 | |
ce9adaa5 | 743 | fs_info = end_io_wq->info; |
4246a0b6 | 744 | end_io_wq->error = bio->bi_error; |
d20f7043 | 745 | |
7b6d91da | 746 | if (bio->bi_rw & REQ_WRITE) { |
9e0af237 LB |
747 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
748 | wq = fs_info->endio_meta_write_workers; | |
749 | func = btrfs_endio_meta_write_helper; | |
750 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
751 | wq = fs_info->endio_freespace_worker; | |
752 | func = btrfs_freespace_write_helper; | |
753 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
754 | wq = fs_info->endio_raid56_workers; | |
755 | func = btrfs_endio_raid56_helper; | |
756 | } else { | |
757 | wq = fs_info->endio_write_workers; | |
758 | func = btrfs_endio_write_helper; | |
759 | } | |
d20f7043 | 760 | } else { |
8b110e39 MX |
761 | if (unlikely(end_io_wq->metadata == |
762 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
763 | wq = fs_info->endio_repair_workers; | |
764 | func = btrfs_endio_repair_helper; | |
765 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
766 | wq = fs_info->endio_raid56_workers; |
767 | func = btrfs_endio_raid56_helper; | |
768 | } else if (end_io_wq->metadata) { | |
769 | wq = fs_info->endio_meta_workers; | |
770 | func = btrfs_endio_meta_helper; | |
771 | } else { | |
772 | wq = fs_info->endio_workers; | |
773 | func = btrfs_endio_helper; | |
774 | } | |
d20f7043 | 775 | } |
9e0af237 LB |
776 | |
777 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
778 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
779 | } |
780 | ||
22c59948 | 781 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 782 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 783 | { |
97eb6b69 | 784 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 785 | |
97eb6b69 | 786 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 CM |
787 | if (!end_io_wq) |
788 | return -ENOMEM; | |
789 | ||
790 | end_io_wq->private = bio->bi_private; | |
791 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 792 | end_io_wq->info = info; |
ce9adaa5 CM |
793 | end_io_wq->error = 0; |
794 | end_io_wq->bio = bio; | |
22c59948 | 795 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
796 | |
797 | bio->bi_private = end_io_wq; | |
798 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
799 | return 0; |
800 | } | |
801 | ||
b64a2851 | 802 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 803 | { |
4854ddd0 | 804 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 805 | info->thread_pool_size, |
4854ddd0 CM |
806 | info->fs_devices->open_devices); |
807 | return 256 * limit; | |
808 | } | |
0986fe9e | 809 | |
4a69a410 CM |
810 | static void run_one_async_start(struct btrfs_work *work) |
811 | { | |
4a69a410 | 812 | struct async_submit_bio *async; |
79787eaa | 813 | int ret; |
4a69a410 CM |
814 | |
815 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
816 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
817 | async->mirror_num, async->bio_flags, | |
818 | async->bio_offset); | |
819 | if (ret) | |
820 | async->error = ret; | |
4a69a410 CM |
821 | } |
822 | ||
823 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
824 | { |
825 | struct btrfs_fs_info *fs_info; | |
826 | struct async_submit_bio *async; | |
4854ddd0 | 827 | int limit; |
8b712842 CM |
828 | |
829 | async = container_of(work, struct async_submit_bio, work); | |
830 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 831 | |
b64a2851 | 832 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
833 | limit = limit * 2 / 3; |
834 | ||
ee863954 DS |
835 | /* |
836 | * atomic_dec_return implies a barrier for waitqueue_active | |
837 | */ | |
66657b31 | 838 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 839 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
840 | wake_up(&fs_info->async_submit_wait); |
841 | ||
bb7ab3b9 | 842 | /* If an error occurred we just want to clean up the bio and move on */ |
79787eaa | 843 | if (async->error) { |
4246a0b6 CH |
844 | async->bio->bi_error = async->error; |
845 | bio_endio(async->bio); | |
79787eaa JM |
846 | return; |
847 | } | |
848 | ||
4a69a410 | 849 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
850 | async->mirror_num, async->bio_flags, |
851 | async->bio_offset); | |
4a69a410 CM |
852 | } |
853 | ||
854 | static void run_one_async_free(struct btrfs_work *work) | |
855 | { | |
856 | struct async_submit_bio *async; | |
857 | ||
858 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
859 | kfree(async); |
860 | } | |
861 | ||
44b8bd7e CM |
862 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
863 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 864 | unsigned long bio_flags, |
eaf25d93 | 865 | u64 bio_offset, |
4a69a410 CM |
866 | extent_submit_bio_hook_t *submit_bio_start, |
867 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
868 | { |
869 | struct async_submit_bio *async; | |
870 | ||
871 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
872 | if (!async) | |
873 | return -ENOMEM; | |
874 | ||
875 | async->inode = inode; | |
876 | async->rw = rw; | |
877 | async->bio = bio; | |
878 | async->mirror_num = mirror_num; | |
4a69a410 CM |
879 | async->submit_bio_start = submit_bio_start; |
880 | async->submit_bio_done = submit_bio_done; | |
881 | ||
9e0af237 | 882 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 883 | run_one_async_done, run_one_async_free); |
4a69a410 | 884 | |
c8b97818 | 885 | async->bio_flags = bio_flags; |
eaf25d93 | 886 | async->bio_offset = bio_offset; |
8c8bee1d | 887 | |
79787eaa JM |
888 | async->error = 0; |
889 | ||
cb03c743 | 890 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 891 | |
7b6d91da | 892 | if (rw & REQ_SYNC) |
5cdc7ad3 | 893 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 894 | |
5cdc7ad3 | 895 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 896 | |
d397712b | 897 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
898 | atomic_read(&fs_info->nr_async_submits)) { |
899 | wait_event(fs_info->async_submit_wait, | |
900 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
901 | } | |
902 | ||
44b8bd7e CM |
903 | return 0; |
904 | } | |
905 | ||
ce3ed71a CM |
906 | static int btree_csum_one_bio(struct bio *bio) |
907 | { | |
2c30c71b | 908 | struct bio_vec *bvec; |
ce3ed71a | 909 | struct btrfs_root *root; |
2c30c71b | 910 | int i, ret = 0; |
ce3ed71a | 911 | |
2c30c71b | 912 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 913 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 914 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
915 | if (ret) |
916 | break; | |
ce3ed71a | 917 | } |
2c30c71b | 918 | |
79787eaa | 919 | return ret; |
ce3ed71a CM |
920 | } |
921 | ||
4a69a410 CM |
922 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
923 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
924 | unsigned long bio_flags, |
925 | u64 bio_offset) | |
22c59948 | 926 | { |
8b712842 CM |
927 | /* |
928 | * when we're called for a write, we're already in the async | |
5443be45 | 929 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 930 | */ |
79787eaa | 931 | return btree_csum_one_bio(bio); |
4a69a410 | 932 | } |
22c59948 | 933 | |
4a69a410 | 934 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
935 | int mirror_num, unsigned long bio_flags, |
936 | u64 bio_offset) | |
4a69a410 | 937 | { |
61891923 SB |
938 | int ret; |
939 | ||
8b712842 | 940 | /* |
4a69a410 CM |
941 | * when we're called for a write, we're already in the async |
942 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 943 | */ |
61891923 | 944 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
4246a0b6 CH |
945 | if (ret) { |
946 | bio->bi_error = ret; | |
947 | bio_endio(bio); | |
948 | } | |
61891923 | 949 | return ret; |
0b86a832 CM |
950 | } |
951 | ||
de0022b9 JB |
952 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
953 | { | |
954 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
955 | return 0; | |
956 | #ifdef CONFIG_X86 | |
bc696ca0 | 957 | if (static_cpu_has(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
958 | return 0; |
959 | #endif | |
960 | return 1; | |
961 | } | |
962 | ||
44b8bd7e | 963 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
964 | int mirror_num, unsigned long bio_flags, |
965 | u64 bio_offset) | |
44b8bd7e | 966 | { |
de0022b9 | 967 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
968 | int ret; |
969 | ||
7b6d91da | 970 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
971 | /* |
972 | * called for a read, do the setup so that checksum validation | |
973 | * can happen in the async kernel threads | |
974 | */ | |
f3f266ab | 975 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
bfebd8b5 | 976 | bio, BTRFS_WQ_ENDIO_METADATA); |
1d4284bd | 977 | if (ret) |
61891923 SB |
978 | goto out_w_error; |
979 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
980 | mirror_num, 0); | |
de0022b9 JB |
981 | } else if (!async) { |
982 | ret = btree_csum_one_bio(bio); | |
983 | if (ret) | |
61891923 SB |
984 | goto out_w_error; |
985 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
986 | mirror_num, 0); | |
987 | } else { | |
988 | /* | |
989 | * kthread helpers are used to submit writes so that | |
990 | * checksumming can happen in parallel across all CPUs | |
991 | */ | |
992 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
993 | inode, rw, bio, mirror_num, 0, | |
994 | bio_offset, | |
995 | __btree_submit_bio_start, | |
996 | __btree_submit_bio_done); | |
44b8bd7e | 997 | } |
d313d7a3 | 998 | |
4246a0b6 CH |
999 | if (ret) |
1000 | goto out_w_error; | |
1001 | return 0; | |
1002 | ||
61891923 | 1003 | out_w_error: |
4246a0b6 CH |
1004 | bio->bi_error = ret; |
1005 | bio_endio(bio); | |
61891923 | 1006 | return ret; |
44b8bd7e CM |
1007 | } |
1008 | ||
3dd1462e | 1009 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1010 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
1011 | struct page *newpage, struct page *page, |
1012 | enum migrate_mode mode) | |
784b4e29 CM |
1013 | { |
1014 | /* | |
1015 | * we can't safely write a btree page from here, | |
1016 | * we haven't done the locking hook | |
1017 | */ | |
1018 | if (PageDirty(page)) | |
1019 | return -EAGAIN; | |
1020 | /* | |
1021 | * Buffers may be managed in a filesystem specific way. | |
1022 | * We must have no buffers or drop them. | |
1023 | */ | |
1024 | if (page_has_private(page) && | |
1025 | !try_to_release_page(page, GFP_KERNEL)) | |
1026 | return -EAGAIN; | |
a6bc32b8 | 1027 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 1028 | } |
3dd1462e | 1029 | #endif |
784b4e29 | 1030 | |
0da5468f CM |
1031 | |
1032 | static int btree_writepages(struct address_space *mapping, | |
1033 | struct writeback_control *wbc) | |
1034 | { | |
e2d84521 MX |
1035 | struct btrfs_fs_info *fs_info; |
1036 | int ret; | |
1037 | ||
d8d5f3e1 | 1038 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1039 | |
1040 | if (wbc->for_kupdate) | |
1041 | return 0; | |
1042 | ||
e2d84521 | 1043 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1044 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
1045 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1046 | BTRFS_DIRTY_METADATA_THRESH); | |
1047 | if (ret < 0) | |
793955bc | 1048 | return 0; |
793955bc | 1049 | } |
0b32f4bb | 1050 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1051 | } |
1052 | ||
b2950863 | 1053 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 1054 | { |
d1310b2e CM |
1055 | struct extent_io_tree *tree; |
1056 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 1057 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 1058 | } |
22b0ebda | 1059 | |
70dec807 | 1060 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1061 | { |
98509cfc | 1062 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1063 | return 0; |
0c4e538b | 1064 | |
f7a52a40 | 1065 | return try_release_extent_buffer(page); |
d98237b3 CM |
1066 | } |
1067 | ||
d47992f8 LC |
1068 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1069 | unsigned int length) | |
d98237b3 | 1070 | { |
d1310b2e CM |
1071 | struct extent_io_tree *tree; |
1072 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1073 | extent_invalidatepage(tree, page, offset); |
1074 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1075 | if (PagePrivate(page)) { |
efe120a0 FH |
1076 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1077 | "page private not zero on page %llu", | |
1078 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1079 | ClearPagePrivate(page); |
1080 | set_page_private(page, 0); | |
09cbfeaf | 1081 | put_page(page); |
9ad6b7bc | 1082 | } |
d98237b3 CM |
1083 | } |
1084 | ||
0b32f4bb JB |
1085 | static int btree_set_page_dirty(struct page *page) |
1086 | { | |
bb146eb2 | 1087 | #ifdef DEBUG |
0b32f4bb JB |
1088 | struct extent_buffer *eb; |
1089 | ||
1090 | BUG_ON(!PagePrivate(page)); | |
1091 | eb = (struct extent_buffer *)page->private; | |
1092 | BUG_ON(!eb); | |
1093 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1094 | BUG_ON(!atomic_read(&eb->refs)); | |
1095 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1096 | #endif |
0b32f4bb JB |
1097 | return __set_page_dirty_nobuffers(page); |
1098 | } | |
1099 | ||
7f09410b | 1100 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1101 | .readpage = btree_readpage, |
0da5468f | 1102 | .writepages = btree_writepages, |
5f39d397 CM |
1103 | .releasepage = btree_releasepage, |
1104 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1105 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1106 | .migratepage = btree_migratepage, |
5a92bc88 | 1107 | #endif |
0b32f4bb | 1108 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1109 | }; |
1110 | ||
d3e46fea | 1111 | void readahead_tree_block(struct btrfs_root *root, u64 bytenr) |
090d1875 | 1112 | { |
5f39d397 CM |
1113 | struct extent_buffer *buf = NULL; |
1114 | struct inode *btree_inode = root->fs_info->btree_inode; | |
090d1875 | 1115 | |
a83fffb7 | 1116 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 | 1117 | if (IS_ERR(buf)) |
6197d86e | 1118 | return; |
d1310b2e | 1119 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1120 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1121 | free_extent_buffer(buf); |
090d1875 CM |
1122 | } |
1123 | ||
c0dcaa4d | 1124 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, |
ab0fff03 AJ |
1125 | int mirror_num, struct extent_buffer **eb) |
1126 | { | |
1127 | struct extent_buffer *buf = NULL; | |
1128 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1129 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1130 | int ret; | |
1131 | ||
a83fffb7 | 1132 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 | 1133 | if (IS_ERR(buf)) |
ab0fff03 AJ |
1134 | return 0; |
1135 | ||
1136 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1137 | ||
1138 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1139 | btree_get_extent, mirror_num); | |
1140 | if (ret) { | |
1141 | free_extent_buffer(buf); | |
1142 | return ret; | |
1143 | } | |
1144 | ||
1145 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1146 | free_extent_buffer(buf); | |
1147 | return -EIO; | |
0b32f4bb | 1148 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1149 | *eb = buf; |
1150 | } else { | |
1151 | free_extent_buffer(buf); | |
1152 | } | |
1153 | return 0; | |
1154 | } | |
1155 | ||
01d58472 | 1156 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, |
0308af44 | 1157 | u64 bytenr) |
0999df54 | 1158 | { |
01d58472 | 1159 | return find_extent_buffer(fs_info, bytenr); |
0999df54 CM |
1160 | } |
1161 | ||
1162 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
a83fffb7 | 1163 | u64 bytenr) |
0999df54 | 1164 | { |
f5ee5c9a | 1165 | if (btrfs_is_testing(root->fs_info)) |
b9ef22de FX |
1166 | return alloc_test_extent_buffer(root->fs_info, bytenr, |
1167 | root->nodesize); | |
ce3e6984 | 1168 | return alloc_extent_buffer(root->fs_info, bytenr); |
0999df54 CM |
1169 | } |
1170 | ||
1171 | ||
e02119d5 CM |
1172 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1173 | { | |
727011e0 | 1174 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1175 | buf->start + buf->len - 1); |
e02119d5 CM |
1176 | } |
1177 | ||
1178 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1179 | { | |
727011e0 | 1180 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1181 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1182 | } |
1183 | ||
0999df54 | 1184 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ce86cd59 | 1185 | u64 parent_transid) |
0999df54 CM |
1186 | { |
1187 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1188 | int ret; |
1189 | ||
a83fffb7 | 1190 | buf = btrfs_find_create_tree_block(root, bytenr); |
c871b0f2 LB |
1191 | if (IS_ERR(buf)) |
1192 | return buf; | |
0999df54 | 1193 | |
ca7a79ad | 1194 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1195 | if (ret) { |
1196 | free_extent_buffer(buf); | |
64c043de | 1197 | return ERR_PTR(ret); |
0f0fe8f7 | 1198 | } |
5f39d397 | 1199 | return buf; |
ce9adaa5 | 1200 | |
eb60ceac CM |
1201 | } |
1202 | ||
01d58472 DD |
1203 | void clean_tree_block(struct btrfs_trans_handle *trans, |
1204 | struct btrfs_fs_info *fs_info, | |
d5c13f92 | 1205 | struct extent_buffer *buf) |
ed2ff2cb | 1206 | { |
55c69072 | 1207 | if (btrfs_header_generation(buf) == |
e2d84521 | 1208 | fs_info->running_transaction->transid) { |
b9447ef8 | 1209 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1210 | |
b9473439 | 1211 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1212 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1213 | -buf->len, | |
1214 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1215 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1216 | btrfs_set_lock_blocking(buf); | |
1217 | clear_extent_buffer_dirty(buf); | |
1218 | } | |
925baedd | 1219 | } |
5f39d397 CM |
1220 | } |
1221 | ||
8257b2dc MX |
1222 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1223 | { | |
1224 | struct btrfs_subvolume_writers *writers; | |
1225 | int ret; | |
1226 | ||
1227 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1228 | if (!writers) | |
1229 | return ERR_PTR(-ENOMEM); | |
1230 | ||
908c7f19 | 1231 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1232 | if (ret < 0) { |
1233 | kfree(writers); | |
1234 | return ERR_PTR(ret); | |
1235 | } | |
1236 | ||
1237 | init_waitqueue_head(&writers->wait); | |
1238 | return writers; | |
1239 | } | |
1240 | ||
1241 | static void | |
1242 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1243 | { | |
1244 | percpu_counter_destroy(&writers->counter); | |
1245 | kfree(writers); | |
1246 | } | |
1247 | ||
707e8a07 DS |
1248 | static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, |
1249 | struct btrfs_root *root, struct btrfs_fs_info *fs_info, | |
143bede5 | 1250 | u64 objectid) |
d97e63b6 | 1251 | { |
7c0260ee | 1252 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
cfaa7295 | 1253 | root->node = NULL; |
a28ec197 | 1254 | root->commit_root = NULL; |
db94535d CM |
1255 | root->sectorsize = sectorsize; |
1256 | root->nodesize = nodesize; | |
87ee04eb | 1257 | root->stripesize = stripesize; |
27cdeb70 | 1258 | root->state = 0; |
d68fc57b | 1259 | root->orphan_cleanup_state = 0; |
0b86a832 | 1260 | |
0f7d52f4 CM |
1261 | root->objectid = objectid; |
1262 | root->last_trans = 0; | |
13a8a7c8 | 1263 | root->highest_objectid = 0; |
eb73c1b7 | 1264 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1265 | root->nr_ordered_extents = 0; |
58176a96 | 1266 | root->name = NULL; |
6bef4d31 | 1267 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1268 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1269 | root->block_rsv = NULL; |
d68fc57b | 1270 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1271 | |
1272 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1273 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1274 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1275 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1276 | INIT_LIST_HEAD(&root->ordered_extents); |
1277 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1278 | INIT_LIST_HEAD(&root->logged_list[0]); |
1279 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1280 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1281 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1282 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1283 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1284 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1285 | spin_lock_init(&root->log_extents_lock[0]); |
1286 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1287 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1288 | mutex_init(&root->log_mutex); |
31f3d255 | 1289 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1290 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1291 | init_waitqueue_head(&root->log_writer_wait); |
1292 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1293 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1294 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1295 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1296 | atomic_set(&root->log_commit[0], 0); |
1297 | atomic_set(&root->log_commit[1], 0); | |
1298 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1299 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1300 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1301 | atomic_set(&root->refs, 1); |
8257b2dc | 1302 | atomic_set(&root->will_be_snapshoted, 0); |
55eeaf05 | 1303 | atomic_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1304 | root->log_transid = 0; |
d1433deb | 1305 | root->log_transid_committed = -1; |
257c62e1 | 1306 | root->last_log_commit = 0; |
7c0260ee | 1307 | if (!dummy) |
06ea65a3 JB |
1308 | extent_io_tree_init(&root->dirty_log_pages, |
1309 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1310 | |
3768f368 CM |
1311 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1312 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1313 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1314 | if (!dummy) |
06ea65a3 JB |
1315 | root->defrag_trans_start = fs_info->generation; |
1316 | else | |
1317 | root->defrag_trans_start = 0; | |
4d775673 | 1318 | root->root_key.objectid = objectid; |
0ee5dc67 | 1319 | root->anon_dev = 0; |
8ea05e3a | 1320 | |
5f3ab90a | 1321 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1322 | } |
1323 | ||
74e4d827 DS |
1324 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1325 | gfp_t flags) | |
6f07e42e | 1326 | { |
74e4d827 | 1327 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1328 | if (root) |
1329 | root->fs_info = fs_info; | |
1330 | return root; | |
1331 | } | |
1332 | ||
06ea65a3 JB |
1333 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1334 | /* Should only be used by the testing infrastructure */ | |
7c0260ee JM |
1335 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info, |
1336 | u32 sectorsize, u32 nodesize) | |
06ea65a3 JB |
1337 | { |
1338 | struct btrfs_root *root; | |
1339 | ||
7c0260ee JM |
1340 | if (!fs_info) |
1341 | return ERR_PTR(-EINVAL); | |
1342 | ||
1343 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
06ea65a3 JB |
1344 | if (!root) |
1345 | return ERR_PTR(-ENOMEM); | |
b9ef22de | 1346 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
7c0260ee | 1347 | __setup_root(nodesize, sectorsize, sectorsize, root, fs_info, |
ef9f2db3 | 1348 | BTRFS_ROOT_TREE_OBJECTID); |
faa2dbf0 | 1349 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1350 | |
1351 | return root; | |
1352 | } | |
1353 | #endif | |
1354 | ||
20897f5c AJ |
1355 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1356 | struct btrfs_fs_info *fs_info, | |
1357 | u64 objectid) | |
1358 | { | |
1359 | struct extent_buffer *leaf; | |
1360 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1361 | struct btrfs_root *root; | |
1362 | struct btrfs_key key; | |
1363 | int ret = 0; | |
6463fe58 | 1364 | uuid_le uuid; |
20897f5c | 1365 | |
74e4d827 | 1366 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1367 | if (!root) |
1368 | return ERR_PTR(-ENOMEM); | |
1369 | ||
707e8a07 DS |
1370 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1371 | tree_root->stripesize, root, fs_info, objectid); | |
20897f5c AJ |
1372 | root->root_key.objectid = objectid; |
1373 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1374 | root->root_key.offset = 0; | |
1375 | ||
4d75f8a9 | 1376 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1377 | if (IS_ERR(leaf)) { |
1378 | ret = PTR_ERR(leaf); | |
1dd05682 | 1379 | leaf = NULL; |
20897f5c AJ |
1380 | goto fail; |
1381 | } | |
1382 | ||
20897f5c AJ |
1383 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1384 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1385 | btrfs_set_header_generation(leaf, trans->transid); | |
1386 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1387 | btrfs_set_header_owner(leaf, objectid); | |
1388 | root->node = leaf; | |
1389 | ||
0a4e5586 | 1390 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1391 | BTRFS_FSID_SIZE); |
1392 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1393 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1394 | BTRFS_UUID_SIZE); |
1395 | btrfs_mark_buffer_dirty(leaf); | |
1396 | ||
1397 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1398 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1399 | |
1400 | root->root_item.flags = 0; | |
1401 | root->root_item.byte_limit = 0; | |
1402 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1403 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1404 | btrfs_set_root_level(&root->root_item, 0); | |
1405 | btrfs_set_root_refs(&root->root_item, 1); | |
1406 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1407 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1408 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1409 | uuid_le_gen(&uuid); |
1410 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1411 | root->root_item.drop_level = 0; |
1412 | ||
1413 | key.objectid = objectid; | |
1414 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1415 | key.offset = 0; | |
1416 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1417 | if (ret) | |
1418 | goto fail; | |
1419 | ||
1420 | btrfs_tree_unlock(leaf); | |
1421 | ||
1dd05682 TI |
1422 | return root; |
1423 | ||
20897f5c | 1424 | fail: |
1dd05682 TI |
1425 | if (leaf) { |
1426 | btrfs_tree_unlock(leaf); | |
59885b39 | 1427 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1428 | free_extent_buffer(leaf); |
1429 | } | |
1430 | kfree(root); | |
20897f5c | 1431 | |
1dd05682 | 1432 | return ERR_PTR(ret); |
20897f5c AJ |
1433 | } |
1434 | ||
7237f183 YZ |
1435 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1436 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1437 | { |
1438 | struct btrfs_root *root; | |
1439 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1440 | struct extent_buffer *leaf; |
e02119d5 | 1441 | |
74e4d827 | 1442 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1443 | if (!root) |
7237f183 | 1444 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1445 | |
707e8a07 DS |
1446 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1447 | tree_root->stripesize, root, fs_info, | |
1448 | BTRFS_TREE_LOG_OBJECTID); | |
e02119d5 CM |
1449 | |
1450 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1451 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1452 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1453 | |
7237f183 | 1454 | /* |
27cdeb70 MX |
1455 | * DON'T set REF_COWS for log trees |
1456 | * | |
7237f183 YZ |
1457 | * log trees do not get reference counted because they go away |
1458 | * before a real commit is actually done. They do store pointers | |
1459 | * to file data extents, and those reference counts still get | |
1460 | * updated (along with back refs to the log tree). | |
1461 | */ | |
e02119d5 | 1462 | |
4d75f8a9 DS |
1463 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1464 | NULL, 0, 0, 0); | |
7237f183 YZ |
1465 | if (IS_ERR(leaf)) { |
1466 | kfree(root); | |
1467 | return ERR_CAST(leaf); | |
1468 | } | |
e02119d5 | 1469 | |
5d4f98a2 YZ |
1470 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1471 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1472 | btrfs_set_header_generation(leaf, trans->transid); | |
1473 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1474 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1475 | root->node = leaf; |
e02119d5 CM |
1476 | |
1477 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1478 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1479 | btrfs_mark_buffer_dirty(root->node); |
1480 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1481 | return root; |
1482 | } | |
1483 | ||
1484 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1485 | struct btrfs_fs_info *fs_info) | |
1486 | { | |
1487 | struct btrfs_root *log_root; | |
1488 | ||
1489 | log_root = alloc_log_tree(trans, fs_info); | |
1490 | if (IS_ERR(log_root)) | |
1491 | return PTR_ERR(log_root); | |
1492 | WARN_ON(fs_info->log_root_tree); | |
1493 | fs_info->log_root_tree = log_root; | |
1494 | return 0; | |
1495 | } | |
1496 | ||
1497 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1498 | struct btrfs_root *root) | |
1499 | { | |
1500 | struct btrfs_root *log_root; | |
1501 | struct btrfs_inode_item *inode_item; | |
1502 | ||
1503 | log_root = alloc_log_tree(trans, root->fs_info); | |
1504 | if (IS_ERR(log_root)) | |
1505 | return PTR_ERR(log_root); | |
1506 | ||
1507 | log_root->last_trans = trans->transid; | |
1508 | log_root->root_key.offset = root->root_key.objectid; | |
1509 | ||
1510 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1511 | btrfs_set_stack_inode_generation(inode_item, 1); |
1512 | btrfs_set_stack_inode_size(inode_item, 3); | |
1513 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 1514 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 1515 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1516 | |
5d4f98a2 | 1517 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1518 | |
1519 | WARN_ON(root->log_root); | |
1520 | root->log_root = log_root; | |
1521 | root->log_transid = 0; | |
d1433deb | 1522 | root->log_transid_committed = -1; |
257c62e1 | 1523 | root->last_log_commit = 0; |
e02119d5 CM |
1524 | return 0; |
1525 | } | |
1526 | ||
35a3621b SB |
1527 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1528 | struct btrfs_key *key) | |
e02119d5 CM |
1529 | { |
1530 | struct btrfs_root *root; | |
1531 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1532 | struct btrfs_path *path; |
84234f3a | 1533 | u64 generation; |
cb517eab | 1534 | int ret; |
0f7d52f4 | 1535 | |
cb517eab MX |
1536 | path = btrfs_alloc_path(); |
1537 | if (!path) | |
0f7d52f4 | 1538 | return ERR_PTR(-ENOMEM); |
cb517eab | 1539 | |
74e4d827 | 1540 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1541 | if (!root) { |
1542 | ret = -ENOMEM; | |
1543 | goto alloc_fail; | |
0f7d52f4 CM |
1544 | } |
1545 | ||
707e8a07 DS |
1546 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1547 | tree_root->stripesize, root, fs_info, key->objectid); | |
0f7d52f4 | 1548 | |
cb517eab MX |
1549 | ret = btrfs_find_root(tree_root, key, path, |
1550 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1551 | if (ret) { |
13a8a7c8 YZ |
1552 | if (ret > 0) |
1553 | ret = -ENOENT; | |
cb517eab | 1554 | goto find_fail; |
0f7d52f4 | 1555 | } |
13a8a7c8 | 1556 | |
84234f3a | 1557 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1558 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
ce86cd59 | 1559 | generation); |
64c043de LB |
1560 | if (IS_ERR(root->node)) { |
1561 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1562 | goto find_fail; |
1563 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1564 | ret = -EIO; | |
64c043de LB |
1565 | free_extent_buffer(root->node); |
1566 | goto find_fail; | |
416bc658 | 1567 | } |
5d4f98a2 | 1568 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1569 | out: |
cb517eab MX |
1570 | btrfs_free_path(path); |
1571 | return root; | |
1572 | ||
cb517eab MX |
1573 | find_fail: |
1574 | kfree(root); | |
1575 | alloc_fail: | |
1576 | root = ERR_PTR(ret); | |
1577 | goto out; | |
1578 | } | |
1579 | ||
1580 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1581 | struct btrfs_key *location) | |
1582 | { | |
1583 | struct btrfs_root *root; | |
1584 | ||
1585 | root = btrfs_read_tree_root(tree_root, location); | |
1586 | if (IS_ERR(root)) | |
1587 | return root; | |
1588 | ||
1589 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1590 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1591 | btrfs_check_and_init_root_item(&root->root_item); |
1592 | } | |
13a8a7c8 | 1593 | |
5eda7b5e CM |
1594 | return root; |
1595 | } | |
1596 | ||
cb517eab MX |
1597 | int btrfs_init_fs_root(struct btrfs_root *root) |
1598 | { | |
1599 | int ret; | |
8257b2dc | 1600 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1601 | |
1602 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1603 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1604 | GFP_NOFS); | |
1605 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1606 | ret = -ENOMEM; | |
1607 | goto fail; | |
1608 | } | |
1609 | ||
8257b2dc MX |
1610 | writers = btrfs_alloc_subvolume_writers(); |
1611 | if (IS_ERR(writers)) { | |
1612 | ret = PTR_ERR(writers); | |
1613 | goto fail; | |
1614 | } | |
1615 | root->subv_writers = writers; | |
1616 | ||
cb517eab | 1617 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1618 | spin_lock_init(&root->ino_cache_lock); |
1619 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1620 | |
1621 | ret = get_anon_bdev(&root->anon_dev); | |
1622 | if (ret) | |
876d2cf1 | 1623 | goto fail; |
f32e48e9 CR |
1624 | |
1625 | mutex_lock(&root->objectid_mutex); | |
1626 | ret = btrfs_find_highest_objectid(root, | |
1627 | &root->highest_objectid); | |
1628 | if (ret) { | |
1629 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1630 | goto fail; |
f32e48e9 CR |
1631 | } |
1632 | ||
1633 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1634 | ||
1635 | mutex_unlock(&root->objectid_mutex); | |
1636 | ||
cb517eab MX |
1637 | return 0; |
1638 | fail: | |
876d2cf1 | 1639 | /* the caller is responsible to call free_fs_root */ |
cb517eab MX |
1640 | return ret; |
1641 | } | |
1642 | ||
35bbb97f JM |
1643 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1644 | u64 root_id) | |
cb517eab MX |
1645 | { |
1646 | struct btrfs_root *root; | |
1647 | ||
1648 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1649 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1650 | (unsigned long)root_id); | |
1651 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1652 | return root; | |
1653 | } | |
1654 | ||
1655 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1656 | struct btrfs_root *root) | |
1657 | { | |
1658 | int ret; | |
1659 | ||
e1860a77 | 1660 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1661 | if (ret) |
1662 | return ret; | |
1663 | ||
1664 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1665 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1666 | (unsigned long)root->root_key.objectid, | |
1667 | root); | |
1668 | if (ret == 0) | |
27cdeb70 | 1669 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1670 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1671 | radix_tree_preload_end(); | |
1672 | ||
1673 | return ret; | |
1674 | } | |
1675 | ||
c00869f1 MX |
1676 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1677 | struct btrfs_key *location, | |
1678 | bool check_ref) | |
5eda7b5e CM |
1679 | { |
1680 | struct btrfs_root *root; | |
381cf658 | 1681 | struct btrfs_path *path; |
1d4c08e0 | 1682 | struct btrfs_key key; |
5eda7b5e CM |
1683 | int ret; |
1684 | ||
edbd8d4e CM |
1685 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1686 | return fs_info->tree_root; | |
1687 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1688 | return fs_info->extent_root; | |
8f18cf13 CM |
1689 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1690 | return fs_info->chunk_root; | |
1691 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1692 | return fs_info->dev_root; | |
0403e47e YZ |
1693 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1694 | return fs_info->csum_root; | |
bcef60f2 AJ |
1695 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1696 | return fs_info->quota_root ? fs_info->quota_root : | |
1697 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1698 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1699 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1700 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1701 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1702 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1703 | ERR_PTR(-ENOENT); | |
4df27c4d | 1704 | again: |
cb517eab | 1705 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1706 | if (root) { |
c00869f1 | 1707 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1708 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1709 | return root; |
48475471 | 1710 | } |
5eda7b5e | 1711 | |
cb517eab | 1712 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1713 | if (IS_ERR(root)) |
1714 | return root; | |
3394e160 | 1715 | |
c00869f1 | 1716 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1717 | ret = -ENOENT; |
581bb050 | 1718 | goto fail; |
35a30d7c | 1719 | } |
581bb050 | 1720 | |
cb517eab | 1721 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1722 | if (ret) |
1723 | goto fail; | |
3394e160 | 1724 | |
381cf658 DS |
1725 | path = btrfs_alloc_path(); |
1726 | if (!path) { | |
1727 | ret = -ENOMEM; | |
1728 | goto fail; | |
1729 | } | |
1d4c08e0 DS |
1730 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1731 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1732 | key.offset = location->objectid; | |
1733 | ||
1734 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1735 | btrfs_free_path(path); |
d68fc57b YZ |
1736 | if (ret < 0) |
1737 | goto fail; | |
1738 | if (ret == 0) | |
27cdeb70 | 1739 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1740 | |
cb517eab | 1741 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1742 | if (ret) { |
4df27c4d YZ |
1743 | if (ret == -EEXIST) { |
1744 | free_fs_root(root); | |
1745 | goto again; | |
1746 | } | |
1747 | goto fail; | |
0f7d52f4 | 1748 | } |
edbd8d4e | 1749 | return root; |
4df27c4d YZ |
1750 | fail: |
1751 | free_fs_root(root); | |
1752 | return ERR_PTR(ret); | |
edbd8d4e CM |
1753 | } |
1754 | ||
04160088 CM |
1755 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1756 | { | |
1757 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1758 | int ret = 0; | |
04160088 CM |
1759 | struct btrfs_device *device; |
1760 | struct backing_dev_info *bdi; | |
b7967db7 | 1761 | |
1f78160c XG |
1762 | rcu_read_lock(); |
1763 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1764 | if (!device->bdev) |
1765 | continue; | |
04160088 | 1766 | bdi = blk_get_backing_dev_info(device->bdev); |
ff9ea323 | 1767 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1768 | ret = 1; |
1769 | break; | |
1770 | } | |
1771 | } | |
1f78160c | 1772 | rcu_read_unlock(); |
04160088 CM |
1773 | return ret; |
1774 | } | |
1775 | ||
04160088 CM |
1776 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1777 | { | |
ad081f14 JA |
1778 | int err; |
1779 | ||
b4caecd4 | 1780 | err = bdi_setup_and_register(bdi, "btrfs"); |
ad081f14 JA |
1781 | if (err) |
1782 | return err; | |
1783 | ||
09cbfeaf | 1784 | bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE; |
04160088 CM |
1785 | bdi->congested_fn = btrfs_congested_fn; |
1786 | bdi->congested_data = info; | |
da2f0f74 | 1787 | bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; |
04160088 CM |
1788 | return 0; |
1789 | } | |
1790 | ||
8b712842 CM |
1791 | /* |
1792 | * called by the kthread helper functions to finally call the bio end_io | |
1793 | * functions. This is where read checksum verification actually happens | |
1794 | */ | |
1795 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1796 | { |
ce9adaa5 | 1797 | struct bio *bio; |
97eb6b69 | 1798 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1799 | |
97eb6b69 | 1800 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1801 | bio = end_io_wq->bio; |
ce9adaa5 | 1802 | |
4246a0b6 | 1803 | bio->bi_error = end_io_wq->error; |
8b712842 CM |
1804 | bio->bi_private = end_io_wq->private; |
1805 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1806 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1807 | bio_endio(bio); |
44b8bd7e CM |
1808 | } |
1809 | ||
a74a4b97 CM |
1810 | static int cleaner_kthread(void *arg) |
1811 | { | |
1812 | struct btrfs_root *root = arg; | |
d0278245 | 1813 | int again; |
da288d28 | 1814 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1815 | |
1816 | do { | |
d0278245 | 1817 | again = 0; |
a74a4b97 | 1818 | |
d0278245 | 1819 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1820 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1821 | goto sleep; |
1822 | ||
90c711ab ZB |
1823 | /* |
1824 | * Do not do anything if we might cause open_ctree() to block | |
1825 | * before we have finished mounting the filesystem. | |
1826 | */ | |
1827 | if (!root->fs_info->open) | |
1828 | goto sleep; | |
1829 | ||
d0278245 MX |
1830 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) |
1831 | goto sleep; | |
1832 | ||
dc7f370c MX |
1833 | /* |
1834 | * Avoid the problem that we change the status of the fs | |
1835 | * during the above check and trylock. | |
1836 | */ | |
babbf170 | 1837 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1838 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1839 | goto sleep; | |
76dda93c | 1840 | } |
a74a4b97 | 1841 | |
c2d6cb16 | 1842 | mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); |
d0278245 | 1843 | btrfs_run_delayed_iputs(root); |
c2d6cb16 FM |
1844 | mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); |
1845 | ||
d0278245 MX |
1846 | again = btrfs_clean_one_deleted_snapshot(root); |
1847 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1848 | ||
1849 | /* | |
05323cd1 MX |
1850 | * The defragger has dealt with the R/O remount and umount, |
1851 | * needn't do anything special here. | |
d0278245 MX |
1852 | */ |
1853 | btrfs_run_defrag_inodes(root->fs_info); | |
67c5e7d4 FM |
1854 | |
1855 | /* | |
1856 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1857 | * with relocation (btrfs_relocate_chunk) and relocation | |
1858 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1859 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1860 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1861 | * unused block groups. | |
1862 | */ | |
1863 | btrfs_delete_unused_bgs(root->fs_info); | |
d0278245 | 1864 | sleep: |
838fe188 | 1865 | if (!again) { |
a74a4b97 | 1866 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1867 | if (!kthread_should_stop()) |
1868 | schedule(); | |
a74a4b97 CM |
1869 | __set_current_state(TASK_RUNNING); |
1870 | } | |
1871 | } while (!kthread_should_stop()); | |
da288d28 FM |
1872 | |
1873 | /* | |
1874 | * Transaction kthread is stopped before us and wakes us up. | |
1875 | * However we might have started a new transaction and COWed some | |
1876 | * tree blocks when deleting unused block groups for example. So | |
1877 | * make sure we commit the transaction we started to have a clean | |
1878 | * shutdown when evicting the btree inode - if it has dirty pages | |
1879 | * when we do the final iput() on it, eviction will trigger a | |
1880 | * writeback for it which will fail with null pointer dereferences | |
1881 | * since work queues and other resources were already released and | |
1882 | * destroyed by the time the iput/eviction/writeback is made. | |
1883 | */ | |
1884 | trans = btrfs_attach_transaction(root); | |
1885 | if (IS_ERR(trans)) { | |
1886 | if (PTR_ERR(trans) != -ENOENT) | |
1887 | btrfs_err(root->fs_info, | |
1888 | "cleaner transaction attach returned %ld", | |
1889 | PTR_ERR(trans)); | |
1890 | } else { | |
1891 | int ret; | |
1892 | ||
1893 | ret = btrfs_commit_transaction(trans, root); | |
1894 | if (ret) | |
1895 | btrfs_err(root->fs_info, | |
1896 | "cleaner open transaction commit returned %d", | |
1897 | ret); | |
1898 | } | |
1899 | ||
a74a4b97 CM |
1900 | return 0; |
1901 | } | |
1902 | ||
1903 | static int transaction_kthread(void *arg) | |
1904 | { | |
1905 | struct btrfs_root *root = arg; | |
1906 | struct btrfs_trans_handle *trans; | |
1907 | struct btrfs_transaction *cur; | |
8929ecfa | 1908 | u64 transid; |
a74a4b97 CM |
1909 | unsigned long now; |
1910 | unsigned long delay; | |
914b2007 | 1911 | bool cannot_commit; |
a74a4b97 CM |
1912 | |
1913 | do { | |
914b2007 | 1914 | cannot_commit = false; |
8b87dc17 | 1915 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1916 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1917 | ||
a4abeea4 | 1918 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1919 | cur = root->fs_info->running_transaction; |
1920 | if (!cur) { | |
a4abeea4 | 1921 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1922 | goto sleep; |
1923 | } | |
31153d81 | 1924 | |
a74a4b97 | 1925 | now = get_seconds(); |
4a9d8bde | 1926 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1927 | (now < cur->start_time || |
1928 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1929 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1930 | delay = HZ * 5; |
1931 | goto sleep; | |
1932 | } | |
8929ecfa | 1933 | transid = cur->transid; |
a4abeea4 | 1934 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1935 | |
79787eaa | 1936 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1937 | trans = btrfs_attach_transaction(root); |
914b2007 | 1938 | if (IS_ERR(trans)) { |
354aa0fb MX |
1939 | if (PTR_ERR(trans) != -ENOENT) |
1940 | cannot_commit = true; | |
79787eaa | 1941 | goto sleep; |
914b2007 | 1942 | } |
8929ecfa | 1943 | if (transid == trans->transid) { |
79787eaa | 1944 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1945 | } else { |
1946 | btrfs_end_transaction(trans, root); | |
1947 | } | |
a74a4b97 CM |
1948 | sleep: |
1949 | wake_up_process(root->fs_info->cleaner_kthread); | |
1950 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1951 | ||
4e121c06 JB |
1952 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1953 | &root->fs_info->fs_state))) | |
1954 | btrfs_cleanup_transaction(root); | |
ce63f891 JK |
1955 | set_current_state(TASK_INTERRUPTIBLE); |
1956 | if (!kthread_should_stop() && | |
1957 | (!btrfs_transaction_blocked(root->fs_info) || | |
1958 | cannot_commit)) | |
1959 | schedule_timeout(delay); | |
1960 | __set_current_state(TASK_RUNNING); | |
a74a4b97 CM |
1961 | } while (!kthread_should_stop()); |
1962 | return 0; | |
1963 | } | |
1964 | ||
af31f5e5 CM |
1965 | /* |
1966 | * this will find the highest generation in the array of | |
1967 | * root backups. The index of the highest array is returned, | |
1968 | * or -1 if we can't find anything. | |
1969 | * | |
1970 | * We check to make sure the array is valid by comparing the | |
1971 | * generation of the latest root in the array with the generation | |
1972 | * in the super block. If they don't match we pitch it. | |
1973 | */ | |
1974 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1975 | { | |
1976 | u64 cur; | |
1977 | int newest_index = -1; | |
1978 | struct btrfs_root_backup *root_backup; | |
1979 | int i; | |
1980 | ||
1981 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1982 | root_backup = info->super_copy->super_roots + i; | |
1983 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1984 | if (cur == newest_gen) | |
1985 | newest_index = i; | |
1986 | } | |
1987 | ||
1988 | /* check to see if we actually wrapped around */ | |
1989 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1990 | root_backup = info->super_copy->super_roots; | |
1991 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1992 | if (cur == newest_gen) | |
1993 | newest_index = 0; | |
1994 | } | |
1995 | return newest_index; | |
1996 | } | |
1997 | ||
1998 | ||
1999 | /* | |
2000 | * find the oldest backup so we know where to store new entries | |
2001 | * in the backup array. This will set the backup_root_index | |
2002 | * field in the fs_info struct | |
2003 | */ | |
2004 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
2005 | u64 newest_gen) | |
2006 | { | |
2007 | int newest_index = -1; | |
2008 | ||
2009 | newest_index = find_newest_super_backup(info, newest_gen); | |
2010 | /* if there was garbage in there, just move along */ | |
2011 | if (newest_index == -1) { | |
2012 | info->backup_root_index = 0; | |
2013 | } else { | |
2014 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2015 | } | |
2016 | } | |
2017 | ||
2018 | /* | |
2019 | * copy all the root pointers into the super backup array. | |
2020 | * this will bump the backup pointer by one when it is | |
2021 | * done | |
2022 | */ | |
2023 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2024 | { | |
2025 | int next_backup; | |
2026 | struct btrfs_root_backup *root_backup; | |
2027 | int last_backup; | |
2028 | ||
2029 | next_backup = info->backup_root_index; | |
2030 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2031 | BTRFS_NUM_BACKUP_ROOTS; | |
2032 | ||
2033 | /* | |
2034 | * just overwrite the last backup if we're at the same generation | |
2035 | * this happens only at umount | |
2036 | */ | |
2037 | root_backup = info->super_for_commit->super_roots + last_backup; | |
2038 | if (btrfs_backup_tree_root_gen(root_backup) == | |
2039 | btrfs_header_generation(info->tree_root->node)) | |
2040 | next_backup = last_backup; | |
2041 | ||
2042 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2043 | ||
2044 | /* | |
2045 | * make sure all of our padding and empty slots get zero filled | |
2046 | * regardless of which ones we use today | |
2047 | */ | |
2048 | memset(root_backup, 0, sizeof(*root_backup)); | |
2049 | ||
2050 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2051 | ||
2052 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2053 | btrfs_set_backup_tree_root_gen(root_backup, | |
2054 | btrfs_header_generation(info->tree_root->node)); | |
2055 | ||
2056 | btrfs_set_backup_tree_root_level(root_backup, | |
2057 | btrfs_header_level(info->tree_root->node)); | |
2058 | ||
2059 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2060 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2061 | btrfs_header_generation(info->chunk_root->node)); | |
2062 | btrfs_set_backup_chunk_root_level(root_backup, | |
2063 | btrfs_header_level(info->chunk_root->node)); | |
2064 | ||
2065 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2066 | btrfs_set_backup_extent_root_gen(root_backup, | |
2067 | btrfs_header_generation(info->extent_root->node)); | |
2068 | btrfs_set_backup_extent_root_level(root_backup, | |
2069 | btrfs_header_level(info->extent_root->node)); | |
2070 | ||
7c7e82a7 CM |
2071 | /* |
2072 | * we might commit during log recovery, which happens before we set | |
2073 | * the fs_root. Make sure it is valid before we fill it in. | |
2074 | */ | |
2075 | if (info->fs_root && info->fs_root->node) { | |
2076 | btrfs_set_backup_fs_root(root_backup, | |
2077 | info->fs_root->node->start); | |
2078 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2079 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2080 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2081 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2082 | } |
af31f5e5 CM |
2083 | |
2084 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2085 | btrfs_set_backup_dev_root_gen(root_backup, | |
2086 | btrfs_header_generation(info->dev_root->node)); | |
2087 | btrfs_set_backup_dev_root_level(root_backup, | |
2088 | btrfs_header_level(info->dev_root->node)); | |
2089 | ||
2090 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2091 | btrfs_set_backup_csum_root_gen(root_backup, | |
2092 | btrfs_header_generation(info->csum_root->node)); | |
2093 | btrfs_set_backup_csum_root_level(root_backup, | |
2094 | btrfs_header_level(info->csum_root->node)); | |
2095 | ||
2096 | btrfs_set_backup_total_bytes(root_backup, | |
2097 | btrfs_super_total_bytes(info->super_copy)); | |
2098 | btrfs_set_backup_bytes_used(root_backup, | |
2099 | btrfs_super_bytes_used(info->super_copy)); | |
2100 | btrfs_set_backup_num_devices(root_backup, | |
2101 | btrfs_super_num_devices(info->super_copy)); | |
2102 | ||
2103 | /* | |
2104 | * if we don't copy this out to the super_copy, it won't get remembered | |
2105 | * for the next commit | |
2106 | */ | |
2107 | memcpy(&info->super_copy->super_roots, | |
2108 | &info->super_for_commit->super_roots, | |
2109 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2110 | } | |
2111 | ||
2112 | /* | |
2113 | * this copies info out of the root backup array and back into | |
2114 | * the in-memory super block. It is meant to help iterate through | |
2115 | * the array, so you send it the number of backups you've already | |
2116 | * tried and the last backup index you used. | |
2117 | * | |
2118 | * this returns -1 when it has tried all the backups | |
2119 | */ | |
2120 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2121 | struct btrfs_super_block *super, | |
2122 | int *num_backups_tried, int *backup_index) | |
2123 | { | |
2124 | struct btrfs_root_backup *root_backup; | |
2125 | int newest = *backup_index; | |
2126 | ||
2127 | if (*num_backups_tried == 0) { | |
2128 | u64 gen = btrfs_super_generation(super); | |
2129 | ||
2130 | newest = find_newest_super_backup(info, gen); | |
2131 | if (newest == -1) | |
2132 | return -1; | |
2133 | ||
2134 | *backup_index = newest; | |
2135 | *num_backups_tried = 1; | |
2136 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2137 | /* we've tried all the backups, all done */ | |
2138 | return -1; | |
2139 | } else { | |
2140 | /* jump to the next oldest backup */ | |
2141 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2142 | BTRFS_NUM_BACKUP_ROOTS; | |
2143 | *backup_index = newest; | |
2144 | *num_backups_tried += 1; | |
2145 | } | |
2146 | root_backup = super->super_roots + newest; | |
2147 | ||
2148 | btrfs_set_super_generation(super, | |
2149 | btrfs_backup_tree_root_gen(root_backup)); | |
2150 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2151 | btrfs_set_super_root_level(super, | |
2152 | btrfs_backup_tree_root_level(root_backup)); | |
2153 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2154 | ||
2155 | /* | |
2156 | * fixme: the total bytes and num_devices need to match or we should | |
2157 | * need a fsck | |
2158 | */ | |
2159 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2160 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2161 | return 0; | |
2162 | } | |
2163 | ||
7abadb64 LB |
2164 | /* helper to cleanup workers */ |
2165 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2166 | { | |
dc6e3209 | 2167 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2168 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2169 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 QW |
2170 | btrfs_destroy_workqueue(fs_info->endio_workers); |
2171 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2172 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); | |
8b110e39 | 2173 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2174 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2175 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); |
2176 | btrfs_destroy_workqueue(fs_info->endio_write_workers); | |
2177 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2178 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2179 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2180 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2181 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2182 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2183 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2184 | btrfs_destroy_workqueue(fs_info->extent_workers); |
7abadb64 LB |
2185 | } |
2186 | ||
2e9f5954 R |
2187 | static void free_root_extent_buffers(struct btrfs_root *root) |
2188 | { | |
2189 | if (root) { | |
2190 | free_extent_buffer(root->node); | |
2191 | free_extent_buffer(root->commit_root); | |
2192 | root->node = NULL; | |
2193 | root->commit_root = NULL; | |
2194 | } | |
2195 | } | |
2196 | ||
af31f5e5 CM |
2197 | /* helper to cleanup tree roots */ |
2198 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2199 | { | |
2e9f5954 | 2200 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2201 | |
2e9f5954 R |
2202 | free_root_extent_buffers(info->dev_root); |
2203 | free_root_extent_buffers(info->extent_root); | |
2204 | free_root_extent_buffers(info->csum_root); | |
2205 | free_root_extent_buffers(info->quota_root); | |
2206 | free_root_extent_buffers(info->uuid_root); | |
2207 | if (chunk_root) | |
2208 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2209 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2210 | } |
2211 | ||
faa2dbf0 | 2212 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2213 | { |
2214 | int ret; | |
2215 | struct btrfs_root *gang[8]; | |
2216 | int i; | |
2217 | ||
2218 | while (!list_empty(&fs_info->dead_roots)) { | |
2219 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2220 | struct btrfs_root, root_list); | |
2221 | list_del(&gang[0]->root_list); | |
2222 | ||
27cdeb70 | 2223 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2224 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2225 | } else { |
2226 | free_extent_buffer(gang[0]->node); | |
2227 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2228 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2229 | } |
2230 | } | |
2231 | ||
2232 | while (1) { | |
2233 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2234 | (void **)gang, 0, | |
2235 | ARRAY_SIZE(gang)); | |
2236 | if (!ret) | |
2237 | break; | |
2238 | for (i = 0; i < ret; i++) | |
cb517eab | 2239 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2240 | } |
1a4319cc LB |
2241 | |
2242 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2243 | btrfs_free_log_root_tree(NULL, fs_info); | |
2244 | btrfs_destroy_pinned_extent(fs_info->tree_root, | |
2245 | fs_info->pinned_extents); | |
2246 | } | |
171f6537 | 2247 | } |
af31f5e5 | 2248 | |
638aa7ed ES |
2249 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2250 | { | |
2251 | mutex_init(&fs_info->scrub_lock); | |
2252 | atomic_set(&fs_info->scrubs_running, 0); | |
2253 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2254 | atomic_set(&fs_info->scrubs_paused, 0); | |
2255 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2256 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2257 | fs_info->scrub_workers_refcnt = 0; | |
2258 | } | |
2259 | ||
779a65a4 ES |
2260 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2261 | { | |
2262 | spin_lock_init(&fs_info->balance_lock); | |
2263 | mutex_init(&fs_info->balance_mutex); | |
2264 | atomic_set(&fs_info->balance_running, 0); | |
2265 | atomic_set(&fs_info->balance_pause_req, 0); | |
2266 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2267 | fs_info->balance_ctl = NULL; | |
2268 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2269 | } | |
2270 | ||
f37938e0 ES |
2271 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, |
2272 | struct btrfs_root *tree_root) | |
2273 | { | |
2274 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2275 | set_nlink(fs_info->btree_inode, 1); | |
2276 | /* | |
2277 | * we set the i_size on the btree inode to the max possible int. | |
2278 | * the real end of the address space is determined by all of | |
2279 | * the devices in the system | |
2280 | */ | |
2281 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
2282 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 ES |
2283 | |
2284 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); | |
2285 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, | |
2286 | fs_info->btree_inode->i_mapping); | |
2287 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; | |
2288 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); | |
2289 | ||
2290 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
2291 | ||
2292 | BTRFS_I(fs_info->btree_inode)->root = tree_root; | |
2293 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2294 | sizeof(struct btrfs_key)); | |
2295 | set_bit(BTRFS_INODE_DUMMY, | |
2296 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
2297 | btrfs_insert_inode_hash(fs_info->btree_inode); | |
2298 | } | |
2299 | ||
ad618368 ES |
2300 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2301 | { | |
2302 | fs_info->dev_replace.lock_owner = 0; | |
2303 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2304 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2305 | rwlock_init(&fs_info->dev_replace.lock); |
2306 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2307 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2308 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2309 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2310 | } |
2311 | ||
f9e92e40 ES |
2312 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2313 | { | |
2314 | spin_lock_init(&fs_info->qgroup_lock); | |
2315 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2316 | fs_info->qgroup_tree = RB_ROOT; | |
2317 | fs_info->qgroup_op_tree = RB_ROOT; | |
2318 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2319 | fs_info->qgroup_seq = 1; | |
2320 | fs_info->quota_enabled = 0; | |
2321 | fs_info->pending_quota_state = 0; | |
2322 | fs_info->qgroup_ulist = NULL; | |
d2c609b8 | 2323 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2324 | mutex_init(&fs_info->qgroup_rescan_lock); |
2325 | } | |
2326 | ||
2a458198 ES |
2327 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2328 | struct btrfs_fs_devices *fs_devices) | |
2329 | { | |
2330 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2331 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2332 | |
2333 | fs_info->workers = | |
cb001095 JM |
2334 | btrfs_alloc_workqueue(fs_info, "worker", |
2335 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2336 | |
2337 | fs_info->delalloc_workers = | |
cb001095 JM |
2338 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2339 | flags, max_active, 2); | |
2a458198 ES |
2340 | |
2341 | fs_info->flush_workers = | |
cb001095 JM |
2342 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2343 | flags, max_active, 0); | |
2a458198 ES |
2344 | |
2345 | fs_info->caching_workers = | |
cb001095 | 2346 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 ES |
2347 | |
2348 | /* | |
2349 | * a higher idle thresh on the submit workers makes it much more | |
2350 | * likely that bios will be send down in a sane order to the | |
2351 | * devices | |
2352 | */ | |
2353 | fs_info->submit_workers = | |
cb001095 | 2354 | btrfs_alloc_workqueue(fs_info, "submit", flags, |
2a458198 ES |
2355 | min_t(u64, fs_devices->num_devices, |
2356 | max_active), 64); | |
2357 | ||
2358 | fs_info->fixup_workers = | |
cb001095 | 2359 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2360 | |
2361 | /* | |
2362 | * endios are largely parallel and should have a very | |
2363 | * low idle thresh | |
2364 | */ | |
2365 | fs_info->endio_workers = | |
cb001095 | 2366 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2367 | fs_info->endio_meta_workers = |
cb001095 JM |
2368 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2369 | max_active, 4); | |
2a458198 | 2370 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2371 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2372 | max_active, 2); | |
2a458198 | 2373 | fs_info->endio_raid56_workers = |
cb001095 JM |
2374 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2375 | max_active, 4); | |
2a458198 | 2376 | fs_info->endio_repair_workers = |
cb001095 | 2377 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2378 | fs_info->rmw_workers = |
cb001095 | 2379 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2380 | fs_info->endio_write_workers = |
cb001095 JM |
2381 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2382 | max_active, 2); | |
2a458198 | 2383 | fs_info->endio_freespace_worker = |
cb001095 JM |
2384 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2385 | max_active, 0); | |
2a458198 | 2386 | fs_info->delayed_workers = |
cb001095 JM |
2387 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2388 | max_active, 0); | |
2a458198 | 2389 | fs_info->readahead_workers = |
cb001095 JM |
2390 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2391 | max_active, 2); | |
2a458198 | 2392 | fs_info->qgroup_rescan_workers = |
cb001095 | 2393 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
2a458198 | 2394 | fs_info->extent_workers = |
cb001095 | 2395 | btrfs_alloc_workqueue(fs_info, "extent-refs", flags, |
2a458198 ES |
2396 | min_t(u64, fs_devices->num_devices, |
2397 | max_active), 8); | |
2398 | ||
2399 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2400 | fs_info->submit_workers && fs_info->flush_workers && | |
2401 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2402 | fs_info->endio_meta_write_workers && | |
2403 | fs_info->endio_repair_workers && | |
2404 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2405 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2406 | fs_info->caching_workers && fs_info->readahead_workers && | |
2407 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2408 | fs_info->extent_workers && | |
2409 | fs_info->qgroup_rescan_workers)) { | |
2410 | return -ENOMEM; | |
2411 | } | |
2412 | ||
2413 | return 0; | |
2414 | } | |
2415 | ||
63443bf5 ES |
2416 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2417 | struct btrfs_fs_devices *fs_devices) | |
2418 | { | |
2419 | int ret; | |
2420 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2421 | struct btrfs_root *log_tree_root; | |
2422 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2423 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2424 | ||
2425 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2426 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2427 | return -EIO; |
2428 | } | |
2429 | ||
74e4d827 | 2430 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2431 | if (!log_tree_root) |
2432 | return -ENOMEM; | |
2433 | ||
2434 | __setup_root(tree_root->nodesize, tree_root->sectorsize, | |
2435 | tree_root->stripesize, log_tree_root, fs_info, | |
2436 | BTRFS_TREE_LOG_OBJECTID); | |
2437 | ||
2438 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
2439 | fs_info->generation + 1); | |
64c043de | 2440 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2441 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2442 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2443 | kfree(log_tree_root); |
0eeff236 | 2444 | return ret; |
64c043de | 2445 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2446 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2447 | free_extent_buffer(log_tree_root->node); |
2448 | kfree(log_tree_root); | |
2449 | return -EIO; | |
2450 | } | |
2451 | /* returns with log_tree_root freed on success */ | |
2452 | ret = btrfs_recover_log_trees(log_tree_root); | |
2453 | if (ret) { | |
34d97007 | 2454 | btrfs_handle_fs_error(tree_root->fs_info, ret, |
63443bf5 ES |
2455 | "Failed to recover log tree"); |
2456 | free_extent_buffer(log_tree_root->node); | |
2457 | kfree(log_tree_root); | |
2458 | return ret; | |
2459 | } | |
2460 | ||
2461 | if (fs_info->sb->s_flags & MS_RDONLY) { | |
2462 | ret = btrfs_commit_super(tree_root); | |
2463 | if (ret) | |
2464 | return ret; | |
2465 | } | |
2466 | ||
2467 | return 0; | |
2468 | } | |
2469 | ||
4bbcaa64 ES |
2470 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info, |
2471 | struct btrfs_root *tree_root) | |
2472 | { | |
a4f3d2c4 | 2473 | struct btrfs_root *root; |
4bbcaa64 ES |
2474 | struct btrfs_key location; |
2475 | int ret; | |
2476 | ||
2477 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; | |
2478 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2479 | location.offset = 0; | |
2480 | ||
a4f3d2c4 DS |
2481 | root = btrfs_read_tree_root(tree_root, &location); |
2482 | if (IS_ERR(root)) | |
2483 | return PTR_ERR(root); | |
2484 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2485 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2486 | |
2487 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2488 | root = btrfs_read_tree_root(tree_root, &location); |
2489 | if (IS_ERR(root)) | |
2490 | return PTR_ERR(root); | |
2491 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2492 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2493 | btrfs_init_devices_late(fs_info); |
2494 | ||
2495 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2496 | root = btrfs_read_tree_root(tree_root, &location); |
2497 | if (IS_ERR(root)) | |
2498 | return PTR_ERR(root); | |
2499 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2500 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2501 | |
2502 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2503 | root = btrfs_read_tree_root(tree_root, &location); |
2504 | if (!IS_ERR(root)) { | |
2505 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
4bbcaa64 ES |
2506 | fs_info->quota_enabled = 1; |
2507 | fs_info->pending_quota_state = 1; | |
a4f3d2c4 | 2508 | fs_info->quota_root = root; |
4bbcaa64 ES |
2509 | } |
2510 | ||
2511 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2512 | root = btrfs_read_tree_root(tree_root, &location); |
2513 | if (IS_ERR(root)) { | |
2514 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2515 | if (ret != -ENOENT) |
2516 | return ret; | |
2517 | } else { | |
a4f3d2c4 DS |
2518 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2519 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2520 | } |
2521 | ||
70f6d82e OS |
2522 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2523 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2524 | root = btrfs_read_tree_root(tree_root, &location); | |
2525 | if (IS_ERR(root)) | |
2526 | return PTR_ERR(root); | |
2527 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2528 | fs_info->free_space_root = root; | |
2529 | } | |
2530 | ||
4bbcaa64 ES |
2531 | return 0; |
2532 | } | |
2533 | ||
ad2b2c80 AV |
2534 | int open_ctree(struct super_block *sb, |
2535 | struct btrfs_fs_devices *fs_devices, | |
2536 | char *options) | |
2e635a27 | 2537 | { |
db94535d CM |
2538 | u32 sectorsize; |
2539 | u32 nodesize; | |
87ee04eb | 2540 | u32 stripesize; |
84234f3a | 2541 | u64 generation; |
f2b636e8 | 2542 | u64 features; |
3de4586c | 2543 | struct btrfs_key location; |
a061fc8d | 2544 | struct buffer_head *bh; |
4d34b278 | 2545 | struct btrfs_super_block *disk_super; |
815745cf | 2546 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2547 | struct btrfs_root *tree_root; |
4d34b278 | 2548 | struct btrfs_root *chunk_root; |
eb60ceac | 2549 | int ret; |
e58ca020 | 2550 | int err = -EINVAL; |
af31f5e5 CM |
2551 | int num_backups_tried = 0; |
2552 | int backup_index = 0; | |
5cdc7ad3 | 2553 | int max_active; |
4543df7e | 2554 | |
74e4d827 DS |
2555 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2556 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2557 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2558 | err = -ENOMEM; |
2559 | goto fail; | |
2560 | } | |
76dda93c YZ |
2561 | |
2562 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2563 | if (ret) { | |
2564 | err = ret; | |
2565 | goto fail; | |
2566 | } | |
2567 | ||
2568 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2569 | if (ret) { | |
2570 | err = ret; | |
2571 | goto fail_srcu; | |
2572 | } | |
2573 | ||
908c7f19 | 2574 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2575 | if (ret) { |
2576 | err = ret; | |
2577 | goto fail_bdi; | |
2578 | } | |
09cbfeaf | 2579 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2580 | (1 + ilog2(nr_cpu_ids)); |
2581 | ||
908c7f19 | 2582 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2583 | if (ret) { |
2584 | err = ret; | |
2585 | goto fail_dirty_metadata_bytes; | |
2586 | } | |
2587 | ||
908c7f19 | 2588 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2589 | if (ret) { |
2590 | err = ret; | |
2591 | goto fail_delalloc_bytes; | |
2592 | } | |
2593 | ||
76dda93c YZ |
2594 | fs_info->btree_inode = new_inode(sb); |
2595 | if (!fs_info->btree_inode) { | |
2596 | err = -ENOMEM; | |
c404e0dc | 2597 | goto fail_bio_counter; |
76dda93c YZ |
2598 | } |
2599 | ||
a6591715 | 2600 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2601 | |
76dda93c | 2602 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2603 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2604 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2605 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2606 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2607 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2608 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2609 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2610 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2611 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2612 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2613 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2614 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2615 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2616 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2617 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2618 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2619 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2620 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2621 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2622 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2623 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2624 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2625 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2626 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2627 | |
0b86a832 | 2628 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2629 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2630 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2631 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2632 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2633 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2634 | BTRFS_BLOCK_RSV_GLOBAL); | |
2635 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2636 | BTRFS_BLOCK_RSV_DELALLOC); | |
2637 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2638 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2639 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2640 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2641 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2642 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2643 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2644 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2645 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2646 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2647 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2648 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2649 | atomic64_set(&fs_info->tree_mod_seq, 0); |
9e7cc91a | 2650 | fs_info->fs_frozen = 0; |
e20d96d6 | 2651 | fs_info->sb = sb; |
95ac567a | 2652 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2653 | fs_info->metadata_ratio = 0; |
4cb5300b | 2654 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2655 | fs_info->free_chunk_space = 0; |
f29021b2 | 2656 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2657 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2658 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2659 | /* readahead state */ |
d0164adc | 2660 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2661 | spin_lock_init(&fs_info->reada_lock); |
c8b97818 | 2662 | |
b34b086c CM |
2663 | fs_info->thread_pool_size = min_t(unsigned long, |
2664 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2665 | |
199c2a9c MX |
2666 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2667 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 | 2668 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2669 | GFP_KERNEL); |
16cdcec7 MX |
2670 | if (!fs_info->delayed_root) { |
2671 | err = -ENOMEM; | |
2672 | goto fail_iput; | |
2673 | } | |
2674 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2675 | |
638aa7ed | 2676 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2677 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2678 | fs_info->check_integrity_print_mask = 0; | |
2679 | #endif | |
779a65a4 | 2680 | btrfs_init_balance(fs_info); |
21c7e756 | 2681 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2682 | |
a061fc8d CM |
2683 | sb->s_blocksize = 4096; |
2684 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2685 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2686 | |
f37938e0 | 2687 | btrfs_init_btree_inode(fs_info, tree_root); |
76dda93c | 2688 | |
0f9dd46c | 2689 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2690 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2691 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2692 | |
11833d66 | 2693 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2694 | fs_info->btree_inode->i_mapping); |
11833d66 | 2695 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2696 | fs_info->btree_inode->i_mapping); |
11833d66 | 2697 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2698 | fs_info->do_barriers = 1; |
e18e4809 | 2699 | |
39279cc3 | 2700 | |
5a3f23d5 | 2701 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2702 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2703 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2704 | mutex_init(&fs_info->transaction_kthread_mutex); |
2705 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2706 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2707 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2708 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2709 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2710 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2711 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2712 | |
ad618368 | 2713 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2714 | btrfs_init_qgroup(fs_info); |
416ac51d | 2715 | |
fa9c0d79 CM |
2716 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2717 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2718 | ||
e6dcd2dc | 2719 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2720 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2721 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2722 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2723 | |
04216820 FM |
2724 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2725 | ||
53b381b3 DW |
2726 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2727 | if (ret) { | |
83c8266a | 2728 | err = ret; |
53b381b3 DW |
2729 | goto fail_alloc; |
2730 | } | |
2731 | ||
707e8a07 | 2732 | __setup_root(4096, 4096, 4096, tree_root, |
2c90e5d6 | 2733 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2734 | |
3c4bb26b | 2735 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2736 | |
2737 | /* | |
2738 | * Read super block and check the signature bytes only | |
2739 | */ | |
a512bbf8 | 2740 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2741 | if (IS_ERR(bh)) { |
2742 | err = PTR_ERR(bh); | |
16cdcec7 | 2743 | goto fail_alloc; |
20b45077 | 2744 | } |
39279cc3 | 2745 | |
1104a885 DS |
2746 | /* |
2747 | * We want to check superblock checksum, the type is stored inside. | |
2748 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2749 | */ | |
2750 | if (btrfs_check_super_csum(bh->b_data)) { | |
05135f59 | 2751 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2752 | err = -EINVAL; |
b2acdddf | 2753 | brelse(bh); |
1104a885 DS |
2754 | goto fail_alloc; |
2755 | } | |
2756 | ||
2757 | /* | |
2758 | * super_copy is zeroed at allocation time and we never touch the | |
2759 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2760 | * the whole block of INFO_SIZE | |
2761 | */ | |
6c41761f DS |
2762 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2763 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2764 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2765 | brelse(bh); |
5f39d397 | 2766 | |
6c41761f | 2767 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2768 | |
1104a885 DS |
2769 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2770 | if (ret) { | |
05135f59 | 2771 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 DS |
2772 | err = -EINVAL; |
2773 | goto fail_alloc; | |
2774 | } | |
2775 | ||
6c41761f | 2776 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2777 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2778 | goto fail_alloc; |
0f7d52f4 | 2779 | |
acce952b | 2780 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2781 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2782 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2783 | |
af31f5e5 CM |
2784 | /* |
2785 | * run through our array of backup supers and setup | |
2786 | * our ring pointer to the oldest one | |
2787 | */ | |
2788 | generation = btrfs_super_generation(disk_super); | |
2789 | find_oldest_super_backup(fs_info, generation); | |
2790 | ||
75e7cb7f LB |
2791 | /* |
2792 | * In the long term, we'll store the compression type in the super | |
2793 | * block, and it'll be used for per file compression control. | |
2794 | */ | |
2795 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2796 | ||
96da0919 | 2797 | ret = btrfs_parse_options(tree_root, options, sb->s_flags); |
2b82032c YZ |
2798 | if (ret) { |
2799 | err = ret; | |
16cdcec7 | 2800 | goto fail_alloc; |
2b82032c | 2801 | } |
dfe25020 | 2802 | |
f2b636e8 JB |
2803 | features = btrfs_super_incompat_flags(disk_super) & |
2804 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2805 | if (features) { | |
05135f59 DS |
2806 | btrfs_err(fs_info, |
2807 | "cannot mount because of unsupported optional features (%llx)", | |
2808 | features); | |
f2b636e8 | 2809 | err = -EINVAL; |
16cdcec7 | 2810 | goto fail_alloc; |
f2b636e8 JB |
2811 | } |
2812 | ||
5d4f98a2 | 2813 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2814 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2815 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2816 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2817 | |
3173a18f | 2818 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2819 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2820 | |
727011e0 CM |
2821 | /* |
2822 | * flag our filesystem as having big metadata blocks if | |
2823 | * they are bigger than the page size | |
2824 | */ | |
09cbfeaf | 2825 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2826 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2827 | btrfs_info(fs_info, |
2828 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2829 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2830 | } | |
2831 | ||
bc3f116f | 2832 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2833 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2834 | stripesize = sectorsize; |
707e8a07 | 2835 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2836 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2837 | |
2838 | /* | |
2839 | * mixed block groups end up with duplicate but slightly offset | |
2840 | * extent buffers for the same range. It leads to corruptions | |
2841 | */ | |
2842 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2843 | (sectorsize != nodesize)) { |
05135f59 DS |
2844 | btrfs_err(fs_info, |
2845 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2846 | nodesize, sectorsize); | |
bc3f116f CM |
2847 | goto fail_alloc; |
2848 | } | |
2849 | ||
ceda0864 MX |
2850 | /* |
2851 | * Needn't use the lock because there is no other task which will | |
2852 | * update the flag. | |
2853 | */ | |
a6fa6fae | 2854 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2855 | |
f2b636e8 JB |
2856 | features = btrfs_super_compat_ro_flags(disk_super) & |
2857 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2858 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
05135f59 DS |
2859 | btrfs_err(fs_info, |
2860 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2861 | features); |
f2b636e8 | 2862 | err = -EINVAL; |
16cdcec7 | 2863 | goto fail_alloc; |
f2b636e8 | 2864 | } |
61d92c32 | 2865 | |
5cdc7ad3 | 2866 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2867 | |
2a458198 ES |
2868 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2869 | if (ret) { | |
2870 | err = ret; | |
0dc3b84a JB |
2871 | goto fail_sb_buffer; |
2872 | } | |
4543df7e | 2873 | |
4575c9cc | 2874 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 | 2875 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
09cbfeaf | 2876 | SZ_4M / PAGE_SIZE); |
4575c9cc | 2877 | |
db94535d | 2878 | tree_root->nodesize = nodesize; |
db94535d | 2879 | tree_root->sectorsize = sectorsize; |
87ee04eb | 2880 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2881 | |
2882 | sb->s_blocksize = sectorsize; | |
2883 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2884 | |
925baedd | 2885 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2886 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2887 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2888 | if (ret) { |
05135f59 | 2889 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 2890 | goto fail_sb_buffer; |
84eed90f | 2891 | } |
0b86a832 | 2892 | |
84234f3a | 2893 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2894 | |
707e8a07 DS |
2895 | __setup_root(nodesize, sectorsize, stripesize, chunk_root, |
2896 | fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
0b86a832 CM |
2897 | |
2898 | chunk_root->node = read_tree_block(chunk_root, | |
2899 | btrfs_super_chunk_root(disk_super), | |
ce86cd59 | 2900 | generation); |
64c043de LB |
2901 | if (IS_ERR(chunk_root->node) || |
2902 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 2903 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 2904 | if (!IS_ERR(chunk_root->node)) |
2905 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2906 | chunk_root->node = NULL; |
af31f5e5 | 2907 | goto fail_tree_roots; |
83121942 | 2908 | } |
5d4f98a2 YZ |
2909 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2910 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2911 | |
e17cade2 | 2912 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2913 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2914 | |
0b86a832 | 2915 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2916 | if (ret) { |
05135f59 | 2917 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 2918 | goto fail_tree_roots; |
2b82032c | 2919 | } |
0b86a832 | 2920 | |
8dabb742 SB |
2921 | /* |
2922 | * keep the device that is marked to be the target device for the | |
2923 | * dev_replace procedure | |
2924 | */ | |
9eaed21e | 2925 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2926 | |
a6b0d5c8 | 2927 | if (!fs_devices->latest_bdev) { |
05135f59 | 2928 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
2929 | goto fail_tree_roots; |
2930 | } | |
2931 | ||
af31f5e5 | 2932 | retry_root_backup: |
84234f3a | 2933 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2934 | |
e20d96d6 | 2935 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2936 | btrfs_super_root(disk_super), |
ce86cd59 | 2937 | generation); |
64c043de LB |
2938 | if (IS_ERR(tree_root->node) || |
2939 | !extent_buffer_uptodate(tree_root->node)) { | |
05135f59 | 2940 | btrfs_warn(fs_info, "failed to read tree root"); |
e5fffbac | 2941 | if (!IS_ERR(tree_root->node)) |
2942 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2943 | tree_root->node = NULL; |
af31f5e5 | 2944 | goto recovery_tree_root; |
83121942 | 2945 | } |
af31f5e5 | 2946 | |
5d4f98a2 YZ |
2947 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2948 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2949 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2950 | |
f32e48e9 CR |
2951 | mutex_lock(&tree_root->objectid_mutex); |
2952 | ret = btrfs_find_highest_objectid(tree_root, | |
2953 | &tree_root->highest_objectid); | |
2954 | if (ret) { | |
2955 | mutex_unlock(&tree_root->objectid_mutex); | |
2956 | goto recovery_tree_root; | |
2957 | } | |
2958 | ||
2959 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2960 | ||
2961 | mutex_unlock(&tree_root->objectid_mutex); | |
2962 | ||
4bbcaa64 ES |
2963 | ret = btrfs_read_roots(fs_info, tree_root); |
2964 | if (ret) | |
af31f5e5 | 2965 | goto recovery_tree_root; |
f7a81ea4 | 2966 | |
8929ecfa YZ |
2967 | fs_info->generation = generation; |
2968 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2969 | |
68310a5e ID |
2970 | ret = btrfs_recover_balance(fs_info); |
2971 | if (ret) { | |
05135f59 | 2972 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
2973 | goto fail_block_groups; |
2974 | } | |
2975 | ||
733f4fbb SB |
2976 | ret = btrfs_init_dev_stats(fs_info); |
2977 | if (ret) { | |
05135f59 | 2978 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
2979 | goto fail_block_groups; |
2980 | } | |
2981 | ||
8dabb742 SB |
2982 | ret = btrfs_init_dev_replace(fs_info); |
2983 | if (ret) { | |
05135f59 | 2984 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
2985 | goto fail_block_groups; |
2986 | } | |
2987 | ||
9eaed21e | 2988 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 2989 | |
b7c35e81 AJ |
2990 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
2991 | if (ret) { | |
05135f59 DS |
2992 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
2993 | ret); | |
b7c35e81 AJ |
2994 | goto fail_block_groups; |
2995 | } | |
2996 | ||
2997 | ret = btrfs_sysfs_add_device(fs_devices); | |
2998 | if (ret) { | |
05135f59 DS |
2999 | btrfs_err(fs_info, "failed to init sysfs device interface: %d", |
3000 | ret); | |
b7c35e81 AJ |
3001 | goto fail_fsdev_sysfs; |
3002 | } | |
3003 | ||
96f3136e | 3004 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3005 | if (ret) { |
05135f59 | 3006 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3007 | goto fail_fsdev_sysfs; |
c59021f8 | 3008 | } |
3009 | ||
c59021f8 | 3010 | ret = btrfs_init_space_info(fs_info); |
3011 | if (ret) { | |
05135f59 | 3012 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3013 | goto fail_sysfs; |
c59021f8 | 3014 | } |
3015 | ||
4bbcaa64 | 3016 | ret = btrfs_read_block_groups(fs_info->extent_root); |
1b1d1f66 | 3017 | if (ret) { |
05135f59 | 3018 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3019 | goto fail_sysfs; |
1b1d1f66 | 3020 | } |
5af3e8cc SB |
3021 | fs_info->num_tolerated_disk_barrier_failures = |
3022 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
3023 | if (fs_info->fs_devices->missing_devices > |
3024 | fs_info->num_tolerated_disk_barrier_failures && | |
3025 | !(sb->s_flags & MS_RDONLY)) { | |
05135f59 DS |
3026 | btrfs_warn(fs_info, |
3027 | "missing devices (%llu) exceeds the limit (%d), writeable mount is not allowed", | |
78fa1770 ZL |
3028 | fs_info->fs_devices->missing_devices, |
3029 | fs_info->num_tolerated_disk_barrier_failures); | |
2365dd3c | 3030 | goto fail_sysfs; |
292fd7fc | 3031 | } |
9078a3e1 | 3032 | |
a74a4b97 CM |
3033 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3034 | "btrfs-cleaner"); | |
57506d50 | 3035 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3036 | goto fail_sysfs; |
a74a4b97 CM |
3037 | |
3038 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3039 | tree_root, | |
3040 | "btrfs-transaction"); | |
57506d50 | 3041 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3042 | goto fail_cleaner; |
a74a4b97 | 3043 | |
3cdde224 JM |
3044 | if (!btrfs_test_opt(tree_root->fs_info, SSD) && |
3045 | !btrfs_test_opt(tree_root->fs_info, NOSSD) && | |
c289811c | 3046 | !fs_info->fs_devices->rotating) { |
05135f59 | 3047 | btrfs_info(fs_info, "detected SSD devices, enabling SSD mode"); |
c289811c CM |
3048 | btrfs_set_opt(fs_info->mount_opt, SSD); |
3049 | } | |
3050 | ||
572d9ab7 | 3051 | /* |
01327610 | 3052 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3053 | * not have to wait for transaction commit |
3054 | */ | |
3055 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3056 | |
21adbd5c | 3057 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3cdde224 | 3058 | if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) { |
21adbd5c | 3059 | ret = btrfsic_mount(tree_root, fs_devices, |
3cdde224 | 3060 | btrfs_test_opt(tree_root->fs_info, |
21adbd5c SB |
3061 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
3062 | 1 : 0, | |
3063 | fs_info->check_integrity_print_mask); | |
3064 | if (ret) | |
05135f59 DS |
3065 | btrfs_warn(fs_info, |
3066 | "failed to initialize integrity check module: %d", | |
3067 | ret); | |
21adbd5c SB |
3068 | } |
3069 | #endif | |
bcef60f2 AJ |
3070 | ret = btrfs_read_qgroup_config(fs_info); |
3071 | if (ret) | |
3072 | goto fail_trans_kthread; | |
21adbd5c | 3073 | |
96da0919 QW |
3074 | /* do not make disk changes in broken FS or nologreplay is given */ |
3075 | if (btrfs_super_log_root(disk_super) != 0 && | |
3cdde224 | 3076 | !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) { |
63443bf5 | 3077 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3078 | if (ret) { |
63443bf5 | 3079 | err = ret; |
28c16cbb | 3080 | goto fail_qgroup; |
79787eaa | 3081 | } |
e02119d5 | 3082 | } |
1a40e23b | 3083 | |
76dda93c | 3084 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa | 3085 | if (ret) |
28c16cbb | 3086 | goto fail_qgroup; |
76dda93c | 3087 | |
7c2ca468 | 3088 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 3089 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3090 | if (ret) |
28c16cbb | 3091 | goto fail_qgroup; |
90c711ab ZB |
3092 | |
3093 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 3094 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 3095 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 3096 | if (ret < 0) { |
05135f59 DS |
3097 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
3098 | ret); | |
d7ce5843 | 3099 | err = -EINVAL; |
bcef60f2 | 3100 | goto fail_qgroup; |
d7ce5843 | 3101 | } |
7c2ca468 | 3102 | } |
1a40e23b | 3103 | |
3de4586c CM |
3104 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3105 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3106 | location.offset = 0; |
3de4586c | 3107 | |
3de4586c | 3108 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3109 | if (IS_ERR(fs_info->fs_root)) { |
3110 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 3111 | goto fail_qgroup; |
3140c9a3 | 3112 | } |
c289811c | 3113 | |
2b6ba629 ID |
3114 | if (sb->s_flags & MS_RDONLY) |
3115 | return 0; | |
59641015 | 3116 | |
3cdde224 | 3117 | if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) && |
511711af | 3118 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3119 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
3120 | ret = btrfs_create_free_space_tree(fs_info); |
3121 | if (ret) { | |
05135f59 DS |
3122 | btrfs_warn(fs_info, |
3123 | "failed to create free space tree: %d", ret); | |
511711af CM |
3124 | close_ctree(tree_root); |
3125 | return ret; | |
3126 | } | |
3127 | } | |
3128 | ||
2b6ba629 ID |
3129 | down_read(&fs_info->cleanup_work_sem); |
3130 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3131 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3132 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
3133 | close_ctree(tree_root); |
3134 | return ret; | |
3135 | } | |
3136 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3137 | |
2b6ba629 ID |
3138 | ret = btrfs_resume_balance_async(fs_info); |
3139 | if (ret) { | |
05135f59 | 3140 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
2b6ba629 ID |
3141 | close_ctree(tree_root); |
3142 | return ret; | |
e3acc2a6 JB |
3143 | } |
3144 | ||
8dabb742 SB |
3145 | ret = btrfs_resume_dev_replace_async(fs_info); |
3146 | if (ret) { | |
05135f59 | 3147 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
8dabb742 SB |
3148 | close_ctree(tree_root); |
3149 | return ret; | |
3150 | } | |
3151 | ||
b382a324 JS |
3152 | btrfs_qgroup_rescan_resume(fs_info); |
3153 | ||
3cdde224 | 3154 | if (btrfs_test_opt(tree_root->fs_info, CLEAR_CACHE) && |
70f6d82e | 3155 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 3156 | btrfs_info(fs_info, "clearing free space tree"); |
70f6d82e OS |
3157 | ret = btrfs_clear_free_space_tree(fs_info); |
3158 | if (ret) { | |
05135f59 DS |
3159 | btrfs_warn(fs_info, |
3160 | "failed to clear free space tree: %d", ret); | |
70f6d82e OS |
3161 | close_ctree(tree_root); |
3162 | return ret; | |
3163 | } | |
3164 | } | |
3165 | ||
4bbcaa64 | 3166 | if (!fs_info->uuid_root) { |
05135f59 | 3167 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3168 | ret = btrfs_create_uuid_tree(fs_info); |
3169 | if (ret) { | |
05135f59 DS |
3170 | btrfs_warn(fs_info, |
3171 | "failed to create the UUID tree: %d", ret); | |
f7a81ea4 SB |
3172 | close_ctree(tree_root); |
3173 | return ret; | |
3174 | } | |
3cdde224 | 3175 | } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3176 | fs_info->generation != |
3177 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3178 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3179 | ret = btrfs_check_uuid_tree(fs_info); |
3180 | if (ret) { | |
05135f59 DS |
3181 | btrfs_warn(fs_info, |
3182 | "failed to check the UUID tree: %d", ret); | |
70f80175 SB |
3183 | close_ctree(tree_root); |
3184 | return ret; | |
3185 | } | |
3186 | } else { | |
3187 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
3188 | } |
3189 | ||
47ab2a6c JB |
3190 | fs_info->open = 1; |
3191 | ||
8dcddfa0 QW |
3192 | /* |
3193 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3194 | * no need to keep the flag | |
3195 | */ | |
3196 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3197 | ||
ad2b2c80 | 3198 | return 0; |
39279cc3 | 3199 | |
bcef60f2 AJ |
3200 | fail_qgroup: |
3201 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3202 | fail_trans_kthread: |
3203 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 3204 | btrfs_cleanup_transaction(fs_info->tree_root); |
faa2dbf0 | 3205 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3206 | fail_cleaner: |
a74a4b97 | 3207 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3208 | |
3209 | /* | |
3210 | * make sure we're done with the btree inode before we stop our | |
3211 | * kthreads | |
3212 | */ | |
3213 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3214 | |
2365dd3c | 3215 | fail_sysfs: |
6618a59b | 3216 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3217 | |
b7c35e81 AJ |
3218 | fail_fsdev_sysfs: |
3219 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3220 | ||
1b1d1f66 | 3221 | fail_block_groups: |
54067ae9 | 3222 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 3223 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
3224 | |
3225 | fail_tree_roots: | |
3226 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3227 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3228 | |
39279cc3 | 3229 | fail_sb_buffer: |
7abadb64 | 3230 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 3231 | fail_alloc: |
4543df7e | 3232 | fail_iput: |
586e46e2 ID |
3233 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3234 | ||
4543df7e | 3235 | iput(fs_info->btree_inode); |
c404e0dc MX |
3236 | fail_bio_counter: |
3237 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3238 | fail_delalloc_bytes: |
3239 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3240 | fail_dirty_metadata_bytes: |
3241 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 3242 | fail_bdi: |
7e662854 | 3243 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
3244 | fail_srcu: |
3245 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3246 | fail: |
53b381b3 | 3247 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3248 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3249 | return err; |
af31f5e5 CM |
3250 | |
3251 | recovery_tree_root: | |
3cdde224 | 3252 | if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT)) |
af31f5e5 CM |
3253 | goto fail_tree_roots; |
3254 | ||
3255 | free_root_pointers(fs_info, 0); | |
3256 | ||
3257 | /* don't use the log in recovery mode, it won't be valid */ | |
3258 | btrfs_set_super_log_root(disk_super, 0); | |
3259 | ||
3260 | /* we can't trust the free space cache either */ | |
3261 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3262 | ||
3263 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3264 | &num_backups_tried, &backup_index); | |
3265 | if (ret == -1) | |
3266 | goto fail_block_groups; | |
3267 | goto retry_root_backup; | |
eb60ceac CM |
3268 | } |
3269 | ||
f2984462 CM |
3270 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3271 | { | |
f2984462 CM |
3272 | if (uptodate) { |
3273 | set_buffer_uptodate(bh); | |
3274 | } else { | |
442a4f63 SB |
3275 | struct btrfs_device *device = (struct btrfs_device *) |
3276 | bh->b_private; | |
3277 | ||
b14af3b4 DS |
3278 | btrfs_warn_rl_in_rcu(device->dev_root->fs_info, |
3279 | "lost page write due to IO error on %s", | |
606686ee | 3280 | rcu_str_deref(device->name)); |
01327610 | 3281 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3282 | * our own ways of dealing with the IO errors |
3283 | */ | |
f2984462 | 3284 | clear_buffer_uptodate(bh); |
442a4f63 | 3285 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3286 | } |
3287 | unlock_buffer(bh); | |
3288 | put_bh(bh); | |
3289 | } | |
3290 | ||
29c36d72 AJ |
3291 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3292 | struct buffer_head **bh_ret) | |
3293 | { | |
3294 | struct buffer_head *bh; | |
3295 | struct btrfs_super_block *super; | |
3296 | u64 bytenr; | |
3297 | ||
3298 | bytenr = btrfs_sb_offset(copy_num); | |
3299 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3300 | return -EINVAL; | |
3301 | ||
3302 | bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); | |
3303 | /* | |
3304 | * If we fail to read from the underlying devices, as of now | |
3305 | * the best option we have is to mark it EIO. | |
3306 | */ | |
3307 | if (!bh) | |
3308 | return -EIO; | |
3309 | ||
3310 | super = (struct btrfs_super_block *)bh->b_data; | |
3311 | if (btrfs_super_bytenr(super) != bytenr || | |
3312 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3313 | brelse(bh); | |
3314 | return -EINVAL; | |
3315 | } | |
3316 | ||
3317 | *bh_ret = bh; | |
3318 | return 0; | |
3319 | } | |
3320 | ||
3321 | ||
a512bbf8 YZ |
3322 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3323 | { | |
3324 | struct buffer_head *bh; | |
3325 | struct buffer_head *latest = NULL; | |
3326 | struct btrfs_super_block *super; | |
3327 | int i; | |
3328 | u64 transid = 0; | |
92fc03fb | 3329 | int ret = -EINVAL; |
a512bbf8 YZ |
3330 | |
3331 | /* we would like to check all the supers, but that would make | |
3332 | * a btrfs mount succeed after a mkfs from a different FS. | |
3333 | * So, we need to add a special mount option to scan for | |
3334 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3335 | */ | |
3336 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3337 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3338 | if (ret) | |
a512bbf8 YZ |
3339 | continue; |
3340 | ||
3341 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3342 | |
3343 | if (!latest || btrfs_super_generation(super) > transid) { | |
3344 | brelse(latest); | |
3345 | latest = bh; | |
3346 | transid = btrfs_super_generation(super); | |
3347 | } else { | |
3348 | brelse(bh); | |
3349 | } | |
3350 | } | |
92fc03fb AJ |
3351 | |
3352 | if (!latest) | |
3353 | return ERR_PTR(ret); | |
3354 | ||
a512bbf8 YZ |
3355 | return latest; |
3356 | } | |
3357 | ||
4eedeb75 HH |
3358 | /* |
3359 | * this should be called twice, once with wait == 0 and | |
3360 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3361 | * we write are pinned. | |
3362 | * | |
3363 | * They are released when wait == 1 is done. | |
3364 | * max_mirrors must be the same for both runs, and it indicates how | |
3365 | * many supers on this one device should be written. | |
3366 | * | |
3367 | * max_mirrors == 0 means to write them all. | |
3368 | */ | |
a512bbf8 YZ |
3369 | static int write_dev_supers(struct btrfs_device *device, |
3370 | struct btrfs_super_block *sb, | |
3371 | int do_barriers, int wait, int max_mirrors) | |
3372 | { | |
3373 | struct buffer_head *bh; | |
3374 | int i; | |
3375 | int ret; | |
3376 | int errors = 0; | |
3377 | u32 crc; | |
3378 | u64 bytenr; | |
a512bbf8 YZ |
3379 | |
3380 | if (max_mirrors == 0) | |
3381 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3382 | ||
a512bbf8 YZ |
3383 | for (i = 0; i < max_mirrors; i++) { |
3384 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3385 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3386 | device->commit_total_bytes) | |
a512bbf8 YZ |
3387 | break; |
3388 | ||
3389 | if (wait) { | |
3390 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3391 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3392 | if (!bh) { |
3393 | errors++; | |
3394 | continue; | |
3395 | } | |
a512bbf8 | 3396 | wait_on_buffer(bh); |
4eedeb75 HH |
3397 | if (!buffer_uptodate(bh)) |
3398 | errors++; | |
3399 | ||
3400 | /* drop our reference */ | |
3401 | brelse(bh); | |
3402 | ||
3403 | /* drop the reference from the wait == 0 run */ | |
3404 | brelse(bh); | |
3405 | continue; | |
a512bbf8 YZ |
3406 | } else { |
3407 | btrfs_set_super_bytenr(sb, bytenr); | |
3408 | ||
3409 | crc = ~(u32)0; | |
b0496686 | 3410 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3411 | BTRFS_CSUM_SIZE, crc, |
3412 | BTRFS_SUPER_INFO_SIZE - | |
3413 | BTRFS_CSUM_SIZE); | |
3414 | btrfs_csum_final(crc, sb->csum); | |
3415 | ||
4eedeb75 HH |
3416 | /* |
3417 | * one reference for us, and we leave it for the | |
3418 | * caller | |
3419 | */ | |
a512bbf8 YZ |
3420 | bh = __getblk(device->bdev, bytenr / 4096, |
3421 | BTRFS_SUPER_INFO_SIZE); | |
634554dc | 3422 | if (!bh) { |
f14d104d DS |
3423 | btrfs_err(device->dev_root->fs_info, |
3424 | "couldn't get super buffer head for bytenr %llu", | |
3425 | bytenr); | |
634554dc JB |
3426 | errors++; |
3427 | continue; | |
3428 | } | |
3429 | ||
a512bbf8 YZ |
3430 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3431 | ||
4eedeb75 | 3432 | /* one reference for submit_bh */ |
a512bbf8 | 3433 | get_bh(bh); |
4eedeb75 HH |
3434 | |
3435 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3436 | lock_buffer(bh); |
3437 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3438 | bh->b_private = device; |
a512bbf8 YZ |
3439 | } |
3440 | ||
387125fc CM |
3441 | /* |
3442 | * we fua the first super. The others we allow | |
3443 | * to go down lazy. | |
3444 | */ | |
e8117c26 WS |
3445 | if (i == 0) |
3446 | ret = btrfsic_submit_bh(WRITE_FUA, bh); | |
3447 | else | |
3448 | ret = btrfsic_submit_bh(WRITE_SYNC, bh); | |
4eedeb75 | 3449 | if (ret) |
a512bbf8 | 3450 | errors++; |
a512bbf8 YZ |
3451 | } |
3452 | return errors < i ? 0 : -1; | |
3453 | } | |
3454 | ||
387125fc CM |
3455 | /* |
3456 | * endio for the write_dev_flush, this will wake anyone waiting | |
3457 | * for the barrier when it is done | |
3458 | */ | |
4246a0b6 | 3459 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3460 | { |
387125fc CM |
3461 | if (bio->bi_private) |
3462 | complete(bio->bi_private); | |
3463 | bio_put(bio); | |
3464 | } | |
3465 | ||
3466 | /* | |
3467 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3468 | * sent down. With wait == 1, it waits for the previous flush. | |
3469 | * | |
3470 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3471 | * capable | |
3472 | */ | |
3473 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3474 | { | |
3475 | struct bio *bio; | |
3476 | int ret = 0; | |
3477 | ||
3478 | if (device->nobarriers) | |
3479 | return 0; | |
3480 | ||
3481 | if (wait) { | |
3482 | bio = device->flush_bio; | |
3483 | if (!bio) | |
3484 | return 0; | |
3485 | ||
3486 | wait_for_completion(&device->flush_wait); | |
3487 | ||
4246a0b6 CH |
3488 | if (bio->bi_error) { |
3489 | ret = bio->bi_error; | |
5af3e8cc SB |
3490 | btrfs_dev_stat_inc_and_print(device, |
3491 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3492 | } |
3493 | ||
3494 | /* drop the reference from the wait == 0 run */ | |
3495 | bio_put(bio); | |
3496 | device->flush_bio = NULL; | |
3497 | ||
3498 | return ret; | |
3499 | } | |
3500 | ||
3501 | /* | |
3502 | * one reference for us, and we leave it for the | |
3503 | * caller | |
3504 | */ | |
9c017abc | 3505 | device->flush_bio = NULL; |
9be3395b | 3506 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3507 | if (!bio) |
3508 | return -ENOMEM; | |
3509 | ||
3510 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3511 | bio->bi_bdev = device->bdev; | |
3512 | init_completion(&device->flush_wait); | |
3513 | bio->bi_private = &device->flush_wait; | |
3514 | device->flush_bio = bio; | |
3515 | ||
3516 | bio_get(bio); | |
21adbd5c | 3517 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3518 | |
3519 | return 0; | |
3520 | } | |
3521 | ||
3522 | /* | |
3523 | * send an empty flush down to each device in parallel, | |
3524 | * then wait for them | |
3525 | */ | |
3526 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3527 | { | |
3528 | struct list_head *head; | |
3529 | struct btrfs_device *dev; | |
5af3e8cc SB |
3530 | int errors_send = 0; |
3531 | int errors_wait = 0; | |
387125fc CM |
3532 | int ret; |
3533 | ||
3534 | /* send down all the barriers */ | |
3535 | head = &info->fs_devices->devices; | |
3536 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3537 | if (dev->missing) |
3538 | continue; | |
387125fc | 3539 | if (!dev->bdev) { |
5af3e8cc | 3540 | errors_send++; |
387125fc CM |
3541 | continue; |
3542 | } | |
3543 | if (!dev->in_fs_metadata || !dev->writeable) | |
3544 | continue; | |
3545 | ||
3546 | ret = write_dev_flush(dev, 0); | |
3547 | if (ret) | |
5af3e8cc | 3548 | errors_send++; |
387125fc CM |
3549 | } |
3550 | ||
3551 | /* wait for all the barriers */ | |
3552 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3553 | if (dev->missing) |
3554 | continue; | |
387125fc | 3555 | if (!dev->bdev) { |
5af3e8cc | 3556 | errors_wait++; |
387125fc CM |
3557 | continue; |
3558 | } | |
3559 | if (!dev->in_fs_metadata || !dev->writeable) | |
3560 | continue; | |
3561 | ||
3562 | ret = write_dev_flush(dev, 1); | |
3563 | if (ret) | |
5af3e8cc | 3564 | errors_wait++; |
387125fc | 3565 | } |
5af3e8cc SB |
3566 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3567 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3568 | return -EIO; |
3569 | return 0; | |
3570 | } | |
3571 | ||
943c6e99 ZL |
3572 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3573 | { | |
8789f4fe ZL |
3574 | int raid_type; |
3575 | int min_tolerated = INT_MAX; | |
943c6e99 | 3576 | |
8789f4fe ZL |
3577 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3578 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3579 | min_tolerated = min(min_tolerated, | |
3580 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3581 | tolerated_failures); | |
943c6e99 | 3582 | |
8789f4fe ZL |
3583 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3584 | if (raid_type == BTRFS_RAID_SINGLE) | |
3585 | continue; | |
3586 | if (!(flags & btrfs_raid_group[raid_type])) | |
3587 | continue; | |
3588 | min_tolerated = min(min_tolerated, | |
3589 | btrfs_raid_array[raid_type]. | |
3590 | tolerated_failures); | |
3591 | } | |
943c6e99 | 3592 | |
8789f4fe ZL |
3593 | if (min_tolerated == INT_MAX) { |
3594 | pr_warn("BTRFS: unknown raid flag: %llu\n", flags); | |
3595 | min_tolerated = 0; | |
3596 | } | |
3597 | ||
3598 | return min_tolerated; | |
943c6e99 ZL |
3599 | } |
3600 | ||
5af3e8cc SB |
3601 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3602 | struct btrfs_fs_info *fs_info) | |
3603 | { | |
3604 | struct btrfs_ioctl_space_info space; | |
3605 | struct btrfs_space_info *sinfo; | |
3606 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3607 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3608 | BTRFS_BLOCK_GROUP_METADATA, | |
3609 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
5af3e8cc SB |
3610 | int i; |
3611 | int c; | |
3612 | int num_tolerated_disk_barrier_failures = | |
3613 | (int)fs_info->fs_devices->num_devices; | |
3614 | ||
2c458045 | 3615 | for (i = 0; i < ARRAY_SIZE(types); i++) { |
5af3e8cc SB |
3616 | struct btrfs_space_info *tmp; |
3617 | ||
3618 | sinfo = NULL; | |
3619 | rcu_read_lock(); | |
3620 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3621 | if (tmp->flags == types[i]) { | |
3622 | sinfo = tmp; | |
3623 | break; | |
3624 | } | |
3625 | } | |
3626 | rcu_read_unlock(); | |
3627 | ||
3628 | if (!sinfo) | |
3629 | continue; | |
3630 | ||
3631 | down_read(&sinfo->groups_sem); | |
3632 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2c458045 ZL |
3633 | u64 flags; |
3634 | ||
3635 | if (list_empty(&sinfo->block_groups[c])) | |
3636 | continue; | |
3637 | ||
3638 | btrfs_get_block_group_info(&sinfo->block_groups[c], | |
3639 | &space); | |
3640 | if (space.total_bytes == 0 || space.used_bytes == 0) | |
3641 | continue; | |
3642 | flags = space.flags; | |
943c6e99 ZL |
3643 | |
3644 | num_tolerated_disk_barrier_failures = min( | |
3645 | num_tolerated_disk_barrier_failures, | |
3646 | btrfs_get_num_tolerated_disk_barrier_failures( | |
3647 | flags)); | |
5af3e8cc SB |
3648 | } |
3649 | up_read(&sinfo->groups_sem); | |
3650 | } | |
3651 | ||
3652 | return num_tolerated_disk_barrier_failures; | |
3653 | } | |
3654 | ||
48a3b636 | 3655 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3656 | { |
e5e9a520 | 3657 | struct list_head *head; |
f2984462 | 3658 | struct btrfs_device *dev; |
a061fc8d | 3659 | struct btrfs_super_block *sb; |
f2984462 | 3660 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3661 | int ret; |
3662 | int do_barriers; | |
a236aed1 CM |
3663 | int max_errors; |
3664 | int total_errors = 0; | |
a061fc8d | 3665 | u64 flags; |
f2984462 | 3666 | |
3cdde224 | 3667 | do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER); |
af31f5e5 | 3668 | backup_super_roots(root->fs_info); |
f2984462 | 3669 | |
6c41761f | 3670 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3671 | dev_item = &sb->dev_item; |
e5e9a520 | 3672 | |
174ba509 | 3673 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3674 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3675 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3676 | |
5af3e8cc SB |
3677 | if (do_barriers) { |
3678 | ret = barrier_all_devices(root->fs_info); | |
3679 | if (ret) { | |
3680 | mutex_unlock( | |
3681 | &root->fs_info->fs_devices->device_list_mutex); | |
34d97007 | 3682 | btrfs_handle_fs_error(root->fs_info, ret, |
5af3e8cc SB |
3683 | "errors while submitting device barriers."); |
3684 | return ret; | |
3685 | } | |
3686 | } | |
387125fc | 3687 | |
1f78160c | 3688 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3689 | if (!dev->bdev) { |
3690 | total_errors++; | |
3691 | continue; | |
3692 | } | |
2b82032c | 3693 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3694 | continue; |
3695 | ||
2b82032c | 3696 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3697 | btrfs_set_stack_device_type(dev_item, dev->type); |
3698 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3699 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3700 | dev->commit_total_bytes); |
ce7213c7 MX |
3701 | btrfs_set_stack_device_bytes_used(dev_item, |
3702 | dev->commit_bytes_used); | |
a061fc8d CM |
3703 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3704 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3705 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3706 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3707 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3708 | |
a061fc8d CM |
3709 | flags = btrfs_super_flags(sb); |
3710 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3711 | ||
a512bbf8 | 3712 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3713 | if (ret) |
3714 | total_errors++; | |
f2984462 | 3715 | } |
a236aed1 | 3716 | if (total_errors > max_errors) { |
efe120a0 | 3717 | btrfs_err(root->fs_info, "%d errors while writing supers", |
d397712b | 3718 | total_errors); |
a724b436 | 3719 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3720 | |
9d565ba4 | 3721 | /* FUA is masked off if unsupported and can't be the reason */ |
34d97007 | 3722 | btrfs_handle_fs_error(root->fs_info, -EIO, |
9d565ba4 SB |
3723 | "%d errors while writing supers", total_errors); |
3724 | return -EIO; | |
a236aed1 | 3725 | } |
f2984462 | 3726 | |
a512bbf8 | 3727 | total_errors = 0; |
1f78160c | 3728 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3729 | if (!dev->bdev) |
3730 | continue; | |
2b82032c | 3731 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3732 | continue; |
3733 | ||
a512bbf8 YZ |
3734 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3735 | if (ret) | |
3736 | total_errors++; | |
f2984462 | 3737 | } |
174ba509 | 3738 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3739 | if (total_errors > max_errors) { |
34d97007 | 3740 | btrfs_handle_fs_error(root->fs_info, -EIO, |
79787eaa JM |
3741 | "%d errors while writing supers", total_errors); |
3742 | return -EIO; | |
a236aed1 | 3743 | } |
f2984462 CM |
3744 | return 0; |
3745 | } | |
3746 | ||
a512bbf8 YZ |
3747 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3748 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3749 | { |
f570e757 | 3750 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3751 | } |
3752 | ||
cb517eab MX |
3753 | /* Drop a fs root from the radix tree and free it. */ |
3754 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3755 | struct btrfs_root *root) | |
2619ba1f | 3756 | { |
4df27c4d | 3757 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3758 | radix_tree_delete(&fs_info->fs_roots_radix, |
3759 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3760 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3761 | |
3762 | if (btrfs_root_refs(&root->root_item) == 0) | |
3763 | synchronize_srcu(&fs_info->subvol_srcu); | |
3764 | ||
1c1ea4f7 | 3765 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3766 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3767 | if (root->reloc_root) { |
3768 | free_extent_buffer(root->reloc_root->node); | |
3769 | free_extent_buffer(root->reloc_root->commit_root); | |
3770 | btrfs_put_fs_root(root->reloc_root); | |
3771 | root->reloc_root = NULL; | |
3772 | } | |
3773 | } | |
3321719e | 3774 | |
faa2dbf0 JB |
3775 | if (root->free_ino_pinned) |
3776 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3777 | if (root->free_ino_ctl) | |
3778 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3779 | free_fs_root(root); |
4df27c4d YZ |
3780 | } |
3781 | ||
3782 | static void free_fs_root(struct btrfs_root *root) | |
3783 | { | |
57cdc8db | 3784 | iput(root->ino_cache_inode); |
4df27c4d | 3785 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3786 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3787 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3788 | if (root->anon_dev) |
3789 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3790 | if (root->subv_writers) |
3791 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3792 | free_extent_buffer(root->node); |
3793 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3794 | kfree(root->free_ino_ctl); |
3795 | kfree(root->free_ino_pinned); | |
d397712b | 3796 | kfree(root->name); |
b0feb9d9 | 3797 | btrfs_put_fs_root(root); |
2619ba1f CM |
3798 | } |
3799 | ||
cb517eab MX |
3800 | void btrfs_free_fs_root(struct btrfs_root *root) |
3801 | { | |
3802 | free_fs_root(root); | |
2619ba1f CM |
3803 | } |
3804 | ||
c146afad | 3805 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3806 | { |
c146afad YZ |
3807 | u64 root_objectid = 0; |
3808 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3809 | int i = 0; |
3810 | int err = 0; | |
3811 | unsigned int ret = 0; | |
3812 | int index; | |
e089f05c | 3813 | |
c146afad | 3814 | while (1) { |
65d33fd7 | 3815 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3816 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3817 | (void **)gang, root_objectid, | |
3818 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3819 | if (!ret) { |
3820 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3821 | break; |
65d33fd7 | 3822 | } |
5d4f98a2 | 3823 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3824 | |
c146afad | 3825 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3826 | /* Avoid to grab roots in dead_roots */ |
3827 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3828 | gang[i] = NULL; | |
3829 | continue; | |
3830 | } | |
3831 | /* grab all the search result for later use */ | |
3832 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3833 | } | |
3834 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3835 | |
65d33fd7 QW |
3836 | for (i = 0; i < ret; i++) { |
3837 | if (!gang[i]) | |
3838 | continue; | |
c146afad | 3839 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3840 | err = btrfs_orphan_cleanup(gang[i]); |
3841 | if (err) | |
65d33fd7 QW |
3842 | break; |
3843 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3844 | } |
3845 | root_objectid++; | |
3846 | } | |
65d33fd7 QW |
3847 | |
3848 | /* release the uncleaned roots due to error */ | |
3849 | for (; i < ret; i++) { | |
3850 | if (gang[i]) | |
3851 | btrfs_put_fs_root(gang[i]); | |
3852 | } | |
3853 | return err; | |
c146afad | 3854 | } |
a2135011 | 3855 | |
c146afad YZ |
3856 | int btrfs_commit_super(struct btrfs_root *root) |
3857 | { | |
3858 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3859 | |
c146afad | 3860 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3861 | btrfs_run_delayed_iputs(root); |
c146afad | 3862 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3863 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3864 | |
3865 | /* wait until ongoing cleanup work done */ | |
3866 | down_write(&root->fs_info->cleanup_work_sem); | |
3867 | up_write(&root->fs_info->cleanup_work_sem); | |
3868 | ||
7a7eaa40 | 3869 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3870 | if (IS_ERR(trans)) |
3871 | return PTR_ERR(trans); | |
d52c1bcc | 3872 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3873 | } |
3874 | ||
3abdbd78 | 3875 | void close_ctree(struct btrfs_root *root) |
c146afad YZ |
3876 | { |
3877 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3878 | int ret; | |
3879 | ||
3880 | fs_info->closing = 1; | |
3881 | smp_mb(); | |
3882 | ||
7343dd61 | 3883 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3884 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3885 | |
803b2f54 SB |
3886 | /* wait for the uuid_scan task to finish */ |
3887 | down(&fs_info->uuid_tree_rescan_sem); | |
3888 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3889 | up(&fs_info->uuid_tree_rescan_sem); | |
3890 | ||
837d5b6e | 3891 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3892 | btrfs_pause_balance(fs_info); |
837d5b6e | 3893 | |
8dabb742 SB |
3894 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3895 | ||
aa1b8cd4 | 3896 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3897 | |
3898 | /* wait for any defraggers to finish */ | |
3899 | wait_event(fs_info->transaction_wait, | |
3900 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3901 | ||
3902 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3903 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3904 | |
21c7e756 MX |
3905 | cancel_work_sync(&fs_info->async_reclaim_work); |
3906 | ||
c146afad | 3907 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
e44163e1 JM |
3908 | /* |
3909 | * If the cleaner thread is stopped and there are | |
3910 | * block groups queued for removal, the deletion will be | |
3911 | * skipped when we quit the cleaner thread. | |
3912 | */ | |
e44163e1 | 3913 | btrfs_delete_unused_bgs(root->fs_info); |
e44163e1 | 3914 | |
acce952b | 3915 | ret = btrfs_commit_super(root); |
3916 | if (ret) | |
04892340 | 3917 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3918 | } |
3919 | ||
87533c47 | 3920 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3921 | btrfs_error_commit_super(root); |
0f7d52f4 | 3922 | |
e3029d9f AV |
3923 | kthread_stop(fs_info->transaction_kthread); |
3924 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3925 | |
f25784b3 YZ |
3926 | fs_info->closing = 2; |
3927 | smp_mb(); | |
3928 | ||
04892340 | 3929 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3930 | |
963d678b | 3931 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3932 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3933 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3934 | } |
bcc63abb | 3935 | |
6618a59b | 3936 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3937 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3938 | |
faa2dbf0 | 3939 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3940 | |
1a4319cc LB |
3941 | btrfs_put_block_group_cache(fs_info); |
3942 | ||
2b1360da JB |
3943 | btrfs_free_block_groups(fs_info); |
3944 | ||
de348ee0 WS |
3945 | /* |
3946 | * we must make sure there is not any read request to | |
3947 | * submit after we stopping all workers. | |
3948 | */ | |
3949 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3950 | btrfs_stop_all_workers(fs_info); |
3951 | ||
47ab2a6c | 3952 | fs_info->open = 0; |
13e6c37b | 3953 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3954 | |
13e6c37b | 3955 | iput(fs_info->btree_inode); |
d6bfde87 | 3956 | |
21adbd5c | 3957 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3cdde224 | 3958 | if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY)) |
21adbd5c SB |
3959 | btrfsic_unmount(root, fs_info->fs_devices); |
3960 | #endif | |
3961 | ||
dfe25020 | 3962 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3963 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3964 | |
e2d84521 | 3965 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3966 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3967 | percpu_counter_destroy(&fs_info->bio_counter); |
04160088 | 3968 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3969 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3970 | |
53b381b3 DW |
3971 | btrfs_free_stripe_hash_table(fs_info); |
3972 | ||
cdfb080e | 3973 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3974 | root->orphan_block_rsv = NULL; |
04216820 FM |
3975 | |
3976 | lock_chunks(root); | |
3977 | while (!list_empty(&fs_info->pinned_chunks)) { | |
3978 | struct extent_map *em; | |
3979 | ||
3980 | em = list_first_entry(&fs_info->pinned_chunks, | |
3981 | struct extent_map, list); | |
3982 | list_del_init(&em->list); | |
3983 | free_extent_map(em); | |
3984 | } | |
3985 | unlock_chunks(root); | |
eb60ceac CM |
3986 | } |
3987 | ||
b9fab919 CM |
3988 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3989 | int atomic) | |
5f39d397 | 3990 | { |
1259ab75 | 3991 | int ret; |
727011e0 | 3992 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3993 | |
0b32f4bb | 3994 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3995 | if (!ret) |
3996 | return ret; | |
3997 | ||
3998 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3999 | parent_transid, atomic); |
4000 | if (ret == -EAGAIN) | |
4001 | return ret; | |
1259ab75 | 4002 | return !ret; |
5f39d397 CM |
4003 | } |
4004 | ||
5f39d397 CM |
4005 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4006 | { | |
06ea65a3 | 4007 | struct btrfs_root *root; |
5f39d397 | 4008 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4009 | int was_dirty; |
b4ce94de | 4010 | |
06ea65a3 JB |
4011 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4012 | /* | |
4013 | * This is a fast path so only do this check if we have sanity tests | |
4014 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
4015 | * outside of the sanity tests. | |
4016 | */ | |
4017 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
4018 | return; | |
4019 | #endif | |
4020 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 4021 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
4022 | if (transid != root->fs_info->generation) |
4023 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 4024 | "found %llu running %llu\n", |
c1c9ff7c | 4025 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 4026 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
4027 | if (!was_dirty) |
4028 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
4029 | buf->len, | |
4030 | root->fs_info->dirty_metadata_batch); | |
1f21ef0a FM |
4031 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4032 | if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { | |
4033 | btrfs_print_leaf(root, buf); | |
4034 | ASSERT(0); | |
4035 | } | |
4036 | #endif | |
eb60ceac CM |
4037 | } |
4038 | ||
b53d3f5d LB |
4039 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
4040 | int flush_delayed) | |
16cdcec7 MX |
4041 | { |
4042 | /* | |
4043 | * looks as though older kernels can get into trouble with | |
4044 | * this code, they end up stuck in balance_dirty_pages forever | |
4045 | */ | |
e2d84521 | 4046 | int ret; |
16cdcec7 MX |
4047 | |
4048 | if (current->flags & PF_MEMALLOC) | |
4049 | return; | |
4050 | ||
b53d3f5d LB |
4051 | if (flush_delayed) |
4052 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 4053 | |
e2d84521 MX |
4054 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
4055 | BTRFS_DIRTY_METADATA_THRESH); | |
4056 | if (ret > 0) { | |
d0e1d66b NJ |
4057 | balance_dirty_pages_ratelimited( |
4058 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 | 4059 | } |
16cdcec7 MX |
4060 | } |
4061 | ||
b53d3f5d | 4062 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 4063 | { |
b53d3f5d LB |
4064 | __btrfs_btree_balance_dirty(root, 1); |
4065 | } | |
585ad2c3 | 4066 | |
b53d3f5d LB |
4067 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
4068 | { | |
4069 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 4070 | } |
6b80053d | 4071 | |
ca7a79ad | 4072 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 4073 | { |
727011e0 | 4074 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 4075 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 4076 | } |
0da5468f | 4077 | |
fcd1f065 | 4078 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 4079 | int read_only) |
4080 | { | |
c926093e | 4081 | struct btrfs_super_block *sb = fs_info->super_copy; |
319e4d06 QW |
4082 | u64 nodesize = btrfs_super_nodesize(sb); |
4083 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
c926093e DS |
4084 | int ret = 0; |
4085 | ||
319e4d06 QW |
4086 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { |
4087 | printk(KERN_ERR "BTRFS: no valid FS found\n"); | |
4088 | ret = -EINVAL; | |
4089 | } | |
4090 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) | |
4091 | printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n", | |
4092 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
21e7626b DS |
4093 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4094 | printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", | |
4095 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4096 | ret = -EINVAL; |
4097 | } | |
21e7626b DS |
4098 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4099 | printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", | |
4100 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4101 | ret = -EINVAL; |
4102 | } | |
21e7626b DS |
4103 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4104 | printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", | |
4105 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4106 | ret = -EINVAL; |
4107 | } | |
4108 | ||
1104a885 | 4109 | /* |
319e4d06 QW |
4110 | * Check sectorsize and nodesize first, other check will need it. |
4111 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
1104a885 | 4112 | */ |
319e4d06 QW |
4113 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || |
4114 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4115 | printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize); | |
4116 | ret = -EINVAL; | |
4117 | } | |
4118 | /* Only PAGE SIZE is supported yet */ | |
09cbfeaf | 4119 | if (sectorsize != PAGE_SIZE) { |
319e4d06 | 4120 | printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n", |
09cbfeaf | 4121 | sectorsize, PAGE_SIZE); |
319e4d06 QW |
4122 | ret = -EINVAL; |
4123 | } | |
4124 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
4125 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
4126 | printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize); | |
4127 | ret = -EINVAL; | |
4128 | } | |
4129 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
4130 | printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n", | |
4131 | le32_to_cpu(sb->__unused_leafsize), | |
4132 | nodesize); | |
4133 | ret = -EINVAL; | |
4134 | } | |
4135 | ||
4136 | /* Root alignment check */ | |
4137 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
c926093e | 4138 | printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", |
cd743fac | 4139 | btrfs_super_root(sb)); |
319e4d06 QW |
4140 | ret = -EINVAL; |
4141 | } | |
4142 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
cd743fac DS |
4143 | printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", |
4144 | btrfs_super_chunk_root(sb)); | |
75d6ad38 DS |
4145 | ret = -EINVAL; |
4146 | } | |
319e4d06 QW |
4147 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { |
4148 | printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", | |
4149 | btrfs_super_log_root(sb)); | |
75d6ad38 DS |
4150 | ret = -EINVAL; |
4151 | } | |
4152 | ||
c926093e DS |
4153 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { |
4154 | printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", | |
4155 | fs_info->fsid, sb->dev_item.fsid); | |
4156 | ret = -EINVAL; | |
4157 | } | |
4158 | ||
4159 | /* | |
4160 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
4161 | * done later | |
4162 | */ | |
99e3ecfc LB |
4163 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { |
4164 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
4165 | btrfs_super_bytes_used(sb)); | |
4166 | ret = -EINVAL; | |
4167 | } | |
b7f67055 | 4168 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { |
99e3ecfc LB |
4169 | btrfs_err(fs_info, "invalid stripesize %u", |
4170 | btrfs_super_stripesize(sb)); | |
4171 | ret = -EINVAL; | |
4172 | } | |
21e7626b | 4173 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
c926093e | 4174 | printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", |
21e7626b | 4175 | btrfs_super_num_devices(sb)); |
75d6ad38 DS |
4176 | if (btrfs_super_num_devices(sb) == 0) { |
4177 | printk(KERN_ERR "BTRFS: number of devices is 0\n"); | |
4178 | ret = -EINVAL; | |
4179 | } | |
c926093e | 4180 | |
21e7626b | 4181 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
c926093e | 4182 | printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", |
21e7626b | 4183 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); |
c926093e DS |
4184 | ret = -EINVAL; |
4185 | } | |
4186 | ||
ce7fca5f DS |
4187 | /* |
4188 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4189 | * and one chunk | |
4190 | */ | |
4191 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
4192 | printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", | |
4193 | btrfs_super_sys_array_size(sb), | |
4194 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
4195 | ret = -EINVAL; | |
4196 | } | |
4197 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4198 | + sizeof(struct btrfs_chunk)) { | |
d2207129 | 4199 | printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", |
ce7fca5f DS |
4200 | btrfs_super_sys_array_size(sb), |
4201 | sizeof(struct btrfs_disk_key) | |
4202 | + sizeof(struct btrfs_chunk)); | |
4203 | ret = -EINVAL; | |
4204 | } | |
4205 | ||
c926093e DS |
4206 | /* |
4207 | * The generation is a global counter, we'll trust it more than the others | |
4208 | * but it's still possible that it's the one that's wrong. | |
4209 | */ | |
21e7626b | 4210 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
c926093e DS |
4211 | printk(KERN_WARNING |
4212 | "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", | |
21e7626b DS |
4213 | btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); |
4214 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
4215 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
c926093e DS |
4216 | printk(KERN_WARNING |
4217 | "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", | |
21e7626b | 4218 | btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); |
c926093e DS |
4219 | |
4220 | return ret; | |
acce952b | 4221 | } |
4222 | ||
48a3b636 | 4223 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 4224 | { |
acce952b | 4225 | mutex_lock(&root->fs_info->cleaner_mutex); |
4226 | btrfs_run_delayed_iputs(root); | |
4227 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4228 | ||
4229 | down_write(&root->fs_info->cleanup_work_sem); | |
4230 | up_write(&root->fs_info->cleanup_work_sem); | |
4231 | ||
4232 | /* cleanup FS via transaction */ | |
4233 | btrfs_cleanup_transaction(root); | |
acce952b | 4234 | } |
4235 | ||
143bede5 | 4236 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4237 | { |
acce952b | 4238 | struct btrfs_ordered_extent *ordered; |
acce952b | 4239 | |
199c2a9c | 4240 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4241 | /* |
4242 | * This will just short circuit the ordered completion stuff which will | |
4243 | * make sure the ordered extent gets properly cleaned up. | |
4244 | */ | |
199c2a9c | 4245 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4246 | root_extent_list) |
4247 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4248 | spin_unlock(&root->ordered_extent_lock); |
4249 | } | |
4250 | ||
4251 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4252 | { | |
4253 | struct btrfs_root *root; | |
4254 | struct list_head splice; | |
4255 | ||
4256 | INIT_LIST_HEAD(&splice); | |
4257 | ||
4258 | spin_lock(&fs_info->ordered_root_lock); | |
4259 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4260 | while (!list_empty(&splice)) { | |
4261 | root = list_first_entry(&splice, struct btrfs_root, | |
4262 | ordered_root); | |
1de2cfde JB |
4263 | list_move_tail(&root->ordered_root, |
4264 | &fs_info->ordered_roots); | |
199c2a9c | 4265 | |
2a85d9ca | 4266 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4267 | btrfs_destroy_ordered_extents(root); |
4268 | ||
2a85d9ca LB |
4269 | cond_resched(); |
4270 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4271 | } |
4272 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4273 | } |
4274 | ||
35a3621b SB |
4275 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4276 | struct btrfs_root *root) | |
acce952b | 4277 | { |
4278 | struct rb_node *node; | |
4279 | struct btrfs_delayed_ref_root *delayed_refs; | |
4280 | struct btrfs_delayed_ref_node *ref; | |
4281 | int ret = 0; | |
4282 | ||
4283 | delayed_refs = &trans->delayed_refs; | |
4284 | ||
4285 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4286 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4287 | spin_unlock(&delayed_refs->lock); |
efe120a0 | 4288 | btrfs_info(root->fs_info, "delayed_refs has NO entry"); |
acce952b | 4289 | return ret; |
4290 | } | |
4291 | ||
d7df2c79 JB |
4292 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4293 | struct btrfs_delayed_ref_head *head; | |
c6fc2454 | 4294 | struct btrfs_delayed_ref_node *tmp; |
e78417d1 | 4295 | bool pin_bytes = false; |
acce952b | 4296 | |
d7df2c79 JB |
4297 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4298 | href_node); | |
4299 | if (!mutex_trylock(&head->mutex)) { | |
4300 | atomic_inc(&head->node.refs); | |
4301 | spin_unlock(&delayed_refs->lock); | |
eb12db69 | 4302 | |
d7df2c79 | 4303 | mutex_lock(&head->mutex); |
e78417d1 | 4304 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
4305 | btrfs_put_delayed_ref(&head->node); |
4306 | spin_lock(&delayed_refs->lock); | |
4307 | continue; | |
4308 | } | |
4309 | spin_lock(&head->lock); | |
c6fc2454 QW |
4310 | list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, |
4311 | list) { | |
d7df2c79 | 4312 | ref->in_tree = 0; |
c6fc2454 | 4313 | list_del(&ref->list); |
d7df2c79 JB |
4314 | atomic_dec(&delayed_refs->num_entries); |
4315 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4316 | } |
d7df2c79 JB |
4317 | if (head->must_insert_reserved) |
4318 | pin_bytes = true; | |
4319 | btrfs_free_delayed_extent_op(head->extent_op); | |
4320 | delayed_refs->num_heads--; | |
4321 | if (head->processing == 0) | |
4322 | delayed_refs->num_heads_ready--; | |
4323 | atomic_dec(&delayed_refs->num_entries); | |
4324 | head->node.in_tree = 0; | |
4325 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
4326 | spin_unlock(&head->lock); | |
4327 | spin_unlock(&delayed_refs->lock); | |
4328 | mutex_unlock(&head->mutex); | |
acce952b | 4329 | |
d7df2c79 JB |
4330 | if (pin_bytes) |
4331 | btrfs_pin_extent(root, head->node.bytenr, | |
4332 | head->node.num_bytes, 1); | |
4333 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 4334 | cond_resched(); |
4335 | spin_lock(&delayed_refs->lock); | |
4336 | } | |
4337 | ||
4338 | spin_unlock(&delayed_refs->lock); | |
4339 | ||
4340 | return ret; | |
4341 | } | |
4342 | ||
143bede5 | 4343 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4344 | { |
4345 | struct btrfs_inode *btrfs_inode; | |
4346 | struct list_head splice; | |
4347 | ||
4348 | INIT_LIST_HEAD(&splice); | |
4349 | ||
eb73c1b7 MX |
4350 | spin_lock(&root->delalloc_lock); |
4351 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4352 | |
4353 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4354 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4355 | delalloc_inodes); | |
acce952b | 4356 | |
4357 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4358 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4359 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4360 | spin_unlock(&root->delalloc_lock); |
acce952b | 4361 | |
4362 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4363 | |
eb73c1b7 | 4364 | spin_lock(&root->delalloc_lock); |
acce952b | 4365 | } |
4366 | ||
eb73c1b7 MX |
4367 | spin_unlock(&root->delalloc_lock); |
4368 | } | |
4369 | ||
4370 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4371 | { | |
4372 | struct btrfs_root *root; | |
4373 | struct list_head splice; | |
4374 | ||
4375 | INIT_LIST_HEAD(&splice); | |
4376 | ||
4377 | spin_lock(&fs_info->delalloc_root_lock); | |
4378 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4379 | while (!list_empty(&splice)) { | |
4380 | root = list_first_entry(&splice, struct btrfs_root, | |
4381 | delalloc_root); | |
4382 | list_del_init(&root->delalloc_root); | |
4383 | root = btrfs_grab_fs_root(root); | |
4384 | BUG_ON(!root); | |
4385 | spin_unlock(&fs_info->delalloc_root_lock); | |
4386 | ||
4387 | btrfs_destroy_delalloc_inodes(root); | |
4388 | btrfs_put_fs_root(root); | |
4389 | ||
4390 | spin_lock(&fs_info->delalloc_root_lock); | |
4391 | } | |
4392 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4393 | } |
4394 | ||
4395 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
4396 | struct extent_io_tree *dirty_pages, | |
4397 | int mark) | |
4398 | { | |
4399 | int ret; | |
acce952b | 4400 | struct extent_buffer *eb; |
4401 | u64 start = 0; | |
4402 | u64 end; | |
acce952b | 4403 | |
4404 | while (1) { | |
4405 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4406 | mark, NULL); |
acce952b | 4407 | if (ret) |
4408 | break; | |
4409 | ||
91166212 | 4410 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4411 | while (start <= end) { |
01d58472 | 4412 | eb = btrfs_find_tree_block(root->fs_info, start); |
707e8a07 | 4413 | start += root->nodesize; |
fd8b2b61 | 4414 | if (!eb) |
acce952b | 4415 | continue; |
fd8b2b61 | 4416 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4417 | |
fd8b2b61 JB |
4418 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4419 | &eb->bflags)) | |
4420 | clear_extent_buffer_dirty(eb); | |
4421 | free_extent_buffer_stale(eb); | |
acce952b | 4422 | } |
4423 | } | |
4424 | ||
4425 | return ret; | |
4426 | } | |
4427 | ||
4428 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
4429 | struct extent_io_tree *pinned_extents) | |
4430 | { | |
4431 | struct extent_io_tree *unpin; | |
4432 | u64 start; | |
4433 | u64 end; | |
4434 | int ret; | |
ed0eaa14 | 4435 | bool loop = true; |
acce952b | 4436 | |
4437 | unpin = pinned_extents; | |
ed0eaa14 | 4438 | again: |
acce952b | 4439 | while (1) { |
4440 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4441 | EXTENT_DIRTY, NULL); |
acce952b | 4442 | if (ret) |
4443 | break; | |
4444 | ||
af6f8f60 | 4445 | clear_extent_dirty(unpin, start, end); |
acce952b | 4446 | btrfs_error_unpin_extent_range(root, start, end); |
4447 | cond_resched(); | |
4448 | } | |
4449 | ||
ed0eaa14 LB |
4450 | if (loop) { |
4451 | if (unpin == &root->fs_info->freed_extents[0]) | |
4452 | unpin = &root->fs_info->freed_extents[1]; | |
4453 | else | |
4454 | unpin = &root->fs_info->freed_extents[0]; | |
4455 | loop = false; | |
4456 | goto again; | |
4457 | } | |
4458 | ||
acce952b | 4459 | return 0; |
4460 | } | |
4461 | ||
49b25e05 JM |
4462 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
4463 | struct btrfs_root *root) | |
4464 | { | |
4465 | btrfs_destroy_delayed_refs(cur_trans, root); | |
49b25e05 | 4466 | |
4a9d8bde | 4467 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 4468 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 4469 | |
4a9d8bde | 4470 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 4471 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 4472 | |
67cde344 MX |
4473 | btrfs_destroy_delayed_inodes(root); |
4474 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4475 | |
49b25e05 JM |
4476 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4477 | EXTENT_DIRTY); | |
6e841e32 LB |
4478 | btrfs_destroy_pinned_extent(root, |
4479 | root->fs_info->pinned_extents); | |
49b25e05 | 4480 | |
4a9d8bde MX |
4481 | cur_trans->state =TRANS_STATE_COMPLETED; |
4482 | wake_up(&cur_trans->commit_wait); | |
4483 | ||
49b25e05 JM |
4484 | /* |
4485 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4486 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4487 | */ | |
4488 | } | |
4489 | ||
48a3b636 | 4490 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4491 | { |
4492 | struct btrfs_transaction *t; | |
acce952b | 4493 | |
acce952b | 4494 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4495 | ||
a4abeea4 | 4496 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4497 | while (!list_empty(&root->fs_info->trans_list)) { |
4498 | t = list_first_entry(&root->fs_info->trans_list, | |
4499 | struct btrfs_transaction, list); | |
4500 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4501 | atomic_inc(&t->use_count); | |
4502 | spin_unlock(&root->fs_info->trans_lock); | |
4503 | btrfs_wait_for_commit(root, t->transid); | |
4504 | btrfs_put_transaction(t); | |
4505 | spin_lock(&root->fs_info->trans_lock); | |
4506 | continue; | |
4507 | } | |
4508 | if (t == root->fs_info->running_transaction) { | |
4509 | t->state = TRANS_STATE_COMMIT_DOING; | |
4510 | spin_unlock(&root->fs_info->trans_lock); | |
4511 | /* | |
4512 | * We wait for 0 num_writers since we don't hold a trans | |
4513 | * handle open currently for this transaction. | |
4514 | */ | |
4515 | wait_event(t->writer_wait, | |
4516 | atomic_read(&t->num_writers) == 0); | |
4517 | } else { | |
4518 | spin_unlock(&root->fs_info->trans_lock); | |
4519 | } | |
4520 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4521 | |
724e2315 JB |
4522 | spin_lock(&root->fs_info->trans_lock); |
4523 | if (t == root->fs_info->running_transaction) | |
4524 | root->fs_info->running_transaction = NULL; | |
acce952b | 4525 | list_del_init(&t->list); |
724e2315 | 4526 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4527 | |
724e2315 JB |
4528 | btrfs_put_transaction(t); |
4529 | trace_btrfs_transaction_commit(root); | |
4530 | spin_lock(&root->fs_info->trans_lock); | |
4531 | } | |
4532 | spin_unlock(&root->fs_info->trans_lock); | |
4533 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4534 | btrfs_destroy_delayed_inodes(root); | |
4535 | btrfs_assert_delayed_root_empty(root); | |
4536 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4537 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4538 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4539 | |
4540 | return 0; | |
4541 | } | |
4542 | ||
e8c9f186 | 4543 | static const struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4544 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4545 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4546 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4547 | /* note we're sharing with inode.c for the merge bio hook */ |
4548 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4549 | }; |