]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
4b4e25f2 | 28 | #include <linux/freezer.h> |
5a0e3ad6 | 29 | #include <linux/slab.h> |
784b4e29 | 30 | #include <linux/migrate.h> |
7a36ddec | 31 | #include <linux/ratelimit.h> |
6463fe58 | 32 | #include <linux/uuid.h> |
803b2f54 | 33 | #include <linux/semaphore.h> |
7e75bf3f | 34 | #include <asm/unaligned.h> |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
0b947aff | 37 | #include "hash.h" |
e089f05c | 38 | #include "transaction.h" |
0f7d52f4 | 39 | #include "btrfs_inode.h" |
0b86a832 | 40 | #include "volumes.h" |
db94535d | 41 | #include "print-tree.h" |
925baedd | 42 | #include "locking.h" |
e02119d5 | 43 | #include "tree-log.h" |
fa9c0d79 | 44 | #include "free-space-cache.h" |
70f6d82e | 45 | #include "free-space-tree.h" |
581bb050 | 46 | #include "inode-map.h" |
21adbd5c | 47 | #include "check-integrity.h" |
606686ee | 48 | #include "rcu-string.h" |
8dabb742 | 49 | #include "dev-replace.h" |
53b381b3 | 50 | #include "raid56.h" |
5ac1d209 | 51 | #include "sysfs.h" |
fcebe456 | 52 | #include "qgroup.h" |
eb60ceac | 53 | |
de0022b9 JB |
54 | #ifdef CONFIG_X86 |
55 | #include <asm/cpufeature.h> | |
56 | #endif | |
57 | ||
e8c9f186 | 58 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 59 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 60 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 61 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 62 | int read_only); |
143bede5 | 63 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 64 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
65 | struct btrfs_root *root); | |
143bede5 | 66 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
acce952b | 67 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
68 | struct extent_io_tree *dirty_pages, | |
69 | int mark); | |
70 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
71 | struct extent_io_tree *pinned_extents); | |
48a3b636 ES |
72 | static int btrfs_cleanup_transaction(struct btrfs_root *root); |
73 | static void btrfs_error_commit_super(struct btrfs_root *root); | |
ce9adaa5 | 74 | |
d352ac68 | 75 | /* |
97eb6b69 DS |
76 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
77 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
78 | * by writes to insert metadata for new file extents after IO is complete. |
79 | */ | |
97eb6b69 | 80 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
81 | struct bio *bio; |
82 | bio_end_io_t *end_io; | |
83 | void *private; | |
84 | struct btrfs_fs_info *info; | |
85 | int error; | |
bfebd8b5 | 86 | enum btrfs_wq_endio_type metadata; |
ce9adaa5 | 87 | struct list_head list; |
8b712842 | 88 | struct btrfs_work work; |
ce9adaa5 | 89 | }; |
0da5468f | 90 | |
97eb6b69 DS |
91 | static struct kmem_cache *btrfs_end_io_wq_cache; |
92 | ||
93 | int __init btrfs_end_io_wq_init(void) | |
94 | { | |
95 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
96 | sizeof(struct btrfs_end_io_wq), | |
97 | 0, | |
98 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, | |
99 | NULL); | |
100 | if (!btrfs_end_io_wq_cache) | |
101 | return -ENOMEM; | |
102 | return 0; | |
103 | } | |
104 | ||
105 | void btrfs_end_io_wq_exit(void) | |
106 | { | |
107 | if (btrfs_end_io_wq_cache) | |
108 | kmem_cache_destroy(btrfs_end_io_wq_cache); | |
109 | } | |
110 | ||
d352ac68 CM |
111 | /* |
112 | * async submit bios are used to offload expensive checksumming | |
113 | * onto the worker threads. They checksum file and metadata bios | |
114 | * just before they are sent down the IO stack. | |
115 | */ | |
44b8bd7e CM |
116 | struct async_submit_bio { |
117 | struct inode *inode; | |
118 | struct bio *bio; | |
119 | struct list_head list; | |
4a69a410 CM |
120 | extent_submit_bio_hook_t *submit_bio_start; |
121 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
122 | int rw; |
123 | int mirror_num; | |
c8b97818 | 124 | unsigned long bio_flags; |
eaf25d93 CM |
125 | /* |
126 | * bio_offset is optional, can be used if the pages in the bio | |
127 | * can't tell us where in the file the bio should go | |
128 | */ | |
129 | u64 bio_offset; | |
8b712842 | 130 | struct btrfs_work work; |
79787eaa | 131 | int error; |
44b8bd7e CM |
132 | }; |
133 | ||
85d4e461 CM |
134 | /* |
135 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
136 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
137 | * the level the eb occupies in the tree. | |
138 | * | |
139 | * Different roots are used for different purposes and may nest inside each | |
140 | * other and they require separate keysets. As lockdep keys should be | |
141 | * static, assign keysets according to the purpose of the root as indicated | |
142 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
143 | * have separate keysets. | |
4008c04a | 144 | * |
85d4e461 CM |
145 | * Lock-nesting across peer nodes is always done with the immediate parent |
146 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
147 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 148 | * |
85d4e461 CM |
149 | * The key is set by the readpage_end_io_hook after the buffer has passed |
150 | * csum validation but before the pages are unlocked. It is also set by | |
151 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 152 | * |
85d4e461 CM |
153 | * We also add a check to make sure the highest level of the tree is the |
154 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
155 | * needs update as well. | |
4008c04a CM |
156 | */ |
157 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
158 | # if BTRFS_MAX_LEVEL != 8 | |
159 | # error | |
160 | # endif | |
85d4e461 CM |
161 | |
162 | static struct btrfs_lockdep_keyset { | |
163 | u64 id; /* root objectid */ | |
164 | const char *name_stem; /* lock name stem */ | |
165 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
166 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
167 | } btrfs_lockdep_keysets[] = { | |
168 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
169 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
170 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
171 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
172 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
173 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 174 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
175 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
176 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
177 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 178 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
85d4e461 | 179 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 180 | }; |
85d4e461 CM |
181 | |
182 | void __init btrfs_init_lockdep(void) | |
183 | { | |
184 | int i, j; | |
185 | ||
186 | /* initialize lockdep class names */ | |
187 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
188 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
189 | ||
190 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
191 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
192 | "btrfs-%s-%02d", ks->name_stem, j); | |
193 | } | |
194 | } | |
195 | ||
196 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
197 | int level) | |
198 | { | |
199 | struct btrfs_lockdep_keyset *ks; | |
200 | ||
201 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
202 | ||
203 | /* find the matching keyset, id 0 is the default entry */ | |
204 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
205 | if (ks->id == objectid) | |
206 | break; | |
207 | ||
208 | lockdep_set_class_and_name(&eb->lock, | |
209 | &ks->keys[level], ks->names[level]); | |
210 | } | |
211 | ||
4008c04a CM |
212 | #endif |
213 | ||
d352ac68 CM |
214 | /* |
215 | * extents on the btree inode are pretty simple, there's one extent | |
216 | * that covers the entire device | |
217 | */ | |
b2950863 | 218 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 219 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 220 | int create) |
7eccb903 | 221 | { |
5f39d397 CM |
222 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
223 | struct extent_map *em; | |
224 | int ret; | |
225 | ||
890871be | 226 | read_lock(&em_tree->lock); |
d1310b2e | 227 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
228 | if (em) { |
229 | em->bdev = | |
230 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 231 | read_unlock(&em_tree->lock); |
5f39d397 | 232 | goto out; |
a061fc8d | 233 | } |
890871be | 234 | read_unlock(&em_tree->lock); |
7b13b7b1 | 235 | |
172ddd60 | 236 | em = alloc_extent_map(); |
5f39d397 CM |
237 | if (!em) { |
238 | em = ERR_PTR(-ENOMEM); | |
239 | goto out; | |
240 | } | |
241 | em->start = 0; | |
0afbaf8c | 242 | em->len = (u64)-1; |
c8b97818 | 243 | em->block_len = (u64)-1; |
5f39d397 | 244 | em->block_start = 0; |
a061fc8d | 245 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 246 | |
890871be | 247 | write_lock(&em_tree->lock); |
09a2a8f9 | 248 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
249 | if (ret == -EEXIST) { |
250 | free_extent_map(em); | |
7b13b7b1 | 251 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 252 | if (!em) |
0433f20d | 253 | em = ERR_PTR(-EIO); |
5f39d397 | 254 | } else if (ret) { |
7b13b7b1 | 255 | free_extent_map(em); |
0433f20d | 256 | em = ERR_PTR(ret); |
5f39d397 | 257 | } |
890871be | 258 | write_unlock(&em_tree->lock); |
7b13b7b1 | 259 | |
5f39d397 CM |
260 | out: |
261 | return em; | |
7eccb903 CM |
262 | } |
263 | ||
b0496686 | 264 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 265 | { |
0b947aff | 266 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
267 | } |
268 | ||
269 | void btrfs_csum_final(u32 crc, char *result) | |
270 | { | |
7e75bf3f | 271 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
272 | } |
273 | ||
d352ac68 CM |
274 | /* |
275 | * compute the csum for a btree block, and either verify it or write it | |
276 | * into the csum field of the block. | |
277 | */ | |
01d58472 DD |
278 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
279 | struct extent_buffer *buf, | |
19c00ddc CM |
280 | int verify) |
281 | { | |
01d58472 | 282 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
607d432d | 283 | char *result = NULL; |
19c00ddc CM |
284 | unsigned long len; |
285 | unsigned long cur_len; | |
286 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
287 | char *kaddr; |
288 | unsigned long map_start; | |
289 | unsigned long map_len; | |
290 | int err; | |
291 | u32 crc = ~(u32)0; | |
607d432d | 292 | unsigned long inline_result; |
19c00ddc CM |
293 | |
294 | len = buf->len - offset; | |
d397712b | 295 | while (len > 0) { |
19c00ddc | 296 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 297 | &kaddr, &map_start, &map_len); |
d397712b | 298 | if (err) |
19c00ddc | 299 | return 1; |
19c00ddc | 300 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 301 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
302 | crc, cur_len); |
303 | len -= cur_len; | |
304 | offset += cur_len; | |
19c00ddc | 305 | } |
607d432d | 306 | if (csum_size > sizeof(inline_result)) { |
31e818fe | 307 | result = kzalloc(csum_size, GFP_NOFS); |
607d432d JB |
308 | if (!result) |
309 | return 1; | |
310 | } else { | |
311 | result = (char *)&inline_result; | |
312 | } | |
313 | ||
19c00ddc CM |
314 | btrfs_csum_final(crc, result); |
315 | ||
316 | if (verify) { | |
607d432d | 317 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
318 | u32 val; |
319 | u32 found = 0; | |
607d432d | 320 | memcpy(&found, result, csum_size); |
e4204ded | 321 | |
607d432d | 322 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 DS |
323 | btrfs_warn_rl(fs_info, |
324 | "%s checksum verify failed on %llu wanted %X found %X " | |
325 | "level %d", | |
01d58472 | 326 | fs_info->sb->s_id, buf->start, |
efe120a0 | 327 | val, found, btrfs_header_level(buf)); |
607d432d JB |
328 | if (result != (char *)&inline_result) |
329 | kfree(result); | |
19c00ddc CM |
330 | return 1; |
331 | } | |
332 | } else { | |
607d432d | 333 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 334 | } |
607d432d JB |
335 | if (result != (char *)&inline_result) |
336 | kfree(result); | |
19c00ddc CM |
337 | return 0; |
338 | } | |
339 | ||
d352ac68 CM |
340 | /* |
341 | * we can't consider a given block up to date unless the transid of the | |
342 | * block matches the transid in the parent node's pointer. This is how we | |
343 | * detect blocks that either didn't get written at all or got written | |
344 | * in the wrong place. | |
345 | */ | |
1259ab75 | 346 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
347 | struct extent_buffer *eb, u64 parent_transid, |
348 | int atomic) | |
1259ab75 | 349 | { |
2ac55d41 | 350 | struct extent_state *cached_state = NULL; |
1259ab75 | 351 | int ret; |
2755a0de | 352 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
353 | |
354 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
355 | return 0; | |
356 | ||
b9fab919 CM |
357 | if (atomic) |
358 | return -EAGAIN; | |
359 | ||
a26e8c9f JB |
360 | if (need_lock) { |
361 | btrfs_tree_read_lock(eb); | |
362 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
363 | } | |
364 | ||
2ac55d41 | 365 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 366 | &cached_state); |
0b32f4bb | 367 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
368 | btrfs_header_generation(eb) == parent_transid) { |
369 | ret = 0; | |
370 | goto out; | |
371 | } | |
94647322 DS |
372 | btrfs_err_rl(eb->fs_info, |
373 | "parent transid verify failed on %llu wanted %llu found %llu", | |
374 | eb->start, | |
29549aec | 375 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 376 | ret = 1; |
a26e8c9f JB |
377 | |
378 | /* | |
379 | * Things reading via commit roots that don't have normal protection, | |
380 | * like send, can have a really old block in cache that may point at a | |
381 | * block that has been free'd and re-allocated. So don't clear uptodate | |
382 | * if we find an eb that is under IO (dirty/writeback) because we could | |
383 | * end up reading in the stale data and then writing it back out and | |
384 | * making everybody very sad. | |
385 | */ | |
386 | if (!extent_buffer_under_io(eb)) | |
387 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 388 | out: |
2ac55d41 JB |
389 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
390 | &cached_state, GFP_NOFS); | |
472b909f JB |
391 | if (need_lock) |
392 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 393 | return ret; |
1259ab75 CM |
394 | } |
395 | ||
1104a885 DS |
396 | /* |
397 | * Return 0 if the superblock checksum type matches the checksum value of that | |
398 | * algorithm. Pass the raw disk superblock data. | |
399 | */ | |
400 | static int btrfs_check_super_csum(char *raw_disk_sb) | |
401 | { | |
402 | struct btrfs_super_block *disk_sb = | |
403 | (struct btrfs_super_block *)raw_disk_sb; | |
404 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
405 | int ret = 0; | |
406 | ||
407 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
408 | u32 crc = ~(u32)0; | |
409 | const int csum_size = sizeof(crc); | |
410 | char result[csum_size]; | |
411 | ||
412 | /* | |
413 | * The super_block structure does not span the whole | |
414 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
415 | * is filled with zeros and is included in the checkum. | |
416 | */ | |
417 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
418 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
419 | btrfs_csum_final(crc, result); | |
420 | ||
421 | if (memcmp(raw_disk_sb, result, csum_size)) | |
422 | ret = 1; | |
423 | } | |
424 | ||
425 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
efe120a0 | 426 | printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", |
1104a885 DS |
427 | csum_type); |
428 | ret = 1; | |
429 | } | |
430 | ||
431 | return ret; | |
432 | } | |
433 | ||
d352ac68 CM |
434 | /* |
435 | * helper to read a given tree block, doing retries as required when | |
436 | * the checksums don't match and we have alternate mirrors to try. | |
437 | */ | |
f188591e CM |
438 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
439 | struct extent_buffer *eb, | |
ca7a79ad | 440 | u64 start, u64 parent_transid) |
f188591e CM |
441 | { |
442 | struct extent_io_tree *io_tree; | |
ea466794 | 443 | int failed = 0; |
f188591e CM |
444 | int ret; |
445 | int num_copies = 0; | |
446 | int mirror_num = 0; | |
ea466794 | 447 | int failed_mirror = 0; |
f188591e | 448 | |
a826d6dc | 449 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
450 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
451 | while (1) { | |
bb82ab88 AJ |
452 | ret = read_extent_buffer_pages(io_tree, eb, start, |
453 | WAIT_COMPLETE, | |
f188591e | 454 | btree_get_extent, mirror_num); |
256dd1bb SB |
455 | if (!ret) { |
456 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 457 | parent_transid, 0)) |
256dd1bb SB |
458 | break; |
459 | else | |
460 | ret = -EIO; | |
461 | } | |
d397712b | 462 | |
a826d6dc JB |
463 | /* |
464 | * This buffer's crc is fine, but its contents are corrupted, so | |
465 | * there is no reason to read the other copies, they won't be | |
466 | * any less wrong. | |
467 | */ | |
468 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
469 | break; |
470 | ||
5d964051 | 471 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 472 | eb->start, eb->len); |
4235298e | 473 | if (num_copies == 1) |
ea466794 | 474 | break; |
4235298e | 475 | |
5cf1ab56 JB |
476 | if (!failed_mirror) { |
477 | failed = 1; | |
478 | failed_mirror = eb->read_mirror; | |
479 | } | |
480 | ||
f188591e | 481 | mirror_num++; |
ea466794 JB |
482 | if (mirror_num == failed_mirror) |
483 | mirror_num++; | |
484 | ||
4235298e | 485 | if (mirror_num > num_copies) |
ea466794 | 486 | break; |
f188591e | 487 | } |
ea466794 | 488 | |
c0901581 | 489 | if (failed && !ret && failed_mirror) |
ea466794 JB |
490 | repair_eb_io_failure(root, eb, failed_mirror); |
491 | ||
492 | return ret; | |
f188591e | 493 | } |
19c00ddc | 494 | |
d352ac68 | 495 | /* |
d397712b CM |
496 | * checksum a dirty tree block before IO. This has extra checks to make sure |
497 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 498 | */ |
d397712b | 499 | |
01d58472 | 500 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 501 | { |
4eee4fa4 | 502 | u64 start = page_offset(page); |
19c00ddc | 503 | u64 found_start; |
19c00ddc | 504 | struct extent_buffer *eb; |
f188591e | 505 | |
4f2de97a JB |
506 | eb = (struct extent_buffer *)page->private; |
507 | if (page != eb->pages[0]) | |
508 | return 0; | |
19c00ddc | 509 | found_start = btrfs_header_bytenr(eb); |
fae7f21c | 510 | if (WARN_ON(found_start != start || !PageUptodate(page))) |
4f2de97a | 511 | return 0; |
01d58472 | 512 | csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
513 | return 0; |
514 | } | |
515 | ||
01d58472 | 516 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
517 | struct extent_buffer *eb) |
518 | { | |
01d58472 | 519 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
2b82032c YZ |
520 | u8 fsid[BTRFS_UUID_SIZE]; |
521 | int ret = 1; | |
522 | ||
0a4e5586 | 523 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
524 | while (fs_devices) { |
525 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
526 | ret = 0; | |
527 | break; | |
528 | } | |
529 | fs_devices = fs_devices->seed; | |
530 | } | |
531 | return ret; | |
532 | } | |
533 | ||
a826d6dc | 534 | #define CORRUPT(reason, eb, root, slot) \ |
efe120a0 FH |
535 | btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ |
536 | "root=%llu, slot=%d", reason, \ | |
c1c9ff7c | 537 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
538 | |
539 | static noinline int check_leaf(struct btrfs_root *root, | |
540 | struct extent_buffer *leaf) | |
541 | { | |
542 | struct btrfs_key key; | |
543 | struct btrfs_key leaf_key; | |
544 | u32 nritems = btrfs_header_nritems(leaf); | |
545 | int slot; | |
546 | ||
547 | if (nritems == 0) | |
548 | return 0; | |
549 | ||
550 | /* Check the 0 item */ | |
551 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
552 | BTRFS_LEAF_DATA_SIZE(root)) { | |
553 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
554 | return -EIO; | |
555 | } | |
556 | ||
557 | /* | |
558 | * Check to make sure each items keys are in the correct order and their | |
559 | * offsets make sense. We only have to loop through nritems-1 because | |
560 | * we check the current slot against the next slot, which verifies the | |
561 | * next slot's offset+size makes sense and that the current's slot | |
562 | * offset is correct. | |
563 | */ | |
564 | for (slot = 0; slot < nritems - 1; slot++) { | |
565 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
566 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
567 | ||
568 | /* Make sure the keys are in the right order */ | |
569 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
570 | CORRUPT("bad key order", leaf, root, slot); | |
571 | return -EIO; | |
572 | } | |
573 | ||
574 | /* | |
575 | * Make sure the offset and ends are right, remember that the | |
576 | * item data starts at the end of the leaf and grows towards the | |
577 | * front. | |
578 | */ | |
579 | if (btrfs_item_offset_nr(leaf, slot) != | |
580 | btrfs_item_end_nr(leaf, slot + 1)) { | |
581 | CORRUPT("slot offset bad", leaf, root, slot); | |
582 | return -EIO; | |
583 | } | |
584 | ||
585 | /* | |
586 | * Check to make sure that we don't point outside of the leaf, | |
587 | * just incase all the items are consistent to eachother, but | |
588 | * all point outside of the leaf. | |
589 | */ | |
590 | if (btrfs_item_end_nr(leaf, slot) > | |
591 | BTRFS_LEAF_DATA_SIZE(root)) { | |
592 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
593 | return -EIO; | |
594 | } | |
595 | } | |
596 | ||
597 | return 0; | |
598 | } | |
599 | ||
facc8a22 MX |
600 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
601 | u64 phy_offset, struct page *page, | |
602 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 603 | { |
ce9adaa5 CM |
604 | u64 found_start; |
605 | int found_level; | |
ce9adaa5 CM |
606 | struct extent_buffer *eb; |
607 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 608 | int ret = 0; |
727011e0 | 609 | int reads_done; |
ce9adaa5 | 610 | |
ce9adaa5 CM |
611 | if (!page->private) |
612 | goto out; | |
d397712b | 613 | |
4f2de97a | 614 | eb = (struct extent_buffer *)page->private; |
d397712b | 615 | |
0b32f4bb JB |
616 | /* the pending IO might have been the only thing that kept this buffer |
617 | * in memory. Make sure we have a ref for all this other checks | |
618 | */ | |
619 | extent_buffer_get(eb); | |
620 | ||
621 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
622 | if (!reads_done) |
623 | goto err; | |
f188591e | 624 | |
5cf1ab56 | 625 | eb->read_mirror = mirror; |
656f30db | 626 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
627 | ret = -EIO; |
628 | goto err; | |
629 | } | |
630 | ||
ce9adaa5 | 631 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 632 | if (found_start != eb->start) { |
94647322 DS |
633 | btrfs_err_rl(eb->fs_info, "bad tree block start %llu %llu", |
634 | found_start, eb->start); | |
f188591e | 635 | ret = -EIO; |
ce9adaa5 CM |
636 | goto err; |
637 | } | |
01d58472 | 638 | if (check_tree_block_fsid(root->fs_info, eb)) { |
94647322 DS |
639 | btrfs_err_rl(eb->fs_info, "bad fsid on block %llu", |
640 | eb->start); | |
1259ab75 CM |
641 | ret = -EIO; |
642 | goto err; | |
643 | } | |
ce9adaa5 | 644 | found_level = btrfs_header_level(eb); |
1c24c3ce | 645 | if (found_level >= BTRFS_MAX_LEVEL) { |
68b663d1 | 646 | btrfs_err(root->fs_info, "bad tree block level %d", |
1c24c3ce JB |
647 | (int)btrfs_header_level(eb)); |
648 | ret = -EIO; | |
649 | goto err; | |
650 | } | |
ce9adaa5 | 651 | |
85d4e461 CM |
652 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
653 | eb, found_level); | |
4008c04a | 654 | |
01d58472 | 655 | ret = csum_tree_block(root->fs_info, eb, 1); |
a826d6dc | 656 | if (ret) { |
f188591e | 657 | ret = -EIO; |
a826d6dc JB |
658 | goto err; |
659 | } | |
660 | ||
661 | /* | |
662 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
663 | * that we don't try and read the other copies of this block, just | |
664 | * return -EIO. | |
665 | */ | |
666 | if (found_level == 0 && check_leaf(root, eb)) { | |
667 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
668 | ret = -EIO; | |
669 | } | |
ce9adaa5 | 670 | |
0b32f4bb JB |
671 | if (!ret) |
672 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 673 | err: |
79fb65a1 JB |
674 | if (reads_done && |
675 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
4bb31e92 | 676 | btree_readahead_hook(root, eb, eb->start, ret); |
4bb31e92 | 677 | |
53b381b3 DW |
678 | if (ret) { |
679 | /* | |
680 | * our io error hook is going to dec the io pages | |
681 | * again, we have to make sure it has something | |
682 | * to decrement | |
683 | */ | |
684 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 685 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 686 | } |
0b32f4bb | 687 | free_extent_buffer(eb); |
ce9adaa5 | 688 | out: |
f188591e | 689 | return ret; |
ce9adaa5 CM |
690 | } |
691 | ||
ea466794 | 692 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 693 | { |
4bb31e92 AJ |
694 | struct extent_buffer *eb; |
695 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
696 | ||
4f2de97a | 697 | eb = (struct extent_buffer *)page->private; |
656f30db | 698 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 699 | eb->read_mirror = failed_mirror; |
53b381b3 | 700 | atomic_dec(&eb->io_pages); |
ea466794 | 701 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 702 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
703 | return -EIO; /* we fixed nothing */ |
704 | } | |
705 | ||
4246a0b6 | 706 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 707 | { |
97eb6b69 | 708 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 709 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
710 | struct btrfs_workqueue *wq; |
711 | btrfs_work_func_t func; | |
ce9adaa5 | 712 | |
ce9adaa5 | 713 | fs_info = end_io_wq->info; |
4246a0b6 | 714 | end_io_wq->error = bio->bi_error; |
d20f7043 | 715 | |
7b6d91da | 716 | if (bio->bi_rw & REQ_WRITE) { |
9e0af237 LB |
717 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
718 | wq = fs_info->endio_meta_write_workers; | |
719 | func = btrfs_endio_meta_write_helper; | |
720 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
721 | wq = fs_info->endio_freespace_worker; | |
722 | func = btrfs_freespace_write_helper; | |
723 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
724 | wq = fs_info->endio_raid56_workers; | |
725 | func = btrfs_endio_raid56_helper; | |
726 | } else { | |
727 | wq = fs_info->endio_write_workers; | |
728 | func = btrfs_endio_write_helper; | |
729 | } | |
d20f7043 | 730 | } else { |
8b110e39 MX |
731 | if (unlikely(end_io_wq->metadata == |
732 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
733 | wq = fs_info->endio_repair_workers; | |
734 | func = btrfs_endio_repair_helper; | |
735 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
736 | wq = fs_info->endio_raid56_workers; |
737 | func = btrfs_endio_raid56_helper; | |
738 | } else if (end_io_wq->metadata) { | |
739 | wq = fs_info->endio_meta_workers; | |
740 | func = btrfs_endio_meta_helper; | |
741 | } else { | |
742 | wq = fs_info->endio_workers; | |
743 | func = btrfs_endio_helper; | |
744 | } | |
d20f7043 | 745 | } |
9e0af237 LB |
746 | |
747 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
748 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
749 | } |
750 | ||
22c59948 | 751 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 752 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 753 | { |
97eb6b69 | 754 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 755 | |
97eb6b69 | 756 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 CM |
757 | if (!end_io_wq) |
758 | return -ENOMEM; | |
759 | ||
760 | end_io_wq->private = bio->bi_private; | |
761 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 762 | end_io_wq->info = info; |
ce9adaa5 CM |
763 | end_io_wq->error = 0; |
764 | end_io_wq->bio = bio; | |
22c59948 | 765 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
766 | |
767 | bio->bi_private = end_io_wq; | |
768 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
769 | return 0; |
770 | } | |
771 | ||
b64a2851 | 772 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 773 | { |
4854ddd0 | 774 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 775 | info->thread_pool_size, |
4854ddd0 CM |
776 | info->fs_devices->open_devices); |
777 | return 256 * limit; | |
778 | } | |
0986fe9e | 779 | |
4a69a410 CM |
780 | static void run_one_async_start(struct btrfs_work *work) |
781 | { | |
4a69a410 | 782 | struct async_submit_bio *async; |
79787eaa | 783 | int ret; |
4a69a410 CM |
784 | |
785 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
786 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
787 | async->mirror_num, async->bio_flags, | |
788 | async->bio_offset); | |
789 | if (ret) | |
790 | async->error = ret; | |
4a69a410 CM |
791 | } |
792 | ||
793 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
794 | { |
795 | struct btrfs_fs_info *fs_info; | |
796 | struct async_submit_bio *async; | |
4854ddd0 | 797 | int limit; |
8b712842 CM |
798 | |
799 | async = container_of(work, struct async_submit_bio, work); | |
800 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 801 | |
b64a2851 | 802 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
803 | limit = limit * 2 / 3; |
804 | ||
ee863954 DS |
805 | /* |
806 | * atomic_dec_return implies a barrier for waitqueue_active | |
807 | */ | |
66657b31 | 808 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 809 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
810 | wake_up(&fs_info->async_submit_wait); |
811 | ||
79787eaa JM |
812 | /* If an error occured we just want to clean up the bio and move on */ |
813 | if (async->error) { | |
4246a0b6 CH |
814 | async->bio->bi_error = async->error; |
815 | bio_endio(async->bio); | |
79787eaa JM |
816 | return; |
817 | } | |
818 | ||
4a69a410 | 819 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
820 | async->mirror_num, async->bio_flags, |
821 | async->bio_offset); | |
4a69a410 CM |
822 | } |
823 | ||
824 | static void run_one_async_free(struct btrfs_work *work) | |
825 | { | |
826 | struct async_submit_bio *async; | |
827 | ||
828 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
829 | kfree(async); |
830 | } | |
831 | ||
44b8bd7e CM |
832 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
833 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 834 | unsigned long bio_flags, |
eaf25d93 | 835 | u64 bio_offset, |
4a69a410 CM |
836 | extent_submit_bio_hook_t *submit_bio_start, |
837 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
838 | { |
839 | struct async_submit_bio *async; | |
840 | ||
841 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
842 | if (!async) | |
843 | return -ENOMEM; | |
844 | ||
845 | async->inode = inode; | |
846 | async->rw = rw; | |
847 | async->bio = bio; | |
848 | async->mirror_num = mirror_num; | |
4a69a410 CM |
849 | async->submit_bio_start = submit_bio_start; |
850 | async->submit_bio_done = submit_bio_done; | |
851 | ||
9e0af237 | 852 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 853 | run_one_async_done, run_one_async_free); |
4a69a410 | 854 | |
c8b97818 | 855 | async->bio_flags = bio_flags; |
eaf25d93 | 856 | async->bio_offset = bio_offset; |
8c8bee1d | 857 | |
79787eaa JM |
858 | async->error = 0; |
859 | ||
cb03c743 | 860 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 861 | |
7b6d91da | 862 | if (rw & REQ_SYNC) |
5cdc7ad3 | 863 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 864 | |
5cdc7ad3 | 865 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 866 | |
d397712b | 867 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
868 | atomic_read(&fs_info->nr_async_submits)) { |
869 | wait_event(fs_info->async_submit_wait, | |
870 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
871 | } | |
872 | ||
44b8bd7e CM |
873 | return 0; |
874 | } | |
875 | ||
ce3ed71a CM |
876 | static int btree_csum_one_bio(struct bio *bio) |
877 | { | |
2c30c71b | 878 | struct bio_vec *bvec; |
ce3ed71a | 879 | struct btrfs_root *root; |
2c30c71b | 880 | int i, ret = 0; |
ce3ed71a | 881 | |
2c30c71b | 882 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 883 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 884 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
885 | if (ret) |
886 | break; | |
ce3ed71a | 887 | } |
2c30c71b | 888 | |
79787eaa | 889 | return ret; |
ce3ed71a CM |
890 | } |
891 | ||
4a69a410 CM |
892 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
893 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
894 | unsigned long bio_flags, |
895 | u64 bio_offset) | |
22c59948 | 896 | { |
8b712842 CM |
897 | /* |
898 | * when we're called for a write, we're already in the async | |
5443be45 | 899 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 900 | */ |
79787eaa | 901 | return btree_csum_one_bio(bio); |
4a69a410 | 902 | } |
22c59948 | 903 | |
4a69a410 | 904 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
905 | int mirror_num, unsigned long bio_flags, |
906 | u64 bio_offset) | |
4a69a410 | 907 | { |
61891923 SB |
908 | int ret; |
909 | ||
8b712842 | 910 | /* |
4a69a410 CM |
911 | * when we're called for a write, we're already in the async |
912 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 913 | */ |
61891923 | 914 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
4246a0b6 CH |
915 | if (ret) { |
916 | bio->bi_error = ret; | |
917 | bio_endio(bio); | |
918 | } | |
61891923 | 919 | return ret; |
0b86a832 CM |
920 | } |
921 | ||
de0022b9 JB |
922 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
923 | { | |
924 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
925 | return 0; | |
926 | #ifdef CONFIG_X86 | |
927 | if (cpu_has_xmm4_2) | |
928 | return 0; | |
929 | #endif | |
930 | return 1; | |
931 | } | |
932 | ||
44b8bd7e | 933 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
934 | int mirror_num, unsigned long bio_flags, |
935 | u64 bio_offset) | |
44b8bd7e | 936 | { |
de0022b9 | 937 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
938 | int ret; |
939 | ||
7b6d91da | 940 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
941 | /* |
942 | * called for a read, do the setup so that checksum validation | |
943 | * can happen in the async kernel threads | |
944 | */ | |
f3f266ab | 945 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
bfebd8b5 | 946 | bio, BTRFS_WQ_ENDIO_METADATA); |
1d4284bd | 947 | if (ret) |
61891923 SB |
948 | goto out_w_error; |
949 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
950 | mirror_num, 0); | |
de0022b9 JB |
951 | } else if (!async) { |
952 | ret = btree_csum_one_bio(bio); | |
953 | if (ret) | |
61891923 SB |
954 | goto out_w_error; |
955 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
956 | mirror_num, 0); | |
957 | } else { | |
958 | /* | |
959 | * kthread helpers are used to submit writes so that | |
960 | * checksumming can happen in parallel across all CPUs | |
961 | */ | |
962 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
963 | inode, rw, bio, mirror_num, 0, | |
964 | bio_offset, | |
965 | __btree_submit_bio_start, | |
966 | __btree_submit_bio_done); | |
44b8bd7e | 967 | } |
d313d7a3 | 968 | |
4246a0b6 CH |
969 | if (ret) |
970 | goto out_w_error; | |
971 | return 0; | |
972 | ||
61891923 | 973 | out_w_error: |
4246a0b6 CH |
974 | bio->bi_error = ret; |
975 | bio_endio(bio); | |
61891923 | 976 | return ret; |
44b8bd7e CM |
977 | } |
978 | ||
3dd1462e | 979 | #ifdef CONFIG_MIGRATION |
784b4e29 | 980 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
981 | struct page *newpage, struct page *page, |
982 | enum migrate_mode mode) | |
784b4e29 CM |
983 | { |
984 | /* | |
985 | * we can't safely write a btree page from here, | |
986 | * we haven't done the locking hook | |
987 | */ | |
988 | if (PageDirty(page)) | |
989 | return -EAGAIN; | |
990 | /* | |
991 | * Buffers may be managed in a filesystem specific way. | |
992 | * We must have no buffers or drop them. | |
993 | */ | |
994 | if (page_has_private(page) && | |
995 | !try_to_release_page(page, GFP_KERNEL)) | |
996 | return -EAGAIN; | |
a6bc32b8 | 997 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 998 | } |
3dd1462e | 999 | #endif |
784b4e29 | 1000 | |
0da5468f CM |
1001 | |
1002 | static int btree_writepages(struct address_space *mapping, | |
1003 | struct writeback_control *wbc) | |
1004 | { | |
e2d84521 MX |
1005 | struct btrfs_fs_info *fs_info; |
1006 | int ret; | |
1007 | ||
d8d5f3e1 | 1008 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
1009 | |
1010 | if (wbc->for_kupdate) | |
1011 | return 0; | |
1012 | ||
e2d84521 | 1013 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 1014 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
1015 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
1016 | BTRFS_DIRTY_METADATA_THRESH); | |
1017 | if (ret < 0) | |
793955bc | 1018 | return 0; |
793955bc | 1019 | } |
0b32f4bb | 1020 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
1021 | } |
1022 | ||
b2950863 | 1023 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 1024 | { |
d1310b2e CM |
1025 | struct extent_io_tree *tree; |
1026 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 1027 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 1028 | } |
22b0ebda | 1029 | |
70dec807 | 1030 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 1031 | { |
98509cfc | 1032 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 1033 | return 0; |
0c4e538b | 1034 | |
f7a52a40 | 1035 | return try_release_extent_buffer(page); |
d98237b3 CM |
1036 | } |
1037 | ||
d47992f8 LC |
1038 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
1039 | unsigned int length) | |
d98237b3 | 1040 | { |
d1310b2e CM |
1041 | struct extent_io_tree *tree; |
1042 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1043 | extent_invalidatepage(tree, page, offset); |
1044 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1045 | if (PagePrivate(page)) { |
efe120a0 FH |
1046 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1047 | "page private not zero on page %llu", | |
1048 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1049 | ClearPagePrivate(page); |
1050 | set_page_private(page, 0); | |
1051 | page_cache_release(page); | |
1052 | } | |
d98237b3 CM |
1053 | } |
1054 | ||
0b32f4bb JB |
1055 | static int btree_set_page_dirty(struct page *page) |
1056 | { | |
bb146eb2 | 1057 | #ifdef DEBUG |
0b32f4bb JB |
1058 | struct extent_buffer *eb; |
1059 | ||
1060 | BUG_ON(!PagePrivate(page)); | |
1061 | eb = (struct extent_buffer *)page->private; | |
1062 | BUG_ON(!eb); | |
1063 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1064 | BUG_ON(!atomic_read(&eb->refs)); | |
1065 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1066 | #endif |
0b32f4bb JB |
1067 | return __set_page_dirty_nobuffers(page); |
1068 | } | |
1069 | ||
7f09410b | 1070 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1071 | .readpage = btree_readpage, |
0da5468f | 1072 | .writepages = btree_writepages, |
5f39d397 CM |
1073 | .releasepage = btree_releasepage, |
1074 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1075 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1076 | .migratepage = btree_migratepage, |
5a92bc88 | 1077 | #endif |
0b32f4bb | 1078 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1079 | }; |
1080 | ||
d3e46fea | 1081 | void readahead_tree_block(struct btrfs_root *root, u64 bytenr) |
090d1875 | 1082 | { |
5f39d397 CM |
1083 | struct extent_buffer *buf = NULL; |
1084 | struct inode *btree_inode = root->fs_info->btree_inode; | |
090d1875 | 1085 | |
a83fffb7 | 1086 | buf = btrfs_find_create_tree_block(root, bytenr); |
5f39d397 | 1087 | if (!buf) |
6197d86e | 1088 | return; |
d1310b2e | 1089 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1090 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1091 | free_extent_buffer(buf); |
090d1875 CM |
1092 | } |
1093 | ||
c0dcaa4d | 1094 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, |
ab0fff03 AJ |
1095 | int mirror_num, struct extent_buffer **eb) |
1096 | { | |
1097 | struct extent_buffer *buf = NULL; | |
1098 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1099 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1100 | int ret; | |
1101 | ||
a83fffb7 | 1102 | buf = btrfs_find_create_tree_block(root, bytenr); |
ab0fff03 AJ |
1103 | if (!buf) |
1104 | return 0; | |
1105 | ||
1106 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1107 | ||
1108 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1109 | btree_get_extent, mirror_num); | |
1110 | if (ret) { | |
1111 | free_extent_buffer(buf); | |
1112 | return ret; | |
1113 | } | |
1114 | ||
1115 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1116 | free_extent_buffer(buf); | |
1117 | return -EIO; | |
0b32f4bb | 1118 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1119 | *eb = buf; |
1120 | } else { | |
1121 | free_extent_buffer(buf); | |
1122 | } | |
1123 | return 0; | |
1124 | } | |
1125 | ||
01d58472 | 1126 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, |
0308af44 | 1127 | u64 bytenr) |
0999df54 | 1128 | { |
01d58472 | 1129 | return find_extent_buffer(fs_info, bytenr); |
0999df54 CM |
1130 | } |
1131 | ||
1132 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
a83fffb7 | 1133 | u64 bytenr) |
0999df54 | 1134 | { |
fccb84c9 | 1135 | if (btrfs_test_is_dummy_root(root)) |
ce3e6984 DS |
1136 | return alloc_test_extent_buffer(root->fs_info, bytenr); |
1137 | return alloc_extent_buffer(root->fs_info, bytenr); | |
0999df54 CM |
1138 | } |
1139 | ||
1140 | ||
e02119d5 CM |
1141 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1142 | { | |
727011e0 | 1143 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1144 | buf->start + buf->len - 1); |
e02119d5 CM |
1145 | } |
1146 | ||
1147 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1148 | { | |
727011e0 | 1149 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1150 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1151 | } |
1152 | ||
0999df54 | 1153 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ce86cd59 | 1154 | u64 parent_transid) |
0999df54 CM |
1155 | { |
1156 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1157 | int ret; |
1158 | ||
a83fffb7 | 1159 | buf = btrfs_find_create_tree_block(root, bytenr); |
0999df54 | 1160 | if (!buf) |
64c043de | 1161 | return ERR_PTR(-ENOMEM); |
0999df54 | 1162 | |
ca7a79ad | 1163 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1164 | if (ret) { |
1165 | free_extent_buffer(buf); | |
64c043de | 1166 | return ERR_PTR(ret); |
0f0fe8f7 | 1167 | } |
5f39d397 | 1168 | return buf; |
ce9adaa5 | 1169 | |
eb60ceac CM |
1170 | } |
1171 | ||
01d58472 DD |
1172 | void clean_tree_block(struct btrfs_trans_handle *trans, |
1173 | struct btrfs_fs_info *fs_info, | |
d5c13f92 | 1174 | struct extent_buffer *buf) |
ed2ff2cb | 1175 | { |
55c69072 | 1176 | if (btrfs_header_generation(buf) == |
e2d84521 | 1177 | fs_info->running_transaction->transid) { |
b9447ef8 | 1178 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1179 | |
b9473439 | 1180 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1181 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1182 | -buf->len, | |
1183 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1184 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1185 | btrfs_set_lock_blocking(buf); | |
1186 | clear_extent_buffer_dirty(buf); | |
1187 | } | |
925baedd | 1188 | } |
5f39d397 CM |
1189 | } |
1190 | ||
8257b2dc MX |
1191 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1192 | { | |
1193 | struct btrfs_subvolume_writers *writers; | |
1194 | int ret; | |
1195 | ||
1196 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1197 | if (!writers) | |
1198 | return ERR_PTR(-ENOMEM); | |
1199 | ||
908c7f19 | 1200 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1201 | if (ret < 0) { |
1202 | kfree(writers); | |
1203 | return ERR_PTR(ret); | |
1204 | } | |
1205 | ||
1206 | init_waitqueue_head(&writers->wait); | |
1207 | return writers; | |
1208 | } | |
1209 | ||
1210 | static void | |
1211 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1212 | { | |
1213 | percpu_counter_destroy(&writers->counter); | |
1214 | kfree(writers); | |
1215 | } | |
1216 | ||
707e8a07 DS |
1217 | static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, |
1218 | struct btrfs_root *root, struct btrfs_fs_info *fs_info, | |
143bede5 | 1219 | u64 objectid) |
d97e63b6 | 1220 | { |
cfaa7295 | 1221 | root->node = NULL; |
a28ec197 | 1222 | root->commit_root = NULL; |
db94535d CM |
1223 | root->sectorsize = sectorsize; |
1224 | root->nodesize = nodesize; | |
87ee04eb | 1225 | root->stripesize = stripesize; |
27cdeb70 | 1226 | root->state = 0; |
d68fc57b | 1227 | root->orphan_cleanup_state = 0; |
0b86a832 | 1228 | |
0f7d52f4 CM |
1229 | root->objectid = objectid; |
1230 | root->last_trans = 0; | |
13a8a7c8 | 1231 | root->highest_objectid = 0; |
eb73c1b7 | 1232 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1233 | root->nr_ordered_extents = 0; |
58176a96 | 1234 | root->name = NULL; |
6bef4d31 | 1235 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1236 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1237 | root->block_rsv = NULL; |
d68fc57b | 1238 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1239 | |
1240 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1241 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1242 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1243 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1244 | INIT_LIST_HEAD(&root->ordered_extents); |
1245 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1246 | INIT_LIST_HEAD(&root->logged_list[0]); |
1247 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1248 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1249 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1250 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1251 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1252 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1253 | spin_lock_init(&root->log_extents_lock[0]); |
1254 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1255 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1256 | mutex_init(&root->log_mutex); |
31f3d255 | 1257 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1258 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1259 | init_waitqueue_head(&root->log_writer_wait); |
1260 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1261 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1262 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1263 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1264 | atomic_set(&root->log_commit[0], 0); |
1265 | atomic_set(&root->log_commit[1], 0); | |
1266 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1267 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1268 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1269 | atomic_set(&root->refs, 1); |
8257b2dc | 1270 | atomic_set(&root->will_be_snapshoted, 0); |
55eeaf05 | 1271 | atomic_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1272 | root->log_transid = 0; |
d1433deb | 1273 | root->log_transid_committed = -1; |
257c62e1 | 1274 | root->last_log_commit = 0; |
06ea65a3 JB |
1275 | if (fs_info) |
1276 | extent_io_tree_init(&root->dirty_log_pages, | |
1277 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1278 | |
3768f368 CM |
1279 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1280 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1281 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
06ea65a3 JB |
1282 | if (fs_info) |
1283 | root->defrag_trans_start = fs_info->generation; | |
1284 | else | |
1285 | root->defrag_trans_start = 0; | |
4d775673 | 1286 | root->root_key.objectid = objectid; |
0ee5dc67 | 1287 | root->anon_dev = 0; |
8ea05e3a | 1288 | |
5f3ab90a | 1289 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1290 | } |
1291 | ||
f84a8bd6 | 1292 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1293 | { |
1294 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1295 | if (root) | |
1296 | root->fs_info = fs_info; | |
1297 | return root; | |
1298 | } | |
1299 | ||
06ea65a3 JB |
1300 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1301 | /* Should only be used by the testing infrastructure */ | |
1302 | struct btrfs_root *btrfs_alloc_dummy_root(void) | |
1303 | { | |
1304 | struct btrfs_root *root; | |
1305 | ||
1306 | root = btrfs_alloc_root(NULL); | |
1307 | if (!root) | |
1308 | return ERR_PTR(-ENOMEM); | |
707e8a07 | 1309 | __setup_root(4096, 4096, 4096, root, NULL, 1); |
27cdeb70 | 1310 | set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); |
faa2dbf0 | 1311 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1312 | |
1313 | return root; | |
1314 | } | |
1315 | #endif | |
1316 | ||
20897f5c AJ |
1317 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1318 | struct btrfs_fs_info *fs_info, | |
1319 | u64 objectid) | |
1320 | { | |
1321 | struct extent_buffer *leaf; | |
1322 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1323 | struct btrfs_root *root; | |
1324 | struct btrfs_key key; | |
1325 | int ret = 0; | |
6463fe58 | 1326 | uuid_le uuid; |
20897f5c AJ |
1327 | |
1328 | root = btrfs_alloc_root(fs_info); | |
1329 | if (!root) | |
1330 | return ERR_PTR(-ENOMEM); | |
1331 | ||
707e8a07 DS |
1332 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1333 | tree_root->stripesize, root, fs_info, objectid); | |
20897f5c AJ |
1334 | root->root_key.objectid = objectid; |
1335 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1336 | root->root_key.offset = 0; | |
1337 | ||
4d75f8a9 | 1338 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1339 | if (IS_ERR(leaf)) { |
1340 | ret = PTR_ERR(leaf); | |
1dd05682 | 1341 | leaf = NULL; |
20897f5c AJ |
1342 | goto fail; |
1343 | } | |
1344 | ||
20897f5c AJ |
1345 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1346 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1347 | btrfs_set_header_generation(leaf, trans->transid); | |
1348 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1349 | btrfs_set_header_owner(leaf, objectid); | |
1350 | root->node = leaf; | |
1351 | ||
0a4e5586 | 1352 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1353 | BTRFS_FSID_SIZE); |
1354 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1355 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1356 | BTRFS_UUID_SIZE); |
1357 | btrfs_mark_buffer_dirty(leaf); | |
1358 | ||
1359 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1360 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1361 | |
1362 | root->root_item.flags = 0; | |
1363 | root->root_item.byte_limit = 0; | |
1364 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1365 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1366 | btrfs_set_root_level(&root->root_item, 0); | |
1367 | btrfs_set_root_refs(&root->root_item, 1); | |
1368 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1369 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1370 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1371 | uuid_le_gen(&uuid); |
1372 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1373 | root->root_item.drop_level = 0; |
1374 | ||
1375 | key.objectid = objectid; | |
1376 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1377 | key.offset = 0; | |
1378 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1379 | if (ret) | |
1380 | goto fail; | |
1381 | ||
1382 | btrfs_tree_unlock(leaf); | |
1383 | ||
1dd05682 TI |
1384 | return root; |
1385 | ||
20897f5c | 1386 | fail: |
1dd05682 TI |
1387 | if (leaf) { |
1388 | btrfs_tree_unlock(leaf); | |
59885b39 | 1389 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1390 | free_extent_buffer(leaf); |
1391 | } | |
1392 | kfree(root); | |
20897f5c | 1393 | |
1dd05682 | 1394 | return ERR_PTR(ret); |
20897f5c AJ |
1395 | } |
1396 | ||
7237f183 YZ |
1397 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1398 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1399 | { |
1400 | struct btrfs_root *root; | |
1401 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1402 | struct extent_buffer *leaf; |
e02119d5 | 1403 | |
6f07e42e | 1404 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1405 | if (!root) |
7237f183 | 1406 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1407 | |
707e8a07 DS |
1408 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1409 | tree_root->stripesize, root, fs_info, | |
1410 | BTRFS_TREE_LOG_OBJECTID); | |
e02119d5 CM |
1411 | |
1412 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1413 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1414 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1415 | |
7237f183 | 1416 | /* |
27cdeb70 MX |
1417 | * DON'T set REF_COWS for log trees |
1418 | * | |
7237f183 YZ |
1419 | * log trees do not get reference counted because they go away |
1420 | * before a real commit is actually done. They do store pointers | |
1421 | * to file data extents, and those reference counts still get | |
1422 | * updated (along with back refs to the log tree). | |
1423 | */ | |
e02119d5 | 1424 | |
4d75f8a9 DS |
1425 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1426 | NULL, 0, 0, 0); | |
7237f183 YZ |
1427 | if (IS_ERR(leaf)) { |
1428 | kfree(root); | |
1429 | return ERR_CAST(leaf); | |
1430 | } | |
e02119d5 | 1431 | |
5d4f98a2 YZ |
1432 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1433 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1434 | btrfs_set_header_generation(leaf, trans->transid); | |
1435 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1436 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1437 | root->node = leaf; |
e02119d5 CM |
1438 | |
1439 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1440 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1441 | btrfs_mark_buffer_dirty(root->node); |
1442 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1443 | return root; |
1444 | } | |
1445 | ||
1446 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1447 | struct btrfs_fs_info *fs_info) | |
1448 | { | |
1449 | struct btrfs_root *log_root; | |
1450 | ||
1451 | log_root = alloc_log_tree(trans, fs_info); | |
1452 | if (IS_ERR(log_root)) | |
1453 | return PTR_ERR(log_root); | |
1454 | WARN_ON(fs_info->log_root_tree); | |
1455 | fs_info->log_root_tree = log_root; | |
1456 | return 0; | |
1457 | } | |
1458 | ||
1459 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1460 | struct btrfs_root *root) | |
1461 | { | |
1462 | struct btrfs_root *log_root; | |
1463 | struct btrfs_inode_item *inode_item; | |
1464 | ||
1465 | log_root = alloc_log_tree(trans, root->fs_info); | |
1466 | if (IS_ERR(log_root)) | |
1467 | return PTR_ERR(log_root); | |
1468 | ||
1469 | log_root->last_trans = trans->transid; | |
1470 | log_root->root_key.offset = root->root_key.objectid; | |
1471 | ||
1472 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1473 | btrfs_set_stack_inode_generation(inode_item, 1); |
1474 | btrfs_set_stack_inode_size(inode_item, 3); | |
1475 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
707e8a07 | 1476 | btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); |
3cae210f | 1477 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1478 | |
5d4f98a2 | 1479 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1480 | |
1481 | WARN_ON(root->log_root); | |
1482 | root->log_root = log_root; | |
1483 | root->log_transid = 0; | |
d1433deb | 1484 | root->log_transid_committed = -1; |
257c62e1 | 1485 | root->last_log_commit = 0; |
e02119d5 CM |
1486 | return 0; |
1487 | } | |
1488 | ||
35a3621b SB |
1489 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1490 | struct btrfs_key *key) | |
e02119d5 CM |
1491 | { |
1492 | struct btrfs_root *root; | |
1493 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1494 | struct btrfs_path *path; |
84234f3a | 1495 | u64 generation; |
cb517eab | 1496 | int ret; |
0f7d52f4 | 1497 | |
cb517eab MX |
1498 | path = btrfs_alloc_path(); |
1499 | if (!path) | |
0f7d52f4 | 1500 | return ERR_PTR(-ENOMEM); |
cb517eab MX |
1501 | |
1502 | root = btrfs_alloc_root(fs_info); | |
1503 | if (!root) { | |
1504 | ret = -ENOMEM; | |
1505 | goto alloc_fail; | |
0f7d52f4 CM |
1506 | } |
1507 | ||
707e8a07 DS |
1508 | __setup_root(tree_root->nodesize, tree_root->sectorsize, |
1509 | tree_root->stripesize, root, fs_info, key->objectid); | |
0f7d52f4 | 1510 | |
cb517eab MX |
1511 | ret = btrfs_find_root(tree_root, key, path, |
1512 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1513 | if (ret) { |
13a8a7c8 YZ |
1514 | if (ret > 0) |
1515 | ret = -ENOENT; | |
cb517eab | 1516 | goto find_fail; |
0f7d52f4 | 1517 | } |
13a8a7c8 | 1518 | |
84234f3a | 1519 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1520 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
ce86cd59 | 1521 | generation); |
64c043de LB |
1522 | if (IS_ERR(root->node)) { |
1523 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1524 | goto find_fail; |
1525 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1526 | ret = -EIO; | |
64c043de LB |
1527 | free_extent_buffer(root->node); |
1528 | goto find_fail; | |
416bc658 | 1529 | } |
5d4f98a2 | 1530 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1531 | out: |
cb517eab MX |
1532 | btrfs_free_path(path); |
1533 | return root; | |
1534 | ||
cb517eab MX |
1535 | find_fail: |
1536 | kfree(root); | |
1537 | alloc_fail: | |
1538 | root = ERR_PTR(ret); | |
1539 | goto out; | |
1540 | } | |
1541 | ||
1542 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1543 | struct btrfs_key *location) | |
1544 | { | |
1545 | struct btrfs_root *root; | |
1546 | ||
1547 | root = btrfs_read_tree_root(tree_root, location); | |
1548 | if (IS_ERR(root)) | |
1549 | return root; | |
1550 | ||
1551 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1552 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1553 | btrfs_check_and_init_root_item(&root->root_item); |
1554 | } | |
13a8a7c8 | 1555 | |
5eda7b5e CM |
1556 | return root; |
1557 | } | |
1558 | ||
cb517eab MX |
1559 | int btrfs_init_fs_root(struct btrfs_root *root) |
1560 | { | |
1561 | int ret; | |
8257b2dc | 1562 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1563 | |
1564 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1565 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1566 | GFP_NOFS); | |
1567 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1568 | ret = -ENOMEM; | |
1569 | goto fail; | |
1570 | } | |
1571 | ||
8257b2dc MX |
1572 | writers = btrfs_alloc_subvolume_writers(); |
1573 | if (IS_ERR(writers)) { | |
1574 | ret = PTR_ERR(writers); | |
1575 | goto fail; | |
1576 | } | |
1577 | root->subv_writers = writers; | |
1578 | ||
cb517eab | 1579 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1580 | spin_lock_init(&root->ino_cache_lock); |
1581 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1582 | |
1583 | ret = get_anon_bdev(&root->anon_dev); | |
1584 | if (ret) | |
8257b2dc | 1585 | goto free_writers; |
f32e48e9 CR |
1586 | |
1587 | mutex_lock(&root->objectid_mutex); | |
1588 | ret = btrfs_find_highest_objectid(root, | |
1589 | &root->highest_objectid); | |
1590 | if (ret) { | |
1591 | mutex_unlock(&root->objectid_mutex); | |
1592 | goto free_root_dev; | |
1593 | } | |
1594 | ||
1595 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1596 | ||
1597 | mutex_unlock(&root->objectid_mutex); | |
1598 | ||
cb517eab | 1599 | return 0; |
8257b2dc | 1600 | |
f32e48e9 CR |
1601 | free_root_dev: |
1602 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
1603 | free_writers: |
1604 | btrfs_free_subvolume_writers(root->subv_writers); | |
cb517eab MX |
1605 | fail: |
1606 | kfree(root->free_ino_ctl); | |
1607 | kfree(root->free_ino_pinned); | |
1608 | return ret; | |
1609 | } | |
1610 | ||
171170c1 ST |
1611 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1612 | u64 root_id) | |
cb517eab MX |
1613 | { |
1614 | struct btrfs_root *root; | |
1615 | ||
1616 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1617 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1618 | (unsigned long)root_id); | |
1619 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1620 | return root; | |
1621 | } | |
1622 | ||
1623 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1624 | struct btrfs_root *root) | |
1625 | { | |
1626 | int ret; | |
1627 | ||
1628 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); | |
1629 | if (ret) | |
1630 | return ret; | |
1631 | ||
1632 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1633 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1634 | (unsigned long)root->root_key.objectid, | |
1635 | root); | |
1636 | if (ret == 0) | |
27cdeb70 | 1637 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1638 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1639 | radix_tree_preload_end(); | |
1640 | ||
1641 | return ret; | |
1642 | } | |
1643 | ||
c00869f1 MX |
1644 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1645 | struct btrfs_key *location, | |
1646 | bool check_ref) | |
5eda7b5e CM |
1647 | { |
1648 | struct btrfs_root *root; | |
381cf658 | 1649 | struct btrfs_path *path; |
1d4c08e0 | 1650 | struct btrfs_key key; |
5eda7b5e CM |
1651 | int ret; |
1652 | ||
edbd8d4e CM |
1653 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1654 | return fs_info->tree_root; | |
1655 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1656 | return fs_info->extent_root; | |
8f18cf13 CM |
1657 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1658 | return fs_info->chunk_root; | |
1659 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1660 | return fs_info->dev_root; | |
0403e47e YZ |
1661 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1662 | return fs_info->csum_root; | |
bcef60f2 AJ |
1663 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1664 | return fs_info->quota_root ? fs_info->quota_root : | |
1665 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1666 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1667 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1668 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1669 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1670 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1671 | ERR_PTR(-ENOENT); | |
4df27c4d | 1672 | again: |
cb517eab | 1673 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1674 | if (root) { |
c00869f1 | 1675 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1676 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1677 | return root; |
48475471 | 1678 | } |
5eda7b5e | 1679 | |
cb517eab | 1680 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1681 | if (IS_ERR(root)) |
1682 | return root; | |
3394e160 | 1683 | |
c00869f1 | 1684 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1685 | ret = -ENOENT; |
581bb050 | 1686 | goto fail; |
35a30d7c | 1687 | } |
581bb050 | 1688 | |
cb517eab | 1689 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1690 | if (ret) |
1691 | goto fail; | |
3394e160 | 1692 | |
381cf658 DS |
1693 | path = btrfs_alloc_path(); |
1694 | if (!path) { | |
1695 | ret = -ENOMEM; | |
1696 | goto fail; | |
1697 | } | |
1d4c08e0 DS |
1698 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1699 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1700 | key.offset = location->objectid; | |
1701 | ||
1702 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1703 | btrfs_free_path(path); |
d68fc57b YZ |
1704 | if (ret < 0) |
1705 | goto fail; | |
1706 | if (ret == 0) | |
27cdeb70 | 1707 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1708 | |
cb517eab | 1709 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1710 | if (ret) { |
4df27c4d YZ |
1711 | if (ret == -EEXIST) { |
1712 | free_fs_root(root); | |
1713 | goto again; | |
1714 | } | |
1715 | goto fail; | |
0f7d52f4 | 1716 | } |
edbd8d4e | 1717 | return root; |
4df27c4d YZ |
1718 | fail: |
1719 | free_fs_root(root); | |
1720 | return ERR_PTR(ret); | |
edbd8d4e CM |
1721 | } |
1722 | ||
04160088 CM |
1723 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1724 | { | |
1725 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1726 | int ret = 0; | |
04160088 CM |
1727 | struct btrfs_device *device; |
1728 | struct backing_dev_info *bdi; | |
b7967db7 | 1729 | |
1f78160c XG |
1730 | rcu_read_lock(); |
1731 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1732 | if (!device->bdev) |
1733 | continue; | |
04160088 | 1734 | bdi = blk_get_backing_dev_info(device->bdev); |
ff9ea323 | 1735 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1736 | ret = 1; |
1737 | break; | |
1738 | } | |
1739 | } | |
1f78160c | 1740 | rcu_read_unlock(); |
04160088 CM |
1741 | return ret; |
1742 | } | |
1743 | ||
04160088 CM |
1744 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1745 | { | |
ad081f14 JA |
1746 | int err; |
1747 | ||
b4caecd4 | 1748 | err = bdi_setup_and_register(bdi, "btrfs"); |
ad081f14 JA |
1749 | if (err) |
1750 | return err; | |
1751 | ||
df0ce26c | 1752 | bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE; |
04160088 CM |
1753 | bdi->congested_fn = btrfs_congested_fn; |
1754 | bdi->congested_data = info; | |
da2f0f74 | 1755 | bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; |
04160088 CM |
1756 | return 0; |
1757 | } | |
1758 | ||
8b712842 CM |
1759 | /* |
1760 | * called by the kthread helper functions to finally call the bio end_io | |
1761 | * functions. This is where read checksum verification actually happens | |
1762 | */ | |
1763 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1764 | { |
ce9adaa5 | 1765 | struct bio *bio; |
97eb6b69 | 1766 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1767 | |
97eb6b69 | 1768 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1769 | bio = end_io_wq->bio; |
ce9adaa5 | 1770 | |
4246a0b6 | 1771 | bio->bi_error = end_io_wq->error; |
8b712842 CM |
1772 | bio->bi_private = end_io_wq->private; |
1773 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1774 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1775 | bio_endio(bio); |
44b8bd7e CM |
1776 | } |
1777 | ||
a74a4b97 CM |
1778 | static int cleaner_kthread(void *arg) |
1779 | { | |
1780 | struct btrfs_root *root = arg; | |
d0278245 | 1781 | int again; |
da288d28 | 1782 | struct btrfs_trans_handle *trans; |
a74a4b97 | 1783 | |
69624913 | 1784 | set_freezable(); |
a74a4b97 | 1785 | do { |
d0278245 | 1786 | again = 0; |
a74a4b97 | 1787 | |
d0278245 | 1788 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1789 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1790 | goto sleep; |
1791 | ||
1792 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) | |
1793 | goto sleep; | |
1794 | ||
dc7f370c MX |
1795 | /* |
1796 | * Avoid the problem that we change the status of the fs | |
1797 | * during the above check and trylock. | |
1798 | */ | |
babbf170 | 1799 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1800 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1801 | goto sleep; | |
76dda93c | 1802 | } |
a74a4b97 | 1803 | |
c2d6cb16 | 1804 | mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); |
d0278245 | 1805 | btrfs_run_delayed_iputs(root); |
c2d6cb16 FM |
1806 | mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); |
1807 | ||
d0278245 MX |
1808 | again = btrfs_clean_one_deleted_snapshot(root); |
1809 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1810 | ||
1811 | /* | |
05323cd1 MX |
1812 | * The defragger has dealt with the R/O remount and umount, |
1813 | * needn't do anything special here. | |
d0278245 MX |
1814 | */ |
1815 | btrfs_run_defrag_inodes(root->fs_info); | |
67c5e7d4 FM |
1816 | |
1817 | /* | |
1818 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1819 | * with relocation (btrfs_relocate_chunk) and relocation | |
1820 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1821 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1822 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1823 | * unused block groups. | |
1824 | */ | |
1825 | btrfs_delete_unused_bgs(root->fs_info); | |
d0278245 | 1826 | sleep: |
9d1a2a3a | 1827 | if (!try_to_freeze() && !again) { |
a74a4b97 | 1828 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1829 | if (!kthread_should_stop()) |
1830 | schedule(); | |
a74a4b97 CM |
1831 | __set_current_state(TASK_RUNNING); |
1832 | } | |
1833 | } while (!kthread_should_stop()); | |
da288d28 FM |
1834 | |
1835 | /* | |
1836 | * Transaction kthread is stopped before us and wakes us up. | |
1837 | * However we might have started a new transaction and COWed some | |
1838 | * tree blocks when deleting unused block groups for example. So | |
1839 | * make sure we commit the transaction we started to have a clean | |
1840 | * shutdown when evicting the btree inode - if it has dirty pages | |
1841 | * when we do the final iput() on it, eviction will trigger a | |
1842 | * writeback for it which will fail with null pointer dereferences | |
1843 | * since work queues and other resources were already released and | |
1844 | * destroyed by the time the iput/eviction/writeback is made. | |
1845 | */ | |
1846 | trans = btrfs_attach_transaction(root); | |
1847 | if (IS_ERR(trans)) { | |
1848 | if (PTR_ERR(trans) != -ENOENT) | |
1849 | btrfs_err(root->fs_info, | |
1850 | "cleaner transaction attach returned %ld", | |
1851 | PTR_ERR(trans)); | |
1852 | } else { | |
1853 | int ret; | |
1854 | ||
1855 | ret = btrfs_commit_transaction(trans, root); | |
1856 | if (ret) | |
1857 | btrfs_err(root->fs_info, | |
1858 | "cleaner open transaction commit returned %d", | |
1859 | ret); | |
1860 | } | |
1861 | ||
a74a4b97 CM |
1862 | return 0; |
1863 | } | |
1864 | ||
1865 | static int transaction_kthread(void *arg) | |
1866 | { | |
1867 | struct btrfs_root *root = arg; | |
1868 | struct btrfs_trans_handle *trans; | |
1869 | struct btrfs_transaction *cur; | |
8929ecfa | 1870 | u64 transid; |
a74a4b97 CM |
1871 | unsigned long now; |
1872 | unsigned long delay; | |
914b2007 | 1873 | bool cannot_commit; |
a74a4b97 CM |
1874 | |
1875 | do { | |
914b2007 | 1876 | cannot_commit = false; |
8b87dc17 | 1877 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1878 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1879 | ||
a4abeea4 | 1880 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1881 | cur = root->fs_info->running_transaction; |
1882 | if (!cur) { | |
a4abeea4 | 1883 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1884 | goto sleep; |
1885 | } | |
31153d81 | 1886 | |
a74a4b97 | 1887 | now = get_seconds(); |
4a9d8bde | 1888 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1889 | (now < cur->start_time || |
1890 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1891 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1892 | delay = HZ * 5; |
1893 | goto sleep; | |
1894 | } | |
8929ecfa | 1895 | transid = cur->transid; |
a4abeea4 | 1896 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1897 | |
79787eaa | 1898 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1899 | trans = btrfs_attach_transaction(root); |
914b2007 | 1900 | if (IS_ERR(trans)) { |
354aa0fb MX |
1901 | if (PTR_ERR(trans) != -ENOENT) |
1902 | cannot_commit = true; | |
79787eaa | 1903 | goto sleep; |
914b2007 | 1904 | } |
8929ecfa | 1905 | if (transid == trans->transid) { |
79787eaa | 1906 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1907 | } else { |
1908 | btrfs_end_transaction(trans, root); | |
1909 | } | |
a74a4b97 CM |
1910 | sleep: |
1911 | wake_up_process(root->fs_info->cleaner_kthread); | |
1912 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1913 | ||
4e121c06 JB |
1914 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1915 | &root->fs_info->fs_state))) | |
1916 | btrfs_cleanup_transaction(root); | |
a0acae0e | 1917 | if (!try_to_freeze()) { |
a74a4b97 | 1918 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1919 | if (!kthread_should_stop() && |
914b2007 JK |
1920 | (!btrfs_transaction_blocked(root->fs_info) || |
1921 | cannot_commit)) | |
8929ecfa | 1922 | schedule_timeout(delay); |
a74a4b97 CM |
1923 | __set_current_state(TASK_RUNNING); |
1924 | } | |
1925 | } while (!kthread_should_stop()); | |
1926 | return 0; | |
1927 | } | |
1928 | ||
af31f5e5 CM |
1929 | /* |
1930 | * this will find the highest generation in the array of | |
1931 | * root backups. The index of the highest array is returned, | |
1932 | * or -1 if we can't find anything. | |
1933 | * | |
1934 | * We check to make sure the array is valid by comparing the | |
1935 | * generation of the latest root in the array with the generation | |
1936 | * in the super block. If they don't match we pitch it. | |
1937 | */ | |
1938 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1939 | { | |
1940 | u64 cur; | |
1941 | int newest_index = -1; | |
1942 | struct btrfs_root_backup *root_backup; | |
1943 | int i; | |
1944 | ||
1945 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1946 | root_backup = info->super_copy->super_roots + i; | |
1947 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1948 | if (cur == newest_gen) | |
1949 | newest_index = i; | |
1950 | } | |
1951 | ||
1952 | /* check to see if we actually wrapped around */ | |
1953 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1954 | root_backup = info->super_copy->super_roots; | |
1955 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1956 | if (cur == newest_gen) | |
1957 | newest_index = 0; | |
1958 | } | |
1959 | return newest_index; | |
1960 | } | |
1961 | ||
1962 | ||
1963 | /* | |
1964 | * find the oldest backup so we know where to store new entries | |
1965 | * in the backup array. This will set the backup_root_index | |
1966 | * field in the fs_info struct | |
1967 | */ | |
1968 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1969 | u64 newest_gen) | |
1970 | { | |
1971 | int newest_index = -1; | |
1972 | ||
1973 | newest_index = find_newest_super_backup(info, newest_gen); | |
1974 | /* if there was garbage in there, just move along */ | |
1975 | if (newest_index == -1) { | |
1976 | info->backup_root_index = 0; | |
1977 | } else { | |
1978 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1979 | } | |
1980 | } | |
1981 | ||
1982 | /* | |
1983 | * copy all the root pointers into the super backup array. | |
1984 | * this will bump the backup pointer by one when it is | |
1985 | * done | |
1986 | */ | |
1987 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1988 | { | |
1989 | int next_backup; | |
1990 | struct btrfs_root_backup *root_backup; | |
1991 | int last_backup; | |
1992 | ||
1993 | next_backup = info->backup_root_index; | |
1994 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1995 | BTRFS_NUM_BACKUP_ROOTS; | |
1996 | ||
1997 | /* | |
1998 | * just overwrite the last backup if we're at the same generation | |
1999 | * this happens only at umount | |
2000 | */ | |
2001 | root_backup = info->super_for_commit->super_roots + last_backup; | |
2002 | if (btrfs_backup_tree_root_gen(root_backup) == | |
2003 | btrfs_header_generation(info->tree_root->node)) | |
2004 | next_backup = last_backup; | |
2005 | ||
2006 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2007 | ||
2008 | /* | |
2009 | * make sure all of our padding and empty slots get zero filled | |
2010 | * regardless of which ones we use today | |
2011 | */ | |
2012 | memset(root_backup, 0, sizeof(*root_backup)); | |
2013 | ||
2014 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2015 | ||
2016 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2017 | btrfs_set_backup_tree_root_gen(root_backup, | |
2018 | btrfs_header_generation(info->tree_root->node)); | |
2019 | ||
2020 | btrfs_set_backup_tree_root_level(root_backup, | |
2021 | btrfs_header_level(info->tree_root->node)); | |
2022 | ||
2023 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2024 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2025 | btrfs_header_generation(info->chunk_root->node)); | |
2026 | btrfs_set_backup_chunk_root_level(root_backup, | |
2027 | btrfs_header_level(info->chunk_root->node)); | |
2028 | ||
2029 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
2030 | btrfs_set_backup_extent_root_gen(root_backup, | |
2031 | btrfs_header_generation(info->extent_root->node)); | |
2032 | btrfs_set_backup_extent_root_level(root_backup, | |
2033 | btrfs_header_level(info->extent_root->node)); | |
2034 | ||
7c7e82a7 CM |
2035 | /* |
2036 | * we might commit during log recovery, which happens before we set | |
2037 | * the fs_root. Make sure it is valid before we fill it in. | |
2038 | */ | |
2039 | if (info->fs_root && info->fs_root->node) { | |
2040 | btrfs_set_backup_fs_root(root_backup, | |
2041 | info->fs_root->node->start); | |
2042 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2043 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2044 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2045 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2046 | } |
af31f5e5 CM |
2047 | |
2048 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2049 | btrfs_set_backup_dev_root_gen(root_backup, | |
2050 | btrfs_header_generation(info->dev_root->node)); | |
2051 | btrfs_set_backup_dev_root_level(root_backup, | |
2052 | btrfs_header_level(info->dev_root->node)); | |
2053 | ||
2054 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
2055 | btrfs_set_backup_csum_root_gen(root_backup, | |
2056 | btrfs_header_generation(info->csum_root->node)); | |
2057 | btrfs_set_backup_csum_root_level(root_backup, | |
2058 | btrfs_header_level(info->csum_root->node)); | |
2059 | ||
2060 | btrfs_set_backup_total_bytes(root_backup, | |
2061 | btrfs_super_total_bytes(info->super_copy)); | |
2062 | btrfs_set_backup_bytes_used(root_backup, | |
2063 | btrfs_super_bytes_used(info->super_copy)); | |
2064 | btrfs_set_backup_num_devices(root_backup, | |
2065 | btrfs_super_num_devices(info->super_copy)); | |
2066 | ||
2067 | /* | |
2068 | * if we don't copy this out to the super_copy, it won't get remembered | |
2069 | * for the next commit | |
2070 | */ | |
2071 | memcpy(&info->super_copy->super_roots, | |
2072 | &info->super_for_commit->super_roots, | |
2073 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2074 | } | |
2075 | ||
2076 | /* | |
2077 | * this copies info out of the root backup array and back into | |
2078 | * the in-memory super block. It is meant to help iterate through | |
2079 | * the array, so you send it the number of backups you've already | |
2080 | * tried and the last backup index you used. | |
2081 | * | |
2082 | * this returns -1 when it has tried all the backups | |
2083 | */ | |
2084 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
2085 | struct btrfs_super_block *super, | |
2086 | int *num_backups_tried, int *backup_index) | |
2087 | { | |
2088 | struct btrfs_root_backup *root_backup; | |
2089 | int newest = *backup_index; | |
2090 | ||
2091 | if (*num_backups_tried == 0) { | |
2092 | u64 gen = btrfs_super_generation(super); | |
2093 | ||
2094 | newest = find_newest_super_backup(info, gen); | |
2095 | if (newest == -1) | |
2096 | return -1; | |
2097 | ||
2098 | *backup_index = newest; | |
2099 | *num_backups_tried = 1; | |
2100 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2101 | /* we've tried all the backups, all done */ | |
2102 | return -1; | |
2103 | } else { | |
2104 | /* jump to the next oldest backup */ | |
2105 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2106 | BTRFS_NUM_BACKUP_ROOTS; | |
2107 | *backup_index = newest; | |
2108 | *num_backups_tried += 1; | |
2109 | } | |
2110 | root_backup = super->super_roots + newest; | |
2111 | ||
2112 | btrfs_set_super_generation(super, | |
2113 | btrfs_backup_tree_root_gen(root_backup)); | |
2114 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2115 | btrfs_set_super_root_level(super, | |
2116 | btrfs_backup_tree_root_level(root_backup)); | |
2117 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2118 | ||
2119 | /* | |
2120 | * fixme: the total bytes and num_devices need to match or we should | |
2121 | * need a fsck | |
2122 | */ | |
2123 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2124 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2125 | return 0; | |
2126 | } | |
2127 | ||
7abadb64 LB |
2128 | /* helper to cleanup workers */ |
2129 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2130 | { | |
dc6e3209 | 2131 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2132 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2133 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 QW |
2134 | btrfs_destroy_workqueue(fs_info->endio_workers); |
2135 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2136 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); | |
8b110e39 | 2137 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2138 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2139 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); |
2140 | btrfs_destroy_workqueue(fs_info->endio_write_workers); | |
2141 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2142 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2143 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2144 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2145 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2146 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2147 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2148 | btrfs_destroy_workqueue(fs_info->extent_workers); |
7abadb64 LB |
2149 | } |
2150 | ||
2e9f5954 R |
2151 | static void free_root_extent_buffers(struct btrfs_root *root) |
2152 | { | |
2153 | if (root) { | |
2154 | free_extent_buffer(root->node); | |
2155 | free_extent_buffer(root->commit_root); | |
2156 | root->node = NULL; | |
2157 | root->commit_root = NULL; | |
2158 | } | |
2159 | } | |
2160 | ||
af31f5e5 CM |
2161 | /* helper to cleanup tree roots */ |
2162 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2163 | { | |
2e9f5954 | 2164 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2165 | |
2e9f5954 R |
2166 | free_root_extent_buffers(info->dev_root); |
2167 | free_root_extent_buffers(info->extent_root); | |
2168 | free_root_extent_buffers(info->csum_root); | |
2169 | free_root_extent_buffers(info->quota_root); | |
2170 | free_root_extent_buffers(info->uuid_root); | |
2171 | if (chunk_root) | |
2172 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2173 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2174 | } |
2175 | ||
faa2dbf0 | 2176 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2177 | { |
2178 | int ret; | |
2179 | struct btrfs_root *gang[8]; | |
2180 | int i; | |
2181 | ||
2182 | while (!list_empty(&fs_info->dead_roots)) { | |
2183 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2184 | struct btrfs_root, root_list); | |
2185 | list_del(&gang[0]->root_list); | |
2186 | ||
27cdeb70 | 2187 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2188 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2189 | } else { |
2190 | free_extent_buffer(gang[0]->node); | |
2191 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2192 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2193 | } |
2194 | } | |
2195 | ||
2196 | while (1) { | |
2197 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2198 | (void **)gang, 0, | |
2199 | ARRAY_SIZE(gang)); | |
2200 | if (!ret) | |
2201 | break; | |
2202 | for (i = 0; i < ret; i++) | |
cb517eab | 2203 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2204 | } |
1a4319cc LB |
2205 | |
2206 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2207 | btrfs_free_log_root_tree(NULL, fs_info); | |
2208 | btrfs_destroy_pinned_extent(fs_info->tree_root, | |
2209 | fs_info->pinned_extents); | |
2210 | } | |
171f6537 | 2211 | } |
af31f5e5 | 2212 | |
638aa7ed ES |
2213 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2214 | { | |
2215 | mutex_init(&fs_info->scrub_lock); | |
2216 | atomic_set(&fs_info->scrubs_running, 0); | |
2217 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2218 | atomic_set(&fs_info->scrubs_paused, 0); | |
2219 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2220 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2221 | fs_info->scrub_workers_refcnt = 0; | |
2222 | } | |
2223 | ||
779a65a4 ES |
2224 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2225 | { | |
2226 | spin_lock_init(&fs_info->balance_lock); | |
2227 | mutex_init(&fs_info->balance_mutex); | |
2228 | atomic_set(&fs_info->balance_running, 0); | |
2229 | atomic_set(&fs_info->balance_pause_req, 0); | |
2230 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2231 | fs_info->balance_ctl = NULL; | |
2232 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2233 | } | |
2234 | ||
f37938e0 ES |
2235 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, |
2236 | struct btrfs_root *tree_root) | |
2237 | { | |
2238 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2239 | set_nlink(fs_info->btree_inode, 1); | |
2240 | /* | |
2241 | * we set the i_size on the btree inode to the max possible int. | |
2242 | * the real end of the address space is determined by all of | |
2243 | * the devices in the system | |
2244 | */ | |
2245 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
2246 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 ES |
2247 | |
2248 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); | |
2249 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, | |
2250 | fs_info->btree_inode->i_mapping); | |
2251 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; | |
2252 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); | |
2253 | ||
2254 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
2255 | ||
2256 | BTRFS_I(fs_info->btree_inode)->root = tree_root; | |
2257 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2258 | sizeof(struct btrfs_key)); | |
2259 | set_bit(BTRFS_INODE_DUMMY, | |
2260 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
2261 | btrfs_insert_inode_hash(fs_info->btree_inode); | |
2262 | } | |
2263 | ||
ad618368 ES |
2264 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2265 | { | |
2266 | fs_info->dev_replace.lock_owner = 0; | |
2267 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2268 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
2269 | mutex_init(&fs_info->dev_replace.lock_management_lock); | |
2270 | mutex_init(&fs_info->dev_replace.lock); | |
2271 | init_waitqueue_head(&fs_info->replace_wait); | |
2272 | } | |
2273 | ||
f9e92e40 ES |
2274 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2275 | { | |
2276 | spin_lock_init(&fs_info->qgroup_lock); | |
2277 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2278 | fs_info->qgroup_tree = RB_ROOT; | |
2279 | fs_info->qgroup_op_tree = RB_ROOT; | |
2280 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2281 | fs_info->qgroup_seq = 1; | |
2282 | fs_info->quota_enabled = 0; | |
2283 | fs_info->pending_quota_state = 0; | |
2284 | fs_info->qgroup_ulist = NULL; | |
2285 | mutex_init(&fs_info->qgroup_rescan_lock); | |
2286 | } | |
2287 | ||
2a458198 ES |
2288 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2289 | struct btrfs_fs_devices *fs_devices) | |
2290 | { | |
2291 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2292 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2293 | |
2294 | fs_info->workers = | |
2295 | btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, | |
2296 | max_active, 16); | |
2297 | ||
2298 | fs_info->delalloc_workers = | |
2299 | btrfs_alloc_workqueue("delalloc", flags, max_active, 2); | |
2300 | ||
2301 | fs_info->flush_workers = | |
2302 | btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); | |
2303 | ||
2304 | fs_info->caching_workers = | |
2305 | btrfs_alloc_workqueue("cache", flags, max_active, 0); | |
2306 | ||
2307 | /* | |
2308 | * a higher idle thresh on the submit workers makes it much more | |
2309 | * likely that bios will be send down in a sane order to the | |
2310 | * devices | |
2311 | */ | |
2312 | fs_info->submit_workers = | |
2313 | btrfs_alloc_workqueue("submit", flags, | |
2314 | min_t(u64, fs_devices->num_devices, | |
2315 | max_active), 64); | |
2316 | ||
2317 | fs_info->fixup_workers = | |
2318 | btrfs_alloc_workqueue("fixup", flags, 1, 0); | |
2319 | ||
2320 | /* | |
2321 | * endios are largely parallel and should have a very | |
2322 | * low idle thresh | |
2323 | */ | |
2324 | fs_info->endio_workers = | |
2325 | btrfs_alloc_workqueue("endio", flags, max_active, 4); | |
2326 | fs_info->endio_meta_workers = | |
2327 | btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); | |
2328 | fs_info->endio_meta_write_workers = | |
2329 | btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); | |
2330 | fs_info->endio_raid56_workers = | |
2331 | btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); | |
2332 | fs_info->endio_repair_workers = | |
2333 | btrfs_alloc_workqueue("endio-repair", flags, 1, 0); | |
2334 | fs_info->rmw_workers = | |
2335 | btrfs_alloc_workqueue("rmw", flags, max_active, 2); | |
2336 | fs_info->endio_write_workers = | |
2337 | btrfs_alloc_workqueue("endio-write", flags, max_active, 2); | |
2338 | fs_info->endio_freespace_worker = | |
2339 | btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); | |
2340 | fs_info->delayed_workers = | |
2341 | btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); | |
2342 | fs_info->readahead_workers = | |
2343 | btrfs_alloc_workqueue("readahead", flags, max_active, 2); | |
2344 | fs_info->qgroup_rescan_workers = | |
2345 | btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); | |
2346 | fs_info->extent_workers = | |
2347 | btrfs_alloc_workqueue("extent-refs", flags, | |
2348 | min_t(u64, fs_devices->num_devices, | |
2349 | max_active), 8); | |
2350 | ||
2351 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2352 | fs_info->submit_workers && fs_info->flush_workers && | |
2353 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2354 | fs_info->endio_meta_write_workers && | |
2355 | fs_info->endio_repair_workers && | |
2356 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2357 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2358 | fs_info->caching_workers && fs_info->readahead_workers && | |
2359 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2360 | fs_info->extent_workers && | |
2361 | fs_info->qgroup_rescan_workers)) { | |
2362 | return -ENOMEM; | |
2363 | } | |
2364 | ||
2365 | return 0; | |
2366 | } | |
2367 | ||
63443bf5 ES |
2368 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2369 | struct btrfs_fs_devices *fs_devices) | |
2370 | { | |
2371 | int ret; | |
2372 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2373 | struct btrfs_root *log_tree_root; | |
2374 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2375 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2376 | ||
2377 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2378 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2379 | return -EIO; |
2380 | } | |
2381 | ||
2382 | log_tree_root = btrfs_alloc_root(fs_info); | |
2383 | if (!log_tree_root) | |
2384 | return -ENOMEM; | |
2385 | ||
2386 | __setup_root(tree_root->nodesize, tree_root->sectorsize, | |
2387 | tree_root->stripesize, log_tree_root, fs_info, | |
2388 | BTRFS_TREE_LOG_OBJECTID); | |
2389 | ||
2390 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
2391 | fs_info->generation + 1); | |
64c043de | 2392 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2393 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2394 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2395 | kfree(log_tree_root); |
0eeff236 | 2396 | return ret; |
64c043de | 2397 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2398 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2399 | free_extent_buffer(log_tree_root->node); |
2400 | kfree(log_tree_root); | |
2401 | return -EIO; | |
2402 | } | |
2403 | /* returns with log_tree_root freed on success */ | |
2404 | ret = btrfs_recover_log_trees(log_tree_root); | |
2405 | if (ret) { | |
a4553fef | 2406 | btrfs_std_error(tree_root->fs_info, ret, |
63443bf5 ES |
2407 | "Failed to recover log tree"); |
2408 | free_extent_buffer(log_tree_root->node); | |
2409 | kfree(log_tree_root); | |
2410 | return ret; | |
2411 | } | |
2412 | ||
2413 | if (fs_info->sb->s_flags & MS_RDONLY) { | |
2414 | ret = btrfs_commit_super(tree_root); | |
2415 | if (ret) | |
2416 | return ret; | |
2417 | } | |
2418 | ||
2419 | return 0; | |
2420 | } | |
2421 | ||
4bbcaa64 ES |
2422 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info, |
2423 | struct btrfs_root *tree_root) | |
2424 | { | |
a4f3d2c4 | 2425 | struct btrfs_root *root; |
4bbcaa64 ES |
2426 | struct btrfs_key location; |
2427 | int ret; | |
2428 | ||
2429 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; | |
2430 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2431 | location.offset = 0; | |
2432 | ||
a4f3d2c4 DS |
2433 | root = btrfs_read_tree_root(tree_root, &location); |
2434 | if (IS_ERR(root)) | |
2435 | return PTR_ERR(root); | |
2436 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2437 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2438 | |
2439 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2440 | root = btrfs_read_tree_root(tree_root, &location); |
2441 | if (IS_ERR(root)) | |
2442 | return PTR_ERR(root); | |
2443 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2444 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2445 | btrfs_init_devices_late(fs_info); |
2446 | ||
2447 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2448 | root = btrfs_read_tree_root(tree_root, &location); |
2449 | if (IS_ERR(root)) | |
2450 | return PTR_ERR(root); | |
2451 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2452 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2453 | |
2454 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2455 | root = btrfs_read_tree_root(tree_root, &location); |
2456 | if (!IS_ERR(root)) { | |
2457 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
4bbcaa64 ES |
2458 | fs_info->quota_enabled = 1; |
2459 | fs_info->pending_quota_state = 1; | |
a4f3d2c4 | 2460 | fs_info->quota_root = root; |
4bbcaa64 ES |
2461 | } |
2462 | ||
2463 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2464 | root = btrfs_read_tree_root(tree_root, &location); |
2465 | if (IS_ERR(root)) { | |
2466 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2467 | if (ret != -ENOENT) |
2468 | return ret; | |
2469 | } else { | |
a4f3d2c4 DS |
2470 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2471 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2472 | } |
2473 | ||
70f6d82e OS |
2474 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2475 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2476 | root = btrfs_read_tree_root(tree_root, &location); | |
2477 | if (IS_ERR(root)) | |
2478 | return PTR_ERR(root); | |
2479 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2480 | fs_info->free_space_root = root; | |
2481 | } | |
2482 | ||
4bbcaa64 ES |
2483 | return 0; |
2484 | } | |
2485 | ||
ad2b2c80 AV |
2486 | int open_ctree(struct super_block *sb, |
2487 | struct btrfs_fs_devices *fs_devices, | |
2488 | char *options) | |
2e635a27 | 2489 | { |
db94535d CM |
2490 | u32 sectorsize; |
2491 | u32 nodesize; | |
87ee04eb | 2492 | u32 stripesize; |
84234f3a | 2493 | u64 generation; |
f2b636e8 | 2494 | u64 features; |
3de4586c | 2495 | struct btrfs_key location; |
a061fc8d | 2496 | struct buffer_head *bh; |
4d34b278 | 2497 | struct btrfs_super_block *disk_super; |
815745cf | 2498 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2499 | struct btrfs_root *tree_root; |
4d34b278 | 2500 | struct btrfs_root *chunk_root; |
eb60ceac | 2501 | int ret; |
e58ca020 | 2502 | int err = -EINVAL; |
af31f5e5 CM |
2503 | int num_backups_tried = 0; |
2504 | int backup_index = 0; | |
5cdc7ad3 | 2505 | int max_active; |
4543df7e | 2506 | |
f84a8bd6 | 2507 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e | 2508 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); |
cb517eab | 2509 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2510 | err = -ENOMEM; |
2511 | goto fail; | |
2512 | } | |
76dda93c YZ |
2513 | |
2514 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2515 | if (ret) { | |
2516 | err = ret; | |
2517 | goto fail; | |
2518 | } | |
2519 | ||
2520 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2521 | if (ret) { | |
2522 | err = ret; | |
2523 | goto fail_srcu; | |
2524 | } | |
2525 | ||
908c7f19 | 2526 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2527 | if (ret) { |
2528 | err = ret; | |
2529 | goto fail_bdi; | |
2530 | } | |
2531 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2532 | (1 + ilog2(nr_cpu_ids)); | |
2533 | ||
908c7f19 | 2534 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2535 | if (ret) { |
2536 | err = ret; | |
2537 | goto fail_dirty_metadata_bytes; | |
2538 | } | |
2539 | ||
908c7f19 | 2540 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2541 | if (ret) { |
2542 | err = ret; | |
2543 | goto fail_delalloc_bytes; | |
2544 | } | |
2545 | ||
76dda93c YZ |
2546 | fs_info->btree_inode = new_inode(sb); |
2547 | if (!fs_info->btree_inode) { | |
2548 | err = -ENOMEM; | |
c404e0dc | 2549 | goto fail_bio_counter; |
76dda93c YZ |
2550 | } |
2551 | ||
a6591715 | 2552 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2553 | |
76dda93c | 2554 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2555 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2556 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2557 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2558 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2559 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2560 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2561 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2562 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2563 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2564 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2565 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2566 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2567 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2568 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2569 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2570 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2571 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2572 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2573 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2574 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2575 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2576 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2577 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2578 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2579 | |
0b86a832 | 2580 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2581 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2582 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2583 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2584 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2585 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2586 | BTRFS_BLOCK_RSV_GLOBAL); | |
2587 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2588 | BTRFS_BLOCK_RSV_DELALLOC); | |
2589 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2590 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2591 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2592 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2593 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2594 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2595 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2596 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2597 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2598 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2599 | atomic_set(&fs_info->qgroup_op_seq, 0); |
fc36ed7e | 2600 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2601 | fs_info->sb = sb; |
95ac567a | 2602 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2603 | fs_info->metadata_ratio = 0; |
4cb5300b | 2604 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2605 | fs_info->free_chunk_space = 0; |
f29021b2 | 2606 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2607 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2608 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2609 | /* readahead state */ |
d0164adc | 2610 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2611 | spin_lock_init(&fs_info->reada_lock); |
c8b97818 | 2612 | |
b34b086c CM |
2613 | fs_info->thread_pool_size = min_t(unsigned long, |
2614 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2615 | |
199c2a9c MX |
2616 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2617 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 MX |
2618 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2619 | GFP_NOFS); | |
2620 | if (!fs_info->delayed_root) { | |
2621 | err = -ENOMEM; | |
2622 | goto fail_iput; | |
2623 | } | |
2624 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2625 | |
638aa7ed | 2626 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2627 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2628 | fs_info->check_integrity_print_mask = 0; | |
2629 | #endif | |
779a65a4 | 2630 | btrfs_init_balance(fs_info); |
21c7e756 | 2631 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2632 | |
a061fc8d CM |
2633 | sb->s_blocksize = 4096; |
2634 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2635 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2636 | |
f37938e0 | 2637 | btrfs_init_btree_inode(fs_info, tree_root); |
76dda93c | 2638 | |
0f9dd46c | 2639 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2640 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2641 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2642 | |
11833d66 | 2643 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2644 | fs_info->btree_inode->i_mapping); |
11833d66 | 2645 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2646 | fs_info->btree_inode->i_mapping); |
11833d66 | 2647 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2648 | fs_info->do_barriers = 1; |
e18e4809 | 2649 | |
39279cc3 | 2650 | |
5a3f23d5 | 2651 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2652 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2653 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2654 | mutex_init(&fs_info->transaction_kthread_mutex); |
2655 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2656 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2657 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2658 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2659 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2660 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2661 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2662 | |
ad618368 | 2663 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2664 | btrfs_init_qgroup(fs_info); |
416ac51d | 2665 | |
fa9c0d79 CM |
2666 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2667 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2668 | ||
e6dcd2dc | 2669 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2670 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2671 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2672 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2673 | |
04216820 FM |
2674 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2675 | ||
53b381b3 DW |
2676 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2677 | if (ret) { | |
83c8266a | 2678 | err = ret; |
53b381b3 DW |
2679 | goto fail_alloc; |
2680 | } | |
2681 | ||
707e8a07 | 2682 | __setup_root(4096, 4096, 4096, tree_root, |
2c90e5d6 | 2683 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2684 | |
3c4bb26b | 2685 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2686 | |
2687 | /* | |
2688 | * Read super block and check the signature bytes only | |
2689 | */ | |
a512bbf8 | 2690 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2691 | if (IS_ERR(bh)) { |
2692 | err = PTR_ERR(bh); | |
16cdcec7 | 2693 | goto fail_alloc; |
20b45077 | 2694 | } |
39279cc3 | 2695 | |
1104a885 DS |
2696 | /* |
2697 | * We want to check superblock checksum, the type is stored inside. | |
2698 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2699 | */ | |
2700 | if (btrfs_check_super_csum(bh->b_data)) { | |
efe120a0 | 2701 | printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); |
1104a885 | 2702 | err = -EINVAL; |
b2acdddf | 2703 | brelse(bh); |
1104a885 DS |
2704 | goto fail_alloc; |
2705 | } | |
2706 | ||
2707 | /* | |
2708 | * super_copy is zeroed at allocation time and we never touch the | |
2709 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2710 | * the whole block of INFO_SIZE | |
2711 | */ | |
6c41761f DS |
2712 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2713 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2714 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2715 | brelse(bh); |
5f39d397 | 2716 | |
6c41761f | 2717 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2718 | |
1104a885 DS |
2719 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2720 | if (ret) { | |
efe120a0 | 2721 | printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); |
1104a885 DS |
2722 | err = -EINVAL; |
2723 | goto fail_alloc; | |
2724 | } | |
2725 | ||
6c41761f | 2726 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2727 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2728 | goto fail_alloc; |
0f7d52f4 | 2729 | |
acce952b | 2730 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2731 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2732 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2733 | |
af31f5e5 CM |
2734 | /* |
2735 | * run through our array of backup supers and setup | |
2736 | * our ring pointer to the oldest one | |
2737 | */ | |
2738 | generation = btrfs_super_generation(disk_super); | |
2739 | find_oldest_super_backup(fs_info, generation); | |
2740 | ||
75e7cb7f LB |
2741 | /* |
2742 | * In the long term, we'll store the compression type in the super | |
2743 | * block, and it'll be used for per file compression control. | |
2744 | */ | |
2745 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2746 | ||
2b82032c YZ |
2747 | ret = btrfs_parse_options(tree_root, options); |
2748 | if (ret) { | |
2749 | err = ret; | |
16cdcec7 | 2750 | goto fail_alloc; |
2b82032c | 2751 | } |
dfe25020 | 2752 | |
f2b636e8 JB |
2753 | features = btrfs_super_incompat_flags(disk_super) & |
2754 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2755 | if (features) { | |
2756 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2757 | "unsupported optional features (%Lx).\n", | |
c1c9ff7c | 2758 | features); |
f2b636e8 | 2759 | err = -EINVAL; |
16cdcec7 | 2760 | goto fail_alloc; |
f2b636e8 JB |
2761 | } |
2762 | ||
707e8a07 DS |
2763 | /* |
2764 | * Leafsize and nodesize were always equal, this is only a sanity check. | |
2765 | */ | |
2766 | if (le32_to_cpu(disk_super->__unused_leafsize) != | |
727011e0 CM |
2767 | btrfs_super_nodesize(disk_super)) { |
2768 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2769 | "blocksizes don't match. node %d leaf %d\n", | |
2770 | btrfs_super_nodesize(disk_super), | |
707e8a07 | 2771 | le32_to_cpu(disk_super->__unused_leafsize)); |
727011e0 CM |
2772 | err = -EINVAL; |
2773 | goto fail_alloc; | |
2774 | } | |
707e8a07 | 2775 | if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { |
727011e0 CM |
2776 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " |
2777 | "blocksize (%d) was too large\n", | |
707e8a07 | 2778 | btrfs_super_nodesize(disk_super)); |
727011e0 CM |
2779 | err = -EINVAL; |
2780 | goto fail_alloc; | |
2781 | } | |
2782 | ||
5d4f98a2 | 2783 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2784 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2785 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2786 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2787 | |
3173a18f | 2788 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
5efa0490 | 2789 | printk(KERN_INFO "BTRFS: has skinny extents\n"); |
3173a18f | 2790 | |
727011e0 CM |
2791 | /* |
2792 | * flag our filesystem as having big metadata blocks if | |
2793 | * they are bigger than the page size | |
2794 | */ | |
707e8a07 | 2795 | if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) { |
727011e0 | 2796 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
efe120a0 | 2797 | printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); |
727011e0 CM |
2798 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2799 | } | |
2800 | ||
bc3f116f | 2801 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f CM |
2802 | sectorsize = btrfs_super_sectorsize(disk_super); |
2803 | stripesize = btrfs_super_stripesize(disk_super); | |
707e8a07 | 2804 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2805 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2806 | |
2807 | /* | |
2808 | * mixed block groups end up with duplicate but slightly offset | |
2809 | * extent buffers for the same range. It leads to corruptions | |
2810 | */ | |
2811 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2812 | (sectorsize != nodesize)) { |
aa8ee312 | 2813 | printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes " |
bc3f116f CM |
2814 | "are not allowed for mixed block groups on %s\n", |
2815 | sb->s_id); | |
2816 | goto fail_alloc; | |
2817 | } | |
2818 | ||
ceda0864 MX |
2819 | /* |
2820 | * Needn't use the lock because there is no other task which will | |
2821 | * update the flag. | |
2822 | */ | |
a6fa6fae | 2823 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2824 | |
f2b636e8 JB |
2825 | features = btrfs_super_compat_ro_flags(disk_super) & |
2826 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2827 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2828 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2829 | "unsupported option features (%Lx).\n", | |
c1c9ff7c | 2830 | features); |
f2b636e8 | 2831 | err = -EINVAL; |
16cdcec7 | 2832 | goto fail_alloc; |
f2b636e8 | 2833 | } |
61d92c32 | 2834 | |
5cdc7ad3 | 2835 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2836 | |
2a458198 ES |
2837 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2838 | if (ret) { | |
2839 | err = ret; | |
0dc3b84a JB |
2840 | goto fail_sb_buffer; |
2841 | } | |
4543df7e | 2842 | |
4575c9cc | 2843 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 | 2844 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
ee22184b | 2845 | SZ_4M / PAGE_CACHE_SIZE); |
4575c9cc | 2846 | |
db94535d | 2847 | tree_root->nodesize = nodesize; |
db94535d | 2848 | tree_root->sectorsize = sectorsize; |
87ee04eb | 2849 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2850 | |
2851 | sb->s_blocksize = sectorsize; | |
2852 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2853 | |
3cae210f | 2854 | if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { |
aa8ee312 | 2855 | printk(KERN_ERR "BTRFS: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2856 | goto fail_sb_buffer; |
2857 | } | |
19c00ddc | 2858 | |
8d082fb7 | 2859 | if (sectorsize != PAGE_SIZE) { |
aa8ee312 | 2860 | printk(KERN_ERR "BTRFS: incompatible sector size (%lu) " |
8d082fb7 | 2861 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); |
941b2ddf KM |
2862 | goto fail_sb_buffer; |
2863 | } | |
2864 | ||
925baedd | 2865 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2866 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2867 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2868 | if (ret) { |
aa8ee312 | 2869 | printk(KERN_ERR "BTRFS: failed to read the system " |
d397712b | 2870 | "array on %s\n", sb->s_id); |
5d4f98a2 | 2871 | goto fail_sb_buffer; |
84eed90f | 2872 | } |
0b86a832 | 2873 | |
84234f3a | 2874 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2875 | |
707e8a07 DS |
2876 | __setup_root(nodesize, sectorsize, stripesize, chunk_root, |
2877 | fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
0b86a832 CM |
2878 | |
2879 | chunk_root->node = read_tree_block(chunk_root, | |
2880 | btrfs_super_chunk_root(disk_super), | |
ce86cd59 | 2881 | generation); |
64c043de LB |
2882 | if (IS_ERR(chunk_root->node) || |
2883 | !extent_buffer_uptodate(chunk_root->node)) { | |
aa8ee312 | 2884 | printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n", |
83121942 | 2885 | sb->s_id); |
e5fffbac | 2886 | if (!IS_ERR(chunk_root->node)) |
2887 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2888 | chunk_root->node = NULL; |
af31f5e5 | 2889 | goto fail_tree_roots; |
83121942 | 2890 | } |
5d4f98a2 YZ |
2891 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2892 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2893 | |
e17cade2 | 2894 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2895 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2896 | |
0b86a832 | 2897 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2898 | if (ret) { |
aa8ee312 | 2899 | printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n", |
d397712b | 2900 | sb->s_id); |
af31f5e5 | 2901 | goto fail_tree_roots; |
2b82032c | 2902 | } |
0b86a832 | 2903 | |
8dabb742 SB |
2904 | /* |
2905 | * keep the device that is marked to be the target device for the | |
2906 | * dev_replace procedure | |
2907 | */ | |
9eaed21e | 2908 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2909 | |
a6b0d5c8 | 2910 | if (!fs_devices->latest_bdev) { |
aa8ee312 | 2911 | printk(KERN_ERR "BTRFS: failed to read devices on %s\n", |
a6b0d5c8 CM |
2912 | sb->s_id); |
2913 | goto fail_tree_roots; | |
2914 | } | |
2915 | ||
af31f5e5 | 2916 | retry_root_backup: |
84234f3a | 2917 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2918 | |
e20d96d6 | 2919 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2920 | btrfs_super_root(disk_super), |
ce86cd59 | 2921 | generation); |
64c043de LB |
2922 | if (IS_ERR(tree_root->node) || |
2923 | !extent_buffer_uptodate(tree_root->node)) { | |
efe120a0 | 2924 | printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", |
83121942 | 2925 | sb->s_id); |
e5fffbac | 2926 | if (!IS_ERR(tree_root->node)) |
2927 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2928 | tree_root->node = NULL; |
af31f5e5 | 2929 | goto recovery_tree_root; |
83121942 | 2930 | } |
af31f5e5 | 2931 | |
5d4f98a2 YZ |
2932 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2933 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2934 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2935 | |
f32e48e9 CR |
2936 | mutex_lock(&tree_root->objectid_mutex); |
2937 | ret = btrfs_find_highest_objectid(tree_root, | |
2938 | &tree_root->highest_objectid); | |
2939 | if (ret) { | |
2940 | mutex_unlock(&tree_root->objectid_mutex); | |
2941 | goto recovery_tree_root; | |
2942 | } | |
2943 | ||
2944 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2945 | ||
2946 | mutex_unlock(&tree_root->objectid_mutex); | |
2947 | ||
4bbcaa64 ES |
2948 | ret = btrfs_read_roots(fs_info, tree_root); |
2949 | if (ret) | |
af31f5e5 | 2950 | goto recovery_tree_root; |
f7a81ea4 | 2951 | |
8929ecfa YZ |
2952 | fs_info->generation = generation; |
2953 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2954 | |
68310a5e ID |
2955 | ret = btrfs_recover_balance(fs_info); |
2956 | if (ret) { | |
aa8ee312 | 2957 | printk(KERN_ERR "BTRFS: failed to recover balance\n"); |
68310a5e ID |
2958 | goto fail_block_groups; |
2959 | } | |
2960 | ||
733f4fbb SB |
2961 | ret = btrfs_init_dev_stats(fs_info); |
2962 | if (ret) { | |
efe120a0 | 2963 | printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", |
733f4fbb SB |
2964 | ret); |
2965 | goto fail_block_groups; | |
2966 | } | |
2967 | ||
8dabb742 SB |
2968 | ret = btrfs_init_dev_replace(fs_info); |
2969 | if (ret) { | |
efe120a0 | 2970 | pr_err("BTRFS: failed to init dev_replace: %d\n", ret); |
8dabb742 SB |
2971 | goto fail_block_groups; |
2972 | } | |
2973 | ||
9eaed21e | 2974 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 2975 | |
b7c35e81 AJ |
2976 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
2977 | if (ret) { | |
2978 | pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret); | |
2979 | goto fail_block_groups; | |
2980 | } | |
2981 | ||
2982 | ret = btrfs_sysfs_add_device(fs_devices); | |
2983 | if (ret) { | |
2984 | pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret); | |
2985 | goto fail_fsdev_sysfs; | |
2986 | } | |
2987 | ||
96f3136e | 2988 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 2989 | if (ret) { |
efe120a0 | 2990 | pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); |
b7c35e81 | 2991 | goto fail_fsdev_sysfs; |
c59021f8 | 2992 | } |
2993 | ||
c59021f8 | 2994 | ret = btrfs_init_space_info(fs_info); |
2995 | if (ret) { | |
efe120a0 | 2996 | printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); |
2365dd3c | 2997 | goto fail_sysfs; |
c59021f8 | 2998 | } |
2999 | ||
4bbcaa64 | 3000 | ret = btrfs_read_block_groups(fs_info->extent_root); |
1b1d1f66 | 3001 | if (ret) { |
efe120a0 | 3002 | printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); |
2365dd3c | 3003 | goto fail_sysfs; |
1b1d1f66 | 3004 | } |
5af3e8cc SB |
3005 | fs_info->num_tolerated_disk_barrier_failures = |
3006 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
3007 | if (fs_info->fs_devices->missing_devices > |
3008 | fs_info->num_tolerated_disk_barrier_failures && | |
3009 | !(sb->s_flags & MS_RDONLY)) { | |
78fa1770 ZL |
3010 | pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n", |
3011 | fs_info->fs_devices->missing_devices, | |
3012 | fs_info->num_tolerated_disk_barrier_failures); | |
2365dd3c | 3013 | goto fail_sysfs; |
292fd7fc | 3014 | } |
9078a3e1 | 3015 | |
a74a4b97 CM |
3016 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
3017 | "btrfs-cleaner"); | |
57506d50 | 3018 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3019 | goto fail_sysfs; |
a74a4b97 CM |
3020 | |
3021 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3022 | tree_root, | |
3023 | "btrfs-transaction"); | |
57506d50 | 3024 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3025 | goto fail_cleaner; |
a74a4b97 | 3026 | |
c289811c CM |
3027 | if (!btrfs_test_opt(tree_root, SSD) && |
3028 | !btrfs_test_opt(tree_root, NOSSD) && | |
3029 | !fs_info->fs_devices->rotating) { | |
efe120a0 | 3030 | printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " |
c289811c CM |
3031 | "mode\n"); |
3032 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
3033 | } | |
3034 | ||
572d9ab7 DS |
3035 | /* |
3036 | * Mount does not set all options immediatelly, we can do it now and do | |
3037 | * not have to wait for transaction commit | |
3038 | */ | |
3039 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3040 | |
21adbd5c SB |
3041 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3042 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
3043 | ret = btrfsic_mount(tree_root, fs_devices, | |
3044 | btrfs_test_opt(tree_root, | |
3045 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
3046 | 1 : 0, | |
3047 | fs_info->check_integrity_print_mask); | |
3048 | if (ret) | |
efe120a0 | 3049 | printk(KERN_WARNING "BTRFS: failed to initialize" |
21adbd5c SB |
3050 | " integrity check module %s\n", sb->s_id); |
3051 | } | |
3052 | #endif | |
bcef60f2 AJ |
3053 | ret = btrfs_read_qgroup_config(fs_info); |
3054 | if (ret) | |
3055 | goto fail_trans_kthread; | |
21adbd5c | 3056 | |
acce952b | 3057 | /* do not make disk changes in broken FS */ |
68ce9682 | 3058 | if (btrfs_super_log_root(disk_super) != 0) { |
63443bf5 | 3059 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3060 | if (ret) { |
63443bf5 | 3061 | err = ret; |
28c16cbb | 3062 | goto fail_qgroup; |
79787eaa | 3063 | } |
e02119d5 | 3064 | } |
1a40e23b | 3065 | |
76dda93c | 3066 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa | 3067 | if (ret) |
28c16cbb | 3068 | goto fail_qgroup; |
76dda93c | 3069 | |
7c2ca468 | 3070 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 3071 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 3072 | if (ret) |
28c16cbb | 3073 | goto fail_qgroup; |
d68fc57b | 3074 | |
5f316481 | 3075 | mutex_lock(&fs_info->cleaner_mutex); |
5d4f98a2 | 3076 | ret = btrfs_recover_relocation(tree_root); |
5f316481 | 3077 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 MX |
3078 | if (ret < 0) { |
3079 | printk(KERN_WARNING | |
efe120a0 | 3080 | "BTRFS: failed to recover relocation\n"); |
d7ce5843 | 3081 | err = -EINVAL; |
bcef60f2 | 3082 | goto fail_qgroup; |
d7ce5843 | 3083 | } |
7c2ca468 | 3084 | } |
1a40e23b | 3085 | |
3de4586c CM |
3086 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
3087 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 3088 | location.offset = 0; |
3de4586c | 3089 | |
3de4586c | 3090 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
3091 | if (IS_ERR(fs_info->fs_root)) { |
3092 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 3093 | goto fail_qgroup; |
3140c9a3 | 3094 | } |
c289811c | 3095 | |
2b6ba629 ID |
3096 | if (sb->s_flags & MS_RDONLY) |
3097 | return 0; | |
59641015 | 3098 | |
511711af CM |
3099 | if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) && |
3100 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3101 | pr_info("BTRFS: creating free space tree\n"); | |
3102 | ret = btrfs_create_free_space_tree(fs_info); | |
3103 | if (ret) { | |
3104 | pr_warn("BTRFS: failed to create free space tree %d\n", | |
3105 | ret); | |
3106 | close_ctree(tree_root); | |
3107 | return ret; | |
3108 | } | |
3109 | } | |
3110 | ||
2b6ba629 ID |
3111 | down_read(&fs_info->cleanup_work_sem); |
3112 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3113 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3114 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
3115 | close_ctree(tree_root); |
3116 | return ret; | |
3117 | } | |
3118 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3119 | |
2b6ba629 ID |
3120 | ret = btrfs_resume_balance_async(fs_info); |
3121 | if (ret) { | |
efe120a0 | 3122 | printk(KERN_WARNING "BTRFS: failed to resume balance\n"); |
2b6ba629 ID |
3123 | close_ctree(tree_root); |
3124 | return ret; | |
e3acc2a6 JB |
3125 | } |
3126 | ||
8dabb742 SB |
3127 | ret = btrfs_resume_dev_replace_async(fs_info); |
3128 | if (ret) { | |
efe120a0 | 3129 | pr_warn("BTRFS: failed to resume dev_replace\n"); |
8dabb742 SB |
3130 | close_ctree(tree_root); |
3131 | return ret; | |
3132 | } | |
3133 | ||
b382a324 JS |
3134 | btrfs_qgroup_rescan_resume(fs_info); |
3135 | ||
70f6d82e OS |
3136 | if (btrfs_test_opt(tree_root, CLEAR_CACHE) && |
3137 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3138 | pr_info("BTRFS: clearing free space tree\n"); | |
3139 | ret = btrfs_clear_free_space_tree(fs_info); | |
3140 | if (ret) { | |
3141 | pr_warn("BTRFS: failed to clear free space tree %d\n", | |
3142 | ret); | |
3143 | close_ctree(tree_root); | |
3144 | return ret; | |
3145 | } | |
3146 | } | |
3147 | ||
4bbcaa64 | 3148 | if (!fs_info->uuid_root) { |
efe120a0 | 3149 | pr_info("BTRFS: creating UUID tree\n"); |
f7a81ea4 SB |
3150 | ret = btrfs_create_uuid_tree(fs_info); |
3151 | if (ret) { | |
efe120a0 | 3152 | pr_warn("BTRFS: failed to create the UUID tree %d\n", |
f7a81ea4 SB |
3153 | ret); |
3154 | close_ctree(tree_root); | |
3155 | return ret; | |
3156 | } | |
4bbcaa64 ES |
3157 | } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) || |
3158 | fs_info->generation != | |
3159 | btrfs_super_uuid_tree_generation(disk_super)) { | |
efe120a0 | 3160 | pr_info("BTRFS: checking UUID tree\n"); |
70f80175 SB |
3161 | ret = btrfs_check_uuid_tree(fs_info); |
3162 | if (ret) { | |
efe120a0 | 3163 | pr_warn("BTRFS: failed to check the UUID tree %d\n", |
70f80175 SB |
3164 | ret); |
3165 | close_ctree(tree_root); | |
3166 | return ret; | |
3167 | } | |
3168 | } else { | |
3169 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
3170 | } |
3171 | ||
47ab2a6c JB |
3172 | fs_info->open = 1; |
3173 | ||
ad2b2c80 | 3174 | return 0; |
39279cc3 | 3175 | |
bcef60f2 AJ |
3176 | fail_qgroup: |
3177 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3178 | fail_trans_kthread: |
3179 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 3180 | btrfs_cleanup_transaction(fs_info->tree_root); |
faa2dbf0 | 3181 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3182 | fail_cleaner: |
a74a4b97 | 3183 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3184 | |
3185 | /* | |
3186 | * make sure we're done with the btree inode before we stop our | |
3187 | * kthreads | |
3188 | */ | |
3189 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3190 | |
2365dd3c | 3191 | fail_sysfs: |
6618a59b | 3192 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3193 | |
b7c35e81 AJ |
3194 | fail_fsdev_sysfs: |
3195 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3196 | ||
1b1d1f66 | 3197 | fail_block_groups: |
54067ae9 | 3198 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 3199 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
3200 | |
3201 | fail_tree_roots: | |
3202 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3203 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3204 | |
39279cc3 | 3205 | fail_sb_buffer: |
7abadb64 | 3206 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 3207 | fail_alloc: |
4543df7e | 3208 | fail_iput: |
586e46e2 ID |
3209 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3210 | ||
4543df7e | 3211 | iput(fs_info->btree_inode); |
c404e0dc MX |
3212 | fail_bio_counter: |
3213 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3214 | fail_delalloc_bytes: |
3215 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3216 | fail_dirty_metadata_bytes: |
3217 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 3218 | fail_bdi: |
7e662854 | 3219 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
3220 | fail_srcu: |
3221 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3222 | fail: |
53b381b3 | 3223 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3224 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3225 | return err; |
af31f5e5 CM |
3226 | |
3227 | recovery_tree_root: | |
af31f5e5 CM |
3228 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
3229 | goto fail_tree_roots; | |
3230 | ||
3231 | free_root_pointers(fs_info, 0); | |
3232 | ||
3233 | /* don't use the log in recovery mode, it won't be valid */ | |
3234 | btrfs_set_super_log_root(disk_super, 0); | |
3235 | ||
3236 | /* we can't trust the free space cache either */ | |
3237 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3238 | ||
3239 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3240 | &num_backups_tried, &backup_index); | |
3241 | if (ret == -1) | |
3242 | goto fail_block_groups; | |
3243 | goto retry_root_backup; | |
eb60ceac CM |
3244 | } |
3245 | ||
f2984462 CM |
3246 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3247 | { | |
f2984462 CM |
3248 | if (uptodate) { |
3249 | set_buffer_uptodate(bh); | |
3250 | } else { | |
442a4f63 SB |
3251 | struct btrfs_device *device = (struct btrfs_device *) |
3252 | bh->b_private; | |
3253 | ||
b14af3b4 DS |
3254 | btrfs_warn_rl_in_rcu(device->dev_root->fs_info, |
3255 | "lost page write due to IO error on %s", | |
606686ee | 3256 | rcu_str_deref(device->name)); |
1259ab75 CM |
3257 | /* note, we dont' set_buffer_write_io_error because we have |
3258 | * our own ways of dealing with the IO errors | |
3259 | */ | |
f2984462 | 3260 | clear_buffer_uptodate(bh); |
442a4f63 | 3261 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3262 | } |
3263 | unlock_buffer(bh); | |
3264 | put_bh(bh); | |
3265 | } | |
3266 | ||
29c36d72 AJ |
3267 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3268 | struct buffer_head **bh_ret) | |
3269 | { | |
3270 | struct buffer_head *bh; | |
3271 | struct btrfs_super_block *super; | |
3272 | u64 bytenr; | |
3273 | ||
3274 | bytenr = btrfs_sb_offset(copy_num); | |
3275 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3276 | return -EINVAL; | |
3277 | ||
3278 | bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE); | |
3279 | /* | |
3280 | * If we fail to read from the underlying devices, as of now | |
3281 | * the best option we have is to mark it EIO. | |
3282 | */ | |
3283 | if (!bh) | |
3284 | return -EIO; | |
3285 | ||
3286 | super = (struct btrfs_super_block *)bh->b_data; | |
3287 | if (btrfs_super_bytenr(super) != bytenr || | |
3288 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3289 | brelse(bh); | |
3290 | return -EINVAL; | |
3291 | } | |
3292 | ||
3293 | *bh_ret = bh; | |
3294 | return 0; | |
3295 | } | |
3296 | ||
3297 | ||
a512bbf8 YZ |
3298 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3299 | { | |
3300 | struct buffer_head *bh; | |
3301 | struct buffer_head *latest = NULL; | |
3302 | struct btrfs_super_block *super; | |
3303 | int i; | |
3304 | u64 transid = 0; | |
92fc03fb | 3305 | int ret = -EINVAL; |
a512bbf8 YZ |
3306 | |
3307 | /* we would like to check all the supers, but that would make | |
3308 | * a btrfs mount succeed after a mkfs from a different FS. | |
3309 | * So, we need to add a special mount option to scan for | |
3310 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3311 | */ | |
3312 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3313 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3314 | if (ret) | |
a512bbf8 YZ |
3315 | continue; |
3316 | ||
3317 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3318 | |
3319 | if (!latest || btrfs_super_generation(super) > transid) { | |
3320 | brelse(latest); | |
3321 | latest = bh; | |
3322 | transid = btrfs_super_generation(super); | |
3323 | } else { | |
3324 | brelse(bh); | |
3325 | } | |
3326 | } | |
92fc03fb AJ |
3327 | |
3328 | if (!latest) | |
3329 | return ERR_PTR(ret); | |
3330 | ||
a512bbf8 YZ |
3331 | return latest; |
3332 | } | |
3333 | ||
4eedeb75 HH |
3334 | /* |
3335 | * this should be called twice, once with wait == 0 and | |
3336 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3337 | * we write are pinned. | |
3338 | * | |
3339 | * They are released when wait == 1 is done. | |
3340 | * max_mirrors must be the same for both runs, and it indicates how | |
3341 | * many supers on this one device should be written. | |
3342 | * | |
3343 | * max_mirrors == 0 means to write them all. | |
3344 | */ | |
a512bbf8 YZ |
3345 | static int write_dev_supers(struct btrfs_device *device, |
3346 | struct btrfs_super_block *sb, | |
3347 | int do_barriers, int wait, int max_mirrors) | |
3348 | { | |
3349 | struct buffer_head *bh; | |
3350 | int i; | |
3351 | int ret; | |
3352 | int errors = 0; | |
3353 | u32 crc; | |
3354 | u64 bytenr; | |
a512bbf8 YZ |
3355 | |
3356 | if (max_mirrors == 0) | |
3357 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3358 | ||
a512bbf8 YZ |
3359 | for (i = 0; i < max_mirrors; i++) { |
3360 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3361 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3362 | device->commit_total_bytes) | |
a512bbf8 YZ |
3363 | break; |
3364 | ||
3365 | if (wait) { | |
3366 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3367 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3368 | if (!bh) { |
3369 | errors++; | |
3370 | continue; | |
3371 | } | |
a512bbf8 | 3372 | wait_on_buffer(bh); |
4eedeb75 HH |
3373 | if (!buffer_uptodate(bh)) |
3374 | errors++; | |
3375 | ||
3376 | /* drop our reference */ | |
3377 | brelse(bh); | |
3378 | ||
3379 | /* drop the reference from the wait == 0 run */ | |
3380 | brelse(bh); | |
3381 | continue; | |
a512bbf8 YZ |
3382 | } else { |
3383 | btrfs_set_super_bytenr(sb, bytenr); | |
3384 | ||
3385 | crc = ~(u32)0; | |
b0496686 | 3386 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3387 | BTRFS_CSUM_SIZE, crc, |
3388 | BTRFS_SUPER_INFO_SIZE - | |
3389 | BTRFS_CSUM_SIZE); | |
3390 | btrfs_csum_final(crc, sb->csum); | |
3391 | ||
4eedeb75 HH |
3392 | /* |
3393 | * one reference for us, and we leave it for the | |
3394 | * caller | |
3395 | */ | |
a512bbf8 YZ |
3396 | bh = __getblk(device->bdev, bytenr / 4096, |
3397 | BTRFS_SUPER_INFO_SIZE); | |
634554dc | 3398 | if (!bh) { |
f14d104d DS |
3399 | btrfs_err(device->dev_root->fs_info, |
3400 | "couldn't get super buffer head for bytenr %llu", | |
3401 | bytenr); | |
634554dc JB |
3402 | errors++; |
3403 | continue; | |
3404 | } | |
3405 | ||
a512bbf8 YZ |
3406 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3407 | ||
4eedeb75 | 3408 | /* one reference for submit_bh */ |
a512bbf8 | 3409 | get_bh(bh); |
4eedeb75 HH |
3410 | |
3411 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3412 | lock_buffer(bh); |
3413 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3414 | bh->b_private = device; |
a512bbf8 YZ |
3415 | } |
3416 | ||
387125fc CM |
3417 | /* |
3418 | * we fua the first super. The others we allow | |
3419 | * to go down lazy. | |
3420 | */ | |
e8117c26 WS |
3421 | if (i == 0) |
3422 | ret = btrfsic_submit_bh(WRITE_FUA, bh); | |
3423 | else | |
3424 | ret = btrfsic_submit_bh(WRITE_SYNC, bh); | |
4eedeb75 | 3425 | if (ret) |
a512bbf8 | 3426 | errors++; |
a512bbf8 YZ |
3427 | } |
3428 | return errors < i ? 0 : -1; | |
3429 | } | |
3430 | ||
387125fc CM |
3431 | /* |
3432 | * endio for the write_dev_flush, this will wake anyone waiting | |
3433 | * for the barrier when it is done | |
3434 | */ | |
4246a0b6 | 3435 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3436 | { |
387125fc CM |
3437 | if (bio->bi_private) |
3438 | complete(bio->bi_private); | |
3439 | bio_put(bio); | |
3440 | } | |
3441 | ||
3442 | /* | |
3443 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3444 | * sent down. With wait == 1, it waits for the previous flush. | |
3445 | * | |
3446 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3447 | * capable | |
3448 | */ | |
3449 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3450 | { | |
3451 | struct bio *bio; | |
3452 | int ret = 0; | |
3453 | ||
3454 | if (device->nobarriers) | |
3455 | return 0; | |
3456 | ||
3457 | if (wait) { | |
3458 | bio = device->flush_bio; | |
3459 | if (!bio) | |
3460 | return 0; | |
3461 | ||
3462 | wait_for_completion(&device->flush_wait); | |
3463 | ||
4246a0b6 CH |
3464 | if (bio->bi_error) { |
3465 | ret = bio->bi_error; | |
5af3e8cc SB |
3466 | btrfs_dev_stat_inc_and_print(device, |
3467 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3468 | } |
3469 | ||
3470 | /* drop the reference from the wait == 0 run */ | |
3471 | bio_put(bio); | |
3472 | device->flush_bio = NULL; | |
3473 | ||
3474 | return ret; | |
3475 | } | |
3476 | ||
3477 | /* | |
3478 | * one reference for us, and we leave it for the | |
3479 | * caller | |
3480 | */ | |
9c017abc | 3481 | device->flush_bio = NULL; |
9be3395b | 3482 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3483 | if (!bio) |
3484 | return -ENOMEM; | |
3485 | ||
3486 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3487 | bio->bi_bdev = device->bdev; | |
3488 | init_completion(&device->flush_wait); | |
3489 | bio->bi_private = &device->flush_wait; | |
3490 | device->flush_bio = bio; | |
3491 | ||
3492 | bio_get(bio); | |
21adbd5c | 3493 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3494 | |
3495 | return 0; | |
3496 | } | |
3497 | ||
3498 | /* | |
3499 | * send an empty flush down to each device in parallel, | |
3500 | * then wait for them | |
3501 | */ | |
3502 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3503 | { | |
3504 | struct list_head *head; | |
3505 | struct btrfs_device *dev; | |
5af3e8cc SB |
3506 | int errors_send = 0; |
3507 | int errors_wait = 0; | |
387125fc CM |
3508 | int ret; |
3509 | ||
3510 | /* send down all the barriers */ | |
3511 | head = &info->fs_devices->devices; | |
3512 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3513 | if (dev->missing) |
3514 | continue; | |
387125fc | 3515 | if (!dev->bdev) { |
5af3e8cc | 3516 | errors_send++; |
387125fc CM |
3517 | continue; |
3518 | } | |
3519 | if (!dev->in_fs_metadata || !dev->writeable) | |
3520 | continue; | |
3521 | ||
3522 | ret = write_dev_flush(dev, 0); | |
3523 | if (ret) | |
5af3e8cc | 3524 | errors_send++; |
387125fc CM |
3525 | } |
3526 | ||
3527 | /* wait for all the barriers */ | |
3528 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3529 | if (dev->missing) |
3530 | continue; | |
387125fc | 3531 | if (!dev->bdev) { |
5af3e8cc | 3532 | errors_wait++; |
387125fc CM |
3533 | continue; |
3534 | } | |
3535 | if (!dev->in_fs_metadata || !dev->writeable) | |
3536 | continue; | |
3537 | ||
3538 | ret = write_dev_flush(dev, 1); | |
3539 | if (ret) | |
5af3e8cc | 3540 | errors_wait++; |
387125fc | 3541 | } |
5af3e8cc SB |
3542 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3543 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3544 | return -EIO; |
3545 | return 0; | |
3546 | } | |
3547 | ||
943c6e99 ZL |
3548 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3549 | { | |
8789f4fe ZL |
3550 | int raid_type; |
3551 | int min_tolerated = INT_MAX; | |
943c6e99 | 3552 | |
8789f4fe ZL |
3553 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3554 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3555 | min_tolerated = min(min_tolerated, | |
3556 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3557 | tolerated_failures); | |
943c6e99 | 3558 | |
8789f4fe ZL |
3559 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3560 | if (raid_type == BTRFS_RAID_SINGLE) | |
3561 | continue; | |
3562 | if (!(flags & btrfs_raid_group[raid_type])) | |
3563 | continue; | |
3564 | min_tolerated = min(min_tolerated, | |
3565 | btrfs_raid_array[raid_type]. | |
3566 | tolerated_failures); | |
3567 | } | |
943c6e99 | 3568 | |
8789f4fe ZL |
3569 | if (min_tolerated == INT_MAX) { |
3570 | pr_warn("BTRFS: unknown raid flag: %llu\n", flags); | |
3571 | min_tolerated = 0; | |
3572 | } | |
3573 | ||
3574 | return min_tolerated; | |
943c6e99 ZL |
3575 | } |
3576 | ||
5af3e8cc SB |
3577 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3578 | struct btrfs_fs_info *fs_info) | |
3579 | { | |
3580 | struct btrfs_ioctl_space_info space; | |
3581 | struct btrfs_space_info *sinfo; | |
3582 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3583 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3584 | BTRFS_BLOCK_GROUP_METADATA, | |
3585 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
5af3e8cc SB |
3586 | int i; |
3587 | int c; | |
3588 | int num_tolerated_disk_barrier_failures = | |
3589 | (int)fs_info->fs_devices->num_devices; | |
3590 | ||
2c458045 | 3591 | for (i = 0; i < ARRAY_SIZE(types); i++) { |
5af3e8cc SB |
3592 | struct btrfs_space_info *tmp; |
3593 | ||
3594 | sinfo = NULL; | |
3595 | rcu_read_lock(); | |
3596 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3597 | if (tmp->flags == types[i]) { | |
3598 | sinfo = tmp; | |
3599 | break; | |
3600 | } | |
3601 | } | |
3602 | rcu_read_unlock(); | |
3603 | ||
3604 | if (!sinfo) | |
3605 | continue; | |
3606 | ||
3607 | down_read(&sinfo->groups_sem); | |
3608 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
2c458045 ZL |
3609 | u64 flags; |
3610 | ||
3611 | if (list_empty(&sinfo->block_groups[c])) | |
3612 | continue; | |
3613 | ||
3614 | btrfs_get_block_group_info(&sinfo->block_groups[c], | |
3615 | &space); | |
3616 | if (space.total_bytes == 0 || space.used_bytes == 0) | |
3617 | continue; | |
3618 | flags = space.flags; | |
943c6e99 ZL |
3619 | |
3620 | num_tolerated_disk_barrier_failures = min( | |
3621 | num_tolerated_disk_barrier_failures, | |
3622 | btrfs_get_num_tolerated_disk_barrier_failures( | |
3623 | flags)); | |
5af3e8cc SB |
3624 | } |
3625 | up_read(&sinfo->groups_sem); | |
3626 | } | |
3627 | ||
3628 | return num_tolerated_disk_barrier_failures; | |
3629 | } | |
3630 | ||
48a3b636 | 3631 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3632 | { |
e5e9a520 | 3633 | struct list_head *head; |
f2984462 | 3634 | struct btrfs_device *dev; |
a061fc8d | 3635 | struct btrfs_super_block *sb; |
f2984462 | 3636 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3637 | int ret; |
3638 | int do_barriers; | |
a236aed1 CM |
3639 | int max_errors; |
3640 | int total_errors = 0; | |
a061fc8d | 3641 | u64 flags; |
f2984462 CM |
3642 | |
3643 | do_barriers = !btrfs_test_opt(root, NOBARRIER); | |
af31f5e5 | 3644 | backup_super_roots(root->fs_info); |
f2984462 | 3645 | |
6c41761f | 3646 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3647 | dev_item = &sb->dev_item; |
e5e9a520 | 3648 | |
174ba509 | 3649 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3650 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3651 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3652 | |
5af3e8cc SB |
3653 | if (do_barriers) { |
3654 | ret = barrier_all_devices(root->fs_info); | |
3655 | if (ret) { | |
3656 | mutex_unlock( | |
3657 | &root->fs_info->fs_devices->device_list_mutex); | |
a4553fef | 3658 | btrfs_std_error(root->fs_info, ret, |
5af3e8cc SB |
3659 | "errors while submitting device barriers."); |
3660 | return ret; | |
3661 | } | |
3662 | } | |
387125fc | 3663 | |
1f78160c | 3664 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3665 | if (!dev->bdev) { |
3666 | total_errors++; | |
3667 | continue; | |
3668 | } | |
2b82032c | 3669 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3670 | continue; |
3671 | ||
2b82032c | 3672 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3673 | btrfs_set_stack_device_type(dev_item, dev->type); |
3674 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3675 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3676 | dev->commit_total_bytes); |
ce7213c7 MX |
3677 | btrfs_set_stack_device_bytes_used(dev_item, |
3678 | dev->commit_bytes_used); | |
a061fc8d CM |
3679 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3680 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3681 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3682 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3683 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3684 | |
a061fc8d CM |
3685 | flags = btrfs_super_flags(sb); |
3686 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3687 | ||
a512bbf8 | 3688 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3689 | if (ret) |
3690 | total_errors++; | |
f2984462 | 3691 | } |
a236aed1 | 3692 | if (total_errors > max_errors) { |
efe120a0 | 3693 | btrfs_err(root->fs_info, "%d errors while writing supers", |
d397712b | 3694 | total_errors); |
a724b436 | 3695 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3696 | |
9d565ba4 | 3697 | /* FUA is masked off if unsupported and can't be the reason */ |
a4553fef | 3698 | btrfs_std_error(root->fs_info, -EIO, |
9d565ba4 SB |
3699 | "%d errors while writing supers", total_errors); |
3700 | return -EIO; | |
a236aed1 | 3701 | } |
f2984462 | 3702 | |
a512bbf8 | 3703 | total_errors = 0; |
1f78160c | 3704 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3705 | if (!dev->bdev) |
3706 | continue; | |
2b82032c | 3707 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3708 | continue; |
3709 | ||
a512bbf8 YZ |
3710 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3711 | if (ret) | |
3712 | total_errors++; | |
f2984462 | 3713 | } |
174ba509 | 3714 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3715 | if (total_errors > max_errors) { |
a4553fef | 3716 | btrfs_std_error(root->fs_info, -EIO, |
79787eaa JM |
3717 | "%d errors while writing supers", total_errors); |
3718 | return -EIO; | |
a236aed1 | 3719 | } |
f2984462 CM |
3720 | return 0; |
3721 | } | |
3722 | ||
a512bbf8 YZ |
3723 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3724 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3725 | { |
f570e757 | 3726 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3727 | } |
3728 | ||
cb517eab MX |
3729 | /* Drop a fs root from the radix tree and free it. */ |
3730 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3731 | struct btrfs_root *root) | |
2619ba1f | 3732 | { |
4df27c4d | 3733 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3734 | radix_tree_delete(&fs_info->fs_roots_radix, |
3735 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3736 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3737 | |
3738 | if (btrfs_root_refs(&root->root_item) == 0) | |
3739 | synchronize_srcu(&fs_info->subvol_srcu); | |
3740 | ||
1a4319cc | 3741 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
3321719e | 3742 | btrfs_free_log(NULL, root); |
3321719e | 3743 | |
faa2dbf0 JB |
3744 | if (root->free_ino_pinned) |
3745 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3746 | if (root->free_ino_ctl) | |
3747 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3748 | free_fs_root(root); |
4df27c4d YZ |
3749 | } |
3750 | ||
3751 | static void free_fs_root(struct btrfs_root *root) | |
3752 | { | |
57cdc8db | 3753 | iput(root->ino_cache_inode); |
4df27c4d | 3754 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3755 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3756 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3757 | if (root->anon_dev) |
3758 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3759 | if (root->subv_writers) |
3760 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3761 | free_extent_buffer(root->node); |
3762 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3763 | kfree(root->free_ino_ctl); |
3764 | kfree(root->free_ino_pinned); | |
d397712b | 3765 | kfree(root->name); |
b0feb9d9 | 3766 | btrfs_put_fs_root(root); |
2619ba1f CM |
3767 | } |
3768 | ||
cb517eab MX |
3769 | void btrfs_free_fs_root(struct btrfs_root *root) |
3770 | { | |
3771 | free_fs_root(root); | |
2619ba1f CM |
3772 | } |
3773 | ||
c146afad | 3774 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3775 | { |
c146afad YZ |
3776 | u64 root_objectid = 0; |
3777 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3778 | int i = 0; |
3779 | int err = 0; | |
3780 | unsigned int ret = 0; | |
3781 | int index; | |
e089f05c | 3782 | |
c146afad | 3783 | while (1) { |
65d33fd7 | 3784 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3785 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3786 | (void **)gang, root_objectid, | |
3787 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3788 | if (!ret) { |
3789 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3790 | break; |
65d33fd7 | 3791 | } |
5d4f98a2 | 3792 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3793 | |
c146afad | 3794 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3795 | /* Avoid to grab roots in dead_roots */ |
3796 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3797 | gang[i] = NULL; | |
3798 | continue; | |
3799 | } | |
3800 | /* grab all the search result for later use */ | |
3801 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3802 | } | |
3803 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3804 | |
65d33fd7 QW |
3805 | for (i = 0; i < ret; i++) { |
3806 | if (!gang[i]) | |
3807 | continue; | |
c146afad | 3808 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3809 | err = btrfs_orphan_cleanup(gang[i]); |
3810 | if (err) | |
65d33fd7 QW |
3811 | break; |
3812 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3813 | } |
3814 | root_objectid++; | |
3815 | } | |
65d33fd7 QW |
3816 | |
3817 | /* release the uncleaned roots due to error */ | |
3818 | for (; i < ret; i++) { | |
3819 | if (gang[i]) | |
3820 | btrfs_put_fs_root(gang[i]); | |
3821 | } | |
3822 | return err; | |
c146afad | 3823 | } |
a2135011 | 3824 | |
c146afad YZ |
3825 | int btrfs_commit_super(struct btrfs_root *root) |
3826 | { | |
3827 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3828 | |
c146afad | 3829 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3830 | btrfs_run_delayed_iputs(root); |
c146afad | 3831 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3832 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3833 | |
3834 | /* wait until ongoing cleanup work done */ | |
3835 | down_write(&root->fs_info->cleanup_work_sem); | |
3836 | up_write(&root->fs_info->cleanup_work_sem); | |
3837 | ||
7a7eaa40 | 3838 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3839 | if (IS_ERR(trans)) |
3840 | return PTR_ERR(trans); | |
d52c1bcc | 3841 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3842 | } |
3843 | ||
3abdbd78 | 3844 | void close_ctree(struct btrfs_root *root) |
c146afad YZ |
3845 | { |
3846 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3847 | int ret; | |
3848 | ||
3849 | fs_info->closing = 1; | |
3850 | smp_mb(); | |
3851 | ||
7343dd61 JM |
3852 | /* wait for the qgroup rescan worker to stop */ |
3853 | btrfs_qgroup_wait_for_completion(fs_info); | |
3854 | ||
803b2f54 SB |
3855 | /* wait for the uuid_scan task to finish */ |
3856 | down(&fs_info->uuid_tree_rescan_sem); | |
3857 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3858 | up(&fs_info->uuid_tree_rescan_sem); | |
3859 | ||
837d5b6e | 3860 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3861 | btrfs_pause_balance(fs_info); |
837d5b6e | 3862 | |
8dabb742 SB |
3863 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3864 | ||
aa1b8cd4 | 3865 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3866 | |
3867 | /* wait for any defraggers to finish */ | |
3868 | wait_event(fs_info->transaction_wait, | |
3869 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3870 | ||
3871 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3872 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3873 | |
21c7e756 MX |
3874 | cancel_work_sync(&fs_info->async_reclaim_work); |
3875 | ||
c146afad | 3876 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
e44163e1 JM |
3877 | /* |
3878 | * If the cleaner thread is stopped and there are | |
3879 | * block groups queued for removal, the deletion will be | |
3880 | * skipped when we quit the cleaner thread. | |
3881 | */ | |
e44163e1 | 3882 | btrfs_delete_unused_bgs(root->fs_info); |
e44163e1 | 3883 | |
acce952b | 3884 | ret = btrfs_commit_super(root); |
3885 | if (ret) | |
04892340 | 3886 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3887 | } |
3888 | ||
87533c47 | 3889 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3890 | btrfs_error_commit_super(root); |
0f7d52f4 | 3891 | |
e3029d9f AV |
3892 | kthread_stop(fs_info->transaction_kthread); |
3893 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3894 | |
f25784b3 YZ |
3895 | fs_info->closing = 2; |
3896 | smp_mb(); | |
3897 | ||
04892340 | 3898 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3899 | |
963d678b | 3900 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3901 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3902 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3903 | } |
bcc63abb | 3904 | |
6618a59b | 3905 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3906 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3907 | |
faa2dbf0 | 3908 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3909 | |
1a4319cc LB |
3910 | btrfs_put_block_group_cache(fs_info); |
3911 | ||
2b1360da JB |
3912 | btrfs_free_block_groups(fs_info); |
3913 | ||
de348ee0 WS |
3914 | /* |
3915 | * we must make sure there is not any read request to | |
3916 | * submit after we stopping all workers. | |
3917 | */ | |
3918 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3919 | btrfs_stop_all_workers(fs_info); |
3920 | ||
47ab2a6c | 3921 | fs_info->open = 0; |
13e6c37b | 3922 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3923 | |
13e6c37b | 3924 | iput(fs_info->btree_inode); |
d6bfde87 | 3925 | |
21adbd5c SB |
3926 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3927 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3928 | btrfsic_unmount(root, fs_info->fs_devices); | |
3929 | #endif | |
3930 | ||
dfe25020 | 3931 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3932 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3933 | |
e2d84521 | 3934 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3935 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3936 | percpu_counter_destroy(&fs_info->bio_counter); |
04160088 | 3937 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3938 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3939 | |
53b381b3 DW |
3940 | btrfs_free_stripe_hash_table(fs_info); |
3941 | ||
cdfb080e | 3942 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3943 | root->orphan_block_rsv = NULL; |
04216820 FM |
3944 | |
3945 | lock_chunks(root); | |
3946 | while (!list_empty(&fs_info->pinned_chunks)) { | |
3947 | struct extent_map *em; | |
3948 | ||
3949 | em = list_first_entry(&fs_info->pinned_chunks, | |
3950 | struct extent_map, list); | |
3951 | list_del_init(&em->list); | |
3952 | free_extent_map(em); | |
3953 | } | |
3954 | unlock_chunks(root); | |
eb60ceac CM |
3955 | } |
3956 | ||
b9fab919 CM |
3957 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3958 | int atomic) | |
5f39d397 | 3959 | { |
1259ab75 | 3960 | int ret; |
727011e0 | 3961 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3962 | |
0b32f4bb | 3963 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3964 | if (!ret) |
3965 | return ret; | |
3966 | ||
3967 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3968 | parent_transid, atomic); |
3969 | if (ret == -EAGAIN) | |
3970 | return ret; | |
1259ab75 | 3971 | return !ret; |
5f39d397 CM |
3972 | } |
3973 | ||
5f39d397 CM |
3974 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3975 | { | |
06ea65a3 | 3976 | struct btrfs_root *root; |
5f39d397 | 3977 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3978 | int was_dirty; |
b4ce94de | 3979 | |
06ea65a3 JB |
3980 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3981 | /* | |
3982 | * This is a fast path so only do this check if we have sanity tests | |
3983 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3984 | * outside of the sanity tests. | |
3985 | */ | |
3986 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3987 | return; | |
3988 | #endif | |
3989 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 3990 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3991 | if (transid != root->fs_info->generation) |
3992 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3993 | "found %llu running %llu\n", |
c1c9ff7c | 3994 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 3995 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3996 | if (!was_dirty) |
3997 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3998 | buf->len, | |
3999 | root->fs_info->dirty_metadata_batch); | |
1f21ef0a FM |
4000 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4001 | if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { | |
4002 | btrfs_print_leaf(root, buf); | |
4003 | ASSERT(0); | |
4004 | } | |
4005 | #endif | |
eb60ceac CM |
4006 | } |
4007 | ||
b53d3f5d LB |
4008 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
4009 | int flush_delayed) | |
16cdcec7 MX |
4010 | { |
4011 | /* | |
4012 | * looks as though older kernels can get into trouble with | |
4013 | * this code, they end up stuck in balance_dirty_pages forever | |
4014 | */ | |
e2d84521 | 4015 | int ret; |
16cdcec7 MX |
4016 | |
4017 | if (current->flags & PF_MEMALLOC) | |
4018 | return; | |
4019 | ||
b53d3f5d LB |
4020 | if (flush_delayed) |
4021 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 4022 | |
e2d84521 MX |
4023 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
4024 | BTRFS_DIRTY_METADATA_THRESH); | |
4025 | if (ret > 0) { | |
d0e1d66b NJ |
4026 | balance_dirty_pages_ratelimited( |
4027 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 | 4028 | } |
16cdcec7 MX |
4029 | } |
4030 | ||
b53d3f5d | 4031 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 4032 | { |
b53d3f5d LB |
4033 | __btrfs_btree_balance_dirty(root, 1); |
4034 | } | |
585ad2c3 | 4035 | |
b53d3f5d LB |
4036 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
4037 | { | |
4038 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 4039 | } |
6b80053d | 4040 | |
ca7a79ad | 4041 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 4042 | { |
727011e0 | 4043 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 4044 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 4045 | } |
0da5468f | 4046 | |
fcd1f065 | 4047 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 4048 | int read_only) |
4049 | { | |
c926093e DS |
4050 | struct btrfs_super_block *sb = fs_info->super_copy; |
4051 | int ret = 0; | |
4052 | ||
21e7626b DS |
4053 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4054 | printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n", | |
4055 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4056 | ret = -EINVAL; |
4057 | } | |
21e7626b DS |
4058 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4059 | printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n", | |
4060 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4061 | ret = -EINVAL; |
4062 | } | |
21e7626b DS |
4063 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
4064 | printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n", | |
4065 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
c926093e DS |
4066 | ret = -EINVAL; |
4067 | } | |
4068 | ||
1104a885 | 4069 | /* |
c926093e DS |
4070 | * The common minimum, we don't know if we can trust the nodesize/sectorsize |
4071 | * items yet, they'll be verified later. Issue just a warning. | |
1104a885 | 4072 | */ |
21e7626b | 4073 | if (!IS_ALIGNED(btrfs_super_root(sb), 4096)) |
c926093e | 4074 | printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n", |
cd743fac | 4075 | btrfs_super_root(sb)); |
21e7626b | 4076 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), 4096)) |
cd743fac DS |
4077 | printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n", |
4078 | btrfs_super_chunk_root(sb)); | |
21e7626b | 4079 | if (!IS_ALIGNED(btrfs_super_log_root(sb), 4096)) |
cd743fac | 4080 | printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n", |
21e7626b | 4081 | btrfs_super_log_root(sb)); |
c926093e | 4082 | |
75d6ad38 DS |
4083 | /* |
4084 | * Check the lower bound, the alignment and other constraints are | |
4085 | * checked later. | |
4086 | */ | |
4087 | if (btrfs_super_nodesize(sb) < 4096) { | |
4088 | printk(KERN_ERR "BTRFS: nodesize too small: %u < 4096\n", | |
4089 | btrfs_super_nodesize(sb)); | |
4090 | ret = -EINVAL; | |
4091 | } | |
4092 | if (btrfs_super_sectorsize(sb) < 4096) { | |
4093 | printk(KERN_ERR "BTRFS: sectorsize too small: %u < 4096\n", | |
4094 | btrfs_super_sectorsize(sb)); | |
4095 | ret = -EINVAL; | |
4096 | } | |
4097 | ||
c926093e DS |
4098 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) { |
4099 | printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n", | |
4100 | fs_info->fsid, sb->dev_item.fsid); | |
4101 | ret = -EINVAL; | |
4102 | } | |
4103 | ||
4104 | /* | |
4105 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
4106 | * done later | |
4107 | */ | |
21e7626b | 4108 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
c926093e | 4109 | printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n", |
21e7626b | 4110 | btrfs_super_num_devices(sb)); |
75d6ad38 DS |
4111 | if (btrfs_super_num_devices(sb) == 0) { |
4112 | printk(KERN_ERR "BTRFS: number of devices is 0\n"); | |
4113 | ret = -EINVAL; | |
4114 | } | |
c926093e | 4115 | |
21e7626b | 4116 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
c926093e | 4117 | printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n", |
21e7626b | 4118 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); |
c926093e DS |
4119 | ret = -EINVAL; |
4120 | } | |
4121 | ||
ce7fca5f DS |
4122 | /* |
4123 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4124 | * and one chunk | |
4125 | */ | |
4126 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
4127 | printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n", | |
4128 | btrfs_super_sys_array_size(sb), | |
4129 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
4130 | ret = -EINVAL; | |
4131 | } | |
4132 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4133 | + sizeof(struct btrfs_chunk)) { | |
d2207129 | 4134 | printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n", |
ce7fca5f DS |
4135 | btrfs_super_sys_array_size(sb), |
4136 | sizeof(struct btrfs_disk_key) | |
4137 | + sizeof(struct btrfs_chunk)); | |
4138 | ret = -EINVAL; | |
4139 | } | |
4140 | ||
c926093e DS |
4141 | /* |
4142 | * The generation is a global counter, we'll trust it more than the others | |
4143 | * but it's still possible that it's the one that's wrong. | |
4144 | */ | |
21e7626b | 4145 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
c926093e DS |
4146 | printk(KERN_WARNING |
4147 | "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n", | |
21e7626b DS |
4148 | btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb)); |
4149 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
4150 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
c926093e DS |
4151 | printk(KERN_WARNING |
4152 | "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n", | |
21e7626b | 4153 | btrfs_super_generation(sb), btrfs_super_cache_generation(sb)); |
c926093e DS |
4154 | |
4155 | return ret; | |
acce952b | 4156 | } |
4157 | ||
48a3b636 | 4158 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 4159 | { |
acce952b | 4160 | mutex_lock(&root->fs_info->cleaner_mutex); |
4161 | btrfs_run_delayed_iputs(root); | |
4162 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4163 | ||
4164 | down_write(&root->fs_info->cleanup_work_sem); | |
4165 | up_write(&root->fs_info->cleanup_work_sem); | |
4166 | ||
4167 | /* cleanup FS via transaction */ | |
4168 | btrfs_cleanup_transaction(root); | |
acce952b | 4169 | } |
4170 | ||
143bede5 | 4171 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4172 | { |
acce952b | 4173 | struct btrfs_ordered_extent *ordered; |
acce952b | 4174 | |
199c2a9c | 4175 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4176 | /* |
4177 | * This will just short circuit the ordered completion stuff which will | |
4178 | * make sure the ordered extent gets properly cleaned up. | |
4179 | */ | |
199c2a9c | 4180 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4181 | root_extent_list) |
4182 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4183 | spin_unlock(&root->ordered_extent_lock); |
4184 | } | |
4185 | ||
4186 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4187 | { | |
4188 | struct btrfs_root *root; | |
4189 | struct list_head splice; | |
4190 | ||
4191 | INIT_LIST_HEAD(&splice); | |
4192 | ||
4193 | spin_lock(&fs_info->ordered_root_lock); | |
4194 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4195 | while (!list_empty(&splice)) { | |
4196 | root = list_first_entry(&splice, struct btrfs_root, | |
4197 | ordered_root); | |
1de2cfde JB |
4198 | list_move_tail(&root->ordered_root, |
4199 | &fs_info->ordered_roots); | |
199c2a9c | 4200 | |
2a85d9ca | 4201 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4202 | btrfs_destroy_ordered_extents(root); |
4203 | ||
2a85d9ca LB |
4204 | cond_resched(); |
4205 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4206 | } |
4207 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4208 | } |
4209 | ||
35a3621b SB |
4210 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4211 | struct btrfs_root *root) | |
acce952b | 4212 | { |
4213 | struct rb_node *node; | |
4214 | struct btrfs_delayed_ref_root *delayed_refs; | |
4215 | struct btrfs_delayed_ref_node *ref; | |
4216 | int ret = 0; | |
4217 | ||
4218 | delayed_refs = &trans->delayed_refs; | |
4219 | ||
4220 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4221 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4222 | spin_unlock(&delayed_refs->lock); |
efe120a0 | 4223 | btrfs_info(root->fs_info, "delayed_refs has NO entry"); |
acce952b | 4224 | return ret; |
4225 | } | |
4226 | ||
d7df2c79 JB |
4227 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4228 | struct btrfs_delayed_ref_head *head; | |
c6fc2454 | 4229 | struct btrfs_delayed_ref_node *tmp; |
e78417d1 | 4230 | bool pin_bytes = false; |
acce952b | 4231 | |
d7df2c79 JB |
4232 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4233 | href_node); | |
4234 | if (!mutex_trylock(&head->mutex)) { | |
4235 | atomic_inc(&head->node.refs); | |
4236 | spin_unlock(&delayed_refs->lock); | |
eb12db69 | 4237 | |
d7df2c79 | 4238 | mutex_lock(&head->mutex); |
e78417d1 | 4239 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
4240 | btrfs_put_delayed_ref(&head->node); |
4241 | spin_lock(&delayed_refs->lock); | |
4242 | continue; | |
4243 | } | |
4244 | spin_lock(&head->lock); | |
c6fc2454 QW |
4245 | list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list, |
4246 | list) { | |
d7df2c79 | 4247 | ref->in_tree = 0; |
c6fc2454 | 4248 | list_del(&ref->list); |
d7df2c79 JB |
4249 | atomic_dec(&delayed_refs->num_entries); |
4250 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4251 | } |
d7df2c79 JB |
4252 | if (head->must_insert_reserved) |
4253 | pin_bytes = true; | |
4254 | btrfs_free_delayed_extent_op(head->extent_op); | |
4255 | delayed_refs->num_heads--; | |
4256 | if (head->processing == 0) | |
4257 | delayed_refs->num_heads_ready--; | |
4258 | atomic_dec(&delayed_refs->num_entries); | |
4259 | head->node.in_tree = 0; | |
4260 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
4261 | spin_unlock(&head->lock); | |
4262 | spin_unlock(&delayed_refs->lock); | |
4263 | mutex_unlock(&head->mutex); | |
acce952b | 4264 | |
d7df2c79 JB |
4265 | if (pin_bytes) |
4266 | btrfs_pin_extent(root, head->node.bytenr, | |
4267 | head->node.num_bytes, 1); | |
4268 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 4269 | cond_resched(); |
4270 | spin_lock(&delayed_refs->lock); | |
4271 | } | |
4272 | ||
4273 | spin_unlock(&delayed_refs->lock); | |
4274 | ||
4275 | return ret; | |
4276 | } | |
4277 | ||
143bede5 | 4278 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4279 | { |
4280 | struct btrfs_inode *btrfs_inode; | |
4281 | struct list_head splice; | |
4282 | ||
4283 | INIT_LIST_HEAD(&splice); | |
4284 | ||
eb73c1b7 MX |
4285 | spin_lock(&root->delalloc_lock); |
4286 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4287 | |
4288 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4289 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4290 | delalloc_inodes); | |
acce952b | 4291 | |
4292 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4293 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4294 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4295 | spin_unlock(&root->delalloc_lock); |
acce952b | 4296 | |
4297 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4298 | |
eb73c1b7 | 4299 | spin_lock(&root->delalloc_lock); |
acce952b | 4300 | } |
4301 | ||
eb73c1b7 MX |
4302 | spin_unlock(&root->delalloc_lock); |
4303 | } | |
4304 | ||
4305 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4306 | { | |
4307 | struct btrfs_root *root; | |
4308 | struct list_head splice; | |
4309 | ||
4310 | INIT_LIST_HEAD(&splice); | |
4311 | ||
4312 | spin_lock(&fs_info->delalloc_root_lock); | |
4313 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4314 | while (!list_empty(&splice)) { | |
4315 | root = list_first_entry(&splice, struct btrfs_root, | |
4316 | delalloc_root); | |
4317 | list_del_init(&root->delalloc_root); | |
4318 | root = btrfs_grab_fs_root(root); | |
4319 | BUG_ON(!root); | |
4320 | spin_unlock(&fs_info->delalloc_root_lock); | |
4321 | ||
4322 | btrfs_destroy_delalloc_inodes(root); | |
4323 | btrfs_put_fs_root(root); | |
4324 | ||
4325 | spin_lock(&fs_info->delalloc_root_lock); | |
4326 | } | |
4327 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4328 | } |
4329 | ||
4330 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
4331 | struct extent_io_tree *dirty_pages, | |
4332 | int mark) | |
4333 | { | |
4334 | int ret; | |
acce952b | 4335 | struct extent_buffer *eb; |
4336 | u64 start = 0; | |
4337 | u64 end; | |
acce952b | 4338 | |
4339 | while (1) { | |
4340 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4341 | mark, NULL); |
acce952b | 4342 | if (ret) |
4343 | break; | |
4344 | ||
4345 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
4346 | while (start <= end) { | |
01d58472 | 4347 | eb = btrfs_find_tree_block(root->fs_info, start); |
707e8a07 | 4348 | start += root->nodesize; |
fd8b2b61 | 4349 | if (!eb) |
acce952b | 4350 | continue; |
fd8b2b61 | 4351 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4352 | |
fd8b2b61 JB |
4353 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4354 | &eb->bflags)) | |
4355 | clear_extent_buffer_dirty(eb); | |
4356 | free_extent_buffer_stale(eb); | |
acce952b | 4357 | } |
4358 | } | |
4359 | ||
4360 | return ret; | |
4361 | } | |
4362 | ||
4363 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
4364 | struct extent_io_tree *pinned_extents) | |
4365 | { | |
4366 | struct extent_io_tree *unpin; | |
4367 | u64 start; | |
4368 | u64 end; | |
4369 | int ret; | |
ed0eaa14 | 4370 | bool loop = true; |
acce952b | 4371 | |
4372 | unpin = pinned_extents; | |
ed0eaa14 | 4373 | again: |
acce952b | 4374 | while (1) { |
4375 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4376 | EXTENT_DIRTY, NULL); |
acce952b | 4377 | if (ret) |
4378 | break; | |
4379 | ||
acce952b | 4380 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
4381 | btrfs_error_unpin_extent_range(root, start, end); | |
4382 | cond_resched(); | |
4383 | } | |
4384 | ||
ed0eaa14 LB |
4385 | if (loop) { |
4386 | if (unpin == &root->fs_info->freed_extents[0]) | |
4387 | unpin = &root->fs_info->freed_extents[1]; | |
4388 | else | |
4389 | unpin = &root->fs_info->freed_extents[0]; | |
4390 | loop = false; | |
4391 | goto again; | |
4392 | } | |
4393 | ||
acce952b | 4394 | return 0; |
4395 | } | |
4396 | ||
49b25e05 JM |
4397 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
4398 | struct btrfs_root *root) | |
4399 | { | |
4400 | btrfs_destroy_delayed_refs(cur_trans, root); | |
49b25e05 | 4401 | |
4a9d8bde | 4402 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 4403 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 4404 | |
4a9d8bde | 4405 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 4406 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 4407 | |
67cde344 MX |
4408 | btrfs_destroy_delayed_inodes(root); |
4409 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4410 | |
49b25e05 JM |
4411 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4412 | EXTENT_DIRTY); | |
6e841e32 LB |
4413 | btrfs_destroy_pinned_extent(root, |
4414 | root->fs_info->pinned_extents); | |
49b25e05 | 4415 | |
4a9d8bde MX |
4416 | cur_trans->state =TRANS_STATE_COMPLETED; |
4417 | wake_up(&cur_trans->commit_wait); | |
4418 | ||
49b25e05 JM |
4419 | /* |
4420 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4421 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4422 | */ | |
4423 | } | |
4424 | ||
48a3b636 | 4425 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4426 | { |
4427 | struct btrfs_transaction *t; | |
acce952b | 4428 | |
acce952b | 4429 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4430 | ||
a4abeea4 | 4431 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4432 | while (!list_empty(&root->fs_info->trans_list)) { |
4433 | t = list_first_entry(&root->fs_info->trans_list, | |
4434 | struct btrfs_transaction, list); | |
4435 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4436 | atomic_inc(&t->use_count); | |
4437 | spin_unlock(&root->fs_info->trans_lock); | |
4438 | btrfs_wait_for_commit(root, t->transid); | |
4439 | btrfs_put_transaction(t); | |
4440 | spin_lock(&root->fs_info->trans_lock); | |
4441 | continue; | |
4442 | } | |
4443 | if (t == root->fs_info->running_transaction) { | |
4444 | t->state = TRANS_STATE_COMMIT_DOING; | |
4445 | spin_unlock(&root->fs_info->trans_lock); | |
4446 | /* | |
4447 | * We wait for 0 num_writers since we don't hold a trans | |
4448 | * handle open currently for this transaction. | |
4449 | */ | |
4450 | wait_event(t->writer_wait, | |
4451 | atomic_read(&t->num_writers) == 0); | |
4452 | } else { | |
4453 | spin_unlock(&root->fs_info->trans_lock); | |
4454 | } | |
4455 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4456 | |
724e2315 JB |
4457 | spin_lock(&root->fs_info->trans_lock); |
4458 | if (t == root->fs_info->running_transaction) | |
4459 | root->fs_info->running_transaction = NULL; | |
acce952b | 4460 | list_del_init(&t->list); |
724e2315 | 4461 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4462 | |
724e2315 JB |
4463 | btrfs_put_transaction(t); |
4464 | trace_btrfs_transaction_commit(root); | |
4465 | spin_lock(&root->fs_info->trans_lock); | |
4466 | } | |
4467 | spin_unlock(&root->fs_info->trans_lock); | |
4468 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4469 | btrfs_destroy_delayed_inodes(root); | |
4470 | btrfs_assert_delayed_root_empty(root); | |
4471 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4472 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4473 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4474 | |
4475 | return 0; | |
4476 | } | |
4477 | ||
e8c9f186 | 4478 | static const struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4479 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4480 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4481 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4482 | /* note we're sharing with inode.c for the merge bio hook */ |
4483 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4484 | }; |