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" |
8b712842 | 42 | #include "async-thread.h" |
925baedd | 43 | #include "locking.h" |
e02119d5 | 44 | #include "tree-log.h" |
fa9c0d79 | 45 | #include "free-space-cache.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" |
eb60ceac | 52 | |
de0022b9 JB |
53 | #ifdef CONFIG_X86 |
54 | #include <asm/cpufeature.h> | |
55 | #endif | |
56 | ||
d1310b2e | 57 | static struct extent_io_ops btree_extent_io_ops; |
d458b054 | 58 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 59 | static void free_fs_root(struct btrfs_root *root); |
fcd1f065 | 60 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 61 | int read_only); |
569e0f35 JB |
62 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
63 | struct btrfs_root *root); | |
143bede5 | 64 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 65 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
66 | struct btrfs_root *root); | |
143bede5 | 67 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
acce952b | 68 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, |
69 | struct extent_io_tree *dirty_pages, | |
70 | int mark); | |
71 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
72 | struct extent_io_tree *pinned_extents); | |
48a3b636 ES |
73 | static int btrfs_cleanup_transaction(struct btrfs_root *root); |
74 | static void btrfs_error_commit_super(struct btrfs_root *root); | |
ce9adaa5 | 75 | |
d352ac68 CM |
76 | /* |
77 | * end_io_wq structs are used to do processing in task context when an IO is | |
78 | * complete. This is used during reads to verify checksums, and it is used | |
79 | * by writes to insert metadata for new file extents after IO is complete. | |
80 | */ | |
ce9adaa5 CM |
81 | struct end_io_wq { |
82 | struct bio *bio; | |
83 | bio_end_io_t *end_io; | |
84 | void *private; | |
85 | struct btrfs_fs_info *info; | |
86 | int error; | |
22c59948 | 87 | int metadata; |
ce9adaa5 | 88 | struct list_head list; |
d458b054 | 89 | struct btrfs_work work; |
ce9adaa5 | 90 | }; |
0da5468f | 91 | |
d352ac68 CM |
92 | /* |
93 | * async submit bios are used to offload expensive checksumming | |
94 | * onto the worker threads. They checksum file and metadata bios | |
95 | * just before they are sent down the IO stack. | |
96 | */ | |
44b8bd7e CM |
97 | struct async_submit_bio { |
98 | struct inode *inode; | |
99 | struct bio *bio; | |
100 | struct list_head list; | |
4a69a410 CM |
101 | extent_submit_bio_hook_t *submit_bio_start; |
102 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
103 | int rw; |
104 | int mirror_num; | |
c8b97818 | 105 | unsigned long bio_flags; |
eaf25d93 CM |
106 | /* |
107 | * bio_offset is optional, can be used if the pages in the bio | |
108 | * can't tell us where in the file the bio should go | |
109 | */ | |
110 | u64 bio_offset; | |
d458b054 | 111 | struct btrfs_work work; |
79787eaa | 112 | int error; |
44b8bd7e CM |
113 | }; |
114 | ||
85d4e461 CM |
115 | /* |
116 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
117 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
118 | * the level the eb occupies in the tree. | |
119 | * | |
120 | * Different roots are used for different purposes and may nest inside each | |
121 | * other and they require separate keysets. As lockdep keys should be | |
122 | * static, assign keysets according to the purpose of the root as indicated | |
123 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
124 | * have separate keysets. | |
4008c04a | 125 | * |
85d4e461 CM |
126 | * Lock-nesting across peer nodes is always done with the immediate parent |
127 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
128 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 129 | * |
85d4e461 CM |
130 | * The key is set by the readpage_end_io_hook after the buffer has passed |
131 | * csum validation but before the pages are unlocked. It is also set by | |
132 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 133 | * |
85d4e461 CM |
134 | * We also add a check to make sure the highest level of the tree is the |
135 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
136 | * needs update as well. | |
4008c04a CM |
137 | */ |
138 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
139 | # if BTRFS_MAX_LEVEL != 8 | |
140 | # error | |
141 | # endif | |
85d4e461 CM |
142 | |
143 | static struct btrfs_lockdep_keyset { | |
144 | u64 id; /* root objectid */ | |
145 | const char *name_stem; /* lock name stem */ | |
146 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
147 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
148 | } btrfs_lockdep_keysets[] = { | |
149 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
150 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
151 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
152 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
153 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
154 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 155 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
156 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
157 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
158 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 159 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
85d4e461 | 160 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 161 | }; |
85d4e461 CM |
162 | |
163 | void __init btrfs_init_lockdep(void) | |
164 | { | |
165 | int i, j; | |
166 | ||
167 | /* initialize lockdep class names */ | |
168 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
169 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
170 | ||
171 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
172 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
173 | "btrfs-%s-%02d", ks->name_stem, j); | |
174 | } | |
175 | } | |
176 | ||
177 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
178 | int level) | |
179 | { | |
180 | struct btrfs_lockdep_keyset *ks; | |
181 | ||
182 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
183 | ||
184 | /* find the matching keyset, id 0 is the default entry */ | |
185 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
186 | if (ks->id == objectid) | |
187 | break; | |
188 | ||
189 | lockdep_set_class_and_name(&eb->lock, | |
190 | &ks->keys[level], ks->names[level]); | |
191 | } | |
192 | ||
4008c04a CM |
193 | #endif |
194 | ||
d352ac68 CM |
195 | /* |
196 | * extents on the btree inode are pretty simple, there's one extent | |
197 | * that covers the entire device | |
198 | */ | |
b2950863 | 199 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 200 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 201 | int create) |
7eccb903 | 202 | { |
5f39d397 CM |
203 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
204 | struct extent_map *em; | |
205 | int ret; | |
206 | ||
890871be | 207 | read_lock(&em_tree->lock); |
d1310b2e | 208 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
209 | if (em) { |
210 | em->bdev = | |
211 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 212 | read_unlock(&em_tree->lock); |
5f39d397 | 213 | goto out; |
a061fc8d | 214 | } |
890871be | 215 | read_unlock(&em_tree->lock); |
7b13b7b1 | 216 | |
172ddd60 | 217 | em = alloc_extent_map(); |
5f39d397 CM |
218 | if (!em) { |
219 | em = ERR_PTR(-ENOMEM); | |
220 | goto out; | |
221 | } | |
222 | em->start = 0; | |
0afbaf8c | 223 | em->len = (u64)-1; |
c8b97818 | 224 | em->block_len = (u64)-1; |
5f39d397 | 225 | em->block_start = 0; |
a061fc8d | 226 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 227 | |
890871be | 228 | write_lock(&em_tree->lock); |
09a2a8f9 | 229 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
230 | if (ret == -EEXIST) { |
231 | free_extent_map(em); | |
7b13b7b1 | 232 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 233 | if (!em) |
0433f20d | 234 | em = ERR_PTR(-EIO); |
5f39d397 | 235 | } else if (ret) { |
7b13b7b1 | 236 | free_extent_map(em); |
0433f20d | 237 | em = ERR_PTR(ret); |
5f39d397 | 238 | } |
890871be | 239 | write_unlock(&em_tree->lock); |
7b13b7b1 | 240 | |
5f39d397 CM |
241 | out: |
242 | return em; | |
7eccb903 CM |
243 | } |
244 | ||
b0496686 | 245 | u32 btrfs_csum_data(char *data, u32 seed, size_t len) |
19c00ddc | 246 | { |
0b947aff | 247 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
248 | } |
249 | ||
250 | void btrfs_csum_final(u32 crc, char *result) | |
251 | { | |
7e75bf3f | 252 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
253 | } |
254 | ||
d352ac68 CM |
255 | /* |
256 | * compute the csum for a btree block, and either verify it or write it | |
257 | * into the csum field of the block. | |
258 | */ | |
19c00ddc CM |
259 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
260 | int verify) | |
261 | { | |
6c41761f | 262 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 263 | char *result = NULL; |
19c00ddc CM |
264 | unsigned long len; |
265 | unsigned long cur_len; | |
266 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
267 | char *kaddr; |
268 | unsigned long map_start; | |
269 | unsigned long map_len; | |
270 | int err; | |
271 | u32 crc = ~(u32)0; | |
607d432d | 272 | unsigned long inline_result; |
19c00ddc CM |
273 | |
274 | len = buf->len - offset; | |
d397712b | 275 | while (len > 0) { |
19c00ddc | 276 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 277 | &kaddr, &map_start, &map_len); |
d397712b | 278 | if (err) |
19c00ddc | 279 | return 1; |
19c00ddc | 280 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 281 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
282 | crc, cur_len); |
283 | len -= cur_len; | |
284 | offset += cur_len; | |
19c00ddc | 285 | } |
607d432d JB |
286 | if (csum_size > sizeof(inline_result)) { |
287 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
288 | if (!result) | |
289 | return 1; | |
290 | } else { | |
291 | result = (char *)&inline_result; | |
292 | } | |
293 | ||
19c00ddc CM |
294 | btrfs_csum_final(crc, result); |
295 | ||
296 | if (verify) { | |
607d432d | 297 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
298 | u32 val; |
299 | u32 found = 0; | |
607d432d | 300 | memcpy(&found, result, csum_size); |
e4204ded | 301 | |
607d432d | 302 | read_extent_buffer(buf, &val, 0, csum_size); |
efe120a0 FH |
303 | printk_ratelimited(KERN_INFO |
304 | "BTRFS: %s checksum verify failed on %llu wanted %X found %X " | |
305 | "level %d\n", | |
306 | root->fs_info->sb->s_id, buf->start, | |
307 | val, found, btrfs_header_level(buf)); | |
607d432d JB |
308 | if (result != (char *)&inline_result) |
309 | kfree(result); | |
19c00ddc CM |
310 | return 1; |
311 | } | |
312 | } else { | |
607d432d | 313 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 314 | } |
607d432d JB |
315 | if (result != (char *)&inline_result) |
316 | kfree(result); | |
19c00ddc CM |
317 | return 0; |
318 | } | |
319 | ||
d352ac68 CM |
320 | /* |
321 | * we can't consider a given block up to date unless the transid of the | |
322 | * block matches the transid in the parent node's pointer. This is how we | |
323 | * detect blocks that either didn't get written at all or got written | |
324 | * in the wrong place. | |
325 | */ | |
1259ab75 | 326 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
327 | struct extent_buffer *eb, u64 parent_transid, |
328 | int atomic) | |
1259ab75 | 329 | { |
2ac55d41 | 330 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
331 | int ret; |
332 | ||
333 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
334 | return 0; | |
335 | ||
b9fab919 CM |
336 | if (atomic) |
337 | return -EAGAIN; | |
338 | ||
2ac55d41 | 339 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
d0082371 | 340 | 0, &cached_state); |
0b32f4bb | 341 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
342 | btrfs_header_generation(eb) == parent_transid) { |
343 | ret = 0; | |
344 | goto out; | |
345 | } | |
7a36ddec | 346 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d | 347 | "found %llu\n", |
c1c9ff7c | 348 | eb->start, parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 349 | ret = 1; |
0b32f4bb | 350 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 351 | out: |
2ac55d41 JB |
352 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
353 | &cached_state, GFP_NOFS); | |
1259ab75 | 354 | return ret; |
1259ab75 CM |
355 | } |
356 | ||
1104a885 DS |
357 | /* |
358 | * Return 0 if the superblock checksum type matches the checksum value of that | |
359 | * algorithm. Pass the raw disk superblock data. | |
360 | */ | |
361 | static int btrfs_check_super_csum(char *raw_disk_sb) | |
362 | { | |
363 | struct btrfs_super_block *disk_sb = | |
364 | (struct btrfs_super_block *)raw_disk_sb; | |
365 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
366 | int ret = 0; | |
367 | ||
368 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
369 | u32 crc = ~(u32)0; | |
370 | const int csum_size = sizeof(crc); | |
371 | char result[csum_size]; | |
372 | ||
373 | /* | |
374 | * The super_block structure does not span the whole | |
375 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
376 | * is filled with zeros and is included in the checkum. | |
377 | */ | |
378 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
379 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
380 | btrfs_csum_final(crc, result); | |
381 | ||
382 | if (memcmp(raw_disk_sb, result, csum_size)) | |
383 | ret = 1; | |
667e7d94 CM |
384 | |
385 | if (ret && btrfs_super_generation(disk_sb) < 10) { | |
efe120a0 FH |
386 | printk(KERN_WARNING |
387 | "BTRFS: super block crcs don't match, older mkfs detected\n"); | |
667e7d94 CM |
388 | ret = 0; |
389 | } | |
1104a885 DS |
390 | } |
391 | ||
392 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
efe120a0 | 393 | printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", |
1104a885 DS |
394 | csum_type); |
395 | ret = 1; | |
396 | } | |
397 | ||
398 | return ret; | |
399 | } | |
400 | ||
d352ac68 CM |
401 | /* |
402 | * helper to read a given tree block, doing retries as required when | |
403 | * the checksums don't match and we have alternate mirrors to try. | |
404 | */ | |
f188591e CM |
405 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
406 | struct extent_buffer *eb, | |
ca7a79ad | 407 | u64 start, u64 parent_transid) |
f188591e CM |
408 | { |
409 | struct extent_io_tree *io_tree; | |
ea466794 | 410 | int failed = 0; |
f188591e CM |
411 | int ret; |
412 | int num_copies = 0; | |
413 | int mirror_num = 0; | |
ea466794 | 414 | int failed_mirror = 0; |
f188591e | 415 | |
a826d6dc | 416 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
417 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
418 | while (1) { | |
bb82ab88 AJ |
419 | ret = read_extent_buffer_pages(io_tree, eb, start, |
420 | WAIT_COMPLETE, | |
f188591e | 421 | btree_get_extent, mirror_num); |
256dd1bb SB |
422 | if (!ret) { |
423 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 424 | parent_transid, 0)) |
256dd1bb SB |
425 | break; |
426 | else | |
427 | ret = -EIO; | |
428 | } | |
d397712b | 429 | |
a826d6dc JB |
430 | /* |
431 | * This buffer's crc is fine, but its contents are corrupted, so | |
432 | * there is no reason to read the other copies, they won't be | |
433 | * any less wrong. | |
434 | */ | |
435 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
436 | break; |
437 | ||
5d964051 | 438 | num_copies = btrfs_num_copies(root->fs_info, |
f188591e | 439 | eb->start, eb->len); |
4235298e | 440 | if (num_copies == 1) |
ea466794 | 441 | break; |
4235298e | 442 | |
5cf1ab56 JB |
443 | if (!failed_mirror) { |
444 | failed = 1; | |
445 | failed_mirror = eb->read_mirror; | |
446 | } | |
447 | ||
f188591e | 448 | mirror_num++; |
ea466794 JB |
449 | if (mirror_num == failed_mirror) |
450 | mirror_num++; | |
451 | ||
4235298e | 452 | if (mirror_num > num_copies) |
ea466794 | 453 | break; |
f188591e | 454 | } |
ea466794 | 455 | |
c0901581 | 456 | if (failed && !ret && failed_mirror) |
ea466794 JB |
457 | repair_eb_io_failure(root, eb, failed_mirror); |
458 | ||
459 | return ret; | |
f188591e | 460 | } |
19c00ddc | 461 | |
d352ac68 | 462 | /* |
d397712b CM |
463 | * checksum a dirty tree block before IO. This has extra checks to make sure |
464 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 465 | */ |
d397712b | 466 | |
b2950863 | 467 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 468 | { |
4eee4fa4 | 469 | u64 start = page_offset(page); |
19c00ddc | 470 | u64 found_start; |
19c00ddc | 471 | struct extent_buffer *eb; |
f188591e | 472 | |
4f2de97a JB |
473 | eb = (struct extent_buffer *)page->private; |
474 | if (page != eb->pages[0]) | |
475 | return 0; | |
19c00ddc | 476 | found_start = btrfs_header_bytenr(eb); |
fae7f21c | 477 | if (WARN_ON(found_start != start || !PageUptodate(page))) |
4f2de97a | 478 | return 0; |
19c00ddc | 479 | csum_tree_block(root, eb, 0); |
19c00ddc CM |
480 | return 0; |
481 | } | |
482 | ||
2b82032c YZ |
483 | static int check_tree_block_fsid(struct btrfs_root *root, |
484 | struct extent_buffer *eb) | |
485 | { | |
486 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
487 | u8 fsid[BTRFS_UUID_SIZE]; | |
488 | int ret = 1; | |
489 | ||
0a4e5586 | 490 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
491 | while (fs_devices) { |
492 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
493 | ret = 0; | |
494 | break; | |
495 | } | |
496 | fs_devices = fs_devices->seed; | |
497 | } | |
498 | return ret; | |
499 | } | |
500 | ||
a826d6dc | 501 | #define CORRUPT(reason, eb, root, slot) \ |
efe120a0 FH |
502 | btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ |
503 | "root=%llu, slot=%d", reason, \ | |
c1c9ff7c | 504 | btrfs_header_bytenr(eb), root->objectid, slot) |
a826d6dc JB |
505 | |
506 | static noinline int check_leaf(struct btrfs_root *root, | |
507 | struct extent_buffer *leaf) | |
508 | { | |
509 | struct btrfs_key key; | |
510 | struct btrfs_key leaf_key; | |
511 | u32 nritems = btrfs_header_nritems(leaf); | |
512 | int slot; | |
513 | ||
514 | if (nritems == 0) | |
515 | return 0; | |
516 | ||
517 | /* Check the 0 item */ | |
518 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
519 | BTRFS_LEAF_DATA_SIZE(root)) { | |
520 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
521 | return -EIO; | |
522 | } | |
523 | ||
524 | /* | |
525 | * Check to make sure each items keys are in the correct order and their | |
526 | * offsets make sense. We only have to loop through nritems-1 because | |
527 | * we check the current slot against the next slot, which verifies the | |
528 | * next slot's offset+size makes sense and that the current's slot | |
529 | * offset is correct. | |
530 | */ | |
531 | for (slot = 0; slot < nritems - 1; slot++) { | |
532 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
533 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
534 | ||
535 | /* Make sure the keys are in the right order */ | |
536 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
537 | CORRUPT("bad key order", leaf, root, slot); | |
538 | return -EIO; | |
539 | } | |
540 | ||
541 | /* | |
542 | * Make sure the offset and ends are right, remember that the | |
543 | * item data starts at the end of the leaf and grows towards the | |
544 | * front. | |
545 | */ | |
546 | if (btrfs_item_offset_nr(leaf, slot) != | |
547 | btrfs_item_end_nr(leaf, slot + 1)) { | |
548 | CORRUPT("slot offset bad", leaf, root, slot); | |
549 | return -EIO; | |
550 | } | |
551 | ||
552 | /* | |
553 | * Check to make sure that we don't point outside of the leaf, | |
554 | * just incase all the items are consistent to eachother, but | |
555 | * all point outside of the leaf. | |
556 | */ | |
557 | if (btrfs_item_end_nr(leaf, slot) > | |
558 | BTRFS_LEAF_DATA_SIZE(root)) { | |
559 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
560 | return -EIO; | |
561 | } | |
562 | } | |
563 | ||
564 | return 0; | |
565 | } | |
566 | ||
facc8a22 MX |
567 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
568 | u64 phy_offset, struct page *page, | |
569 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 570 | { |
ce9adaa5 CM |
571 | u64 found_start; |
572 | int found_level; | |
ce9adaa5 CM |
573 | struct extent_buffer *eb; |
574 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 575 | int ret = 0; |
727011e0 | 576 | int reads_done; |
ce9adaa5 | 577 | |
ce9adaa5 CM |
578 | if (!page->private) |
579 | goto out; | |
d397712b | 580 | |
4f2de97a | 581 | eb = (struct extent_buffer *)page->private; |
d397712b | 582 | |
0b32f4bb JB |
583 | /* the pending IO might have been the only thing that kept this buffer |
584 | * in memory. Make sure we have a ref for all this other checks | |
585 | */ | |
586 | extent_buffer_get(eb); | |
587 | ||
588 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
589 | if (!reads_done) |
590 | goto err; | |
f188591e | 591 | |
5cf1ab56 | 592 | eb->read_mirror = mirror; |
ea466794 JB |
593 | if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { |
594 | ret = -EIO; | |
595 | goto err; | |
596 | } | |
597 | ||
ce9adaa5 | 598 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 599 | if (found_start != eb->start) { |
efe120a0 | 600 | printk_ratelimited(KERN_INFO "BTRFS: bad tree block start " |
193f284d | 601 | "%llu %llu\n", |
c1c9ff7c | 602 | found_start, eb->start); |
f188591e | 603 | ret = -EIO; |
ce9adaa5 CM |
604 | goto err; |
605 | } | |
2b82032c | 606 | if (check_tree_block_fsid(root, eb)) { |
efe120a0 | 607 | printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n", |
c1c9ff7c | 608 | eb->start); |
1259ab75 CM |
609 | ret = -EIO; |
610 | goto err; | |
611 | } | |
ce9adaa5 | 612 | found_level = btrfs_header_level(eb); |
1c24c3ce | 613 | if (found_level >= BTRFS_MAX_LEVEL) { |
efe120a0 | 614 | btrfs_info(root->fs_info, "bad tree block level %d", |
1c24c3ce JB |
615 | (int)btrfs_header_level(eb)); |
616 | ret = -EIO; | |
617 | goto err; | |
618 | } | |
ce9adaa5 | 619 | |
85d4e461 CM |
620 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
621 | eb, found_level); | |
4008c04a | 622 | |
ce9adaa5 | 623 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 624 | if (ret) { |
f188591e | 625 | ret = -EIO; |
a826d6dc JB |
626 | goto err; |
627 | } | |
628 | ||
629 | /* | |
630 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
631 | * that we don't try and read the other copies of this block, just | |
632 | * return -EIO. | |
633 | */ | |
634 | if (found_level == 0 && check_leaf(root, eb)) { | |
635 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
636 | ret = -EIO; | |
637 | } | |
ce9adaa5 | 638 | |
0b32f4bb JB |
639 | if (!ret) |
640 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 641 | err: |
79fb65a1 JB |
642 | if (reads_done && |
643 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
4bb31e92 | 644 | btree_readahead_hook(root, eb, eb->start, ret); |
4bb31e92 | 645 | |
53b381b3 DW |
646 | if (ret) { |
647 | /* | |
648 | * our io error hook is going to dec the io pages | |
649 | * again, we have to make sure it has something | |
650 | * to decrement | |
651 | */ | |
652 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 653 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 654 | } |
0b32f4bb | 655 | free_extent_buffer(eb); |
ce9adaa5 | 656 | out: |
f188591e | 657 | return ret; |
ce9adaa5 CM |
658 | } |
659 | ||
ea466794 | 660 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 661 | { |
4bb31e92 AJ |
662 | struct extent_buffer *eb; |
663 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
664 | ||
4f2de97a | 665 | eb = (struct extent_buffer *)page->private; |
ea466794 | 666 | set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); |
5cf1ab56 | 667 | eb->read_mirror = failed_mirror; |
53b381b3 | 668 | atomic_dec(&eb->io_pages); |
ea466794 | 669 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
4bb31e92 | 670 | btree_readahead_hook(root, eb, eb->start, -EIO); |
4bb31e92 AJ |
671 | return -EIO; /* we fixed nothing */ |
672 | } | |
673 | ||
ce9adaa5 | 674 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
675 | { |
676 | struct end_io_wq *end_io_wq = bio->bi_private; | |
677 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 678 | |
ce9adaa5 | 679 | fs_info = end_io_wq->info; |
ce9adaa5 | 680 | end_io_wq->error = err; |
fccb5d86 | 681 | btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); |
d20f7043 | 682 | |
7b6d91da | 683 | if (bio->bi_rw & REQ_WRITE) { |
53b381b3 | 684 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) |
fccb5d86 QW |
685 | btrfs_queue_work(fs_info->endio_meta_write_workers, |
686 | &end_io_wq->work); | |
53b381b3 | 687 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) |
fccb5d86 QW |
688 | btrfs_queue_work(fs_info->endio_freespace_worker, |
689 | &end_io_wq->work); | |
53b381b3 | 690 | else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
fccb5d86 QW |
691 | btrfs_queue_work(fs_info->endio_raid56_workers, |
692 | &end_io_wq->work); | |
cad321ad | 693 | else |
fccb5d86 QW |
694 | btrfs_queue_work(fs_info->endio_write_workers, |
695 | &end_io_wq->work); | |
d20f7043 | 696 | } else { |
53b381b3 | 697 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) |
fccb5d86 QW |
698 | btrfs_queue_work(fs_info->endio_raid56_workers, |
699 | &end_io_wq->work); | |
53b381b3 | 700 | else if (end_io_wq->metadata) |
fccb5d86 QW |
701 | btrfs_queue_work(fs_info->endio_meta_workers, |
702 | &end_io_wq->work); | |
d20f7043 | 703 | else |
fccb5d86 QW |
704 | btrfs_queue_work(fs_info->endio_workers, |
705 | &end_io_wq->work); | |
d20f7043 | 706 | } |
ce9adaa5 CM |
707 | } |
708 | ||
0cb59c99 JB |
709 | /* |
710 | * For the metadata arg you want | |
711 | * | |
712 | * 0 - if data | |
713 | * 1 - if normal metadta | |
714 | * 2 - if writing to the free space cache area | |
53b381b3 | 715 | * 3 - raid parity work |
0cb59c99 | 716 | */ |
22c59948 CM |
717 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
718 | int metadata) | |
0b86a832 | 719 | { |
ce9adaa5 | 720 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
721 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
722 | if (!end_io_wq) | |
723 | return -ENOMEM; | |
724 | ||
725 | end_io_wq->private = bio->bi_private; | |
726 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 727 | end_io_wq->info = info; |
ce9adaa5 CM |
728 | end_io_wq->error = 0; |
729 | end_io_wq->bio = bio; | |
22c59948 | 730 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
731 | |
732 | bio->bi_private = end_io_wq; | |
733 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
734 | return 0; |
735 | } | |
736 | ||
b64a2851 | 737 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 738 | { |
4854ddd0 | 739 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 740 | info->thread_pool_size, |
4854ddd0 CM |
741 | info->fs_devices->open_devices); |
742 | return 256 * limit; | |
743 | } | |
0986fe9e | 744 | |
d458b054 | 745 | static void run_one_async_start(struct btrfs_work *work) |
4a69a410 | 746 | { |
4a69a410 | 747 | struct async_submit_bio *async; |
79787eaa | 748 | int ret; |
4a69a410 CM |
749 | |
750 | async = container_of(work, struct async_submit_bio, work); | |
79787eaa JM |
751 | ret = async->submit_bio_start(async->inode, async->rw, async->bio, |
752 | async->mirror_num, async->bio_flags, | |
753 | async->bio_offset); | |
754 | if (ret) | |
755 | async->error = ret; | |
4a69a410 CM |
756 | } |
757 | ||
d458b054 | 758 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 CM |
759 | { |
760 | struct btrfs_fs_info *fs_info; | |
761 | struct async_submit_bio *async; | |
4854ddd0 | 762 | int limit; |
8b712842 CM |
763 | |
764 | async = container_of(work, struct async_submit_bio, work); | |
765 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 766 | |
b64a2851 | 767 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
768 | limit = limit * 2 / 3; |
769 | ||
66657b31 | 770 | if (atomic_dec_return(&fs_info->nr_async_submits) < limit && |
b64a2851 | 771 | waitqueue_active(&fs_info->async_submit_wait)) |
4854ddd0 CM |
772 | wake_up(&fs_info->async_submit_wait); |
773 | ||
79787eaa JM |
774 | /* If an error occured we just want to clean up the bio and move on */ |
775 | if (async->error) { | |
776 | bio_endio(async->bio, async->error); | |
777 | return; | |
778 | } | |
779 | ||
4a69a410 | 780 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
781 | async->mirror_num, async->bio_flags, |
782 | async->bio_offset); | |
4a69a410 CM |
783 | } |
784 | ||
d458b054 | 785 | static void run_one_async_free(struct btrfs_work *work) |
4a69a410 CM |
786 | { |
787 | struct async_submit_bio *async; | |
788 | ||
789 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
790 | kfree(async); |
791 | } | |
792 | ||
44b8bd7e CM |
793 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
794 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 795 | unsigned long bio_flags, |
eaf25d93 | 796 | u64 bio_offset, |
4a69a410 CM |
797 | extent_submit_bio_hook_t *submit_bio_start, |
798 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
799 | { |
800 | struct async_submit_bio *async; | |
801 | ||
802 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
803 | if (!async) | |
804 | return -ENOMEM; | |
805 | ||
806 | async->inode = inode; | |
807 | async->rw = rw; | |
808 | async->bio = bio; | |
809 | async->mirror_num = mirror_num; | |
4a69a410 CM |
810 | async->submit_bio_start = submit_bio_start; |
811 | async->submit_bio_done = submit_bio_done; | |
812 | ||
5cdc7ad3 QW |
813 | btrfs_init_work(&async->work, run_one_async_start, |
814 | run_one_async_done, run_one_async_free); | |
4a69a410 | 815 | |
c8b97818 | 816 | async->bio_flags = bio_flags; |
eaf25d93 | 817 | async->bio_offset = bio_offset; |
8c8bee1d | 818 | |
79787eaa JM |
819 | async->error = 0; |
820 | ||
cb03c743 | 821 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 822 | |
7b6d91da | 823 | if (rw & REQ_SYNC) |
5cdc7ad3 | 824 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 825 | |
5cdc7ad3 | 826 | btrfs_queue_work(fs_info->workers, &async->work); |
9473f16c | 827 | |
d397712b | 828 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
829 | atomic_read(&fs_info->nr_async_submits)) { |
830 | wait_event(fs_info->async_submit_wait, | |
831 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
832 | } | |
833 | ||
44b8bd7e CM |
834 | return 0; |
835 | } | |
836 | ||
ce3ed71a CM |
837 | static int btree_csum_one_bio(struct bio *bio) |
838 | { | |
839 | struct bio_vec *bvec = bio->bi_io_vec; | |
840 | int bio_index = 0; | |
841 | struct btrfs_root *root; | |
79787eaa | 842 | int ret = 0; |
ce3ed71a CM |
843 | |
844 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 845 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a | 846 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
79787eaa JM |
847 | ret = csum_dirty_buffer(root, bvec->bv_page); |
848 | if (ret) | |
849 | break; | |
ce3ed71a CM |
850 | bio_index++; |
851 | bvec++; | |
852 | } | |
79787eaa | 853 | return ret; |
ce3ed71a CM |
854 | } |
855 | ||
4a69a410 CM |
856 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
857 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
858 | unsigned long bio_flags, |
859 | u64 bio_offset) | |
22c59948 | 860 | { |
8b712842 CM |
861 | /* |
862 | * when we're called for a write, we're already in the async | |
5443be45 | 863 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 864 | */ |
79787eaa | 865 | return btree_csum_one_bio(bio); |
4a69a410 | 866 | } |
22c59948 | 867 | |
4a69a410 | 868 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
869 | int mirror_num, unsigned long bio_flags, |
870 | u64 bio_offset) | |
4a69a410 | 871 | { |
61891923 SB |
872 | int ret; |
873 | ||
8b712842 | 874 | /* |
4a69a410 CM |
875 | * when we're called for a write, we're already in the async |
876 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 877 | */ |
61891923 SB |
878 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
879 | if (ret) | |
880 | bio_endio(bio, ret); | |
881 | return ret; | |
0b86a832 CM |
882 | } |
883 | ||
de0022b9 JB |
884 | static int check_async_write(struct inode *inode, unsigned long bio_flags) |
885 | { | |
886 | if (bio_flags & EXTENT_BIO_TREE_LOG) | |
887 | return 0; | |
888 | #ifdef CONFIG_X86 | |
889 | if (cpu_has_xmm4_2) | |
890 | return 0; | |
891 | #endif | |
892 | return 1; | |
893 | } | |
894 | ||
44b8bd7e | 895 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
896 | int mirror_num, unsigned long bio_flags, |
897 | u64 bio_offset) | |
44b8bd7e | 898 | { |
de0022b9 | 899 | int async = check_async_write(inode, bio_flags); |
cad321ad CM |
900 | int ret; |
901 | ||
7b6d91da | 902 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
903 | /* |
904 | * called for a read, do the setup so that checksum validation | |
905 | * can happen in the async kernel threads | |
906 | */ | |
f3f266ab CM |
907 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, |
908 | bio, 1); | |
1d4284bd | 909 | if (ret) |
61891923 SB |
910 | goto out_w_error; |
911 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
912 | mirror_num, 0); | |
de0022b9 JB |
913 | } else if (!async) { |
914 | ret = btree_csum_one_bio(bio); | |
915 | if (ret) | |
61891923 SB |
916 | goto out_w_error; |
917 | ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
918 | mirror_num, 0); | |
919 | } else { | |
920 | /* | |
921 | * kthread helpers are used to submit writes so that | |
922 | * checksumming can happen in parallel across all CPUs | |
923 | */ | |
924 | ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
925 | inode, rw, bio, mirror_num, 0, | |
926 | bio_offset, | |
927 | __btree_submit_bio_start, | |
928 | __btree_submit_bio_done); | |
44b8bd7e | 929 | } |
d313d7a3 | 930 | |
61891923 SB |
931 | if (ret) { |
932 | out_w_error: | |
933 | bio_endio(bio, ret); | |
934 | } | |
935 | return ret; | |
44b8bd7e CM |
936 | } |
937 | ||
3dd1462e | 938 | #ifdef CONFIG_MIGRATION |
784b4e29 | 939 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
940 | struct page *newpage, struct page *page, |
941 | enum migrate_mode mode) | |
784b4e29 CM |
942 | { |
943 | /* | |
944 | * we can't safely write a btree page from here, | |
945 | * we haven't done the locking hook | |
946 | */ | |
947 | if (PageDirty(page)) | |
948 | return -EAGAIN; | |
949 | /* | |
950 | * Buffers may be managed in a filesystem specific way. | |
951 | * We must have no buffers or drop them. | |
952 | */ | |
953 | if (page_has_private(page) && | |
954 | !try_to_release_page(page, GFP_KERNEL)) | |
955 | return -EAGAIN; | |
a6bc32b8 | 956 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 957 | } |
3dd1462e | 958 | #endif |
784b4e29 | 959 | |
0da5468f CM |
960 | |
961 | static int btree_writepages(struct address_space *mapping, | |
962 | struct writeback_control *wbc) | |
963 | { | |
e2d84521 MX |
964 | struct btrfs_fs_info *fs_info; |
965 | int ret; | |
966 | ||
d8d5f3e1 | 967 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
968 | |
969 | if (wbc->for_kupdate) | |
970 | return 0; | |
971 | ||
e2d84521 | 972 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 973 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
974 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
975 | BTRFS_DIRTY_METADATA_THRESH); | |
976 | if (ret < 0) | |
793955bc | 977 | return 0; |
793955bc | 978 | } |
0b32f4bb | 979 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
980 | } |
981 | ||
b2950863 | 982 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 983 | { |
d1310b2e CM |
984 | struct extent_io_tree *tree; |
985 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 986 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 987 | } |
22b0ebda | 988 | |
70dec807 | 989 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 990 | { |
98509cfc | 991 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 992 | return 0; |
0c4e538b | 993 | |
f7a52a40 | 994 | return try_release_extent_buffer(page); |
d98237b3 CM |
995 | } |
996 | ||
d47992f8 LC |
997 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
998 | unsigned int length) | |
d98237b3 | 999 | { |
d1310b2e CM |
1000 | struct extent_io_tree *tree; |
1001 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
1002 | extent_invalidatepage(tree, page, offset); |
1003 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 1004 | if (PagePrivate(page)) { |
efe120a0 FH |
1005 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
1006 | "page private not zero on page %llu", | |
1007 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
1008 | ClearPagePrivate(page); |
1009 | set_page_private(page, 0); | |
1010 | page_cache_release(page); | |
1011 | } | |
d98237b3 CM |
1012 | } |
1013 | ||
0b32f4bb JB |
1014 | static int btree_set_page_dirty(struct page *page) |
1015 | { | |
bb146eb2 | 1016 | #ifdef DEBUG |
0b32f4bb JB |
1017 | struct extent_buffer *eb; |
1018 | ||
1019 | BUG_ON(!PagePrivate(page)); | |
1020 | eb = (struct extent_buffer *)page->private; | |
1021 | BUG_ON(!eb); | |
1022 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
1023 | BUG_ON(!atomic_read(&eb->refs)); | |
1024 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 1025 | #endif |
0b32f4bb JB |
1026 | return __set_page_dirty_nobuffers(page); |
1027 | } | |
1028 | ||
7f09410b | 1029 | static const struct address_space_operations btree_aops = { |
d98237b3 | 1030 | .readpage = btree_readpage, |
0da5468f | 1031 | .writepages = btree_writepages, |
5f39d397 CM |
1032 | .releasepage = btree_releasepage, |
1033 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1034 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1035 | .migratepage = btree_migratepage, |
5a92bc88 | 1036 | #endif |
0b32f4bb | 1037 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1038 | }; |
1039 | ||
ca7a79ad CM |
1040 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1041 | u64 parent_transid) | |
090d1875 | 1042 | { |
5f39d397 CM |
1043 | struct extent_buffer *buf = NULL; |
1044 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1045 | int ret = 0; |
090d1875 | 1046 | |
db94535d | 1047 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1048 | if (!buf) |
090d1875 | 1049 | return 0; |
d1310b2e | 1050 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1051 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1052 | free_extent_buffer(buf); |
de428b63 | 1053 | return ret; |
090d1875 CM |
1054 | } |
1055 | ||
ab0fff03 AJ |
1056 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1057 | int mirror_num, struct extent_buffer **eb) | |
1058 | { | |
1059 | struct extent_buffer *buf = NULL; | |
1060 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1061 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1062 | int ret; | |
1063 | ||
1064 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1065 | if (!buf) | |
1066 | return 0; | |
1067 | ||
1068 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1069 | ||
1070 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1071 | btree_get_extent, mirror_num); | |
1072 | if (ret) { | |
1073 | free_extent_buffer(buf); | |
1074 | return ret; | |
1075 | } | |
1076 | ||
1077 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1078 | free_extent_buffer(buf); | |
1079 | return -EIO; | |
0b32f4bb | 1080 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1081 | *eb = buf; |
1082 | } else { | |
1083 | free_extent_buffer(buf); | |
1084 | } | |
1085 | return 0; | |
1086 | } | |
1087 | ||
0999df54 CM |
1088 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1089 | u64 bytenr, u32 blocksize) | |
1090 | { | |
f28491e0 | 1091 | return find_extent_buffer(root->fs_info, bytenr); |
0999df54 CM |
1092 | } |
1093 | ||
1094 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1095 | u64 bytenr, u32 blocksize) | |
1096 | { | |
f28491e0 | 1097 | return alloc_extent_buffer(root->fs_info, bytenr, blocksize); |
0999df54 CM |
1098 | } |
1099 | ||
1100 | ||
e02119d5 CM |
1101 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1102 | { | |
727011e0 | 1103 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1104 | buf->start + buf->len - 1); |
e02119d5 CM |
1105 | } |
1106 | ||
1107 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1108 | { | |
727011e0 | 1109 | return filemap_fdatawait_range(buf->pages[0]->mapping, |
8aa38c31 | 1110 | buf->start, buf->start + buf->len - 1); |
e02119d5 CM |
1111 | } |
1112 | ||
0999df54 | 1113 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1114 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1115 | { |
1116 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1117 | int ret; |
1118 | ||
0999df54 CM |
1119 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1120 | if (!buf) | |
1121 | return NULL; | |
0999df54 | 1122 | |
ca7a79ad | 1123 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
0f0fe8f7 FDBM |
1124 | if (ret) { |
1125 | free_extent_buffer(buf); | |
1126 | return NULL; | |
1127 | } | |
5f39d397 | 1128 | return buf; |
ce9adaa5 | 1129 | |
eb60ceac CM |
1130 | } |
1131 | ||
d5c13f92 JM |
1132 | void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
1133 | struct extent_buffer *buf) | |
ed2ff2cb | 1134 | { |
e2d84521 MX |
1135 | struct btrfs_fs_info *fs_info = root->fs_info; |
1136 | ||
55c69072 | 1137 | if (btrfs_header_generation(buf) == |
e2d84521 | 1138 | fs_info->running_transaction->transid) { |
b9447ef8 | 1139 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1140 | |
b9473439 | 1141 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
e2d84521 MX |
1142 | __percpu_counter_add(&fs_info->dirty_metadata_bytes, |
1143 | -buf->len, | |
1144 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1145 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1146 | btrfs_set_lock_blocking(buf); | |
1147 | clear_extent_buffer_dirty(buf); | |
1148 | } | |
925baedd | 1149 | } |
5f39d397 CM |
1150 | } |
1151 | ||
8257b2dc MX |
1152 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1153 | { | |
1154 | struct btrfs_subvolume_writers *writers; | |
1155 | int ret; | |
1156 | ||
1157 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1158 | if (!writers) | |
1159 | return ERR_PTR(-ENOMEM); | |
1160 | ||
1161 | ret = percpu_counter_init(&writers->counter, 0); | |
1162 | if (ret < 0) { | |
1163 | kfree(writers); | |
1164 | return ERR_PTR(ret); | |
1165 | } | |
1166 | ||
1167 | init_waitqueue_head(&writers->wait); | |
1168 | return writers; | |
1169 | } | |
1170 | ||
1171 | static void | |
1172 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1173 | { | |
1174 | percpu_counter_destroy(&writers->counter); | |
1175 | kfree(writers); | |
1176 | } | |
1177 | ||
143bede5 JM |
1178 | static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
1179 | u32 stripesize, struct btrfs_root *root, | |
1180 | struct btrfs_fs_info *fs_info, | |
1181 | u64 objectid) | |
d97e63b6 | 1182 | { |
cfaa7295 | 1183 | root->node = NULL; |
a28ec197 | 1184 | root->commit_root = NULL; |
db94535d CM |
1185 | root->sectorsize = sectorsize; |
1186 | root->nodesize = nodesize; | |
1187 | root->leafsize = leafsize; | |
87ee04eb | 1188 | root->stripesize = stripesize; |
123abc88 | 1189 | root->ref_cows = 0; |
0b86a832 | 1190 | root->track_dirty = 0; |
c71bf099 | 1191 | root->in_radix = 0; |
d68fc57b YZ |
1192 | root->orphan_item_inserted = 0; |
1193 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1194 | |
0f7d52f4 CM |
1195 | root->objectid = objectid; |
1196 | root->last_trans = 0; | |
13a8a7c8 | 1197 | root->highest_objectid = 0; |
eb73c1b7 | 1198 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1199 | root->nr_ordered_extents = 0; |
58176a96 | 1200 | root->name = NULL; |
6bef4d31 | 1201 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1202 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1203 | root->block_rsv = NULL; |
d68fc57b | 1204 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1205 | |
1206 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1207 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1208 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1209 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1210 | INIT_LIST_HEAD(&root->ordered_extents); |
1211 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1212 | INIT_LIST_HEAD(&root->logged_list[0]); |
1213 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1214 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1215 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1216 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1217 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1218 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1219 | spin_lock_init(&root->log_extents_lock[0]); |
1220 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1221 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1222 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1223 | init_waitqueue_head(&root->log_writer_wait); |
1224 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1225 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1226 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1227 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1228 | atomic_set(&root->log_commit[0], 0); |
1229 | atomic_set(&root->log_commit[1], 0); | |
1230 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1231 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1232 | atomic_set(&root->orphan_inodes, 0); |
b0feb9d9 | 1233 | atomic_set(&root->refs, 1); |
8257b2dc | 1234 | atomic_set(&root->will_be_snapshoted, 0); |
7237f183 | 1235 | root->log_transid = 0; |
d1433deb | 1236 | root->log_transid_committed = -1; |
257c62e1 | 1237 | root->last_log_commit = 0; |
06ea65a3 JB |
1238 | if (fs_info) |
1239 | extent_io_tree_init(&root->dirty_log_pages, | |
1240 | fs_info->btree_inode->i_mapping); | |
017e5369 | 1241 | |
3768f368 CM |
1242 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1243 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1244 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1245 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
06ea65a3 JB |
1246 | if (fs_info) |
1247 | root->defrag_trans_start = fs_info->generation; | |
1248 | else | |
1249 | root->defrag_trans_start = 0; | |
58176a96 | 1250 | init_completion(&root->kobj_unregister); |
6702ed49 | 1251 | root->defrag_running = 0; |
4d775673 | 1252 | root->root_key.objectid = objectid; |
0ee5dc67 | 1253 | root->anon_dev = 0; |
8ea05e3a | 1254 | |
5f3ab90a | 1255 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1256 | } |
1257 | ||
f84a8bd6 | 1258 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) |
6f07e42e AV |
1259 | { |
1260 | struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); | |
1261 | if (root) | |
1262 | root->fs_info = fs_info; | |
1263 | return root; | |
1264 | } | |
1265 | ||
06ea65a3 JB |
1266 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1267 | /* Should only be used by the testing infrastructure */ | |
1268 | struct btrfs_root *btrfs_alloc_dummy_root(void) | |
1269 | { | |
1270 | struct btrfs_root *root; | |
1271 | ||
1272 | root = btrfs_alloc_root(NULL); | |
1273 | if (!root) | |
1274 | return ERR_PTR(-ENOMEM); | |
1275 | __setup_root(4096, 4096, 4096, 4096, root, NULL, 1); | |
1276 | root->dummy_root = 1; | |
1277 | ||
1278 | return root; | |
1279 | } | |
1280 | #endif | |
1281 | ||
20897f5c AJ |
1282 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1283 | struct btrfs_fs_info *fs_info, | |
1284 | u64 objectid) | |
1285 | { | |
1286 | struct extent_buffer *leaf; | |
1287 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1288 | struct btrfs_root *root; | |
1289 | struct btrfs_key key; | |
1290 | int ret = 0; | |
6463fe58 | 1291 | uuid_le uuid; |
20897f5c AJ |
1292 | |
1293 | root = btrfs_alloc_root(fs_info); | |
1294 | if (!root) | |
1295 | return ERR_PTR(-ENOMEM); | |
1296 | ||
1297 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1298 | tree_root->sectorsize, tree_root->stripesize, | |
1299 | root, fs_info, objectid); | |
1300 | root->root_key.objectid = objectid; | |
1301 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1302 | root->root_key.offset = 0; | |
1303 | ||
1304 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, | |
1305 | 0, objectid, NULL, 0, 0, 0); | |
1306 | if (IS_ERR(leaf)) { | |
1307 | ret = PTR_ERR(leaf); | |
1dd05682 | 1308 | leaf = NULL; |
20897f5c AJ |
1309 | goto fail; |
1310 | } | |
1311 | ||
20897f5c AJ |
1312 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1313 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1314 | btrfs_set_header_generation(leaf, trans->transid); | |
1315 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1316 | btrfs_set_header_owner(leaf, objectid); | |
1317 | root->node = leaf; | |
1318 | ||
0a4e5586 | 1319 | write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), |
20897f5c AJ |
1320 | BTRFS_FSID_SIZE); |
1321 | write_extent_buffer(leaf, fs_info->chunk_tree_uuid, | |
b308bc2f | 1322 | btrfs_header_chunk_tree_uuid(leaf), |
20897f5c AJ |
1323 | BTRFS_UUID_SIZE); |
1324 | btrfs_mark_buffer_dirty(leaf); | |
1325 | ||
1326 | root->commit_root = btrfs_root_node(root); | |
1327 | root->track_dirty = 1; | |
1328 | ||
1329 | ||
1330 | root->root_item.flags = 0; | |
1331 | root->root_item.byte_limit = 0; | |
1332 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1333 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1334 | btrfs_set_root_level(&root->root_item, 0); | |
1335 | btrfs_set_root_refs(&root->root_item, 1); | |
1336 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1337 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1338 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1339 | uuid_le_gen(&uuid); |
1340 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1341 | root->root_item.drop_level = 0; |
1342 | ||
1343 | key.objectid = objectid; | |
1344 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1345 | key.offset = 0; | |
1346 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1347 | if (ret) | |
1348 | goto fail; | |
1349 | ||
1350 | btrfs_tree_unlock(leaf); | |
1351 | ||
1dd05682 TI |
1352 | return root; |
1353 | ||
20897f5c | 1354 | fail: |
1dd05682 TI |
1355 | if (leaf) { |
1356 | btrfs_tree_unlock(leaf); | |
1357 | free_extent_buffer(leaf); | |
1358 | } | |
1359 | kfree(root); | |
20897f5c | 1360 | |
1dd05682 | 1361 | return ERR_PTR(ret); |
20897f5c AJ |
1362 | } |
1363 | ||
7237f183 YZ |
1364 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1365 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1366 | { |
1367 | struct btrfs_root *root; | |
1368 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1369 | struct extent_buffer *leaf; |
e02119d5 | 1370 | |
6f07e42e | 1371 | root = btrfs_alloc_root(fs_info); |
e02119d5 | 1372 | if (!root) |
7237f183 | 1373 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1374 | |
1375 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1376 | tree_root->sectorsize, tree_root->stripesize, | |
1377 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1378 | ||
1379 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1380 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1381 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1382 | /* |
1383 | * log trees do not get reference counted because they go away | |
1384 | * before a real commit is actually done. They do store pointers | |
1385 | * to file data extents, and those reference counts still get | |
1386 | * updated (along with back refs to the log tree). | |
1387 | */ | |
e02119d5 CM |
1388 | root->ref_cows = 0; |
1389 | ||
5d4f98a2 | 1390 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 | 1391 | BTRFS_TREE_LOG_OBJECTID, NULL, |
5581a51a | 1392 | 0, 0, 0); |
7237f183 YZ |
1393 | if (IS_ERR(leaf)) { |
1394 | kfree(root); | |
1395 | return ERR_CAST(leaf); | |
1396 | } | |
e02119d5 | 1397 | |
5d4f98a2 YZ |
1398 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1399 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1400 | btrfs_set_header_generation(leaf, trans->transid); | |
1401 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1402 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1403 | root->node = leaf; |
e02119d5 CM |
1404 | |
1405 | write_extent_buffer(root->node, root->fs_info->fsid, | |
0a4e5586 | 1406 | btrfs_header_fsid(), BTRFS_FSID_SIZE); |
e02119d5 CM |
1407 | btrfs_mark_buffer_dirty(root->node); |
1408 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1409 | return root; |
1410 | } | |
1411 | ||
1412 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1413 | struct btrfs_fs_info *fs_info) | |
1414 | { | |
1415 | struct btrfs_root *log_root; | |
1416 | ||
1417 | log_root = alloc_log_tree(trans, fs_info); | |
1418 | if (IS_ERR(log_root)) | |
1419 | return PTR_ERR(log_root); | |
1420 | WARN_ON(fs_info->log_root_tree); | |
1421 | fs_info->log_root_tree = log_root; | |
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1426 | struct btrfs_root *root) | |
1427 | { | |
1428 | struct btrfs_root *log_root; | |
1429 | struct btrfs_inode_item *inode_item; | |
1430 | ||
1431 | log_root = alloc_log_tree(trans, root->fs_info); | |
1432 | if (IS_ERR(log_root)) | |
1433 | return PTR_ERR(log_root); | |
1434 | ||
1435 | log_root->last_trans = trans->transid; | |
1436 | log_root->root_key.offset = root->root_key.objectid; | |
1437 | ||
1438 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1439 | btrfs_set_stack_inode_generation(inode_item, 1); |
1440 | btrfs_set_stack_inode_size(inode_item, 3); | |
1441 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
1442 | btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); | |
1443 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); | |
7237f183 | 1444 | |
5d4f98a2 | 1445 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1446 | |
1447 | WARN_ON(root->log_root); | |
1448 | root->log_root = log_root; | |
1449 | root->log_transid = 0; | |
d1433deb | 1450 | root->log_transid_committed = -1; |
257c62e1 | 1451 | root->last_log_commit = 0; |
e02119d5 CM |
1452 | return 0; |
1453 | } | |
1454 | ||
35a3621b SB |
1455 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1456 | struct btrfs_key *key) | |
e02119d5 CM |
1457 | { |
1458 | struct btrfs_root *root; | |
1459 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1460 | struct btrfs_path *path; |
84234f3a | 1461 | u64 generation; |
db94535d | 1462 | u32 blocksize; |
cb517eab | 1463 | int ret; |
0f7d52f4 | 1464 | |
cb517eab MX |
1465 | path = btrfs_alloc_path(); |
1466 | if (!path) | |
0f7d52f4 | 1467 | return ERR_PTR(-ENOMEM); |
cb517eab MX |
1468 | |
1469 | root = btrfs_alloc_root(fs_info); | |
1470 | if (!root) { | |
1471 | ret = -ENOMEM; | |
1472 | goto alloc_fail; | |
0f7d52f4 CM |
1473 | } |
1474 | ||
db94535d | 1475 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb | 1476 | tree_root->sectorsize, tree_root->stripesize, |
cb517eab | 1477 | root, fs_info, key->objectid); |
0f7d52f4 | 1478 | |
cb517eab MX |
1479 | ret = btrfs_find_root(tree_root, key, path, |
1480 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1481 | if (ret) { |
13a8a7c8 YZ |
1482 | if (ret > 0) |
1483 | ret = -ENOENT; | |
cb517eab | 1484 | goto find_fail; |
0f7d52f4 | 1485 | } |
13a8a7c8 | 1486 | |
84234f3a | 1487 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1488 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1489 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1490 | blocksize, generation); |
cb517eab MX |
1491 | if (!root->node) { |
1492 | ret = -ENOMEM; | |
1493 | goto find_fail; | |
1494 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1495 | ret = -EIO; | |
1496 | goto read_fail; | |
416bc658 | 1497 | } |
5d4f98a2 | 1498 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1499 | out: |
cb517eab MX |
1500 | btrfs_free_path(path); |
1501 | return root; | |
1502 | ||
1503 | read_fail: | |
1504 | free_extent_buffer(root->node); | |
1505 | find_fail: | |
1506 | kfree(root); | |
1507 | alloc_fail: | |
1508 | root = ERR_PTR(ret); | |
1509 | goto out; | |
1510 | } | |
1511 | ||
1512 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1513 | struct btrfs_key *location) | |
1514 | { | |
1515 | struct btrfs_root *root; | |
1516 | ||
1517 | root = btrfs_read_tree_root(tree_root, location); | |
1518 | if (IS_ERR(root)) | |
1519 | return root; | |
1520 | ||
1521 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
e02119d5 | 1522 | root->ref_cows = 1; |
08fe4db1 LZ |
1523 | btrfs_check_and_init_root_item(&root->root_item); |
1524 | } | |
13a8a7c8 | 1525 | |
5eda7b5e CM |
1526 | return root; |
1527 | } | |
1528 | ||
cb517eab MX |
1529 | int btrfs_init_fs_root(struct btrfs_root *root) |
1530 | { | |
1531 | int ret; | |
8257b2dc | 1532 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1533 | |
1534 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1535 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1536 | GFP_NOFS); | |
1537 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1538 | ret = -ENOMEM; | |
1539 | goto fail; | |
1540 | } | |
1541 | ||
8257b2dc MX |
1542 | writers = btrfs_alloc_subvolume_writers(); |
1543 | if (IS_ERR(writers)) { | |
1544 | ret = PTR_ERR(writers); | |
1545 | goto fail; | |
1546 | } | |
1547 | root->subv_writers = writers; | |
1548 | ||
cb517eab MX |
1549 | btrfs_init_free_ino_ctl(root); |
1550 | mutex_init(&root->fs_commit_mutex); | |
1551 | spin_lock_init(&root->cache_lock); | |
1552 | init_waitqueue_head(&root->cache_wait); | |
1553 | ||
1554 | ret = get_anon_bdev(&root->anon_dev); | |
1555 | if (ret) | |
8257b2dc | 1556 | goto free_writers; |
cb517eab | 1557 | return 0; |
8257b2dc MX |
1558 | |
1559 | free_writers: | |
1560 | btrfs_free_subvolume_writers(root->subv_writers); | |
cb517eab MX |
1561 | fail: |
1562 | kfree(root->free_ino_ctl); | |
1563 | kfree(root->free_ino_pinned); | |
1564 | return ret; | |
1565 | } | |
1566 | ||
171170c1 ST |
1567 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1568 | u64 root_id) | |
cb517eab MX |
1569 | { |
1570 | struct btrfs_root *root; | |
1571 | ||
1572 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1573 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1574 | (unsigned long)root_id); | |
1575 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1576 | return root; | |
1577 | } | |
1578 | ||
1579 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1580 | struct btrfs_root *root) | |
1581 | { | |
1582 | int ret; | |
1583 | ||
1584 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); | |
1585 | if (ret) | |
1586 | return ret; | |
1587 | ||
1588 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1589 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1590 | (unsigned long)root->root_key.objectid, | |
1591 | root); | |
1592 | if (ret == 0) | |
1593 | root->in_radix = 1; | |
1594 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1595 | radix_tree_preload_end(); | |
1596 | ||
1597 | return ret; | |
1598 | } | |
1599 | ||
c00869f1 MX |
1600 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1601 | struct btrfs_key *location, | |
1602 | bool check_ref) | |
5eda7b5e CM |
1603 | { |
1604 | struct btrfs_root *root; | |
1605 | int ret; | |
1606 | ||
edbd8d4e CM |
1607 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1608 | return fs_info->tree_root; | |
1609 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1610 | return fs_info->extent_root; | |
8f18cf13 CM |
1611 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1612 | return fs_info->chunk_root; | |
1613 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1614 | return fs_info->dev_root; | |
0403e47e YZ |
1615 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1616 | return fs_info->csum_root; | |
bcef60f2 AJ |
1617 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1618 | return fs_info->quota_root ? fs_info->quota_root : | |
1619 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1620 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1621 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1622 | ERR_PTR(-ENOENT); | |
4df27c4d | 1623 | again: |
cb517eab | 1624 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1625 | if (root) { |
c00869f1 | 1626 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1627 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1628 | return root; |
48475471 | 1629 | } |
5eda7b5e | 1630 | |
cb517eab | 1631 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1632 | if (IS_ERR(root)) |
1633 | return root; | |
3394e160 | 1634 | |
c00869f1 | 1635 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1636 | ret = -ENOENT; |
581bb050 | 1637 | goto fail; |
35a30d7c | 1638 | } |
581bb050 | 1639 | |
cb517eab | 1640 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1641 | if (ret) |
1642 | goto fail; | |
3394e160 | 1643 | |
3f870c28 KN |
1644 | ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, |
1645 | location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); | |
d68fc57b YZ |
1646 | if (ret < 0) |
1647 | goto fail; | |
1648 | if (ret == 0) | |
1649 | root->orphan_item_inserted = 1; | |
1650 | ||
cb517eab | 1651 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1652 | if (ret) { |
4df27c4d YZ |
1653 | if (ret == -EEXIST) { |
1654 | free_fs_root(root); | |
1655 | goto again; | |
1656 | } | |
1657 | goto fail; | |
0f7d52f4 | 1658 | } |
edbd8d4e | 1659 | return root; |
4df27c4d YZ |
1660 | fail: |
1661 | free_fs_root(root); | |
1662 | return ERR_PTR(ret); | |
edbd8d4e CM |
1663 | } |
1664 | ||
04160088 CM |
1665 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1666 | { | |
1667 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1668 | int ret = 0; | |
04160088 CM |
1669 | struct btrfs_device *device; |
1670 | struct backing_dev_info *bdi; | |
b7967db7 | 1671 | |
1f78160c XG |
1672 | rcu_read_lock(); |
1673 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1674 | if (!device->bdev) |
1675 | continue; | |
04160088 CM |
1676 | bdi = blk_get_backing_dev_info(device->bdev); |
1677 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1678 | ret = 1; | |
1679 | break; | |
1680 | } | |
1681 | } | |
1f78160c | 1682 | rcu_read_unlock(); |
04160088 CM |
1683 | return ret; |
1684 | } | |
1685 | ||
ad081f14 JA |
1686 | /* |
1687 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1688 | * bdi again. | |
1689 | */ | |
04160088 CM |
1690 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1691 | { | |
ad081f14 JA |
1692 | int err; |
1693 | ||
1694 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1695 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1696 | if (err) |
1697 | return err; | |
1698 | ||
4575c9cc | 1699 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1700 | bdi->congested_fn = btrfs_congested_fn; |
1701 | bdi->congested_data = info; | |
1702 | return 0; | |
1703 | } | |
1704 | ||
8b712842 CM |
1705 | /* |
1706 | * called by the kthread helper functions to finally call the bio end_io | |
1707 | * functions. This is where read checksum verification actually happens | |
1708 | */ | |
d458b054 | 1709 | static void end_workqueue_fn(struct btrfs_work *work) |
ce9adaa5 | 1710 | { |
ce9adaa5 | 1711 | struct bio *bio; |
8b712842 | 1712 | struct end_io_wq *end_io_wq; |
ce9adaa5 | 1713 | int error; |
ce9adaa5 | 1714 | |
8b712842 CM |
1715 | end_io_wq = container_of(work, struct end_io_wq, work); |
1716 | bio = end_io_wq->bio; | |
ce9adaa5 | 1717 | |
8b712842 CM |
1718 | error = end_io_wq->error; |
1719 | bio->bi_private = end_io_wq->private; | |
1720 | bio->bi_end_io = end_io_wq->end_io; | |
1721 | kfree(end_io_wq); | |
8b712842 | 1722 | bio_endio(bio, error); |
44b8bd7e CM |
1723 | } |
1724 | ||
a74a4b97 CM |
1725 | static int cleaner_kthread(void *arg) |
1726 | { | |
1727 | struct btrfs_root *root = arg; | |
d0278245 | 1728 | int again; |
a74a4b97 CM |
1729 | |
1730 | do { | |
d0278245 | 1731 | again = 0; |
a74a4b97 | 1732 | |
d0278245 | 1733 | /* Make the cleaner go to sleep early. */ |
babbf170 | 1734 | if (btrfs_need_cleaner_sleep(root)) |
d0278245 MX |
1735 | goto sleep; |
1736 | ||
1737 | if (!mutex_trylock(&root->fs_info->cleaner_mutex)) | |
1738 | goto sleep; | |
1739 | ||
dc7f370c MX |
1740 | /* |
1741 | * Avoid the problem that we change the status of the fs | |
1742 | * during the above check and trylock. | |
1743 | */ | |
babbf170 | 1744 | if (btrfs_need_cleaner_sleep(root)) { |
dc7f370c MX |
1745 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1746 | goto sleep; | |
76dda93c | 1747 | } |
a74a4b97 | 1748 | |
d0278245 MX |
1749 | btrfs_run_delayed_iputs(root); |
1750 | again = btrfs_clean_one_deleted_snapshot(root); | |
1751 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
1752 | ||
1753 | /* | |
05323cd1 MX |
1754 | * The defragger has dealt with the R/O remount and umount, |
1755 | * needn't do anything special here. | |
d0278245 MX |
1756 | */ |
1757 | btrfs_run_defrag_inodes(root->fs_info); | |
1758 | sleep: | |
9d1a2a3a | 1759 | if (!try_to_freeze() && !again) { |
a74a4b97 | 1760 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1761 | if (!kthread_should_stop()) |
1762 | schedule(); | |
a74a4b97 CM |
1763 | __set_current_state(TASK_RUNNING); |
1764 | } | |
1765 | } while (!kthread_should_stop()); | |
1766 | return 0; | |
1767 | } | |
1768 | ||
1769 | static int transaction_kthread(void *arg) | |
1770 | { | |
1771 | struct btrfs_root *root = arg; | |
1772 | struct btrfs_trans_handle *trans; | |
1773 | struct btrfs_transaction *cur; | |
8929ecfa | 1774 | u64 transid; |
a74a4b97 CM |
1775 | unsigned long now; |
1776 | unsigned long delay; | |
914b2007 | 1777 | bool cannot_commit; |
a74a4b97 CM |
1778 | |
1779 | do { | |
914b2007 | 1780 | cannot_commit = false; |
8b87dc17 | 1781 | delay = HZ * root->fs_info->commit_interval; |
a74a4b97 CM |
1782 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1783 | ||
a4abeea4 | 1784 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1785 | cur = root->fs_info->running_transaction; |
1786 | if (!cur) { | |
a4abeea4 | 1787 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1788 | goto sleep; |
1789 | } | |
31153d81 | 1790 | |
a74a4b97 | 1791 | now = get_seconds(); |
4a9d8bde | 1792 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 DS |
1793 | (now < cur->start_time || |
1794 | now - cur->start_time < root->fs_info->commit_interval)) { | |
a4abeea4 | 1795 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1796 | delay = HZ * 5; |
1797 | goto sleep; | |
1798 | } | |
8929ecfa | 1799 | transid = cur->transid; |
a4abeea4 | 1800 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1801 | |
79787eaa | 1802 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1803 | trans = btrfs_attach_transaction(root); |
914b2007 | 1804 | if (IS_ERR(trans)) { |
354aa0fb MX |
1805 | if (PTR_ERR(trans) != -ENOENT) |
1806 | cannot_commit = true; | |
79787eaa | 1807 | goto sleep; |
914b2007 | 1808 | } |
8929ecfa | 1809 | if (transid == trans->transid) { |
79787eaa | 1810 | btrfs_commit_transaction(trans, root); |
8929ecfa YZ |
1811 | } else { |
1812 | btrfs_end_transaction(trans, root); | |
1813 | } | |
a74a4b97 CM |
1814 | sleep: |
1815 | wake_up_process(root->fs_info->cleaner_kthread); | |
1816 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1817 | ||
4e121c06 JB |
1818 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
1819 | &root->fs_info->fs_state))) | |
1820 | btrfs_cleanup_transaction(root); | |
a0acae0e | 1821 | if (!try_to_freeze()) { |
a74a4b97 | 1822 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa | 1823 | if (!kthread_should_stop() && |
914b2007 JK |
1824 | (!btrfs_transaction_blocked(root->fs_info) || |
1825 | cannot_commit)) | |
8929ecfa | 1826 | schedule_timeout(delay); |
a74a4b97 CM |
1827 | __set_current_state(TASK_RUNNING); |
1828 | } | |
1829 | } while (!kthread_should_stop()); | |
1830 | return 0; | |
1831 | } | |
1832 | ||
af31f5e5 CM |
1833 | /* |
1834 | * this will find the highest generation in the array of | |
1835 | * root backups. The index of the highest array is returned, | |
1836 | * or -1 if we can't find anything. | |
1837 | * | |
1838 | * We check to make sure the array is valid by comparing the | |
1839 | * generation of the latest root in the array with the generation | |
1840 | * in the super block. If they don't match we pitch it. | |
1841 | */ | |
1842 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1843 | { | |
1844 | u64 cur; | |
1845 | int newest_index = -1; | |
1846 | struct btrfs_root_backup *root_backup; | |
1847 | int i; | |
1848 | ||
1849 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1850 | root_backup = info->super_copy->super_roots + i; | |
1851 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1852 | if (cur == newest_gen) | |
1853 | newest_index = i; | |
1854 | } | |
1855 | ||
1856 | /* check to see if we actually wrapped around */ | |
1857 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1858 | root_backup = info->super_copy->super_roots; | |
1859 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1860 | if (cur == newest_gen) | |
1861 | newest_index = 0; | |
1862 | } | |
1863 | return newest_index; | |
1864 | } | |
1865 | ||
1866 | ||
1867 | /* | |
1868 | * find the oldest backup so we know where to store new entries | |
1869 | * in the backup array. This will set the backup_root_index | |
1870 | * field in the fs_info struct | |
1871 | */ | |
1872 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1873 | u64 newest_gen) | |
1874 | { | |
1875 | int newest_index = -1; | |
1876 | ||
1877 | newest_index = find_newest_super_backup(info, newest_gen); | |
1878 | /* if there was garbage in there, just move along */ | |
1879 | if (newest_index == -1) { | |
1880 | info->backup_root_index = 0; | |
1881 | } else { | |
1882 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1883 | } | |
1884 | } | |
1885 | ||
1886 | /* | |
1887 | * copy all the root pointers into the super backup array. | |
1888 | * this will bump the backup pointer by one when it is | |
1889 | * done | |
1890 | */ | |
1891 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1892 | { | |
1893 | int next_backup; | |
1894 | struct btrfs_root_backup *root_backup; | |
1895 | int last_backup; | |
1896 | ||
1897 | next_backup = info->backup_root_index; | |
1898 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1899 | BTRFS_NUM_BACKUP_ROOTS; | |
1900 | ||
1901 | /* | |
1902 | * just overwrite the last backup if we're at the same generation | |
1903 | * this happens only at umount | |
1904 | */ | |
1905 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1906 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1907 | btrfs_header_generation(info->tree_root->node)) | |
1908 | next_backup = last_backup; | |
1909 | ||
1910 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1911 | ||
1912 | /* | |
1913 | * make sure all of our padding and empty slots get zero filled | |
1914 | * regardless of which ones we use today | |
1915 | */ | |
1916 | memset(root_backup, 0, sizeof(*root_backup)); | |
1917 | ||
1918 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1919 | ||
1920 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1921 | btrfs_set_backup_tree_root_gen(root_backup, | |
1922 | btrfs_header_generation(info->tree_root->node)); | |
1923 | ||
1924 | btrfs_set_backup_tree_root_level(root_backup, | |
1925 | btrfs_header_level(info->tree_root->node)); | |
1926 | ||
1927 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1928 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1929 | btrfs_header_generation(info->chunk_root->node)); | |
1930 | btrfs_set_backup_chunk_root_level(root_backup, | |
1931 | btrfs_header_level(info->chunk_root->node)); | |
1932 | ||
1933 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1934 | btrfs_set_backup_extent_root_gen(root_backup, | |
1935 | btrfs_header_generation(info->extent_root->node)); | |
1936 | btrfs_set_backup_extent_root_level(root_backup, | |
1937 | btrfs_header_level(info->extent_root->node)); | |
1938 | ||
7c7e82a7 CM |
1939 | /* |
1940 | * we might commit during log recovery, which happens before we set | |
1941 | * the fs_root. Make sure it is valid before we fill it in. | |
1942 | */ | |
1943 | if (info->fs_root && info->fs_root->node) { | |
1944 | btrfs_set_backup_fs_root(root_backup, | |
1945 | info->fs_root->node->start); | |
1946 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1947 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1948 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1949 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1950 | } |
af31f5e5 CM |
1951 | |
1952 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1953 | btrfs_set_backup_dev_root_gen(root_backup, | |
1954 | btrfs_header_generation(info->dev_root->node)); | |
1955 | btrfs_set_backup_dev_root_level(root_backup, | |
1956 | btrfs_header_level(info->dev_root->node)); | |
1957 | ||
1958 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1959 | btrfs_set_backup_csum_root_gen(root_backup, | |
1960 | btrfs_header_generation(info->csum_root->node)); | |
1961 | btrfs_set_backup_csum_root_level(root_backup, | |
1962 | btrfs_header_level(info->csum_root->node)); | |
1963 | ||
1964 | btrfs_set_backup_total_bytes(root_backup, | |
1965 | btrfs_super_total_bytes(info->super_copy)); | |
1966 | btrfs_set_backup_bytes_used(root_backup, | |
1967 | btrfs_super_bytes_used(info->super_copy)); | |
1968 | btrfs_set_backup_num_devices(root_backup, | |
1969 | btrfs_super_num_devices(info->super_copy)); | |
1970 | ||
1971 | /* | |
1972 | * if we don't copy this out to the super_copy, it won't get remembered | |
1973 | * for the next commit | |
1974 | */ | |
1975 | memcpy(&info->super_copy->super_roots, | |
1976 | &info->super_for_commit->super_roots, | |
1977 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1978 | } | |
1979 | ||
1980 | /* | |
1981 | * this copies info out of the root backup array and back into | |
1982 | * the in-memory super block. It is meant to help iterate through | |
1983 | * the array, so you send it the number of backups you've already | |
1984 | * tried and the last backup index you used. | |
1985 | * | |
1986 | * this returns -1 when it has tried all the backups | |
1987 | */ | |
1988 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1989 | struct btrfs_super_block *super, | |
1990 | int *num_backups_tried, int *backup_index) | |
1991 | { | |
1992 | struct btrfs_root_backup *root_backup; | |
1993 | int newest = *backup_index; | |
1994 | ||
1995 | if (*num_backups_tried == 0) { | |
1996 | u64 gen = btrfs_super_generation(super); | |
1997 | ||
1998 | newest = find_newest_super_backup(info, gen); | |
1999 | if (newest == -1) | |
2000 | return -1; | |
2001 | ||
2002 | *backup_index = newest; | |
2003 | *num_backups_tried = 1; | |
2004 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2005 | /* we've tried all the backups, all done */ | |
2006 | return -1; | |
2007 | } else { | |
2008 | /* jump to the next oldest backup */ | |
2009 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2010 | BTRFS_NUM_BACKUP_ROOTS; | |
2011 | *backup_index = newest; | |
2012 | *num_backups_tried += 1; | |
2013 | } | |
2014 | root_backup = super->super_roots + newest; | |
2015 | ||
2016 | btrfs_set_super_generation(super, | |
2017 | btrfs_backup_tree_root_gen(root_backup)); | |
2018 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2019 | btrfs_set_super_root_level(super, | |
2020 | btrfs_backup_tree_root_level(root_backup)); | |
2021 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2022 | ||
2023 | /* | |
2024 | * fixme: the total bytes and num_devices need to match or we should | |
2025 | * need a fsck | |
2026 | */ | |
2027 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2028 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2029 | return 0; | |
2030 | } | |
2031 | ||
7abadb64 LB |
2032 | /* helper to cleanup workers */ |
2033 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2034 | { | |
dc6e3209 | 2035 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2036 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2037 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 QW |
2038 | btrfs_destroy_workqueue(fs_info->endio_workers); |
2039 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2040 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); | |
d05a33ac | 2041 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2042 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); |
2043 | btrfs_destroy_workqueue(fs_info->endio_write_workers); | |
2044 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2045 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2046 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2047 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2048 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2049 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2050 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
7abadb64 LB |
2051 | } |
2052 | ||
2e9f5954 R |
2053 | static void free_root_extent_buffers(struct btrfs_root *root) |
2054 | { | |
2055 | if (root) { | |
2056 | free_extent_buffer(root->node); | |
2057 | free_extent_buffer(root->commit_root); | |
2058 | root->node = NULL; | |
2059 | root->commit_root = NULL; | |
2060 | } | |
2061 | } | |
2062 | ||
af31f5e5 CM |
2063 | /* helper to cleanup tree roots */ |
2064 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2065 | { | |
2e9f5954 | 2066 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2067 | |
2e9f5954 R |
2068 | free_root_extent_buffers(info->dev_root); |
2069 | free_root_extent_buffers(info->extent_root); | |
2070 | free_root_extent_buffers(info->csum_root); | |
2071 | free_root_extent_buffers(info->quota_root); | |
2072 | free_root_extent_buffers(info->uuid_root); | |
2073 | if (chunk_root) | |
2074 | free_root_extent_buffers(info->chunk_root); | |
af31f5e5 CM |
2075 | } |
2076 | ||
171f6537 JB |
2077 | static void del_fs_roots(struct btrfs_fs_info *fs_info) |
2078 | { | |
2079 | int ret; | |
2080 | struct btrfs_root *gang[8]; | |
2081 | int i; | |
2082 | ||
2083 | while (!list_empty(&fs_info->dead_roots)) { | |
2084 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2085 | struct btrfs_root, root_list); | |
2086 | list_del(&gang[0]->root_list); | |
2087 | ||
2088 | if (gang[0]->in_radix) { | |
cb517eab | 2089 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2090 | } else { |
2091 | free_extent_buffer(gang[0]->node); | |
2092 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2093 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2094 | } |
2095 | } | |
2096 | ||
2097 | while (1) { | |
2098 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2099 | (void **)gang, 0, | |
2100 | ARRAY_SIZE(gang)); | |
2101 | if (!ret) | |
2102 | break; | |
2103 | for (i = 0; i < ret; i++) | |
cb517eab | 2104 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2105 | } |
1a4319cc LB |
2106 | |
2107 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2108 | btrfs_free_log_root_tree(NULL, fs_info); | |
2109 | btrfs_destroy_pinned_extent(fs_info->tree_root, | |
2110 | fs_info->pinned_extents); | |
2111 | } | |
171f6537 | 2112 | } |
af31f5e5 | 2113 | |
ad2b2c80 AV |
2114 | int open_ctree(struct super_block *sb, |
2115 | struct btrfs_fs_devices *fs_devices, | |
2116 | char *options) | |
2e635a27 | 2117 | { |
db94535d CM |
2118 | u32 sectorsize; |
2119 | u32 nodesize; | |
2120 | u32 leafsize; | |
2121 | u32 blocksize; | |
87ee04eb | 2122 | u32 stripesize; |
84234f3a | 2123 | u64 generation; |
f2b636e8 | 2124 | u64 features; |
3de4586c | 2125 | struct btrfs_key location; |
a061fc8d | 2126 | struct buffer_head *bh; |
4d34b278 | 2127 | struct btrfs_super_block *disk_super; |
815745cf | 2128 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2129 | struct btrfs_root *tree_root; |
4d34b278 ID |
2130 | struct btrfs_root *extent_root; |
2131 | struct btrfs_root *csum_root; | |
2132 | struct btrfs_root *chunk_root; | |
2133 | struct btrfs_root *dev_root; | |
bcef60f2 | 2134 | struct btrfs_root *quota_root; |
f7a81ea4 | 2135 | struct btrfs_root *uuid_root; |
e02119d5 | 2136 | struct btrfs_root *log_tree_root; |
eb60ceac | 2137 | int ret; |
e58ca020 | 2138 | int err = -EINVAL; |
af31f5e5 CM |
2139 | int num_backups_tried = 0; |
2140 | int backup_index = 0; | |
5cdc7ad3 QW |
2141 | int max_active; |
2142 | int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; | |
70f80175 SB |
2143 | bool create_uuid_tree; |
2144 | bool check_uuid_tree; | |
4543df7e | 2145 | |
f84a8bd6 | 2146 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); |
6f07e42e | 2147 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); |
cb517eab | 2148 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2149 | err = -ENOMEM; |
2150 | goto fail; | |
2151 | } | |
76dda93c YZ |
2152 | |
2153 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2154 | if (ret) { | |
2155 | err = ret; | |
2156 | goto fail; | |
2157 | } | |
2158 | ||
2159 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
2160 | if (ret) { | |
2161 | err = ret; | |
2162 | goto fail_srcu; | |
2163 | } | |
2164 | ||
e2d84521 MX |
2165 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); |
2166 | if (ret) { | |
2167 | err = ret; | |
2168 | goto fail_bdi; | |
2169 | } | |
2170 | fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * | |
2171 | (1 + ilog2(nr_cpu_ids)); | |
2172 | ||
963d678b MX |
2173 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); |
2174 | if (ret) { | |
2175 | err = ret; | |
2176 | goto fail_dirty_metadata_bytes; | |
2177 | } | |
2178 | ||
c404e0dc MX |
2179 | ret = percpu_counter_init(&fs_info->bio_counter, 0); |
2180 | if (ret) { | |
2181 | err = ret; | |
2182 | goto fail_delalloc_bytes; | |
2183 | } | |
2184 | ||
76dda93c YZ |
2185 | fs_info->btree_inode = new_inode(sb); |
2186 | if (!fs_info->btree_inode) { | |
2187 | err = -ENOMEM; | |
c404e0dc | 2188 | goto fail_bio_counter; |
76dda93c YZ |
2189 | } |
2190 | ||
a6591715 | 2191 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 2192 | |
76dda93c | 2193 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2194 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2195 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2196 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2197 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2198 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2199 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2200 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2201 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2202 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2203 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2204 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 2205 | spin_lock_init(&fs_info->free_chunk_lock); |
f29021b2 | 2206 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2207 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2208 | spin_lock_init(&fs_info->buffer_lock); |
f29021b2 | 2209 | rwlock_init(&fs_info->tree_mod_log_lock); |
7585717f | 2210 | mutex_init(&fs_info->reloc_mutex); |
de98ced9 | 2211 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2212 | |
58176a96 | 2213 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 2214 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2215 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2216 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
0b86a832 | 2217 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2218 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2219 | BTRFS_BLOCK_RSV_GLOBAL); | |
2220 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, | |
2221 | BTRFS_BLOCK_RSV_DELALLOC); | |
2222 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); | |
2223 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2224 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2225 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2226 | BTRFS_BLOCK_RSV_DELOPS); | |
cb03c743 | 2227 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 2228 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 2229 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 2230 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 2231 | atomic_set(&fs_info->defrag_running, 0); |
fc36ed7e | 2232 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2233 | fs_info->sb = sb; |
6f568d35 | 2234 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 2235 | fs_info->metadata_ratio = 0; |
4cb5300b | 2236 | fs_info->defrag_inodes = RB_ROOT; |
2bf64758 | 2237 | fs_info->free_chunk_space = 0; |
f29021b2 | 2238 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2239 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
0a2b2a84 | 2240 | fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64); |
90519d66 AJ |
2241 | /* readahead state */ |
2242 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
2243 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 2244 | |
b34b086c CM |
2245 | fs_info->thread_pool_size = min_t(unsigned long, |
2246 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2247 | |
199c2a9c MX |
2248 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2249 | spin_lock_init(&fs_info->ordered_root_lock); | |
16cdcec7 MX |
2250 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
2251 | GFP_NOFS); | |
2252 | if (!fs_info->delayed_root) { | |
2253 | err = -ENOMEM; | |
2254 | goto fail_iput; | |
2255 | } | |
2256 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2257 | |
a2de733c AJ |
2258 | mutex_init(&fs_info->scrub_lock); |
2259 | atomic_set(&fs_info->scrubs_running, 0); | |
2260 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2261 | atomic_set(&fs_info->scrubs_paused, 0); | |
2262 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
c404e0dc | 2263 | init_waitqueue_head(&fs_info->replace_wait); |
a2de733c | 2264 | init_waitqueue_head(&fs_info->scrub_pause_wait); |
a2de733c | 2265 | fs_info->scrub_workers_refcnt = 0; |
21adbd5c SB |
2266 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2267 | fs_info->check_integrity_print_mask = 0; | |
2268 | #endif | |
a2de733c | 2269 | |
c9e9f97b ID |
2270 | spin_lock_init(&fs_info->balance_lock); |
2271 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2272 | atomic_set(&fs_info->balance_running, 0); |
2273 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2274 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2275 | fs_info->balance_ctl = NULL; |
837d5b6e | 2276 | init_waitqueue_head(&fs_info->balance_wait_q); |
a2de733c | 2277 | |
a061fc8d CM |
2278 | sb->s_blocksize = 4096; |
2279 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2280 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2281 | |
76dda93c | 2282 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2283 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2284 | /* |
2285 | * we set the i_size on the btree inode to the max possible int. | |
2286 | * the real end of the address space is determined by all of | |
2287 | * the devices in the system | |
2288 | */ | |
2289 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2290 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2291 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2292 | ||
5d4f98a2 | 2293 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2294 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2295 | fs_info->btree_inode->i_mapping); |
0b32f4bb | 2296 | BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; |
a8067e02 | 2297 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2298 | |
2299 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2300 | |
76dda93c YZ |
2301 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2302 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2303 | sizeof(struct btrfs_key)); | |
72ac3c0d JB |
2304 | set_bit(BTRFS_INODE_DUMMY, |
2305 | &BTRFS_I(fs_info->btree_inode)->runtime_flags); | |
778ba82b | 2306 | btrfs_insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2307 | |
0f9dd46c | 2308 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2309 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2310 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2311 | |
11833d66 | 2312 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2313 | fs_info->btree_inode->i_mapping); |
11833d66 | 2314 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2315 | fs_info->btree_inode->i_mapping); |
11833d66 | 2316 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2317 | fs_info->do_barriers = 1; |
e18e4809 | 2318 | |
39279cc3 | 2319 | |
5a3f23d5 | 2320 | mutex_init(&fs_info->ordered_operations_mutex); |
9ffba8cd | 2321 | mutex_init(&fs_info->ordered_extent_flush_mutex); |
e02119d5 | 2322 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2323 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2324 | mutex_init(&fs_info->transaction_kthread_mutex); |
2325 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2326 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2327 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2328 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2329 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2330 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
e922e087 SB |
2331 | fs_info->dev_replace.lock_owner = 0; |
2332 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2333 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
2334 | mutex_init(&fs_info->dev_replace.lock_management_lock); | |
2335 | mutex_init(&fs_info->dev_replace.lock); | |
fa9c0d79 | 2336 | |
416ac51d | 2337 | spin_lock_init(&fs_info->qgroup_lock); |
f2f6ed3d | 2338 | mutex_init(&fs_info->qgroup_ioctl_lock); |
416ac51d AJ |
2339 | fs_info->qgroup_tree = RB_ROOT; |
2340 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2341 | fs_info->qgroup_seq = 1; | |
2342 | fs_info->quota_enabled = 0; | |
2343 | fs_info->pending_quota_state = 0; | |
1e8f9158 | 2344 | fs_info->qgroup_ulist = NULL; |
2f232036 | 2345 | mutex_init(&fs_info->qgroup_rescan_lock); |
416ac51d | 2346 | |
fa9c0d79 CM |
2347 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2348 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2349 | ||
e6dcd2dc | 2350 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2351 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2352 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2353 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2354 | |
53b381b3 DW |
2355 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2356 | if (ret) { | |
83c8266a | 2357 | err = ret; |
53b381b3 DW |
2358 | goto fail_alloc; |
2359 | } | |
2360 | ||
0b86a832 | 2361 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2362 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2363 | |
3c4bb26b | 2364 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2365 | |
2366 | /* | |
2367 | * Read super block and check the signature bytes only | |
2368 | */ | |
a512bbf8 | 2369 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2370 | if (!bh) { |
2371 | err = -EINVAL; | |
16cdcec7 | 2372 | goto fail_alloc; |
20b45077 | 2373 | } |
39279cc3 | 2374 | |
1104a885 DS |
2375 | /* |
2376 | * We want to check superblock checksum, the type is stored inside. | |
2377 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2378 | */ | |
2379 | if (btrfs_check_super_csum(bh->b_data)) { | |
efe120a0 | 2380 | printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); |
1104a885 DS |
2381 | err = -EINVAL; |
2382 | goto fail_alloc; | |
2383 | } | |
2384 | ||
2385 | /* | |
2386 | * super_copy is zeroed at allocation time and we never touch the | |
2387 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2388 | * the whole block of INFO_SIZE | |
2389 | */ | |
6c41761f DS |
2390 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2391 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2392 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2393 | brelse(bh); |
5f39d397 | 2394 | |
6c41761f | 2395 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2396 | |
1104a885 DS |
2397 | ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); |
2398 | if (ret) { | |
efe120a0 | 2399 | printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); |
1104a885 DS |
2400 | err = -EINVAL; |
2401 | goto fail_alloc; | |
2402 | } | |
2403 | ||
6c41761f | 2404 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2405 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2406 | goto fail_alloc; |
0f7d52f4 | 2407 | |
acce952b | 2408 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2409 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2410 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2411 | |
af31f5e5 CM |
2412 | /* |
2413 | * run through our array of backup supers and setup | |
2414 | * our ring pointer to the oldest one | |
2415 | */ | |
2416 | generation = btrfs_super_generation(disk_super); | |
2417 | find_oldest_super_backup(fs_info, generation); | |
2418 | ||
75e7cb7f LB |
2419 | /* |
2420 | * In the long term, we'll store the compression type in the super | |
2421 | * block, and it'll be used for per file compression control. | |
2422 | */ | |
2423 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2424 | ||
2b82032c YZ |
2425 | ret = btrfs_parse_options(tree_root, options); |
2426 | if (ret) { | |
2427 | err = ret; | |
16cdcec7 | 2428 | goto fail_alloc; |
2b82032c | 2429 | } |
dfe25020 | 2430 | |
f2b636e8 JB |
2431 | features = btrfs_super_incompat_flags(disk_super) & |
2432 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2433 | if (features) { | |
2434 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2435 | "unsupported optional features (%Lx).\n", | |
c1c9ff7c | 2436 | features); |
f2b636e8 | 2437 | err = -EINVAL; |
16cdcec7 | 2438 | goto fail_alloc; |
f2b636e8 JB |
2439 | } |
2440 | ||
727011e0 CM |
2441 | if (btrfs_super_leafsize(disk_super) != |
2442 | btrfs_super_nodesize(disk_super)) { | |
2443 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2444 | "blocksizes don't match. node %d leaf %d\n", | |
2445 | btrfs_super_nodesize(disk_super), | |
2446 | btrfs_super_leafsize(disk_super)); | |
2447 | err = -EINVAL; | |
2448 | goto fail_alloc; | |
2449 | } | |
2450 | if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2451 | printk(KERN_ERR "BTRFS: couldn't mount because metadata " | |
2452 | "blocksize (%d) was too large\n", | |
2453 | btrfs_super_leafsize(disk_super)); | |
2454 | err = -EINVAL; | |
2455 | goto fail_alloc; | |
2456 | } | |
2457 | ||
5d4f98a2 | 2458 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2459 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
69e380d1 | 2460 | if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2461 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
727011e0 | 2462 | |
3173a18f | 2463 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
efe120a0 | 2464 | printk(KERN_ERR "BTRFS: has skinny extents\n"); |
3173a18f | 2465 | |
727011e0 CM |
2466 | /* |
2467 | * flag our filesystem as having big metadata blocks if | |
2468 | * they are bigger than the page size | |
2469 | */ | |
2470 | if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { | |
2471 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) | |
efe120a0 | 2472 | printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); |
727011e0 CM |
2473 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2474 | } | |
2475 | ||
bc3f116f CM |
2476 | nodesize = btrfs_super_nodesize(disk_super); |
2477 | leafsize = btrfs_super_leafsize(disk_super); | |
2478 | sectorsize = btrfs_super_sectorsize(disk_super); | |
2479 | stripesize = btrfs_super_stripesize(disk_super); | |
e2d84521 | 2480 | fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2481 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f CM |
2482 | |
2483 | /* | |
2484 | * mixed block groups end up with duplicate but slightly offset | |
2485 | * extent buffers for the same range. It leads to corruptions | |
2486 | */ | |
2487 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
2488 | (sectorsize != leafsize)) { | |
efe120a0 | 2489 | printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes " |
bc3f116f CM |
2490 | "are not allowed for mixed block groups on %s\n", |
2491 | sb->s_id); | |
2492 | goto fail_alloc; | |
2493 | } | |
2494 | ||
ceda0864 MX |
2495 | /* |
2496 | * Needn't use the lock because there is no other task which will | |
2497 | * update the flag. | |
2498 | */ | |
a6fa6fae | 2499 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2500 | |
f2b636e8 JB |
2501 | features = btrfs_super_compat_ro_flags(disk_super) & |
2502 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2503 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2504 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2505 | "unsupported option features (%Lx).\n", | |
c1c9ff7c | 2506 | features); |
f2b636e8 | 2507 | err = -EINVAL; |
16cdcec7 | 2508 | goto fail_alloc; |
f2b636e8 | 2509 | } |
61d92c32 | 2510 | |
5cdc7ad3 | 2511 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2512 | |
5cdc7ad3 QW |
2513 | fs_info->workers = |
2514 | btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, | |
2515 | max_active, 16); | |
c8b97818 | 2516 | |
afe3d242 QW |
2517 | fs_info->delalloc_workers = |
2518 | btrfs_alloc_workqueue("delalloc", flags, max_active, 2); | |
771ed689 | 2519 | |
a44903ab QW |
2520 | fs_info->flush_workers = |
2521 | btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); | |
61b49440 | 2522 | |
e66f0bb1 QW |
2523 | fs_info->caching_workers = |
2524 | btrfs_alloc_workqueue("cache", flags, max_active, 0); | |
bab39bf9 | 2525 | |
a8c93d4e QW |
2526 | /* |
2527 | * a higher idle thresh on the submit workers makes it much more | |
61b49440 CM |
2528 | * likely that bios will be send down in a sane order to the |
2529 | * devices | |
2530 | */ | |
a8c93d4e QW |
2531 | fs_info->submit_workers = |
2532 | btrfs_alloc_workqueue("submit", flags, | |
2533 | min_t(u64, fs_devices->num_devices, | |
2534 | max_active), 64); | |
53863232 | 2535 | |
dc6e3209 QW |
2536 | fs_info->fixup_workers = |
2537 | btrfs_alloc_workqueue("fixup", flags, 1, 0); | |
fccb5d86 QW |
2538 | |
2539 | /* | |
2540 | * endios are largely parallel and should have a very | |
2541 | * low idle thresh | |
2542 | */ | |
2543 | fs_info->endio_workers = | |
2544 | btrfs_alloc_workqueue("endio", flags, max_active, 4); | |
2545 | fs_info->endio_meta_workers = | |
2546 | btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); | |
2547 | fs_info->endio_meta_write_workers = | |
2548 | btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); | |
2549 | fs_info->endio_raid56_workers = | |
2550 | btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); | |
d05a33ac QW |
2551 | fs_info->rmw_workers = |
2552 | btrfs_alloc_workqueue("rmw", flags, max_active, 2); | |
fccb5d86 QW |
2553 | fs_info->endio_write_workers = |
2554 | btrfs_alloc_workqueue("endio-write", flags, max_active, 2); | |
2555 | fs_info->endio_freespace_worker = | |
2556 | btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); | |
5b3bc44e QW |
2557 | fs_info->delayed_workers = |
2558 | btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); | |
736cfa15 QW |
2559 | fs_info->readahead_workers = |
2560 | btrfs_alloc_workqueue("readahead", flags, max_active, 2); | |
fc97fab0 QW |
2561 | fs_info->qgroup_rescan_workers = |
2562 | btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); | |
61b49440 | 2563 | |
a8c93d4e | 2564 | if (!(fs_info->workers && fs_info->delalloc_workers && |
fccb5d86 QW |
2565 | fs_info->submit_workers && fs_info->flush_workers && |
2566 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2567 | fs_info->endio_meta_write_workers && | |
2568 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
e66f0bb1 | 2569 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
dc6e3209 | 2570 | fs_info->caching_workers && fs_info->readahead_workers && |
fc97fab0 QW |
2571 | fs_info->fixup_workers && fs_info->delayed_workers && |
2572 | fs_info->qgroup_rescan_workers)) { | |
5cdc7ad3 QW |
2573 | err = -ENOMEM; |
2574 | goto fail_sb_buffer; | |
2575 | } | |
4543df7e | 2576 | |
4575c9cc | 2577 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2578 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2579 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2580 | |
db94535d CM |
2581 | tree_root->nodesize = nodesize; |
2582 | tree_root->leafsize = leafsize; | |
2583 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2584 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2585 | |
2586 | sb->s_blocksize = sectorsize; | |
2587 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2588 | |
3cae210f | 2589 | if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { |
efe120a0 | 2590 | printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2591 | goto fail_sb_buffer; |
2592 | } | |
19c00ddc | 2593 | |
8d082fb7 | 2594 | if (sectorsize != PAGE_SIZE) { |
efe120a0 | 2595 | printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) " |
8d082fb7 | 2596 | "found on %s\n", (unsigned long)sectorsize, sb->s_id); |
941b2ddf KM |
2597 | goto fail_sb_buffer; |
2598 | } | |
2599 | ||
925baedd | 2600 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2601 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2602 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2603 | if (ret) { |
efe120a0 | 2604 | printk(KERN_WARNING "BTRFS: failed to read the system " |
d397712b | 2605 | "array on %s\n", sb->s_id); |
5d4f98a2 | 2606 | goto fail_sb_buffer; |
84eed90f | 2607 | } |
0b86a832 CM |
2608 | |
2609 | blocksize = btrfs_level_size(tree_root, | |
2610 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2611 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2612 | |
2613 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2614 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2615 | ||
2616 | chunk_root->node = read_tree_block(chunk_root, | |
2617 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2618 | blocksize, generation); |
416bc658 JB |
2619 | if (!chunk_root->node || |
2620 | !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { | |
efe120a0 | 2621 | printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n", |
83121942 | 2622 | sb->s_id); |
af31f5e5 | 2623 | goto fail_tree_roots; |
83121942 | 2624 | } |
5d4f98a2 YZ |
2625 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2626 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2627 | |
e17cade2 | 2628 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2629 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2630 | |
0b86a832 | 2631 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2632 | if (ret) { |
efe120a0 | 2633 | printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n", |
d397712b | 2634 | sb->s_id); |
af31f5e5 | 2635 | goto fail_tree_roots; |
2b82032c | 2636 | } |
0b86a832 | 2637 | |
8dabb742 SB |
2638 | /* |
2639 | * keep the device that is marked to be the target device for the | |
2640 | * dev_replace procedure | |
2641 | */ | |
2642 | btrfs_close_extra_devices(fs_info, fs_devices, 0); | |
dfe25020 | 2643 | |
a6b0d5c8 | 2644 | if (!fs_devices->latest_bdev) { |
efe120a0 | 2645 | printk(KERN_CRIT "BTRFS: failed to read devices on %s\n", |
a6b0d5c8 CM |
2646 | sb->s_id); |
2647 | goto fail_tree_roots; | |
2648 | } | |
2649 | ||
af31f5e5 | 2650 | retry_root_backup: |
db94535d CM |
2651 | blocksize = btrfs_level_size(tree_root, |
2652 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2653 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2654 | |
e20d96d6 | 2655 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2656 | btrfs_super_root(disk_super), |
84234f3a | 2657 | blocksize, generation); |
af31f5e5 CM |
2658 | if (!tree_root->node || |
2659 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
efe120a0 | 2660 | printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", |
83121942 | 2661 | sb->s_id); |
af31f5e5 CM |
2662 | |
2663 | goto recovery_tree_root; | |
83121942 | 2664 | } |
af31f5e5 | 2665 | |
5d4f98a2 YZ |
2666 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2667 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2668 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2669 | |
cb517eab MX |
2670 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2671 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2672 | location.offset = 0; | |
2673 | ||
2674 | extent_root = btrfs_read_tree_root(tree_root, &location); | |
2675 | if (IS_ERR(extent_root)) { | |
2676 | ret = PTR_ERR(extent_root); | |
af31f5e5 | 2677 | goto recovery_tree_root; |
cb517eab | 2678 | } |
0b86a832 | 2679 | extent_root->track_dirty = 1; |
cb517eab | 2680 | fs_info->extent_root = extent_root; |
0b86a832 | 2681 | |
cb517eab MX |
2682 | location.objectid = BTRFS_DEV_TREE_OBJECTID; |
2683 | dev_root = btrfs_read_tree_root(tree_root, &location); | |
2684 | if (IS_ERR(dev_root)) { | |
2685 | ret = PTR_ERR(dev_root); | |
af31f5e5 | 2686 | goto recovery_tree_root; |
cb517eab | 2687 | } |
5d4f98a2 | 2688 | dev_root->track_dirty = 1; |
cb517eab MX |
2689 | fs_info->dev_root = dev_root; |
2690 | btrfs_init_devices_late(fs_info); | |
3768f368 | 2691 | |
cb517eab MX |
2692 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; |
2693 | csum_root = btrfs_read_tree_root(tree_root, &location); | |
2694 | if (IS_ERR(csum_root)) { | |
2695 | ret = PTR_ERR(csum_root); | |
af31f5e5 | 2696 | goto recovery_tree_root; |
cb517eab | 2697 | } |
d20f7043 | 2698 | csum_root->track_dirty = 1; |
cb517eab | 2699 | fs_info->csum_root = csum_root; |
d20f7043 | 2700 | |
cb517eab MX |
2701 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
2702 | quota_root = btrfs_read_tree_root(tree_root, &location); | |
2703 | if (!IS_ERR(quota_root)) { | |
bcef60f2 AJ |
2704 | quota_root->track_dirty = 1; |
2705 | fs_info->quota_enabled = 1; | |
2706 | fs_info->pending_quota_state = 1; | |
cb517eab | 2707 | fs_info->quota_root = quota_root; |
bcef60f2 AJ |
2708 | } |
2709 | ||
f7a81ea4 SB |
2710 | location.objectid = BTRFS_UUID_TREE_OBJECTID; |
2711 | uuid_root = btrfs_read_tree_root(tree_root, &location); | |
2712 | if (IS_ERR(uuid_root)) { | |
2713 | ret = PTR_ERR(uuid_root); | |
2714 | if (ret != -ENOENT) | |
2715 | goto recovery_tree_root; | |
2716 | create_uuid_tree = true; | |
70f80175 | 2717 | check_uuid_tree = false; |
f7a81ea4 SB |
2718 | } else { |
2719 | uuid_root->track_dirty = 1; | |
2720 | fs_info->uuid_root = uuid_root; | |
70f80175 SB |
2721 | create_uuid_tree = false; |
2722 | check_uuid_tree = | |
2723 | generation != btrfs_super_uuid_tree_generation(disk_super); | |
f7a81ea4 SB |
2724 | } |
2725 | ||
8929ecfa YZ |
2726 | fs_info->generation = generation; |
2727 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2728 | |
68310a5e ID |
2729 | ret = btrfs_recover_balance(fs_info); |
2730 | if (ret) { | |
efe120a0 | 2731 | printk(KERN_WARNING "BTRFS: failed to recover balance\n"); |
68310a5e ID |
2732 | goto fail_block_groups; |
2733 | } | |
2734 | ||
733f4fbb SB |
2735 | ret = btrfs_init_dev_stats(fs_info); |
2736 | if (ret) { | |
efe120a0 | 2737 | printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", |
733f4fbb SB |
2738 | ret); |
2739 | goto fail_block_groups; | |
2740 | } | |
2741 | ||
8dabb742 SB |
2742 | ret = btrfs_init_dev_replace(fs_info); |
2743 | if (ret) { | |
efe120a0 | 2744 | pr_err("BTRFS: failed to init dev_replace: %d\n", ret); |
8dabb742 SB |
2745 | goto fail_block_groups; |
2746 | } | |
2747 | ||
2748 | btrfs_close_extra_devices(fs_info, fs_devices, 1); | |
2749 | ||
5ac1d209 JM |
2750 | ret = btrfs_sysfs_add_one(fs_info); |
2751 | if (ret) { | |
efe120a0 | 2752 | pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); |
5ac1d209 JM |
2753 | goto fail_block_groups; |
2754 | } | |
2755 | ||
c59021f8 | 2756 | ret = btrfs_init_space_info(fs_info); |
2757 | if (ret) { | |
efe120a0 | 2758 | printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); |
2365dd3c | 2759 | goto fail_sysfs; |
c59021f8 | 2760 | } |
2761 | ||
1b1d1f66 JB |
2762 | ret = btrfs_read_block_groups(extent_root); |
2763 | if (ret) { | |
efe120a0 | 2764 | printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); |
2365dd3c | 2765 | goto fail_sysfs; |
1b1d1f66 | 2766 | } |
5af3e8cc SB |
2767 | fs_info->num_tolerated_disk_barrier_failures = |
2768 | btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); | |
292fd7fc SB |
2769 | if (fs_info->fs_devices->missing_devices > |
2770 | fs_info->num_tolerated_disk_barrier_failures && | |
2771 | !(sb->s_flags & MS_RDONLY)) { | |
efe120a0 FH |
2772 | printk(KERN_WARNING "BTRFS: " |
2773 | "too many missing devices, writeable mount is not allowed\n"); | |
2365dd3c | 2774 | goto fail_sysfs; |
292fd7fc | 2775 | } |
9078a3e1 | 2776 | |
a74a4b97 CM |
2777 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2778 | "btrfs-cleaner"); | |
57506d50 | 2779 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 2780 | goto fail_sysfs; |
a74a4b97 CM |
2781 | |
2782 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2783 | tree_root, | |
2784 | "btrfs-transaction"); | |
57506d50 | 2785 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2786 | goto fail_cleaner; |
a74a4b97 | 2787 | |
c289811c CM |
2788 | if (!btrfs_test_opt(tree_root, SSD) && |
2789 | !btrfs_test_opt(tree_root, NOSSD) && | |
2790 | !fs_info->fs_devices->rotating) { | |
efe120a0 | 2791 | printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " |
c289811c CM |
2792 | "mode\n"); |
2793 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2794 | } | |
2795 | ||
3818aea2 QW |
2796 | /* Set the real inode map cache flag */ |
2797 | if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE)) | |
2798 | btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE); | |
2799 | ||
21adbd5c SB |
2800 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2801 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2802 | ret = btrfsic_mount(tree_root, fs_devices, | |
2803 | btrfs_test_opt(tree_root, | |
2804 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2805 | 1 : 0, | |
2806 | fs_info->check_integrity_print_mask); | |
2807 | if (ret) | |
efe120a0 | 2808 | printk(KERN_WARNING "BTRFS: failed to initialize" |
21adbd5c SB |
2809 | " integrity check module %s\n", sb->s_id); |
2810 | } | |
2811 | #endif | |
bcef60f2 AJ |
2812 | ret = btrfs_read_qgroup_config(fs_info); |
2813 | if (ret) | |
2814 | goto fail_trans_kthread; | |
21adbd5c | 2815 | |
acce952b | 2816 | /* do not make disk changes in broken FS */ |
68ce9682 | 2817 | if (btrfs_super_log_root(disk_super) != 0) { |
e02119d5 CM |
2818 | u64 bytenr = btrfs_super_log_root(disk_super); |
2819 | ||
7c2ca468 | 2820 | if (fs_devices->rw_devices == 0) { |
efe120a0 | 2821 | printk(KERN_WARNING "BTRFS: log replay required " |
d397712b | 2822 | "on RO media\n"); |
7c2ca468 | 2823 | err = -EIO; |
bcef60f2 | 2824 | goto fail_qgroup; |
7c2ca468 | 2825 | } |
e02119d5 CM |
2826 | blocksize = |
2827 | btrfs_level_size(tree_root, | |
2828 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2829 | |
6f07e42e | 2830 | log_tree_root = btrfs_alloc_root(fs_info); |
676e4c86 DC |
2831 | if (!log_tree_root) { |
2832 | err = -ENOMEM; | |
bcef60f2 | 2833 | goto fail_qgroup; |
676e4c86 | 2834 | } |
e02119d5 CM |
2835 | |
2836 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2837 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2838 | ||
2839 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2840 | blocksize, |
2841 | generation + 1); | |
416bc658 JB |
2842 | if (!log_tree_root->node || |
2843 | !extent_buffer_uptodate(log_tree_root->node)) { | |
efe120a0 | 2844 | printk(KERN_ERR "BTRFS: failed to read log tree\n"); |
416bc658 JB |
2845 | free_extent_buffer(log_tree_root->node); |
2846 | kfree(log_tree_root); | |
2847 | goto fail_trans_kthread; | |
2848 | } | |
79787eaa | 2849 | /* returns with log_tree_root freed on success */ |
e02119d5 | 2850 | ret = btrfs_recover_log_trees(log_tree_root); |
79787eaa JM |
2851 | if (ret) { |
2852 | btrfs_error(tree_root->fs_info, ret, | |
2853 | "Failed to recover log tree"); | |
2854 | free_extent_buffer(log_tree_root->node); | |
2855 | kfree(log_tree_root); | |
2856 | goto fail_trans_kthread; | |
2857 | } | |
e556ce2c YZ |
2858 | |
2859 | if (sb->s_flags & MS_RDONLY) { | |
79787eaa JM |
2860 | ret = btrfs_commit_super(tree_root); |
2861 | if (ret) | |
2862 | goto fail_trans_kthread; | |
e556ce2c | 2863 | } |
e02119d5 | 2864 | } |
1a40e23b | 2865 | |
76dda93c | 2866 | ret = btrfs_find_orphan_roots(tree_root); |
79787eaa JM |
2867 | if (ret) |
2868 | goto fail_trans_kthread; | |
76dda93c | 2869 | |
7c2ca468 | 2870 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b | 2871 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 ID |
2872 | if (ret) |
2873 | goto fail_trans_kthread; | |
d68fc57b | 2874 | |
5d4f98a2 | 2875 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2876 | if (ret < 0) { |
2877 | printk(KERN_WARNING | |
efe120a0 | 2878 | "BTRFS: failed to recover relocation\n"); |
d7ce5843 | 2879 | err = -EINVAL; |
bcef60f2 | 2880 | goto fail_qgroup; |
d7ce5843 | 2881 | } |
7c2ca468 | 2882 | } |
1a40e23b | 2883 | |
3de4586c CM |
2884 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2885 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 2886 | location.offset = 0; |
3de4586c | 2887 | |
3de4586c | 2888 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
2889 | if (IS_ERR(fs_info->fs_root)) { |
2890 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2891 | goto fail_qgroup; |
3140c9a3 | 2892 | } |
c289811c | 2893 | |
2b6ba629 ID |
2894 | if (sb->s_flags & MS_RDONLY) |
2895 | return 0; | |
59641015 | 2896 | |
2b6ba629 ID |
2897 | down_read(&fs_info->cleanup_work_sem); |
2898 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
2899 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 2900 | up_read(&fs_info->cleanup_work_sem); |
2b6ba629 ID |
2901 | close_ctree(tree_root); |
2902 | return ret; | |
2903 | } | |
2904 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 2905 | |
2b6ba629 ID |
2906 | ret = btrfs_resume_balance_async(fs_info); |
2907 | if (ret) { | |
efe120a0 | 2908 | printk(KERN_WARNING "BTRFS: failed to resume balance\n"); |
2b6ba629 ID |
2909 | close_ctree(tree_root); |
2910 | return ret; | |
e3acc2a6 JB |
2911 | } |
2912 | ||
8dabb742 SB |
2913 | ret = btrfs_resume_dev_replace_async(fs_info); |
2914 | if (ret) { | |
efe120a0 | 2915 | pr_warn("BTRFS: failed to resume dev_replace\n"); |
8dabb742 SB |
2916 | close_ctree(tree_root); |
2917 | return ret; | |
2918 | } | |
2919 | ||
b382a324 JS |
2920 | btrfs_qgroup_rescan_resume(fs_info); |
2921 | ||
f7a81ea4 | 2922 | if (create_uuid_tree) { |
efe120a0 | 2923 | pr_info("BTRFS: creating UUID tree\n"); |
f7a81ea4 SB |
2924 | ret = btrfs_create_uuid_tree(fs_info); |
2925 | if (ret) { | |
efe120a0 | 2926 | pr_warn("BTRFS: failed to create the UUID tree %d\n", |
f7a81ea4 SB |
2927 | ret); |
2928 | close_ctree(tree_root); | |
2929 | return ret; | |
2930 | } | |
f420ee1e SB |
2931 | } else if (check_uuid_tree || |
2932 | btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { | |
efe120a0 | 2933 | pr_info("BTRFS: checking UUID tree\n"); |
70f80175 SB |
2934 | ret = btrfs_check_uuid_tree(fs_info); |
2935 | if (ret) { | |
efe120a0 | 2936 | pr_warn("BTRFS: failed to check the UUID tree %d\n", |
70f80175 SB |
2937 | ret); |
2938 | close_ctree(tree_root); | |
2939 | return ret; | |
2940 | } | |
2941 | } else { | |
2942 | fs_info->update_uuid_tree_gen = 1; | |
f7a81ea4 SB |
2943 | } |
2944 | ||
ad2b2c80 | 2945 | return 0; |
39279cc3 | 2946 | |
bcef60f2 AJ |
2947 | fail_qgroup: |
2948 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
2949 | fail_trans_kthread: |
2950 | kthread_stop(fs_info->transaction_kthread); | |
54067ae9 | 2951 | btrfs_cleanup_transaction(fs_info->tree_root); |
7b5ff90e | 2952 | del_fs_roots(fs_info); |
3f157a2f | 2953 | fail_cleaner: |
a74a4b97 | 2954 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2955 | |
2956 | /* | |
2957 | * make sure we're done with the btree inode before we stop our | |
2958 | * kthreads | |
2959 | */ | |
2960 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 2961 | |
2365dd3c AJ |
2962 | fail_sysfs: |
2963 | btrfs_sysfs_remove_one(fs_info); | |
2964 | ||
1b1d1f66 | 2965 | fail_block_groups: |
54067ae9 | 2966 | btrfs_put_block_group_cache(fs_info); |
1b1d1f66 | 2967 | btrfs_free_block_groups(fs_info); |
af31f5e5 CM |
2968 | |
2969 | fail_tree_roots: | |
2970 | free_root_pointers(fs_info, 1); | |
2b8195bb | 2971 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 2972 | |
39279cc3 | 2973 | fail_sb_buffer: |
7abadb64 | 2974 | btrfs_stop_all_workers(fs_info); |
16cdcec7 | 2975 | fail_alloc: |
4543df7e | 2976 | fail_iput: |
586e46e2 ID |
2977 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2978 | ||
4543df7e | 2979 | iput(fs_info->btree_inode); |
c404e0dc MX |
2980 | fail_bio_counter: |
2981 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
2982 | fail_delalloc_bytes: |
2983 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
2984 | fail_dirty_metadata_bytes: |
2985 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
ad081f14 | 2986 | fail_bdi: |
7e662854 | 2987 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2988 | fail_srcu: |
2989 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2990 | fail: |
53b381b3 | 2991 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 2992 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 2993 | return err; |
af31f5e5 CM |
2994 | |
2995 | recovery_tree_root: | |
af31f5e5 CM |
2996 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2997 | goto fail_tree_roots; | |
2998 | ||
2999 | free_root_pointers(fs_info, 0); | |
3000 | ||
3001 | /* don't use the log in recovery mode, it won't be valid */ | |
3002 | btrfs_set_super_log_root(disk_super, 0); | |
3003 | ||
3004 | /* we can't trust the free space cache either */ | |
3005 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3006 | ||
3007 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3008 | &num_backups_tried, &backup_index); | |
3009 | if (ret == -1) | |
3010 | goto fail_block_groups; | |
3011 | goto retry_root_backup; | |
eb60ceac CM |
3012 | } |
3013 | ||
f2984462 CM |
3014 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3015 | { | |
f2984462 CM |
3016 | if (uptodate) { |
3017 | set_buffer_uptodate(bh); | |
3018 | } else { | |
442a4f63 SB |
3019 | struct btrfs_device *device = (struct btrfs_device *) |
3020 | bh->b_private; | |
3021 | ||
efe120a0 | 3022 | printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to " |
606686ee JB |
3023 | "I/O error on %s\n", |
3024 | rcu_str_deref(device->name)); | |
1259ab75 CM |
3025 | /* note, we dont' set_buffer_write_io_error because we have |
3026 | * our own ways of dealing with the IO errors | |
3027 | */ | |
f2984462 | 3028 | clear_buffer_uptodate(bh); |
442a4f63 | 3029 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3030 | } |
3031 | unlock_buffer(bh); | |
3032 | put_bh(bh); | |
3033 | } | |
3034 | ||
a512bbf8 YZ |
3035 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3036 | { | |
3037 | struct buffer_head *bh; | |
3038 | struct buffer_head *latest = NULL; | |
3039 | struct btrfs_super_block *super; | |
3040 | int i; | |
3041 | u64 transid = 0; | |
3042 | u64 bytenr; | |
3043 | ||
3044 | /* we would like to check all the supers, but that would make | |
3045 | * a btrfs mount succeed after a mkfs from a different FS. | |
3046 | * So, we need to add a special mount option to scan for | |
3047 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3048 | */ | |
3049 | for (i = 0; i < 1; i++) { | |
3050 | bytenr = btrfs_sb_offset(i); | |
8068a47e AJ |
3051 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3052 | i_size_read(bdev->bd_inode)) | |
a512bbf8 | 3053 | break; |
8068a47e AJ |
3054 | bh = __bread(bdev, bytenr / 4096, |
3055 | BTRFS_SUPER_INFO_SIZE); | |
a512bbf8 YZ |
3056 | if (!bh) |
3057 | continue; | |
3058 | ||
3059 | super = (struct btrfs_super_block *)bh->b_data; | |
3060 | if (btrfs_super_bytenr(super) != bytenr || | |
3cae210f | 3061 | btrfs_super_magic(super) != BTRFS_MAGIC) { |
a512bbf8 YZ |
3062 | brelse(bh); |
3063 | continue; | |
3064 | } | |
3065 | ||
3066 | if (!latest || btrfs_super_generation(super) > transid) { | |
3067 | brelse(latest); | |
3068 | latest = bh; | |
3069 | transid = btrfs_super_generation(super); | |
3070 | } else { | |
3071 | brelse(bh); | |
3072 | } | |
3073 | } | |
3074 | return latest; | |
3075 | } | |
3076 | ||
4eedeb75 HH |
3077 | /* |
3078 | * this should be called twice, once with wait == 0 and | |
3079 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
3080 | * we write are pinned. | |
3081 | * | |
3082 | * They are released when wait == 1 is done. | |
3083 | * max_mirrors must be the same for both runs, and it indicates how | |
3084 | * many supers on this one device should be written. | |
3085 | * | |
3086 | * max_mirrors == 0 means to write them all. | |
3087 | */ | |
a512bbf8 YZ |
3088 | static int write_dev_supers(struct btrfs_device *device, |
3089 | struct btrfs_super_block *sb, | |
3090 | int do_barriers, int wait, int max_mirrors) | |
3091 | { | |
3092 | struct buffer_head *bh; | |
3093 | int i; | |
3094 | int ret; | |
3095 | int errors = 0; | |
3096 | u32 crc; | |
3097 | u64 bytenr; | |
a512bbf8 YZ |
3098 | |
3099 | if (max_mirrors == 0) | |
3100 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3101 | ||
a512bbf8 YZ |
3102 | for (i = 0; i < max_mirrors; i++) { |
3103 | bytenr = btrfs_sb_offset(i); | |
3104 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
3105 | break; | |
3106 | ||
3107 | if (wait) { | |
3108 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
3109 | BTRFS_SUPER_INFO_SIZE); | |
634554dc JB |
3110 | if (!bh) { |
3111 | errors++; | |
3112 | continue; | |
3113 | } | |
a512bbf8 | 3114 | wait_on_buffer(bh); |
4eedeb75 HH |
3115 | if (!buffer_uptodate(bh)) |
3116 | errors++; | |
3117 | ||
3118 | /* drop our reference */ | |
3119 | brelse(bh); | |
3120 | ||
3121 | /* drop the reference from the wait == 0 run */ | |
3122 | brelse(bh); | |
3123 | continue; | |
a512bbf8 YZ |
3124 | } else { |
3125 | btrfs_set_super_bytenr(sb, bytenr); | |
3126 | ||
3127 | crc = ~(u32)0; | |
b0496686 | 3128 | crc = btrfs_csum_data((char *)sb + |
a512bbf8 YZ |
3129 | BTRFS_CSUM_SIZE, crc, |
3130 | BTRFS_SUPER_INFO_SIZE - | |
3131 | BTRFS_CSUM_SIZE); | |
3132 | btrfs_csum_final(crc, sb->csum); | |
3133 | ||
4eedeb75 HH |
3134 | /* |
3135 | * one reference for us, and we leave it for the | |
3136 | * caller | |
3137 | */ | |
a512bbf8 YZ |
3138 | bh = __getblk(device->bdev, bytenr / 4096, |
3139 | BTRFS_SUPER_INFO_SIZE); | |
634554dc | 3140 | if (!bh) { |
efe120a0 | 3141 | printk(KERN_ERR "BTRFS: couldn't get super " |
634554dc JB |
3142 | "buffer head for bytenr %Lu\n", bytenr); |
3143 | errors++; | |
3144 | continue; | |
3145 | } | |
3146 | ||
a512bbf8 YZ |
3147 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
3148 | ||
4eedeb75 | 3149 | /* one reference for submit_bh */ |
a512bbf8 | 3150 | get_bh(bh); |
4eedeb75 HH |
3151 | |
3152 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
3153 | lock_buffer(bh); |
3154 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
442a4f63 | 3155 | bh->b_private = device; |
a512bbf8 YZ |
3156 | } |
3157 | ||
387125fc CM |
3158 | /* |
3159 | * we fua the first super. The others we allow | |
3160 | * to go down lazy. | |
3161 | */ | |
e8117c26 WS |
3162 | if (i == 0) |
3163 | ret = btrfsic_submit_bh(WRITE_FUA, bh); | |
3164 | else | |
3165 | ret = btrfsic_submit_bh(WRITE_SYNC, bh); | |
4eedeb75 | 3166 | if (ret) |
a512bbf8 | 3167 | errors++; |
a512bbf8 YZ |
3168 | } |
3169 | return errors < i ? 0 : -1; | |
3170 | } | |
3171 | ||
387125fc CM |
3172 | /* |
3173 | * endio for the write_dev_flush, this will wake anyone waiting | |
3174 | * for the barrier when it is done | |
3175 | */ | |
3176 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
3177 | { | |
3178 | if (err) { | |
3179 | if (err == -EOPNOTSUPP) | |
3180 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
3181 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
3182 | } | |
3183 | if (bio->bi_private) | |
3184 | complete(bio->bi_private); | |
3185 | bio_put(bio); | |
3186 | } | |
3187 | ||
3188 | /* | |
3189 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
3190 | * sent down. With wait == 1, it waits for the previous flush. | |
3191 | * | |
3192 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
3193 | * capable | |
3194 | */ | |
3195 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
3196 | { | |
3197 | struct bio *bio; | |
3198 | int ret = 0; | |
3199 | ||
3200 | if (device->nobarriers) | |
3201 | return 0; | |
3202 | ||
3203 | if (wait) { | |
3204 | bio = device->flush_bio; | |
3205 | if (!bio) | |
3206 | return 0; | |
3207 | ||
3208 | wait_for_completion(&device->flush_wait); | |
3209 | ||
3210 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
efe120a0 | 3211 | printk_in_rcu("BTRFS: disabling barriers on dev %s\n", |
606686ee | 3212 | rcu_str_deref(device->name)); |
387125fc | 3213 | device->nobarriers = 1; |
5af3e8cc | 3214 | } else if (!bio_flagged(bio, BIO_UPTODATE)) { |
387125fc | 3215 | ret = -EIO; |
5af3e8cc SB |
3216 | btrfs_dev_stat_inc_and_print(device, |
3217 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
387125fc CM |
3218 | } |
3219 | ||
3220 | /* drop the reference from the wait == 0 run */ | |
3221 | bio_put(bio); | |
3222 | device->flush_bio = NULL; | |
3223 | ||
3224 | return ret; | |
3225 | } | |
3226 | ||
3227 | /* | |
3228 | * one reference for us, and we leave it for the | |
3229 | * caller | |
3230 | */ | |
9c017abc | 3231 | device->flush_bio = NULL; |
9be3395b | 3232 | bio = btrfs_io_bio_alloc(GFP_NOFS, 0); |
387125fc CM |
3233 | if (!bio) |
3234 | return -ENOMEM; | |
3235 | ||
3236 | bio->bi_end_io = btrfs_end_empty_barrier; | |
3237 | bio->bi_bdev = device->bdev; | |
3238 | init_completion(&device->flush_wait); | |
3239 | bio->bi_private = &device->flush_wait; | |
3240 | device->flush_bio = bio; | |
3241 | ||
3242 | bio_get(bio); | |
21adbd5c | 3243 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
3244 | |
3245 | return 0; | |
3246 | } | |
3247 | ||
3248 | /* | |
3249 | * send an empty flush down to each device in parallel, | |
3250 | * then wait for them | |
3251 | */ | |
3252 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3253 | { | |
3254 | struct list_head *head; | |
3255 | struct btrfs_device *dev; | |
5af3e8cc SB |
3256 | int errors_send = 0; |
3257 | int errors_wait = 0; | |
387125fc CM |
3258 | int ret; |
3259 | ||
3260 | /* send down all the barriers */ | |
3261 | head = &info->fs_devices->devices; | |
3262 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3263 | if (dev->missing) |
3264 | continue; | |
387125fc | 3265 | if (!dev->bdev) { |
5af3e8cc | 3266 | errors_send++; |
387125fc CM |
3267 | continue; |
3268 | } | |
3269 | if (!dev->in_fs_metadata || !dev->writeable) | |
3270 | continue; | |
3271 | ||
3272 | ret = write_dev_flush(dev, 0); | |
3273 | if (ret) | |
5af3e8cc | 3274 | errors_send++; |
387125fc CM |
3275 | } |
3276 | ||
3277 | /* wait for all the barriers */ | |
3278 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3279 | if (dev->missing) |
3280 | continue; | |
387125fc | 3281 | if (!dev->bdev) { |
5af3e8cc | 3282 | errors_wait++; |
387125fc CM |
3283 | continue; |
3284 | } | |
3285 | if (!dev->in_fs_metadata || !dev->writeable) | |
3286 | continue; | |
3287 | ||
3288 | ret = write_dev_flush(dev, 1); | |
3289 | if (ret) | |
5af3e8cc | 3290 | errors_wait++; |
387125fc | 3291 | } |
5af3e8cc SB |
3292 | if (errors_send > info->num_tolerated_disk_barrier_failures || |
3293 | errors_wait > info->num_tolerated_disk_barrier_failures) | |
387125fc CM |
3294 | return -EIO; |
3295 | return 0; | |
3296 | } | |
3297 | ||
5af3e8cc SB |
3298 | int btrfs_calc_num_tolerated_disk_barrier_failures( |
3299 | struct btrfs_fs_info *fs_info) | |
3300 | { | |
3301 | struct btrfs_ioctl_space_info space; | |
3302 | struct btrfs_space_info *sinfo; | |
3303 | u64 types[] = {BTRFS_BLOCK_GROUP_DATA, | |
3304 | BTRFS_BLOCK_GROUP_SYSTEM, | |
3305 | BTRFS_BLOCK_GROUP_METADATA, | |
3306 | BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; | |
3307 | int num_types = 4; | |
3308 | int i; | |
3309 | int c; | |
3310 | int num_tolerated_disk_barrier_failures = | |
3311 | (int)fs_info->fs_devices->num_devices; | |
3312 | ||
3313 | for (i = 0; i < num_types; i++) { | |
3314 | struct btrfs_space_info *tmp; | |
3315 | ||
3316 | sinfo = NULL; | |
3317 | rcu_read_lock(); | |
3318 | list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { | |
3319 | if (tmp->flags == types[i]) { | |
3320 | sinfo = tmp; | |
3321 | break; | |
3322 | } | |
3323 | } | |
3324 | rcu_read_unlock(); | |
3325 | ||
3326 | if (!sinfo) | |
3327 | continue; | |
3328 | ||
3329 | down_read(&sinfo->groups_sem); | |
3330 | for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { | |
3331 | if (!list_empty(&sinfo->block_groups[c])) { | |
3332 | u64 flags; | |
3333 | ||
3334 | btrfs_get_block_group_info( | |
3335 | &sinfo->block_groups[c], &space); | |
3336 | if (space.total_bytes == 0 || | |
3337 | space.used_bytes == 0) | |
3338 | continue; | |
3339 | flags = space.flags; | |
3340 | /* | |
3341 | * return | |
3342 | * 0: if dup, single or RAID0 is configured for | |
3343 | * any of metadata, system or data, else | |
3344 | * 1: if RAID5 is configured, or if RAID1 or | |
3345 | * RAID10 is configured and only two mirrors | |
3346 | * are used, else | |
3347 | * 2: if RAID6 is configured, else | |
3348 | * num_mirrors - 1: if RAID1 or RAID10 is | |
3349 | * configured and more than | |
3350 | * 2 mirrors are used. | |
3351 | */ | |
3352 | if (num_tolerated_disk_barrier_failures > 0 && | |
3353 | ((flags & (BTRFS_BLOCK_GROUP_DUP | | |
3354 | BTRFS_BLOCK_GROUP_RAID0)) || | |
3355 | ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) | |
3356 | == 0))) | |
3357 | num_tolerated_disk_barrier_failures = 0; | |
53b381b3 DW |
3358 | else if (num_tolerated_disk_barrier_failures > 1) { |
3359 | if (flags & (BTRFS_BLOCK_GROUP_RAID1 | | |
3360 | BTRFS_BLOCK_GROUP_RAID5 | | |
3361 | BTRFS_BLOCK_GROUP_RAID10)) { | |
3362 | num_tolerated_disk_barrier_failures = 1; | |
3363 | } else if (flags & | |
15b0a89d | 3364 | BTRFS_BLOCK_GROUP_RAID6) { |
53b381b3 DW |
3365 | num_tolerated_disk_barrier_failures = 2; |
3366 | } | |
3367 | } | |
5af3e8cc SB |
3368 | } |
3369 | } | |
3370 | up_read(&sinfo->groups_sem); | |
3371 | } | |
3372 | ||
3373 | return num_tolerated_disk_barrier_failures; | |
3374 | } | |
3375 | ||
48a3b636 | 3376 | static int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 3377 | { |
e5e9a520 | 3378 | struct list_head *head; |
f2984462 | 3379 | struct btrfs_device *dev; |
a061fc8d | 3380 | struct btrfs_super_block *sb; |
f2984462 | 3381 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3382 | int ret; |
3383 | int do_barriers; | |
a236aed1 CM |
3384 | int max_errors; |
3385 | int total_errors = 0; | |
a061fc8d | 3386 | u64 flags; |
f2984462 CM |
3387 | |
3388 | do_barriers = !btrfs_test_opt(root, NOBARRIER); | |
af31f5e5 | 3389 | backup_super_roots(root->fs_info); |
f2984462 | 3390 | |
6c41761f | 3391 | sb = root->fs_info->super_for_commit; |
a061fc8d | 3392 | dev_item = &sb->dev_item; |
e5e9a520 | 3393 | |
174ba509 | 3394 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 3395 | head = &root->fs_info->fs_devices->devices; |
d7306801 | 3396 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
387125fc | 3397 | |
5af3e8cc SB |
3398 | if (do_barriers) { |
3399 | ret = barrier_all_devices(root->fs_info); | |
3400 | if (ret) { | |
3401 | mutex_unlock( | |
3402 | &root->fs_info->fs_devices->device_list_mutex); | |
3403 | btrfs_error(root->fs_info, ret, | |
3404 | "errors while submitting device barriers."); | |
3405 | return ret; | |
3406 | } | |
3407 | } | |
387125fc | 3408 | |
1f78160c | 3409 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3410 | if (!dev->bdev) { |
3411 | total_errors++; | |
3412 | continue; | |
3413 | } | |
2b82032c | 3414 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3415 | continue; |
3416 | ||
2b82032c | 3417 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3418 | btrfs_set_stack_device_type(dev_item, dev->type); |
3419 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
3420 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
3421 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
3422 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
3423 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3424 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3425 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 3426 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 3427 | |
a061fc8d CM |
3428 | flags = btrfs_super_flags(sb); |
3429 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3430 | ||
a512bbf8 | 3431 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
3432 | if (ret) |
3433 | total_errors++; | |
f2984462 | 3434 | } |
a236aed1 | 3435 | if (total_errors > max_errors) { |
efe120a0 | 3436 | btrfs_err(root->fs_info, "%d errors while writing supers", |
d397712b | 3437 | total_errors); |
a724b436 | 3438 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
79787eaa | 3439 | |
9d565ba4 SB |
3440 | /* FUA is masked off if unsupported and can't be the reason */ |
3441 | btrfs_error(root->fs_info, -EIO, | |
3442 | "%d errors while writing supers", total_errors); | |
3443 | return -EIO; | |
a236aed1 | 3444 | } |
f2984462 | 3445 | |
a512bbf8 | 3446 | total_errors = 0; |
1f78160c | 3447 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3448 | if (!dev->bdev) |
3449 | continue; | |
2b82032c | 3450 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3451 | continue; |
3452 | ||
a512bbf8 YZ |
3453 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
3454 | if (ret) | |
3455 | total_errors++; | |
f2984462 | 3456 | } |
174ba509 | 3457 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3458 | if (total_errors > max_errors) { |
79787eaa JM |
3459 | btrfs_error(root->fs_info, -EIO, |
3460 | "%d errors while writing supers", total_errors); | |
3461 | return -EIO; | |
a236aed1 | 3462 | } |
f2984462 CM |
3463 | return 0; |
3464 | } | |
3465 | ||
a512bbf8 YZ |
3466 | int write_ctree_super(struct btrfs_trans_handle *trans, |
3467 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 3468 | { |
f570e757 | 3469 | return write_all_supers(root, max_mirrors); |
cfaa7295 CM |
3470 | } |
3471 | ||
cb517eab MX |
3472 | /* Drop a fs root from the radix tree and free it. */ |
3473 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3474 | struct btrfs_root *root) | |
2619ba1f | 3475 | { |
4df27c4d | 3476 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3477 | radix_tree_delete(&fs_info->fs_roots_radix, |
3478 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3479 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3480 | |
3481 | if (btrfs_root_refs(&root->root_item) == 0) | |
3482 | synchronize_srcu(&fs_info->subvol_srcu); | |
3483 | ||
1a4319cc | 3484 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
3321719e | 3485 | btrfs_free_log(NULL, root); |
3321719e | 3486 | |
581bb050 LZ |
3487 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
3488 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3489 | free_fs_root(root); |
4df27c4d YZ |
3490 | } |
3491 | ||
3492 | static void free_fs_root(struct btrfs_root *root) | |
3493 | { | |
82d5902d | 3494 | iput(root->cache_inode); |
4df27c4d | 3495 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
1cb048f5 FDBM |
3496 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3497 | root->orphan_block_rsv = NULL; | |
0ee5dc67 AV |
3498 | if (root->anon_dev) |
3499 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3500 | if (root->subv_writers) |
3501 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3502 | free_extent_buffer(root->node); |
3503 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3504 | kfree(root->free_ino_ctl); |
3505 | kfree(root->free_ino_pinned); | |
d397712b | 3506 | kfree(root->name); |
b0feb9d9 | 3507 | btrfs_put_fs_root(root); |
2619ba1f CM |
3508 | } |
3509 | ||
cb517eab MX |
3510 | void btrfs_free_fs_root(struct btrfs_root *root) |
3511 | { | |
3512 | free_fs_root(root); | |
2619ba1f CM |
3513 | } |
3514 | ||
c146afad | 3515 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3516 | { |
c146afad YZ |
3517 | u64 root_objectid = 0; |
3518 | struct btrfs_root *gang[8]; | |
3519 | int i; | |
3768f368 | 3520 | int ret; |
e089f05c | 3521 | |
c146afad YZ |
3522 | while (1) { |
3523 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
3524 | (void **)gang, root_objectid, | |
3525 | ARRAY_SIZE(gang)); | |
3526 | if (!ret) | |
3527 | break; | |
5d4f98a2 YZ |
3528 | |
3529 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 3530 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
3531 | int err; |
3532 | ||
c146afad | 3533 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3534 | err = btrfs_orphan_cleanup(gang[i]); |
3535 | if (err) | |
3536 | return err; | |
c146afad YZ |
3537 | } |
3538 | root_objectid++; | |
3539 | } | |
3540 | return 0; | |
3541 | } | |
a2135011 | 3542 | |
c146afad YZ |
3543 | int btrfs_commit_super(struct btrfs_root *root) |
3544 | { | |
3545 | struct btrfs_trans_handle *trans; | |
a74a4b97 | 3546 | |
c146afad | 3547 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 3548 | btrfs_run_delayed_iputs(root); |
c146afad | 3549 | mutex_unlock(&root->fs_info->cleaner_mutex); |
9d1a2a3a | 3550 | wake_up_process(root->fs_info->cleaner_kthread); |
c71bf099 YZ |
3551 | |
3552 | /* wait until ongoing cleanup work done */ | |
3553 | down_write(&root->fs_info->cleanup_work_sem); | |
3554 | up_write(&root->fs_info->cleanup_work_sem); | |
3555 | ||
7a7eaa40 | 3556 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3557 | if (IS_ERR(trans)) |
3558 | return PTR_ERR(trans); | |
d52c1bcc | 3559 | return btrfs_commit_transaction(trans, root); |
c146afad YZ |
3560 | } |
3561 | ||
3562 | int close_ctree(struct btrfs_root *root) | |
3563 | { | |
3564 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3565 | int ret; | |
3566 | ||
3567 | fs_info->closing = 1; | |
3568 | smp_mb(); | |
3569 | ||
803b2f54 SB |
3570 | /* wait for the uuid_scan task to finish */ |
3571 | down(&fs_info->uuid_tree_rescan_sem); | |
3572 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3573 | up(&fs_info->uuid_tree_rescan_sem); | |
3574 | ||
837d5b6e | 3575 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3576 | btrfs_pause_balance(fs_info); |
837d5b6e | 3577 | |
8dabb742 SB |
3578 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3579 | ||
aa1b8cd4 | 3580 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3581 | |
3582 | /* wait for any defraggers to finish */ | |
3583 | wait_event(fs_info->transaction_wait, | |
3584 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3585 | ||
3586 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3587 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3588 | |
c146afad | 3589 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3590 | ret = btrfs_commit_super(root); |
3591 | if (ret) | |
efe120a0 | 3592 | btrfs_err(root->fs_info, "commit super ret %d", ret); |
acce952b | 3593 | } |
3594 | ||
87533c47 | 3595 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
68ce9682 | 3596 | btrfs_error_commit_super(root); |
0f7d52f4 | 3597 | |
e3029d9f AV |
3598 | kthread_stop(fs_info->transaction_kthread); |
3599 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3600 | |
f25784b3 YZ |
3601 | fs_info->closing = 2; |
3602 | smp_mb(); | |
3603 | ||
bcef60f2 AJ |
3604 | btrfs_free_qgroup_config(root->fs_info); |
3605 | ||
963d678b | 3606 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
efe120a0 | 3607 | btrfs_info(root->fs_info, "at unmount delalloc count %lld", |
963d678b | 3608 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3609 | } |
bcc63abb | 3610 | |
5ac1d209 JM |
3611 | btrfs_sysfs_remove_one(fs_info); |
3612 | ||
c146afad | 3613 | del_fs_roots(fs_info); |
d10c5f31 | 3614 | |
1a4319cc LB |
3615 | btrfs_put_block_group_cache(fs_info); |
3616 | ||
2b1360da JB |
3617 | btrfs_free_block_groups(fs_info); |
3618 | ||
96192499 JB |
3619 | btrfs_stop_all_workers(fs_info); |
3620 | ||
13e6c37b | 3621 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3622 | |
13e6c37b | 3623 | iput(fs_info->btree_inode); |
d6bfde87 | 3624 | |
21adbd5c SB |
3625 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3626 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3627 | btrfsic_unmount(root, fs_info->fs_devices); | |
3628 | #endif | |
3629 | ||
dfe25020 | 3630 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3631 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3632 | |
e2d84521 | 3633 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3634 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3635 | percpu_counter_destroy(&fs_info->bio_counter); |
04160088 | 3636 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3637 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3638 | |
53b381b3 DW |
3639 | btrfs_free_stripe_hash_table(fs_info); |
3640 | ||
1cb048f5 FDBM |
3641 | btrfs_free_block_rsv(root, root->orphan_block_rsv); |
3642 | root->orphan_block_rsv = NULL; | |
3643 | ||
eb60ceac CM |
3644 | return 0; |
3645 | } | |
3646 | ||
b9fab919 CM |
3647 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3648 | int atomic) | |
5f39d397 | 3649 | { |
1259ab75 | 3650 | int ret; |
727011e0 | 3651 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3652 | |
0b32f4bb | 3653 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3654 | if (!ret) |
3655 | return ret; | |
3656 | ||
3657 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3658 | parent_transid, atomic); |
3659 | if (ret == -EAGAIN) | |
3660 | return ret; | |
1259ab75 | 3661 | return !ret; |
5f39d397 CM |
3662 | } |
3663 | ||
3664 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3665 | { |
0b32f4bb | 3666 | return set_extent_buffer_uptodate(buf); |
5f39d397 | 3667 | } |
6702ed49 | 3668 | |
5f39d397 CM |
3669 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3670 | { | |
06ea65a3 | 3671 | struct btrfs_root *root; |
5f39d397 | 3672 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3673 | int was_dirty; |
b4ce94de | 3674 | |
06ea65a3 JB |
3675 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3676 | /* | |
3677 | * This is a fast path so only do this check if we have sanity tests | |
3678 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3679 | * outside of the sanity tests. | |
3680 | */ | |
3681 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3682 | return; | |
3683 | #endif | |
3684 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
b9447ef8 | 3685 | btrfs_assert_tree_locked(buf); |
31b1a2bd JL |
3686 | if (transid != root->fs_info->generation) |
3687 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " | |
d397712b | 3688 | "found %llu running %llu\n", |
c1c9ff7c | 3689 | buf->start, transid, root->fs_info->generation); |
0b32f4bb | 3690 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 MX |
3691 | if (!was_dirty) |
3692 | __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, | |
3693 | buf->len, | |
3694 | root->fs_info->dirty_metadata_batch); | |
eb60ceac CM |
3695 | } |
3696 | ||
b53d3f5d LB |
3697 | static void __btrfs_btree_balance_dirty(struct btrfs_root *root, |
3698 | int flush_delayed) | |
16cdcec7 MX |
3699 | { |
3700 | /* | |
3701 | * looks as though older kernels can get into trouble with | |
3702 | * this code, they end up stuck in balance_dirty_pages forever | |
3703 | */ | |
e2d84521 | 3704 | int ret; |
16cdcec7 MX |
3705 | |
3706 | if (current->flags & PF_MEMALLOC) | |
3707 | return; | |
3708 | ||
b53d3f5d LB |
3709 | if (flush_delayed) |
3710 | btrfs_balance_delayed_items(root); | |
16cdcec7 | 3711 | |
e2d84521 MX |
3712 | ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, |
3713 | BTRFS_DIRTY_METADATA_THRESH); | |
3714 | if (ret > 0) { | |
d0e1d66b NJ |
3715 | balance_dirty_pages_ratelimited( |
3716 | root->fs_info->btree_inode->i_mapping); | |
16cdcec7 MX |
3717 | } |
3718 | return; | |
3719 | } | |
3720 | ||
b53d3f5d | 3721 | void btrfs_btree_balance_dirty(struct btrfs_root *root) |
35b7e476 | 3722 | { |
b53d3f5d LB |
3723 | __btrfs_btree_balance_dirty(root, 1); |
3724 | } | |
585ad2c3 | 3725 | |
b53d3f5d LB |
3726 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) |
3727 | { | |
3728 | __btrfs_btree_balance_dirty(root, 0); | |
35b7e476 | 3729 | } |
6b80053d | 3730 | |
ca7a79ad | 3731 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3732 | { |
727011e0 | 3733 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
0b32f4bb | 3734 | return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
6b80053d | 3735 | } |
0da5468f | 3736 | |
fcd1f065 | 3737 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
acce952b | 3738 | int read_only) |
3739 | { | |
1104a885 DS |
3740 | /* |
3741 | * Placeholder for checks | |
3742 | */ | |
fcd1f065 | 3743 | return 0; |
acce952b | 3744 | } |
3745 | ||
48a3b636 | 3746 | static void btrfs_error_commit_super(struct btrfs_root *root) |
acce952b | 3747 | { |
acce952b | 3748 | mutex_lock(&root->fs_info->cleaner_mutex); |
3749 | btrfs_run_delayed_iputs(root); | |
3750 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3751 | ||
3752 | down_write(&root->fs_info->cleanup_work_sem); | |
3753 | up_write(&root->fs_info->cleanup_work_sem); | |
3754 | ||
3755 | /* cleanup FS via transaction */ | |
3756 | btrfs_cleanup_transaction(root); | |
acce952b | 3757 | } |
3758 | ||
569e0f35 JB |
3759 | static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, |
3760 | struct btrfs_root *root) | |
acce952b | 3761 | { |
3762 | struct btrfs_inode *btrfs_inode; | |
3763 | struct list_head splice; | |
3764 | ||
3765 | INIT_LIST_HEAD(&splice); | |
3766 | ||
3767 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
199c2a9c | 3768 | spin_lock(&root->fs_info->ordered_root_lock); |
acce952b | 3769 | |
569e0f35 | 3770 | list_splice_init(&t->ordered_operations, &splice); |
acce952b | 3771 | while (!list_empty(&splice)) { |
3772 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3773 | ordered_operations); | |
3774 | ||
3775 | list_del_init(&btrfs_inode->ordered_operations); | |
199c2a9c | 3776 | spin_unlock(&root->fs_info->ordered_root_lock); |
acce952b | 3777 | |
3778 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 3779 | |
199c2a9c | 3780 | spin_lock(&root->fs_info->ordered_root_lock); |
acce952b | 3781 | } |
3782 | ||
199c2a9c | 3783 | spin_unlock(&root->fs_info->ordered_root_lock); |
acce952b | 3784 | mutex_unlock(&root->fs_info->ordered_operations_mutex); |
acce952b | 3785 | } |
3786 | ||
143bede5 | 3787 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 3788 | { |
acce952b | 3789 | struct btrfs_ordered_extent *ordered; |
acce952b | 3790 | |
199c2a9c | 3791 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
3792 | /* |
3793 | * This will just short circuit the ordered completion stuff which will | |
3794 | * make sure the ordered extent gets properly cleaned up. | |
3795 | */ | |
199c2a9c | 3796 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
3797 | root_extent_list) |
3798 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
3799 | spin_unlock(&root->ordered_extent_lock); |
3800 | } | |
3801 | ||
3802 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
3803 | { | |
3804 | struct btrfs_root *root; | |
3805 | struct list_head splice; | |
3806 | ||
3807 | INIT_LIST_HEAD(&splice); | |
3808 | ||
3809 | spin_lock(&fs_info->ordered_root_lock); | |
3810 | list_splice_init(&fs_info->ordered_roots, &splice); | |
3811 | while (!list_empty(&splice)) { | |
3812 | root = list_first_entry(&splice, struct btrfs_root, | |
3813 | ordered_root); | |
1de2cfde JB |
3814 | list_move_tail(&root->ordered_root, |
3815 | &fs_info->ordered_roots); | |
199c2a9c | 3816 | |
2a85d9ca | 3817 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
3818 | btrfs_destroy_ordered_extents(root); |
3819 | ||
2a85d9ca LB |
3820 | cond_resched(); |
3821 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
3822 | } |
3823 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 3824 | } |
3825 | ||
35a3621b SB |
3826 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
3827 | struct btrfs_root *root) | |
acce952b | 3828 | { |
3829 | struct rb_node *node; | |
3830 | struct btrfs_delayed_ref_root *delayed_refs; | |
3831 | struct btrfs_delayed_ref_node *ref; | |
3832 | int ret = 0; | |
3833 | ||
3834 | delayed_refs = &trans->delayed_refs; | |
3835 | ||
3836 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 3837 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 3838 | spin_unlock(&delayed_refs->lock); |
efe120a0 | 3839 | btrfs_info(root->fs_info, "delayed_refs has NO entry"); |
acce952b | 3840 | return ret; |
3841 | } | |
3842 | ||
d7df2c79 JB |
3843 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
3844 | struct btrfs_delayed_ref_head *head; | |
e78417d1 | 3845 | bool pin_bytes = false; |
acce952b | 3846 | |
d7df2c79 JB |
3847 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
3848 | href_node); | |
3849 | if (!mutex_trylock(&head->mutex)) { | |
3850 | atomic_inc(&head->node.refs); | |
3851 | spin_unlock(&delayed_refs->lock); | |
acce952b | 3852 | |
d7df2c79 | 3853 | mutex_lock(&head->mutex); |
e78417d1 | 3854 | mutex_unlock(&head->mutex); |
d7df2c79 JB |
3855 | btrfs_put_delayed_ref(&head->node); |
3856 | spin_lock(&delayed_refs->lock); | |
3857 | continue; | |
3858 | } | |
3859 | spin_lock(&head->lock); | |
3860 | while ((node = rb_first(&head->ref_root)) != NULL) { | |
3861 | ref = rb_entry(node, struct btrfs_delayed_ref_node, | |
3862 | rb_node); | |
3863 | ref->in_tree = 0; | |
3864 | rb_erase(&ref->rb_node, &head->ref_root); | |
3865 | atomic_dec(&delayed_refs->num_entries); | |
3866 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 3867 | } |
d7df2c79 JB |
3868 | if (head->must_insert_reserved) |
3869 | pin_bytes = true; | |
3870 | btrfs_free_delayed_extent_op(head->extent_op); | |
3871 | delayed_refs->num_heads--; | |
3872 | if (head->processing == 0) | |
3873 | delayed_refs->num_heads_ready--; | |
3874 | atomic_dec(&delayed_refs->num_entries); | |
3875 | head->node.in_tree = 0; | |
3876 | rb_erase(&head->href_node, &delayed_refs->href_root); | |
3877 | spin_unlock(&head->lock); | |
3878 | spin_unlock(&delayed_refs->lock); | |
3879 | mutex_unlock(&head->mutex); | |
acce952b | 3880 | |
d7df2c79 JB |
3881 | if (pin_bytes) |
3882 | btrfs_pin_extent(root, head->node.bytenr, | |
3883 | head->node.num_bytes, 1); | |
3884 | btrfs_put_delayed_ref(&head->node); | |
acce952b | 3885 | cond_resched(); |
3886 | spin_lock(&delayed_refs->lock); | |
3887 | } | |
3888 | ||
3889 | spin_unlock(&delayed_refs->lock); | |
3890 | ||
3891 | return ret; | |
3892 | } | |
3893 | ||
143bede5 | 3894 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 3895 | { |
3896 | struct btrfs_inode *btrfs_inode; | |
3897 | struct list_head splice; | |
3898 | ||
3899 | INIT_LIST_HEAD(&splice); | |
3900 | ||
eb73c1b7 MX |
3901 | spin_lock(&root->delalloc_lock); |
3902 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 3903 | |
3904 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
3905 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
3906 | delalloc_inodes); | |
acce952b | 3907 | |
3908 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
3909 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
3910 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 3911 | spin_unlock(&root->delalloc_lock); |
acce952b | 3912 | |
3913 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 3914 | |
eb73c1b7 | 3915 | spin_lock(&root->delalloc_lock); |
acce952b | 3916 | } |
3917 | ||
eb73c1b7 MX |
3918 | spin_unlock(&root->delalloc_lock); |
3919 | } | |
3920 | ||
3921 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
3922 | { | |
3923 | struct btrfs_root *root; | |
3924 | struct list_head splice; | |
3925 | ||
3926 | INIT_LIST_HEAD(&splice); | |
3927 | ||
3928 | spin_lock(&fs_info->delalloc_root_lock); | |
3929 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
3930 | while (!list_empty(&splice)) { | |
3931 | root = list_first_entry(&splice, struct btrfs_root, | |
3932 | delalloc_root); | |
3933 | list_del_init(&root->delalloc_root); | |
3934 | root = btrfs_grab_fs_root(root); | |
3935 | BUG_ON(!root); | |
3936 | spin_unlock(&fs_info->delalloc_root_lock); | |
3937 | ||
3938 | btrfs_destroy_delalloc_inodes(root); | |
3939 | btrfs_put_fs_root(root); | |
3940 | ||
3941 | spin_lock(&fs_info->delalloc_root_lock); | |
3942 | } | |
3943 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 3944 | } |
3945 | ||
3946 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3947 | struct extent_io_tree *dirty_pages, | |
3948 | int mark) | |
3949 | { | |
3950 | int ret; | |
acce952b | 3951 | struct extent_buffer *eb; |
3952 | u64 start = 0; | |
3953 | u64 end; | |
acce952b | 3954 | |
3955 | while (1) { | |
3956 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 3957 | mark, NULL); |
acce952b | 3958 | if (ret) |
3959 | break; | |
3960 | ||
3961 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3962 | while (start <= end) { | |
fd8b2b61 JB |
3963 | eb = btrfs_find_tree_block(root, start, |
3964 | root->leafsize); | |
69a85bd8 | 3965 | start += root->leafsize; |
fd8b2b61 | 3966 | if (!eb) |
acce952b | 3967 | continue; |
fd8b2b61 | 3968 | wait_on_extent_buffer_writeback(eb); |
acce952b | 3969 | |
fd8b2b61 JB |
3970 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
3971 | &eb->bflags)) | |
3972 | clear_extent_buffer_dirty(eb); | |
3973 | free_extent_buffer_stale(eb); | |
acce952b | 3974 | } |
3975 | } | |
3976 | ||
3977 | return ret; | |
3978 | } | |
3979 | ||
3980 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3981 | struct extent_io_tree *pinned_extents) | |
3982 | { | |
3983 | struct extent_io_tree *unpin; | |
3984 | u64 start; | |
3985 | u64 end; | |
3986 | int ret; | |
ed0eaa14 | 3987 | bool loop = true; |
acce952b | 3988 | |
3989 | unpin = pinned_extents; | |
ed0eaa14 | 3990 | again: |
acce952b | 3991 | while (1) { |
3992 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 3993 | EXTENT_DIRTY, NULL); |
acce952b | 3994 | if (ret) |
3995 | break; | |
3996 | ||
3997 | /* opt_discard */ | |
5378e607 LD |
3998 | if (btrfs_test_opt(root, DISCARD)) |
3999 | ret = btrfs_error_discard_extent(root, start, | |
4000 | end + 1 - start, | |
4001 | NULL); | |
acce952b | 4002 | |
4003 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
4004 | btrfs_error_unpin_extent_range(root, start, end); | |
4005 | cond_resched(); | |
4006 | } | |
4007 | ||
ed0eaa14 LB |
4008 | if (loop) { |
4009 | if (unpin == &root->fs_info->freed_extents[0]) | |
4010 | unpin = &root->fs_info->freed_extents[1]; | |
4011 | else | |
4012 | unpin = &root->fs_info->freed_extents[0]; | |
4013 | loop = false; | |
4014 | goto again; | |
4015 | } | |
4016 | ||
acce952b | 4017 | return 0; |
4018 | } | |
4019 | ||
49b25e05 JM |
4020 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
4021 | struct btrfs_root *root) | |
4022 | { | |
724e2315 JB |
4023 | btrfs_destroy_ordered_operations(cur_trans, root); |
4024 | ||
49b25e05 | 4025 | btrfs_destroy_delayed_refs(cur_trans, root); |
49b25e05 | 4026 | |
4a9d8bde | 4027 | cur_trans->state = TRANS_STATE_COMMIT_START; |
d7096fc3 | 4028 | wake_up(&root->fs_info->transaction_blocked_wait); |
49b25e05 | 4029 | |
4a9d8bde | 4030 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
d7096fc3 | 4031 | wake_up(&root->fs_info->transaction_wait); |
49b25e05 | 4032 | |
67cde344 MX |
4033 | btrfs_destroy_delayed_inodes(root); |
4034 | btrfs_assert_delayed_root_empty(root); | |
49b25e05 | 4035 | |
49b25e05 JM |
4036 | btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, |
4037 | EXTENT_DIRTY); | |
6e841e32 LB |
4038 | btrfs_destroy_pinned_extent(root, |
4039 | root->fs_info->pinned_extents); | |
49b25e05 | 4040 | |
4a9d8bde MX |
4041 | cur_trans->state =TRANS_STATE_COMPLETED; |
4042 | wake_up(&cur_trans->commit_wait); | |
4043 | ||
49b25e05 JM |
4044 | /* |
4045 | memset(cur_trans, 0, sizeof(*cur_trans)); | |
4046 | kmem_cache_free(btrfs_transaction_cachep, cur_trans); | |
4047 | */ | |
4048 | } | |
4049 | ||
48a3b636 | 4050 | static int btrfs_cleanup_transaction(struct btrfs_root *root) |
acce952b | 4051 | { |
4052 | struct btrfs_transaction *t; | |
acce952b | 4053 | |
acce952b | 4054 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
4055 | ||
a4abeea4 | 4056 | spin_lock(&root->fs_info->trans_lock); |
724e2315 JB |
4057 | while (!list_empty(&root->fs_info->trans_list)) { |
4058 | t = list_first_entry(&root->fs_info->trans_list, | |
4059 | struct btrfs_transaction, list); | |
4060 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
4061 | atomic_inc(&t->use_count); | |
4062 | spin_unlock(&root->fs_info->trans_lock); | |
4063 | btrfs_wait_for_commit(root, t->transid); | |
4064 | btrfs_put_transaction(t); | |
4065 | spin_lock(&root->fs_info->trans_lock); | |
4066 | continue; | |
4067 | } | |
4068 | if (t == root->fs_info->running_transaction) { | |
4069 | t->state = TRANS_STATE_COMMIT_DOING; | |
4070 | spin_unlock(&root->fs_info->trans_lock); | |
4071 | /* | |
4072 | * We wait for 0 num_writers since we don't hold a trans | |
4073 | * handle open currently for this transaction. | |
4074 | */ | |
4075 | wait_event(t->writer_wait, | |
4076 | atomic_read(&t->num_writers) == 0); | |
4077 | } else { | |
4078 | spin_unlock(&root->fs_info->trans_lock); | |
4079 | } | |
4080 | btrfs_cleanup_one_transaction(t, root); | |
4a9d8bde | 4081 | |
724e2315 JB |
4082 | spin_lock(&root->fs_info->trans_lock); |
4083 | if (t == root->fs_info->running_transaction) | |
4084 | root->fs_info->running_transaction = NULL; | |
acce952b | 4085 | list_del_init(&t->list); |
724e2315 | 4086 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 4087 | |
724e2315 JB |
4088 | btrfs_put_transaction(t); |
4089 | trace_btrfs_transaction_commit(root); | |
4090 | spin_lock(&root->fs_info->trans_lock); | |
4091 | } | |
4092 | spin_unlock(&root->fs_info->trans_lock); | |
4093 | btrfs_destroy_all_ordered_extents(root->fs_info); | |
4094 | btrfs_destroy_delayed_inodes(root); | |
4095 | btrfs_assert_delayed_root_empty(root); | |
4096 | btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); | |
4097 | btrfs_destroy_all_delalloc_inodes(root->fs_info); | |
acce952b | 4098 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 4099 | |
4100 | return 0; | |
4101 | } | |
4102 | ||
d1310b2e | 4103 | static struct extent_io_ops btree_extent_io_ops = { |
ce9adaa5 | 4104 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 4105 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 4106 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
4107 | /* note we're sharing with inode.c for the merge bio hook */ |
4108 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 4109 | }; |